Datasets:

pszemraj

/

scientific_lay_summarisation-plos-norm

like 6

Fanconi Anemia (FA) is a rare autosomal recessive disorder characterized by hypersensitivity to inter-strand crosslinks (ICLs). FANCD2, a central factor of the FA pathway, is essential for the repair of double strand breaks (DSBs) generated during fork collapse at ICLs. While lesions different from ICLs can also trigger fork collapse, the contribution of FANCD2 to the resolution of replication-coupled DSBs generated independently from ICLs is unknown. Intriguingly, FANCD2 is readily activated after UV irradiation, a DNA-damaging agent that generates predominantly intra-strand crosslinks but not ICLs. Hence, UV irradiation is an ideal tool to explore the contribution of FANCD2 to the DNA damage response triggered by DNA lesions other than ICL repair. Here we show that, in contrast to ICL-causing agents, UV radiation compromises cell survival independently from FANCD2. In agreement, FANCD2 depletion does not increase the amount of DSBs generated during the replication of UV-damaged DNA and is dispensable for UV-induced checkpoint activation. Remarkably however, FANCD2 protects UV-dependent, replication-coupled DSBs from aberrant processing by non-homologous end joining, preventing the accumulation of micronuclei and chromatid aberrations including non-homologous chromatid exchanges. Hence, while dispensable for cell survival, FANCD2 selectively safeguards chromosomal stability after UV-triggered replication stress. Fanconi anemia (FA) is a rare recessive disorder characterized by increased spontaneous rearrangements of chromosomes, tumorigenesis and cell death [1,2]. Initial signs of FA include bone or skeleton defects, renal dysfunction, short stature and very frequently abnormal hyper- and hypo-pigmentation of the skin and café_au_lait spots [3]. FA is characterized by bone marrow failure and high risk of developing myeloid leukemias and squamous cell carcinomas [4]. Cells derived from FA patients are strikingly sensitive to DNA interstrand crosslinks (ICLs), i. e. cross-links between two DNA strands. Consequently, much of our current understanding of FA comes from studies that utilize ICL-causing agents, such as mitomycin C (MMC), diepoxybutane or cisplatin, as sources of DNA damage [1,2]. To date, 17 genes with described mutations in patients were defined as components of the FA pathway that are all required for ICL repair [5]. ICL removal is generally accomplished when the replication fork abuts the DNA lesion. ICL-stalled replication forks undergo a programmed collapse, which is regulated by all FA proteins [6]. Firstly, FANCD2 is loaded onto the ICL, a process that requires the FA core complex, the D2 partner FANCI and D2 monoubiquitination [7]. Indeed, FANCD2-FANCI bind preferentially to a variety of branched DNA structures formed by ICL repair intermediates [8,9]. Moreover, the crystal structure of FANCI with DNA suggests that the ID2 complex could accommodate the X-shaped DNA structures formed by replication forks that collide with ICLs [10]. Secondly, FANCD2 recruits the XESS nuclease complex (including the nucleases XPF-ERCC1 and SLX1 and the scaffold protein SLX4) and the FAN1 and SNM1A nucleases [8]. Thirdly, these enzymes co-ordinately incise the DNA 3´and 5´of the lesion, thus unhooking the ICL. Finally, FANCD2 masters the resolution of such DNA repair intermediate by coordinating the activation of translesion DNA synthesis (TLS), homologous recombination repair (HRR) and possibly Nucleotide Excision Repair (NER) [1,2]. Collectively, solid evidence demonstrates that FANCD2 is crucial to ICL repair. Upon γIR, a source of replication-independent DSBs, ATM activates FANCD2 by phosphorylation [11]. However, FANCD2-deficient cells are only moderately sensitive to γIR and X-rays, another source of replication-independent DSBs [12–15]. In addition, FANCD2 does not play a predominant role in the repair of DSBs generated by restriction enzymes, but it is key to the resolution of ICL-dependent replication-coupled DSBs [16]. These results led to the assumption that FANCD2 is specifically required for the resolution of all replication-coupled but not direct DSBs. However, it is yet unclear whether FANCD2 resolves DSBs generated at replication forks stalled by lesions others than ICLs. It has been shown that the activation of FANCD2 during unperturbed S phase [17] suggests that FANCD2 participates in mechanisms unrelated to DSB repair. Indeed, FANCD2 prevents the nucleolytic degradation of nascent DNA triggered by hydroxyurea (HU) or aphidicolin (APH) and promotes fork restart immediately after drug removal [18–22]. Hence, FANCD2 not only promotes DSB repair by HRR but also attenuates DSB formation by protecting persistently stalled replication forks and promoting their reactivation. Intriguingly, FANCD2 is activated by UV irradiation, a DNA-damaging agent which rarely causes ICL accumulation [23,24] with no persistent stalling of replication forks at doses of 20 J/m2 or lower [25,26]. In contrast to ICL repair, the removal of UV-induced lesions does not require coordination between TLS and NER as both processes can occur independently from each other in UV-treated cells [27]. Moreover, NER efficiency is not altered in FA-defective backgrounds [28]. Importantly, FANCD2-deficient cells show normal spontaneous and UV-C-induced point mutation frequency [29] and null or very low sensitivity to UV-light [30–33]. Nonetheless, it is intriguing that the hypo/hyperpigmentation and the café_au_lait spots that characterize the FA disease are skin-associated defects. We thus reasoned that the function of FANCD2 after UV irradiation could be revealed by exploring processes that may not necessarily trigger cell death. We found that the UV irradiation of FANCD2-depleted cells with doses as low as 1. 5 J/m2 cause a striking increase of genomic instability markers, such as aberrant chromatid exchanges and micronuclei (MN) formation. The generation of both aberrations require DSBs [34,35]. While UV irradiation is not expected to directly cause DSBs, replication-associated one-ended DSBs (also known as double strand ends–DSEs) could accumulate when elongating forks encounter UV lesions [36,37]. Our results demonstrate that FANCD2 does not majorly modulate DSB accumulation. On the contrary, FANCD2 guarantees the correct processing of replication-coupled DSBs after UV irradiation. In particular, FANCD2 promotes the recruitment of the HRR factor RAD51 to UV-damaged DNA and the resolution of replication-associated DSBs by HRR. When FANCD2 is depleted, unleashed non-homologous end joining (NHEJ) increases genomic instability after UV irradiation. Hence, FANCD2 operates beyond ICL processing, and such function might apply to all replication-coupled DSBs generated after different genotoxic insults. As reported by others [23,38], UV irradiation induces focal organization (S1A and S1B Fig) and monoubiquitination (S1C and S1D Fig) of FANCD2, both in U2OS and PD20 cells expressing FANCD2 (PD20+D2). However, FANCD2 depletion (Fig 1A) did not alter the cell cycle distribution after UV irradiation (whereas it did alter the cell cycle profile after MMC treatment, Fig 1B). Moreover, both the transient depletion of FANCD2 in U2OS cells (Fig 1A higher panel) and the permanent loss of FANCD2 in PD20 cells obtained from patients (Fig 1A lower panel) did not affect cell survival both in short (2 days) and long (8–10 days) term assays (however we observed hypersensitivity to MMC in FANCD2-depleted samples, Fig 1C–1D). To corroborate that the doses of UV radiation used can impact the clonogenic potential of U2OS cells, we depleted Pol η (to impair TLS). We observed that TLS-Pol η-depletion reduced the colony formation ability of UV-irradiated cells (S1E Fig), therefore demonstrating that UV hypersensitivity can be revealed in our settings. The undetectable contribution of FANCD2 to UV resistance is in agreement with four previous reports that found no effect of FANCD2 depletion on cell survival after UV irradiation [30–33] and other manuscripts that showed similar results when depleting FA core proteins (see Discussion). We reasoned that, while not affecting cell survival, FANCD2 depletion could jeopardize the stability of the genome after UV irradiation. To evaluate this possibility, we first analyzed MN formation at the lowest dose required to achieve a detectable difference between untreated and UV-irradiated samples (5 J/m2). Strikingly, when depleting FANCD2, the frequency of MN increased in UV-irradiated U2OS (Fig 2A and 2B) and in PD20 cells, when compared to control GM00637 fibroblasts or PD20 reconstituted counterparts, respectively (Fig 2C and 2D). MN are formed when DSB are processed in a manner that excludes fragments of chromosomes from nuclei during/ after karyokinesis and before cytokinesis [34]. Although not widely accepted, UV irradiation has been reported as a source of DSB formation [39–41]. We therefore inferred that the increase in MN in UV-treated FANCD2-depleted cells results from an increase in the number of DSBs and/or because of aberrant DSB processing. After UV irradiation, the most likely sources of DSBs are replication-coupled, one-ended double-strand ends generated at collapsed replication forks. The deficient resolution of replication-coupled DSBs increases replication-derived chromatid aberrations, which are specifically generated in S phase [42]. Thus, we evaluated the role of FANCD2 on chromatid aberrations after UV irradiation. We first determined the lowest dose required to upregulate these aberrations in control samples (1. 5 J/m2, Fig 2E and 2F). Interestingly, despite modest FANCD2 ubiquitination at low UV doses (S1F Fig), chromatidic breaks/gaps were upregulated in such conditions when FANCD2 was depleted (Fig 2E). Moreover, aberrations such as chromatid exchanges (mono and poly-radial chromosomes), which have been exclusively associated to replication-coupled DSBs [35,42] robustly increased in FANCD2-depleted but not in control cells (Fig 2F). Importantly, only chromatid (generated in S/G2 phase) but not chromosome (generated in G1/G0) exchanges [42] accumulated in UV-irradiated FANCD2-depleted samples (S2 Fig). Altogether, Figs 2 and S2 indicate that FANCD2 activation is required to avoid aberrant processing of replication-coupled DSBs after UV irradiation. The afore-mentioned results indicate that FANCD2 either prevents DSB formation or regulates their processing once they are formed. To explore the first possibility, we first analyzed PCNA monoubiquitination and Pol η recruitment to replication factories, two hallmarks of UV-triggered TLS, a well-characterized mechanism that aids DNA replication across UV-triggered DNA lesions and could thus prevent UV-induced DSB formation [43]. [43]. It has been previously demonstrated that FA core components [29,44] but not FANCD2 [29] promote TLS events after UV irradiation. However one report indicates that FANCD2 depletion reduces the ratio of Pol η focus formation over total Pol η signal in UV-irradiated (20 J/m2) Hela cells [45]. Based on these previous reports, we reasoned that the depletion of FANCD2 could modulate TLS markers in our settings. When analysing PCNA ubiquitination in FANCD2-depleted samples and PD20 cells (Figs 3A and S3A), alterations were not evident at time points (6 hours) in which TLS events are expected to be fully active [46] or at later (24hrs) time points at doses used in MN formation and induction of chromosomal aberrations (≤5 J/m2 –Figs 1 and 2). To further explore a potential modulation of TLS activity after FANCD2 knockdown we also evaluated the recruitment of TLS Pol η to replication factories, which is another parameter of TLS activation [47]. Here we observed that the proportion of cells with Pol η foci was not modulated by FANCD2 depletion in our experimental settings at 5 J/m2 (Fig 3B). This is in agreement with the previously reported negligible contribution of FANCD2 to PCNA ubiquitination, Rev1 recruitment to replication factories and the unaltered TLS-dependent mutagenesis of UV-irradiated FANCD2-depleted samples [29,48]. Hence, two central TLS parameters were not modulated by FANCD2 knockdown at UV doses such as 5 J/m2, which do alter the genomic stability of FANCD2-depleted samples. We then explored checkpoint activation, which is up-regulated by replication fork stalling and/or by increased DSBs levels. Chk1 phosphorylation is readily induced after low doses of UV irradiation [49] and increases when FA core components are depleted, possibly as a consequence of TLS defects [44]. In contrast, Chk1 phosphorylation at Ser 345 was transiently reduced 6hs -but not 24hs- post-UV in FANCD2-depleted U2OS and in PD20 cells (Figs 3C and S3B). Moreover, the extent and the timing of p53 activation and p21 downregulation after UV irradiation [50] were not modified when FANCD2 was depleted (Fig 3C). Together, these results suggest that there is no persistent reprogramming of TLS and Chk1 signals in FANCD2-depleted cells. We then asked whether the total number of DSBs increases in UV-irradiated samples after FANCD2 transient or permanent knockdown. Supporting the notion of a constant number of DSBs, we found that the activating phosphorylation of the histone variant H2AX (γH2AX—S139) [51] increased in a manner that depended on the UV dose but not on the levels of FANCD2 (Fig 4D). Moreover, when specifically focusing on the 5 J/m2 dose, the intensity of the γH2AX signal modestly increased with respect to sham-irradiated controls with no significant changes after FANCD2 depletion (Fig 4A). The percentage of cells with γH2AX foci (Figs 4B, S4A and S4C) and the number of γH2AX foci per cell (Fig 4C) were also unaffected by FANCD2 depletion. These results are opposite to those obtained when we analyzed FANCD2-depleted cells treated with the ICLs inducer, MMC (S4B and S4C Fig). Since γH2AX foci can be formed in the absence of DSBs [52] we evaluated other markers of DSBs such as the phosphorylation of ATM kinase at S1981 or of KAP1 at S824 [7,53]. p-ATM did not increase and rather decreased in FANCD2-depleted samples (U2OS in Fig 4D and PD20 cells in S3C Fig). Similarly, pKAP1 levels did not increase in UV-irradiated FANCD2-depleted samples (Figs 4D and S3C). Our results are thus in agreement with a recent report from the Vaziri group showing Tunnel negative staining of FANCD2-depleted samples [54]. Collectively, these data suggest that FANCD2 depletion does not increase the levels of DSBs both before and after UV irradiation. To confirm this hypothesis, we set up a Pulse Field Gel Electrophoresis (PFGE) analysis to directly measure DSB formation. We observed no significant differences between control and FANCD2-depleted samples both 6 and 24 hours post-UV irradiation (Fig 4E). Similar results were obtained in PD20 cells (S4C Fig). While it might be argued that PFGE might have low sensitivity to detect small amounts of DSBs, our experimental setup proved to be sensitive enough to detect DSBs even at the lowest doses of UV irradiation (Fig 4E). Moreover, as a control of our PFGE experimental setup, we confirmed that FANCD2 prevents DSB accumulation in MMC-treated (S4E Fig) but not in UV-irradiated samples (S4D Fig). Collectively, the experiments in Figs 3 and 4 and S3 and S4 suggest that the infrequent DSBs that accumulate after UV irradiation are not upregulated by FANCD2 depletion. Rad51 is a highly conserved protein that promotes homology search and strand invasion events during HRR [55]. Rad51 recruitment to chromatin is thus a hallmark of HRR activation. We therefore analyzed the local recruitment of Rad51 to unshielded regions within UV irradiated nuclei to explore the FAND2 contribution to UV-dependent HRR (Fig 5A–5C). Interestingly, transient or permanent depletion of FANCD2 in U2OS and PD20 cells impaired Rad51 recruitment to unshielded nuclear regions (Fig 5B and 5C). To evaluate the functional contribution of FANCD2 to UV-induced HRR, we explored the frequency of homologous recombination events evidenced as the exchange of large DNA regions between sister chromatids (sister chromatid exchange-SCE) [56]. The defective accumulation of SCEs indicates defects in HRR activation at replication-associated DSBs [57]. Notably, the number of SCE decreased in UV-irradiated FANCD2-depleted samples (Fig 5D). This result suggests that FANCD2 directs the processing of UV-triggered DSBs generated at collapsed forks into HRR resolution. Cells choose to repair DSBs by HRR or NHEJ mainly depending on its replicative status [58]. Therefore, we evaluated the effect of FANCD2 depletion on the recruitment to γH2AX foci of a factor that is recruited to DSBs committed to NHEJ, the BRCT-containing protein 53BP1 [59,60]. Interestingly, the percentage of 53BP1 foci colocalizing with γH2AX foci was upregulated in UV-irradiated FANCD2-depleted samples (Fig 5E). Consistently, the total number of cells with 53BP1 foci increased in UV-treated FANCD2-depleted samples, albeit less markedly than after MMC treatment (to allow an easier comparison, the percentages of cells with 53BP1 foci in UV and MMC treated-cells are shown as overlapped bars in Fig 5F). Moreover, the cells with increased 53BP1 foci were almost exclusively those that were transiting S phase at the time of UV irradiation (S5A–S5C Fig), demonstrating that FANCD2 may prevent NHEJ events in S-phase. It is important to mention that other functions of 53BP1 such as the shielding of fragile DNA in G1 phase were recently documented [61,62]. Such 53BP1 structures are generated because of defective chromosomal segregation and are characterized by fewer but larger 53BP1 foci in G1 [61,62]. To evidence such 53BP1 foci we performed an EdU incorporation right before fixation and focused our analysis in EdU negative samples (S5D Fig). In such experimental settings, the percentage of cells with 53BP1 foci were unaffected by UV-irradiation in cells depleted from FANCD2 (S5E Fig). Hence, FANCD2 promotes the recruitment of HRR factors but not of NHEJ factors to UV-damaged DNA in cells transiting S phase. To evaluate the contribution of NHEJ to the genomic instability of FANCD2-depleted cells, we transiently downregulated the NHEJ core component XRCC4 [63]. The depletion of XRCC4 in U2OS cells (S6A Fig) had no effect on the number of cells with 53BP1 foci (S6B Fig), the levels of DSBs (Figs 6A and S6C), the clonogenic potential (S6D Fig) or the accumulation of chromosomal abnormalities (Fig 6B–6D) in sham- or UV-irradiated samples. As NHEJ is a pathway that resolves replication-independent DSBs [15], this result indicates that UV is not a source of such DSBs, while UV might trigger replication-coupled DSBs. The simultaneous depletion of XRCC4 and FANCD2 did not affect DSB accumulation (Figs 6A and S6C) in comparison to FANCD2- or sham-depleted samples, thus reinforcing the notion that FANCD2 depletion does not contribute to DSB formation. Cell survival was also unaffected by simultaneous depletion of XRCC4 and FANCD2 (S6D Fig). However, the percentage of cells with 53BP1 foci increased (S6B Fig), thus suggesting a potential delay in the processing of DSB at such foci in XRCC4- and FANCD2-depleted cells. Remarkably, XRCC4 depletion rescued the accumulation of chromatid aberrations and MN formation caused by FANCD2 depletion (Fig 6B–6D). Similarly, XRCC4 depletion rescued MN accumulation in PD20 cells (S7A–S7C Fig). Importantly, the prevention of MN accumulation after UV irradiation depended predominantly on FANCD2 ubiquitination as PD20 cells expressing the FANCD2 K561R mutant (PD20+D2 KRo) had MN levels similar to those in PD20 cells (S7B and S7C Fig). Moreover, the increased UV-associated genomic instability of PD20 cells expressing FANCD2 K561R was also rescued by XRCC4 depletion (S7B and S7C Fig). Finally, MN accumulated primarily in EdU-positive cells (S7D and S7E Fig), i. e. cells transiting S phase at the time of UV irradiation (see timeline in S7D Fig). Altogether, these results demonstrate that FANCD2 is crucial to the repair of replication-derived DSBs generated independently from ICLs. In contrast to FANCD2 function during ICL repair, FANCD2-dependent DSB repair pathway choice after UV irradiation is irrelevant to cell survival but it is key to safeguarding genomic stability. It has been previously reported that the elimination of the FA pathway has a modest or null effect on UV sensitivity. In fact, with the exception of FANCM, cells deficient in FANCC, FANCA, FANCE, FANCL, FANCD2 and FANCJ are not or modestly hypersensitive to UV light ([23,30–33,44,64–67] and this work). Remarkably however, we have unveiled a function of FANCD2 after UV irradiation. In particular, we show that FANCD2 preserves genomic stability by modulating the correct processing of DSBs generated during the replication of UV-damaged DNA. Whereas DSBs are not caused directly by UV irradiation [68], the accumulation of DSBs have been previously reported after 8–10 J/m2 [39–41]. Indeed, we demonstrate herein that DSBs are formed at UV doses of ≤5 J/m2. For example, UV doses as low as 1. 5 J/m2 induce SCE in control samples and complex aberrations (radials) in FANCD2-depleted samples. The formation of SCEs, aberrant chromatid exchanges and MN in binucleated cells require not only DSBs, but also DNA replication [69]. Therefore, UV-triggered DSBs are most likely generated as a consequence of DNA replication across damaged DNA. In fact, ATM phosphorylation after UV irradiation takes place predominantly in S phase [41]. Moreover, 53BP1 foci and MN in FANCD2-depleted cells accumulated almost exclusively in cells transiting S-phase at the time of UV irradiation (see S5 and S7 Figs). Thus, UV irradiation generates DSBs, most likely at collapsed replication forks. It should be noted that occasional ICLs, which depend on an alternative conformation of DNA that approximates pyrimidines from different strands, were also reported after UV irradiation [70–72]. In fact, when irradiating plasmidic DNA in vitro, a dose of 1000 J/m2 (260 nm) was required to accumulate ~0. 07 ICLs/kbp [24]. While we cannot formally discard their contribution, it is unlikely that such a sporadic event could predominate over other types of fork collapses (at frequent UV lesions such as unrepaired cyclobutane pyrimide dimers and 6–4 photoproducts). Moreover, FANCD2 differentially contributes to the replication of UV- and MMC-damaged DNA (see next section), thus reinforcing a difference in the fork-collapsing event after both treatments. Despite their increased genomic instability, FANCD2-depleted cells did not show increased DSB levels. In fact, PFGE did not reveal substantial changes in the accumulation of DSB in FANCD2–depleted samples. Consistently, KAP1 phosphorylation (a DSB marker) was not upregulated, and ATM and Chk1 phosphorylation were transiently downregulated. These results indicate that FANCD2 might process DNA repair intermediates at collapsed forks, generating substrates for ATM and Chk1 activation. Such speculation is supported by recent results indicating that after ICLs, DNA ends are resected into HRR-proficient substrates that promote robust ATM activation [73]. While we cannot further speculate on the signals leading to impaired ATM and Chk1 activation in UV-irradiated FANCD2-depleted samples, it is evident that in agreement with our PFGE results, the lack of upregulated phosphorylation of KAP1, ATM and Chk1 argues against a role of FANCD2 in the prevention of DSB accumulation after UV irradiation. Our results indicate that the main role of FANCD2 after UV irradiation is to direct DSBs into HRR repair (Figs 2 and 6). This implies that FANCD2 may be crucial to the repair of replication-coupled DSBs that arise from sources other than ICLs. Indeed, our results suggest that the functions of FANCD2 in the cellular response to UV irradiation and ICL accumulation partially overlap. However, the responses are not equivalent. This conclusion is supported by the following observations: A) FANCD2 depletion does not trigger cell death after UV irradiation, which is strikingly different from the significant increase in cell death observed after MMC treatment in FANCD2-depleted cells. Moreover, while NHEJ deficiency either rescues or exacerbates cell death in FANCD2 deficient samples treated with ICL inducers [74–76], we revealed an insignificant effect of XRCC4 depletion in the survival of UV-irradiated FANCD2-depleted cells. B) NHEJ depletion abrogates all the chromatid aberrations caused by UV irradiation in FANCD2-depleted cells. While similar results were reported in other systems using MMC [74,75], the simultaneous elimination of FANCD2 and KU80 after MMC and cisplatin treatments in mammalian cells not only fails to abrogate, but instead further increases chromosomal instability [76]. Hence, while in response to UV- and ICL-damaged DNA FANCD2 facilitates HRR, the quality and/or quantity of DSBs may not be equivalent in both scenarios. In fact, HRR most likely takes place after the convergence of two opposite replication forks at the ICL [8]. Therefore, the HRR substrate during ICL repair may resemble a canonical double-ended DSB, which could be repaired by NHEJ without causing a massive chromosomal rearrangement. In contrast, fork collapse induced by UV irradiation may generate DSEs which, in FANCD2-depleted backgrounds, may induce gross chromosomal rearrangements when processed by NHEJ. Alternatively, different nucleases may be recruited to DSBs after UV irradiation or ICLs. In this respect, it should be mentioned that FANCD2 not only recruits nucleases to ICLs but also to DNA lesions generated by HU [21]. While the nuclease in charge of the processing of UV-triggered DSBs remains unidentified, we postulate that FANCD2 mediates the processing of collapsed forks into HHR-proficient substrates. In fact, as mentioned before, defective Chk1 activation may indicate defective processing of DNA in the absence of FANCD2. Since FA patients are not normally exposed to ICLs agents, a major concern of clinical relevance is to identify life-threatening sources of stress in FA patients. The group of K. Patel has elegantly shown that aldehydes are an endogenous source of ICLs [77] and that the enzyme Aldh2 is essential to prevent the accumulation of aldehyde-derived ICLs [78]. Tissues with low levels of Aldh2, e. g. the hematopoietic linage, rely heavily on the FA pathway to process ICLs generated from endogenous aldehydes [79,80]. Hence, endogenous ICLs represent important triggers for oncogenesis in FA patients. But whether they represent the sole trigger for genomic instability in FA patients is still unresolved. While previous studies have proposed that the contribution of FANCD2 to the resolution of DSBs might be specifically linked to inter-strand ICLs [12,14,16,81], our report demonstrates that replication-coupled DSBs unrelated to ICLs may require FANCD2 for their repair through the HRR pathway. In addition, unanticipated HRR-independent functions of FANCD2 have been recently identified. Pioneer work from K. Schlacher and M. Jasin showed that FANCD2 and BRCA2/FANCD1 prevent degradation of nascent DNA in HU-treated cells [18,19]. It has also been shown that after HU, and in a core-independent manner, FANCD2 in concert with the Bloom helicase (BLM) restart stalled replication forks while suppressing origin firing [20,22]. In FANCD2 depleted samples, increased aberrant rearrangements of chromosomes were reported in [18,19] and increased frequencies of MN where reported in [20,22]. It is therefore possible that the defects in chromosomal integrity observed after UV irradiation are the indirect consequence of DSB-independent functions of FANCD2 at replicating DNA after UV irradiation. However, a number of evidences disfavour such hypothesis. First, the DSBs-independent contribution of FANCD2 after HU has been associated with persistent fork stalling [19], which is not frequent event after UV irradiation doses used in this study [25,26]. Second, the chromosomal integrity of FANCD2-depleted cells after UV is restored when NHEJ is silenced, therefore suggesting that the main function of FANCD2 is related to the processing of DSBs rather than to DNA replication events taking place prior to DSB formation. Third, the events taking place prior to DSBs processing after HU are independent of FANCD2 ubiquitination [82] whereas the UV-triggered events, which take place after DSB formation, are dependent on FANCD2 ubiquitination (see S7 Fig). Moreover, it is also reasonable to speculate that after HU treatment, the accumulation of at least some aberrations in FANCD2-depleted cells, e. g. the non-homologous exchanges [19] require the elimination of the FANCD2-mediated facilitation of DSB resolution by HRR (in addition to the disruption of FANCD2 functions at nascent DNA). Hence, while it is conceivable that during the replication of UV-damaged DNA FANCD2 participates in more than one (HRR-dependent and independent) process, results in Fig 6 demonstrate that the inhibition of NHEJ is a function of FANCD2, which must obligatorily be disrupted to generate many -if not all- the chromosome aberrations observed after UV irradiation. Remarkably, uncontrolled NHEJ at replication-coupled DSBs might also be the source of the chromosomal abnormalities reported in FANCD2-depleted samples subjected to replication-stressing agents such as HPV 16 E6/E7 expression [83], HU/APH treatments [19,21,84], PARP inhibition [85], R-loop accumulation [86], and dysregulated Pol κ recruitment to replication forks [87]. It is unclear to us why genomic stability but not cell survival is affected by FANCD2 depletion after UV irradiation. Similar results were reported after HU treatment [19]. It is possible that the DSBs generated by UV irradiation and HU are infrequent and therefore only tangentially contribute to cell death. Alternatively, while unresolved DSBs could be extremely toxic, their resolution, even when aberrant (e. g. in a FANCD2 depleted sample), may suffice to prevent cell death. Indeed, our data reveals multiple backup mechanisms that promote resolution of DSBs. Hence, when forks collapse, resolution mechanisms that promote cell survival may prevail even when genomic stability is compromised with multiple rearrangements. Our results suggest that low levels of replication-associated DSBs may be an important oncogenic factor if FANCD2 is not available to direct them into an error free pathway. FANCD2 is also required for the spontaneous levels of SCEs in uveal melanoma [88], thus we speculate that even during unperturbed replication FANCD2 regulates the pathway choice for DSBs repair. We propose a surveillance role for FANCD2 that is required to resolve replication-associated DSBs arising from any stress source and which might be relevant for the etiology of cancer in FA patients. The following cells were used: U2OS cells (ATCC), GM00637 (Coriell Repositories), FANCD2-deficient PD20 cells (GM16633—Coriell Repositories) and two reconstituted counterparts, PD20 + D2 (GM16634—Coriell Repositories, a microcell hybrid expressing low levels of wt FANCD2) and PD20 + D2O (PD20 cells expressing full-length FANCD2 cDNA), and PD20 K561R (overexpressed FANCD2 mutant with mutated K561 lysine). PD20, PD20 K561R and PD20 + D2 cells were a gift from J. Surralles (Universidad de Barcelona, Spain) and PD20 + D2O from T. Huang (New York University). All cells were grown in Dulbecco’s modified Eagle’s medium (Invitrogen) supplemented with 10% fetal calf serum. Transfections were performed using Jet Prime (Polyplus). GFP-Pol η was a gift from A. Lehmann. UVC irradiation was performed using a CL-1000 ultraviolet cross-linker equipped with 254 nm tubes (UVP) or a XX-15S UV bench lamp from UVP. For local irradiation, a polycarbonate filter with 5 μm pores (Millipore # TMTP01300) was positioned in direct contact with cells, which were then treated with 100 J/m2 -equivalent to a much lower dose than the one reported in [89]. siRNA duplexes (Thermo-Fisher Scientific) were the following: siFANCD2: 5-UUGGAGGAGAUUGAUGGUCUA-3 [90], siXRCC4: 5-AUAUGUUGGUGAACUGAGA-3 [91] siPol η: 5-CUGGUUGUGAGCAUUCGUGUA-3 has been recently described [92] and in our laboratory was designed by using the Invitrogen Block-iT RNAi Designer program validated with Dharmacon siRNA design software. siLuc: 5-CGUACGCGGAAUACUUCGA-3 [93]. For the immunodetection of FANCD2, Rad51,53BP1 and γH2AX, cells were fixed in 2% paraformaldehyde (PFA) /sucrose and permeablized with 0. 1% Triton X-100 in phosphate buffered saline (PBS). Well-assembled GFP- Pol η foci were quantified after fixation with ice-cold methanol followed by a 30-second incubation with ice-cold acetone as previously described by us [93]. EdU was detected following manufacturer’s instructions (Click-iT EdU kit– C10338). Blocking was performed overnight in PBS 2% donkey serum (Sigma). Coverslips were incubated for 1 h in primary antibodies: α FANCD2 (Novus), α Rad51 (Calbiochem), α γH2AX (Ser 139, Upstate), α 53BP1 (Santa Cruz). Secondary α-mouse/rabbit-conjugated Cy2/Cy3 antibodies (Jackson Immuno Research) and α -rabbit Alexa 488 (Invitrogen) were used. GFP-Pol η was detected by GFP auto-fluorescence. Nuclei were stained with DAPI (Sigma). Images were obtained with a Zeiss Axioplan confocal microscope or a Zeiss Axio Imager. A2. When quantifying GFP-pol η nuclear focal structures, cells with more than 10 foci were considered positive. When quantifying cells with Rad51 recruitment to locally irradiated areas of nuclei revealed by DAPI staining, only fields with γH2AX (+) staining were analyzed. Rad51 was always recruited to γH2AX (+) regions for all conditions tested. γH2AX staining was positive in 50% of the nuclei for all conditions tested. To quantify γ-H2AX intensity 100x images were analyzed with ImageJ. Approximately 30 pictures per condition were evaluated (300 cells); DAPI images were used as a pattern to define the position of nuclei on the images. The γ-H2AX intensity was determined in 300 nuclei/sample in arbitrary units, which were expressed as a fold increase with respect to the untreated control (siLuc non-irradiated). Western blots were performed using the following antibodies: α FANCD2 (Santa Cruz Biotechnology; FI17), α Ku70 (Santa Cruz Biotechnology; A9), α PCNA (Santa Cruz Biotechnology; PC10), α phospho- (S1981) -ATM (Millipore), α ATM (GeneTex 2C1), α phospho- (S345) -Chk1 (Cell Signalling), α Chk1 (Santa Cruz Biotechnology, G4), α p21 (Santa Cruz Biotechnology, C19), α p53 (DO-1 and 1801) and α γH2AX (Upstate). α phospho (S824) KAP1 (Bethyl Laboratories), α KAP1 (Bethyl Laboratories), α Pol η (Santa Cruz Biotechnology; H-300). Incubation with secondary antibodies (Sigma) and ECL detection (Amersham GE Healthcare) were performed according to the manufacturers' instructions. Western blot images were taken with Image QuantLAS4000 (GE Healthcare), which allows capture and quantification of images within a linear range. These images were then quantified with the ImageJ software. While U20S cells can be used in clonogenic assays, PD20 cells did not resist such harsh treatment in our experimental settings. Clonogenic assays performed in U2OS cells involved an initial siRNA transfection step in 35-mm dishes, followed by replating 200 cells per 60-mm plate (2 plates per condition) and UV irradiation 24 hours later. 8–10 days later, colony formation was visualized by crystal violet staining. Colonies with more than 40 cells were scored as positive. For PD20 (and U2OS) cells, a viability kit was used at earlier time points (up to 72 hours). Transfected or PD20 and PD20 + D2 cells were plated in 96-well plates; 24 hours later cells were UV irradiated or treated with Mitomicyn C (MMC, Roche). When using MMC, treatment was interrupted 15 hrs later and samples were washed and incubated with fresh growing medium. The analysis was performed at the indicated hours after release. PD20 cells were subjected to the Cell Viability Assay following manufacturer’s instructions (CellTiter-Glo Luminescent Cell Viability Assay G-7570, Promega). Cells were fixed with ice-cold ethanol and resuspended in PBS containing RNase I (100 mg/ml, Sigma) and propidium iodide (50 mg/ml, Sigma). Samples were subjected to fluorescence activated cell sorting (FACS, Calibur, Becton Dickinson), and data was analyzed using the Summit 4. 3 software (DAKO Cytomation). U2OS and PD20 cells were plated at low density, UV irradiated 24 hours later and incubated with cytochalasin B (4. 5 ug/ml, Sigma) for 40 h (U2OS) and 24 hrs (PD20). Cells were washed 1 min with hypotonic buffer (KCl 0. 0075 M), twice with PBS and fixed with paraformaldehyde (PFA) /sucrose 2% for 20 min. Phalloidin and DAPI staining served to visualize whole cells and nuclei respectively. 300 binucleated cells were analyzed and the frequency was calculated as MN/binucleated cells. Metaphase chromosome spreads were generated introducing minor modifications to protocols previously used by us [94]. Briefly, U2OS transfected cells were replated and UV irradiated (1. 5 J/m2). Before harvesting, cells were treated with Colcemid (0. 08 μg/ml, KaryoMAX, Invitrogen) for 20 h. Cell pellets were incubated in hypotonic buffer (KCl 0. 0075 M) at 37°C for 4 min, followed by fixation in Carnoy’s fixative (3: 1 methanol: glacial acetic acid). Cells were dropped onto slides and air-dried before staining with 6% w/v Giemsa in Sorensen’s buffer (2: 1 67 mM KH2PO4: 67 mM Na2HPO4, pH 6. 8) for 2 min. Samples were analyzed in an Applied Imaging Cytovision 3. 7. 50 metaphase spreads were used to quantify chromosomal gaps, breaks and exchanges. This protocol was set up to enrich samples with cells transiting the first cell cycle after UV irradiation. Transfected U2OS cells (with siLuc and siD2) were UV irradiated (1. 5 J/m2). To generate the differential staining of sister chromatids, cells were incubated with the thymine analogue 5-bromo-2´-deoxyuridine (BrdU, 20 μM, Becton Dickinson) for two complete cell cycles. Colcemid (0. 08 μg/ml, KaryoMAX, Invitrogen) was added 20 h before harvest. Metaphase chromosome spreads were prepared as mentioned above (see Chromosomal aberration analysis). Slides were air dried for 5 days, stained with Hoechst (5 μg/ml, Invitrogen), irradiated with a sun lamp (Ultra-Vitalux, OSRAM) for 7 min and finally stained with 6% w/v Giemsa in Sorensen’s buffer for 2 min. The treatment with Hoechst dye and Giemsa allows the newly synthesized DNA within a chromatid to be recognized, since BrdU incorporation results in much weaker staining. Sister-chromatid exchanges (SCE) were scored analysing chromosomes in 50 metaphase spreads. To prepare agarose plugs we used the protocol reported in [52] with minor modifications. Briefly, samples were UV irradiated, 6 or 24 h later 1 x 105 cells were melted into 1. 0% Pulsed Field Certified Agarose (Bio-Rad Laboratories). Agarose plugs were digested in 0. 5 M EDTA-1% N-laurylsarcosyl-proteinase K (1 mg/ml, Invitrogen) at 50°C for 48 h and washed four times in TE buffer and loaded onto a separation gel (1. 0% Pulsed Field Certified Agarose). Electrophoresis was performed on CHEF DR II equipment (Bio-Rad Laboratories) as previously described in [52]. A second electrophoresis protocol was also used [49], with minor modifications: 9 h, 120°, 5. 5 V/cm, 30–18 s switch time; 6 h, 120°, 4. 5 V/cm, 18–9 s switch time; 6 h, 120°, 4 V/cm, 9–5 s switch time, for 24 hr. A 2h-bleomycin (100 μg/mL, Gador) treatment was used as a positive control. Ethidium bromide–stained gels were visualized in a White Ultraviolet Transilluminator (UVP) or with Image Quant LAS4000, which allows capture and quantification of images within a linear range. PFGE images were then quantified with the ImageJ software. Cells were lysed and total RNA was extracted using Trizol Reagent (Invitrogen). 1 μg of total RNA was used as template for cDNA synthesis using ImProm-II Reverse Transcription System (Promega) and oligo-dT. Quantitative real-time PCR was performed in a MX3005P qPCR instrument (Stratagene) with Taq DNA polymerase (Invitrogen) and SyberGreen and ROX as reference dyes (Invitrogen). All amplification reactions approached 100% efficiency as determined by standard curves. Three independent biological samples were analyzed and one representative set of results is shown. Primers used for Quantitative Real Time PCR analysis: XRCC4 (f) 5′-AAGATGTCTCATTCAGACTTG-3′ (r) 5′ CCGCTTATAAAGATCAGTCTC-3′ [95]. GADPH: (f) 5’-AGCCTCCCGCTTCGCTCTCT-3’ (r) 5’-GAGCGATGTGGCTCGGCTGG-3’. [96] GraphPad Prism 5 software was used to analyse SCE, for cytogenetic experiments and foci formation experiments we used the Student' s t test. Other calculations and graphics were performed by using Microsoft Excel 2010.

Here we show that irradiation with low doses of UV light causes modest accumulation of replication-coupled double strand breaks (DSBs), i. e. collapsed forks. Remarkably, the Fanconi Anemia protein FANCD2 is central to prevent the aberrant processing of UV-triggered DSBs and the generation of micronuclei and chromosome fusions but is dispensable to modulate cell death. Specifically, FANCD2 promotes homologous recombination-dependent repair of UV-triggered DSBs, thus preventing their aberrant processing by non-homologous end joining. Hence, the homologous recombination-dependent tumor suppressor function of FANCD2 is not restricted to inter-strand crosslinks but instead extends to replication-coupled DSBs that arise from a broader range of genotoxic stimuli.

Abstract Introduction Results Discussion Materials and Methods

2016

Chromosomal Integrity after UV Irradiation Requires FANCD2-Mediated Repair of Double Strand Breaks

The use of the bacterium Wolbachia is an attractive alternative method to control vector populations. In mosquitoes, as in members of the Culex pipiens complex, Wolbachia induces a form of embryonic lethality called cytoplasmic incompatibility, a sperm-egg incompatibility occurring when infected males mate either with uninfected females or with females infected with incompatible Wolbachia strain (s). Here we explore the feasibility of the Incompatible Insect Technique (IIT), a species-specific control approach in which field females are sterilized by inundative releases of incompatible males. We show that the Wolbachia wPip (Is) strain, naturally infecting Cx. p. pipiens mosquitoes from Turkey, is a good candidate to control Cx. p. quinquefasciatus populations on four islands of the south-western Indian Ocean (La Réunion, Mauritius, Grande Glorieuse and Mayotte). The wPip (Is) strain was introduced into the nuclear background of Cx. p. quinquefasciatus mosquitoes from La Réunion, leading to the LR[wPip (Is) ] line. Total embryonic lethality was observed in crosses between LR[wPip (Is) ] males and all tested field females from the four islands. Interestingly, most crosses involving LR[wPip (Is) ] females and field males were also incompatible, which is expected to reduce the impact of any accidental release of LR[wPip (Is) ] females. Cage experiments demonstrate that LR[wPip (Is) ] males are equally competitive with La Réunion males resulting in demographic crash when LR[wPip (Is) ] males were introduced into La Réunion laboratory cages. These results, together with the geographic isolation of the four south-western Indian Ocean islands and their limited land area, support the feasibility of an IIT program using LR[wPip (Is) ] males and stimulate the implementation of field tests for a Cx. p. quinquefasciatus control strategy on these islands. The last few years have witnessed an increasing interest in the alpha-proteobacterium Wolbachia (Rickettsiales) for the biological control of insect pest populations [for reviews see [1]–[5]. Wolbachia is the most common intracellular bacterium yet described [6], [7], present in more than 65% of insect species and found in all major insect families [8]. Some medically important mosquitoes are naturally infected by Wolbachia, such as the common house mosquito Culex pipiens [9], [10] and the Asian tiger mosquito Aedes albopictus [11], or can otherwise be artificially infected, such as the yellow fever mosquito Ae. aegypti [12]–[14]. Wolbachia is vertically inherited from a female host to its progeny through the egg cytoplasm, males being a dead end in terms of transmission [4], [15]. Wolbachia is usually termed a ‘reproductive parasite’ in the sense that it optimizes its transmission by manipulating its host' s reproductive biology [15], [16]. In mosquitoes, Wolbachia induces a form of embryonic death called cytoplasmic incompatibility (CI) [9]. This phenomenon results from sperm-egg incompatibility occurring when Wolbachia-infected males mate with uninfected females or females infected with an incompatible Wolbachia strain [17]. Therefore, CI has been investigated as a mechanism to control field populations [1], [18], [19], [20]–[22], or to drive transgenes into field populations [2], [3], [10], [23]. In addition, recent investigations showed that Wolbachia can affect virus transmission both by reducing the lifespan of the infected vector and by interfering with the arthropod-borne parasite [14], [24]–[26]. Mosquitoes of the Cx. pipiens complex are of special interest for Wolbachia-based control strategies. The most common members of the complex are the subspecies Cx. p. quinquefasciatus (Say) and Cx. p. pipiens (Linnaeus) (also considered as true species, depending on the authors), representing the southern and northern mosquito populations, which are ubiquitous in tropical and temperate regions, respectively [27]. This mosquito is the main vector of lymphatic filarial in Comoros and Madagascar [28] as well as a known vector for many arboviruses worldwide [29]. This is the case, for example, of the West Nile Virus (WNV), recrudescent in Mediterranean countries [30], [31] and in the United States where thousands of cases have been identified in the last decade [32], [33]. This species also transmits the Rift Valley Fever (RVF) virus, currently expanding in the Indian Ocean [34], [35]. Members of the Cx. pipiens complex are naturally infected with different Wolbachia strains, referred as wPip strains. The prevalence is high in natural populations with wPip infections near to fixation [10], [36], [37]. Recent multi-loci typing approaches revealed that the wPip strains cluster into five distinct phylogenetic groups (referred as wPip-I to V) which form a robust monophyletic clade within the B group of Wolbachia [38]. The Cx. pipiens complex exhibits the largest variation of CI crossing types observed in arthropods thus far [39]–[43]. When Cx. pipiens individuals are infected by different Wolbachia strains (here arbitrarily named wPip (1) and wPip (2) ), their crosses can be (a) compatible and produce viable offspring; (b) incompatible in both directions and produce infertile eggs (a phenomenon called bidirectional CI); or (c) incompatible in one direction only (unidirectional CI, e. g. the cross between wPip (1) males and wPip (2) females is incompatible, while the reciprocal cross is compatible). The presence of incompatible wPip infections in the Cx. pipiens system makes unnecessary the artificial introduction of exogenous Wolbachia strains, and encourages the development of a Wolbachia-based control strategy. Here, we examined the feasibility of an ‘Incompatible Insect Technique’ (IIT) strategy targeting Cx. pipiens natural populations. IIT derives from the ‘Sterile Insect technique’ (SIT) notably used in the control of the New World screwworm Cochliomyia hominivorax [44]. In both SIT and IIT, mating of released sterilizing males with native females leads to a decrease in the females' reproductive potential and ultimately, if males are released in sufficient numbers over a sufficient period of time, to the local elimination or eradication of the pest population [3], [20], [22], [45]. In the SIT program, males are sterilized with irradiation or chemicals, which might weaken the fitness of sterilized insects, making them less competitive than field males for mating [46], [47]. In the IIT strategy, Cx. pipiens males are infected by a wPip strain incompatible with the wPip strain (s) infecting field females. In this case, the released incompatible males are not expected to suffer any reduction in mating when competing with field males. The IIT strategy has been successfully applied in a field trial assay targeting Cx. pipiens populations in Burma [48] as well as in cage experiments with the Polynesian tiger mosquito Ae. polynesiensis [49] and the medfly Ceratitis capitata [19]. We focused on natural populations of Cx. p. quinquefasciatus collected on five islands in the south-western Indian Ocean (SWIO): La Réunion, Mauritius, Mayotte, Madagascar and Grande Glorieuse. Prior studies have demonstrated that Cx. p. quinquefasciatus lines from La Réunion are infected with closely related wPip strains which express complete CI (ca. 100% embryo mortality) with Cx. pipiens lines from distant geographic areas and infected by genetically different wPip strains [43]. The Cx. p. pipiens Is line from Turkey, infected by the wPip (Is) strain, is of particular interest: all crosses between Is males and females from La Réunion are incompatible and almost all reciprocal crosses are incompatible as well [43]. This complete bidirectional CI makes the wPip (Is) strain a good candidate for an IIT program. In this study, we obtained robust data that encourage the use of wPip (Is) -infected Cx. p. quinquefasciatus males in an IIT program which could be implemented on the SWIO islands. First, the regional genetic diversity of wPip infections is low as all identified wPip strains belong to the wPip-I group; this indicates that immigration of mosquitoes into the controlled area is unlikely to introduce a new wPip strain compatible with wPip (Is) -infected males. Second, the wPip (Is) strain, from the wPip- IV group, was introduced into the nuclear background of Cx. p. quinquefasciatus mosquitoes, leading to a line (LR[wPip (Is) ]) expressing complete CI with wild females sampled from all 5 SWIO Islands. Last, CI properties expressed by this line are optimal as (i) there is no effect of males ageing on CI expression, (ii) LR[wPip (Is) ] males show similar body size and longevity as males from La Réunion Island, suggesting good competitiveness of incompatible males vs. wild males, which was further confirmed in cage confrontations and (iii) LR[wPip (Is) ] mosquitoes are mainly bidirectionally incompatible with La Réunion, Mauritius, Mayotte and Grande Glorieuse field mosquitoes: this lowers the risk of Wolbachia replacement possibly induced by accidental releases of LR[wPip (Is) ] females. Two laboratory lines of Cx. pipiens mosquitoes naturally infected by Wolbachia were used in the experiments: the isofemale line Is, a Cx. p. pipiens line from Turkey infected by the wPip (Is) strain, and the Cx. p. quinquefasciatus LR line, infected by the wPip (LR) strain, and established from several hundred field-caught larvae in La Réunion island (Table 1 and Figure 1). In addition, one uninfected line, LR-TC, was generated by curing Wolbachia of mosquitoes from the LR line with antibiotic, following the protocol described in [50]. Briefly, ca. 5,000 LR larvae were reared for three generations in a solution containing tetracycline hydrochloride at concentrations of 10−4,2×10−4 and 4×10−4 M for the first, second and third-instar larvae, respectively. Mosquitoes from LR-TC were next reared for at least two generations in the absence of tetracycline before experiments, to prevent any possible side-effects of the treatment. Field Cx. p. quinquefasciatus larvae and pupae were collected during the summers 2007–2011 in 29 natural breeding sites on five islands of the Indian Ocean: La Réunion (16 populations), Mauritius (four populations), Mayotte (three populations) Madagascar (five populations) and Grande Glorieuse (one population) (Table 1 and Figure 1). Specimens were brought to the laboratory for emergence and identification. Individuals were either directly stored in 70% EtOH for molecular analyses or kept alive for crossing experiments. All mosquitoes were reared in 65 dm3 cages kept at ca. 25±2°C with 12 h/12 h light/dark cycle. Larvae were fed ad libitum with a mixture of shrimp powder and rabbit pellets, and adults with a honey solution. Mosquito DNA was extracted using a CetylTrimethylAmmonium Bromide (CTAB) protocol [51]. The wPip infections were characterized through the analysis of one Wolbachia marker, the ankyrin domains encoding gene, ank2 [52] (primers are listed in Table S1). This marker differentiated wPip strains from groups I and IV on the basis of the size of the PCR amplified fragments: 313 bp and 511 bp fragment for group I and IV, respectively. For field samples, the ank2 PCR products from two specimens per sample site were sequenced to confirm their identity with La Réunion ank2 allele [Genbank AM397068; [43]]. The examination of the mosquito nuclear genome was assessed by PCR/RFLP tests based on Cx. pipiens ace-2 and Ester2 genes (primers are in Table S1). The ace-2 gene is located on chromosome I and encodes acetylcholinesterase 2 (AChE2) [53]. The Ester2 gene is located on chromosome II and encodes a carboxylester hydrolase [54]. A PCR/RFLP test on ace-2 using the ScaI restriction enzyme (37°C, 3 hours; see [55]) allows the discrimination between the Is (two fragments: 230 and 470 bp) and the LR (three fragments: 120,230 and 350 bp) nuclear genomes. We developed a PCR/RFLP test on Ester2 using the AvaII enzyme (37°C, 3 hours) that also generated different restriction fragments for the Is (three fragments: 37,519 and 544 bp) and LR (four fragments: 91,176,313 and 520 bp) nuclear genomes. All PCRs were performed with ca. 20 ng of genomic DNA solution in a 40 µl final volume reaction for 35 cycles (94°C, 5 min; 94°C, 30 sec; 52°C, 30 sec; 72°C, 1 min). Direct sequencing of PCR products was performed on an ABI Prism 3130 sequencer using the BigDye Terminator Kit (Applied Biosystems) after purification with the QIAquick gel extraction kit (QIAGEN, Valencia, CA). Sequence alignment and analyses were done using MEGA software [56]. The cytoplasm of the Is line, including the wPip (Is) strain, was introduced into the LR nuclear background through eight generations of backcrossing, a procedure that should result in at least 99% genome replacement of the Is line by the LR nuclear genome. A first cross was performed using 200 virgin Is females and 250 LR-TC males. For the following generations, 200 hybrid females were backcrossed with 250 LR-TC males. Using this protocol, we obtained the LR[wPip (Is) ] line which carries the LR nuclear genome and the wPip (Is) strain. We examined the crossing relationships between mosquito lines through crossing experiments. Mass crosses were carried out using 35–200 two-day-old males and an equivalent number of females that had been individually separated at the pupal stage (age was assessed from the emergence of adults; day 0 = emergence). We also tested the effect of male aging on CI by comparing crossing relationships of young males (two-day-old) to that of older males (24-day-old). For all crosses, females were allowed to blood feed 5 days after caging. Egg-rafts were collected and stored separately until hatching at 25°C±2°C. Hatching rates (HR) were scored 72 h after egg-raft collection to determine the CI phenotype. All unhatched egg-rafts were checked for fertilization through observation of embryonic development following the procedure of [57]. The longevity of the LR[wPip (Is) ] and LR males was compared. We obtained males from larvae reared in standardized laboratory conditions at ca. 25°C±2°C. For each line, three containers containing 300 first-instar larvae with 1 L of water were set up. The water of each container was changed every two days and food provided ad libitum. Pupae were randomly sampled from the three containers to minimize possible rearing bias. Pupae were placed separately in 5 mL vials for emergence. Freshly-emerged males were kept in their vials until they died, and mortality was checked twice a day. No food was provided to the adults but they had access to the water in their tube. Survival data were fitted to the Cox proportional hazards models (coxph, survival package) [58] and a ratio for each line was estimated as their instantaneous risk of death relative to each other. These analyses were performed using R software (www. r-project. org). One posterior leg was removed on dead specimens and the tibia was measured with a micrometer (NIKON Digital Counter CM-6S). Four cages were set up to compare the mating performance of both LR[wPip (Is) ] and LR males. Each cage contained an equal number of two-day-old virgin LR females and LR males (1∶1), as well as different numbers of two-day-old virgin LR[wPip (Is) ] males so that different ratios of the three types of mosquitoes could be tested (1∶1∶0,1∶1∶1,1∶1∶5 and 1∶1∶10). Thus the total number of adults for each of these confrontations was 200,300,350 and 600, respectively. For each confrontation, all the mosquitoes were introduced into the cage at the same time. Females were allowed to blood feed five days after caging and their egg-rafts were collected daily to score HR. To assess the stability of the expression of CI over the mosquito lifespan, a second blood meal was given to females 15 days after the first one, and new collections of egg-rafts were then made. We first examined the genetic diversity of wPip strains found in natural populations of Cx. p. quinquefasciatus from La Réunion, Mauritius, Mayotte, Madagascar and Grande Glorieuse. The main purpose of this investigation was to assess the possibility of controlling mosquito populations in each of these four islands with wPip (Is) -infected males. We examined 650 Cx. p. quinquefasciatus field specimens from 29 populations: La Réunion (16 populations, n = 384 individuals), Mauritius (4 populations, n = 91 individuals), Mayotte (3 populations, n = 69 individuals), Madagascar (5 populations, n = 105 individuals) and Grande Glorieuse (1 population, n = 24) (Table 1 and Figure 1). The genotyping of wPip infections in these samples was performed using only the ank2 gene which was recently shown to discriminate wPip strains into five distinct phylogenetic groups (referred as wPip-I to wPip-V) [38]. PCR assays using ank2 indicated the occurrence of wPip infection in all Cx. p. quinquefasciatus field specimens, as observed in other geographic areas for this species [10], [36], [37], and all shared the same ank2 allele as indicated by the length of ank2 PCR products (313 bp). This similarity was further confirmed by sequencing the ank2 gene of two individuals per population from Mauritius, Mayotte, Madagascar and Grande Glorieuse. All sequences were found to be strictly identical to that found in the wPip strains infecting all 10 laboratory isofemale lines from La Réunion and to other wPip strains belonging to the wPip-I group [38]. This result shows that wPip strains from La Réunion, Mauritius, Mayotte, Madagascar and Grande Glorieuse are genetically closely related and are genetically different from the wPip (Is) strain belonging to the wPip-IV group. Males from the Is line belong to Cx. p. pipiens subspecies and may not be optimally adapted to the tropical environment of the Indian Ocean where Cx. p. quinquefasciatus is found. More specifically the two subspecies are known to differ by behavioral and physiological characters including mating behavior [27]. To circumvent this problem, we introduced the wPip (Is) strain into the Cx. p. quinquefasciatus nuclear background from La Réunion. First a LR line was established from a large number (>5,000) of field-caught Cx. p. quinquefasciatus from three localities of La Réunion in order to have a good representation of the local genetic diversity. This line was then cured of its Wolbachia by tetracycline treatment of larvae during three generations (LR-TC line). Finally wPip (Is) from the Is line was introduced into the nuclear background of the LR-TC line by successive backcrossing. The LR[wPip (Is) ] line thus created shares the same nuclear genetic background as the LR line but is infected by the wPip (Is) strain (Figure S1). This was verified by PCR/RFLP tests on ace-2 and Ester2 Cx. pipiens nuclear genes (Figure S2A and S2B) and by analyzing the allelic profiles of the ank2 gene of the infecting Wolbachia (Figure S2C). Crossing experiments between LR[wPip (Is) ] and Is lines were conducted to check that Cx. p. quinquefasciatus nuclear background has not altered the CI phenotype of the wPip (Is) strain. This aspect needs to be investigated since the host nuclear genome has been reported to affect the penetration of the CI phenotype induced by a Wolbachia strain [59]–[61]. Our data show that both lines behave similarly: LR[wPip (Is) ] and Is showed bidirectional CI with LR while LR[wPip (Is) ] and Is were mutually compatible (Table 2). The intensity of CI was very high, with 98–100% of the embryos that did not hatch in incompatible crosses. In addition, crosses between infected and uninfected lines showed unidirectional CI: males from all infected lines (LR[wPip (Is) ], Is and LR) induced complete CI (100% embryo mortality) when crossed with uninfected females (LR-TC), the reverse crosses (i. e. uninfected males and infected females) being always compatible. Overall, no significant difference of hatching rate (HR) was found when the LR[wPip (Is) ] and Is lines were compared (Wilcoxon test; all P>0. 14). This shows that the CI phenotype of the wPip (Is) strain was not altered by the LR genetic background, and that the CI phenotype is controlled by the wPip infection rather than by nuclear genes, which is in accordance with most studies involving species of the Cx. pipiens complex [43], [62]. The effect of male ageing on CI intensity was also tested as, in a few host species including some mosquitoes, CI intensity has been shown to decrease with male aging [63]–[67]. Such an effect could impede the use of LR[wPip (Is) ] males to sterilize field females. To investigate this aspect, we crossed two-day and 24-day old LR[wPip (Is) ] males with two-day old LR females. No viable embryo was obtained in incompatible crosses with both young and old LR[wPip (Is) ] males (Table 3). Thus CI is expressed with the same intensity throughout the LR[wPip (Is) ] males' lifespan, a result also observed in diverse Cx. pipiens laboratory lines [10], [68]. LR[wPip (Is) ] males were crossed with field females from five populations: Samuel (La Réunion; n = 75 females), Salines (Mauritius; n = 37), Tsoundzou (Mayotte; n = 75) Mada (Madagascar; n = 44) and Grande Glorieuse (Grande Glorieuse; n = 97 females). All crosses were incompatible, displaying >99% embryo mortality (Table 4). Thus, LR[wPip (Is) ] males express high CI intensity with field females from the four islands, as observed with females of the LR line. Crossing relationships between LR[wPip (Is) ] females and field males were also investigated to determine how the LR[wPip (Is) ] line may evolve in Cx. p. quinquefasciatus field populations in the case of accidental release of LR[wPip (Is) ] females. LR[wPip (Is) ] females were incompatible with all males from Samuel (n = 36 males), Salines (n = 37) and Grande Glorieuse (n = 40) (Table 4). This shows that LR[wPip (Is) ] expresses bidirectional CI with field specimens from these populations. However, males from Tsoundzou (n = 16) were polymorphic for their CI properties, the majority (n = 14) expressing complete CI with LR[wPip (Is) ] females and a few (n = 2) being compatible (HR = 0. 895±0. 035) (Table 4). This shows that LR[wPip (Is) ] expresses either bidirectional CI or unidirectional CI with field specimens from Tsoundzou. Thus, two crossing types coexist in Mayotte, but it is likely that the bidirectional CI crossing type is the most frequent one. Males from Mada were also polymorphic for their CI properties but, in contrast to Tsoundzou males, most Mada males were compatible with LR[wPip (Is) ] females (n = 18, HR = 0. 804±0. 283) while only two males expressed CI. So the unidirectional CI type was the most frequent in the Mada population. Inferior competitive ability of LR[wPip (Is) ] males compared with field males may limit the efficiency of an IIT program. Thus, the performances of LR[wPip (Is) ] and LR males, reared in standardized conditions, were examined for different life history traits. Longevity of LR[wPip (Is) ] and LR males (n = 154 and n = 238, respectively) was investigated in conditions where males had to survive by metabolizing nutritional reserves accumulated during their larval life (see material and methods) [69]. No significant difference was found (χ2 = 0. 04, P = 0. 84; Figure 2), suggesting that the infection by wPip (Is) did not alter mosquito metabolism. There was also no significant difference between LR[wPip (Is) ] and LR male tibia length (n = 30 and n = 30; Wilcoxon two-sided test, P = 0. 34; Figure 3), a parameter known to be positively correlated with mosquitoes' adult size and reproductive success [70]. This suggests that LR[wPip (Is) ] and LR males most probably exhibit similar mating performance. To further test this assumption, mating competition between LR[wPip (Is) ] and LR males was investigated in laboratory cages. Four cages containing different ratios of LR females to LR males to LR[wPip (Is) ] males (1∶1∶0,1∶1∶1,1∶1∶5 and 1∶1∶10) were set up. Note that as CI occurring between LR[wPip (Is) ] males and LR females is complete, it was easy to distinguish egg-rafts produced from compatible (LR males×LR females) or incompatible (LR[wPip (Is) ] males×LR females) crosses. Two successive collections of egg-rafts were obtained for each cage by giving females two distinct blood meals. There was no significant variation in the proportion of incompatible egg-rafts between the first and the second series of egg-rafts (Fisher exact test, all P>0. 57). As expected, when only LR males were present, all the egg-rafts were compatible (Table 5). In the other cages, no significant difference between LR[wPip (Is) ] and LR males' mating capacity was found. Indeed, the number of incompatible egg-rafts observed was not significantly different from expected values assuming an equal competitiveness of LR[wPip (Is) ] and LR males and random mating (Binomial test, all P>0. 18; Table 5). For instance, with an identical ratio of LR[wPip (Is) ] and LR males (1∶1), ca. 50% of the egg-rafts produced by LR females were incompatible. When the LR[wPip (Is) ] males' ratio was higher than that of LR males, i. e. at 1∶5 and at 1∶10, we observed ca. five and ten times more incompatible egg-rafts than compatible ones. Taken together, these results showed that LR[wPip (Is) ] males are as fit as LR males, at least in our laboratory conditions. These experiments also established that LR females cannot discriminate between compatible LR males and incompatible LR[wPip (Is) ] males, a result consistent with previous observations of random mating between Cx. pipiens mosquitoes infected by incompatible Wolbachia strains [37], [48], [71]. The study presented here supports the feasibility of an IIT strategy using the LR[wPip (Is) ] males and targeting field Cx. p. quinquefasciatus populations, a species of medical and veterinary concern in the SWIO islands. This method now needs to be further tested in semi-field conditions in order to optimize several key parameters, i. e. the number of males to be released as well as the timing of releases. Recently, new semi-field cages were developed to measure the impact of the life-shortening Wolbachia wMelPop strain on populations of Aedes aegypti [80]. Such cages provide a realistic transitional platform between laboratory and field conditions. The risk of accidental releases of females needs also to be limited by developing an efficient sexing method to prevent any unintentional Wolbachia replacement.

Mosquitoes of the Culex pipiens complex are important vectors of human pathogens including filarial parasites and many currently expanding arboviruses. The absence of effective vaccines and the evolution of insecticide resistance stress the urgent need for the development of novel control strategies. One strategy that is receiving increasing attention is based upon the use of the intracellular bacteria Wolbachia, which induce a form of sterility known as cytoplasmic incompatibility in mosquitoes. Here, we show that a Wolbachia strain, named wPip (Is) and naturally infecting Cx. p. pipiens from Turkey, can be used in the Incompatible Insect Technique (IIT) to sterilize Cx. p. quinquefasciatus females from several islands of the southwestern Indian Ocean (SWIO). The wPip (Is) strain was introduced into SWIO Cx. p. quinquefasciatus nuclear background leading to the LR[wPip (Is) ] line. Males from this latter line were found to sterilize all wild females tested, and no difference in mating competition was observed between LR[wPip (Is) ] and wild males. These results encourage the development of an IIT program based on the wPip (Is) strain to control mosquito populations in the SWIO.

Abstract Introduction Materials and Methods Results Discussion

vector biology biology microbiology

2011

Cytoplasmic Incompatibility as a Means of Controlling Culex pipiens quinquefasciatus Mosquito in the Islands of the South-Western Indian Ocean

Rabies is a fatal zoonosis that still causes nearly 70,000 human deaths every year. In Europe, the oral rabies vaccination (ORV) of red foxes (Vulpes vulpes) was developed in the late 1970s and has demonstrated its effectiveness in the eradication of the disease in Western and some Central European countries. Following the accession of the three Baltic countries—Estonia, Latvia and Lithuania—to the European Union in 2004, subsequent financial support has allowed the implementation of regular ORV campaigns since 2005–2006. This paper reviews ten years of surveillance efforts and ORV campaigns in these countries resulting in the near eradication of the disease. The various factors that may have influenced the results of vaccination monitoring were assessed using generalized linear models (GLMs) on bait uptake and on herd immunity. As shown in previous studies, juveniles had lower bait uptake level than adults. For the first time, raccoon dogs (Nyctereutes procyonoides) were shown to have significantly lower bait uptake proportion compared with red foxes. This result suggests potentially altered ORV effectiveness in this invasive species compared to the red foxes. An extensive phylogenetic analysis demonstrated that the North-East European (NEE) rabies phylogroup is endemic in all three Baltic countries. Although successive oral vaccination campaigns have substantially reduced the number of detected rabies cases, sporadic detection of the C lineage (European part of Russian phylogroup) underlines the risk of reintroduction via westward spread from bordering countries. Vaccine induced cases were also reported for the first time in non-target species (Martes martes and Meles meles). Rabies disease is a fatal mammalian encephalomyelitis caused by the rabies virus of the genus Lyssavirus (family Rhabdoviridae) [1]. The virus is distributed worldwide, with the exception of the Antarctic, Australia and several islands and although all species of mammals are susceptible to this virus, it infects principally carnivores and bats [2]. In Europe, the genus lyssavirus evolves through five virus species (four of them circulate in bats only): the classic rabies virus (RABV) affecting non-flying terrestrial mammals only, the european bat lyssaviruses type 1 and type 2 (EBLV-1 and EBLV-2) and the more recently detected Bokelo bat lyssavirus (BBLV) and Lleida bat lyssavirus not yet taxonomically assessed [3]. RABV has spread in Europe since antiquity as a dog and wolf-mediated disease [4]. In the 1940s, likely due to spillover from domestic animals, a new epizootic maintained by a single species, the red fox, emerged in Eastern Europe with an assumed ground zero in Kaliningrad [5]. The front moved from Poland to Germany spreading through Europe with a speed of approximately 30–60 km per year, reaching France in 1968 and Italy in 1980 [6]. Large rivers, lakes and high mountain chains acted as obstacles to the spread; bridges facilitated the crossing of rivers. Intensive fox destruction campaigns alone cannot stop the spread of the virus [7], prompting oral rabies vaccination (ORV) programs that rapidly proved to be the only efficient technique for controlling the disease. The first ORV field trial was conducted in 1978 in Switzerland [8] and was gradually extended to surrounding countries, such as Belgium, France and Germany. In the 1980s, fox rabies control in European Union became a public health issue. Since 1989, the European Commission has provided funding to Member States for national eradication programmes, thereby improving surveillance and encouraging regular implementation of oral vaccination campaigns on large scales in coordination with neighbouring countries. This strategy leaded to the successful elimination of rabies in most Western and Central European countries [9,10]. In Europe, approximately half of the historical rabies endemic countries are now free of rabies (Austria, Belgium, Czech Republic, Finland, France, Germany, Italy, Luxembourg, Switzerland and the Netherlands). In the Baltics, the three countries were recently officially recognized free of rabies according to OIE (World Organisation for Animal Health) criteria [11–13]. In the last three years, some sporadic cases have been reported in some countries (Bulgaria, Hungary, Slovakia and Slovenia) and the disease is still endemic in several Eastern European countries (eastern Poland, Romania, Ukraine, Belarus and Russia, source: http: //www. who-rabies-bulletin. org/Queries/Surveillance. aspx). In the Baltic States, represented by Lithuania, Estonia and Latvia, sylvatic rabies emerged in the 1950s-1960s [14]. Since this time, a surveillance of the disease were progressively implemented and positive cases have been observed mainly in red foxes and raccoon dogs [15–17]. Although the red fox is known to be highly susceptible to RABV and is the main reservoir and vector of rabies throughout Europe, the Baltic countries has the particularity to host a second vector and reservoir, i. e. the raccoon dog [14]. Raccoon dog is one of the most successful alien carnivores in Europe. Native to East of Asia, this species was introduced in the eastern part of Russia via fur industry during the first half of the 20th century and has spread throughout Europe, becoming common in the Baltics and some other northeastern European countries. After it was first observed in the 1950’s in the Baltics, ten years were required to colonize the entire countries [18]. Foxes and raccoon dogs are both opportunistic omnivores, often share the same habitats and overlap their home ranges increasing the probability of contacts between the two species. Moreover, their combined densities could allow rabies epizootics to persist in a certain area [19]. The existence of this important rabies transmitter in this area challenged health authorities and questioned on its potential impact on the success of conventional ORV method used to control rabies in Western Europe. ORV programs were experimented differently according to the Baltic State. While no ORV was implemented in Estonia until 2005, in Latvia, ORV was firstly initiated in 1991 using chicken head vaccine baits. ORV using manufactured baits started in 1998 and has been performed twice a year since 1999, but regular purchase of the necessary amount of vaccine baits for annual nationwide vaccination was not possible because of financial reasons. The vaccination area was enlarged every year to cover the whole territory by 2001–2003 and vaccines were distributed manually [20]. In Lithuania, ORV was tested for the first time in 1983 with fish or meat baits containing a vaccine made of a derived ERA (Evelyn Rokitnicki Abelseth) laboratory fixed virus strain produced in Russia. In 1993 ORV was occasionally assessed on three districts [21]. Between 1995 and 2000, following the Lithuanian National Rabies prevention programme, ORV was performed generally manually and a large range of vaccines was used (Street Alabama Duffering (SAD) Bern, SAD P5/88 (Rabifox), (Street Alabama Gif (SAG) 1) over variable geographic areas. Following the accession of three Baltic countries to the European Union in 2004, subsequent financial support allowed the implementation of regular oral vaccination campaigns in the three countries since 2006 and ORV are still ongoing. This paper reviews ten years of surveillance efforts and oral vaccination campaigns conducted in the frame of European Commission programmes. Through the epidemiological analysis of rabies surveillance in these countries and an in-deep analysis of the ORV monitoring results, this paper emphasizes determinants of success and draws lessons for the future. These findings could provide valuable insights into the strategy required for rabies elimination and may help guide future implementation of oral vaccination programmes. Covering approximately 175,000 km2, the Baltic States lie in the northeastern part of Europe and comprise the countries of Estonia (45,227 km2), Latvia (64,589 km2) and Lithuania (65,303 km2) (Fig 1). The Baltic States are bounded on the west and north by the Baltic Sea, which gives the region its name, on the east by Russia (511 km of common border), on the southeast by Belarus (818 km), and on the southwest by Poland (104 km) and an exclave of Russia named Kaliningrad (255 km). The topography of this area is relatively flat (culminating points in the three countries are around 300 m), characterized by numerous lakes and ponds, especially in the north, and hills in Lithuania. The most commonly encountered landscape is the temperate forest covering between 35 and 50% of the territories. All suspect non-flying mammals exhibiting clinical signs suggestive of rabies or showing abnormal behaviour, animals found dead in the field including road kills and those to which humans have been exposed (bites, scratches, licking of wounds or contamination of mucous membranes with saliva) are defined as indicator animals and are submitted for diagnosis [19]. The sampling scheme focusing on these animals, covering the whole country territory, is herein defined by expert committees of the WHO (World Health Organization) and EFSA (European Food Safety Authority) as the surveillance system [2,19]. All collected samples were shipped and analyzed in the respective National Reference Laboratories of each Baltic country (Estonian Veterinary and Food Laboratory for Estonia; Institute of Food Safety, Animal Health and Environment" BIOR" for Latvia and National Food and Veterinary Risk Assessment Institute of Lithuania for Lithuania). Brain tissues were analyzed for viral antigens using the Fluorescent Anti-body Test (FAT) which is the gold standard technique for rabies diagnosis [22,23]. For all three countries, FAT-negative results of animals involved in human exposure and FAT-inconclusive results were confirmed using the rabies tissue culture infection test (RTCIT) [24], Reverse Transcription Polymerase Chain Reaction (RT-PCR) [25] or Real-Time Polymerase Chain Reaction (RT-qPCR) [26,27]. The first wildlife ORV campaign in Estonia was organized in autumn 2005 and covered 57% of Estonian lands in the northern part of the country as part of a PHARE Twining Light Project (Fig 1) [16]. Vaccination programmes covering the entire territory (excluding urban areas, roads, water bodies and wet fields) representing approximately 43,000 km2 were carried out from 2006 to 2010. Bait distribution was performed twice a year, in spring (May, early June) and in autumn (September, October) as recommended by WHO and EFSA [2,19]. Baits were distributed at a rate of 20 baits per km2 using small fixed-wing aircraft flying at an altitude of 100–150 m, speeds of 150–200 km/h and in parallel flight lines (global positioning system (GPS) routes followed by the plane) distanced of 600 m [16,28]. Since spring 2011, ORV campaigns have been conducted only in a buffer zone of 9,325 km2 adjacent to neighbouring infected countries (Russia and Latvia) to ensure a sufficient level of immunity among raccoon dog and red fox populations. No automatic dropping device was used in the airplanes and no additional manual distribution was carried out in the field. A single vaccine bait type was selected through a tendering process, the modified live attenuated SAG2 vaccine (RABIGEN, Virbac Laboratories, France) (Fig 2). In Latvia, following a PHARE Twinning Light project, ORV campaigns were carried out in 2005 for the first time via aircraft in half of the country (the size of vaccination area was 28,000 km2 and it was delimited by natural barriers) twice a year with 23 baits per km2. Starting from 2006, two vaccination campaigns were implemented in the entire territory (64,589 km2) (except in 2008 and autumn 2011 when ORV campaigns were incomplete and in spring 2014 where no ORV was carried out). Since 2006, between 21 and 31 baits per km2 have been distributed using flight line distances of 1000 m until 2008,1000 m and 500 m alternately between 2008 and 2011, and 500 m since 2011. An automatic dropping device has been used since 2007 to distribute the baits. The type of vaccine purchased varied according to the procurement procedure. In general, two vaccines were used within the period 2005–2011—SAD B19 vaccine (FUCHSORAL, IDT Biologica GmbH, Germany) and SAD Bern (LYSVULPEN, Bioveta, Czech Republic). Since 2012, only the Lysvulpen vaccine has been in use (Fig 2). In Lithuania, ORV programmes have been implemented since 2006 using aircrafts over the whole country (65,000 km2) except lakes, urban areas and the Ignalina nuclear power-station. The no-fly area surrounding the Ignalina power plant was covered by manual distribution of baits. Like in other countries, the vaccination strategy has been implemented biannually (one vaccination in spring between March and May and one vaccination in autumn between October and December). Parallel flight lines generally separated by 1000 m (since 2011,500 m in areas on the Belarus border) at an altitude of 150–200 m and speed 150–200 km/h were used to distribute 20 baits per km2 [29] (Fig 2). Since 2006, only the Lysvuplen vaccines have been distributed except in 2011 and 2012 when Fuchsoral vaccines were used. In addition to the sampling scheme designed for rabies surveillance, a second sampling plan defined as monitoring of ORV was set up in vaccinated areas to evaluate the efficacy of ORV campaigns in terms of bait consumption (bait-uptake) and herd immunity [2,19,30]. This sampling focused on the collection of animals (red foxes and raccoon dogs) targeted by oral vaccines. These animals sampled by hunter associations are therefore considered as not suspected for rabies. Herd immunity level was assessed by enzyme-linked immunosorbent assays (ELISAs) [31]. Two commercial anti-rabies ELISA kit were used within the study: the BioPro ELISA (BioPro, Czech Republic) and the Platelia Rabies II kit (Bio-Rad, France). Their technologies differ by their coating aspect. The BioPro ELISA is a blocking ELISA using the crude glycoprotein to coat the plates and a positivity threshold (expressed as a percentage of blocking) of 40% [32,33]. The Platelia Rabies II kit is an indirect test using a purified rabies glycoprotein for the coating [34]. Serum titers were expressed as Equivalent Units per milliliter (EU/mL) with a cut-off of positivity fixed at 0. 5 EU/mL in Estonia and Lithuania and 0. 125 EU/mL in Latvia. The BioPro Rabies ELISA Ab kit was used in Latvia only. Bait uptake was investigated by collecting red fox and raccoon dog jaws and by analysing the tetracycline (TTC) specific fluorescence in thin tooth sections under ultraviolet light [35,36]. Indeed, after its inclusion in the coating of the bait and its consumption by the targeted animal, the tetracycline molecule, used as a bait uptake marker, is incorporated into bones and teeth. This interaction creates a line that can be detected using epi-fluorescence microscopy. Each animal sampled for monitoring were analyzed for both serological analysis and tetracycline detection when possible (depending of the organs let intact by the shot of the hunter). Studied animals from surveillance and monitoring scheme were originated from the field, died of natural causes and during the hunting/vaccination program developed and launched by the ministry of each country. These sampling processes were realised in compliance with the legislation of each country and under the recommendations of international institution (WHO [2] and EFSA [19]). In Europe, such process does not require any specific ethical approval as animals are received only dead in laboratories. Hunting plans are organised in the frame of control programmes of the disease and organised by Member States. A panel of 165 field rabies viruses was collected in Baltic countries between 2004 and 2013 for this study. The isolates investigated from domestic and wild animals were extracted from brains of animals samples in Estonia (n = 43), Latvia (n = 42) and Lithuania (n = 80). The samples were isolated from 12 different wild and domestic animal species: Nyctereutes procyonoides (65), Vulpes vulpes (44), Canis familiaris (14), Bos taurus (11), Felis catus (ten), Procyon lotor (four), Meles meles (three), Equus caballus (two), Martes martes (two), Ovis aries (one), Canis (one), Lynx lynx (one) and six non-determined species (S1 Table). The samples were initially tested using the FAT prior to genetic characterization [23]. For all the samples, forward and reverse sequences were assembled and edited using the ContigExpress program of Vector NTI software, version 11. 5. 3. (Invitrogen, France). Alignments were edited using Genedoc software, version 2. 7. 000 [41]. The same software was used to translate the gene sequence. Percentage identities and similarity scores were determined using the BIOEDIT program version 7. 2. 5. [42]. After the alignment of sequenced amplified PCR products, 106 identical sequences (56 from Lithuania, 23 from Latvia and 27 from Estonia) showing 100% nucleotide identity for the N gene (460 nt) were removed from the phylogeny. Fifty-nine partial N gene sequences (24 from Lithuania, 19 from Latvia and 16 from Estonia) were available for subsequent analysis. The dataset contained 93 sequences (361 nucleotides, positions 109 to 470 compared with the challenge virus standard (CVS) -11 strain GenBank no. GQ918139). Fifty nine representative Baltic samples, eight isolates from neighbouring countries (six from Poland and two from Russia), two from Ukraine, seven from Europe, two fixed strains (D42112 and HQ829841), three representatives of rabies vaccine strains (EF206708, EF206709 and EF206719), six referenced Artic and Artic-like isolates and six reference strains used as outgroup were included in the dataset (S1 Table). A total of 24,919 animals were diagnosed for rabies from 2005 to 2014 in the Baltics. Around 70 to 80% of all detected positive cases were found in red foxes and raccoon dogs (For Estonia, 35% foxes and 48% racoon dogs; for Latvia, 40% foxes and 30% raccoon dogs; for Lithuania 31% foxes and 40% raccoon dogs). In the three countries, the maximum number of detected rabies cases was observed during the 2005–2006 period (Fig 2). The highest number of detected cases was recorded at the same semester of the implantation of the first ORV in Estonia and Latvia, while one semester after the first ORV in Lithuania. The ORV induced indisputably a decrease of the number of positives cases in the three countries (excepted in Lithuania between the second semester of 2006 and the first semester of 2007). Regarding the maximum number of cases observed in each country, 90% reduction of detected cases was reached after two ORV campaigns in Estonia, eight in Latvia and four in Lithuania. The last rabies case (field strain) was notified in 2011 in Estonia, in 2012 in Latvia and in 2013 in Lithuania. When starting ORV, surveillance effort (number of indicator animals sampled per 100 km2 in the whole country) ranged from 1. 7 to 1. 3 in Estonia (2005 and 2006), 1. 7 to 1. 6 in Latvia (2005 and 2006) and 5. 8 in Lithuania (but caution must be taken in the interpretation of this number because animals collected for monitoring were also included for this latter country). Since no rabies cases were detected, the pressure of surveillance appeared also comparable between the three countries ranging from 0. 42–0. 30 in Estonia (2012–2014), 0. 42–0. 30 in Latvia (2013–2014) and 0. 48 in Lithuania (2014). These data thus support the comparability of the number of positive cases in the different countries in recent years. The two phylogenetic analyses of the partial N gene sequences performed using either PhyML or Mr Bayes produced trees with identical topology. The phylogenetic analysis showed that 163 of the 165 studied Baltic sequences belonged to the lineage formed by the classical rabies virus within the cosmopolitan lineage, with a bootstrap value of 86% (Fig 8). No Arctic or Arctic-like variants were identified in the panel of viruses studied from the Baltic States. The majority of the Baltic rabies isolates grouped with the North-East European lineage (NEE), forming one strongly supported group (bootstrap value of 83). The NEE group consisted of 52 samples from the Baltic States and 21 published viral sequences (S1 Table) [38,49–51]. Both wild and domestic species fell in the NEE group. The NEE Group showed less than 1% nucleotide divergence and 3% amino acid divergence among all Baltic isolates. Nucleotide sequence analysis showed 100% of nucleotide identity between a red fox sample (no. 11584) isolated in 2006 in Estonia and three samples from Lithuania (a red fox isolated in 2007, a raccoon dog and a cattle both isolated in 2009). The same perfect identity was obtained between the Estonian isolate no. 11584 and two samples from Latvia (a raccoon dog and a dog both isolated in 2008). Five sequences from the Baltic States clustered with C group [48] formed with two published sequences, one from Russia and one from Ukraine (bootstrap support of 96%). Four species were included in this group: red fox (n = 2), raccoon dog (n = 1), cattle (n = 1) and dog (n = 1). Within C group, sequences showed more than 99. 9% of nucleotide similarity. 100% nucleotide identity was shown between a red fox sample (MT3-TA11-00267) isolated in Estonia in 2011 and two samples from Lithuania; a dog (no. 864) in 2012 and a raccoon dog (no. 4740) isolated in 2010. The PhyML tree also showed that a badger (Meles meles) (no. DR 784) and a marten (Martes martes) (no. 24771) isolated in Latvia in 2013 and in Lithuania in 2008 respectively, belonged to the group formed by the rabies virus vaccines (bootstrap of 100) (Fig 8). The vaccine-induced case isolated in Lithuania was found in the Alytus district in the south of the country, an area vaccinated since 2006 with Lysvulpen baits, whereas the vaccine-induced case isolated in Latvia was found in the Aloja district, in the north of the country, an area also vaccinated with Lysvulpen baits since 2011. Nucleotide analysis of the partial N gene sequenced of the two isolates showed 100% of nucleotide identity with the two referenced SAD-derived laboratory vaccine virus strains (EF206719 and EF206709) and there was 99. 4% nucleotide similarity with the SAD Bern vaccine strain (EF206708). The case found in Latvia was located in an area vaccinated with the Lysvulpen baits 25 km away from Estonia where the Rabigen baits were used. As the partial N gene sequence analysis did not discriminate among the two SAD-derived laboratory vaccine virus strains, the amplification of partial G gene sequence on the DR784 isolate was undertaken to identify the vaccine strain. The comparison between DR784 isolate and three vaccine strains, SAD B19 (EF206709), SAG2 (EF206719) and SAD Bern (EF206708), showed 100% nucleotide identity with SAD B19 and 99. 8% of similarity with the SAG2 and SAD Bern vaccines. The isolate DR784 was characterized by the presence of arginine in codon 333 (G gene). The sequence was clearly distinct from SAG2 (EF206719), characterized by two mutations in codon 333 yielding glutamine (Gln) at this position instead of arginine (Arg). Surveillance data indicated a drastic reduction (90%) in the number of detected cases following 1 to 4 years of ORV. These results corroborate those from other European countries where 90% reduction of rabies detected cases were observed within 10 years, and in many cases less than 5 years following first ORV [52–54]. Depending on the country, the time to complete elimination (i. e. remaining 10%) is more or less longer to achieve. While eradication requires an additional 10 or more campaigns until no more cases are detected in Freuling at al. [10] we found that 2 to 8 campaigns were necessary. Variation in the reduction of the number of cases detected after each ORV depends on multiple factors such as the geographical location of the infected country, the initial epidemiological situation, the tools and strategy used in the control programmes and indubitably the implementation of an appropriate surveillance scheme. As soon as ORV was implemented on the whole territories, the proportion of positive cases started to decrease in the three countries. As suggested previously in Brochier et al. [55], and Aubert [56] for fox rabies and Townsend et al. [57] for dog rabies, an inadequate vaccination area can compromise success and considerably extend the time to elimination. For Lithuania, the animals collected in vaccinated areas for the monitoring of ORV were also diagnosed for rabies. Because this sampling focuses on the animal population targeted by oral vaccines and not suspected for rabies (in contrary to classic rabies surveillance plans where only suspect animals are collected), Lithuanian surveillance data probably overestimates the number of negative samples compare to other countries. The comparison of the percentage of positive cases between countries became consequently hazardous. For this reason, combining data issued from surveillance sampling and monitoring sampling should, insofar as possible, be avoided [2,22]. Appropriate surveillance schemes focus on indicator animals collected at anytime, anywhere throughout the country and no sample size can be defined for proving the absence or the presence of rabies in wildlife regardless of the reservoir species. In contrast, the monitoring schemes are based on sampling foxes and raccoon dogs shot by hunters in vaccinated areas after ORV campaigns [30]. The oral vaccines used at the present time in Europe for raccoon dogs were developed to control rabies specifically in foxes. An experimental study evaluated the safety and efficacy of SAG 2 baits on caged raccoon dogs [58]. Either direct instillation or bait ingestion using a virus dose containing at least 10 times the field vaccine dose proved vaccine safety during the 60 days of observation of animals. More than 6 months after oral vaccination with the field dose, all animals were challenged with a street rabies virus. All vaccinated animals developed high rabies neutralizing antibody titers and survived a virulent challenge, demonstrating the effectiveness of the vaccine bait according to the European Pharmacopeia monograph. These results suggest that SAG2 vaccine baits are suitable for this species. Another study conducted on the SADP5/88 vaccine (derived from SAD Bern and no longer in use) in which two different doses of the vaccine were administrated showed satisfactory protection of challenged animals [59]. Paradoxically, to our knowledge, there are very few experimental studies using vaccines used in Baltic countries on raccoon dogs to assess their efficacy and safety prior to their release in the field. For the first time, bait uptake results suggest a significant difference in the frequency of uptake of red foxes and raccoon dogs, with a lower proportion of tetracycline-positive raccoon dogs compared with red foxes. This result can be attributed to the difference in behavior of the two species and particularly to the hibernation of raccoon dogs in the Baltics during the cold season (November–March) [60], which may influence the epidemiology of the disease and access to vaccines distributed during this period. The impact of hibernation was suggested in a model of rabies transmission in both raccoon dogs and red foxes [61]. As suggested by our results, strategies to control rabies in countries where this species is an important transmitter should better focus on the raccoon dog biology. As example, ORV could also target raccoon dogs after they emerge from hibernation. All countries implemented ORV according to the EU recommendations[19]. Bait uptake levels in Baltic countries rapidly reached 80% of the target population. Our study showed a constantly increasing bait uptake throughout the study period, suggesting cumulative exposure to distributed baits [19]. Data analysis in Estonia and Lithuania confirmed previous studies, showing a significantly lower bait uptake in juveniles than in adults [28,62,63]. As a matter of fact, difficulties in reaching juveniles during ORV campaigns were observed. This was observed especially in early spring [19] because cubs are in dens or cannot be vaccinated because too young to eat the bait. Latvia has used two types of vaccines, Lysvulpen and Fuchsoral vaccines. Analysis of factors that potentially affect bait uptake showed a significant influence of the type of bait used, with higher bait uptake when the vaccine Lysvuplen was used. The type of bait influence was independent from the year as further analysis, omitting the first years of vaccination with Fuchsoral baits, still considered the bait type as a significant variable explaining the TTC variations. Given that, according to the manufacturer’s specifications, both vaccines contain 150 mg of tetracycline in the bait matrix, the reason for this difference is unknown. More investigations on bait matrix composition and palatability are needed. Neutralizing antibodies are the most reliable parameter for assessing the efficacy of vaccination because they are closely correlated with protection against rabies infection [64]. The assessment of the rabies antibody level is theoretically the best means for evaluating vaccination coverage because individual variation in immune reactions is taken into account. ELISAs allow large-scale screening because they are rapid, easy to perform, do not require live rabies virus or cell culture, and can be performed in any laboratory. These tests have been demonstrated as particularly suitable for assessing the effectiveness of oral vaccination in field samples [31,65,66]. The evolution of herd immunity level did not show any specific pattern, showing an unsteady evolution in all three countries. The surprising absence of any immunological trend may reflect the lack of sensitivity or reliability of some commercial ELISA kits, as has already been demonstrated recently [34,67,68]. Although the overall average bait uptake in this study was 80%, seroconversion level was approximately 50%. The same large discrepancies observed between uptake and seroconverion were attributed the lack of sensitivity of a commercial kit on field samples compared to blood samples taken from experimentally infected foxes and raccoon dogs, probably due to the reduced quality of the sera (haemolysis, bacterial contamination due to field condition) [28,67,69]. Latvia was the only country that used two different kinds of ELISA kits (Bio-Rad and BioPro) to evaluate vaccination coverage in red foxes and raccoon dogs. Further analysis demonstrated that significantly different levels of seroconversion were found for the two different kits. BioPro ELISA results showed lower seroconversion level than those of the Bio-Rad ELISA kit. These discrepancies are inconsistent with previous studies in which the seroconverion were found lower using Bio-Rad compared to BioPro kits due to the lower sensitivity of the first test [33,70]. Our results may be explained by the fact that a different cut-off value from the 0. 5 conventional one’s was used for the Bio-Rad kit (0. 125 instead of the 0. 5 used in Wasniewski et al.). These results must be also nuanced by the fact that Latvia encountered specific events in the same period when using the Bio-Pro test in 2012, a year during which an epidemic sarcoptic mange occurred. Immunological reactivity due to sarcoptic mange could potentially have interfered with the rabies vaccination, leading to a lower response and a decrease in the level of rabies antibodies. A sharp decrease in the number of marked animals was observed in Estonia during the last four campaigns as soon as the ORV area was reduced to a buffer zone of 9 325 km2 (20 km along the Southern border and 30–50 km in eastern part of the country). This drop could be explained, inter alia, by an “edge effect” due to the small size of the vaccinated areas. The areas being small, the perimeter-to-surface ratio is higher and the probability of sampling an unvaccinated animal in bordering areas is higher than for a large ORV areas. Moreover, the proportion of raccoon dogs in the monitoring sample has increased every year, ultimately constituting more than ¾ of all animals tested. This example highlights the importance of considering the structure of the monitoring sample in the determinism of the overall and final bait uptake level. Thus, comparison of monitoring data between countries and their interpretation should be assessed by taking into account the species (raccoon dogs vs red foxes) and the age class (juvenile vs adult) of the sampled population. The molecular epidemiology of RABV in the Baltic countries showed the presence of three types of RABV variants in the Baltic States, the North-East European group (NEE) (158/165 isolates), the Russian group (C) (5/165 isolates) and two vaccine-induced rabies cases. These results confirm that the terrain for rabies hosts infected with Baltic variants is broad [71], ranging from Eastern to the Central Europe. More precisely, the NEE group has been reported in Eastern part of Russia and from Finland to Romania [49] including the Baltic States [28,72,73], Slovakia, Poland and Ukraine [38,48,74], while C group has been reported from the European part of Russia [48] to different parts of Ukraine [74]. Although the C group is the most widely reported RABV variant in Russia [75] including regions of western Siberia, Kazakhstan and Tuva, four other variants have been previously described in Russia [48]. This study is the first to report the presence of C variant in North-East Europe with three cases in Lithuania reported between 2009 and 2013, one case in Estonia in 2011 and one case in Latvia in 2012. The occurrence of the C variant in Baltic States could be the result of a westward spread of rabies-infected hosts from Russia or from Belarus to the Baltic States. Animal-to-animal transmission of rabies virus or human-mediated transports of latently infected animals could explain the movement of rabies infected hosts across borders. There are numerous studies illustrating rabies virus transmission by human-mediated animal movements [76], wildlife-mediated movement of rabies [50] or movements of infected animals a cross frozen seas [77]. In Russia, six wild canid species (red fox, raccoon dog, artic fox, steppe fox, jackal and wolf) are vector of the disease. In Eastern Europe and in north-eastern Europe, most wildlife cases are reported in red foxes and raccoon dogs. The NEE variant is particularly associated with raccoon dogs in north-western Russia and north-eastern Europe, while C group were previously associated with the red fox and the steppe fox in Russia [48]. In this study, no phylogenetic distinction was reported between the red fox and raccoon dog isolates, whatever the variant analysed (C and NEE groups) and whatever the phylogenetic method used. Perfect identity observed between one isolate (red fox) in Estonia in 2006 and five strains (two raccoon dogs, one fox, one cattle and a dog) isolated in Latvia and in Lithuania between 2007 and 2009 suggests that the variant circulating in fox and raccoon dog populations have the same origin. Dogs may have served as an early reservoir for interspecies rabies virus transmission generating viral lineages that then spread to other species [78]. Due to the risk of residual pathogenicity of oral rabies vaccines related to the viral strain’s attenuation level, all rabies virus samples isolated in areas where attenuated rabies virus vaccines are used should be typed in order to distinguish between vaccine and field virus strains [2,19,22]. For the first time, we demonstrated that two field Baltic isolates (a marten from Lithuania in 2008 and a badger from Latvia in 2013), clustered with the group forming the rabies vaccines, SAG2, SAD B19 and SAD Bern. Clearly, the two vaccine-induced rabies cases were closely related to SAD B19 strains, although both cases were found in an area vaccinated with SAD Bern Lysvulpen baits. Previous study results indicated that the SAD Bern Lysvulpen vaccine shows higher similarity to the strains belonging to the SAD B19 vaccine [79]. Such findings led to a change in the viral strain description for the national marketing authorization dossier of this vaccine, http: //www. uskvbl. cz/en/authorisation-a-approval/marketing-authorisation-of-vmps/list-of-vmps/authorised-by-national-and-mrdc-procedures. This is also the first reporting of a vaccine-associated virus detected in badgers and in martens. To date, few vaccine-induced rabies cases have been documented in target species. Muller et al. [80] reported six vaccine-induced rabies cases in foxes caused by SAD B19 and SADP5/88 in vaccinated areas in Germany and Austria, respectively. In Slovenia, a young fox was also shown closely related to SAD B19 in 2012 [81]. The most likely explanation for these vaccine associated cases isolated in non target species is either residual pathogenicity of the virus vaccine despite vaccine attenuation or reversion to virulence. RNA viruses are known to have high mutation rates due to the lack of proofreading by RNA polymerases and could have occasionally reversed to more virulent viruses. The second hypotheses would be a transmission from a red fox or raccoon dog initially infected by a vaccine strain. Potential transmission of vaccine strain has indeed been recently questioned when finding vaccine strain in salivary gland of a naturally infected fox [81].

This paper reviews ten years of rabies epidemiology in the three Baltic countries. Both surveillance efforts and oral rabies vaccination campaigns have resulted in the near eradication of the disease. Multivariate analysis assessed with generalized linear models (GLM) suggested lower oral vaccination effectiveness in raccoon dogs compared with red foxes, highlighting the importance of adapting oral vaccination strategy to each vector of the disease. Although eradication of the disease is almost achieved, the detection of some cases belonging with the Russian rabies lineage emphasizes a risk of rabies reintroduction in the Baltic States due to westward spread from bordering countries. This study show also the first vaccine-induced cases detected in non-target species (Martes martes and Meles meles).

Abstract Introduction Materials and Methods Results Discussion

medicine and health sciences pathology and laboratory medicine pathogens immunology tropical diseases microbiology vertebrates geographical locations estonia dogs animals mammals viruses vaccines preventive medicine rabies raccoons rna viruses neglected tropical diseases vaccination and immunization rabies virus public and occupational health infectious diseases foxes latvia zoonoses medical microbiology microbial pathogens people and places lyssavirus viral pathogens biology and life sciences viral diseases europe organisms

2016

Rabies in the Baltic States: Decoding a Process of Control and Elimination

Spatial modeling is increasingly utilized to elucidate relationships between demographic, environmental, and socioeconomic factors, and infectious disease prevalence data. However, there is a paucity of studies focusing on spatio-temporal modeling that take into account the uncertainty of diagnostic techniques. We obtained Schistosoma japonicum prevalence data, based on a standardized indirect hemagglutination assay (IHA), from annual reports from 114 schistosome-endemic villages in Dangtu County, southeastern part of the People' s Republic of China, for the period 1995 to 2004. Environmental data were extracted from satellite images. Socioeconomic data were available from village registries. We used Bayesian spatio-temporal models, accounting for the sensitivity and specificity of the IHA test via an equation derived from the law of total probability, to relate the observed with the ‘true’ prevalence. The risk of S. japonicum was positively associated with the mean land surface temperature, and negatively correlated with the mean normalized difference vegetation index and distance to the nearest water body. There was no significant association between S. japonicum and socioeconomic status of the villages surveyed. The spatial correlation structures of the observed S. japonicum seroprevalence and the estimated infection prevalence differed from one year to another. Variance estimates based on a model adjusted for the diagnostic error were larger than unadjusted models. The generated prediction map for 2005 showed that most of the former and current infections occur in close proximity to the Yangtze River. Bayesian spatial-temporal modeling incorporating diagnostic uncertainty is a suitable approach for risk mapping S. japonicum prevalence data. The Yangtze River and its tributaries govern schistosomiasis transmission in Dangtu County, but spatial correlation needs to be taken into consideration when making risk prediction at small scales. Schistosomiasis japonica is a zoonotic disease caused by the digenetic trematode Schistosoma japonicum. Historically, the disease was endemic in 12 provinces of the People' s Republic of China, with more than 10 million individuals infected [1]–[3]. Sustained control efforts implemented over the past 50 years have confined S. japonicum to seven provinces and brought down the number of infected people to less than 1 million [1]–[3]. The mean infection intensity has also decreased significantly [2]. However, surveillance and interventions are still warranted in 435 counties according to the 2005 annual report on the epidemiologic status of schistosomiasis in the People' s Republic of China [4]. Geographic information system (GIS) and remote sensing technologies are increasingly utilized for risk mapping and prediction of schistosomiasis [5], [6]. Over the past decade, several studies have explored the relationship between the occurrence of schistosomiasis, its intermediate host snails and environmental factors, particularly land surface temperature (LST) and normalized difference vegetation index (NDVI) [7]–[14]. Socioeconomic factors and water contact patterns were also studied [11], [15]–[18]. The flexibility in modeling and parameter estimation renders Bayesian spatial modeling particularly attractive for risk factor analysis and mapping [19]–[21]. Early statistical methods employed for data analysis followed independent rather than spatially-correlated approaches. More recently, spatial modeling using Bayesian Markov chain Monte Carlo (MCMC) simulation-based inference has been applied to estimate the relation between environmental predictors, socioeconomic factors, and schistosomiasis. This approach allows the prediction of the prevalence and intensity of infection at non-sampled locations, taking into account the spatial correlation present in the data [11], [12], [21]–[25]. However, none of the above-mentioned studies pertaining to the spatial or spatio-temporal distribution of disease risk has taken into account the uncertainty of the diagnostic technique. In the case of schistosomiasis japonica, both serological (e. g. , enzyme-linked immunosorbent assay (ELISA), indirect hemagglutination assay (IHA) [26]) and parasitological methods (e. g. , Kato-Katz technique [27], miracidium hatching test [28]) are used in epidemiologic surveys. None of these diagnostic approaches has 100% sensitivity, however [28]–[31]. Although an enhanced sampling effort (e. g. , multiple stool examinations) and simultaneous use of different diagnostics improve the sensitivity [32], [33] this strategy is not feasible in routine surveys due to logistic and financial constraints. In the early 1990s, the Chinese schistosomiasis control programme embarked on a two-pronged diagnostic approach. Local residents in S. japonicum-endemic areas are first screened with a serological test, followed by stool examination of seropositive individuals [29]. According to expert opinion, the sensitivity of ELISA ranges from 90% to 95%, and the specificity from 85% to 90%. In the case of the Kato-Katz technique, the estimated sensitivity and specificity are 20–70%, and 95–100%, respectively [32]. In this study, we employed a Bayesian approach to investigate the spatio-temporal patterns of S. japonicum infection, and to identify environmental and socioeconomic risk factors. In our models, we explicitly took into account the diagnostic uncertainty. The study was carried out in Dangtu, one of 14 S. japonicum-endemic counties in Anhui province, southeastern part of the People' s Republic of China. The first local case of schistosomiasis japonica was confirmed in 1953. Dangtu is situated on the lower reaches of the Yangtze River and stretches from 118°22′ to 118°53′E longitude and from 31°17′ to 31°42′N latitude. All three commonly recognized S. japonicum ecotypes are found in Dangtu, i. e. , (i) plains with waterway networks, (ii) marshlands and lakes, and (iii) hilly and mountainous regions. S. japonicum prevalence data were obtained from the annual county reports, covering the period from 1995 to 2004. Each year in September, field teams of the schistosomiasis control station in Dangtu sampled and surveyed the 114 schistosome-endemic villages as part of the national control program of schistosomiasis, which was approved by the institutional review board of the National Institute of Parasitic Diseases, Chinese Center for Disease Control and Prevention in Shanghai. The sampling frequency was in accordance with the prior classification of the respective village. Hence, villages with ongoing transmission were surveyed annually, villages where transmission was under control (prevalence <1%) were sampled every 2–3 years, and villages which had reached the criteria for transmission interruption (no human or animal cases within the past 5 years, no intermediate host snails observed in the previous year) were only surveyed if new snail habitats had been identified. During the 10-year surveillance period covered here, between 43 (in 1999 and 2002) and 68 (in 1998) villages were surveyed annually (median: 49). In sampled villages, all residents aged 5 to 65 years were invited to participate. One of the study requirements was that at least 80% of the eligible individuals should be tested. A two-pronged diagnostic approach was adopted; individuals were first screened by the IHA, followed by stool examination of seropositives. Parasitological diagnosis usually relied on the Kato-Katz technique [27]. Those found with S. japonicum eggs in their stool were treated with praziquantel. The median number of IHA tests performed per village was 778 (lower and upper quintiles: 302 and 1250). In this study, data from the Kato-Katz thick smear examinations were not used for further analysis, since some of the seropositives were not followed-up by the Kato-Katz technique due to recent treatments with praziquantel, and lack of compliance. The geographic coordinates of the village committee houses in the S. japonicum-endemic villages were collected using hand-held global positioning system (GPS) receivers (Garmin Corp. ; Olathe, KS, USA) and used as a proxi for the location of the village. Figure 1 shows the 114 S. japonicum-endemic villages in Dangtu county in relation to identified water bodies. Most endemic villages are located in the vicinity of water bodies or in the marshlands. Only four villages are situated in the northeastern hilly and mountainous region. A SPOT5 image with a spatial resolution of 2. 5 m and covering the whole study area, taken on February 9,2004, was purchased from China Remote Sensing Satellite Ground Station (Beijing, People' s Republic of China). This image was chosen because of its high quality (e. g. , cloud cover <10%). With regard to water bodies, no major changes occurred from 1995 to 2004. Water bodies were identified using an unsupervised classification function of ERDAS IMAGINE version 8. 6 (ERDAS LLC. ; Atlanta, GA, USA). The shortest straight-line distance between each village and the closest water body was calculated in ArcGIS version 8. 3 (ESRI; Redlands, CA, USA). For each year, one cloud-free Landsat-5 TM image with a spatial resolution of 30 m was purchased from China Remote Sensing Satellite Ground Station, covering the period from 1995 to 2004 (4 scenes were acquired in April, 3 in March, 2 in June, and 1 in August). LST and NDVI were extracted using the tools offered by ERDAS (http: //gi. leica-geosystems. com). For each scene, the mean LST and NDVI within a 2-km buffer zone around the centroids of the study villages were calculated in ArcGIS. Village-specific socioeconomic data were obtained from the annually-updated village registries. The available indices included annual average per-capita income and the proportion of households with tap water and improved sanitation. Dangtu county was partitioned into 0. 25×0. 25 km grid cells for the generation of a smooth prediction map for 2005. The minimum distance from each grid cell centroid to the nearest water body was calculated in ArcGIS. For each cell, the mean LST and the mean NDVI were extracted from the 2005 Landsat scene. LST and NDVI data were standardized by subtracting the arithmetic mean calculated from data within a 2-km buffer zone around the centroids of the study villages for each scene and then dividing the standard deviation using SAS version 8. 0 (SAS Institute Inc. ; Cary, NC). Villages were stratified into five wealth quintiles, based on the annual average per-capita income. The relationship between S. japonicum seroprevalence and village-specific environmental and socioeconomic surrogate measures was examined using scatter plots. A Bayesian approach was utilized to explore the spatio-temporal patterns of the S. japonicum seroprevalence data. The relationship between seroprevalence and environmental and socioeconomic covariates was also examined. We applied two different model specifications. The first set of models assumed no diagnostic error of the IHA. The second set of models explicitly took into account the diagnostic error, thus correcting for the estimated ‘true’ sensitivity and specificity of the IHA. For 2005, we created a smoothed predictive map of the S. japonicum prevalence. Let nit and zit be the number of examined and positive subjects by IHA, respectively, of village i (i = 1, …, N) in year t (t = 1, …, T). We assumed that zit follows a binomial distribution, that is zit ∼ Binomial (pit, nit), where pit is the seroprevalence following the standard formulation of the logistic regression model. We introduced covariate effects on the logit transformation of pit, that is, where α is the intercept, βk denotes a regression coefficient, and Xitk is the environmental or socioeconomic covariate. The standard assumption of this formulation is that the observations are independent. However, our data are spatially correlated because common environmental factors concurrently influence the infection risk in neighboring villages. Similarly, the data are temporally correlated because they have been obtained through repeated cross-sectional surveys. Ignoring these correlations, we would overestimate the significance of the covariates. To account for the spatio-temporal correlation, we introduced village-specific and year-specific random effects, ui and vt, respectively, as follows: . We defined a latent stationary and isotropic spatial process [34] on ui, by assuming that u = (u1, u2, …, uN) T has a multivariate normal distribution with variance-covariance matrix Σ, that is, u∼MVN (0, Σ). We defined Σ by an exponential correlation function, i. e. , Σlm = σ2exp (−ϕdlm), where dlm is the shortest straight-line distance between villages l and m, σ2 models the geographic variability, and ϕ is a smoothing parameter controlling the rate of decline of the spatial correlation with distance throughout the study period. For the exponential correlation function we have adopted the minimum distance at which correlation becomes less than 5%, which is defined by 3/ϕ and expressed in meters. Similar to previous spatio-temporal modeling of schistosomiasis [12], we defined a first-order autoregressive process (AR (1) ) on vt, assuming that temporal correlation ρ exists only with the preceding year [35]. An alternative spatio-temporal structure was modeled by assuming that spatial correlations evolve over time (space-time interaction) that is, where ut = (u1t, u2t, …, uNt) T is the spatio-temporal random effect such that with the parameter ϕt controlling the rate of decline of spatial correlation with distance in year t. We assessed the significance of covariates by including only environmental, or only socioeconomic, or both types of covariates. The model detailed before was based on the assumption that the IHA reliably diagnoses a S. japonicum infection, i. e. , its sensitivity and specificity are 100%. However, since IHA and other diagnostic tests have shortcomings [29], we made an attempt to incorporate the diagnostic error of IHA into our modeling framework. Expert opinions on the diagnostic performance of IHA were gathered by means of a questionnaire survey, as described elsewhere [32]. The experts' consensus was that the sensitivity and specificity of IHA is 80–95% and 70–80%, respectively. These values were fed into the model as prior information. Let πit be the underlying true prevalence of S. japonicum infection for village i in year t, and pit the observed prevalence of infection. Following the model specifications of Booth and colleagues [33] and Wang et al. [32], we assumed that the number of seropositives zit has a binomial distribution that is zit ∼ Binomial (pit, nit), and related the observed and true prevalence via the equation pit = πitsjt + (1−πit) (1−cjt). This equation is derived from the law of total probability, where sjt and cjt are the sensitivity and specificity of IHA for village j (j = 1, …, J) in year t, respectively, where j is a group of adjacent villages. The models described previously were fitted, but with underlying prevalence πit instead of the seroprevalence pit. The same database was used throughout the study. We randomly selected 93 out of the 114 S. japonicum-endemic villages (82%), and used the surveys conducted between 1995 and 2004 for fitting the models, employing 408 out of the available 508 surveys. The remaining 100 surveys carried out in the other 21 villages over the same period served for model validation. In a first step, we compared the goodness-of-fit of the models by using the deviance information criterion (DIC) [36]. The model with the smallest DIC value was considered the best-fitting one. Next, we evaluated the predictive abilities of different models by calculating five different Bayesian credible intervals (BCIs) with probability coverage equal to 5%, 25%, 50%, 75%, and 95% of the posterior predictive distribution at the test locations, as suggested elsewhere [19]. Models with a high percentage of records falling into the narrowest BCIs were considered to have good predictive abilities. Following a Bayesian model formulation, we adopted vague normal prior distributions for each regression coefficient βk and intercept α, vague inverse gamma priors for variances, and a uniform prior ranging from −1 to 1 for temporal correlation ρ. Informative beta prior distributions derived from expert opinion that is, beta (67. 18,9. 60) and beta (224. 25,74. 75), were used for sensitivity sjt and specificity cjt, respectively. We assumed that the prior for the spatial correlation ranged from 0. 01 to 0. 99 at the minimal distance between villages (0. 6 km) and from 0 to 0. 2 at maximal distance (49 km), thus uniform priors ranging from 0. 017 to 7. 675 were used for the spatial decay parameters ϕ and ϕt. Two-chain MCMC was used for parameter estimation. Model convergence was assessed by visually inspecting the time series plot for each parameter, and Gelman-Rubin statistics [37]. The inference of the parameters was based on 15,000 iterations of both chains after the burn-in phase. Model fit was carried out in WinBUGS 1. 4. 1 (Imperial College and MRC, London, UK). Figure 2 shows the observed seroprevalence in the study villages, according to survey year. Commonly, high seroprevalences were observed in villages located in close proximity to large rivers. In 27% of the village surveys the seroprevalence was zero, whereas a mean seroprevalence ≥10% was found in 41% of the surveys. Table 1 summarizes the goodness-of-fit and the predictive ability of the models which did not take into account the diagnostic error of IHA. The smaller DIC values of the spatio-temporal models indicate that they fitted the data better than the non-spatial ones. The predictive ability of the models could be improved significantly by considering spatio-temporal random effects. Moreover, the percentage of testing records falling into smaller BCIs of the posterior predictive distribution was considerably higher in the spatio-temporal models than in the non-spatial ones. Models considering the temporal evolution of spatial correlation also appeared to better fit the data than those assuming independent spatial and temporal processes. Considering also socioeconomic information did not further improve the model. Hence, the model with environmental covariates and variable spatial correlation was considered the best-fitting one. As shown in Table 2, incorporating the sensitivity and specificity of IHA as model parameters, resulted in smaller DIC values in the annual differences in the spatial correlation. When models also considered socioeconomic information there was no further improvement. Actually, the percentages of testing records falling into smaller BCIs were larger in a similar model that only considered environmental covariates. Thus, the model without explicit consideration of socioeconomic data was considered the best-fitting one. However, its predictive ability was inferior to that of the best-performing model which did not take into account the diagnostic error of IHA (4% versus 31% of the test records falling into the 5% BCI). Table 3 summarizes the results of the best-fitting spatio-temporal models regarding the observed S. japonicum seroprevalence and the ‘true’ infection prevalence. The prevalence increased with the mean LST (regression coefficients: 0. 201 and 0. 669 for seroprevalence and ‘true’ infection prevalence, respectively), and was negatively correlated with the mean NDVI (regression coefficient: −0. 327 and −1. 044, respectively). The seroprevalence was also inversely related to the distance to the closest water body (regression coefficient: −0. 277 and −1. 069, respectively). The estimated variances using the model with adjusting for the diagnostic error were larger, as suggested by larger 95% BCIs. The relationship between the serostatus and socioeconomic covariates was not further explored since the selected variables neither improved the goodness-of-fit nor the prediction ability of the models. The best-fitting spatio-temporal models indicated that the spatial correlation structures of the observed seroprevalence and the ‘true’ prevalence differed from one year to another, albeit not significantly (Table 3). Generally, the spatial correlation of the seroprevalence declined at a slower pace than that of the ‘true’ prevalence (smaller values of the parameter ϕ indicate a slower decay of the correlation with distance). For the measured seroprevalence, the shortest distance at which the spatial correlation was below 5% was determined in 1995 (5. 9 km; 95% confidence interval (CI): 0. 5–17. 8 km). The maximum value of 55. 6 km (95% CI: 21. 0–144. 4 km) was modeled for 2003. For the underlying ‘true’ prevalence, the respective distances were 0. 7 km (95% CI: 0. 4–3. 0 km in 2001) and 3. 7 km (95% CI: 0. 4–20. 7 km in 1999; Figure 3). The model for the measured seroprevalence further indicated a fast decline of the spatial correlation with distance in 1995,1998, and 2001, and a slower decay over the respective ensuing two years. The S. japonicum prevalence in Dangtu county was predicted for 2005, based on the spatial correlation structures observed in the preceding year. The predicted seroprevalence in the county ranged from 0. 05% to 22. 9% (posterior median). Most of the predicted high-seroprevalence areas are located in close proximity to water bodies, especially the Yangtze River, and in the southeast of the county (data not shown). The predicted ‘true’ S. japonicum prevalence ranged from nil to 3. 7% (posterior median). The locations for which a relatively high ‘true’ prevalence was predicted are again located in the vicinity of water bodies (Figures 4a and 4c). The distribution of the prediction error is depicted in Figures 4b and 4d. In this study, we estimated the ‘true’ S. japonicum prevalence in a schistosome-endemic county of the People' s Republic of China by explicitly taking into consideration the diagnostic error of a widely used serological test, i. e. IHA. Additionally, we explored the spatial distribution over time, and produced a predictive risk map for the year 2005. Since antibody-based immunological tests, such as IHA and ELISA, cannot distinguish between an active and a recently cleared infection, these techniques result in low specificity in areas where chemotherapy is provided on a regular basis [31]. Thus, the analysis of uncorrected seroprevalence data would only be suggestive of the overall infection pressure [38]. In order to better understand the epidemiologic characteristics of schistosomiasis japonica, we accounted for the lack of sensitivity and specificity of the standard serological test employed in our study setting by using a Bayesian approach, and compared the outcome with that of the uncorrected model that assumed 100% sensitivity and specificity. In recent years, significant progress has been made with Bayesian spatio-temporal models. Thus our understanding of the epidemiology of infectious diseases in general [39], [40], and schistosomiasis in particular [22], has been improved. We used two types of spatio-temporal models; one assumed independent spatial and temporal random effects, and the second assumed that spatial correlations evolved over time (space-time interaction). Similar approaches have been successfully employed before [12], [41], [42]. We considered a stationary spatial process, although recent investigations suggest that non-stationarity is a more reasonable approach [19], [21]. The reasons were as follows. First, Dangtu county is small, spanning 50 km at most. Second, the local environment in this setting is rather uniform, and the study area mainly consists of plain regions with waterways, marshlands and lakes. In future analyses, it would be interesting to investigate anisotropic processes. Remotely-sensed environmental data are increasingly utilized in schistosomiasis research [5], [11], [43], [44]. Temperature and vegetation coverage are among the most frequently investigated environmental features, as they can be readily derived from satellite images. Their utility for an enhanced understanding of the local epidemiology of schistosomiasis has been demonstrated extensively [8], [11], [44]. In this study, LST and NDVI were extracted from Landsat-5 TM images, and averaged values for each village for individual survey years were calculated for 2-km buffer zones around the centroid of each village. The 2-km buffer zone approximately corresponds to an average village in Dangtu, and most daily activities take place within such a range. Prevailing weather conditions did not allow us to obtain all remotely-sensed data in the same month, i. e. , April, the first month of the local transmission season [12]. To remedy this issue, we standardized the indices. Three important findings emerged from our study. First, LST was positively associated with S. japonicum prevalence, whereas the NDVI and distance to water bodies were negatively associated. These observations are consistent with previous findings [12], [23]. However, the non-spatial models revealed that the prediction ability of these covariates was poor whether or not the diagnostic error of IHA was taken into account. It is thus conceivable that the environmental factors explained the local S. japonicum prevalence to a small degree only. The effects of socioeconomic factors such as the annual average per-capita income, the proportion of households with piped water supply, and the proportion of households with access to improved sanitation were even smaller, contrasting results for S. mansoni in Côte d' Ivoire [11], [45]. Possible explanations for this finding are that socioeconomic factors could be disconnected from the epidemiology of schistosomiasis at small spatial scales, and improved water supply and sanitation do not necessarily change the water contact pattern of villagers [15]. A model incorporating socioeconomic variables measured at the individual level rather than at the village level as done here, might result in a better fit. Second, the spatial correlation of the seroprevalence and the estimated ‘true’ prevalence of S. japonicum occurred over greater distances for the former than the later. Our study is the first to compare the range of spatial correlation of the seroprevalence with that of the underlying prevalence. Additional investigations in different settings are warranted to verify this finding and explore possible reasons. Spatial correlation has also been documented for S. haematobium and S. mansoni in different African settings [11], [23]. The importance of the spatial correlation was underscored by the finding that the predictive ability of the model was greatly improved when spatio-temporal random effects were incorporated. The inclusion of the uncertainty about IHA sensitivity and specificity lowered the predictive ability, and increased the prediction errors since additional sources of errors were considered and the spatial correlation occurred over shorter distances. Whilst the spatial correlation varied from one year to another, no strong temporal trend was observed in our study. One possible reason is that the duration of our inquiry (i. e. , 10 years) is not long enough for capturing prevailing temporal patterns. Third, smoothed risk maps for 2005 were created based on the spatial correlation found in the preceding year. Since no significant temporal trend was detected from 1995 to 2004, it was decided to use the most recent data only. It is evident that most human infections were predicted to occur in close proximity to the Yangtze River and its tributaries. It has already been noted before that Oncomelania hupensis in the waterways connected to the Yangtze River are difficult to eliminate, and that snails can readily re-colonize cleared areas [2]. The prediction maps highlighted the areas (villages) at high risk of S. japonicum infection, and emphasized the important role of the Yangtze River in the transmission of schistosomiasis in Dangtu county. Implications for the local schistosomiasis control program are that control measures should be targeted to those villages at highest risk. One limitation of our study is that about 20% of the eligible population (aged 5–65 years) in the sampled villages was not surveyed. It is hard to predict whether non-compliance biased our risk profiles. Another limitation is that non-surveyed villages were excluded from the analysis in the corresponding year (s) and their effects on the estimates were not taken into consideration in the models, since there might be too many parameters to be estimated. In conclusion, we have presented an in-depth study on the spatio-temporal pattern of S. japonicum within a single county. Importantly, we explicitly took into account the diagnostic error of the serological screening test, and employed a Bayesian modeling approach, through which the underlying ‘true’ prevalence of S. japonicum infection could be estimated and predicted. There is considerable spatial correlation and annual variability of S. japonicum infection. Hence, for small-scale prediction, accounting for the spatial correlation seems more important than considering the risk factors included in our study. Finally, the Yangtze River and its tributaries play an essential role in the local epidemiology of schistosomiasis japonica.

Schistosomiasis is a serious public health problem in the People' s Republic of China and elsewhere, and mapping of risk areas is important for guiding control interventions. Here, a 10-year surveillance database from Dangtu County in the southeastern part of the People' s Republic of China was utilized for modeling the spatial and temporal distribution of infections in relation to environmental features and socioeconomic factors. Disease surveillance was done on the basis of a serological test, and we explicitly considered the imperfect sensitivity and specificity of the test when modeling the ‘true’ infection prevalence of Schistosoma japonicum. We then produced a risk map for S. japonicum transmission, which can assist decision making for local control interventions. Our work emphasizes the importance of accounting for the uncertainty in the diagnosis of schistosomiasis, and the potential of predicting the spatial and temporal distribution of the disease when using a Bayesian modeling framework. Our study can therefore serve as a template for future risk profiling of neglected tropical diseases studies, particularly when exploring spatial and temporal disease patterns in relation to environmental and socioeconomic factors, and how to account for the influence of diagnostic uncertainty.

Abstract Introduction Materials and Methods Results Discussion

infectious diseases/neglected tropical diseases public health and epidemiology/epidemiology infectious diseases/helminth infections public health and epidemiology/social and behavioral determinants of health mathematics/mathematical computing infectious diseases/tropical and travel-associated diseases public health and epidemiology/environmental health infectious diseases/epidemiology and control of infectious diseases

2008

Bayesian Spatio-Temporal Modeling of Schistosoma japonicum Prevalence Data in the Absence of a Diagnostic ‘Gold’ Standard

Staphylococcus aureus is a human commensal that can also cause systemic infections. This transition requires evasion of the immune response and the ability to exploit different niches within the host. However, the disease mechanisms and the dominant immune mediators against infection are poorly understood. Previously it has been shown that the infecting S. aureus population goes through a population bottleneck, from which very few bacteria escape to establish the abscesses that are characteristic of many infections. Here we examine the host factors underlying the population bottleneck and subsequent clonal expansion in S. aureus infection models, to identify underpinning principles of infection. The bottleneck is a common feature between models and is independent of S. aureus strain. Interestingly, the high doses of S. aureus required for the widely used “survival” model results in a reduced population bottleneck, suggesting that host defences have been simply overloaded. This brings into question the applicability of the survival model. Depletion of immune mediators revealed key breakpoints and the dynamics of systemic infection. Loss of macrophages, including the liver Kupffer cells, led to increased sensitivity to infection as expected but also loss of the population bottleneck and the spread to other organs still occurred. Conversely, neutrophil depletion led to greater susceptibility to disease but with a concomitant maintenance of the bottleneck and lack of systemic spread. We also used a novel microscopy approach to examine abscess architecture and distribution within organs. From these observations we developed a conceptual model for S. aureus disease from initial infection to mature abscess. This work highlights the need to understand the complexities of the infectious process to be able to assign functions for host and bacterial components, and why S. aureus disease requires a seemingly high infectious dose and how interventions such as a vaccine may be more rationally developed. Staphylococcus aureus is a leading opportunistic human pathogen renowned for its ability to evade the immune system and cause a variety of different infections [1]. S. aureus infections can vary from superficial skin lesions, through deep seated abscesses to life threatening sepsis [1]. The diversity of disease modalities has made an understanding of the underlying principles of infection challenging. S. aureus primarily occurs as a human commensal, mostly in the nares from whence it is able to seed infection. Many S. aureus infections are iatrogenic, and these are commonly associated with the colonisation of indwelling medical devices [2]. Typically during an infection, after invasion, an immune reaction is initiated by macrophages and these release cytokines to summon neutrophils [3]. Fibrosis also occurs, as well as the death of many of the invading immune cells leading to the pus filled abscesses associated with S. aureus infections. S. aureus can also regularly escape local infection sites and disseminate further. If it enters the bloodstream this can lead to sepsis as well as invasion of other organs whereby further local infections can occur. Thus S. aureus infection is a highly dynamic process with broad dissemination and repeated metastases. Phagocytosis by professional phagocytes such as macrophages and neutrophils is the primary mode by which S. aureus is controlled by the immune system. However, S. aureus has multiple mechanisms to subvert the immune response [4,5], including the production of a variety of specific toxins, the ability to survive intracellularly within immune cells and the capacity to evade capture by phagocytes. As S. aureus is effectively ubiquitous in the human environment, our immune systems are exposed to this organism as evidenced by the significant bloodstream titre of antibodies against a variety of S. aureus antigens [4,6–8], however these do not necessarily afford protection against infection. It is therefore the subject of much debate as to what are the key immune factors that control S. aureus due to the variety of possible infections as well as the many immune components involved [9,10]. This is of great translational importance as the correlates of protection against infection will determine crucial developments such as a vaccine to prevent disease [9,10]. In order to develop the knowledge base to underpin such an intervention it is necessary to first understand the correlates of disease, which in itself is dependent on mapping disease progression from initial infection to resolution in favour of the host or the pathogen. To understand the factors that are important for disease progression it is first necessary to determine the population, spatial and temporal dynamics of the infectious agent within the host. This will highlight the immune breakpoints through which the pathogen must pass to achieve a productive infection. A key way to demonstrate this has been through the creation of multiple tagged variant strains of the pathogen of equivalent fitness and passaging them, at specific ratios, through the relevant host model of disease [11–15]. Subsequent analysis of overall pathogen numbers and the ratios of the tagged strains allows determination of the within host population dynamics. Coupling this with temporal and spatial (specific organ) analysis provides a map of the spread and proliferation of the infectious agent, and highlights the potential for population bottlenecks from which clonal expansion may occur. In model animal systems this can also be combined with the manipulation of the host and/or pathogen to determine the role of multiple factors in disease. This type of study has been very revealing for a range of pathogens such as Streptococcus pneumoniae, Yersinia pseudotuberculosis and Salmonella enterica [12,14,16]. We have used this approach with S. aureus to demonstrate a population bottleneck during systemic infection in murine and zebrafish embryo models [13,17]. An initial discovery in the zebrafish model was used to inform the design of murine experiments. Using an intravenous administration of a mix of 3 antibiotic resistance marker tagged S. aureus Newman HG (NewHG) clones it was found that the bacteria are initially largely sequestered in the liver and spleen. Bacterial numbers subsequently slowly decreased in the liver with concomitant increasing numbers in the kidneys. In the liver and spleen the initial ratio of strains is apparent but surprisingly there is a significant alteration in strain ratio in the kidneys consistent with individual bacteria founding the characteristic abscesses [13]. This clonal expansion increases in all organs during infection but is most pronounced in the kidneys. These results show that a population bottleneck can occur in the mouse alluding to a metastatic spread of bacteria from a primary site of infection to the kidneys. In zebrafish, manipulation of professional phagocytes demonstrated the population bottleneck to be likely neutrophil mediated (with the involvement of macrophages) [17,18]. Since it is mediated by immune cells we would also define this population bottleneck as an immune bottleneck. This provides an initial framework onto which to map within host population dynamics and the influence of host immune defences. The basis for key developments in our understanding of disease relies on the use of animal models, the aim of which is to reflect facets of human disease. There are a multiplicity of S. aureus infection models available, ranging from insects (e. g. Drosophila & Galleria), to zebrafish and mammals (e. g. mice and rabbits) [9,17,19,20]. Mammalian models are considered the most apposite for S. aureus infections as they share several important characteristics with humans, for instance body temperature, innate and adaptive immune systems and disease pathology. However mammalian models typically require a high initial inoculum to ensure disease initiation and many S. aureus virulence determinants are human specific [9,21]. Consequently, it is difficult to equate animal model data with the human situation, particularly in the translational context where for instance, despite promising animal model data, vaccine trials have failed in the clinic [9,10]. Given these caveats, animal models are still important conduits for discovery of basic disease parameters. However it is important to evaluate the relative merits of the available models before relying on them to make useful conclusions. In this study we have used 3 models of systemic disease: firstly the zebrafish model to provide initial data to inform murine studies and then the mouse sepsis and survival models (15,18–20). The sepsis model is well established and gives a clear progression of bacterial population dynamics [22]. The survival model has been standard for many studies and is fundamentally similar to the sepsis model in that bacteria are commonly injected in to the bloodstream. However the dose is relatively much higher to cause the subjects to succumb to infection rather than survive and resolve as in the sepsis model [23]. The survival, or lethal challenge model, has been in general use and is of regulatory importance [21,23,24]. Here, we have evaluated infection dynamics in the above systemic models of infection with a suite of otherwise isogenic, antibiotic resistant variants of a range of S. aureus strains. This has revealed population bottlenecks within organ pathogen spatial distribution and established the role of specific immune effector cells in the dynamics of infection. This has led to the formulation of a new conceptual model for disease progression. To evaluate potential strain-specific and generic pathogenesis parameters the population dynamics of a range of different S. aureus strains (Newman, NewHG, SH1000 and USA300) in the zebrafish embryo systemic model was first determined. Three matched antibiotic resistant S. aureus strains in the USA300 (JE2) strain background: erythromycin/lincomycin-resistant, EryR (EPPS1); kanamycin-resistant, KanR (EPPS2); and tetracycline-resistant, TetR (EPPS3), and the Newman strain background EryR (EPPS4), KanR (EPPS5), TetR (EPPS6) were produced by transduction of the relevant markers from the previously constructed NewHG strains [13]. The established zebrafish embryo model of systemic disease was then used [17], whereby a 1: 1: 1 mixture of the variants of each strain was injected into embryos 30 hours post fertilisation (c. 1500 CFU (colony forming units) in total), followed by monitoring of host survival. Bacteria were harvested from embryos and their total numbers and proportions of each S. aureus strain variant determined. Each marked variant was equally likely to predominate (group variation was not significantly different, Bartlett’s test for equal variance USA300: p = 0. 1252, Newman: p = 0. 2364), demonstrating that there were no relative fitness costs associated with the individual antibiotic resistance markers which would undermine the analysis [11] (S1 Fig). The proportion of the different strains was used to generate the species evenness index for the population [25], which defines how evenly matched different populations of organisms are within an environment. A population evenness of 1 for a given population (whether it was a host or organ) means the strain variants are evenly distributed (1: 1: 1 ratio) whereas 0 means the entire population consists of one strain variant. We chose the population evenness metric as it is a commonly accepted in ecological studies [26,27]. It is suitable as it is based on Shannon’s diversity index which is equally sensitive to very rare and very common species in a sample. Our samples inherently had these properties as the populations varied from evenly mixed to completely dominated by 1 variant. We found that the variants which randomly came to predominate readily occurred in both the Newman and USA300 infected hosts, i. e. the population evenness of the injected population started at near 1 and became 0 over the course of the experiment (Fig 1). Additionally, there was a statistically significant correlation between time of death and decreasing population evenness for both strain backgrounds (linear regression, USA300: P<0. 0001, F = 57. 39, R2 = 0. 3766, Newman: p = 0. 0006, F = 13. 69, R2 = 0. 2504, Fig 1). The decrease in population evenness showed that the population had an increased chance of being clonal over time. This meant that the bacterial population in those zebrafish had passed through a likely population bottleneck. This corresponds to the previous observation of clonality in SH1000 and NewHG in the zebrafish model. The marked strains were then used for determination of the population dynamics in the established murine sepsis model of infection [22]. A total infectious dose of approximately 1x107 CFU was injected intravenously via the tail for each combination of marked variants in the 4 strain backgrounds (Newman, USA300, SH1000 and NewHG constructed here and previously [13]). Mice were culled at 2,18,48 and 72 hours post infection. The heart, lungs, spleen, left and right kidneys and liver were extracted from each subject, homogenised and the CFUs (colony forming units) of the different marked variants determined. As for the zebrafish model the species evenness index was determined. Overall the occurrence of the TetR, EryR and KanR populations was not significantly different in the mice regardless of strain (S1 Fig, Bartlett’s test for equal variance, SH1000: p = 0. 5598, NewHG: p = 0. 8478, Newman: p = 0. 9631 and USA300: p = 0. 9330), demonstrating that in all strain backgrounds the antibiotic resistance markers did not impart a fitness cost in the mouse model. The occurrence and distribution of S. aureus for NewHG, Newman, SH1000 and USA300 are shown in Figs 2 and S2. As the liver is likely the source of within host dissemination, how is this mediated? Previously we have postulated that phagocytes form the focus for the population bottleneck and it has been demonstrated that macrophages and neutrophils are required for the control of S. aureus infections in the mouse and zebrafish models [17,28,29]. In particular, it has been shown that after iv injection, S. aureus is initially phagocytosed by Kupffer cells (liver macrophages) [30]. Neutrophils have also been demonstrated to act a potential “Trojan horses” carrying viable bacteria within themselves [5]. Macrophages were depleted using clodronate vesicles [31] and, as expected, mice are consequently more susceptible to S. aureus infection and so the dose was reduced to 1x105 CFU of 1: 1: 1 labelled NewHG. Treated mice had a significantly higher bacterial burden in the liver compared to the control vesicle treated subjects (ANOVA, p<0. 0001 Multiple comparisons show that the clodronate treated is significantly different from the blank controls, Fig 4). An additional replicate produced similar results (S5 Fig). Occasionally other organs, including the kidneys, were colonised. Interestingly, population evenness was significantly increased compared to the control (Kruskal-Wallis, p = 0. 0002, Multiple comparisons show that the clodronate treated group is significantly different from the blank controls). This increased population evenness is due to the formation of multiple small abscesses visible on the liver rather than a single clone emerging. Similarly, to macrophage depletion, neutropenia (generated using anti-Ly6G antibodies) resulted in greatly increased susceptibility to S. aureus infection with an inoculum of 5x105 CFU being used (Fig 4). Neutropenic mice demonstrated detectable bacteria almost exclusively in the liver and in contrast to macrophage depletion these populations showed no difference in population evenness compared to the control, indicating clonal expansion (Kruskal-Wallis, p = 0. 0002, Multiple comparisons show that the anti-Ly6G treated group is significantly different from the clodronate groups but not the blank controls). The infected livers appeared pale with no observable surface abscesses. We initially tried a lower dose of 1x105 CFU, although there was neutrophil depletion, the resulting CFUs were no greater than the controls (again only the livers contained S. aureus by day 3 post infection) so we increased the dose used to the 5x105 CFU results presented (additional results are shown in S5 Fig). We also depleted the neutrophils using cyclophosphamide and an inoculum of 1x105 CFU. This again showed pale livers, increased CFU in the livers and mostly clonal in contrast to the macrophage depletion study (S5 Fig) and supporting our anti-Ly6G results (Liver CFU ANOVA, p<0. 0001, Multiple comparisons show that the cyclophosphamide treated group is significantly different from both the blank controls and clodronate). However, the depletion of neutrophils was not as complete (cyclophosphamide is generally toxic and although it depletes neutrophils in particular it also affects lymphocytes, monocytes and other fast growing cells) so the anti-Ly6G mediated depletion was preferred for demonstrating the effect of neutrophils. Whilst both macrophages and neutrophils are required for host defence compared to the control and their loss results in greater bacterial loads (ANOVA, P<0. 0001 Multiple comparisons show that the clodronate and anti-Ly6G treated groups are significantly different from the blank controls) the macrophages, and not neutrophils, appear to be the basis for clonal expansion as their depletion results in increased population evenness. Macrophage depletion also results in greater bacterial loads than neutrophil loads when the same amount of S. aureus is administered, further indicating that macrophages are the likely bottleneck. Conversely, neutrophils appear to be involved in the seeding of organs from an already derived clonal population within the liver. Kidney abscesses in the sepsis model are largely clonal in that they are derived from an individual founding cell [13,18]. However, outwardly the kidneys can show a multi-lobed abscess structure (Fig 5A). Two methods of within-organ analysis were used, both using a mixed population of differentially labelled bacteria to allow clonality to be determined post hoc. Firstly mice were injected with a 1: 1 ratio of 1x107 CFU NewHG GFP KanR/NewHG mCherry EryR, and at 5 days post infection (the time by which kidney abscesses had developed), the mice were sacrificed. Infected kidneys were serially and sequentially sectioned for CFU determination, histology and microscopy analysis of labelled bacterial populations (Fig 5B). This allowed a reconstruction of the 3D abscess architecture. Sample organs were sectioned every 300μm and 3x8μm sections of tissue were stained with either DAPI or Hematoxylin and Eosin (Fig 5F and 5G). The 300μm of tissue in between these sections was homogenised and plated for CFU determination (Fig 5D). Secondly, an optical clearing technique was employed to reveal the in situ distribution of fluorescently labelled bacteria within the organ (Fig 5H and S1 Video). Here lightsheet microscopy allowed a 3D rendering of the bacterial distribution to be determined [32,33]. Overall the histological sectioning revealed that there was an uneven distribution of bacteria within organs, with distinct foci that correlated with both CFU and visualisation of fluorescent bacteria (Fig 5). The foci of infection consist of solid aggregates of S. aureus or scattered individuals that co-localise with areas of neutrophil infiltration (S7 Fig). Where mature abscesses were present they almost exclusively consisted of one fluorescent variant (Fig 5C and 5D). However, this technique resulted in the destruction of the tissue and we used new techniques to confirm the distribution of S. aureus in the kidneys. We used lightsheet microscopy combined with the optical clearing of the organs, which maintains the fluorescence of the bacteria whilst making an abscess observable in 3D. Organ clearing and lightsheet microscopy has been used in other systems to great effect due to preserving structures in situ and we anticipate this technique will be of great use in S. aureus research. The lightsheet microscopy additionally showed (beyond the details revealed by histology) that within abscesses the solid aggregates in the kidneys were following the structure of kidney tubules in the cortex whilst the neutrophil infiltration sites were outside the tubules (Fig 5H and S1 Video). This would not have been demonstrated by histology alone. Based on these observations it seems likely the S. aureus is trapped in the tubules, grows into solid aggregates and then some can subsequently escape into the regions between the tubules where neutrophils engage the bacteria. Even in infected organs that contain abscesses that overall are formed from more than one marked strain, there is a clear differential distribution of clones suggesting that pervasive abscesses which have multiple foci are seeded from a single bacterium which has divided and spread to form extended foci of infection in the surrounding tissue. The mouse survival model has been used widely to test the efficacy of novel treatments, prophylaxis and the role of bacterial and host factors in disease [21,23,34–36]. It is technically similar to the mouse sepsis model but primarily differs in that an increased dose is given leading to mortality as a primary output. Strains NewHG, Newman, SH1000 and USA300 were compared using the survival model. For all strains a high inoculum was used (around 1x108 CFU except USA300 which is similarly lethal at a lower dose of 3x107 CFU), made up of equal proportions of the 3 antibiotic resistant marked variants in each case (Figs 6 and S3). Infection became systemic in every subject, with bacteria across multiple organs, and typically resulted in mouse cull within 2–3 days. Clonality (low population evenness) was much rarer at these high doses in all organs, in all subjects, and across all 4 bacterial strains (Fig 6). It was also notable that the hearts, lungs and spleens now had high numbers of S. aureus. To determine whether the sepsis and survival model were ends of a spectrum or independent of each other, intermediate doses of the strains were administered to delay, and reduce, the number of subjects reaching the morbidity endpoint (for all strains around 3x107 CFU except USA300 which is similarly lethal at a lower dose of 1x107 CFU). This resulted in a lower rate of mortality (S4 Fig), with those subjects culled up to and including day 5 having more systemic infections, whereas after this time only the liver and kidneys had bacterial loads as the infection resolves. Lower dose survival treatments had much greater levels of clonality than the equivalent higher dose survival treatment. Both kidneys are more frequently concurrently infected as well. Overall there is a significant correlation between increasing initial dose of S. aureus and the decreasing occurrence of clonality (linear regression, p = 0. 0386, F = 6. 959, R2 = 0. 5370, Fig 7). There is a significant correlation between increasing clonality and increased survival by day 4 (linear regression, p=0. 0049, F=18. 82, R2=0. 7583, Fig 7). There is also a significant correlation between increasing initial dose of S. aureus and decreasing survival (linear regression, p = 0. 0196, F = 9. 982, R2 = 0. 6246). The order of increasing virulence in the survival model was SH1000, Newman, NewHG and USA300 consistent with previous reports [37–39]. As an opportunistic pathogen, S. aureus is able to cause a wide range of human diseases from the superficial to potentially life threatening. There are multiple animal models of S. aureus infection that all aim to recapitulate those events that shape the interaction between the pathogen and the human host. However, increasingly there is evidence that many of the virulence determinants are human specific and thus are not relevant to the widely used animal models. Also the route and mode of human infection is difficult to replicate in models. Murine models of S. aureus infection are commonplace and have been one of the main tools in developing our understanding of the disease processes. Models of sepsis (bacteraemia) are well-established and are characterised by kidney abscesses as a primary outcome [21,23,24]. Establishment of sepsis requires a high inoculum as is apparent in other murine models. This has been suggested to be due to a bolus of bacteria being required to initiate disease [13,15]. Recently we have shown that, in fact, there is a population bottleneck whereby likely single bacteria found kidney abscesses [13,18]. This therefore requires an explanation as to what the series of events that precedes abscess formation are, and the mechanism (s) involved. The individual bacteria that found lesions occur randomly from the population in that they do not have a genetic advantage [13,18]. Thus the initiation of abscesses is a stochastic event that is enhanced by a large inoculum. Here we aimed to investigate within-host population dynamics and determine the cellular basis of bottlenecks in S. aureus infection models. The temporal and spatial dynamics of sepsis in zebrafish and murine models have been determined using sets of marked strains in a range of backgrounds. This enabled a model for the dynamics of infection in the mouse to be established (Fig 8). We found for all strains that the liver was the key destination organ for the initial mixed population inoculum. Previously S. aureus has been shown to be phagocytosed by Kupffer cells in the liver and these form a primary immunological defence (2,22). Subsequently expansion of individual clones occurs in the liver or the bacterial load is cleared. Kupffer cells are effective agents for the control of S. aureus but if this immunological bottleneck fails then bacteria are able to grow to form abscesses. Depletion of Kupffer cells results in greatly increased host susceptibility to S. aureus infection manifested by abscesses mostly in the livers, but also in the kidneys. Kidney abscesses develop during the infection but these organs are not an initial colonisation site. Interestingly kidneys do not have mixed populations that then become clonal but rather this happens at abscess initiation suggesting that they are founded by single organisms or that the founders were already clonal. Given the complex population dynamics in the liver we suggest that it is the transfer of S. aureus from the liver that gives rise to kidney colonisation. But how therefore do the bacteria traffic between organs? A clue to this comes from the generation of neutropenia within the mouse. Interestingly loss of neutrophils results in no loss of clonality within the liver. Thus the neutrophils are not the primary bottleneck within the liver. However, subsequent abscess formation was abrogated within the kidneys. The importance of neutrophils for dissemination correlates with other research that shows that S. aureus can live intracellularly within phagosomes and be transported by them in the blood, forming a mobile reservoir to infect other organs [5,40,41]. Also, treatment with antibiotics that do not penetrate the neutrophil phagosome do not prevent internal dissemination of S. aureus [30,42]. However, conventionally neutrophils that mature and enter infected tissue do not re-enter the blood stream [28,43]. The most parsimonious explanation is that neutrophils that are already circulating in the blood (whose population greatly increases in response to infection) take up S. aureus that escape into the blood stream from the lesions in the liver. This could be due to abscesses/microlesions shedding S. aureus into the blood stream, as is known to occur in clinical infections [44]. Kidney CFU alone is used as a readout in many studies of S. aureus disease, for which one might express caution as if bacteria have passed through the liver bottleneck then they can clonally expand and so the subtleties of S aureus infection prior to this point will be lost. Liver CFU provides a useful adjunct to kidney numbers. It might seem surprising that a kidney with many apparent surface abscesses could in fact be seeded by an individual founder. It is also important to note that it is unlikely that apparently clonal organs are founded by multiple bacteria of the same type as left and right kidneys are often clonal but with different clones. In order to determine how a single organism could establish a disseminated abscess throughout a kidney we began to map the 3D architecture of the abscess. We developed two methods to address this conundrum. Serial organ sectioning revealed foci of infection with the characteristic abscess structure of infiltration of phagocytes but these extended throughout the kidney forming a seemingly linked series of lesions. Organ clearing is a new approach to study infecting organisms within host tissue and here it revealed a pervasive spread of S. aureus as solid aggregates within cortex tubules and as diffuse groupings with infiltration of phagocytes outside these tubules [33,45]. The scheme presented in Fig 8 suggests the presence of two immunological bottlenecks within the murine sepsis model. Firstly, at the level of Kupffer cells in the liver and then at the neutrophils when disseminating from the liver. Zebrafish embryos do not have Kupffer cells at the time of infection. In the zebrafish the population bottleneck occurs in the phagocytes but is mainly believed to be due to the neutrophils[18]. Our mouse models therefore apparently diverge from the zebrafish model on the relative importance of the two major classes of professional phagocytes. This study provides data that phagocytes present a key breakpoint during infection in that these cells are absolutely required for host resistance but also permit proliferation as “Trojan Horses” as has been previously proposed [40]. The question arises however as to whether this hypothesis translates into a human infection both in terms of the immune system and the occurrence of population bottlenecks. The murine sepsis model shares several characteristics that are known to occur within human infections such as how S. aureus can disseminate in the blood stream to different organs from an initial infection site and the formation of localised abscesses in separate organs. It is known that professional phagocytes are key to controlling S. aureus infections but that S. aureus can also hide intracellularly [5,28]. What we have found correlates with human infections. In our model both neutrophils and macrophages are important for controlling S. aureus infection and the same is true in humans as shown by various genetic disorders in these can result in increased S. aureus infections [46]. This is however more commonly associated with neutrophil disorders (which may reflect how macrophages are generally less dispensable as they are required for tissue development and homeostasis). One particular point is that our model shows that the liver is the site of the primary bottleneck. Generally, in humans, the liver is not a site particularly associated with S. aureus infection (compared to soft tissue infections or deep infections such as osteomyelitis) so it would seem unlikely it would have the same role in humans. However, patients with chronic granulomatous disease (lacking the respiratory burst part of phagocytosis) have a particularly high risk of staphylococcal liver abscesses [46,47]. This implies that in normal people S. aureus can be cleared from the liver as a normal process, which would be by the action of Kupffer cells, the resident professional phagocytes. Our results suggest that the killing activity of macrophages on S. aureus should be worthy of further study but also the ability of neutrophils to transfer S. aureus between organs should also be of great interest as has been shown by other methods [5,30]. Genome studies have shown that there are repeated genetic population bottlenecks and that these occur both during transmission and the shift from colonisation to disease [48–50]. In human infections, population bottlenecks would be interacting with the inherent variation of the naturally colonising S. aureus population as infections tend to be derived from the previous resident S. aureus [48]. The existence of stochastic niches and resource limitation (here access to host nutrients) according to ecological studies are intrinsically supposed to maintain increased population diversity particularly if they are better at competing for different resources [51]. Population bottlenecks, as the host is infected, would form stochastic niches and the exact site would exert different selection pressures. They could therefore work to maintain population diversity and increased strain divergence as the population bottleneck would randomly exclude different subpopulations upon each infection. This may be reflected in the diversity of virulence factors that different strains of S. aureus possess and the maintenance of different clones of S. aureus [52]. Population bottlenecks would also favour variants that are better at getting through it rather than selecting for mutants fitter after subsequent diversification. This concurs with the diversity seen in Young during colonisation and the more limited subsequent diversification [48]. A population bottleneck results in a notable distribution of the observable population. There are members of the administered dose that pass through and come to dominate the population and then there are those that are excluded, which ultimately means there tends to be two groups of data: mixed populations that have not yet passed through the bottleneck and populations that have become, or are rapidly becoming, clonal (as one strain is expanding relative to the others). This can be seen in our data as organs with a mixed population with all 3 strains present and organs with clonal populations. It is reasonable to make assessments based on the proportion of a population that was mixed or clonal as we have done but it would be much harder to assess if one population was becoming clonal more quickly than another (e. g. if one strain or another was better at expanding out of the macrophage bottleneck). Population bottlenecks during infection represent key breakpoints for interventions. Despite apparent success in animal models, vaccine development for S. aureus has not translated into successful human trials [10,53]. Active vaccination is by its nature prophylaxis and thus the target should be to prevent the ability of S. aureus to pass through bottlenecks and proliferate. The knowledge that intracellular organisms form this nexus within phagocytes provides a key parameter from which to design future vaccine formulations. Recent evidence highlighting the targeting of intraphagocytic S. aureus supports such an approach [42]. This is consistent with what has been found previously, in that S. aureus can survive within neutrophils; and treating mice with antibiotics that do not affect intracellular S. aureus does not prevent dissemination [5,30]. Our work, as well as highlighting the cellular locations of immunological bottlenecks also sheds light on the role of the infectious dose on the outcome of infection and the population dynamics therein. The survival challenge model is a standard in the field and has been used in many studies [21,24]. Measuring population dynamics within this model showed a clear diminution of population bottlenecks correlating with an increased infectious dose leading to host morbidity. High infectious doses result in systemic infections across all major organs. The relevance of this model is therefore bought into question as clearly those immunological processes that control the dynamics of infection have been overwhelmed. This observation is therefore of great importance for the design of intervention studies where one is modelling human infection and not uncontrolled bacteraemia. Using lower doses is more realistic as well as allowing population bottlenecks to occur as infectious doses of S. aureus are likely to be inherently lower in humans compared to our mouse models. In Grice et al punch biopsies were used which gave total bacterial numbers of at least 106 CFU/cm2[54]. In a study of vascular catheters, a range of bacteria were found with numbers up to 107 CFU [55]. In a given infection it is likely that only a fraction of these would be able to disseminate from the skin or implant. This is reflected in bacteraemia where the count is typically <10 CFU per ml [56]. The number of bacteria that can transfer to infect organs is much less than >5 x 107 CFU per ml murine blood if 108 CFU is injected (assuming 1. 5 ml of blood per 25g mouse). The threat of antimicrobial resistance to human healthcare is real and increasing. Disease is a complex and dynamic interplay between host and a population of infectious agent. Here we have highlighted key parameters of how S. aureus disease progresses and provide a framework for determination of efficacy of new interventions. Animal work (both mice and zebrafish) was carried out according to guidelines and legislation set out in the UK Animals (Scientific Procedures) Act 1986, under Project Licenses PPL 40/3699 and PPL 40/3574. Ethical approval was granted by the University of Sheffield Local Ethical Review Panel. Staphylococcus aureus strains (S1 Table) were grown using brain heart infusion (BHI) liquid or solid medium (Oxoid) at 37°C, supplemented with the following antibiotics where appropriate: kanamycin 50 μg/ml, tetracycline 5 μg/ml or erythromycin 5 μg/ml plus lincomycin 25 μg/ml (Sigma-Aldrich). S. aureus TetR, EryR and KanR strains used included those constructed previously [13]. The resistance markers were transferred into further strain backgrounds using ϕ11 transduction [57]. ϕ11 variants of suicide vector pMUTIN4 were used to integrate various antibiotic resistance cassettes downstream of the lysA gene in S. aureus in the original strains [13]. The genomic region surrounding lysA is conserved in the Newman and USA300 genomes. The suicide vector pKASBAR (and pKASBARkan) was used to integrate constitutive GFP and mCherry fluorescence markers with promoter Pma1M from the E. faecalis pmv158GFP and pmv158mCherry plasmids into the S. aureus lipase gene in RN4220 [58,59]. The fluorescent markers were then transferred into SH1000, Newman, NewHG and USA300 by ϕ11 transduction. This resulted in strains which were GFP KanR or mCherry TetR. A fragment of c. 1200 bp containing the GFP encoding gene and its promoter sequence was amplified by PCR from pMV158GFP plasmid using the following primers: The PCR fragment was cloned into pKASBARkan by Gibson assembly, with the plasmid cut by EcoRI and BamHI [58]. The resulting plasmid pKASBARGFPkan was introduced into E. coli NEB5-alpha by electroporation with selection for ampicillin resistance. The correct plasmid was confirmed by sequencing and used to transform S. aureus competent cells RN4220 containing a helper plasmid pCL112Δ19 and then selected for with Kanamycin[60]. The marker was then transferred into the NewHG strain (and other backgrounds) by ϕ11 transduction. A fragment of c. 1200 bp containing the mCherry encoding gene and its promoter sequence was amplified by PCR from pMV158mCherry plasmid using the following primers: The PCR fragment was cloned into pKASBAR by Gibson assembly with the plasmid cut by EcoRI and BamHI[58]. The resulting plasmid pKASBARmCherry was introduced into E. coli NEB5-alpha by electroporation with selection for ampicillin resistance. The correct plasmid was confirmed by sequencing and used to transform S. aureus competent cells RN4220 containing a helper plasmid pCL112Δ19 and then selected for with tetracycline [60]. The marker was then transferred into NewHG (and other backgrounds) by ϕ11 transduction. The generated strains however showed only weak tetracycline resistance and were therefore supplemented by EryR from SJF3673 strain collection (lysA: : ery lysA+) by ϕ11 transduction. Transductants with supplemented EryR were verified by PCR. London wild-type (LWT) zebrafish embryos (bred in the MRC CDBG aquarium facilities at the University of Sheffield; see Ethics Statement) were used for all experiments and were incubated in E3 medium at 28. 3°C according to standard protocols [61]. Anaesthetized embryos at 30 hours post fertilization were embedded in 3% w/v methylcellulose and injected individually using microcapillary pipettes filled with bacterial suspension of known concentration into the blood circulation, as previously described [17]. Following infection, embryos were kept individually in 100 μl E3 medium, and observed frequently up to 90 hours post infection; dead embryos removed and CFU from these embryos recorded at each time point. 6–7 week old Female BALB/c mice were purchased from Envigo (formerly Harlan (UK) ) and maintained at the University of Sheffield using standard husbandry procedures. The mice were acclimatised for 1 week. The mice were then inoculated in the tail vein with 100 μl of S. aureus suspension in endotoxin-free PBS (Sigma) diluted from frozen stocks. Viable bacteria in the inoculum were plated on TSB (plus appropriate antibiotics) after serial decimal dilution to confirm the accuracy of the bacterial dose. Mice were monitored and sacrificed at various time-points according to experimental design. All mice were injected with 1x107 CFU S. aureus consisting of a 1: 1: 1 ratio of KanR, EryR & TetR variants of the different strains (SH1000, Newman, NewHG, USA300). 5 mice were sacrificed at the following time points post infection: 2hrs, 18hrs, 48hrs, 72hrs (end of procedure) The mice were injected with the following doses of S. aureus to reflect different levels of challenge that result in both low and high levels of mice that reach the severity limits. The higher dose set: SH1000: 1x108 CFU, NewHG: 1x108 CFU, Newman: 9x107 CFU, USA300: 3x107 CFU The lower dose set: SH1000: 3x107 CFU, NewHG: 3x107 CFU, Newman: 3x107 CFU, USA300: 1x107 CFU. Macrophages were depleted using clodronate liposomes following previously published protocols (NvR, http: //www. clodronateliposomes. org/) [30,31]. The mice were injected iv with 1ml of liposomes per 100g on day 1. The mice were then injected with 1x105 CFU S. aureus on day 2. Mice were sacrificed on day 5 (3 days post infection). Blank liposomes were used as a control. Macrophage depletion was confirmed in pilot studies using histology sections of the liver stained with F4/80 from Serotec AbD Serotec MCA497, followed by an anti-rat red fluorophore. The results confirming macrophage depletion are shown in S6 Fig. Neutrophils were depleted using anti-Ly6G mouse antibodies following previously published protocols [62]. For antibody based neutrophil depletion Invivo anti-Ly6G mouse antibody (1A8, BioXcell) was used [62]. The mice were injected with 1. 5mg/mouse (200ul per mouse) on day 1 with the mice being injected with S. aureus on day 2. Mice were sacrificed on day 5 (3 days post infection). For cyclophosphamide depletion mice were injected intraperitoneally with 150mg/kg cyclophosphamide monohydrate (Sigma, also known as Cyclophosphamide & Cytoxan) on day 1 and 100mg/kg cyclophosphamide intraperitoneally on day 4. The mice were injected with 20mg/ml cyclophosphamide reconstituted in etox free water. The mice were then injected with S. aureus on day 5. Mice were sacrificed on day 8 (3 days post infection). Neutrophil depletion was confirmed using flow cytometry. 100μl of blood was collected via tail bleeding at the same time as S. aureus was injected and 100μl was collected at the end of the experiment via terminal anaesthesia and heart puncture. These blood samples were each mixed with 20μl of Heparin. The blood samples were stained with APC Rat Anti-Mouse Ly6G antibody (BD biosciences, cat. No: 560599) according to the BD bioscience protocol for Immunofluorescent Staining of Mouse and Rat Leukocytes and Fixation with fixation buffer (cat. No: 554655). (http: //www. bdbiosciences. com/eu/resources/s/mouseratleukocytes). The samples were then processed using the BD LSRII flow cytometer. The results confirming neutrophil depletion are shown in S6 Fig. In order to recover bacteria from host tissues, whole zebrafish embryos or mouse organs were individually homogenized in a suitable volume of PBS using the PreCellys 24-Dual (Peqlab) [13]. Homogenates were serially diluted in PBS and plated on TSB (S. aureus) or TSB supplemented with appropriate antibiotics (tetracycline, erythromycin or kanamycin) to determine bacterial numbers. The limit of detection was defined as <100 CFU as defined previously and these results were treated as 0 CFU [13]. Mice were injected with a 1: 1 ratio of 1x107 CFU NewHG GFP KanR/NewHG mCherry EryR, and at 5 days post infection the mice were sacrificed, organs with notable abscesses were sectioned and histology performed. The selected organs were sectioned evenly throughout. Every 300μm, 3 sections (8μm each) of tissue were stained with either DAPI, Hematoxylin and Eosin, or Gram stain. The 300μm sections in between were homogenised separately and plated out for bacterial enumeration as described above. DAPI stained tissue sections were analysed using a Nikon Dual Cam fluorescent microscope whilst the Hematoxylin and Eosin or Gram stained slides were analysed using an Aperio digital microscope slide scanner. Other organs were optically cleared using the ScaleS protocol and were visualised using a Zeiss Z1 lightsheet microscope [33]. Survival experiments were evaluated using the Kaplan-Meier method. Comparisons between curves were performed using the log rank test. For comparisons between CFU groups, a two-tailed, unpaired ANOVA was used. For comparisons of strain ratios where 3 strains were tested (e. g. TetR, EryR and KanR), species evenness was calculated per sample (Species evenness is Shannon’s diversity index H divided by the natural logarithm of species richness ln (S) ) and then compared using a (non-parametric) Kruskal-Wallis test. Bartlett’s test for equal variance was used to test if the mixed strains were equally fit (population spread should be equal). For comparing correlations linear regression was used and the mean was presented on corresponding graphs. All statistical analysis was performed using Prism version 6. 0 (GraphPad) and statistical significance was assumed at p<0. 05. Error bars indicate mean ± One Standard Deviation.

Staphylococcus aureus is a major human pathogen that causes a wide variety of infections. In animal infection models, high doses of S. aureus are generally required to establish infections. We have recently shown in animal models that this is due to very few bacteria within the infecting population going on to cause disease. This population bottleneck during infection is an ideal target for treatment development. Here we have shown that this bottleneck is common to a range of different S. aureus types and infection models. The mouse survival model is commonly used for testing antistaphylococcal treatments. We show that the large infectious dose it requires diminishes the bacterial population bottleneck. This suggests the host immune system is simply overwhelmed and brings into question the applicability of such a model. We found that both macrophages and neutrophils are important for resistance to S. aureus infection but these affect the population bottleneck differently. Macrophages in the liver mediate the initial population bottleneck whereas neutrophils enable the subsequent spread of bacteria to other organs. This has allowed a model of S. aureus disease progression to be established and our study gives insights into how novel treatments could be established to counter it.

Abstract Introduction Results Discussion Materials & methods

blood cells medicine and health sciences fish liver immune cells pathology and laboratory medicine pathogens immunology microbiology vertebrates staphylococcus aureus animals animal models osteichthyes model organisms signs and symptoms experimental organism systems kidneys bacteria neutrophils bacterial pathogens research and analysis methods abscesses white blood cells animal cells staphylococcus medical microbiology microbial pathogens mouse models zebrafish eukaryota diagnostic medicine anatomy cell biology biology and life sciences cellular types renal system macrophages organisms

2018

Staphylococcus aureus infection dynamics

Novel drugs are required for the elimination of infections caused by filarial worms, as most commonly used drugs largely target the microfilariae or first stage larvae of these infections. Previous studies, conducted in vitro, have shown that inhibition of Hsp90 kills adult Brugia pahangi. As numerous small molecule inhibitors of Hsp90 have been developed for use in cancer chemotherapy, we tested the activity of several novel Hsp90 inhibitors in a fluorescence polarization assay and against microfilariae and adult worms of Brugia in vitro. The results from all three assays correlated reasonably well and one particular compound, NVP-AUY922, was shown to be particularly active, inhibiting Mf output from female worms at concentrations as low as 5. 0 nanomolar after 6 days exposure to drug. NVP-AUY922 was also active on adult worms after a short 24 h exposure to drug. Based on these in vitro data, NVP-AUY922 was tested in vivo in a mouse model and was shown to significantly reduce the recovery of both adult worms and microfilariae. These studies provide proof of principle that the repurposing of currently available Hsp90 inhibitors may have potential for the development of novel agents with macrofilaricidal properties. Infections caused by the parasitic filarial nematodes Wuchereria bancrofti, Brugia malayi and Onchocerca volvulus remain a significant cause of pathology in the tropics. The adult stages of these pathogens are extremely difficult to kill with currently available drugs. Treatment relies upon two compounds, ivermectin (IVM) or diethylcarbamazine (DEC), both of which largely target the larval stage of the life cycle (the microfilariae, Mf). In the Global Campaign for the Elimination of Lymphatic Filariasis, either DEC or IVM is combined with albendazole. While this approach effectively disrupts transmission [1], Mf repopulate the circulation, necessitating the repeated administration of drug. As the reproductive life span of the adult female worm is estimated to be around 10 years for the lymphatic species [2] and longer for Onchocerca volvulus [3], programs aimed at eradication of these parasites are faced with a considerable challenge, as treatment must be continued over this long timescale. At least for O. volvulus, the repeated administration of ivermectin over many years is associated with treatment failures [4], although whether these truly reflect resistance remains the subject of debate. Consequently, drugs that target adult filarial worms would be a major advantage in control programs aimed at eliminating these parasites [5]. Heat shock protein 90 (Hsp90) has emerged in recent years as a validated target for the therapy of various tumors [6], resulting in the development of many Hsp90-specific small molecule inhibitors. Hsp90 is essential in all eukaryotes and several recent studies have demonstrated the activity of specific inhibitors against a variety of tropical pathogens, such as Plasmodium [7], [8], Trypanosoma sp [9] Leishmania sp [10] and the filarial worm Brugia [11], [12]. The repurposing of compounds designed for one purpose to control of tropical infections is an attractive proposition [13], generating considerable enthusiasm in the pharmaceutical industry. Starting the search for new therapeutics for these diseases with drug-like molecules offers several short cuts, as these have already passed the basic criteria for development, have usually been optimized for their drug-like qualities and have often undergone toxicity testing. Here we compare the efficacy of several classes of Hsp90 inhibitor against the lymphatic filarial nematode Brugia. The prototype Hsp90 inhibitor is geldanamycin (GA), a fermentation product of Streptomyces species that binds at the N-terminal ATP domain of Hsp90 disrupting its function [14]. Hsp90 acts as a molecular chaperone helping to fold and stabilize a variety of different proteins, the so-called ‘client’ proteins, many of which are involved in signal transduction [6]. The realization that Hsp90 client proteins, such as those encoded by oncogenes, were unable to attain their active conformation and were degraded following exposure to GA led to studies in animal models of various cancers. However, GA suffers from several target-unrelated limitations as an in vivo chemotherapeutic agent because of its chemical structure, as it contains a benzoquinone ring, rendering it hepatotoxic [15]. GA has been extensively modified to limit these liabilities and some of the resulting derivatives are still undergoing clinical assessment (reviewed in [16]). However, most recent efforts have been directed at developing synthetic small molecule inhibitors of distinct chemical scaffold, such as the purine-scaffold series [17], that bind at the same site as GA but lack the target-unrelated liabilities. These molecules have undergone considerable modification and one compound, PU-H71, shows potential in the clinic [18], [19]. Several additional N-terminal Hsp90 inhibitors have been identified in high throughput screens, including the pyrazole, isoxazole and triazole resorcinol classes such as VER-50589, NVP-AUY922 and STA-9090 (ganetespib), respectively [20], [21]. NVP-AUY922 is progressing through Phase I/II clinical trials while STA-9090 has advanced to Phase III [22], [23]. An additional class of compound, the Serenex series, also progressed to phase I/II clinical trials (reviewed in [22]). In this paper we report the efficacy of five inhibitors, representing four different classes of compound, on adult Brugia in vitro and compare the results with those from screens based on Mf viability and a fluorescence polarization assay. We focus on the most active compound, NVP-AUY922, and describe its in vitro effects on three life cycle stages of Brugia and its efficacy against adult worms in vivo. All animal protocols were carried out in accordance with the guidelines of the UK Home Office, under the Animal (Scientific Procedures) Act 1986, following approval by the University of Glasgow Ethical Review Panel. Experiments were performed under the authority of the UK Home Office, project numbers 60/4448 and 60/3792. The Brugia pahangi life cycle was maintained by serial passage through mosquitoes (Aedes aegypti, Refm) and jirds, Meriones unguiculatus, as described previously [24]. Adult worms of B. pahangi were obtained from infected jirds after 3–4 months, exactly as described previously [11] and were frozen in liquid nitrogen, ground in a pestle and mortar to a fine powder and re-suspended in an appropriate volume of HFB assay buffer (20 mM HEPES, pH 7. 3,50 mM KCl, 5 mM MgCl2,20 mM Na2MoO4,1% NP40). Protein concentrations were estimated using the BioRad protein assay. At this point lysates were freeze-dried for shipping to the USA. The FP assay was set up essentially as described previously [12], [25]. In brief, assays were performed in black 96-well half-volume non-binding microtiter plates (Corning #3686) in a total volume of 50 µl per well. Assay buffer (HFB2) contained 20 mM HEPES, pH 7. 3,50 mM KCl, 2 mM EDTA, 0. 01% Triton-X100,0. 1 mg/ml bovine gamma globulin (Sigma #G5009, Saint Louis, MO), 2 mM DTT (Sigma, Saint Louis, MO) and protease inhibitor cocktail (Roche #11836170, Indianapolis, IN). The equilibrium binding of Cy3B-GA and recombinant human Hsp90α (Enzo Life Sciences, Farmingdale, NY USA) or parasite lysate was determined by creating a two-fold dilution series of protein/extract for an eleven-point curve with the first column containing no protein. The dilution series was incubated with 6 nM Cy3B-GA in triplicate at 4°C with gentle shaking for different periods of time and FP measurements taken on a Safire2 plate reader (Tecan, San Jose, CA) with excitation and emission wavelengths of 530 nm and 585 nm, respectively, and a bandwidth of 20 nm. All FP values are expressed in millipolarization (mP) units with the mP value of free Cy3B-GA probe set to 50. Equilibrium binding constants were determined by nonlinear regression using a one-site binding model (GraphPad Prism software). The relative binding affinities of inhibitors to human or parasite-derived Hsp90 was determined using competitive FP binding assays. Human Hsp90 was used at a concentration resulting in 50% maximal binding of 6 nM Cy3B-GA (2. 4 nM for human Hsp90α). For parasite-derived Hsp90, an amount of parasite lysate resulting in 50% of maximal Cy3B-GA binding was used. The drugs tested in the FP assay were GA and 17-AAG, CCTC018159, VER-49009, VER-50589, NBP-AUY922, NVP-BEP800, CAY 10607, BIIB021, PU-H71, SNX-2112, SNX-9203 and HSP990. Stock solutions of each compound were prepared in DMSO at a concentration of 10 mM and 3-fold serial dilutions prepared in DMSO for eleven point curves. Drugs were then diluted 100-fold into HFB2 assay buffer containing 12 nM Cy3B-GA in 96-well storage plates to create 2X drug solutions. Drug solutions (25 µl/well) were then transferred in duplicate to 96-well black assay plates (Corning#3686) containing 25 µl HFB2 with 2X the final desired concentration of Hsp90. The final concentration of Cy3B-GA was 6 nM and the final DMSO concentration in all wells was 0. 5%. Free Cy3B-GA (mP set to 50) and buffer only (background) wells were included as controls on each plate. Plates were incubated at 8°C with gentle shaking for 20 h. FP measurements were taken and the inhibitor concentration at which 50% of bound Cy3b-GA was displaced (IC50) was determined using nonlinear regression with a four parameter logistic equation (GraphPad Prism software). The five new compounds selected for in vitro testing on B. pahangi were NVP-AUY922, NVP-BEP800, SNX-2112, SNX-9203 and BIIB021. GA was used as a positive control in some experiments. All compounds were supplied by Selleck Chemicals (www. selleckchem. com), with the exception of GA, which was supplied by MBL International Corporation (Woburn, MA). Drugs were dissolved in DMSO to give a stock solution of 10 mM, then aliquoted and stored at −20°C. Working concentrations of drugs were prepared on the day of use by dilution in tissue culture medium. For in vivo studies, NVP-AUY922 was purchased from LC Laboratories (www. LCLabs. com) and dissolved in DMSO at 50 mg/ml. As adult Brugia worms are limited in numbers, initial experiments assessed the effect of each drug on Mf viability. Mf were purified from infected animals essentially as described previously [26]. In brief, following lavage of the peritoneal cavity of an infected animal with Hanks Balanced Salt Solution (HBSS) pre-warmed to 37°C, Mf were collected by centrifugation and then purified from host cells by centrifugation through lymphoprep (Sigma). This procedure was repeated twice. Mf were collected from the pellet, washed twice in HBSS and once in worm culture medium (WCM) which comprised RPMI 1640, (Invitrogen Cat No: 52400), containing 5% heat inactivated fetal calf serum, 1% glucose, 100 units/ml penicillin and 100 µg/ml streptomycin (all Invitrogen). Mf were then dispensed into the wells of a 24-well plate to give approximately 200 Mf in 2. 0 ml, using a single well for each drug concentration. All procedures were carried out using aseptic techniques. The five novel compounds, plus GA and medium alone controls, were tested three times against Mf over the full range of concentrations. For adult worm assays, adult female B. pahangi, 3–4 months old, were incubated individually in 2. 0 ml of WCM overnight in 24-well plates and pre-screened for Mf production. Any worms that failed to produce Mf overnight were discarded. For the drug experiments, six female worms of B. pahangi for each concentration of drug were cultured individually in 24-well plates in 2. 0 ml of WCM containing drug, or carrier alone (DMSO) at a concentration equal to that in the highest concentration of drug. In some experiments, GA was used as a positive control. Initially, all five compounds were tested over selected concentrations starting at 2 µM to 100 nM (see Results for details). For Hsp90 inhibitors, Mf output by individual female worms is a sensitive indicator of adult worm viability and, in most experiments, was assessed at 72 h. In addition, adult worms were examined microscopically on a daily basis for 7–10 days to determine whether lower concentrations of drugs had any effects over a longer period of incubation. Results are expressed as mean Mf output ± SD over a 72 h period. Statistical significance between groups was calculated using the Mann Whitney test with P values<0. 05 being considered significant. In two additional experiments, adult worms were exposed to a short 24 h incubation in 250,25 or 10 nM NVP-AUY922 or DMSO in medium alone, using six worms per concentration as described above. After 24 h in drug, adult worms were removed, washed out of drug and incubated in medium alone. Mf output was counted after 24 h in drug and again after 24 h or 48 h in medium alone. Plates were maintained for up to 10 days and the condition of adult parasites noted at regular intervals. In one experiment the effect of GA on Mf output by adult B. malayi worms was compared with B. pahangi. B. malayi worms were kindly provided by Prof. R. Maizels (University of Edinburgh). In this experiment, Mf were counted after 48 h of culture in GA at 2. 0 µM and 1. 0 µM GA. In all in vitro experiments, plates were viewed daily and the motility and condition of the parasites noted by two independent observers, of whom one was unaware of the contents of the well. L3 stages were harvested from mosquitoes nine days post-infection. L3 were picked individually with a fine glass pipette, counted and washed three times in HBSS containing 1000 units/ml of penicillin and 1000 µg/ml streptomycin by sedimentation at room temperature. 20–30 L3 per well were plated out in duplicate in 24-well plates in 2. 0 ml of WCM containing drug, or carrier alone (DMSO) and cultured at 37°C for up to 7 days. NVP-AUY922 was tested at a range of concentrations: 500,250,100,50,25,10,5. 0,1. 0,0. 5 and 0. 1 nM. This experiment was repeated twice. Adult worms were removed from the peritoneal cavity of infected jirds and rinsed in HBSS. 10 adult female worms were transplanted into the peritoneal cavity of each of ten male BALB/c mouse using standard methods [27], [28]. Adult worms were transferred into the peritoneal cavity of anaesthetized mice using a blunted glass hook. Not all mice received the full quota of 10 worms (see Table 1). The stock solution of drug was diluted to 5 mg/ml in sterile PBS containing Tween 20, to a final concentration 5%, and DMSO to a final concentration of 10%, as detailed previously [29]. Five mice were treated with 50 mg/Kg NVP-AUY922 and five with sterile PBS/Tween 20/DMSO by intra-peritoneal injection at three time points: day 0 (7 days post-transplantation), day 3 and day 7. This dose was selected on the basis of previous studies in a mouse xenograft model [29]. Adult worms and Mf were recovered 9 days after the last dose of drug by peritoneal lavage with pre-warmed HBSS. The condition of any adult worms recovered was noted. Mf in the first 12 ml of peritoneal washings were pelleted by centrifugation and fixed in 2% formalin in water and stored at 4°C until counted. Mice were weighed at each time point to monitor any possible weight loss. Several classes of inhibitor were screened against B. pahangi or B. malayi lysates, as well as human Hsp90α, using the FP assay originally developed as a high throughput screen for Hsp90 inhibitors in tumor cells, as previously applied to Brugia [12]. This assay is based on the ability of small molecules to inhibit the binding of Cy3B labeled GA to Hsp90. Table 2 shows the range of IC50 values for the binding of selected compounds to B. pahangi, B. malayi or human Hsp90α. In general, most compounds bound to parasite-derived and human Hsp90 with broadly similar affinities. Several compounds bound worm Hsp90 with high affinity, including NVP-AUY922 and VER-50589, each of which bound Brugia Hsp90 at low nanomolar concentrations (IC50 1–2 nM). A second group of compounds including GA, PU-H71, SNX-2112, SNX-9203, BIIB021 and VER-49009 bound Brugia Hsp90 with an IC50 in the range of 10–20 nM. Most drugs bound to Hsp90 from both species of Brugia with broadly similar affinity; CAY10607 was an exception in this respect, showing a 10-fold higher affinity for B. pahangi Hsp90 compared to B. malayi Hsp90. We initially compared the efficacy of Hsp90 inhibitors against Mf as this life cycle stage is extremely abundant, while the numbers of adult worms are more restricted. In these experiments six drugs, representing five different chemotypes were selected including NVP-AUY922, NVP-BEP800, SNX-2112, SNX-9203, BIIB021 and GA (see Fig. 1 for drug structures), were tested over ten doubling dilutions from 4. 0 µM to 0. 976 nM and compared with the DMSO vehicle. The results of these experiments were clear-cut: NVP-AUY922 was by far the most effective compound tested, killing 100% of Mf by day 7 at all concentrations down to and including 1. 95 nM. At the lowest concentration of NVP-AUY922 tested (0. 976 nM), ∼75% of the Mf were dead after 7 days incubation. In contrast, a concentration of 500 nM NVP-BEP800 was required to kill the majority of Mf by day 7, with ∼50% death at 250 nM. Lower concentrations of NVP-BEP800 affected Mf motility but did not kill substantial numbers of worms. For the SNX compounds, SNX-2112 killed approximately 50% of Mf at 500 nM by day 7, while at 250 nM, Mf were very sluggish but still alive. SNX-9203 was slightly more effective at lower concentrations, killing ∼90% of Mf at 500 nM after 7 days exposure, while at 250 nM most worms were alive but were much less motile than controls. Finally, BIIB021 was active only at 4. 0 µM, while lower concentrations affected worm motility but did not kill the worms within a 7-day period. In comparison, GA killed ∼90% of the Mf at 31. 25 nM, while at 15. 6 nM ∼50% of the Mf died, similar to the effective dose reported previously [11]. Thus, NVP-AUY922 was the most efficient inhibitor of Hsp90 as judged by Mf killing followed by GA, while BIIB021 was the least effective and the SNX compounds and NVP-BEP800 had broadly similar effects on Mf viability (see Table 3 for summary). We have previously described a sensitive assay to record the effect of GA and the purine scaffold inhibitors on adult worm viability that involves counting Mf output over a designated time of exposure to inhibitor [11], [12]. In initial experiments, all five new compounds were screened against adult female worms at concentrations ranging from 2. 0 µM to 100 nM and Mf output and adult worm viability monitored over a 7-day period (data not shown). Three of the five compounds tested (NVP-AUY922, NVP-BEP800 and SNX-9203) had a significant effect on Mf output at all doses tested including 100 nM, while SNX-2112 was significant only at 500 nM and BIIB021 at 2. 0 µM. In additional experiments, NVP-BEP800 and the SNX compounds were further titrated from 500 nM to 1. 0 nM to estimate the minimal effective concentration. For all three drugs a concentration of 500 nM was required to consistently show a significant effect on Mf output by adult worms in replicate experiments. Further experiments dealt only with the most effective compound NVP-AUY922, which was titrated over a range of concentrations from 500 nM to 1. 0 nM. Following 72 h of exposure to all concentrations from 500 nM to 10 nM NVP-AUY922, a significant inhibitory effect on Mf output was observed (P = 0. 0087 for 10 nM vs DMSO, see Fig. 2). Continuing the cultures for a further three days in the presence of drug resulted in a significant decrease in Mf output at 5. 0 nM (P = 0. 0043) but not at 1. 0 nM. By 6 days of exposure all adult worms were dead at concentrations of 100 nM NVP-AUY922 and above, and, in a typical experiment (shown in Fig. 2), 4/6 worms were dead at 50 nM drug and 3/6 were dead at 25 nM drug. The remaining worms although alive, were extremely sluggish. Although the motility of the adult worms was affected by exposure to lower concentrations of NVP-AUY922 (10 and 5. 0 nM), they did not die at these concentrations. Worms that were dying tended to burst and release their uterine contents and wells contained many embryonic stages in addition to Mf. Confirmation that NVP-AUY922 has a direct macrofilaricidal effect was obtained by exposing male worms to a range of concentrations from 5. 0 µM to 100 nM. By day 7 of exposure 100% of male worms exposed to 100 nM drug were dead. While for human filarial parasites, drugs that target adult worms are the priority, it was of interest to determine whether NVP-AUY922 also killed the infective form of the parasite, the L3. In these experiments, L3 were harvested directly from mosquitoes, washed and exposed to varying concentrations of NVP-AUY922 from 500 nM to 0. 1 nM in WCM. After 6 days exposure to drug, 100% of L3 were dead at all concentration down to and including 10 nM. At 5. 0 nM approximately 30% of parasites were dead and the remainder were moving very slowly. At this time point there was no significant mortality in control wells and, at concentrations of 1 nM or below, no effect on the L3 was observed. Thus NVP-AUY922 is toxic to L3 stages at relatively low concentrations. In the experiments described above, adult worms or Mf were continuously exposed to drug and viability assessed. However, as any drug that might be used against filarial worms in vivo would be required to exert its effect over a limited period of time, we carried out two additional experiments to determine the outcome of exposing adult female worms to NVP-AUY922 for a 24 h period only. In these experiments six adult female worms per group were cultured individually with 250,25 or 10 nM NVP-AUY922 or the appropriate concentration of DMSO alone for 24 h. Worms were then washed free of drug and cultured in medium alone for a further 7–9 days. After 24 h in drug, significantly fewer Mf were produced at 250 nM and at 25 nM NVP-AUY922 (P = 0. 0043 for both concentrations) but no significant difference was observed in Mf output in worms incubated with 10 nM drug. After 24 h in drug and 24 h in medium alone, Mf production was almost completely inhibited from worms cultured in 250 nM drug (mean of 11±23 Mf) or 25 nM drug (mean of 100±60 Mf) compared to DMSO controls (1658±198, P<0. 05 for both 250 nm and 25 nM drug). Although worms exposed to 10 nM drug for 24 h produced fewer Mf (mean 910±674 Mf) than DMSO controls this difference failed to reach significance (P = 0. 0635). However following 24 h in drug and 48 h in medium alone, adult worms exposed to 10 nM NVP-AUY922 for 24 h produced significantly fewer Mf than control worms (P = 0. 0317) (see Fig. 3). Adult worms were clearly affected by a short-term exposure to 250 nM NVP-AUY922, being much less motile than control worms after 24 h in medium alone. By 48 h they were elongate and by 9 days in medium alone, they were barely moving, but they were still alive. Worms exposed to 25 nM NVP-AUY922 were also noticeably more sluggish than control worms at day 9, while no obvious difference was observed between those incubated in 10 nM drug and DMSO controls. The experiment was discontinued at this point. Most of our experiments have been carried out using B. pahangi, a close relative of the human parasite, B. malayi. We compared the efficacy of GA at two concentrations (1. 0 and 2. 0 µM) on B. malayi in parallel experiments with B. pahangi. Not surprisingly, given the degree of amino acid identity between Hsp90 in the two species (93. 5% identical), both were equally sensitive to Hsp90 inhibition. In this experiment, the reduction in Mf output at any one concentration of drug was almost identical: at 2. 0 µM GA, there was a 93% reduction in Mf output in B. malayi and 92% with B. pahangi while at 1. 0 µM, there was a 74% reduction in Mf output with B. malayi and a 78% reduction with B. pahangi after 48 h of drug exposure (P = 0. 0043 for all concentrations of GA versus DMSO control). As NVP-AUY922 appeared to be extremely active at low concentrations after a short exposure in vitro, it was pertinent to determine whether it would have activity in vivo against adult worms transplanted into the peritoneal cavity of BALB/c mice. Three animals in the treated group were given 10 adult worms, while the remaining two mice received 9 worms, while in the control group, one animal received 9 worms, one received 8 worms and the remaining three mice received 10 worms. Following adult worm transplant, mice were randomly assigned to a treatment group (five per group) and received either 50 mg/Kg NVP-AUY922 at three time points by intra-peritoneal injection or an injection of PBS/Tween 20/DMSO. Mice were weighed at each treatment and prior to recovery of adult worms, but no weight loss was observed in drug-treated animals over the time course of the experiment. Adult worms were recovered 9 days after the last injection of drug, at which point there was a significant reduction in worm recovery from treated mice compared to control animals (P = 0. 0109) (see Table 1). All control animals contained live, motile adult worms with recoveries varying from 30–78% of transplanted worms (see Table 1 for details). In contrast, very few live worms were recovered from NVP-AUY922 treated mice (recovery of live worms ranged from 0–11%, Table 1). Adult parasites recovered from all animals were placed in HBSS at 37°C for 2–3 hours and examined again. There was no evidence to suggest that the adult worms recovered from drug-treated animals regained their motility over this time period. In three out of five treated animals, only a few dead Mf were observed in the peritoneal washings, while the remaining two animals contained low numbers of Mf that were very slow moving (P = 0. 0079, NVP-AUY922 versus control). In this paper we extend our observations on the effect of Hsp90 inhibitors on adult Brugia worms in vitro and in vivo. A panel of commercially available Hsp90 inhibitors, all designed and optimized for binding to human Hsp90, was profiled in FP binding assays. The broadly similar binding affinities between human and parasite proteins highlight the evolutionarily conserved structure of the nucleotide-binding domain that is targeted by these inhibitors. Selectivity towards parasite Hsp90 would obviously be preferable, but this endeavor would require a structure-based design effort, as recently described for trypanosome Hsp83 (Hsp90) [30]. In the spirit of exploring the potential for a direct repurposing strategy, the translation from parasite Hsp90 binding to filaricidal activity was examined. Five clinically viable compounds belonging to four different drug classes were tested in vitro for their ability to kill Mf and inhibit Mf output from adult female worms. The results from all three systems were reasonably consistent and highlight the efficacy of one specific inhibitor, NVP-AUY922. This compound showed significant activity against adult female worms at a concentration of 25 nM after 6 days exposure, significantly inhibiting Mf output and killing 50% of adult worms. Although lower concentrations of NVP-AUY922 (down to 5. 0 nM) showed a significant effect on Mf output, adult worms were not killed at this concentration over the time scale of the experiment. NVP-AUY922 showed a high affinity for Brugia Hsp90 in the FP screen, further validating the application of this assay as a high throughput screen. It was also the most active compound tested in the Mf killing assay. As a single infected animal can produce millions of Mf, these results indicate that using Mf could provide a simple, inexpensive and relatively high throughput pre-screen for compounds with macrofilaricidal activity, while acknowledging that not all compounds that effect Mf will also kill adult worms. The translation from Brugia Hsp90 binding to micro- and macro-filaricidal activity depends upon the chemotype being tested and this is evident in the compounds used here. While NVP-AUY922 bound Brugia Hsp90 at ∼1 nM and had an EC50 against Mf viability in the same range, the other compounds exhibited variable translation between the two assays. For example, NVP-BEP800 bound quite well to Brugia Hsp90 (∼7 nM IC50) but required much higher concentrations to kill microfilaria (∼250 nM EC50) during an extended 7-day incubation. There are many possible explanations for this discrepancy, including failure of the compound to fully equilibrate within the microfilariae during the experiment. It is possible that this particular compound does not penetrate well or is actively exported from the parasites. As more detailed studies would be required to address this issue, we continued to focus on compounds with the best activity against live parasites. We assessed the effect of these inhibitors on adult female worms by quantifying Mf output, as a measure of worm viability. Mf output is an active process in filarial worms and relies on the presence of live Mf in utero and on the activity of the vulva, a muscular opening close to the anterior end of the worm. When Hsp90 is inhibited, cessation of Mf output is a sensitive surrogate measure of adult female worm health and appears to be one of the first signs that worm viability is compromised. Once Mf production ceases following exposure to Hsp90 inhibitors, it does not resume, at least over the time scale of our in vitro experiments. In addition, previous studies [11] demonstrated that embryogenesis in adult B. pahangi was disrupted upon inhibition of Hsp90 with GA. When RNAi was used to knockdown hsp90 (daf-21) in the free-living nematode Caenorhabditis elegans, one of the most penetrant phenotypes observed was a protruding vulva and sterility in the F1 generation [31]. In C. elegans, the vulva is a complex structure, the development of which is dependent upon a number of signaling pathways (reviewed in [32]). The vulva is innervated by neurons and egg-laying in C. elegans (the equivalent of Mf release in filarial worms) is regulated by multiple factors and molecules, including neuropeptides, kinases and members of the TGF-β signaling pathway [33]. Much less is known of the factors that regulate Mf production in filarial nematodes, but the vulva may be particularly sensitive to Hsp90 inhibitors, perhaps acting on various kinases required for signaling in this structure. Alternatively, the rapid inhibition of Mf output observed (within 24 h at high concentrations of inhibitor) may reflect a general demise in adult female worms upon inhibition of Hsp90. In the related filarial nematode O. volvulus, ivermectin is reported to inhibit Mf output, while not killing the adult worm. In cattle infected with O. ochengi, embryogenesis is disrupted by treatment with ivermectin and an accumulation of dead and dying Mf are observed [34], similar to that described for female Brugia exposed to GA [11]. However, the results of the present study show that inhibition of Hsp90 by NVP-AUY922 not only affects Mf release but also directly affects the viability of adult worms. This property was clearly demonstrated following in vitro exposure of both male and female adults to the inhibitor and in vivo administration of the inhibitor to mice following transplant of female worms. Hsp90 functions in a complex with other proteins to fold and/or stabilize a wide variety of client proteins, most of which have been identified from mammalian cells or yeast. A list of Hsp90 client proteins is curated by the Picard Lab (http: //www. picard. ch). However, we know little about the key client proteins clients in Brugia, except by homology with known interacting proteins identified from other systems. A better understanding of the Hsp90 interactome in Brugia may help explain the effects observed upon chemical inhibition of Hsp90. As all our previous studies have focused on B. pahangi, we also tested B. malayi in this study and showed that GA had very similar effects upon both species. Similar data was recently published [35] showing that GA and four derivatives of GA, were active in vitro on adult B. malayi and the trematode parasite Schistosoma japonicum. In that study, the lowest concentration tested was 500 nM, but encouragingly all compounds were active against adult B. malayi at this concentration. Interestingly, NVP-AUY922 also showed potent activity against the L3 stage of B. pahangi. While drugs that target the L3 stage of filarial nematodes are not the highest priority for human use, they are of significant interest in the veterinary field, where prophylaxis of the dog heartworm, Dirofilaria immitis, is a major area of concern in veterinary practice. Here, the macrocyclic lactones, such as ivermectin and moxidectin, have been the mainstay of control for many years. Given the recent observation on treatment failures with ivermectin in some dogs in the Southern states of the USA [36], there is a clear need for the development of novel compounds to protect susceptible animals. As a preliminary to in vivo testing, we investigated whether a short exposure to the most effective compound identified, NVP-AUY922, affected adult worm viability in vitro. These experiments showed that Mf output was significantly reduced following a 24 h exposure to all concentrations of drug including 10 nM (the lowest concentration tested in this study), although it required an additional 48 h incubation in medium alone to detect this effect. However, although Mf output was significantly different between drug exposed and control worms, a direct effect was observed at 250 nM concentration after a 24 h exposure, with adult female worms being almost motionless after 9 days in culture. While a 24 h exposure to 25 nM NVP-AUY922 resulted in a significant decrease in motility, the adult female worms were still alive at day 9. Whether such motility-impaired worms would be viable in vivo in an immuno-competent host is doubtful. These data also highlight one of the characteristics of Hsp90 inhibitors: the time taken to kill adult worms. They do not kill rapidly, which may be a useful attribute for in vivo use, as some of the pathogenesis of lymphatic filariasis is associated with death of the adult worms, and the subsequent release of antigen and the Wolbachia endosymbiont [37]. NVP-AUY922 belongs to the isoxazole resorcinol class of Hsp90 inhibitors and has shown some promise as an anti-tumor agent in a mouse xenograft tumor model [29], [38]. In the first of these studies, pharmacokinetic analysis showed that the plasma levels of drug reached a maximum of 52,506 nM at 0. 25 h post-dosing following a single dose of 50 mg/Kg delivered by the intra-peritoneal route. As a 24 h exposure to 250 nM NVP-AUY922 had a significant deleterious effect on adult female worms, we were encouraged to test this compound in vivo, using an adoptive transplant system in which viable adult worms are transplanted into the peritoneal cavity of mice [27], [28]. In previous experiments with NVP-AUY922 administered in a similar dosing regimen to nude mice containing xeno-grafted tumors, weight loss was observed [38]. However our immune-competent animals showed no loss in weight or other obvious deleterious effects over the time scale of the experiment. Adult worms were recovered nine days after the last dose of drug, as the in vitro data indicated that it might take some time for the drug to act. Additionally, previous studies with suramin, one of the few compounds with activity against adult filarial worms, demonstrated that it took approximately 6 weeks to reduce adult worm recovery in the jird model [39]. A total of eight out of 48 transplanted female worms were recovered from drug treated animals, but of these only two were healthy with the others being coated in cells, indicating that they were in the process of being cleared by the immune system. In contrast, 26 female worms out xof 47 transplanted were recovered from animal treated with vehicle alone. Moreover, only two of the drug-treated mice contained live Mf, and these appeared very sluggish compared to controls. Further studies will be required to examine alternative routes of drug administration and to explore in more detail the mechanism by which these inhibitors exert their macrofilaricidal effect. The clinically viable Hsp90 inhibitors tested here are targeted and potent towards human Hsp90 and thus may have unwanted side effects. In this respect, the therapeutic potential of any of these inhibitors for filarial disease will depend on the pharmacodynamic differences between host toxicities and parasite killing. There are pharmacokinetic data in both animals and humans available for many of the Hsp90 inhibitors. We have used this data in combination with in vitro worm killing assays to predict the doses of NVP-AUY922 that would be successful in killing parasites in vivo. While efficacy in cancer treatment likely requires chronic dosing of near maximum tolerated doses, the treatment of parasitic diseases would ideally necessitate a short course of treatment. It is not currently clear whether a short treatment course with current inhibitors would be able to clear parasites, while maintaining appropriate safety margins. Additional studies will be required to carefully delineate the therapeutic window of short treatment courses. As noted earlier, the development of parasite-selective inhibitors with reduced potential for host toxicity would be of enormous benefit and should also be pursued. The ubiquitous nature of Hsp90 in normal as well as transformed cells, has led some to question the potential of this molecule as a drug target. However, for the past 20 years, the National Cancer Institute has advocated Hsp90 as a drug target, since GA was first shown to exhibit anti-tumor properties. There are currently seventeen Hsp90 inhibitors in clinical trials and a growing arsenal of novel Hsp90 inhibitors, of structurally diverse scaffold (reviewed in [40]). Current emphasis is on combination therapies in which Hsp90 inhibitors are combined with other anti-tumor drugs [41]. Whether a similar approach would further enhance the activity of these inhibitors against adult filarial parasites remains to be determined. However, we believe that these studies provide data to further strengthen the contention that inhibition of Hsp90 is a valid target for the chemotherapy of lymphatic filariasis, while acknowledging that a significant medicinal chemistry effort would be required to optimize the activity of these inhibitors.

Adult filarial worms are long-lived nematode parasites that have proved very difficult to kill with existing drugs. Current campaigns for the control or elimination of these parasites are largely based on treatment with drugs such as diethylcarbamazine or ivermectin, that preferentially kill the first stage larvae of the parasite, the microfilariae. As microfilariae repopulate the body from unaffected adult worms, repeated dosing with these drugs is required over the long reproductive life span of the worm. The availability of compounds with macrofilaricidal activity would help facilitate the goal of controlling filarial infections. Hsp90 is a recognized target in tumor cells: consequently many oncology programs have developed small molecule inhibitors of Hsp90, several of which are commercially available. Here we provide proof of principle that inhibition of Hsp90 is lethal to adult Brugia worms in vivo, as well as in vitro, suggesting that these compounds may have potential for further development as macrofilaricidal drugs.

Abstract Introduction Methods Results Discussion

oncology medicine chemotherapy and drug treatment cancer treatment biology microbiology parasitology

2014

A Repurposing Strategy for Hsp90 Inhibitors Demonstrates Their Potency against Filarial Nematodes

Spikelets are small spike-like depolarizations that can be measured in somatic intracellular recordings. Their origin in pyramidal neurons remains controversial. To explain spikelet generation, we propose a novel single-cell mechanism: somato-dendritic input generates action potentials at the axon initial segment that may fail to activate the soma and manifest as somatic spikelets. Using mathematical analysis and numerical simulations of compartmental neuron models, we identified four key factors controlling spikelet generation: (1) difference in firing threshold, (2) impedance mismatch, and (3) electrotonic separation between the soma and the axon initial segment, as well as (4) input amplitude. Because spikelets involve forward propagation of action potentials along the axon while they avoid full depolarization of the somato-dendritic compartments, we conjecture that this mode of operation saves energy and regulates dendritic plasticity while still allowing for a read-out of results of neuronal computations. Brain functions rely on computations in single neurons, but some basic features of neural processing still remain unclear. Here, we focus on spikelets, which are brief, spike-like depolarizations of small amplitude (< 20 mV). Spikelets can be measured in somatic intracellular recordings in diverse neuron types, including cortical interneurons (e. g. , [1]) and pyramidal cells [2–4]. Due to their all-or-none appearance and spike-like shape, spikelets are considered to reflect action potentials (APs) occurring in electrotonically distinct compartments. These APs might originate either in the dendrites or in the axon of the same cell, or in another neuron that is either coupled ephaptically or through gap junctions. Since spikelets influence somatic voltage dynamics, including AP generation [2], identifying the origin of spikelets is important for understanding neural computations. The origin of spikelets in hippocampal [2,3, 5] and neocortical [4] pyramidal neurons is not well understood. The original hypothesis of spikelets resulting from dendritic spikes [6] could not be supported by subsequent studies [7]. Instead, axo-axonal [8,9] and somato-dendritic [10,11] gap-junction coupling of pyramidal neurons has been suggested as the spikelet origin, however, the supporting experimental evidence is scarce, raising the question whether there are other mechanisms for generating spikelets in pyramidal neurons. In vitro, somatic spikelets can be evoked with distal axonal stimulation if an antidromically propagating AP [12] does not suffice to activate the somatic sodium channels. This can happen because of somatic hyperpolarization, (prolonged) somatic depolarization, or fast repeated axonal stimulation [13–16]. However, in-vivo inputs are usually considered to arrive at the soma orthodromically. Indeed, spontaneous antidromic spikelets (also called “ectopic”) have been identified mainly under pathological conditions, such as epilepsy [17]. Additionally, antidromic spikelets are expected to occur when neurons would be coupled through axo-axonal gap junctions [8]. Here, we present a novel hypothesis for the origin of spikelets in pyramidal neurons. Using a computational approach, we demonstrate that spikelets can be evoked orthodromically with somato-dendritic inputs, which initiate APs at the distal axon initial segment (AIS). Under certain conditions, these APs in the AIS fail to fully activate the soma and appear there as spikelets. Consequently, the possibility of a forward propagating AP without it propagating back to the soma and into the dendrites presents a powerful mechanism for control of dendritic plasticity while ensuring the read-out of neural computations. To investigate mechanisms underlying spikelet occurrence, we first used a previously published multi-compartmental model of a reconstructed layer V pyramidal neuron ([16]; Fig 1A). This model includes a detailed sodium channel distribution at the AIS and a hyperpolarized voltage shift of 13 mV in the activation and inactivation functions of the low-threshold NaV1. 6 channels, present in the AIS and axon. To increase the incidence of spikelets, we modestly reduced the density of sodium channels (see Methods for details). The model cell was stimulated at the soma with stochastic excitatory and inhibitory synaptic point conductances [18] representing in vivo-like background activity. The resulting somatic voltage traces (Fig 1B, top) showed both APs and spikelets (stars). All APs were shoulder-APs (sh-APs; [2]) characterized by two components in the rising phase. The first component (the shoulder) was slower and resembled the waveform of spikelets (Fig 1C); the second, faster component included the peak of the AP. To reveal the origin of spikelets and sh-APs in our model, we compared voltage traces in the soma and the AIS (Fig 1B). The APs and spikelets recorded at the soma were initiated as full APs at the distal AIS (Fig 1D). Accordingly, both the shoulders of the sh-APs and the spikelets reflected axonal APs invading the soma [14,19]. Next, we aligned APs to the times of crossing a voltage threshold in the soma, and spikelets to the times of crossing the same voltage threshold in the AIS (Fig 1E, see also Methods). This alignment revealed a variable delay between the shoulder and the peak of the AP (Fig 1E, left) and demonstrated the all-or-none nature of the spikelet waveform (Fig 1E and 1F), as observed experimentally (Fig 1G; [2]). To understand why APs initiated at the AIS sometimes failed to elicit a somatic AP, we calculated both AP-triggered and spikelet-triggered averages of the synaptic input (Fig 1H). Excitation slowly increased ca. 5 ms before the onset of both APs and spikelets but dropped sharply prior to spikelet initiation; inhibition was stronger during spikelets compared to APs (Fig 1H2). Together, this input resulted in a weaker and briefer depolarizing synaptic drive for the initiation of spikelets compared to APs (Fig 1H3). We found that fast sodium channel inactivation, known to modulate spiking thresholds [20], was not a major factor influencing spikelet generation in our model (S1 Fig). Spikelets can thus be generated in a computational model of a single pyramidal neuron experiencing in vivo-like synaptic input: APs initiated at the AIS may fail to activate the soma and appear there as spikelets. Failure of AP propagation from the AIS to the soma (Fig 1) suggests that there is a strong voltage attenuation from axon to soma such that the somatic voltage does not reach the spiking threshold. To identify cell properties that could underlie such attenuation, we mathematically analyzed a passive-membrane model consisting of an axonal cable connected to a single somato-dendritic compartment (Fig 2A; see Methods for details). In particular, we computed the attenuation for sinusoidal input currents at several frequencies as a function of all model parameters (Fig 2B–2G; see Methods for equations). A central factor influencing signal attenuation is the electrotonic distance between the soma and the AIS. Attenuation thus increases with increasing physical distance (Fig 2B), increasing axial resistivity (Fig 2C), and decreasing axonal diameter (Fig 2D). Importantly, the attenuation is typically much larger in the antidromic (axon-to-soma) than in the orthodromic (soma-to-axon) direction because the large somato-dendritic compartment provides a substantially stronger current sink for the passively propagated signal than the thin axon, i. e. , there is a strong impedance mismatch between the two. Consistently, increasing the somato-dendritic surface area increased the attenuation of the antidromic signal whereas it did not affect the orthodromic propagation (Fig 2E). However, this did not reveal the nature of the current sink since the membrane resistance and the membrane capacitance are co-varied when changing the surface area. The specific membrane resistance, when varied separately in a range realistic for a pyramidal neuron (> 1 kΩ cm2), did not influence the antidromic attenuation for frequencies > 100 Hz (Fig 2F); in contrast, the antidromic attenuation of high-frequency (> 100 Hz) inputs was strongly influenced by the membrane capacitance (Fig 2G). For a fast, transient signal such as an AP, particularly the high-frequency components determine its shape. Indeed, in our model, the axon-to-soma attenuation of an AP waveform (black dashed lines in Fig 2B–2G) was very similar to the attenuation of a 300 Hz sine wave. Hence, apart from the electrotonic distance between soma and AIS, the capacitance of the somato-dendritic compartment strongly influences the attenuation of APs propagating from axon to soma. In general, the attenuation is asymmetric, i. e. , much larger in the axon-to-soma than in the soma-to-axon direction, which constitutes a favorable condition for spikelet generation. We next tested whether the asymmetric voltage attenuation is indeed a key component underlying the generation of spikelets through somato-dendritic input. For this, we turned to a model consisting of a dendrite, a soma, and an axon that all expressed active conductances (Fig 3A; see Methods for details). Similarly to the detailed compartmental model in Fig 1, the sodium channels at the distal AIS and in the axon were set to activate and inactivate at more hyperpolarized voltages than the sodium channels in the dendrite, the soma, and the proximal AIS [16,21]. However, the model in Fig 3 is much simpler than the complex model in Fig 1, which enabled us to explore its parameter space. To study the response of the model neuron with a simple stimulus, we applied rectangular current pulses (50 ms) to the soma for a range of input strengths. When an AP at the AIS was evoked, the corresponding somatic maximum response amplitude was recorded and plotted in a continuous color code (Fig 3B–3H). However, the somatic response amplitudes typically appeared in three well-separated clusters (examples in Fig 3B and S2 Fig B): (i) Spikelets (yellow) resulted from the weakest inputs that generated APs at the AIS but failed to evoke a somatic AP. (ii) The sh-APs (red) were evoked by larger somatic inputs and resulted from APs at the AIS that evoked a somatic AP. The shoulders of the sh-APs matched the spikelet waveform (see phase plots in Fig 3B, right). (iii) Finally, strong enough inputs could lead to full-blown APs (fb-APs; orange), which did not display a shoulder. The fb-APs resulted from AP initiation at the soma before or concurrent with AP initiation at the AIS. Consequently, fb-APs lacked the rapid onset (“kink”) typical for spikelets and sh-APs (Fig 3B, right) and the fb-AP amplitudes (from maximum curvature to maximum voltage) appeared smaller than the amplitudes of sh-APs because the maximum curvature occurred at higher voltages (Fig 3B, right). So similarly to the detailed model from Fig 1, input amplitude determined whether a spikelet or an AP was generated at the soma (see also S2 Fig): passive somatic depolarization from the input current added up to the somatic depolarization due to the AP propagated from the AIS, and if it reached the (fixed) somatic threshold, an AP was generated at the soma. Otherwise, a somatic spikelet appeared. To quantify how the somatic response type (spikelet, sh-AP, or fb-AP) depends on the somatic stimulus amplitude and the model parameters, we performed extensive numerical simulations of the active model with reduced morphology (Fig 3C–3H). These simulations indicated that the occurrence of spikelets required a certain degree of electrotonic separation between the soma and the AIS (Fig 3C and 3D) to allow for sufficient attenuation from axon to soma, as was suggested by the analytical results from the passive-membrane model (see Fig 2B–2D). Furthermore, spikelet generation needed a high enough somatic input capacitance (Fig 3E), in agreement with the analytical result that membrane capacitance was the primary current sink for APs propagating from AIS to soma (Fig 2F and 2G). Also as predicted, spikelet activity depended only weakly on the membrane resistance in a range that is plausible for pyramidal neurons (Fig 3F). Besides the passive membrane characteristics, also active properties of sodium channels were fundamental to the generation of somatic spikelets (Fig 3G and 3H). Lowering somato-dendritic sodium channel densities increased the somatic firing threshold and thereby promoted spikelet occurrence (Fig 3G). This result is in agreement with the reduced sodium channel densities boosting spikelet generation in the multi-compartment model in Fig 1. Another way to increase the firing-threshold difference between the soma and the AIS and thereby facilitate spikelet occurrence was to introduce a voltage shift in the activation function between the somato-dendritic and the axonal sodium channels (Fig 3H). The voltage shift had to be large enough such that an AP initiated at the AIS did not reach the voltage threshold in the soma. In summary, the simulation results of the active model with reduced morphology confirm that spikelets can be evoked through sufficiently small somatic input. In addition to strong and asymmetric voltage attenuation, the generation of spikelets requires a substantially lower AP threshold in the AIS compared to the soma. Spikelets of axonal origin can be evoked with distal axonal stimulation when the antidromically propagating AP does not suffice to cross the somatic spiking threshold. Such antidromic spikelets could also result from axo-axonic coupling by gap junctions [8]. Since the antidromic spikelets have different functional consequences than the orthodromic spikelets shown in Figs 1 and 3, it is important to be able to distinguish the two phenomena. To compare the properties of orthodromic and antidromic spikelets, the detailed model neuron with fluctuating somatic inputs from Fig 1 was additionally stimulated with brief current pulses to the distal axon (Fig 4A), which evoked axonal APs propagating antidromically towards the soma. The resulting spikelets were classified as antidromic (evoked with the distal axonal stimulus) and orthodromic (evoked with the somatic stimulus). Classification was based on the relative timing of the AP occurring at the distal AIS and in the axon (Fig 4B; see Methods). The two spikelet types were similar in shape and amplitude (Fig 4B and 4C), but the averaged antidromic spikelet displayed a more hyperpolarized somatic threshold and started abruptly from the baseline without a preceding depolarization (Fig 4C1), which is also typical for experimentally recorded antidromic APs [15]. For the antidromic spikelets in our computational model, the somatic excitatory and inhibitory conductances as well as the effective synaptic reversal potential did not show any modulation, which is in line with its distal axonal origin and its independence from somatic activity (Fig 4C2 and 4C3). Although the physiological occurrence of antidromic spikelets is disputed [22], we hypothesized that spikelets with similar properties can occur in pyramidal cells when the axon is attached to a dendrite instead of the soma [23]. To simulate this scenario, we adapted the morphology of the detailed model cell used in Figs 1 and 4 (Fig 5; see Methods), and excitatory postsynaptic conductances (EPSGs) were delivered to the axon-carrying dendrite, additionally to the somatic fluctuating inputs (Fig 5A). The resulting spikelets (Fig 5B) were classified according to the relative timing of the spikelet and the EPSG (see Methods). Both types of spikelets had comparable shapes and phase plots (Fig 5B). Spikelets evoked with stimuli to the axon-carrying dendrite exhibited a hyperpolarized average onset; nevertheless, some depolarization preceding these spikelets was visible in the somatic traces because the underlying input was located close enough to the soma (≈ 25 μm). However, spikelets evoked with stimuli to the axon-carrying dendrite were basically independent of somatic synaptic conductances (Fig 5C), and these spikelets are therefore reminiscent of the antidromic spikelets described in Fig 4. Alternatively, when the model presented in Fig 1 was additionally stimulated with brief current pulses at the proximal apical dendrite, the thresholds and waveforms of spikelets resulting from the dendritic stimulus were virtually identical to spikelets triggered by the fluctuating background stimulus applied to the soma (Fig 6). The average background conductances (Fig 6C2) and the effective synaptic drive (Fig 6C3) were less modulated for the dendritically evoked spikelets than for the spikelets evoked with the background stimulus. The number of dendritically evoked spikelets was substantially smaller than for inputs located at the distal axon or at the axon-attached dendrite because of an interplay between the dendritic and somatic stimulus in spikelet generation: The dendritic stimulus added to the background somatic input and triggered spikelets if the soma had the right level of depolarization. If the soma was too depolarized at the time point when the dendritic stimulus arrives, somatic APs were evoked; if the soma was too hyperpolarized, the compound input did not suffice to trigger an AP at the AIS. To summarize our results, spikelets can be generated within a single pyramidal neuron in three ways (Fig 7A, Sp1–Sp3). Each type of spikelet has characteristic features, which may allow to infer the origin of spikelets in experimental somatic voltage traces. Two key distinguishing features of spikelets are the somatic voltage threshold (Fig 7B) and the slope of the voltage a few milliseconds before the threshold is reached (Fig 7C). As a reference we consider the orthodromic APs, which exhibit the highest somatic firing threshold and are preceded by the steepest depolarization compared to the three types of spikelets: Orthodromic spikelets (Sp1) show a slightly smaller threshold and are preceded by a less steep depolarization, consistent with the finding that they required weaker inputs than APs. Antidromic spikelets (Sp2), which were evoked in our simulations with distal axonal stimulation, are characterized by the lowest thresholds and the highest somatic threshold variability. They arise abruptly at the soma: the averaged voltage trace shows no preceding depolarization. Finally, spikelets evoked by inputs to the axon-carrying dendrite (Sp3) lie somewhere in between the orthodromic and antidromic spikelets, regarding the average somatic threshold and the preceding depolarization; their orthodromic-like versus antidromic-like appearance depends on the electrotonic separation of the soma and the axon-carrying dendrite. Action potentials are the basis of neural function, yet some of their fundamental features are still not well understood, as highlighted by the recent focus on the rapidness of the AP onset [19,24,25]. It is generally assumed that an AP initiated in the AIS of a pyramidal neuron always leads to an AP in the soma. We argue here that this view needs to be corrected. Under certain conditions, APs initiated in the AIS by somato-dendritic inputs fail to fully activate the soma and appear there as spikelets. In simulations we showed that spikelets can result from APs that were evoked at the AIS with somato-dendritic inputs and propagated down the axon, but that did not trigger a somato-dendritic AP. This AP failure occurred for a sufficiently large difference in spiking thresholds between the soma and the AIS, together with a strong impedance mismatch (causing asymmetric voltage attenuation) and some degree of electrotonic separation between the soma and the AIS. In this way, a weak depolarizing input could pass through the soma and initiate an AP at the AIS, which, in turn, was not able to depolarize the soma to the firing threshold. Thus, a spikelet appeared at the soma instead of an AP. This mechanism reproduced several key features of spikelets reported in the experimental literature [2,4, 5]: the fast dynamics and rapid onset of spikelets as well as the match between the spikelet waveform and the shoulder of a sh-AP. This single-cell mechanism is also in line with the observation that APs and spikelets recorded in a single hippocampal place cell exhibit virtually identical place fields [2]. In contrast, in the electrotonic-coupling (gap junction) scenario of pairs of pyramidal cells [8–10], the place fields of spikelets and APs measured in a single cell are expected to differ due to lack of topography in hippocampus [26]. We found that the fast dynamics and amplitudes of spikelets observed in pyramidal neurons can be compatible with gap junction coupling only if the somato-dendritic gap junctions are very strong and located at proximal sites (S3 Fig). In previous experimental studies, spikelets could be evoked with dendritic stimulation or dendritic EPSPs [4,6], which led the authors to conclude that somatic spikelets arise from dendritic spikes. However, our modelling results suggest that although spikelets can be evoked with somato-dendritic inputs, they rather originate in the axon. Depending on the state of the proximal axonal sodium channels, the AP is initiated either in the AIS, as we considered in this study, or further down the axon. Consistently, a recent experimental study demonstrated an axonal origin of spikelets occurring during dendritic plateau-driven complex spiking in CA1 pyramidal neurons [27]. Also in other central neurons, spikelets occurring during somatic bursts can originate in the axon, for example, in inferior olivary neurons [28] and in cerebellar Purkinje neurons [29]. Antidromic spikelets also result from axonal APs, but these are evoked by distal axonal inputs [30] or by APs propagating through putative axo-axonal gap junctions [8]. Compared to the orthodromic spikelets, antidromic spikelets are characterized by hyperpolarized thresholds and they arise abruptly without a preceding depolarization (Fig 4C1). However, the best experimental distinguishing criterion is the fact that, because of their distal origin, they survive moderate levels of somatic hyperpolarization, as has been demonstrated, for example, in layer V pyramidal neurons in vitro [16]. Orthodromic spikelets do not occur when the somatically injected hyperpolarizing current is larger than the synaptic driving current measured at the soma, since the synaptic depolarizing input has to pass through the soma to trigger an AP at the AIS. In contrast, antidromic spikelets can be evoked even when the synaptic driving current is somewhat smaller than the somatically injected hyperpolarizing current. Spikelets evoked by inputs to the axon-carrying dendrite (Fig 5) would also be abolished by a certain level of somatic hyperpolarization, because of the relatively small electrotonic distance between the soma and the axon origin [23]. Consistent with an orthodromic origin of spikelets is the experimental observation that spikelets are suppressed by hyperpolarizing somatic current injections, leading to the conclusion that spikelets “are not generated far from the soma” [4]. Our proposed spikelet hypothesis relies on AP initiation at the AIS. Indeed, APs in hippocampal [31] and neocortical pyramidal neurons [16,32] are typically initiated in the distal portion of the AIS, about 20 − 40 μm away from the axon hillock. This site is preferred for AP initiation because of its decreased capacitive load from the soma [33] and increased sodium channel density, especially of the NaV1. 6 channel subtype [34], which activates at more hyperpolarized membrane potentials than the somatic sodium channel subtype NaV1. 2 [21]. However, it is still disputed whether the axonal sodium channel density is substantially higher (up to 50-times higher, [35]) than the somatic sodium channel density or whether the axonal and somatic sodium channels have similar densities [21,36]. The model neuron used in Figs 1 and 4–6 is characterized by a high ratio between the axonal and somatic sodium channel densities (up to a factor 40, [16]), which contributes to the large threshold difference between the axon and the soma, thus favoring spikelet generation. The question then arises how spikelet generation is affected when the sodium channel density ratio is smaller. The model used in Fig 3 employed a much smaller density ratio of 5 between the soma and the distal AIS (0. 02 and 0. 1 S/cm2, respectively). Fig 3G illustrates that spikelets occurred when the somatic sodium channel density was less than half the value at the distal AIS (i. e. , < 0. 05 S/cm2). In vivo, a fraction of somatic sodium channels is inactivated due to ongoing activity, which decreases the effective sodium channel density and promotes spikelet occurrence. However, the range of density ratios that support spikelet generation is not absolute, but depends on other parameters influencing somatic voltage threshold, like the voltage shift between the activation of somatic and axonal sodium channels (Fig 3H). In the present study, we used the standard sodium channel models that were fitted to neocortical (Figs 1 and 4–6, [16]) and hippocampal (Fig 3, [37]) pyramidal neurons. However, the dynamics of these model channels is slow compared to what has been found in more recent experiments [38,39]. Interestingly, simulations by Fleidervish et al. demonstrated that the faster, more realistic, sodium channel activation generated larger axo-somatic delays and larger voltage gradients than the classic, slower, sodium channel models [36]. As this axo-somatic gradient is vital for spikelet generation, we expect faster Na-channel gating to support spikelet generation. Experimental recordings featuring spikelets typically contain two types of APs: shoulder-APs with an initial slower phase corresponding to the spikelet, and full-blown APs, characterized by a single rising phase without a shoulder [2]. The shoulder of sh-APs is considered to result from the AP evoked at the AIS (e. g. , [19]). Then, the question about the origin of fb-APs arises. In our detailed compartmental model (Fig 1), all APs are evoked at the AIS and exhibit a shoulder. In the simple model shown in Fig 3, fb-APs can be generated with strong stimuli and for large electrotonic distances between the soma and the AIS, which allows somatic AP initiation to precede or co-occur with AP initiation at the AIS. However, unlike experimentally recorded fb-APs, they arise smoothly from the subthreshold depolarization and do not exhibit a rapid onset that is present in simulated and experimentally recorded spikelets and sh-APs. According to the “compartmentalization hypothesis of AP initiation” [25], the AP onset rapidness is caused by axonal AP initiation. This suggests that experimentally recorded fb-APs with rapid onset are not generated at the soma. Consistently, somatic AP initiation due to serotonin inhibition of AIS channels can result in gradually rising APs without a rapid onset [40]. Therefore, we hypothesize that fb-APs are either generated at the AIS and the shoulder is “masked” by fast somato-dendritic activation or they are initiated in the apical dendrites and no shoulder is visible because of the smooth morphologic transition between the primary apical dendrite and the soma. An intriguing issue concerns the rare observation of spikelets in vitro. Our analyses suggest that pyramidal neurons are positioned at the edge of a regime that allows spikelet generation. In the complex model from [16] used in Figs 1 and 4–6 for example, a modest decrease in sodium channel density strongly increased spikelet occurrence. One reason for such a decrease in functional sodium channel availability might be slow sodium channel inactivation [41]. In vitro, there is less slow sodium channel inactivation: a larger fraction of sodium channels might be available for spiking due to a lower average membrane potential and a lower firing activity, which keeps the fraction of inactivated sodium channels low. Additionally, sodium channel availability is regulated by various neuromodulators, acting via activity-dependent phosphorylation [42]. This might be especially relevant in vivo, where a variety of homeostatic mechanisms are expected to maintain spiking activity in neural circuits [43]. In our models, fast sodium channel inactivation was not a main factor influencing spikelet generation (S1 and S2 Figs). It cannot be ruled out, however, that fast sodium inactivation does play a significant role in real neurons under certain in vivo conditions. Another important factor for spikelet generation is the somato-dendritic current sink, which is reduced in brain slices because of “dendritic pruning”, i. e. , dendritic processes cut by the slicing procedure [44]. The typical thickness of slices is a few hundred microns (e. g. , 300 μm, [8,16]), which roughly matches the spatial extent of a pyramidal neuron’s dendritic tree (e. g. , [45]). For patch-clamp recordings, cells close to the slice surface are preferentially used, which is where one expects significant damage to proximal dendrites [44]. A pyramidal cell’s input capacitance is in the range of hundreds of picofarads [46], and considerable changes of this value are predicted to strongly affect spikelet occurrence (Fig 3E). In contrast, an artificial capacitance increase of about 4—10 pF by an uncompensated patch electrode [47] is small compared to a pyramidal cell’s input capacitance and, thus, should not influence spikelet incidence significantly. The presented hypothesis predicts that all-or-none somatic spikelets in pyramidal neurons are associated with APs at the AIS or further down in the axon [27]. This mechanism could be tested experimentally with simultaneous recordings of the somatic and axonal membrane voltages, which, however, might be difficult in vivo. An alternative would be to establish a reliable spikelet model in vitro. We propose to recreate in vitro a state of a pyramidal cell that retains the in vivo properties of sodium channels, for example by prolonged stimulation with fluctuating inputs and/or application of relevant neurotransmitters and neuromodulators naturally present in the cerebrospinal fluid in vivo [48]. Additionally, it might be necessary to record from neurons located in the middle of a slice, to minimize the dendritic loss and the resulting decrease in the somato-dendritic current sink. Interestingly, unlike in mammalian cells, spikelets are easily evoked in turtle pyramidal neurons in vitro with weak somatic or dendritic stimuli [49,50]. The amplitudes and waveforms of these spikelets closely resemble those in mammalian pyramidal neurons. Dual somatic and axonal recordings suggested an axonal origin of these spikelets [50]. We hypothesize that there might be two important differences between turtle and mammalian neurons that support in vitro spikelet firing in turtles. First, the slower and wider APs in turtles suggest that the effective (peri-) somatic sodium channel densities might be smaller in turtle than in mammalian pyramidal neurons. Second, the somata of turtle neurons are substantially larger than the somata of mammalian neurons, and most of the dendrites are single branches extending from the soma [50]. This might result in an increased capacitive somato-dendritic current sink and augment the impedance mismatch between the axon and the soma. The spikelets we described here are APs that propagate forward down the axon but not backward into the soma and the dendrites. What could be a functional role of such “output-only APs”? From an energetic point of view, spikelet firing saves energy since it avoids activation of sodium currents in the soma and the dendritic tree. Output-only APs thus minimize their contribution to activity-dependent metabolism [51,52]. Moreover, spikelets might be a means of reading out the result of neuronal computations without triggering dendritic plasticity through backpropagating APs [53]. Hence, spikelets potentially represent a mode of operation that is functionally highly relevant. To further unravel the role spikelets may play in neural computations, more theoretical and experimental studies are needed. Developing a CA1 pyramidal neuron model with a realistic AIS composition incorporating state-of-the-art sodium channel models is vital for a quantitative study of spikelet generation and properties, as the prevailing experimental work on spikelets has been carried out in these neurons. In order to construct such a model, further experimental studies of AIS composition and function in CA1 pyramidal neurons are necessary. Future studies could also address the putative role of axo-axonic synapses in spikelet generation, which provide powerful inhibition at the proximal AIS that can prevent antidromically evoked APs from invading the soma [12]. It would be important to see whether these synapses can control the propagation of orthodromically initiated APs and give rise to somatic spikelets, given the small distances between the soma and the distal AIS and the requirement for precise timing of inhibition: Too early inhibition would shunt the subthreshold depolarization and prevent AP initiation in the first place, whereas too late inhibition would be ineffective to stop the propagating AP (see also [54]). Also the influence of sodium channel neuromodulation on spikelet occurrence [42] and generation of full-blown APs in cells exhibiting spikelets are important topics for our understanding of spikelets in pyramidal neurons. This knowledge should allow to assess the computational consequences of spikelet firing at the single-cell and network level. For the results in Figs 1,4, 5 and 6 we used a previously published detailed model of a reconstructed layer V pyramidal neuron [16, ModelDB accession number 123897], implemented in NEURON [55]. Compared to the original model, we made two modifications. First, a small geometrical discontinuity at the AIS was corrected. In the original model, the AIS tapers from 1. 7 μm to 1. 22 μm. However, the diameter at the end of the axon hillock, i. e. , at the hillock-AIS boundary, is 1. 3 μm. We removed this sudden jump in the diameter so that the diameters at the end of the axon hillock and at the beginning of the AIS are equal at a value of 1. 3 μm (then tapering smoothly to 1. 22 μm, at the end of AIS). Second, the density of the NaV1. 2 subtype was decreased in soma, axon hillock, and AIS to 80%, and in dendrites to 60% of the original values. These changes only weakly influenced the AP properties and firing patterns (Table 1). The largest effects were observed for spikelet frequency and maximum AP slope. The decrease in maximum AP slope was desired, as it reflects the smaller AP slopes reported in vivo. Overall, the properties of APs generated in this model (Table 1) fit well into the range reported for pyramidal neurons in the experimental literature [2,5, 24,32,56]. The compartmental model cell was stimulated with two fluctuating synaptic point conductances placed at the soma [18] with the following parameters (values given in parentheses): reversal potential of the excitatory (Ee = 0 mV) and inhibitory (Ei = −75 mV) conductance, average excitatory (ge0 = 0. 01 μS) and inhibitory (gi0 = 0. 0573 μS) conductance, standard deviation of the excitatory (stde = 0. 014 μS) and inhibitory (stdi = 0. 02 μS) conductance and time constant of the excitatory (τe = 2. 728 ms) and inhibitory (τi = 10. 49 ms) conductance. As a result, the somatic membrane voltage fluctuated with a standard deviation of 8. 09 mV, producing a somatic AP firing rate of 5. 79 s−1 and a spikelet firing rate of 0. 63 s−1 (Fig 1). The somatic APs and spikelets were detected using a voltage-threshold criterion at the AIS and at the soma (both − 10mV). For both types of events, the threshold at the AIS had to be crossed. If the threshold at the soma was crossed within a time window from 1 ms before to 5 ms after the AIS threshold crossing, such an event was classified as an AP. Otherwise, the event was a spikelet. We also used a double-threshold criterion for the somatic voltage derivative (dV/dt) to confirm that no event was missed by the above voltage-threshold criterion and that indeed all somatic APs and spikelets were associated with an AP at the AIS: events that crossed the first threshold (20 V/s), but not the second threshold (100 V/s) were classified as spikelets, whereas somatic APs had to cross both thresholds within 2 ms. In Fig 1E, the APs were aligned in time to the point of crossing a somatic voltage threshold of -10 mV, whereas spikelets were aligned to the point of crossing a voltage threshold of -10 mV at the AIS. In Fig 1H, all events were aligned to the point of crossing the voltage threshold at the AIS to allow for a comparison of inputs between APs and spikelets. In Fig 1H, the effective synaptic reversal potential was calculated as (ge (t) Ee + gi (t) Ei) / (ge (t) + gi (t) ), i. e. , the excitatory and inhibitory reversal potentials weighted with the respective conductances. In Fig 4, in addition to the somatic conductance inputs as in Fig 1, the model cell was also stimulated with brief current pulses (0. 5 nA for 2 ms) delivered every 500 ms at the most distal axonal compartment. Somatic spikelets were classified as orthodromic (i. e. , evoked with somatic inputs) or antidromic (i. e. , evoked with distal axonal inputs) based on the relative timing of the AP at the distal AIS and in the axon. For orthodromic spikelets, the AP at the distal AIS preceded the AP in the axon; for antidromic spikelets, the AP at the distal AIS followed the AP in the axon. In Fig 5, the morphology of the model cell was altered: the axon hillock was omitted and the AIS was attached to a basal dendrite (“dendrite3[2] (0. 5) ”) 20. 5 μm away from the soma. In addition to the somatic conductance inputs as in Fig 1, an EPSG (τrise = 0. 5 ms, τdecay = 2 ms, peak conductance = 0. 02 μS, Esyn = 0 mV) was delivered every 500 ms to the axon-carrying dendrite, distally to the AIS-connecting site (“dendrite3[3] (0. 1) ”). Spikelets evoked with dendritic EPSGs were distinguished from the orthodromic spikelets (evoked with somatic inputs) as spikelets occurring within a 2 ms window after the dendritic EPSG. In Fig 6, in addition to the somatic conductance inputs as in Fig 1, the model cell was also stimulated with a brief current pulse (2 nA for 1 ms) delivered every 20 ms at the proximal apical dendrite (“dendrite11[2] (0) ”) 47 μm away from soma. In 200 s of simulation, 2,106 somatic APs and 91 somatic spikelets were generated. We classified the spikelets as evoked with the dendritic input if the somatic spikelet was evoked within 2 ms from dendritic stimulus onset (N = 43); if the spikelet occurred 10 ms or later after the onset of the dendritic stimulus, the spikelet was classified as triggered by the somatic background stimulus (N = 41). In S3 Fig, we simulated two identical cells (as in Fig 1) coupled by a gap junction. The gap junction was modelled as an ohmic resistor, allowing to transmit voltage changes between the coupled cells [57]. In cell 1, an AP was evoked with a somatic current step (2 nA applied for 15 ms), and a spikelet was recorded in cell 2. The strength of the gap junction was varied between 22 and 82 MΩ in 5 MΩ steps (corresponding to gap junctional conductance of 12–45 nS). The gap junction was placed at the soma or at several positions along the main apical dendrite (at a distance of ≈ 8,24,47,78, or 109 μm from soma). The leak reversal and initial membrane voltages were set to -80 mV instead of the original leak reversal of -70 mV because otherwise the closest and strongest gap junctions could only generate an AP and not a spikelet in cell 2. The amplitude of spikelets was measured from the maximum of the 2nd derivative (the “kink”) to the maximum amplitude. We mathematically analyzed a model consisting of a semi-infinite cable with an RC-circuit as a boundary condition, representing the axon and the entire somato-dendritic compartment, respectively (Fig 2). The system is mathematically equivalent to the lumped-soma model introduced by Rall [58]. Our model describes the dynamics of the voltage V along the axon at distance x from the soma in response to current input at location x = y using the linear cable equation: λ 2 δ 2 δ x 2 V (x, t) - τ δ δ t V (x, t) - V (x, t) = g (x, t) for x > 0 (1) where τ is the membrane time constant (in ms), λ is the axonal length constant (in cm), and g (x, t) is the input to the model. The boundary condition to include the somato-dendritic compartment at x = 0 is τ δ δ t V (0, t) = λ ρ δ δ x V (0, t) - V (0, t) (2) where the dimensionless parameter ρ denotes the ratio of the total somato-dendritic membrane resistance to the input resistance of the axon. The semi-infinite cable boundary condition is lim x → ∞ V (x, t) = 0. (3) For notational convenience we consider the resting potential in this linear system to be 0 mV. The parameters τ, λ, and ρ are determined by physiological parameters. Setting the specific membrane resistance Rm = 104 Ω cm2, specific membrane capacitance Cm = 1 μF/cm2, axial resistivity Ra = 150 Ω cm, surface area of the somato-dendritic compartment Asd = 2 ⋅ 10−4 cm2 and diameter of the axon da = 10−4 cm yields τ = RmCm = 10 ms, λ = R m d a 4 R a = 0. 041 cm and ρ = π d a 3 / 2 2 A s d = 0. 064. The purpose of the mathematical model was to compute the frequency-dependent attenuation of voltage signals between the axon and the somato-dendritic compartment. One approach is to use a complex-valued input current in the original partial differential equation and solve for the voltage responses of the axon and the somato-dendritic compartment. Here, we will instead proceed using a real-valued input current and use the Fourier transforms of the above partial differential equation and boundary conditions: λ 2 δ 2 δ x 2 V ^ (x, ω) - b (ω) 2 V ^ (x, ω) = g ^ (x, ω) for x > 0 (4) with the boundary conditions δ δ x V ^ (0, ω) - b (ω) 2 λ ρ V ^ (0, ω) = 0 (5) and lim x → ∞ V ^ (x, ω) = 0, (6) where V ^ (x, ω) and g ^ (x, ω) are the Fourier transforms of V (x, t) and g (x, t), respectively, ω = 2πf with frequency f (in Hertz), and b (ω) 2 = 1 + iωτ. We next calculated the voltage response of the model to the real-valued sinusoidal input current at location x = y: g (x, t) = R m π d a I 0 cos (ω 0 t) δ (x - y) (7) with radial frequency ω = ω0 ≥ 0 and amplitude I0. The Fourier transform of the input term is g ^ (x, ω) = R m π d a I 0 δ (ω - ω 0) δ (x - y), (8) where we neglected the negative-frequency terms. We then solved the above second-order, nonhomogeneous ODE by first considering solutions of the form V ^ h (x, ω) = c 1 exp (− b (ω) x / λ) + c 2 exp (b (ω) x / λ) for the homogeneous version of the ODE and use this to find a particular solution V ^ n h (x, ω) for the nonhomogeneous ODE; subsequently the constants c1 and c2 were determined by considering the boundary conditions [59, section 6. 2]. The sinusoidal voltage response at location 0 ≤ x ≤ y is V ^ (x, ω 0) = I 0 R ∞ b 0 ρ cosh (b 0 x / λ) + b 0 sinh (b 0 x / λ) (b 0 + ρ) exp (b 0 y / λ), (9) and for x ≥ y it is V ^ (x, ω 0) = I 0 R ∞ b 0 ρ cosh (b 0 x / λ) + b 0 sinh (b 0 x / λ) (b 0 + ρ) exp (b 0 y / λ) - sinh (b 0 (x - y) / λ) (10) where b0 = b (ω0) is the principal square root (i. e. , with positive real part) of 1 + i ω 0 τ and R ∞ = 2 π d a - 3 / 2 R m R a is the input resistance of a semi-infinite cable. The steady-state voltage attenuation from axon to soma is then given by the ratio of the voltage response amplitude at the axonal injection site to the somatic voltage response amplitude: A a x o n → s o m a (y, ω 0) = V ^ (y, ω 0) V ^ (0, ω 0) = cosh (b 0 y / λ) + b 0 ρ sinh (b 0 y / λ), (11) where |z| denotes the absolute value of the complex number z. Similarly, the frequency-dependent voltage attenuation from soma to axon for a somatic input (i. e. , y = 0 and x ≥ y) can be computed, which is equal to the attenuation in an (semi-) infinite cable: Asoma→axon (x, ω0) =| V^ (0, ω0) V^ (x, ω0) |=| exp (b0x/λ) |. (12) In Fig 2B–2G, the natural logarithm of the attenuation was plotted. The axonal stimulation/recording site was y = 50 μm away from the soma (except in Fig 2B where it was varied). The passive-membrane model was also simulated numerically with the NEURON module embedded in Python [60] to compare the antidromic (axon-to-soma) attenuation of pure sine waves with the attenuation of an AP waveform. Here, identical parameters were used as in the analytical calculations (see above). The axon length was set to 2 mm, corresponding to an electrotonic length of 4. 9 λ. The AP waveform was delivered via a voltage clamp at a 1 μm long axonal compartment located 50 μm away from the soma. We used an AP waveform recorded at the AIS of the detailed model (Fig 1D, middle). The input capacitance in Fig 2G was calculated from a small, prolonged voltage-clamp step by dividing the integrated transient charge by the voltage-clamp step size [61]. Results presented in Fig 3 used an active compartmental model of a simplified neuron morphology. The model consisted of a dendritic cable (length × diameter: 900 μm × 6 μm), an axonal cable (1,060 μm × 1 μm), and a cylindrical somatic compartment (40 μm × 20 μm). The axonal cable included a proximal AIS (30 μm), a distal AIS (30 μm), and the axon (1,000 μm). The passive model properties were uniform along the model neuron: specific membrane capacitance 1 μF/cm2, specific membrane resistance 10 kΩ cm2, and axial resistivity 150 Ω cm. The resting membrane potential equaled the leak reversal potential, which was set to -70 mV. The active model properties included transient sodium and delayed rectifier potassium conductances. Channel models were taken from [37, ModelDB accession number 2796], with parameter values corresponding to hippocampal pyramidal neurons. Active currents were present in all compartments (densities given in parentheses): Na-channel conductance in the soma and the dendrite (0. 02 S/cm2), in the proximal AIS and the axon (0. 04 S/cm2), and in the distal AIS (0. 1 S/cm2); K-channel conductance in the soma and the dendrite (0. 05 S/cm2), in the proximal and distal AIS (0. 25 S/cm2), and in the axon (0. 125 S/cm2). Additionally, the activation and inactivation curves of the Na-channels in the distal AIS and in the axon were shifted by 10 mV in hyperpolarizing direction compared to the activation and inactivation curves of Na-channels in the dendrite, the soma, and the proximal AIS. To elicit spiking activity in the model, rectangular current stimuli of 50 ms duration were applied at the soma. The resulting somatic event amplitude was measured from the voltage at the maximum of its second derivative (i. e. , maximum curvature) to the peak voltage. However, if there was no AP occurring at the AIS (detected as not crossing a voltage threshold of -20 mV), the somatic amplitude was not plotted (white regions in the heat maps). The input capacitance (Fig 3E) was calculated in the same way as in the passive-membrane model (see above). Voltage traces shown in S2 Fig were generated in a model with default parameters, except the length of the proximal AIS, which was set to 100 μm instead of the default 30 μm, so that all event types (spikelet, sh-AP, fb-AP) could be produced. In S2C Fig, the dynamics of sodium channel inactivation was “frozen” to the steady-state value at -70 mV by setting the time constant of inactivation to a very large value (105 ms). Numerical simulations were performed using the NEURON simulation environment [55], with the NEURON module embedded in Python [60].

Action potentials (APs) are digital, all-or-none signals by which neurons communicate with each other. Therefore, APs are the basis of neural function, yet some of their fundamental features are still not well understood. Here we focus on pyramidal cells, which are the principal neurons in neocortex and hippocampus. According to textbook knowledge, an AP in pyramidal neurons is initiated at the axon initial segment and propagates along the axon to the next cell. Concurrently, the AP also propagates back to the soma and into the dendrites where it might trigger synaptic plasticity, which is the basis of learning and memory. However, besides APs, pyramidal cells sometimes also show somatic spikelets—small depolarizations with an AP-like shape—whose origin remains unclear. Here, we propose that spikelets occur when an AP initiated at the axon initial segment only propagates down the axon, but fails to activate sodium currents in the soma and dendrites. As a result, spikelet firing saves energy, and moreover, might be a means to control synaptic plasticity and thereby control learning and memory.

Abstract Introduction Results Discussion Methods

cell physiology medicine and health sciences action potentials nervous system depolarization membrane potential junctional complexes electrophysiology neuroscience gap junctions ion channels nerve fibers neuronal dendrites sodium channels animal cells axons proteins biophysics physics biochemistry cellular neuroscience cell biology anatomy synapses physiology neurons biology and life sciences cellular types physical sciences neurophysiology

2017

Spikelets in Pyramidal Neurons: Action Potentials Initiated in the Axon Initial Segment That Do Not Activate the Soma

The bloodstream forms of Trypanosoma brucei, the causative agent of sleeping sickness, rely solely on glycolysis for ATP production. It is generally accepted that pyruvate is the major end-product excreted from glucose metabolism by the proliferative long-slender bloodstream forms of the parasite, with virtually no production of succinate and acetate, the main end-products excreted from glycolysis by all the other trypanosomatid adaptative forms, including the procyclic insect form of T. brucei. A comparative NMR analysis showed that the bloodstream long-slender and procyclic trypanosomes excreted equivalent amounts of acetate and succinate from glucose metabolism. Key enzymes of acetate production from glucose-derived pyruvate and threonine are expressed in the mitochondrion of the long-slender forms, which produces 1. 4-times more acetate from glucose than from threonine in the presence of an equal amount of both carbon sources. By using a combination of reverse genetics and NMR analyses, we showed that mitochondrial production of acetate is essential for the long-slender forms, since blocking of acetate biosynthesis from both carbon sources induces cell death. This was confirmed in the absence of threonine by the lethal phenotype of RNAi-mediated depletion of the pyruvate dehydrogenase, which is involved in glucose-derived acetate production. In addition, we showed that de novo fatty acid biosynthesis from acetate is essential for this parasite, as demonstrated by a lethal phenotype and metabolic analyses of RNAi-mediated depletion of acetyl-CoA synthetase, catalyzing the first cytosolic step of this pathway. Acetate produced in the mitochondrion from glucose and threonine is synthetically essential for the long-slender mammalian forms of T. brucei to feed the essential fatty acid biosynthesis through the “acetate shuttle” that was recently described in the procyclic insect form of the parasite. Consequently, key enzymatic steps of this pathway, particularly acetyl-CoA synthetase, constitute new attractive drug targets against trypanosomiasis. Trypanosoma brucei is a unicellular eukaryote, belonging to the protozoan order Kinetoplastida that causes sleeping sickness in humans and economically important livestock diseases [1]. This parasite undergoes a complex life cycle during transmission from the bloodstream of a mammalian host (bloodstream forms of the parasite - BSF) to the alimentary tract (procyclic form - PF) and salivary glands (epimastigote and metacyclic forms) of a blood feeding insect vector, the tsetse fly. In the bloodstream of the mammalian host, the pleomorphic BSF strains proliferate as “long-slender” BSF (LS-BSF) and differentiate into the non-proliferative “short-stumpy” trypanosomes (SS-BSF), which are preadapted for differentiation into PF in the insect midgut [2]. The environmental changes encountered by the parasite require significant morphological and metabolic adaptations, as exemplified by important qualitative and quantitative differences in glucose metabolism between BSF and PF [3], [4]. PF living in the tsetse fly midgut – where glucose is scarce or absent – have developed an elaborate energy metabolism based on amino acids, such as proline. However, when grown in standard glucose-rich conditions, they prefer glucose to proline as a carbon source [5], [6]. PF converts glucose into the partially oxidized and excreted end-products, acetate and succinate, with most of the glycolysis taking place in specialized peroxisomes called glycosomes [7]. In the course of glycolysis, phosphoenolpyruvate (PEP) is produced in the cytosol, where it is located at a branching point to feed the glycosomal ‘succinate branch’ and the mitochondrial ‘acetate and succinate branches’ (see Fig. 1). For the “succinate branches”, PEP must re-enter the glycosomes where it is converted into malate and succinate within that compartment. Malate, which moves from the glycosomes into the mitochondrion, can also be converted into succinate therein. Additionally, PEP can be converted in the cytosol into pyruvate to feed the ‘acetate branch’ (steps 1–4 in Fig. 1). In the mitochondrion, pyruvate is converted by the pyruvate dehydrogenase complex (PDH, EC 1. 2. 4. 1, step 1) into acetyl-CoA and then into acetate by two different enzymes, i. e. acetate∶succinate CoA transferase (ASCT, EC 2. 8. 3. 8, step 2) and acetyl-CoA thioesterase (ACH, EC 3. 1. 2. 1, step 3) [8]–[10]. In PF, acetate production plays an important role for mitochondrial ATP production by the ASCT/SCoAS cycle (steps 2 and 4), while ACH is not involved in ATP production [10]. Acetate can also be produced from threonine, a major carbon source of PF present in the in vitro medium [6], [11], [12]. This amino acid is converted into acetate by threonine-3-dehydrogenase (TDH, EC 1. 1. 1. 103, step 5), acetyl-CoA∶glycine C acetyltransferase (EC 2. 3. 1. 29, step 6) and probably ASCT and/or ACH. We recently showed that PF uses a new metabolic pathway only observed in PF trypanosomes so far, named the “acetate shuttle”, which transfers acetyl-CoA from the mitochondrion to the cytosol to feed the essential cytosolic fatty acid biosynthesis [13]. In this shuttle, acetate produced in the mitochondrion from acetyl-CoA is exported in the cytosol and converted back into acetyl-CoA by the cytosolic acetyl-CoA synthetase (AMP-dependent enzyme, AceCS, EC 6. 2. 1. 1, step 7). In contrast to PF, BSF trypanosomes rely only on glucose for their energy production, with a 5- to 10-fold higher rate of glucose consumption [14]. It is generally accepted that the proliferative LS-BSF grown under aerobiosis convert glucose exclusively into pyruvate [15], [16], although excretion of trace amounts of other incompletely oxidized end-products such as glycerol, succinate and alanine have been reported [14], [17]. These minor glycolytic end-products are thought to be produced by “contaminating” non-proliferative SS-BSF trypanosomes that have developed a more elaborated central metabolism with a number of PF traits, including production of acetate and succinate from glycolysis [18], [19]. Consequently, recent reports consider that LS-BSF trypanosomes do not produce acetate from glucose. In the seventies, acetate production was reported from the threonine degradation in both PF and BSF trypanosomes [12], however, this metabolic pathway was not further investigated. Here we investigated the role of glucose and threonine degradation in acetate production in the monomorphic 427 BSF strain, which proliferates as LS-BSF trypanosomes and has lost the ability to differentiate into non-proliferative SS-BSF [20], [21]. This BSF cell line produces and excretes acetate, as a minor end-product of glucose metabolism, with a metabolic flux in the same range as observed for PF. Glucose and threonine contribute almost equally to acetate production, which is essential for the viability of proliferative BSF trypanosomes, as demonstrated by reverse genetics approaches. Our data reveal unexpected metabolic similarities between PF and LS-BSF trypanosomes. The bloodstream form of T. brucei 427 90-13 (TetR-HYG T7RNAPOL-NEO), a 427 221a line (MiTat 1. 2) designed for the conditional expression of genes, was cultured at 37°C in IMDM (Iscove' s Modified Dulbecco' s Medium, Life Technologies) supplemented with 10% (v/v) heat-inactivated fetal calf serum (FCS), 0. 25 mM ß-mercaptoethanol, 36 mM NaHCO3,1 mM hypoxanthine, 0. 16 mM thymidine, 1 mM sodium pyruvate, 0. 05 mM bathocuprone and 2 mM L-cysteine [22]. To prepare threonine-depleted IMDM medium all the compounds constituting the medium, except threonine, were purchased from Sigma-Aldrich. The procyclic form of T. brucei EATRO1125 was cultured at 27°C in SDM79 medium containing 10% (v/v) heat-inactivated fetal calf serum and 35 µg/mL hemin [23]. Replacement of the threonine-3-dehydrogenase (TDH: Tb927. 6. 2790) by the puromycin (PAC) and blasticidin (BSD) resistance markers via homologous recombination was performed with DNA fragments containing a resistance marker gene flanked by the TDH UTR sequences. The TDH knock out was generated in the 427 90-13 BSF parental cell line, which constitutively expresses the T7 RNA polymerase gene and the tetracycline repressor under the control of a T7 RNA polymerase promoter for tetracycline inducible expression (TetR-HYG T7RNAPOL-NEO) [24]. Transfection and selection of drug-resistant clones were performed as previously reported using the Nucleofactor system [25]. The first and second TDH alleles were replaced by puromycin- and blasticidin-resistant genes, respectively. Transfected cells were selected in IMDM medium containing hygromycin B (5 µg/mL), neomycin (2. 5 µg/mL), puromycin (0. 1 µg/mL) and blasticidin (10 µg/mL). The selected cell line (TetR-HYG T7RNAPOL-NEO Δtdh: : PAC/Δtdh: : BSD) is called Δtdh. Accession numbers (http: //www. genedb. org/genedb/tryp/) of genes targeted by RNAi, acetyl-CoA synthetase (AMP-dependent enzyme, AceCS) and E2 subunit of the pyruvate dehydrogenase complex (PDH-E2), are Tb927. 8. 2520 and Tb927. 10. 7570, respectively. RNAi-mediated inhibition of gene expression in the 427 90-13 BSF parental cell line was performed by expression of stem-loop “sense/anti-sense” RNA molecules of the targeted sequences introduced into the pHD1336 (kindly provided by C. Clayton, ZMBH, Heidelberg, Germany, cclayton@zmbh. uni-heidelberg. de). The AceCS-SAS and PDH-E2-SAS “sense/anti-sense” constructs were first generated in the pLew100 vector (kindly provided by E. Wirtz and G. Cross) [24] as described before [5], [13]. Then the AceCS-SAS and PDH-E2-SAS HindIII-BamHI cassettes extracted from the pLew100 plasmids were inserted in HindIII-BamHI digested pHD1336 vector, which contains the blasticidin resistance gene. The RNAi-harboring RNAiAceCS and RNAiPDH single mutant cell lines were produced by transfection of the 427 90-13 cell line with the NotI-linearized pHD-AceCS-SAS and pHD-PDH-E2-SAS plasmids, respectively, and selected in IMDM medium containing hygromycin B (5 µg/mL), neomycin (2. 5 µg/mL) and blasticidin (10 µg/mL). For transfection of the Δtdh cell line with the pLew-PDH-E2-SAS construct, all of the four antibiotics used to select the Δtdh cell line, in addition to phleomycin (2. 5 µg/mL), were included in the medium to select double mutant cell lines. A recombinant fragment containing the full-length TDH gene was inserted into the NdeI and BamHI restriction sites of the pET28a expression vectors (Novagen) to express in BL21 Escherichia coli the TDH protein preceded by a N-terminal histidine tag (6 histidine codons). Cells were harvested by centrifugation, and recombinant proteins purified by nickel chelation chromatography (Novagen) from the insoluble fraction according to the manufacturer' s instructions. The anti-TDH immune serum was raised in rabbits by five injections at 15-day intervals of 100 µg of TDH-His recombinant nickel-purified proteins, emulsified with complete (first injection) or incomplete Freund' s adjuvant (Proteogenix S. A.). Antibodies raised against the T. brucei TDH protein expressed in E. coli recognize a single 36. 5 kDa protein in western blots, corresponding to the calculated TDH molecular weight (36. 96 kDa). Total protein extracts of bloodstream or procyclic forms of T. brucei (5×106 cells) were separated by SDS PAGE (10%) and immunoblotted on Immobilon-P filters (Millipore) [26]. Immunodetection was performed as described [26], [27] using as primary antibodies, the mouse anti-sera against AceCS diluted 1∶100 [13], PDH-E2 diluted 1∶500 [28] or the heat shock protein 60 (hsp60) diluted 1∶10,000 [29], or the rabbit anti-sera against TDH diluted 1∶500, acetate∶succinate CoA-transferase (ASCT) diluted 1∶100 [9], acetyl-CoA thioesterase (ACH) diluted 1∶500 [10] or glycerol-3-phosphate dehydrogenase (GPDH, EC 1. 1. 1. 8) diluted 1∶100 [30]. Goat anti-rabbit Ig/peroxidase (1∶10,000 dilution) or goat anti-mouse Ig/peroxidase were used as secondary antibody and revelation was performed using the SuperSignal West Pico Chemiluminescent Substrate as described by the manufacturer (Thermo Scientific). Images were acquired and analyzed with a KODAK Image Station 4,000 MM and quantitative analyses were performed with the KODAK MI application. Cells were washed in PBS and lysed by sonication (5 sec at 4°C) in hypotonic lysis buffer (5 mM Na2HPO4,0. 3 mM KH2PO4). Determination of TDH and PDH enzymatic activities was performed using a spectrophotometric assay as described before [12], [31], [32]. To stain mitochondria of the wild-type cell lines, 200 nM MitoTracker Red CMXRos (Invitrogen) were added to the culture, followed by a 20 min incubation and washes in PBS. Then wild-type cells were fixed with 4% formaldehyde in PBS, permeabilized with 1% Triton X-100, and spread on poly-L-lysine-coated slides. The slides were then incubated for 45 min in PBS containing 5% BSA, followed by incubation in PBS with 2% BSA and the primary antiserum, 1∶50 diluted rabbit anti-TDH, mouse anti-AceCS or mouse anti-PDH-E1α. After washing with PBS, the slides were incubated with 2 µg/mL Alexa 594 anti-rabbit IgG conjugate or Alexa Fluor 594 anti-mouse IgG conjugate (Molecular Probes). Slides were then washed and mounted in the SlowFade antifade reagent (Invitrogen). Cells were visualized with a Leica DM5500B microscope, and images were captured by an ORCA-R2 camera (Hamamatsu) and Leica MM AF Imaging System software (MetaMorph). The bloodstream forms (2. 5×107 cells, ∼0. 25 mg of proteins) or procyclic form (5×107 cells, ∼0. 25 mg of proteins) of T. brucei were collected by centrifugation at 1,400 g for 10 min, washed once/twice with phosphate-buffered saline (PBS) and incubated for 5 h at 37°C in 2. 5 mL of incubation buffer (PBS supplemented with 5 g/L NaHCO3, pH 7. 4), with [U-13C]-glucose (4 mM) in the presence or the absence of threonine (4 mM). The same experiments were performed with regular 12C glucose as the only carbon source. The integrity of the cells during the incubation was checked by microscopic observation. 50 µL of maleate (20 mM) were added as internal reference to a 500 µL aliquot of the collected supernatant and proton NMR (1H-NMR) spectra were performed at 125. 77 MHz on a Bruker DPX500 spectrometer equipped with a 5 mm broadband probe head. Measurements were recorded at 25°C with an ERETIC method. This method provides an electronically synthesized reference signal [33]. Acquisition conditions were as follows: 90° flip angle, 5,000 Hz spectral width, 32 K memory size, and 9. 3 sec total recycle time. Measurements were performed with 256 scans for a total time close to 40 min. Before each experiment, the phase of the ERETIC peak was precisely adjusted. Resonances of the obtained spectra were integrated and results were expressed relative to ERETIC peak integration. The linear production of pyruvate and acetate throughout the experiment was confirmed by 1H-NMR quantification of the end-products excreted by the wild type trypanosomes incubated for 6 h in PBS containing 4 mM [U-13C]-glucose (data not shown). Cells in the late exponential phase (5×107 cells) were incubated for 16 h in 10 mL of modified IMDM medium without threonine, pyruvate, leucine, isoleucine, valine, containing 25 mM glucose, 100 µM acetate and 40 µCi of [1-14C]-acetate (55. 3 mCi/mmol). Cells were checked microscopically for viability several times during incubation. Subsequently, lipids were extracted by chloroform∶methanol (2∶1, v/v) for 30 min at room temperature, and then washed three times with 0. 9% NaCl. The washed lipid extracts were then evaporated and lipids were dissolved in 1 mL of methanol∶H2SO4 (40∶1, v/v). Trans-esterification of the fatty acids of the lipids was performed at 80°C for 60 min. After cooling the samples, 400 µL of hexane (99% pure) and 1. 5 mL of H2O were added, and the mixture was homogenized vigorously for 20 sec. The samples were then centrifuged for 5 min at 1,000 g to separate the phases, and the hexane upper phases containing fatty acid methyl ester (FAMEs) were recovered without contact with the lower phases. FAMEs were loaded onto HPTLC plates developed in hexane/ethylether/acetic acid (90∶15∶2, v/v) and were separated (RF 0. 90). They were identified by co-migration with known standards. Their radio-labeling was then determined with a STORM 860 (GE Healthcare). The values were normalized with the amounts of total esters in each sample and detected by densitometry analysis using a TLC scanner 3 (CAMAG, Muttenz, Switzerland) as already described [13]. Eight- to ten-week-old female BALB/c mice bred at the SAS Centre d' Elevage Depré (Saint Doulchard, France) were housed under conventional conditions, with food and water administered ad libitum, according to institutional guidelines. Twelve mice per group were immunocompromised by intraperitoneal injection of 300 mg/kg Endoxan 48 h prior to infection and then infected with a single intraperitoneal injection of 104 parasites suspended in 0. 3 mL of fresh IMDM medium. Where appropriate, 1 mg/mL doxycycline and 50 g/L saccharose were added every 48 h to the drinking water starting three days prior to infection. Four experimental groups were studied: animals infected with wild-type parasites without (group 1) or with doxycycline (group 2) in the drinking water, animals infected with the c RNAiPDH. ni cell line (group 3) and animals infected with the RNAiPDH. i cell line cultured for 48 h in the presence of doxycycline to pre-induce down-regulation of PDH-E2 expression and then kept with doxycycline in the drinking water (group 4). To prevent the phenotypic reversion commonly observed in BSF mutants, the injected RNAiPDH. i cell line was selected from a fresh transfection and maintained in vitro up to 4 weeks post-transfection before injecting the animals. Efficient down-regulation of PDH-E2 expression was confirmed by western blot and the threonine-dependency of the selected cell line was confirmed in vitro. The health status of the animals was monitored on a daily basis and parasitaemias were counted daily. Experiments, maintenance and care of mice complied with guidelines of the European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes (CETS n°123). Experiments were approved by the Department for the protection of animals and plants of the Préfecture de la Gironde (Identification number A33-063-324). PF depend on acetate produced in the mitochondrion to feed fatty acid biosynthesis through the essential enzyme AceCS [13]. A western blot analysis showed that AceCS (74 kDa) was expressed at the same level in the BSF and PF, and an immunofluorescence analysis using the anti-AceCS immune serum showed a homogeneous diffuse pattern characteristic of a cytoplasmic localization (Fig. 2), suggesting that this pathway may also exist in BSF. AceCS is essential for BSF viability, as demonstrated by the death of the RNAiAceCS. i cell line three days post-induction of down-regulation of the AceCS gene expression (. ni and. i stands for uninduced and tetracycline-induced, respectively) (Fig. 3A). To investigate the role of AceCS, radiolabel incorporation into fatty acids from [1-14C]-acetate was measured for the parental and RNAiAceCS cell lines incubated in the IMDM medium. Label incorporation into fatty acids was reduced 2. 1- and 8. 1-fold one and two days after tetracycline addition, respectively, which correlates with the reduction of AceCS expression (Fig. 3B). Altogether these data demonstrate that proliferative BSF require acetate to feed the essential fatty acid biosynthetic pathway, as previously observed in PF [13]. The IMDM medium does not contain acetate, except the minor contribution of the 10% FCS supplement (∼5 µM) [34]. Consequently, LS-BSF may produce acetate from the catabolic pathways previously identified in PF, i. e. mitochondrial production of acetyl-CoA from glucose and threonine degradation through PDH and TDH, respectively [3], followed by conversion of acetyl-CoA into acetate by two mitochondrial enzymes, ASCT [9] and ACH [10]. A western blot analysis showed that ASCT (54 kDa), ACH (40. 5 kDa), PDH-E2 (E2 subunit of PDH, 49. 6 kDa) and TDH (39. 5 kDa) are expressed in the 427 BSF strain, which has lost the ability to differentiate into SS-BSF (Fig. 2A). This was confirmed by determination of the PDH and TDH activities in LS-BSF, which were 4-fold and 6. 7-fold lower than PF, respectively (Fig. 2A). Immunofluorescence analyses revealed colocalization of PDH-E1α (E1α subunit of PDH) and TDH with the mitochondrion-specific dye MitoTracker Red CMXRos (Invitrogen) (Fig. 2B). The mitochondrial localization of TDH is consistent with a 24-amino-acid N-terminal mitochondrial targeting signal predicted by MitoProt (http: //ihg. gsf. de/ihg/mitoprot. html) with a high probability (0. 82). Since BSF express the whole set of enzymes required for acetate production, we then used a combination of reverse genetics on PDH-E2 and TDH and metabolic profiling by NMR to investigate mitochondrial acetate production in BSF. It is widely considered that LS-BSF excrete only pyruvate from glucose metabolism, while PF mainly produce acetate and succinate. To compare glucose metabolism in these two forms, 2. 5×107 LS-BSF and 5×107 PF (equivalent to 0. 25 mg of proteins) were incubated in 2. 5 mL of PBS containing 4 mM glucose (Fig. 4) or [U-13C]-glucose (Fig. 5A and Table 1). 13C-enriched end-products excreted in the medium from [U-13C]-glucose metabolism were quantified by 1H-NMR (Table 1). It is to note that quantification errors are significant, in particular for molecules representing less than 5% of all excreted end-products. As expected, BSF mainly converted glucose into pyruvate (7761 nmol/h/108 cells), which accounts for 85. 1% of the excreted end-products. In addition, BSF excreted significant amounts of alanine, acetate and succinate, which represent 9. 2%, 4. 9% and 0. 8% of the excreted end-products from glucose metabolism, respectively (Table 1). Surprisingly, the rate of excretion of 13C-enriched acetate and succinate from [U-13C]-glucose was only 2-fold lower in BSF than in PF (446 versus 789 nmol of acetate/h/108 cells and 71 versus 156 nmol of succinate/h/108 cells, respectively) (Table 1, see Fig. 4). The unexpected similar rate of acetate and succinate excretion in both trypanosome forms is probably due to the ∼10-fold higher glycolytic rate in BSF (the rate of glycolytic end-product excretion was 9. 4-fold higher in BSF compared to PF - Table 1). Consequently, the high glycolytic rate in BSF combined with the dominant conversion of glucose into pyruvate (85. 1% of the excreted end-products) may have led to underestimation of the role of acetate and succinate production in LS-BSF, although their rate of production were in the same range in PF. To confirm acetate production from glucose metabolism by LS-BSF, we conducted RNAi-mediated down-regulation of expression of the PDH-E2 gene. The RNAiPDH. i cell line showed no growth phenotype upon tetracycline induction (Fig. 6A), although the PDH-E2 protein was no longer detectable by western blot two days post-induction (Fig. 6A, inset). Metabolite profiling of the RNAiPDH. i cell line incubated in the presence of 4 mM of [U-13C]-glucose showed a 13. 7-fold reduction of acetate production from glucose compared to the RNAiPDH. ni cells (55 versus 755 nmol/h/108 cells) (Table 1). It is to note that, for unknown reasons, the uninduced RNAiPDH. ni cell line produces ∼1. 7-times more acetate from glucose than the parental cells. Since both BSF and PF have been reported to produce acetate from threonine [12], which is present in the IMDM medium (0. 9 mM), we investigated the threonine degradation pathway in BSF. Incubation of the parasites in threonine-depleted medium, which contains only ∼15 µM of the amino acid coming from FCS [35], did not affect growth of the wild-type and RNAiPDH. ni cells, while growth of the RNAiPDH. i mutant was abolished (Fig. 6D). To confirm that glucose and threonine degradations contribute to acetate production, both pathways were interrupted by down-regulating PDH-E2 expression in the TDH null background (Δtdh/RNAiPDH cell line). First, both TDH alleles were replaced by the puromycin (PAC) and blasticidin (BSD) markers in the Δtdh cell line, with no effect on growth rate (Fig. 6C). Deletion of both TDH alleles was confirmed by PCR analyses (Fig. 6B), western blot analyses and enzymatic assays (insets of Fig. 6C). Second, RNAi-mediated down-regulation of PDH-E2 was performed in the Δtdh background. Growth of the Δtdh/RNAiPDH. i cell lines was abolished three days post-induction before cell death seven days later (Fig. 6E). Addition of 4 mM acetate in the medium does not rescue growth of the Δtdh/RNAiPDH. i mutant (data not shown) suggesting that acetate and/or acetyl-coA need to be produced inside the mitochondrion to feed the essential mitochondrial fatty acid pathway, may be through the production of the precursor butyryl-CoA [36]. This result confirms that abolition of mitochondrial acetyl-CoA/acetate production from both glucose and threonine is lethal for BSF grown in standard medium. To address this question we developed a metabolite profiling assay based on the ability of 1H-NMR spectrometry to distinguish 13C-enriched molecules from 12C ones. Cells were incubated in PBS with equal amounts (4 mM) of [U-13C]-glucose and unenriched threonine in order to perform a quantitative analysis of threonine-derived and glucose-derived acetate production by 1H-NMR. When [U-13C]-glucose was the only carbon source in the incubation medium, the excreted [13C]-acetate (annotated A13 in Fig. 5) was represented by two doublets with chemical shifts at around 2. 0 ppm and 1. 75 ppm, respectively (see Fig. 5A). It is to be noted that threonine metabolism cannot be analyzed independently since glucose is essential for BSF. Addition of threonine to the [U-13C]-glucose/PBS medium induced production of threonine-derived [12C]-acetate (386 nmol/h/108 cells) in addition to [13C]-glucose-derived [13C]-acetate (532 nmol/h/108 cells) (Fig. 5B and Table 2). This shows that in the presence of equal amounts of both carbon sources, glucose contributes ∼1. 4-fold more than threonine to acetate production. 1H-NMR metabolite profiling of the single and double mutants confirmed the involvement of both glucose and threonine in acetate production. As expected, production of [13C]-glucose-derived [13C]-acetate was ∼50-times lower in the RNAiPDH. i than in the RNAiPDH. ni cells (16 versus 847 nmol/h/108 cells), while threonine-derived acetate production was not affected. Conversely, production of threonine-derived acetate was abolished in the Δtdh mutant, while [13C]-glucose-derived [13C]-acetate was not affected (Table 2 and Fig. 5C–D). Finally, production of acetate from both carbon sources was affected in the Δtdh/RNAiPDH. i double mutant cell line (Table 2 and Fig. 5E). BALB/c mice immunocompromised by Endoxan treatment were injected with wild-type and RNAiPDH cells and kept with or without doxycycline, a stable tetracycline analog, in the drinking water to down-regulate expression of PDH-E2. Animal survival and the blood parasite levels were monitored. No differences were observed between the four groups of animals, in which parasite density started to rise at day three post-infection. All mice were dead at days 6–7 post-infection (data not shown). This shows that acetate production from glucose is not necessary for the viability of T. brucei in vivo, suggesting that a possible acetate source (threonine) that is present in the blood is absent in the threonine-depleted in vitro culture medium. As mentioned above, mammalian blood contains approximately 150 µM threonine [35], [37], which is 10-times higher than in the threonine-depleted IMDM medium. The RNAiPDH. i cell line died in IMDM medium containing 15,37. 5 and 75 µM threonine, while addition of 150 µM of the amino acid restored its growth in vitro (Fig. 6D), suggesting that the homeostatic threonine blood concentration (150 µM) is sufficient to provide BSF with the required acetyl-CoA/acetate molecules. Altogether, this demonstrates that BSF trypanosomes have developed two complementary and self-sufficient ways to maintain the essential production of acetate in the blood of mammalian hosts. LS-BSF trypanosomes are well known for their glucose-dependency to satisfy ATP requirements [38]. Indeed, net production of all cellular ATP is fulfilled by the last glycolytic step catalyzed by pyruvate kinase, which produces pyruvate, the excreted glycolytic end-product. Excretion of significant amounts of other partially oxidized end-products of glycolysis, such as glycerol, succinate and alanine, has been previously reported [14], [17], [39], [40]. However, in the late seventies emerged a general dogma whereby pyruvate was considered the exclusive glycolytic end-product excreted from LS-BSF under aerobic conditions [15], [16], because the minor end-products were assigned to either non-growing conditions or contamination with non-dividing SS-BSF [18], [19]. Here, we used as an experimental model the 427 BSF strain, which has lost the ability to differentiate into SS-BSF, in order to focus our analysis of glucose metabolism on LS-BSF trypanosomes. This laboratory-adapted monomorphic strain is insensitive to the stumpy inductor factor, but, it successfully differentiates in vitro into bona fide SS-BSF, for instance when expression of the protein kinase target of rapamycin (TOR4) is inhibited [41]. This suggests that the 427 strain can be considered as a slender-like BSF that has lost the ability to respond to the stumpy inductor factor, and as such is the relevant model to study the metabolism of proliferative BSF. Our metabolic analyses showed that LS-BSF can produce almost as much succinate and acetate from glucose as PF incubated in the same conditions. This suggests that most, if not all, enzymes involved in the “succinate and acetate branches” previously characterized in PF are also expressed in LS-BSF. To produce acetate, PDH (step 1 in Fig. 1) converts pyruvate into acetyl-CoA, which is the substrate of ASCT (step 2) and ACH (step 3) for acetate production. ASCT expression is low in BSF (Fig. 2 and [19]), while ACH is relatively abundant (Fig. 2) with an ACH activity ∼2-fold higher than PF (data not shown). Three of the four PDH subunits have been investigated so far and are expressed in BSF (PDH-E1α and PDH-E2, see Fig. 2; PDH-E3, [42]), with a PDH enzymatic activity only 4-fold lower than in PF (Fig. 2). This relatively high PDH activity is in agreement with a recent comparative SILAC proteomics analysis showing that PDH-E1α, PDH-E1ß, PDH-E2 and PDH-E3 are 5. 3-, 7. 6-, 5. 2- and 8. 1-fold more abundant in PF than LS-BSF, respectively [43]. The same proteomics analysis in LS-BSF also detected most, if not all, of the enzymes involved in succinate production from phosphoenolpyruvate, although at a lower level of expression than in PF (between 3- and 20-fold). Altogether, this clearly demonstrates that LS-BSF have maintained the capacity to produce and excrete acetate and succinate from glycolysis. The relatively high rate of acetate and succinate production (only ∼2-fold higher in PF), while the enzymes involved in the corresponding metabolic pathways are 5- to 20-times more abundant in PF, may be due to the considerably higher glycolytic flux in BSF. We determined that the excretion rate of glycolytic end-products is 9. 5-fold higher in BSF than in PF (9115 versus 968 nmol of excreted end-products/h/108 cells), which is consistent with the previously measured 5- to 10-fold difference in glycolytic flux [14]. A recent analysis of the glycolytic flux in the same BSF strain incubated in growing conditions (IMDM containing 20 mM glucose) showed a higher rate of pyruvate production compared to our analysis performed in PBS containing 4 mM glucose and threonine (19. 2 versus 12. 0 µmol/h/108 cells) [44]. The reduced glycolytic flux observed in PBS conditions certainly reflects the difference between non-growing conditions (PBS) and exponential growth (IMDM) with a doubling time in the range of 5 h [44]. Also, the 35% reduction in total end-product fluxes for the wild type cells depending on the available substrates (PBS/glucose versus PBS/glucose/threonine) shows that metabolic fluxes are dependent on the exact context of substrates present (Table 1). It is important to note that our experimental procedures do not reflect physiological conditions, since trypanosomes were incubated at high density in PBS containing 4 mM glucose. Consequently, these minor glycolytic end-products might be excreted at a lower rate, or not at all, by LS-BSF trypanosomes in vivo. A recent quantitative analysis of the fate of glucose in exponentially growing 427 LS-BSF in vitro (the same strain analysis here) showed that pyruvate is the only excreted glycolytic end-product [44]. Glucose, pyruvate and glycerol were analysed in that study. Although they report an almost complete carbon balance between glucose uptake and pyruvate excretion, their analysis leaves room for small fluxes towards products they did not analyse such as acetate and succinate. We here report these end-products, with fluxes to acetate and succinate together representing ∼5% of the excreted glycolytic end products. The exact fraction of total carbon going to these end-products is difficult to assign, due to the errors in quantification of fluxes and because our results did not enable the calculation of a carbon balance between carbon uptake and excretion. Whatever the rate of acetate excretion from glucose metabolism in exponentially growing LS-BSF is, its production is essential for growth, as exemplified by the death of the RNAiPDH. i cell line incubated in the absence of threonine, the other acetate source. This was confirmed by inducing cell death upon blocking acetate production from both carbon sources in the Δtdh/RNAiPDH. i double mutant, while growth of the corresponding single mutants in standard IMDM medium was not affected (Fig. 6). The relevance of acetate production from glycolysis for LS-BSF is further strengthened by (i) the same high rate of growth of the wild-type parasite, even in the absence of threonine (Fig. 6), as recently observed by the development of a new minimal medium that supports growth of BSF [37] and (ii) the impossibility to rescue growth of the Δtdh/RNAiPDH. i double mutant by addition of sodium acetate in the medium, which cannot substitutes glucose-derived acetate production. The above-mentioned reverse genetic experiments combined with 1H-NMR metabolic analyses also clearly demonstrate that glucose and threonine are the only significant carbon sources contributing to the essential production of acetate in the 427 BSF strain. To our knowledge, this is the first report showing acetate production from glucose in LS-BSF, while threonine has been described before as an acetate-source in BSF [12], [45]. Inhibition of a single acetate-production pathway in the RNAiPDH. i and Δtdh cell lines grown in standard IMDM medium does not affect growth of LS-BSF, indicating that a single active pathway is sufficient for growth in vitro. This is probably also true in vivo, since (i) glucose concentration is higher in the blood than in our experimental conditions (5 versus 4 mM) and (ii) the relatively low homeostatic concentration of threonine in mammalian blood (150 µM) is sufficient for acetate production, as demonstrated by the absence of growth effect of the RNAiPDH. i mutant in medium containing at least 150 µM threonine (Fig. 6D). When incubated with equal amounts of threonine and glucose (4 mM), the parasite produces ∼1. 4-fold more acetate from glycolysis. Mammalian blood contains ∼30-fold more glucose than threonine (5 mM versus 150 µM), which strengthens the view that the contribution of glucose to acetate production is relevant in vivo. This contrasts with an equivalent recent analysis performed on the procyclic form of T. brucei, which prefers threonine for acetate and fatty acid productions, with a ∼2. 5-fold higher contribution of threonine compared to glucose when incubated with 4 mM of both carbon sources [46]. In PF trypanosomes, the mitochondrial production of acetate is essential to feed de novo fatty acid biosynthesis through the “acetate shuttle” [13]. In this shuttle, acetate produced in the mitochondrion reaches the cytosol, where a part of it is converted by AceCS into acetyl-CoA to produce malonyl-CoA, the elongator for fatty acid biosynthesis. It is noteworthy that both the microsomal elongase-dependent and mitochondrial type II fatty acid synthase pathways use malonyl-CoA to elongate fatty acids [47]. As observed in PF, AceCS is essential for incorporation of [1-14C]-acetate into LS-BSF fatty acids (Fig. 3), indicating that AceCS is required for de novo fatty acid synthesis. In addition, Gilbert et al. previously showed that both glucose and threonine are used as carbon sources for fatty acid synthesis [47]. Altogether, this demonstrates that, like PF, proliferative BSF trypanosomes use the “acetate shuttle” to feed fatty acid biosynthetic pathways. The essential role of mitochondrial fatty acid synthesis has been documented in BSF, however, RNAi-mediated down regulation of elongase genes involved in the microsomal fatty acid biosynthesis does not affect growth of the parasite [36], [48], [49], [50]. Consequently, the lethal phenotype observed for the LS-BSF RNAiAceCS mutant is probably due to the dramatic reduction of cytosolic malonyl-CoA production required to feed mitochondrial fatty acid biosynthesis, which contribute to ∼10% of cellular fatty acid production [36]. As mentioned above, our detection of several “minor” glycolytic end-products excreted by BSF trypanosomes is consistent with most, if not all, early reports [14], [17], [39], [40] and the recent quantitative flux analysis in BSF 427 [44]. The high glycolytic flux combined with the almost exclusive conversion of glucose into pyruvate has probably caused the community to overlook these minor end-products and the metabolic pathways leading to their production, although their biosynthesis may be essential for the parasite, as described here for acetate. For instance, we also observed that LS-BSF excretes alanine from glucose (twice more than acetate, Table 1), as reported before [17]. Alanine is produced from pyruvate by transfer of an amino group coming from various possible amino acid sources. A recent metabolomic analysis showed that glutamate and hydrophobic keto acids accumulate in the BSF spent media, suggesting that glutamine and hydrophobic amino acids are possible substrates of alanine aminotransferase for alanine production [37]. The relevance of alanine production from glycolysis is strengthened by the reported accumulation of hydrophobic keto acids in the plasma and urine of infected rodents [51], [52] and the requirement of alanine aminotransferase activity for in vitro growth of the parasite [53]. Succinate production from glycolysis may also be of importance for biosynthetic pathways in LS-BSF, as exemplified by the requirement of fumarate for de novo synthesis of pyrimidine through the unusual fumarate-dependent dihydroorotate dehydrogenase [54]. Altogether, this highlights the need to reconsider and further investigate the metabolic pathways leading to minor glycolytic end-products, which are still considered negligible compared to the total carbon flux from glucose [44], in order to reveal new essential metabolic pathways that could be targeted to develop new trypanocidal molecules. Beyond glycolysis, other overlooked metabolic pathways of the central metabolism need to be revisited in LS-BSF, as exemplified by the recent observation that RNAi down-regulation of the tricarboxylic acid enzyme succinyl-CoA synthetase induces one of the most spectacular death phenotypes observed in BSF, with 100% cell death within less than 20 h post-induction [55], while this pathway is considered repressed in BSF.

Many protists, including parasitic helminthes, trichomonads and trypanosomatids, produce acetate in their mitochondrion or mitochondrion-like organelle, which is excreted as a main metabolic end-product of their energy metabolism. We have recently demonstrated that mitochondrial production of acetate is essential for fatty acid biosynthesis and ATP production in the procyclic insect form of T. brucei. However, acetate metabolism has not been investigated in the long-slender bloodstream forms of the parasite, the proliferative forms responsible for the sleeping sickness. In contrast to the current view, we showed that the bloodstream forms produce almost as much acetate from glucose than the procyclic parasites. Acetate production from glucose and threonine is synthetically essential for growth and de novo synthesis of fatty acids of the bloodstream trypanosomes. These data highlight that the central metabolism of the bloodstream forms contains unexpected essential pathways, although minor in terms of metabolic flux, which could be targeted for the development of trypanocidal drugs.

Abstract Introduction Materials and Methods Results Discussion

2013

Revisiting the Central Metabolism of the Bloodstream Forms of Trypanosoma brucei: Production of Acetate in the Mitochondrion Is Essential for Parasite Viability

DNA molecules are highly charged semi-flexible polymers that are involved in a wide variety of dynamical processes such as transcription and replication. Characterizing the binding landscapes around DNA molecules is essential to understanding the energetics and kinetics of various biological processes. We present a curvilinear coordinate system that fully takes into account the helical symmetry of a DNA segment. The latter naturally allows to characterize the spatial organization and motions of ligands tracking the minor or major grooves, in a motion reminiscent of sliding. Using this approach, we performed umbrella sampling (US) molecular dynamics (MD) simulations to calculate the three-dimensional potentials of mean force (3D-PMFs) for a Na+ cation and for methyl guanidinium, an arginine analog. The computed PMFs show that, even for small ligands, the free energy landscapes are complex. In general, energy barriers of up to ~5 kcal/mol were measured for removing the ligands from the minor groove, and of ~1. 5 kcal/mol for sliding along the minor groove. We shed light on the way the minor groove geometry, defined mainly by the DNA sequence, shapes the binding landscape around DNA, providing heterogeneous environments for recognition by various ligands. For example, we identified the presence of dissociation points or “exit ramps” that naturally would terminate sliding. We discuss how our findings have important implications for understanding how proteins and ligands associate and slide along DNA. DNA is a charged, semi-flexible polymer, carrying two elementary negative charges per base-pair. Mobile counterions screen the inter-strand electrostatic repulsion, hence, stabilizing the double helix. Moreover, since DNA molecules are highly rigid, mobile ions also influence DNA’s mechanical properties, for example, favoring bending and substantially influencing DNA’s conformational preferences [1,2]. The ionic atmosphere around DNA is distinctly nonhomogeneous, where different counterions preferentially associate, for example, with the DNA grooves or the strands [3–5]. The helical nature of DNA plays a critical role in determining the distribution of counterions, providing a variety of local environments. For example, the minor groove enhanced electrostatic potential provides binding sites to positively charged ligands. Furthermore, the width of the minor groove, which is highly correlated with its electrostatic potential, is a readout mechanism determined by the sequence dependent geometry of the DNA molecule [6]. The organization and dynamics of condensed ions can also give rise to complex interactions between DNA molecules, that lead, for example, to aggregation [7–10]. Despite substantial prior work on elucidating the counterionic atmosphere around DNA, the topography and roughness of the energy landscape for ligand binding to DNA are still not well understood. However, such binding landscape features are important, since they determine how binding partners, such as counterions, drugs or proteins, move around the DNA chain. In the eukariotic cell nucleus, DNA molecules associate with a variety of counterions, proteins and other molecules. By associating with these molecules, the DNA chain condenses into organized chromatin structures. In this compact state, a twofold challenge needs to be overcome: the DNA-binding proteins need to associate tightly to their targets, while also being able to seek out the specific site in an efficient way among a myriad of non-specific decoys. Even though numerous studies have shed light on the molecular basis of specific protein-DNA complex formation, much less is known regarding the mechanisms allowing DNA-binding proteins to find their specific targets. Experimental [11–13], theoretical [14,15] and computational [16–20] studies have suggested that some of the processes involved in the search procedure are: 1) one-dimensional sliding of the protein along DNA (intramolecular translocation), 2) direct transfer from one DNA segment to another and 3) jumping from one DNA segment by dissociation and re-association [11,21–23]. The first process, protein sliding or translocation, corresponds to a one-dimensional diffusion process where proteins or other ligands first associate non-specifically to DNA molecules, followed by thermally induced motions along the DNA chains. For the term sliding to make sense, the protein needs to move significant distances on a DNA surface (in either direction) before dissociating. This basically necessitates relatively low free energy barriers with respect to the longitudinal motions along DNA and relatively high (or moderate) free energy barriers to prevent dissociation [14,15]. Recent studies allowed visualization of single transcription factors sliding along extended DNA molecules, offering a step towards understanding the one dimensional translocation or diffusion search processes [22,24,25]. These experiments indirectly suggest that the protein’s translocation motion is coupled to the rotation along the DNA’s axis. It is believed that, while sliding, an ensemble of rapidly fluctuating nonspecific protein-DNA interactions allow the ligand to maintain continuous contact with the major groove, minor groove, or both [26,27]. These interactions are presumably mostly electrostatic, however, as the protein finds its DNA target site, they switch to sequence specific interactions, such as hydrogen bonding and van der Waals interactions. For example, insertion of arginine residues into the narrow minor groove, where the electrostatic potential is strongly enhanced, is a widely used mode of non-specific protein-DNA recognition [28,29]. In DNA regulatory processes, which are thought to be highly dynamic, the helical symmetry of the DNA segment plays a critical role in providing a unique physico-chemical environment for ligand binding to a specific DNA sequence. These different environments are widely exploited by proteins and other regulatory molecules in order to tightly regulate transcription and translation. Unlike most previous modeling efforts, the current study focuses on how counterions and charged residues might move along the DNA, where its local helical geometry is emphasized and fully taken into account. Our computational approach uses molecular dynamics (MD) simulations to determine the three dimensional potential of mean force (3D-PMF) of these charged ligands. For this purpose, we developed a helical coordinate system that allowed us to constrain the ligand to track the minor groove. We studied two small charged ligands known to localize to the minor groove: a Na+ ion and methyl-guanidinium, an arginine side chain analog. In prior works, tracking of helical paths was used as reaction coordinate [20]. However, a helical coordinate system tiling the three-dimensional space is introduced in this work for the first time, to the best of our knowledge, . Focusing on two small ligands allowed us to achieve two goals: 1) investigate the accuracy and efficiency of 3D-PMF simulations when using the newly introduced helical coordinates, and, furthermore, 2) map out the fine-scale roughness of the binding free energy landscape of the minor grove. The latter investigation would be difficult to carry out with protein size probes, that would introduce significant disturbances into the minor groove. Fine-grained mapping of the binding free energy landscape provides important mechanistic information about the role of non-specific interactions in protein sliding [30,31], as elaborated below. The computed free energy landscapes directly illustrate the binding sites and energetic barriers that may be encountered by mobile ligands or proteins. The helical coordinate system presented in this work makes it possible to straightforwardly compare the energetics of association and dissociation processes to those of sliding motions along the helical path. Furthermore, we determined the changes in the solvent structure at the interface between the ligand and the DNA molecule, highlighting the rapidly fluctuating nature of DNA-solvent-ligand nonspecific interactions. Overall, our simulation results shed light on the way the roughness of the free-energy landscape in the vicinity of a DNA segment modulates binding and diffusion of ligands. Using all-atom umbrella sampling (US) MD simulations with the above-mentioned helical coordinate system, we calculated the PMF for a Na+ ion tracking the DNA’s minor groove to probe the roughness of the binding free energy landscape (Fig. 2). Notably, the free-energy landscape varies significantly along one helical turn, indicating a rough free energy surface, with large energy barriers and free energy minima. For example, in most cases, the sites of lowest free-energy are deeply buried in the minor groove (i. e. closer to the DNA’s axis). However, large free energy barriers were also identified (≳ 5 kcal/mol), possibly due to steric hindrance, as the ligands comes closer to the DNA’s axis or backbone. The large free-energy barrier in the first-quadrant (i. e. ϕ in the range from 0 to 90o) are due to the deformation of the DNA which causes a steric clash between the Na+ cation and the backbone between T5 and C6 of the 3’ to 5’ strand (S4 Fig.). On the other hand, at intermediate radii from the DNA’s axis (∼ 12 Å), the energy landscape becomes smoother (Fig. 3. A). Binding sites along the studied region are not necessarily well localized and can extend for several base pairs (Fig. 3. A). For the specific DNA sequence studied in this work, the computed energy landscape suggests that the Na+ ions localize to a 5 base-pair segment. Previous studies have determined that Na+ ions bind to AT rich regions for periods of 50 ns [3]. These binding sites are the main contributors to slowly exchanging Na+ ions between the DNA’s minor groove and the surrounding bulk solvent [33]. Our calculations are in agreement with these findings, but suggest that once buried into the minor groove, Na+ ions may easily slide along the helical path, moving slightly away from the helical axis during sliding, when needed. The radial distribution function (gOW−Na+) of water oxygens around the Na+ ions (Fig. 3. B) shows that the minor groove is able to accommodate hydrated Na+ ions. The Na+ ions probed inside the minor groove maintain most of their first hydration shell (first peak, Fig. 3. B) and are characterized by partial dehydration of the second hydration shell (second peak, Fig. 3. B). This result suggests that the water molecules mediate the interactions between the Na+ cations and the DNA molecule and, consequently, contribute to the localization of the ions to the minor groove. This is consistent with having a shared hydration spine along the minor groove [34], where the water coordination of the Na+ and the DNA molecules allows the movement of the cations along the helical path, without necessarily unbinding and rebinding. We computed the 3D-PMF for the arginine analog, methyl guanidinium, to illustrate the positional free-energy of this ligand in the minor groove. At first sight, the 2D projection of the 3D-PMF (top view, Fig. 4. A) shows that the surface of the minor groove represents an intricate landscape with various local free-energy minima, with free-energy barriers of up to 5 kcal/mol. The 1D-PMFs computed at different angular locations (Fig. 4. B) show that the free-energy barriers to removing the ligand from the minor groove can vary significantly, ranging from 1. 5 to 5 kcal/mol. These free-energy differences can be interpreted as the unbinding energies at different locations. On the other hand, the free-energy profiles of the ligand sliding along the helical path (Fig. 5. A) show energy barriers ranging from 0. 5 to 1. 5 kcal/mol. Additionally, the minimum free energy path along the studied turn (Fig. 5. B) indicates that a small free ligand, such as methyl guanidinium, would localize to a free energy minimum having a depth of 1. 5 kcal/mol. Notably, the free-energy profiles vary significantly at different radii, indicating that, at an intermediate radius (i. e. ∼ 9. 3 Å) the free-energy profile for sliding has the smallest barriers (≲ 0. 8 kcal/mol). Consequently, in general, the energy barriers are smaller in the angular direction than in the radial direction, making sliding the preferred mechanism for moving methyl guanidinium along the DNA chain. Fig. 6. A shows the 2D projections of the electrostatic potential inside the minor groove along one DNA turn, analogous to the 2D projections of the 3D-PMF (Fig. 5. B). The Pearson’s correlation between the free-energy (Fig. 4. A) and potential energy (Fig. 6. A) is r≃0. 91. This result supports the view that the enhanced electrostatic potential in the minor groove is a key determinant of the free-energy landscape. Furthermore, the computed DNA’s minor groove width (Fig. 6. B) shows that, along the studied turn, there are two narrower segments. Qualitatively, the location and depth of the free-energy minima (Fig. 5. B) correlate with the narrow regions of the minor groove (correlation coefficient r≃0. 82, Fig. 6. B) and electrostatic potentials (Fig. 6. A), in agreement with Honig and co-workers [28,35]. However, at some different radii, the free energy profiles (Fig. 5. A) do not correlate well with the widths of the minor groove. This might have important implications for proteins binding and sliding along DNA, as further elaborated below. In addition, in the studied spatial region around the DNA segment, the ranges of the free energy differences and the electrostatic potentials differ from each other, being 5. 3 kcal/mol and 3. 5 kcal/mol, respectively. This range difference is most likely associated with the inaccurate treatment of complex hydration effects in the minor groove when using continuum electrostatic approaches, as well as the importance of non-electrostatic interactions. Given the high charge density of the DNA backbone, it is expected that the binding free-energy (Fig. 5) and electrostatic potential (Fig. 6) landscapes are not completely smooth at large distances (≳ 12 Å). Our simulations indicated that the presence of the methyl-guanidinium ligand did not noticeably affect the DNA’s minor groove geometry. However, it may be possible that the timescale for groove deformations is significantly longer than the simulated time for each of our US windows. As the methyl-guanidinium ligand is removed from the minor groove, the head group shows partial dehydration, as shown by the radial distribution function gOW−cation (Fig. 7). This non-trivial hydration profile suggests that the largest dehydration occurs at an intermediate range of radii between 11 and 12 Å off the DNA’s axis (Fig. 7. B), which coincides with the position of the DNA’s phosphate groups. Consequently, the methyl-guanidinium ligands become partially dehydrated as they get into the minor groove, but once further buried into the minor groove, water molecules relocalize to the first hydration shell. Fig. 7. A also indicates partial dehydration of the second hydration shell for methyl-guanidinium cations buried in the minor groove. These hydration patterns reveal the presence of water mediated interactions [36,37] between the methyl-guanidinium ligand and the DNA molecule. These observations suggest that the free-energy barriers of radial ligand movement not only depend on the electrostatic potential inside the minor groove but also on the sizes and the hydration levels of the ligand. The methyl-guanidinium ligand is capable of forming direct and water-mediated hydrogen bonds with the DNA molecule. We determined that, on average, the ligand forms ∼ 0. 7–1 direct hydrogen bonds with DNA, having an average life time of ∼ 25 ps. The number of hydrogen bonds decreases slightly as the ligand is removed from the minor groove, and is higher at the intermediate radii (i. e. ρ ∼ 9. 3 − 10. 0 Å). Water mediated hydrogen bonds were found to be very short lived, with an average lifetime of ∼ 2. 8 ps. These results corroborate the idea that protein-DNA interactions in the minor groove are non-specific and rapidly fluctuating. In proteins, this may allow the ligand to maintain contact with the DNA molecule without the entropic cost of fixing the conformation of the side chains [27,38]. By computing the 3D-PMFs discussed above, we have characterized the physicochemical environment of the minor groove for two small charged ligands: Na+ and methyl-guanidinium. These maps can be used to shed light on how proteins might slide along the DNA axis while searching for their binding target. For example, the calculated PMFs show that, even for very small ligands, the surface of the minor groove is very rough and that the fine structure of the landscape varies notably at different locations. For example, smoother free energy profiles were obtained at intermediate radii (Figs. 3. A and 5. A). It is expected that small ligands, such as the studied here, will localize to the sites of lowest free energy (Fig. 5. B), even if these binding sites are buried deep inside the minor groove. However, this scenario would change for larger ligands or proteins, where the inner binding sites might not be accessible due to steric or geometrical constraints. For particular systems, such as glycosylases, experimental evidence support the concept that a single ‘wedge residue’ can be used to scan DNA [31]. In the case of sliding proteins, having a smooth free energy landscape would prove advantageous and favor moving along the helical path rather than dissociation and re-association. The methods presented here can be used to study proteins, such as transcription factors, sliding along DNA. Defining a reaction coordinate to track the minor or major grooves, or both, is natural and simple in the helical coordinate systems. Furthermore, using a coordinate system that is congruent with the geometry of DNA allows for straightforward studying of the preferred paths for proteins and other ligands. Additionally, our work corroborates that the pre-existing geometry of the minor groove, determined mainly by the sequence, is a major determinant of the presence of binding sites for positively charged ligands. We identified that the binding sites are located in the narrow AT rich regions, where the electrostatic potential is strongly enhanced. Therefore, in the context of charged ligands bound to the DNA’s minor groove, a sliding mechanism can be described as follows: the roughness of the free energy landscape strongly depends on 1) the degree to which the ligand is buried and 2) the geometry of the minor groove. In particular, if the ligand is constrained to move at the intermediate radii, the height of the free energy barriers are comparable to the thermal fluctuations and, consequently, sliding along DNA helical path, rather than dissociation, is energetically favorable. In the current implementation, it is required that the DNA’s axis is aligned to the z-axis, thus not being applicable in situations that require DNA translation or bending. For applications that require allowing the DNA molecules to translate or bend, our approach can be straightforwardly generalized in such a way that the helical coordinate system is defined locally, based on the instantaneous orientation of the axis of some DNA segment, while at the same time allowing for dynamic adjustment of this orientation during the simulation time course. We have presented in this work a helical coordinate system, which facilitates studying the physicochemical environments surrounding DNA molecules, allowing natural tracking of DNA’s minor and major grooves. This coordinate system is fully congruent with the helical symmetry of DNA molecules and is general enough for studying the energetics of ligand or protein interactions with various DNA segments in molecular detail. As a key advantage, the helical coordinates can be used to directly obtain 3D-PMFs of ligand association to DNA, in order to determine whether sliding or unbinding is more energetically favorable at each spatial location. The 3D-PMFs can be used to determined the preferred paths of diffusive proteins and other ligands in the DNA’s vicinity. The computed PMFs indicate that the spatially resolved binding landscapes around DNA chain segments are far from smooth, even for small ligands, showing rich fine structures at different positions. We determined that ligands need to overcome free energy barriers of up to ∼ 5 kcal/mol when dissociating. On the other hand, while sliding, ligand encounter free energy barriers of up to ∼ 1. 5 kcal/mol. Consequently, in general, sliding is favored over dissociation. Nevertheless, due to the roughness of the energy landscape and the non-homogeneity of the energy barrier locations, we hypothesize that variations in the geometry of the minor groove can be exploited to create dissociation points or “exit ramps”, to halt sliding. We found that the smallest free energy barriers are encountered by ligands that slide at the intermediate radii from the DNA axis, with the barriers being less than 1. 5 and 0. 8 kcal/mol for Na+ and methyl guanidinium, respectively. We also confirmed prior suggestions that DNA’s free energy landscape is highly modulated by its electrostatic potential, the latter being mostly determined by the sequence-dependent geometry of the minor groove. Additionally, we found that both ligands were only partial dehydrated inside the minor groove and that water-mediated interactions between the ligands and the DNA may play a critical role in favoring sliding over dissociation. Our study provides a general framework for characterizing the free energy landscapes surrounding DNA molecules and for making quantitative predictions of the energetics and molecular basis of different types of diffusive motions in the proximity of DNA chains. All of our simulations were carried out using the LAMMPS MD software [39], the amber parmbsc0 forcefield for nucleic acids [40], the Joung and Cheatham ions parameters [41] and the TIP3P water model. Starting from a canonical B-form 20-base pair DNA oligomer, [d (CGCGAGGTTTAAACCTCGCG) ]2, we solvated the system in a 50 × 50 × 70 Å box of water with periodic boundary conditions, that were applied throughout all the simulations. In addition, each strand of the DNA molecule was covalently liked to itself over the periodic boundaries of the system [42,43]. This setup allowed us to suppress all deformation modes whose wavelengths were bigger than the system size. Na+ and Cl− ions were added to compensate charge and to represent a 0. 15 M concentration that mimics the physiological environment. The total number of atoms in the system was 15,000. The system was first minimized by 2000 cycles while constraining the DNA heavy atoms followed 1000 steps of unconstrained energy minimization using the steepest descent method. The system was then heated to 300K for 1 ns. This was followed by a round of MD simulations at constant pressure (NPT) for 1 ns using a Langevin piston pressure control to obtain a pressure of 1 atm for density equilibration. After density equilibration, the average size of the box was 48. 4 × 48. 1 × 67. 1 Å. The final round of equilibrations was performed at 300K for 50 ns to ensure the equilibration of the ions. All equilibration and production runs were performed at constant volume, using the average volume measured in the NPT simulations. To maintain a constant temperature, a Langevin thermostat was applied with a damping constant of 0. 2 ps−1. In equilibration and production runs the SHAKE algorithm was applied to constrain all bonds involving hydrogen atoms. Electrostatic interactions were modeled using the Particle Mesh Ewald (PME) method [44] and van der Waals interactions were truncated at 12 Å. The production run time-step was set to 2 fs and frames were saved every 2 ps. Prior works have shown that, for comparable systems (15,000—19,000 atoms), 50 ns are enough to equilibrate the NaCl atmosphere around the DNA molecule [1,5]. During the equilibration, the DNA molecules remained in the B-form conformation. A second model included a charged arginine side chain analog, methyl-guanidinium, bound to DNA. This ligand was modeled using the amber ff99SB* force-field [45,46] for the arginine residue. The initial coordinates of the arginine analog were obtained from the conserved -N terminal structure of the Antennapedia homeodomain (pdb access code 9ANT) [26]. Taking advantage of the helical symmetry of the DNA molecules, we constructed a helical coordinates system (ρ, ϕ, ξ) that specifies the relative position of the ligands with respect to a DNA molecule. The DNA’s axis was aligned to the z-axis, such that the ρ and ϕ coordinates are equivalent in magnitude to the r and θ coordinates used in cylindrical coordinates (r, θ, z). ξ is the family of helical surfaces given by ξ = z−pϕ/2π, with p the pitch of the helical system (Fig. 1. A). It is also useful to define the pitch angle α such that tanα = p/2πρ (Fig. 1. A) and p ¯ =p/2π. This coordinate system was initially introduced by Waldron [47] for applications in electromagnetic theory and takes the helix to be right-handed. In helical coordinates, at ρ = constant, the components of a vector V = (Vρ, Vϕ, Vξ), are expressed in the following way, V ϕ = V θ sec α (1) V ξ = V z − V θ tan α (2) through the corresponding cylindrical components (Vr, Vθ, Vz) (Fig. 1. B). It is important to note that the radial component of the vector Vρ is perpendicular to Vϕ and Vξ, but Vϕ and Vξ coordinates are not orthogonal. Analogously, the transformation between helical and cartesian (x, y, z) coordinates systems is given by: ρ = x 2 + y 2 (3) ϕ = tan − 1 (y / x) (4) ξ = z − p ¯ tan − 1 (y / x) (5) For the helical coordinate system we define a covariant basis (eρ, eϕ, eξ) and a contravariant basis (bρ, bϕ, bξ). The scale factors are given by: hρ = 1, hϕ = ρ/ cos α and hξ = 1. The covariant basis vectors are defined as: e ρ = (cos ϕ, sin ϕ, 0) (6) e ϕ = (− cos α sin ϕ, cos α cos ϕ, sin α) (7) e ξ = (0,0, 1) (8) The (non-unit) contravariant basis vectors are given by: b ρ = (cos ϕ, sin ϕ, 0) (9) b ϕ = (− sin ϕ cos α, cos ϕ cos α, 0) (10) b ξ = (tan α sin ϕ ρ, − tan α cos ϕ ρ, 1) (11) Following the definition of the gradient, ∇ Ψ = 1 h i ∂ Ψ ∂ q i b i (12) we expressed the helical coordinates gradient as, ∇ Ψ = b ρ ∂ ρ Ψ + b ϕ cos α ρ ∂ ϕ Ψ + b ξ ∂ ξ Ψ (13) Note that Eq. 13 becomes the gradient in cylindrical coordinates in the limit α → 0. The helical coordinate system described above was implemented in LAMMPS. We used potential of mean force (PMF) umbrella sampling [48] techniques to characterize the free energy surface governing how counterions and charged residues move along the minor groove. To restrain the ligand’s center of mass, we introduced the potential: Ψ (ρ, ϕ, ξ) = k ρ 2 (ρ − ρ 0) 2 + k ϕ 2 (ρ ϕ − ρ 0 ϕ 0) 2 + k ξ 2 (ξ − ξ 0) 2, (14) where ρ, ϕ and ξ correspond to the coordinates of the center of mass, ρ0, ϕ0 and ξ0 are the target positions and kρ, kϕ and kξ the respective force constants. For this potential, the force (F) acting on the center of mass of the ligands were obtained by using the gradient in helical coordinates (Eq. 13), F = −∇Ψ. For each ligand we computed the PMF for one full turn (ϕ = 2π) around the DNA molecule, equivalent to a rotation along 10. 5 base pairs. From the simulated segment, we computed the binding free-energy map for the mid-section (underlined sequence): [d (CGCGAGGTTTAAACCTCGCG) ]2. The Na+ center of mass was constrained using force constants of kρ = 25 kcal/mol/Å2, kϕ = 2. 5 kcal/mol/ (rad2 ⋅Å2) and kξ = 10 kcal/mol/Å2. Umbrella increments were set to 0. 4 for ρ and 0. 25 rad for ϕ, for ρ ranging from 10 to 15 Å and ϕ ranging from 0 to 2π. The methyl-guanidinium center of mass was constrained using forced constants of kρ = 10 kcal/mol/Å2, kϕ = 1 kcal/mol/ (rad2 ⋅Å2) and kξ = 5 kcal/mol/Å2. Umbrella increments were set to 0. 4 Å for ρ, 0. 25 rad for ϕ for ρ ranging from 8. 6 to 12. 8 Å and ϕ ranging from 0 to 2π. Force constants were chosen to ensure the proper overlap of the histograms (see S2 Fig.) and that energy contributions associated with the restraining potential in each of the three coordinates (ρ, ϕ, ξ) are comparable. For Na+, US production runs were 5 ns long for each window. For methyl-guanidinium the US production runs were 8 ns for each window. The 3D-PMF maps were obtained from the US simulations using the WHAM algorithm [49] extended to three dimensions to account the for the helical coordinate system. The PMF’s were calculated considering only the translational degrees of freedom of the ligand’s center of mass restrained in the helical coordinate system (ρ, ϕ, ξ). The rotational degrees of freedom, as well as the DNA’s and solvent’s degrees of freedom were averaged in the PMF. To test the convergence of the binding free-energy estimates, we repeated the WHAM calculations using only fraction of the US trajectories (S3 Fig.). The free-energy errors were estimated as integrated standard deviations of the mean using the bootstrap algorithm [50]. In each case we generated 2000 bootstraps. For visualization purposes, the 3D-PMF’s were projected into 2D-PMF’s by, at every angular position, obtaining the free energies to a “ribbon” passing through the middle of the minor groove’s solvent accessible volume (S1 Fig.). It is important to note that the helical coordinate system is not periodic. This is, for a probe with coordinates (ρ, ϕ, ξ), after a 2π rotation, the coordinates are (ρ, ϕ, ξ + p). The geometry of the DNA molecule was analyzed using the program Curves [51], which allowed us to obtain the width of the minor groove along the studied turn. The electrostatic potential inside the minor groove was determined from the solutions of the non-linear Poisson-Boltzmann equation using the Adaptive Poisson-Boltzmann (APBS) software [52] at 0. 15 M salt concentration. The atomic partial charges and radii were obtain from the Amber force field [40]. A 129 × 129 × 193 grid was used, with a grid spacing of 1. 5 Å. A dielectric value of ϵ = 2. 00 was assigned to the interior of the DNA molecule (calculate with a 1. 4 Å probe sphere), whereas a dielectric value of ϵ = 78. 54 was assigned to the solvent. Boundary condition values were determined using the Debye-Hückel approximation. To compute the correlation between the binding free-energy and the electrostatic potential we first used interpolation [53] to build a grid with a 0. 13 Å spacing, equivalent to the one from the free-energy maps. The Pearson’s correlation coefficient was computed between the maps with equivalent spacing. We identified all of the methyl-guanidinium and DNA’s minor groove hydrogen bond donors and acceptors and recorded all hydrogen bonds formed within every 2 picosecond frame along each trajectory. A geometric definition of a hydrogen bond was used: two heavy atoms are considered to be bonded if (1) their donor-acceptor distance is less than 3. 5 Å and (2) the acceptor-donor-hydrogen angle is less than 60o. Additionally, we computed the radial distribution function (g (r) ) between the Na+ ions and the oxygen atoms of the water molecules as well as the g (r) between the polar head of the methyl-guanidinum (defined as the center of mass of the NE, CZ, NH1, and NH2 atoms) and the oxygen atoms of the water molecules. For Na+ we computed the g (r) of cations localized to the minor groove via the umbrella potentials and for cations in the bulk in all trajectories. The hydration number of these ligands was determined by integrating, over volume, the water molecules in the first hydration shell, which is equivalent to integrating the first peak in the g (r).

Protein-DNA and ion-DNA interactions are key for many essential biological activities such as DNA condensation, replication, transcription and repair. All these processes require DNA-binding proteins to associate tightly to their specific targets, while also being able to find these sites in an efficient way. To facilitate the search of their DNA targets, DNA-binding proteins often first interact non-specifically with DNA molecules and then slide along the DNA. In this paper, we quantitatively describe the energetics of sliding and binding of two small ligands to the DNA’s minor groove. We show that the minor groove geometry shapes the free-energy landscape surrounding the DNA molecule providing heterogeneous binding environment that can be exploited by DNA-binding proteins in this search for their specific recognition sites.

Abstract Introduction Results Discussion Materials and Methods

2015

DNA Exit Ramps Are Revealed in the Binding Landscapes Obtained from Simulations in Helical Coordinates

Theileria parasites invade and transform bovine leukocytes causing either East Coast fever (T. parva), or tropical theileriosis (T. annulata). Susceptible animals usually die within weeks of infection, but indigenous infected cattle show markedly reduced pathology, suggesting that host genetic factors may cause disease susceptibility. Attenuated live vaccines are widely used to control tropical theileriosis and attenuation is associated with reduced invasiveness of infected macrophages in vitro. Disease pathogenesis is therefore linked to aggressive invasiveness, rather than uncontrolled proliferation of Theileria-infected leukocytes. We show that the invasive potential of Theileria-transformed leukocytes involves TGF-b signalling. Attenuated live vaccine lines express reduced TGF-b2 and their invasiveness can be rescued with exogenous TGF-b. Importantly, infected macrophages from disease susceptible Holstein-Friesian (HF) cows express more TGF-b2 and traverse Matrigel with great efficiency compared to those from disease-resistant Sahiwal cattle. Thus, TGF-b2 levels correlate with disease susceptibility. Using fluorescence and time-lapse video microscopy we show that Theileria-infected, disease-susceptible HF macrophages exhibit increased actin dynamics in their lamellipodia and podosomal adhesion structures and develop more membrane blebs. TGF-b2-associated invasiveness in HF macrophages has a transcription-independent element that relies on cytoskeleton remodelling via activation of Rho kinase (ROCK). We propose that a TGF-b autocrine loop confers an amoeboid-like motility on Theileria-infected leukocytes, which combines with MMP-dependent motility to drive invasiveness and virulence. Cellular transformation is a complex, multi-step process and leukocyte transformation by Theileria is no exception, as parasite infection activates several different leukocyte-signalling pathways, the combination of which leads to full host cell transformation [1]. However, Theileria-induced leukocyte transformation is unusual in that it is rapid and appears to be entirely reversible with the host cell losing its transformed phenotype upon drug-induced parasite death [2]. Just like most cancer cells however, Theileria-induced pathogenesis (virulence) is associated with the invasive capacity of transformed leukocytes, which is lost upon attenuation of vaccine lines [3]. Attenuation of virulence has been ascribed to decreased matrix-metallo-proteinase-9 (MMP9) production and loss of AP-1 transcriptional activity [4]. Consistently, functional inactivation of AP-1 resulted in reduced tumour formation, when infected and transformed B cells were injected into Rag2gC mice [5]. Host leukocyte tropism differs with T. parva infecting all subpopulations of lymphocytes whereas T. annulata infects monocytes/macrophages, dendritic cells and B lymphocytes [1]. Despite this, the diseases they cause (called tropical theileriosis with T. annulata infection and East Coast fever with T. parva infection) are both severe, as susceptible animals usually die within three weeks of infection. The geographical distribution of their respective tick vector species determines areas where disease is widespread. Tropical theileriosis affects over 250 million animals and extends over the Mediterranean basin, the Middle East, India and the Far East, whereas East Coast fever is prevalent in eastern, central and southern Africa. It is noteworthy that in endemic areas indigenous breeds of cattle are more resistant to disease. For example, when Bos indicus Sahiwals are experimentally infected with T. annulata they exhibit fewer clinical symptoms and recover from a parasite dose that is fatal in the European Holstein-Friesian (HF) B. taurus breed [6]–[7]. Theileria-infected leukocytes are capable of producing IL-1 and IL-6 [8], as well as GM-CSF [9] and TNF [10]. Nonetheless, no differences in the level of expression of the pro-inflammatory cytokines TNF, IL-1b, or IL-6 were detected between disease-resistant Sahiwal- versus HF-infected macrophages [11]. Some additional inherent genetic trait of Sahiwal animals must therefore underlie their disease-resistance. Although transcriptome analysis of 3–5 times passaged Sahiwal and HF macrophages following infection with T. annulata revealed significant breed differences in both the resting and infected gene expression profiles, no clear candidate genetic trait was revealed [12]. Transforming growth factor beta (TGF-b) is a family of cytokines and both TGF-b1 and TGF-b2 can bind with high affinity to the TGF-b type II receptor (TGF-RII) leading to the recruitment of TGF-RI. The constitutive kinase activity of TGF-RII phosphorylates and activates TGF-RI, which in turns recruits and activates Smad2 and Smad3, which bind Smad4, and the whole complex translocates to the nucleus and induces the transcription of target genes [13]. The TGF-b signalling pathway can be negatively regulated [14] and an increasing number of non-Smad-mediated TGF-b signalling pathways have been described [15]. TGF-b can also regulate cytoskeleton dynamics via transcription-dependent and transcription-independent processes [16]. It is likely that all these different pathways contribute in different ways to the pleiotropic effects of TGF-b (see http: //www. cell. com/enhanced/taylor). TGF-b can exert opposite effects on cell growth: in most non-transformed cells TGF-b is usually growth inhibitory, but it can increase motility of certain mesenchymal cells and monocytes, but however, at some point in the transformation process TGF-b becomes pro-metastatic [17]–[18], for example in ovarian cancer [19]. We show here that TGF-b plays a role in infected host leukocyte invasiveness. Importantly, the high level of TGF-b2 production by Theileria-infected HF-transformed macrophages renders them more invasive than those of disease-resistant Sahiwal animals. In addition, vaccination against tropical theileriosis uses live attenuated T. annulata-infected macrophages and attenuation leads to the loss of both TGF-b2 transcription and alteration in the expression of a set of TGF-b-target genes, and a drop in TGF-b-mediated invasion. Thus, Theileria-dependent TGF-b2 induction is a virulence trait that underscores susceptibility to tropical theileriosis. As Theileria-transformed leukocytes are known to secrete a number of different cytokines we examined whether infection by T. annulata sporozites of the same parasite strain (Hissar) could induce TGF-b in macrophages 72h post-invasion, as described [12]. Prior to infection both Sahiwal and HF macrophages produced low levels of TGF-b transcript with slightly higher amounts of TGF-b1 (Fig. 1A). Interestingly, Theileria infection induces preferentially TGF-b2 in both Sahiwal and HF macrophages, and importantly, the induction after 72h is much greater in disease-susceptible HF cells. We next examined the levels of TGF-b transcripts in a series of T. annulata-transformed cell lines derived from Sahiwal and HF animals. Again, TGF-b1 and TGF-b2 mRNA could be detected in all 10 transformed cell lines (Fig. 1B). Similar to freshly invaded cells the relative mRNA levels of TGF-b1 did not differ significantly across the T. annulata infected cell lines and there was no evidence of a breed-specific difference in TGF-b1 expression (Fig. 1B grey bars, p = 0. 710). In contrast, the relative TGF-b2 mRNA levels exhibit statistically significant differences (Fig. 1B black bars, p<0. 001), with TGF-b2 mRNA levels being higher in HF cell-lines. Additional qRT-PCR experiments revealed that the levels of expression of TGF-RI, TGF-RII and TGF-RIII were equivalent in HF versus Sahiwal T. annulata-infected cell lines (data not shown). Thus, disease susceptibility correlates to the level of TGF-b2 transcripts that are expressed 7-fold (p<0. 001) more by T. annulata-transformed macrophages of HF origin. In T. parva-transformed B cells a TGF-b-mediated signalling pathway is active and invasion is partially TGF-b-dependent (Fig. S1). We therefore compared the invasive capacity of T. annulata-transformed HF versus Sahiwal macrophages (Figure 2). We found that disease-susceptible HF macrophages displayed 30% greater capacity (p<0. 005) to traverse Matrigel than infected Sahiwal macrophages and that traversal is again TGF-b-dependent (Fig. 2A). The invasive capacity of the H7 cell line was reduced to levels equivalent to S3 cells upon treatment with the TGF-R inhibitor (Fig. 2A) and conversely, when S3 cells were stimulated with conditioned medium from H7 cultures, S3 cells displayed increased invasiveness (Fig. 2B). Moreover, the reduced invasive capacity of disease-resistant S3 macrophages could be restored to above virulent levels by addition of either TGF-b1, or TGF-b2 (Fig. 2C), demonstrating that the TGF-b signalling pathways are intact in these cells. Theileria infection therefore preferentially induces up-regulation of TGF-b2 and increased invasiveness of transformed leukocytes. T. annulata-infected cell lines can be attenuated for virulence by multiple in vitro passages to generate live vaccines that are used to protect against tropical theileriosis [3]. The molecular basis of attenuation is not known, but our above observations on preferential TGF-b2 induction and augmented host cell invasiveness suggest that attenuation might be lead to reduced TGF-b2 transcription and TGF-b2-mediated invasion. To directly test this prediction we examined the Ode vaccine line derived from an infected HF cow in India [20] and estimated the level of TGF-b2 transcripts and TGF-b-target gene transcription in virulent (early passage) and attenuated (late passage) infected macrophages (Figure 3). As predicted, attenuation leads to a significant decrease in the amount of TGF-b2 message and surprisingly, a slight increase in TGF-b1 transcripts (Fig. 3A). This strongly suggests that upon attenuation the parasite' s ability to induce host cell TGF-b2 has been impaired. Reduced levels of TGF-b2 should lead to an alteration in the transcription profiles of known TGF-b-target genes and to see if this is indeed the case, we performed microarray analyses and hierarchical clustering of transcript levels. The microarray representing 26,751 bovine genes included 1,158 targets of TGF-b (http: //www. netpath. org/) and 76 of these genes were identified as differentially expressed upon attenuation. The heat-map, where low gene expression level is depicted as blue, intermediate as yellow and high expression as red, is present in Fig. 3B. Among the down-regulated TGF-target genes, five were chosen at random and their expression verified by qRT-PCR using mRNA from early and late passage Ode (Fig. 3C). In each case their transcription was reduced upon attenuation and could be restored by adding exogenous recombinant TGF-b2. Consistently, their expression was high in disease-susceptible infected HF macrophages that produce more TGF-b2 (see Fig. 1) and low in Sahiwal macrophages, but could be augmented by exogenous TGF-b2 stimulation (Fig. S2). Attenuation of virulence therefore, leads to ablation of TGF-b2 signalling and an alteration in the profile of expression of a set of TGF-target genes. The observation that early passage Ode cells express higher levels of TGF-b2 message and have altered expression of 76 TGF-b-target genes led us to compare early with late passage Ode and examine the contribution of TGF-b to their invasive capacity (Figure 4). As previously described [4], attenuation of Ode leads to a significant drop in invasive capacity (***p<0. 005) and receptor blockade by SB431542 gives an estimate (***p<0. 005) of the contribution of TGF-b to early passage Ode virulence (Fig. 4A). The potential contribution of TGF-b to virulence has been ablated by attenuation, since the invasive capacity late passage Ode is insensitive to receptor blockade (Fig. 4A, right). When conditioned medium from early passage Ode is given to late passage Ode there is a marked (**p<0. 05) regain in invasion (Fig. 4B). Virulent Ode therefore, secretes factors into the medium that contribute to invasiveness, one of which is clearly TGF-b2, and this capacity is lost upon attenuation. The partial inhibition of invasion by early passage Ode by SB431542 might also suggest that although virulent following 65/70 in vitro passages some attenuation of TGF-b2 induction might be occurring. We next studied whether the TGF-b-mediated invasion programme might have a consequence on cellular adhesion and invasion structures such as lamellipodia, podosomes and membrane blebs. We first investigated by time-lapse video microscopy lamellipodia morphology of T. annulata-infected macrophages cultured without (control), or with SB431542 (Fig. S3 and Movies S1 and S2) and found that the size of lamellipodia (shown boxed) decreased upon SB431542 treatment (Fig. S3A). Visualisation of the actin cytoskeleton with Texas red-labelled phalloidin showed that decreased lamellipodia size correlated with reduced actin dynamics (Fig. S3B and C) and suggested that TGF-controls cortical actin dynamics in infected macrophages. We next compared the basal and central cortical actin cytoskeleton of S3 and H7 cells cultured on a gelatin/fibronectin matrix, which facilitates adhesion of these cells. In S3 cells, we observed only small podosomal adhesion structures that were rarely clustered and no actin-rich membrane blebs (Fig. 5A). In contrast, in H7 cells podosomal adhesion structures were markedly enlarged and clustered and the majority of cells displayed actin-rich membrane blebs. Membrane blebbing was confirmed by live-cell imaging (Fig. S4A), which highlights the dynamics of bleb formation. Individual blebs expand within one to three seconds and persist for approximately 30–120 seconds, which is a time frame typically observed in bleb formation and retraction [21]. Reducing the serum concentration from 10% to 0. 5% (starvation) resulted in a significant decrease in the number of membrane blebs (Fig. 5B and C). Membrane blebbing in starved cells was rescued by the exogenous addition of TGF-b2. The TGF-b-induced membrane blebs were blocked by the Rho-kinase (ROCK) inhibitor H-1152 and in the presence of serum the formation of actin-rich membrane blebs was significantly reduced after treatment with the TGF-R inhibitor, but completely blocked in the presence of H-1152 (Fig. 5C). Thus, one consequence of increased TGF-b2 production is increased cortical actin dynamics, which likely gives rise to enhanced podosomal adhesion structures (invadosomes) and membrane blebs in macrophages derived from disease susceptible HF cattle. We next investigated the functional significance of membrane blebs for cell motility in 3-D matrices. In the low rigidity fibrillar collagen or high-density Matrigel matrices, H7 macrophages acquired an amoeboid pattern of motility with characteristic polarized formation of membrane blebs (Fig. 5D and Fig. S4B and C). Membrane bleb formation required ROCK activity and inhibition of ROCK prevents local contractibility, polarized bleb formation and forward movement of the cell. Taken together, these data show that exposure of H7 macrophages to TGF-b2, results in ROCK-dependent membrane blebbing that drives motility in 3-D matrices. We have shown here that Theileria-induced leukocyte transformation results in the constitutive induction of a TGF-b autocrine loop that augments the invasive potential of infected leukocytes. We could find no evidence for a contribution of TGF-b signalling to host cell survival, or proliferation (data not shown), implying that leukocyte infection by Theileria essentially confers on TGF-b a pro-metastatic role. Smads and p53 are known to associate and collaborate in the induction of a subset of TGF-b target genes [22]. Recently, p53 has been described as being sequestered in the cytosol of Theileria-transformed leukocytes, as part of a parasite-induced survival mechanism [23]. Although not addressed by Haller et al it is possible that cytosol located p53 might ablate nuclear translocation of Smads, thus counteracting the anti-proliferative affect of TGF-b in Theileria-transformed leukocytes? Importantly, comparison of disease-susceptible HF transformed macrophages to disease-resistant Sahiwal ones, showed that the degree to which Theileria (the same parasite strain) induces TGF-b2 influences the invasive potential of the infected and transformed host cell. The likelihood of developing a life-threatening cancer-like disease therefore appears to be due in part to the inherent genetic propensity of HF macrophages to produce high levels of TGF-b2 upon infection that might render the transformed macrophages more invasive. T. annulata-transformation of HF macrophages leads to the induction of higher amounts of TGF-b2 the levels of induced TGF-b1 mRNA being the same as in Sahiwal macrophages. As TGF-RI and -RII and -RIII [24] are expressed to the same extent in the two types of macrophages (data not shown) it suggests that only the amount of TGF-b2 is crucial. The predisposition of Theileria transformation to induce TGF-b2 over TGF-b1 in HF versus Sahiwal macrophages implies that there could be disease-associated sensitivity to infection linked to TGF-b2 over-production. Species-specific promoter differences, or some other unknown breed difference may explain the greater propensity of Theileria to induce TGF-b2 over TGF-b1 transcripts in HF compared to Sahiwal cattle. However, promoter sequence differences seem unlikely to underlie Theileria' s ability to induce TGF-b2 over TGF-b1 transcripts, or explain the drop in TGF-b2 levels upon attenuation of the Ode vaccine line as here, both virulent and attenuated infected macrophages are of HF origin [20]. Loss of virulence of the Ode vaccine line upon long-term in vitro passage is clearly associated with a decrease in TGF-b2 transcripts and it would appear that the parasite' s ability to specifically activate host cell transcription of TGF-b2 is impaired and one possibility is that attenuation is associated with altered epigenetic regulation of TGF-b2 promoter activity. Microarray analyses indicate that upon attenuation of virulence, not only do TGF-b2 levels drop, but also 76 TGF-b-target genes display altered transcription. It would appear then that preferential TGF-b2 induction following Theileria infection initiates a host cell genetic programme that contributes to more aggressive invasiveness of transformed HF macrophages. We believe the same TGF-b2-initiated genetic programme also contributes to the invasiveness, albeit reduced, of disease-resistant Sahiwal macrophages, as Theileria infection also preferentially induces TGF-b2, just to a lesser extent. We have used fluorescence and time-lapse video microscopy to examine the morphology of Theileria-infected Sahiwal and HF infected macrophages and the effect of TGF-b and Rho kinase (ROCK) on actin dynamics and lamellipodia formation. Theileria infected, disease-susceptible HF macrophages show increased actin dynamics in their lamellipodia and podosomal adhesion structures and a remarkable propensity to develop membrane blebs. Figure 5D shows the dynamic behaviour of infected cells embedded in 3-D matrices (see also Fig. S3 and movies S3 and S4), where either fibrillar collagen, or matrigel was used giving two 3-D matrices of low (collagen) and high (matrigel) rigidity. Membrane blebbing of motile H7 cells occurs in both matrices in a polarized fashion at the leading edge, clearly suggesting that membrane blebbing is required for infected cell motility in 3-D. Movie S3 shows bleb-driven membrane protrusions, which results in forward movement of the cell (see also kymographs of movie S3, Fig. 5D). Moreover, ROCK activity is required, because inhibition of ROCK with H-1152 impairs polarized bleb formation and forward movement in fibrillar collagen and matrigel. Spatio-temporal control of Rho-ROCK activity is also required for cell polarization and lamellipodia formation in 2-D [25]. Since TGF-b acts upstream of Rho-ROCK in infected cells, we would therefore predict that spatio-temporal activation of Rho-ROCK – controlled by TGF-b signalling – is required for lamellipodia formation as well. Combined, increased bleb and lamellipodia formation could give rise to more invadosomes on infected virulent macrophages in a similar manner to TGF-b-mediated increased adhesion of immortalised hepatocyte cell lines [26]. We showed that augmented invasiveness by TGF-b2 in disease-susceptible HF macrophages has a transcription-independent element that relies on cytoskeleton remodelling via activation of Rho and its downstream target ROCK [27]–[28]. Given the important role played by ROCK in increased blebbing of infected host cells and the recent description that Rho/ROCK signals amoeboid-like motility [29], it is tempting to speculate that the TGF-b autocrine loop confers on Theileria-infected leukocytes an amoeboid-like motility that contributes to invasiveness and makes them more virulent. However, since prolonged TGF-b stimulation can result in decreased Rho activity due to the action of p190RhoGap [30], Theileria infected cells must have developed a mechanism to balance TGF-b-induced RhoGAP activities. It is possible that the parasite might also regulate the expression, activity or localization of Rho-family GEFs, as the Rho activator GEF-H1/Lfc has been shown to be a TGF-b1 target gene [31] and an analogous mechanism involving TGF-b2 might be exploited by Theileria parasites? Alternatively, the parasite could function by excluding negative regulators of Rho from specific subcellular compartments, analogous to the exclusion of the RhoGAP Myo9b from lamellipodia of macrophages [25]. Whatever the underlying mechanism of inducing TGF-b2 levels in Theileria-infected leukocytes, its induction and the genetic programme it initiates is clearly correlated with the invasive phenotype of transformed macrophages of disease-sensitive hosts. This implies that overall invasiveness of Theileria-transformed leukocytes is made up of amoeboid (TGF-b- & ROCK-dependent) and mixed amoeboid/proteolytic (MMP-dependent [4]) motility; reviewed in [32]. These Theileria-based observations also suggest that in some cases the propensity of human leukaemia patients to develop life-threatening cancer could be due to the inherent genetic predisposition of their tumours to produce high levels of TGF-b2, rather than TGF-b1, and the genetic programme this initiates on promoting an additional amoeboid-like invasive phenotype of their tumours. TpMD409. B2 is a T. parva Muguga-infected B-cell clone (B2) and its B-cell characteristics have been previously confirmed [33]. The cell lines S1–S5 and H7–H10 have been described previously [6]. In vitro infection of uninfected S and H cells by Theileria sporozoites was done as described [12]. The characterisation of the Ode vaccine line has been reported [20] and in this study virulent/early Ode corresponds to passage 65–70 and attenuated to passage 318–322. It is possible that passages 65–70 have already become slightly attenuated. All cultures were maintained in RPMI-1640 medium supplemented with 10% foetal calf serum (FCS) and 50 uM b-mercaptoethanol. Cell cultures were passaged 24h before harvesting to maintain the cells in the exponential growth phase. The TGF-bRI/ALK5 inhibitor SB431542 (Sigma #S4317) was added at 10uM for 18h. The Rho kinase (ROCK) inhibitor H-1152 (Alexis Biochemicals, #ALX-270-423) was added at 10uM. Recombinant bovine TGF-b1 and TGF-b2 (rboTGF-b1 and rboTGF-b2; both NIBSC, Potters Bar. UK) were added in the culture media at 10ng/ml and incubated for different times (15 min or 30 min). All transfections were carried out by electroporation as previously described [34]. The CAGA-luc (the Smad 3/4 binding element-luciferase construct) was transiently transfected into B2 cells with the inhibitory Smad7 plasmid (Flag-Smad7-pcdef3), or an empty vector (pcdef3). The major late minimal promoter (MLP - a minimal promoter consisting of the TATA box and the initiator sequence of the adenovirus major late promoter) was cloned into pGL3 (Promega) to generate the MLP-luc plasmid that was used as minimal promoter negative control [35]–[36]. Efficiencies of transfections were normalized to renilla activity by co-transfection of a pRL-TK renilla reporter plasmid (Promega #E6241). Luciferase assay was performed 24h after transfection using the Dual-Luciferase Reporter Assay System (Promega, #E1980) in a microplate luminometer (Centro LB 960, Berthold). Relative luminescence was represented as the ratio firefly/renilla luminescence. Nuclear extracts of from T. parva-infected (B2) cells were prepared as described [37]. 20ug of proteins were separated in a denaturing 8% SDS-PAGE gel and electro-transferred onto a nitrocellulose membrane (Scheicher and Schuell). Antibodies used in immuno-blotting were as follows: anti-phospho-Smad2 (Cell Signaling #3101), anti-Smad2 (BioVision, #3462-100), anti-PARP (Clone C2-10, Pharmingen #556362). Total RNA was isolated from each of the T. annulata infected cell lines using the RNeasy mini kit (Qiagen) according to the manufacturer' s instructions. The quality and quantity of the resulting RNA was determined using a Nanodrop spectrophotometer and gel electrophoresis and for microarray screens on an Agilent 2100 Bioanalyser (Agilent Technologies). mRNA was reverse transcribed to first-strand cDNA and the relative levels of each transcript were quantified by real-time PCR using SYBR Green detection. The detection of a single product was verified by dissociation curve analysis and relative quantities of mRNA calculated using the method described by [38]. GAPDH, HPRT1 and C13orf8 relative quantities were used to normalise mRNA levels. For list of primers used, see Table 1. Host gene expression was investigated using a custom-designed microarray (Roche NimbleGen Inc. , Madison, WI), which represented every bovine RNA reference sequence currently deposited in the NCBI database (n = 26,751). Each gene was represented by two identical sets of five 60-mer-oligonucleotide probes that were isothermal with respect to melting temperature. cDNA was generated from 10 ug total RNA using oligo (dT) primer and tagged with 3′-Cy3 dye and labelled cDNA was hybridised to the array. Gene expression values were calculated from an RMA-normalised dataset [39] and differentially expressed genes were identified using Rank Product Analysis [40]. Genes were defined as differentially expressed using the criteria of a false discovery rate of less than 5% and a fold change of greater than two. Selected gene sets were subjected to hierarchical clustering based on Euclidean distance between expression values and the results were illustrated using a heat-map (ArrayStar3, DNASTAR Inc. , Madison, WI). The invasive capacity of the bovine leukocytes was assessed using in vitro Matrigel migration chambers, as described [5]. After 26h of incubation at 37°C, filters were washed twice in PBS and the Matrigel was eliminated. In some cases, during this 12h period cells were also incubated with the TGF-b inhibitor SB431542 (10uM), or with recombinant TGF-b protein (10ng/ml). When added TGF-b was maintained in the top chamber, meaning that overall cells were incubated with TGF-b for 36h. Cells were then counted under the microscope (40× objective) to obtain a statistically significant mean number of cells per field (at least 10 fields per filter). The experiment was performed at least in triplicate. Time-lapse imaging using video microscopy was performed with cells growing on glass bottom culture dishes (Willco Wells, the Netherlands) using a Nikon Eclipse TE2000-U inverted microscope equipped with a climate-controlled chamber. Data acquisition and image processing was performed using NIS software of Nikon Instruments. DIC images were acquired in intervals ranging from 0. 5 msec to 1 min for 30 min and assembled in movies; acceleration of movies is approximately 600-fold. Kymographs were acquired along a one pixel wide line using NIS software. A 40× Plan Achromat Objective of Nikon was used for longworking distance image acquisition in Matrigel. Theileria-infected macrophages were seeded onto glass coverslips, or glass bottom culture dishes (Willco Wells, the Netherlands) and maintained in growth medium for 48h without or with 10uM SB431542 (SIGMA). SB431542 was replenished after 24h. Cells were either processed for live cell imaging or fixed in 3. 5% formaldehyde, 15 min. Actin cytoskeletons of fixed and Triton X-100 permeabilized cells were visualized with Texas red-labelled phalloidin. Lamellipodia area and integrated fluorescence intensities in lamellipodia were determined using photoshop CS3 software. Glass coverslips were coated with 0. 1% poly-L-lysine for 15 min at room temperature and then fixed with 0. 5% glutharaldehyde for 15min. After 3 washes with PBS, coverslips were inverted onto droplet containing 2mg/ml (0. 2%) gelatin (MERCK) in H2O for 15min. After 3 washes with PBS, coverslips were inverted onto droplet containing 25ug/ml bovine plasma fibronectin (SIGMA) in PBS and then incubated 1h at room temperature. Coverslips were then transferred into 24 well plate washed once with PBS and kept in PBS until seeding of 50,000 cells per well. After 18h cells were fixed in 3. 5% formaldehyde, 15min. Embedding in collagen: 3×105 cells in 50ul medium were added to a mixture of 68µl sodium bicarbonate (7. 5%, SIGMA), 240ul 10× PBS and 2ml PureCol (3mg/ml, Inamed). The resulting gelatin solution with the concentration of 2. 4mg gelatin/ml was distributed into live-cell imaging wells or dishes and transferred to 37°C for collagen polymerization. Embedding in Matrigel: 1×105 cells in medium were mixed on ice with 250ul growth factor reduced Matrigel (BD biosciences) and was distributed into live-cell imaging wells or dishes and transferred to 37°C for polymerization.

Theileria annulata causes tropical theileriosis that is endemic in cattle in North Africa, the Middle East, India and China. T. parva causes East Coast fever that is prevalent in East and Southern Africa. In endemic countries indigenous cattle are more resistant to pathology, but produce little meat and milk and attempts to improve output by importing European and American breeds have failed due to a high susceptibility to these diseases that are often rapidly fatal. We examined T. annulata-transformed macrophages isolated from disease resistant Sahiwal compared to disease-susceptible Holstein-Friesian (HF) cattle, for their capacity to traverse synthetic extra-cellular matrix in vitro. The invasive capacity of all transformed macrophages was TGF-b dependent, but those of disease-susceptible HF animals invaded better i. e. they were more aggressive. The greater invasive capacity of HF transformed macrophages matched their increased production of TGF-b2, since levels of TGF-b1, and all three TGF-b receptors, were the same as in transformed macrophages isolated from disease-resistant Sahiwal animals. TGF-b2 production therefore likely renders Theileria-transformed leukocytes more pathogenic and consistently, in a live attenuated line used to vaccinate against tropical theileriosis transcripts of TGF-b2 and those of a significant number of TGF-target genes drop and consequently, TGF-b-mediated invasiveness decreases.

Abstract Introduction Results Discussion Materials and Methods

infectious diseases/protozoal infections cell biology/extra-cellular matrix cell biology/leukocyte signaling and gene expression

2010

TGF-b2 Induction Regulates Invasiveness of Theileria-Transformed Leukocytes and Disease Susceptibility

No-go Decay (NGD) is a process that has evolved to deal with stalled ribosomes resulting from structural blocks or aberrant mRNAs. The process is distinguished by an endonucleolytic cleavage prior to degradation of the transcript. While many of the details of the pathway have been described, the identity of the endonuclease remains unknown. Here we identify residues of the small subunit ribosomal protein Rps3 that are important for NGD by affecting the cleavage reaction. Mutation of residues within the ribosomal entry tunnel that contact the incoming mRNA leads to significantly reduced accumulation of cleavage products, independent of the type of stall sequence, and renders cells sensitive to damaging agents thought to trigger NGD. These phenotypes are distinct from those seen in combination with other NGD factors, suggesting a separate role for Rps3 in NGD. Conversely, ribosomal proteins ubiquitination is not affected by rps3 mutations, indicating that upstream ribosome quality control (RQC) events are not dependent on these residues. Together, these results suggest that Rps3 is important for quality control on the ribosome and strongly supports the notion that the ribosome itself plays a central role in the endonucleolytic cleavage reaction during NGD. The elongation phase of translation is an imperfect process, during which the ribosome moves with irregular speed along the mRNA template [1]. By and large the elongation speed is determined by sequence and structural features of the coding sequence. For instance, the identity of the A-site codon is known to have a drastic effect on the rate of protein synthesis depending on the availability of its partner tRNA and the nature of the codon-anticodon base-pairing interaction [2,3]. Furthermore, the chemical characteristics of the locally-encoded amino acids have been shown to regulate the rate of protein synthesis based on the manner they interact with the exit tunnel of the ribosome [4]. mRNAs are also known to harbor local secondary structures that can slow down the ribosome as it unwinds them [5,6]. Regardless of the underlying mechanism, the fluctuating rate of protein synthesis along an mRNA molecule appears to serve important biological functions such as promoting appropriate co-translational protein folding and ensuring that the encoded protein is targeted to the correct destination in the cell [7–12]. In contrast to this “programmed” regulation of ribosome traffic, the ribosome often encounters unwanted obstacles that severely hinder its progression and in some cases stall protein synthesis all together [13,14]. Most of these impediments are typically associated with defects in the mRNA, including stable secondary structures, stretches of rare and inhibitory codons, as well as truncations and chemical damage [3,15–17], [18]. Because multiple ribosomes are typically translating a single mRNA at any given point, one stalled ribosome is likely to impede the progression of multiple upstream ribosomes. As a result, if left unresolved, these stalling events have the potential to severely reduce cellular fitness. Notably, the stalling of the ribosome itself is not such a detriment to the cell as is the loss of valuable ribosomes from the translating net pool [13,14]. In eukaryotes, the evolutionary solution to this predicament was No-Go Decay (NGD) [15] as a means to dissociate stalled ribosomes [19–21]. It is thought that over time, this mechanism was expanded on to include mRNA surveillance to dispose of the aberrant mRNA. In particular, the mRNA undergoes an endonucleolytic cleavage upstream of the stall site. The resulting deadenylated 5’-end and uncapped 3’-end pieces are then rapidly degraded by the exosome and Xrn1, respectively [3,15–17]. Initial studies on NGD in yeast focused on the two factors Dom34 (Pelota in mammals) and Hbs1 [15,22,23]. These factors are homologs of the termination factors eRF1 and eRF3, respectively. Early reports of NGD hinted at a role for the factors in mediating the endonucleolytic cleavage of the mRNA near the stalled ribosome [15,24]. However, later studies by the same group and others showed the cleavage to take place in the absence of the factors [22] leaving the question of the role of the factors in the process unanswered. Interestingly prior to the discovery of NGD, genetics studies suggested that Dom34 and Hbs1 are important in maintaining ribosome homeostasis of the cell [25]. To this end, both factors become essential or near-essential when ribosomes are depleted either by knocking down certain ribosomal proteins or under conditions when ribosomes are sequestered [25–27]. These observations are consistent with biochemical studies using a yeast translation reconstituted system, which showed the factors to be responsible for dissociating ribosomes into their respective subunits [18]. This splitting activity of Dom34-Hbs1 was also found to be much more efficient in the presence of Rli1 (ABCE1 in mammals) [20,21]. In vivo data also supported this model for the role of the three factors in dissociating ribosomes [16]. Hence, this rescuing/recycling activity of these factors rationalizes the effect of their deletion on ribosome availability, especially under stress conditions. In addition to ribosome rescue and degradation of the aberrant RNA, NGD is closely linked to a newly discovered protein-quality-control process termed ribosome quality control (RQC). This process is responsible for degrading the incomplete nascent protein resulting from stalled translation [28–34]. RQC proceeds after the splitting action of Dom34-Hbs1-Rli1, which results in a peptidyl-tRNA-associated large-ribosome subunit. This atypical form of the 60S subunit is recognized by the E3 ligase Ltn1 (Listerin in mammals) alongside Rqc2 (formerly Tae2) [30,33,35]. Ltn1 ligates ubiquitin chains to the nascent peptide as it is attached to the tRNA on the large subunit. The ubiquitinated nascent peptide is then extracted and delivered to the proteasome for degradation through the action of Rqc2 and Cdc48 (and its adaptor proteins Ufd1 and Npl4). Two additional factors, the ribosome-associated Asc1 and the E3 ligase Hel2 (Rack1 and Znf598 in mammals, respectively), also appear to be important for proper RQC function. Both factors are important for ribosomal protein ubiquitination and appear to play a role during stalling [36,37]. In particular, deletion of either factor results in increased readthrough of stall sequences [38,39]. How regulatory ribosomal protein ubiquitination interconnects with RQC and NGD is currently poorly understood. Even though the consequences of ribosome stalling in eukaryotes was initially described in the context of its impact on mRNA steady state levels [15], as detailed above we know far more about its entanglement with ribosome rescue and quality control of the associated nascent peptide. More specifically, degradation of the mRNA is initiated by endonucleolytic cleavage, but the identity of the endonuclease remains elusive. This in turn has precluded further critical mechanistic dissections of NGD. Some of these outstanding important questions are: 1) How does the endonuclease recognize stalled ribosomes? 2) Is it associated with the ribosome? 3) Does it have a specificity for certain mRNAs 4) How is its function activated? 5) Can NGD be used to regulate gene expression? Work from our group recently provided some clues about the cleavage reaction. Using reporters and genetic manipulation of yeast we showed that the physical act of ribosome collision is important for initiating the process of RNA degradation and ribosome rescue during no-go decay (NGD) [40]. High-resolution mapping of the cleavage products also provided some important clues about the potential role of the ribosome in the reaction. Namely, cleavage appears to take place well upstream of the lead stalled ribosome with the closest most prominent one being ~45 nt upstream of the stall site. As ribosomes are likely to be stacked on the mRNA, this suggested the possibility that the cleavage is taking place inside the ribosome [18,41]. Multiple regions of the ribosome make intimate contact with the mRNA. Most noteworthy among these is the mRNA entry tunnel, which encompasses residues of the ribosomal proteins Rps3/uS3 and Rps2/uS5 [42]. In eukaryotes additional contacts are made by helices 18 and 14 of the 18S rRNA, whereas in bacteria these contacts are carried out by Rps4/uS4 (orthologous to Rps9 in yeast and humans) [42–44]. In the entry tunnel, Rps3’s contacts with the mRNA stand out because they appear to be almost universally conserved and form an integral part of the helicase domain of the ribosome [42]. Furthermore, the protein has been implicated in translation initiation during the rearrangement of the small subunit that allows for the opening of the ribosomal mRNA binding channel and subsequent scanning of the mRNA [45] as well as start-codon selection [46]. Here we show the entry tunnel of the ribosome to play an important role during NGD. Mutation of the residues of RPS3 that form part of the entry tunnel, which have also been implicated in the helicase activity of the ribosome, were found to significantly reduce the accumulation of cleavage products. This effect on cleavage efficiency to a large extent was independent of the identity of the stall site. Combining these mutations with factors involved in other aspects of NGD revealed that the entry tunnel is also likely to be important in ribosome rescue. Our findings provide new insights into how quality control mechanisms evolved to integrate into fundamental biological machines. To address a potential role for Rps3 in the cleavage reaction, we introduced a number of mutations to the protein and assessed their effect on cleavage of stalling reporters. Our choice of residues for the mutations was motivated by three criteria: they had to be conserved, made intimate contacts with the mRNA and have basic or acidic side chains (Fig 1A and 1B). This led us to Arg116 (R116) and Arg117 (R117). In addition to these, we also analyzed two residues that have been suggested to be important for Rps3’s extra-ribosomal activity in DNA repair [47–51], Asp154 (D154) and Lys200 (K200). Mutation of these residues abolishes the 8-oxoguanosine glycoslase and AP/endonuclease activities of the protein [51]. The variant-yeast strains were generated by introducing mutations to the chromosomal copy of RPS3 (see Methods) in different backgrounds of deletions and mutations. All in all, we generated the following mutants: Arg116 and Arg117 were substituted by Ala residues (R116A/R117A), Asp154 was substituted by an Ala residue (D154A), Lys200 was substituted by an Asn residue (K200N) and finally we generated a double mutant D154A/K200N. Of these the R116A/R117A mutation was notable as the side chain of these residues are projected into the entry tunnel of the ribosome and make electrostatic interactions with the mRNA (Fig 1B). Next, we assessed the effect of these mutations on the cleavage of NGD substrates. We initially used an NGD reporter, which harbors a stable stem loop in the PGK1 coding sequence and was originally designed by Parker and colleagues. The stem loop presents a robust obstacle for the ribosome and is subject to an endonucleolytic cleavage as evidenced by the accumulation of 5’ and 3’ fragments when the exosome and Xrn1 are inactivated, respectively [15]. Indeed, similar to what was observed by us and others [15,16,40], in the ski2Δ strain- which is defective for 3’-5’ mRNA degradation- northern analysis of cells expressing PGK1-SL revealed substantial accumulation of 5’-fragments (Fig 1C). The D154A and K200N mutations in RPS3, which have been suggested to be important for an AP endonuclease activity [51], had no observable effect on the cleavage efficiency and appear to play no role in NGD. In contrast, the R116A/R117A mutations appear to reduce the accumulation of cleavage fragments and also increased heterogeneity among these products (Fig 1C). Interestingly, the mutations also appear to affect the steady-state levels of endogenous PGK1 transcript (Fig 1C). Regardless, these observations suggest that residues of Rps3 that interact with the mRNA in the entry tunnel are important during NGD. The effects of the R116A/R117A mutations on the cleavage reaction were further studied in the context of other deletions that alter different aspects of NGD. Namely, we introduced these mutations into dom34Δ and xrn1Δ strains in addition to the wild-type parent strain. As expected, expression of the PGK1-SL in these strains does not result in the accumulation of 5’-fragments and the R116A/R117A mutations have no effect. As a control, these fragments were seen in the ski2Δ background and the rps3 mutations significantly reduced their levels (Fig 1D). Production of the 3’-fragments, as expected, was seen in the absence of XRN1 and their levels diminished in the presence of the RPS3 mutations, albeit to a lower extent than that seen for the 5’ fragments (Fig 1E). These latter observations suggested that the R116A/R117A mutations do not completely inhibit cleavage and that they may affect other aspects of NGD. To provide further support for a role for the entry tunnel residues of Rps3 during NGD, we next examined the effect of the mutations on the stability of the PGK1-SL mRNA. Our reporters are expressed under the control of the GAL1 promoter, and as a result transcriptional-shutoff by shifting cells to glucose-containing media was used to measure the decay rate of the reporter mRNAs. As a control, we initially measured the decay rate of a non-NGD reporter (PGK1), which does not harbor any stalling sequence. The mutations were found to have little effect on the decay rate of the PGK1 mRNA reporter (Fig 2A); we measured half-lives of 28 ± 1. 9 and 26 ± 4. 4 minutes in the WT and the RPS3-mutant stains, respectively. As expected, the PGK1-SL mRNA decays with a faster rate relative to its PGK1 parent (Fig 2B). Its half-life of 4. 7 ± 0. 2 minutes is similar to previously published reports [15]. Here the RPS3 mutations result in a moderate but reproducible increase in reporter half-life to 6. 0 ± 1. 3 minutes, suggesting greater stabilization of the PGK1-SL mRNA (Fig 2B). Hence, these findings add additional support for the entry tunnel of the ribosome playing a role in mRNA-surveillance during NGD, whereby loss of interactions with the mRNA leads to stabilization of mRNAs harboring stalls. So far, our analysis has focused on one type of stall—a stable RNA secondary structure in the form of a stem loop. Since the mutations under investigation here are important for the helicase function of the ribosome, any effect we saw on the cleavage reaction could be explained by defects in the unwinding activity of the ribosome and not in NGD. To rule out this potential explanation, we used two other reporters that had 12 stretches of the inhibitory arginine CGA or lysine AAA codons. Both are known to efficiently block translation and are not predicted to form secondary structures [3,15,23]. These new reporters were introduced to wild-type or mutant RPS3 yeast strains in the ski2Δ background. As expected, the CGA and AAA reporters accumulated 5’-fragments in the wild-type RPS3 strain, whereas the control UUU reporter did not (Fig 3). Similar to what we observed for the SL reporter, the R116A/R117A mutations significantly reduced the 5’-fragments levels for the CGA and AAA reporters, suggesting that the entry tunnel residues affect the accumulation of cleavage fragments independent of the type of stall (Fig 3). Interestingly, however, unlike the SL reporter, for which we observe an almost complete loss of cleavage products when RPS3 was mutated, cleavage fragments resulting from the CGA and AAA reporters were still visible but instead were heterogeneous in nature (Fig 3). This also made it difficult to perform any meaningful quantification. This is likely due to cleavage fragments produced by inefficient initial cleavage reactions, which lead to ribosome queuing upstream of the lead stalled ribosome. Ski7, a component of the exosome in yeast, has been implicated in non-stop decay (NSD) [52–54]; given the similarities between NSD and NGD, the mutations in RPS3 could potentially affect the function of Ski7. To address this possibility, we deleted SKI7 from the wild-type, dom34Δ and ski2Δ stains in the absence and presence of the RPS3 mutations and assessed its effect on NGD efficiency from the SL reporter. We observed no significant changes to the accumulation of the 5’-fragments due to the SKI7 deletion suggesting that the entry tunnel residues do not affect the function of the factor (S1 Fig). As mentioned earlier, in addition to Rps3, the mRNA entry tunnel of the small subunit also encompasses conserved residues of the ribosomal protein Rps2 [42,55]. Namely the side-chain of Glu120 of the yeast protein protrudes into the entry tunnel and is likely to interact with the mRNA downstream of the A site (S1 Fig). Consequently, we determined whether this residue contributes to NGD or not. We mutated Glu120 to Ala in the ski2Δ strain and evaluated its effect on NGD cleavage efficiency. In contrast to the RPS3 mutations, the RPS2 mutation had no noticeable effect on the cleavage reaction; we observed comparable levels of 5’-fragments accumulation from the SL reporters in the RPS2 wild-type and mutant strains (S1 Fig). It thus appears that the changes to NGD we observe in the presence of the RPS3 mutations are the result of Rps3-dependent effects, and likely not from general alterations to the mRNA-entry tunnel. Initial reports of NGD suggested that Dom34 plays a role in the cleavage reaction due to the loss of the cleavage products accumulation when the factor is deleted [15,24]. Later studies, however, showed that the protein together with Hbs1 and ABCE1 dissociates stalled ribosomes [19]. In its absence ribosomes pile up on the mRNA leading to multiple cleavage events upstream of the lead stalled ribosome, which run as a long smear on a gel that appears to result in loss of cleavage efficiency [16]. Furthermore, overexpression of certain ribosomal proteins restored cleavage in the absence of DOM34, suggesting that the protein is involved in maintaining ribosome homeostasis [22]. To gain further insights into the role of the entry-tunnel residues in ribosome rescue, we deleted DOM34 from our RPS3-mutant strains and assessed its effect on the accumulation of 5’-fragments from the PGK1-SL reporter. As had been seen by others, deletion of DOM34 appeared to result in a loss of cleavage [16]. Interestingly the same deletion in the presence of the R116A/R117A mutations appears to restore cleavage with one caveat; the fragments are much more heterogeneous relative to those observed under normal conditions (Fig 4A). In particular, the products were observed to form a long smear on agarose gels. It seems that, under conditions where ribosome rescue is inhibited, mutation of the entry tunnel residues leads to a spreading of cleavage events well upstream of the stall site. To provide further support for this notion, we examined the effect of mutations in ASC1 on cleavage in conjunction with the RPS3 mutations. Asc1 is a ribosome-associated protein that has been implicated in multiple aspects of ribosome quality control processes including NGD [38,56–58]. For instance, cryoEM structures of a Dom34-Hbs1-bound ribosome revealed the factor to interact with Dom34 suggesting that it is critical for NGD [59,60]. In addition, recent data from the Inada group showed that the factor is important for sequential endonucleolytic cleavage during non-stop decay (NSD) in the absence of DOM34 [58]. Instead of deleting ASC1- which harbors a snoRNA gene in its intron- from our rps3 strains, we opted to introduce the R38D/K40E mutations into the chromosomal copy of the gene. These mutations are known to affect the association of the factor with the ribosome and phenocopy its deletion in NGD [61]. Similar to the effect we saw in the dom34Δ background, the ASC1 mutations resulted in the accumulation of heterogeneous 5’-fragments from the PGK1-SL NGD substrate in the presence of the R116A/R117A mutations (Fig 4B). To verify that the effect on NGD we observe with the RPS3 mutants are not due to decreased association of Asc1 with the ribosome, we carried out polysome analysis and used western analysis to look at the binding of Asc1 to ribosomes. As can be seen in Fig 4C, ribosomal occupancy by wild-type Asc1 is not significantly altered by the mutations in RPS3; similar to the wild-type, the protein was found to primarily associate with the polysomes in the presence of the RPS3 mutation (top panels). As a control, the R38D/K40E mutant was observed in the light fractions of the sucrose gradient, that is not ribosome-associated, regardless of RPS3 status (bottom panels). We should note, though, Asc1 participates in a multitude of processes on the ribosome including translation of short ORFs, stall clearance and ribosomal protein ubiquitination [37,38,56–58,62]. As a result, any interpretation of its consequence on NGD is likely to be complicated by the larger context of its effect on ribosome function. How inhibition of ribosome rescue either by deletion of DOM34 or mutation of ASC1 restores cleavage efficiency to entry-tunnel mutants, albeit with a distinct signature of heterogeneous product accumulation, is difficult to interpret. One plausible explanation is that the R116A/R117A mutations inhibit the accumulation of cleavage fragments and under normal conditions ribosome rescue is fast enough to dissociate stalled ribosomes, which results in the observed disappearance of cleavage products. When rescue slows down due to reduced cleavage kinetics, ribosomes accumulate on the mRNA, initiating cleavage further upstream of the stall sequence. Our Northern analysis of the NGD-cleavage products suggested that the R116A/R117A mutations affect cleavage fragments accumulation and result in ribosome queueing upstream of the stall site. This pile-up of ribosomes, in turn, results in cleavage reactions even farther upstream leading to diffusion of the NGD intermediates. We provided further support for these ideas by conducting high-throughput sequencing to map the 3’-end of the 5’-NGD fragments. Briefly, total RNA was isolated from strains harboring either the RPS3 mutants, dom34Δ, or ASC1 mutants in the ski2Δ background, each expressing one of the three NGD reporters- SL, (CGA) 12 and (AAA) 12. An adenylated DNA oligonucleotide was ligated to the 3’-end of the RNA samples, which was used to prime reverse transcription. The resulting cDNA was then amplified using a PGK1-specific 5’-primer and subjected to high-throughput sequencing using the Illumina Hiseq 2500 platform (GEO accession: GSE117652). Similar to what we have reported earlier [40], for otherwise wild-type cells, the 5’-fragments resulting from the PGK1-SL reporter mapped well upstream of the stall in all strains regardless of the mutational background (Fig 5). However, mapping of the fragments from the R116A/R117A mutant cells revealed extensive spreading of the cleavage events (Fig 5B). More specifically, whereas in the wild-type RPS3 background we observe one predominant peak near the ~150-nt upstream mark, in the rps3 mutant background, no predominant peak was observed (Fig 5B). Instead, fragments mapped throughout a 500-nt region upstream of the stall site and multiple peaks were observed with a near 30-nt periodicity. Interestingly, in the dom34Δ and the asc1 cells, the fragments displayed distinct mapping patterns relative to the wild-type and rps3 cells as well as to each other. Similar to what was observed for the rps3 mutant cells, in the dom34Δ cells the predominant peak at ~150-nt is lost, but here the distance between the peaks increased to 40–60 nt (Fig 5C). This is consistent with the role of Dom34 in rescuing ribosomes that run to the end of the transcript following endonucleolytic cleavage on NGD reporters. Since multiple ribosomes appear to be required for efficient cleavage, the reaction would be expected to occur every ~45-nt- with the lead ribosome protecting 15-nt, while the one behind protects 30-nt. In clear distinction to both the rps3 and the dom34Δ cells, mapping of the 5’-fragments from the SL reporter was not as diffuse in the asc1 mutant cells. Instead, only one additional predominant peak (relative to the wild-type cells) was observed at ~250 nt upstream (Fig 5D). Differences in cleavage patterns from the WT, rps3 and dom34Δ cells were also evident for 5’-fragments obtained from the (CGA) 12 reporter, and to a lesser extent for (AAA) 12 reporter (S2 Fig). We note that for both the (CGA) 12 and (AAA) 12 reporters, fewer reads were mapped in the rps3 cells, presumably due to decreased cleavage efficiency. These differences between the R116A/R117A mutant, and the DOM34 and ASC1 mutants suggest that the entry tunnel of Rps3 affects different aspects of NGD relative to these factors. It is also consistent with our model that these residues are important for the endonuclease function. Recently we showed that ribosome collision appears to play an important role in initiating NGD during stalling [17]. In particular, decreasing ribosome concentration, and hence ribosome density per mRNA, by deleting certain ribosomal protein paralogues was found to reduce cleavage of NGD targets [40]. As a result, we wondered whether the mutations of the entry tunnel residues had similar effects on ribosome density. To address this potential explanation, we compared the polysome profile of the rps3 cells to the wild-type ones. Our analysis revealed that the mutations in RPS3 had little effect on ribosome density (Fig 6). The ratio of polysomes to monosomes in the mutant is largely similar to that observed in the wild-type background. In contrast, similar analysis of the dom34Δ cells- as has been seen before [27]- revealed elevated levels of 80S monosomes relative to polysomes (Fig 6). The finding that the RPS3 mutations do not seem to affect ribosome density has two immediate ramifications: 1) the observed inhibition of NGD in the presence of these mutations does not result from changes to ribosome collisions; 2) consistent with our mapping analysis, the mutations are not likely affecting the function of Dom34. As discussed earlier, ubiquitination of ribosomal proteins by Hel2 (Znf598 in humans) has recently been recognized as an important feature of ribosome stalling. This modification promotes stalling on inhibitory codons as deletion of HEL2 results in significant bypassing of stalls by the ribosome [36–38,63,64]. Relevant to our studies is the observation that Rps3 is one of the targets for Hel2-mediated ubiquitination on K212, but it is currently unclear if its modification is important for stalling [64]. Nevertheless, if the entry tunnel mutations somehow affect Hel2 function, this could in principle explain their effect on NGD. As a result, we set out to assess stalling-induced ribosomal protein ubiquitination in the presence of R116A/R117A mutations. We took advantage of our previous observation that the addition of cycloheximide to an intermediary concentration, whereby ribosome collisions presumably occur at a global level, results in robust ribosomal protein ubiquitination [62]. We added cycloheximide to a final concentration of 2 μg/mL to wild-type, rps3 R116A/R117A, dom34Δ and double mutant cells; and isolated ribosomes. Ubiquitination patterns of ribosomal proteins resulting from cycloheximide addition, as assessed by western-blotting, was nearly identical among all strains (Fig 7). However, we noted that deletion of DOM34 had a discernible effect on the ubiquitination levels suggesting that Dom34 might affect Hel2 function (Fig 7). The rps3 mutations on their own, however, had no observable effect on the efficiency of ribosomal proteins ubiquitination. Hence, it is very unlikely that the effect of the entry-tunnel mutations on NGD are due to differences in ribosomal protein ubiquitination during stalling. We reasoned that if the entry-tunnel residues of Rps3 are affecting NGD, then mutating them should result in increased sensitivity to cycloheximide especially at intermediate concentrations, at which ribosome collisions will occur and hence NGD is triggered. Growth of the rps3 strain was compared to the wild-type one in the presence of varying concentrations of cycloheximide (Fig 8A and 8B). To distinguish between effects on the growth rate versus lag time, we determined the first derivative of the growth curve to measure the instantaneous growth rate. The maxima of the resulting curves report on the maximal growth rate, whereas the distance between the maxima reports on the lag. As expected, the mutations had no effect on the growth rate or lag period in the absence of the drug and at very low and high concentrations (S3 Fig). In contrast and in agreement with our model, the addition of cycloheximide at intermediate concentrations (0. 02–0. 32 μg/mL) significantly increased the lag period for the R116A/R117A mutant. This effect was most noticeable at the 0. 16 μg/mL concentration, for which we observed a lag-time difference between the wild-type and the mutant cells of more than 4 hours (S3 Fig). Our data suggests that the entry-tunnel residues are important for dealing with intermittent collision events, and likely the ensuing process of ribosome rescue. Previous work from our lab revealed that RNA oxidation strongly stalls translation in vitro [17]. In particular, the introduction of a single 8-oxoguanosine adduct to the mRNA reduced the rate of peptide-bond formation by almost three orders of magnitude in a bacterial reconstituted system and prevented the formation of full-length protein products in wheat-germ and rabbit-reticulocyte extracts. We also provided evidence that showed oxidized mRNA is subject to NGD. Because our rps3 mutations appear to affect NGD, they should also in principle result in increased sensitivity to agents that react with RNA to produce adducts such as 8-oxoguanosine. We used the chemical 4-Nitroquinoline 1-oxide (4NQO), a UV mimetic and known to produce reactive oxygen species, to introduce 8-oxoguanosine into RNA [65] in living yeast. Wild-type and rps3-mutant cells were grown to mid-logarithmic before being challenged with 5 μg/mL of 4NQO for 30 minutes. Cells were washed with fresh media, diluted and their growth monitored. In the absence of any drugs, the rps3 mutant displayed a growth rate nearly identical to that of the wild-type (6. 6 ±0. 23 versus 6. 3 ±0. 07 hours). After incubation with 4NQO, the mutant displayed a notable lag in its growth of 1. 4 hours (10. 4 ±0. 18 versus 9. 0 ±0. 79) (Fig 9A and 9B). We note that although the effects we saw are modest, they are reproducible and suggest that mutations of the entry-tunnel residues render cells sensitive to damaging agents. These effects are also reminiscent of the effects that we and others have documented for dom34Δ and xrn1Δ strains [17]. These findings together with the observation that mutations in RPS3 result in increased sensitivity towards cycloheximide provide further support for a role for the factor in NGD. NGD is a conserved eukaryotic process that responds to stalled ribosomes [14]. The process is characterized by an endonucleolytic cleavage of the aberrant mRNA upstream of the lead ribosome [15] and as yet the identity of the culprit endonuclease remains unknown. As a result, there is a critical gap in our understanding of some of the mechanistic details of the process. Nonetheless, multiple studies have provided important hints about the enzyme. For instance, mapping experiments suggested that the endonuclease is ribosome-associated [40,41]. In particular, cleavage takes place in frame with the ribosome and is phased by ~30 nt, the mRNA-length protected by the ribosome. Furthermore, the reaction appears to likely take place between stacked ribosomes [40]. These studies hinted at a role for the ribosome itself in activating or recruiting the endonuclease. Here we provided further evidence for this notion. More specifically, we find the entry tunnel of the ribosomal protein Rps3 to be important for the cleavage reaction. Mutation of the key-entry-tunnel residues Arg116 and Arg117 were found to drastically affect the outcome of the cleavage event; we observe a significant reduction in the accumulation of 5’-fragments from a number of NGD reporters when these residues are mutated to Ala. Consistent with these findings, although subtle, the half-life of the SL reporter increases in the presence of the mutations suggesting that these mutations may stabilize NGD reporters. Mapping of the cleavage products also revealed spreading of the cleavage reaction in the presence of the mutations. We note that Rps3 is known to interact with two key NGD factors: Dom34 and ribosome-associated Asc1 [60,66]. Although deletion or mutation of these factors affects the cleavage pattern in the rps3 background, as evidenced by northern analysis, the effect of the mutations on NGD do not appear to phenocopy those observed in the dom34Δ and asc1 strains, which is apparent in the high-throughput mapping data. Furthermore, the mutations do not alter Asc1 occupancy on the ribosome. Collectively our data suggests that the entry-tunnel region of Rps3, and hence the ribosome, has a function in NGD upon stalling. In agreement with this proposal, mutations of this region render cells sensitive to intermediate concentrations of cycloheximide and the nucleic-acid damaging agent 4NQO; both stall the ribosome and likely trigger NGD. Apart from the decoding center nucleotides, the Arg116 and Arg117 residues of the entry tunnel of the ribosome come closest to the mRNA. Indeed, some of the first studies on this region showed it to be important for unwinding the mRNA and make up part of the helicase domain of the ribosome [42]. While our data do not show the residues to be required for cleavage to take place–we still observe accumulation of NGD fragments in the presence of the mutations–they clearly affect the pattern of the cleavage reaction. It is feasible that the electrostatic interaction between the side chains and the phosphodiester backbone of the mRNA is important for locking the mRNA in place for the endonuclease to carry out its cleavage reaction. When these residues are mutated to Ala residues, the mRNA is more dynamic and its accessibility to the enzyme’s active site is severely affected. Alternatively, these residues might be important for recruiting or activating the endonuclease and as a result, changing their identities inhibits the cleavage reaction, although it is not clear how residues buried deep in the ribosome could be used efficiently to recruit exogenous protein factors. Instead, we favor a model whereby the endonuclease is intimately associated with the ribosome and it is activated upon stalling. In agreement with this, previous work has indicated that during non-stop decay, when the ribosome runs to the end of an mRNA, the endonucleolytic cleavage takes place near the exit tunnel of the ribosome [16,41,67] as evidenced partly by the accumulation of 15–18 nt fragments. Similarly, during a novel form of mRNA degradation termed ribothrypsis, it was suggested that an endonucleolytic cleavage event takes place near the exit tunnel [68]. Interestingly, recent structural data from human cells has revealed the position of multiple ribosomal proteins and associated factors at collided di-ribosomes–events that trigger NGD [69]. It appears that this higher order structure brings an entry- and exit-tunnel face of adjacent ribosomes in close proximity, which could potentially allow for interactions between otherwise distally positioned components. These include RACK1/Asc1 on the stalled ribosome with uS3, eS10, and uS10 on the collided ribosome, as well as eS26 and eS28 facing uS4 and rRNA helix 16 on the stalled and collided ribosomes, respectively. It will be exciting to see how modifications to these factors may affect endonuclease activity. In an endonuclease-independent consequence, the residues and their interaction with the mRNA could play a role in recruiting Dom34 and Hbs1 to the ribosome. Biochemical and structural studies have suggested that Hbs1 is recruited to a ribosome with little to no mRNA downstream of the A site [20,21,66,70]. The N-terminal of Hbs1 binds in the RNA entry tunnel, interacting with Rps3 [66]. It was hypothesized that Hbs1 cannot bind in the presence of mRNA in the entry tunnel [19,20,60,66]. Additional recent structural studies also revealed a potential role for Dom34 in sensing the mRNA channel, whereby it uses a unique β-loop to protrude into the mRNA channel to sense its absence [60]. Together these two mechanisms ensure that ribosome dissociation only occurs when the ribosome reaches the end of the mRNA, such as during NSD or on the behind ribosomes following cleavage during NGD. It is possible that the mutations in the entry tunnel of Rps3 make the mRNA more dynamic, preventing a clash with Dom34 and Hbs1. In turn, this allows the factors to bind and dissociate the ribosomes before cleavage could take place. In agreement with this model, deletion of DOM34 in the presence of the rps3 mutations restores cleavage efficiency, and with increased heterogeneity, as expected, due to widespread ribosome queueing. This model, however, does not explain why the cleavage patterns in the double mutant do not look similar to those observed in the dom34Δ mutant. Therefore, the effects of the rps3 mutations appear to be more complex and they are likely to alter different aspects of NGD including the cleavage and the dissociation reactions. In contrast, the mutations do not appear to affect the RQC pathway, as we observe comparable ribosomal protein ubiquitination patterning and efficiency upon inducing ribosome collisions regardless of the status of Rps3. Perhaps not surprising given its proximity to the mRNA, Rps3 plays a number of roles on the ribosome during translation. It has been shown to be important for providing the helicase activity to the ribosome; in bacteria Rps3/uS3, together with Rps4/uS4 and Rps5/uS5, encircle the incoming mRNA within the entry tunnel. When Arg131 and Arg132 in bacteria (corresponding to Arg116 and Arg117 in yeast) were mutated to Alanine, the efficiency of unwinding an RNA duplex by the ribosome was reduced [42]. Residues of Rps4 were also shown to contribute to helicase activity, but the process overall is coupled to and dependent on movements during translocation [71]. Rps3 is known to interact with other ribosomal proteins, including ribosome-bound Asc1/RACK1 [60,66]. In addition to its aforementioned role in NGD, Asc1 is known to be involved in preventing readthrough of inhibitory codons and reading-frame maintenance [72]. In eukaryotes, the C-terminal tail of Rps3 lies further inside the mRNA channel, proximal to Asc1 [43]. It is tempting to speculate that conformational changes that involve Rps3 could be communicated to Asc1, which then may initiate additional steps in NGD. However, the convergence of phenotypes among Rps3, Asc1 and Dom34 highlight the potential for redundancy or simply subtle differences of function between these and related factors. This is also evident during non-functional 18S rRNA decay (NRD), where both Asc1 and Rps3 have recently been identified as players in the pathway [73]. The post-translationally modified C-terminal tail of Rps3 is required for 18S NRD and, as Asc1 can collaborate with either Dom34 or Hbs1, it was suggested that multiple overlapping pathways function to deal with damaged rRNA. At another step in the translation cycle, Rps3 also contributes to stabilizing the incoming mRNA during initiation. Again, yeast residues Arg116 and Arg117 were shown to promote binding of the mRNA to eIF3 dependent pre-initiation complexes (PICs) and in particular, when the exit channel is empty, they were absolutely required [46]. This demonstrates the diverse functionality of Rps3 that is likely due in part to its position at the entry tunnel where it interacts with and can survey incoming transcripts. Collectively our findings provide further evidence for the central role of the ribosome in mRNA-surveillance pathways beyond just recognizing the aberrant mRNA and initiating the downstream events. The observation that mutations deep into the ribosome lead to dramatic changes to NGD bolsters arguments by us and others that the endonuclease is likely to be an integral part of the machine. This in turn could explain why it has been difficult to identify the endonuclease. It would be interesting to examine how quality control mechanisms evolved to integrate into fundamental biological machines. Further delineation of the details of this mechanism will also contribute to the understanding of how cells identify and degrade defective biological molecules. Finally, similar to NMD, NGD is likely to have been coopted to regulate gene expression. Indeed, recent reports have shown conditional deletion of Pelota (the human orthologue of Dom34) results in abnormal cellular differentiation [74]. The identification of the endonuclease is more than likely to provide further and important appreciation of the pervasiveness of this mode of gene regulation through NGD. Cells were grown at 30°C in YPD or in defined media when expressing reporter plasmids. Yeast strains were made using standard PCR-based disruption techniques in the background BY4741 (MATa (his3Δ1 leu2Δ0 met15Δ0 ura3Δ0). SKI7 knockout strains were generated with a LEU2 cassette, amplified using oligos complementary to the insertion site. RPS3 mutant strains were constructed by first cloning a fragment encoding RPS3-HIS3-rpS3 3’UTR, generated by fusion PCR, into the BamHI/XhoI sites in pPROEX-HTb. Point mutations in RPS3 were introduced by site directed mutagenesis and a cassette encoding the entire region was PCR amplified and used to transform the target yeast strains. RPS2 (E120A) strains were made using the same method and ASC1 (R38D, K40E) strains were made similarly, except using BamHI/XbaI sites in pET28a. HIS3 and LEU2 coding regions were amplified from plasmids pFAGa-6xGLY-FLAG-HIS3 and pAG415 [75] respectively. Plasmids encoding the PGK1 gene or PGK1-SL under control of the GAL1 promoter were obtained from R. Parker [15]. PGK1- (CGA) 12, PGK1- (AAA) 12 and PGK1- (UUU) 12 were made by annealing complementary oligos and ligating them to XbaI digested PGK1 plasmid [40]. Culture was grown overnight in a defined media (-Ura) with glucose. Cells were washed twice in media containing 2% Raffinose and 2% galactose, diluted to OD 0. 1 in the same media and grown to an OD of 0. 5–0. 8 to permit expression of the gal-driven reporters. RNA was isolated using hot phenol extraction followed by two sets of chloroform extraction and ethanol precipitation. 2μg of total RNA was resolved on 1. 2% formaldehyde agarose gel, followed by transfer to positively-charged nylon membrane (GE Lifesciences) using a vacuum blotter (Biorad). Next, nucleic acids were UV cross-linked to the membrane and baked at 80°C for 15 minutes. Membranes were then pre-hybridized in Rapid-Hyb buffer (GE Lifesciences) for 30 minutes in a hybridization oven. Radiolabeled DNA probe, which was labeled using polynucleotide kinase and [γ-32P]ATP, was added to the buffer and incubated overnight. Membranes were washed with nonstringent buffer (2 × SSC, 0. 1% SDS) three times, in some cases followed by three washes in stringent buffer (0. 2 × SSC, 0. 1% SDS), all at hybridization temperature. Membranes were exposed to a phosphorimager screen and analyzed using a Biorad Personal Molecular Imager. All Northern analyses were performed using at least three biological replicates. Representative images are shown. Cells expressing PGK1-SL were grown overnight in defined media (-Ura) plus glucose. Cultures were then washed in -Ura media, resuspended at OD 0. 1 in 50 mL -Ura plus galactose, and grown for 18–20 hours to allow expression of the reporter plasmid. Cells were collected at OD 0. 5–0. 6, washed once and resuspended in 11 mL pre-warmed -Ura media. A 1 mL aliquot was saved for the t0 timepoint and 1 mL 40% glucose added to the remainder. Cells were incubated at 30°C while shaking and aliquots taken at the indicated timepoints. For each sample, cells were pelleted, media was removed, and tubes were frozen on dry ice. RNA was isolated using a hot phenol method followed by two rounds of chloroform extraction and ethanol precipitation. 2 μg of total RNA for each sample was analyzed by Northern blot. Yeast cultures were grown to mid-log phase before addition of cycloheximide to a final concentration of 100 μg/mL. The culture was chilled by adding an equal volume of ice and centrifuged at 4°C. Cells were then resuspended in polysome lysis buffer (20 mM Tris pH 7. 5,140 mM KCl, 5 mM MgCl2,0. 5 mM DTT, 1% Triton-100,100 μg/mL cycloheximide, 200 μg/mL heparin), washed once and lysed with glass beads using a FastPrep (MP Biomedical). Supernatant from cleared lysate corresponding to 1 mg of total RNA was layered over a 10–50% sucrose gradient and centrifuged at 37,000 rpm for 160 min in an SW41Ti (Beckman) swinging bucket rotor. Gradients were fractionated using a Brandel tube-piercing system combined with continuous absorbance reading at A254 nm. Proteins were precipitated by the addition of TCA to 10% after a twofold dilution with water, and resuspended in HU buffer (8 M Urea, 5% SDS, 200 mM Tris pH 6. 8,100 mM DTT). Proteins were resolved on 15% SDS PAGE gels and transferred to PVDF membranes using a semi-dry transfer apparatus (BioRad). The membranes were blocked with milk in PBST for ~ 30 minutes at room temperature followed by incubation with primary antibody overnight at 4°C. After washing with PBST, the membrane was incubated with the appropriate HRP-conjugated secondary antibody for ~ 1hr at room temperature before washing 3–4 × with PBST. Detection was carried out on a GE ImageQuant LAS 4000 using the Pierce SuperSignal West Pico Chemiluminescent Substrate. The following antibodies were used: mouse anti-PGK1[22C5D8] (ab113687) and rabbit anti-rpS9 (ab117861) from Abcam; rabbit anti-ASC1 was a gift from Wendy Gilbert (Yale University) [61]; mouse anti-rpL4 was a gift from Heather True (Washington University in St. Louis); goat anti-mouse IgG HRP (31430) and goat anti-rabbit IgG HRP (31460) from Thermo Scientific. Total RNA from the indicated strains was ligated to a short adenylated DNA oligonucleotide, 5' rAppCTGTAGGCACCATCAAT/3ddC/ 3' , at its 3’ end using truncated T4 RNA ligase 2 (NEB). For each sample, total RNA from at least two biological replicates was included. Reverse transcription using a primer complementary to the adaptor was performed, and then cDNA was amplified with a 5’-primer that annealed at position 585 of PGK1. Primers were designed for the Illumina HiSeq platform and samples were column purified to remove primers before sequencing. Single-read HiSeq 2500 sequencing was performed by the Genome TechnologyAccess Center (GTAC) at Washington University. Raw data was analyzed for quality using the Fastx toolkit (http: //hannonlab. cshl. edu/fastx_toolkit/index. html), trimmed using cutadapt [76] and aligned to our reference reporter sequence using NovoAlign (http: //www. novocraft. com/). Sequencing results are available at GEO (accession #GSE117652). Sensitivity assays were conducted essentially as described [74]. Yeast cells were grown to mid-log-phase (OD600 of 0. 5–0. 7), collected, washed and resuspended in YPD to a final density of OD 0. 8. 5 μl of the cell suspension was added to 195 μl of YPD with CHX at various concentrations, from 0–10 μg/mL. All samples were prepared in biological triplicates as well as technical duplicates in 96-well polystyrene microplates. The plate was incubated at 30°C with shaking on a microplate scanning spectrophotometer (Biotek). Cell density was monitored every 10 min over 24–48 h at 600nm. To assay sensitivity to 4NQO, after growing cells to mid-log (OD 0. 5–0. 7) cultures were treated with and without 5 μg/mL 4QNO for 30 minutes. Cells were collected, washed and adjusted to OD 0. 8. Samples were plated and growth monitored as above.

In all organisms, optimum cellular fitness depends on the ability of cells to recognize and degrade aberrant molecules. Messenger RNA is subject to alterations and, as a result, often presents roadblocks for the translating ribosomes. It is not surprising, then, that organisms evolved pathways to resolve these valuable stuck ribosomes. In eukaryotes, this process is called no-go decay (NGD) because it is coupled with decay of mRNAs that are associated with ribosomes that do not ‘go’. This decay process initiates with cleavage of the mRNA near the stall site, but some important details about this reaction are lacking. Here, we show that the ribosome itself is very central to the cleavage reaction. In particular, we identified a pair of residues of a ribosomal protein to be important for cleavage efficiency. These observations are consistent with prior structural studies showing that the residues make intimate contacts with the incoming mRNA in the entry tunnel. Altogether our data provide important clues about this quality-control pathway and suggest that the endonuclease not only recognizes stalled ribosomes but may have coevolved with the translation machinery to take advantage of certain residues of the ribosome to fulfill its function.

Abstract Introduction Results Discussion Methods

northern analysis deletion mutation molecular probe techniques messenger rna polyribosomes mutation fungi molecular biology techniques cellular structures and organelles gel electrophoresis research and analysis methods electrophoretic techniques proteins ubiquitination molecular biology ribosomes yeast biochemistry rna eukaryota cell biology nucleic acids post-translational modification genetics biology and life sciences organisms

2018

Interactions between the mRNA and Rps3/uS3 at the entry tunnel of the ribosomal small subunit are important for no-go decay

Viruses in the Flavivirus genus of the Flaviviridae family are arthropod-transmitted and contribute to staggering numbers of human infections and significant deaths annually across the globe. To identify cellular factors with antiviral activity against flaviviruses, we screened a cDNA library using an iterative approach. We identified a mammalian Hsp40 chaperone protein (DNAJC14) that when overexpressed was able to mediate protection from yellow fever virus (YFV) -induced cell death. Further studies revealed that DNAJC14 inhibits YFV at the step of viral RNA replication. Since replication of bovine viral diarrhea virus (BVDV), a member of the related Pestivirus genus, is also known to be modulated by DNAJC14, we tested the effect of this host factor on diverse Flaviviridae family members. Flaviviruses, including the pathogenic Asibi strain of YFV, Kunjin, and tick-borne Langat virus, as well as a Hepacivirus, hepatitis C virus (HCV), all were inhibited by overexpression of DNAJC14. Mutagenesis showed that both the J-domain and the C-terminal domain, which mediates self-interaction, are required for anti-YFV activity. We found that DNAJC14 does not block YFV nor HCV NS2-3 cleavage, and using non-inhibitory mutants demonstrate that DNAJC14 is recruited to YFV replication complexes. Immunofluorescence analysis demonstrated that endogenous DNAJC14 rearranges during infection and is found in replication complexes identified by dsRNA staining. Interestingly, silencing of endogenous DNAJC14 results in impaired YFV replication suggesting a requirement for DNAJC14 in YFV replication complex assembly. Finally, the antiviral activity of overexpressed DNAJC14 occurs in a time- and dose-dependent manner. DNAJC14 overexpression may disrupt the proper stoichiometry resulting in inhibition, which can be overcome upon restoration of the optimal ratios due to the accumulation of viral nonstructural proteins. Our findings, together with previously published work, suggest that the members of the Flaviviridae family have evolved in unique and important ways to interact with this host Hsp40 chaperone molecule. The Flavivirus, Pestivirus and Hepacivirus genera of the Flaviviridae family each include important human and/or animal pathogens [1]. A major human pathogen hepatitis C virus (HCV) is a member of the Hepacivirus genus, while Pestiviruses bovine viral diarrhea (BVDV), border disease and classical swine fever viruses each have significant economic consequences in the livestock industry. Within the Flaviviridae family, members of the Flavivirus genus, which includes over 50 viral species, have perhaps the most significant impact on human health [2], [3]. Viruses in this genus, including yellow fever (YF), dengue (DEN), West Nile (WN), Japanese encephalitis (JE) and tick-borne encephalitis (TBE) viruses, contribute to staggering numbers of human infections and significant death rates across the globe. Viruses in this genus are usually transmitted via arthropod vectors and as such human infection depends on climate and geographical factors affecting the ranges of the transmitting arthropod and the likelihood of arthropod-human contact. Rising global temperatures, increased human population densities, human movement and increased dispersal of ticks and mosquitoes have contributed to increased numbers of epidemics in new geographical locations; this trend is likely to continue. While successful vaccines have been developed for prevention of YFV, JEV and TBEV infection, none are available for other pathogenic flaviviruses. Efforts to create and implement such vaccines have been hampered by the presence of multiple serotypes (DEN), the large geographical areas involved, and the sporadic nature of infection [4], [5]. Even for vaccine-preventable flavivirus infections, the cost associated with immunizing all at-risk people is enormous. Moreover, there are currently no drugs available for the specific treatment of any flaviviral disease. While several viral proteins are attractive targets for the development of small molecule inhibitors, the potential for rapid evolution of the flavivirus RNA genome suggests that resistance may be a significant problem. Disrupting critical interactions of viral proteins with host factors, or inducing expression of host proteins able to inhibit viral replication, are alternative approaches to developing effective anti-flaviviral therapies, and may limit the emergence of escape mutants. Unfortunately our understanding of host factor involvement in promoting or inhibiting flaviviral replication remains incomplete. Members of the Flavivirus genus share a common genome organization and replication strategy [1]. After virion entry and fusion of the viral and host membranes within the endosome, the ∼11,000 nt viral positive sense genomic RNA is translated in association with host cell membranes to form a single polyprotein which is co- and post-translationally cleaved by both host and viral proteases. The structural proteins, C, prM and E, are located in the N-terminal region of the polyprotein, followed by the nonstructural proteins, NS1, NS2A, NS2B, NS3, NS4A, NS4B and NS5. After appropriate cleavage and assembly of replication complexes, the genomic RNA is replicated by NS5, the viral RNA-dependent RNA polymerase, in association with other viral nonstructural and host proteins to generate new progeny genomes. Virion morphogenesis then follows via encapsidation and budding into the ER lumen. Virions mature during transport through the secretory pathway and are released into the extracellular milieu via exocytosis. Given the common replication strategy of the Flavivirus genus members, different species may exploit or be susceptible to many of the same host factors or environmental conditions. We hypothesized that identification of a cellular antiviral factor with activity against one flavivirus species may provide information on targets for broad-spectrum therapeutics. In order to identify antiviral cellular factors active against YFV, we conducted an iterative screen of a cDNA library from interferon-α treated cells. We identified DNAJC14, an Hsp40 family member, as able to mediate protection from YFV-induced cell death. Further studies demonstrate that DNAJC14 is recruited to YFV replication complexes and that its overexpression inhibits viral RNA accumulation. The C-terminus of DNAJC14, which mediates its multimerization, is required for antiviral activity. Furthermore, we found that silencing of endogenous DNAJC14 inhibits YFV replication. Overall our findings suggest that DNAJC14 plays an important role in regulating YFV replication complex assembly. Retroviral vectors pV1, pV1-GFP, pTrip-EGFP and pTrip-TagRFP have been described [6], [7], [8], [9]. Derivatives were generated using standard methods; all polymerase chain reaction (PCR) generated sequences were verified by sequencing and primer sequences are available upon request. Derivatives of pV1 were constructed to express human DNAJC14 (hDNAJC14) and mutants, each containing a carboxyl-terminal myc tag; the myc tag (EQKLISEEDL) was introduced during PCR by inclusion of myc-encoding nucleotide sequences in the antisense primers. DNAJC14- or truncation mutant-encoding DNA was amplified, using the Expand Long Template PCR System (Roche), from hDNAJC14 cDNA plasmid (Open Biosystems), and after digestion with SfiI was cloned into similarly digested pV1. Plasmid pV1-hDNAJC14-FL encodes full-length hDNAJC14 (amino acids 1–702) with a carboxyl-terminal myc tag. The human N-terminal truncation mutant (NT1) corresponding to the truncated hamster cDNA isolated in the screen starts from the amino acid corresponding to residue 305 of hDNAJC14, and was designed to have the identical amino terminus as the hamster sequence (305MVQFLSQS—); the corresponding human wildtype sequence has a phenylalanine residue at position 306. Additional mutants generated include NT2: 250AGFWWLIE—; NT3: 291MGVWTGRL—; NT4: 320FTRFLKLL—; NT5: 349LVGLGDRL— and as necessary contained an extra methionine residue for translation initiation. C-terminal truncated mutants end at the following amino acids prior to the myc epitope tag: CT1: —HISFGSRI625; CT2: —DLKEAMNT534; CT3: —EEVARLLT433; CT4: —RFLVGLGD354; CT5: —AEELCQLG248. The NT5CT1 mutant contains both the NT5 and CT1 truncations. Point mutations were introduced using site directed mutagenesis with the appropriate oligos by standard techniques. Plasmids pTrip-EGFP-hDNAJC14-NT5 and pTrip-RFP-hDNAJC14-FL or pTrip-RFP-hDNAJC14-NT1 were generated by PCR amplification of the NT5, full-length (FL) or NT1 hDNAJC14 sequences. After BsrG1 and XhoI digestion the sequences were ligated into similarly digested pTRIP-EGFP or pTRIP-TagRFP. These plasmids express hDNAJC14 (or mutants) fused in frame to the C-terminus of EGFP or RFP. To generate a doxycycline-inducible cell line expressing hDNAJC14-NT5, sequences encoding NT5 were generated by PCR and were inserted into pcDNA4/TO/myc-His B (Invitrogen) via HindIII and XbaI digestion to generate pcDNA4/TO/hDNAJC14-NT5. Plasmid pTrip-RFP-hNZAP was generated by cloning DNA encoding amino acids 1 to 252 of human zinc-finger antiviral protein [10], generated by PCR, as an in frame fusion with the carboxyl terminus of RFP in the pTrip-TagRFP vector. Plasmids pFlag-HCV-NS2-3-WT or pFlag-HCV-NS2-3-H171A were described previously [11] and express Flag-tagged HCV NS2-3181 or the inactive H171A mutant form of the NS2 protease, respectively. Plasmids pFlag-YF-NS2-3 (181) -WT or pFlag-YF-NS2-3 (181) -S138A were generated by amplification of YFV NS2-3 fragments from plasmids pACNR-YF17D [12] and pET-BS (+) /Sig2A-5356-R2107/2506E, S1622A [13], respectively, and cloning the KpnI/XhoI digested products into similarly digested pcDNA3. 1. The sense primer contained appropriate sequences to encode a Flag epitope tag on the N-termini of the respective proteins. The plasmids express Flag-tagged YFV NS2-3181 or the inactive S138A form of the NS3 protease, respectively. Plasmids pACNR-FLYF17Dx [12] and pACNR-FLYF17Da [14] contain the sequences of YFV 17D downstream of the SP6 promoter with XhoI and AflII linearization sites, respectively. Plasmid pACNR-FLYF-Asibi (the details of which will be described in another paper) contains the sequences of YFV Asibi downstream of the SP6 promoter. Plasmid pYF17D (5′C25Venus2AUbi) was constructed by inserting Venus, a variant of yellow fluorescent protein (YFP), into the YFV 17D open reading frame (ORF) using standard molecular techniques. All generated PCR products and plasmids were verified by restriction digests and by sequencing (primers available upon request). First, the Venus cassette was amplified from plasmid Venus/pCS2 (kindly provided by Dr. Atsushi Miyawaki) and cloned in frame after the first 25 amino acids of the YFV Capsid in pNEB193/YF5′ [15] using the SacI and AgeI sites. Next, the foot-and-mouth disease virus (FMDV) 2A peptide, which mediates cleavage following its own carboxy-terminus, and a ubiquitin (Ubi) monomer were amplified from pTM3-HCV-Ubi-NS5B (C. Lin and C. M. Rice, unpublished) and inserted downstream of Venus in pNEB193/YF5′ by assembly PCR. YFV 17D amino acids 1–514, containing silent mutations in sequences encoding the first 25 amino acids to avoid recombination, were similarly assembled downstream of FMDV 2A in pNEB193/YF5′. Finally, to generate pYF17D (5′C25Venus2AUbi), Venus and YFV 17D sequences were removed from pNEB193/YF5′ using SrfI and NsiI and cloned into pCC1-YF17D [16], which contains the entire YFV genome. Thus in pYF17D (5′C25Venus2AUbi), the Venus/2AUbi cassette is inserted in frame after the first 25 amino acids of Capsid, followed by the complete YFV polyprotein. The presence of FMDV 2A and Ubi downstream of Venus ensures complete cleavage from the YFV polyprotein, thereby ensuring the authentic YFV amino terminus. This strategy has the potential advantage of allowing expression of foreign inserts without disrupting the YFV 17D polyprotein, which may have unpredictable deleterious effects on replication. YFV replicon plasmids pYF-R. luc2A-RP and pYF-luc-IRES-RP-ΔDD [17] were kindly provided by Richard J. Kuhn (Purdue University). Plasmid pYF-R. luc2A-RP-ΔDD (expressing a polymerase defective YFV luciferase-expressing replicon) was constructed by swapping the NdeI/XhoI fragment from pYF-luc-IRES-RP-ΔDD into pYF-R. luc2A-RP. All cell lines were maintained at 37°C in humidified chambers containing 5% CO2. SW13 (human adrenal carcinoma) cells were cultured in Minimum Essential Medium (MEM) Alpha Medium (MEMa, Invitrogen) supplemented with 10% fetal bovine serum (FBS, Invitrogen). Huh7. 5 cells [18] were cultured in Dulbecco' s Modified Eagle Medium (DMEM, Invitrogen) supplemented with nonessential amino acids (Invitrogen) and 10% FBS. HEK293T and Vero cells were cultured in DMEM supplemented with 10% FBS. T-REx-293-LacZ cells inducibly expressing myc-tagged LacZ were previously described [19]. T-REx-293-NT5 cells inducibly expressing hDNAJC14-NT5 were obtained by transfection of T-REx-293 cells with pcDNA4/TO/hDNAJC14-NT5 and selection in medium containing zeocin. The selected bulk population was then cultured in DMEM supplemented with 10% FBS, 5 µg/ml blasticidin, and 0. 5 mg/ml zeocin. For induction, doxycycline was added to a final concentration of 1 µg/ml. BHK-J cells, a previously described [20] line of BHK-21 hamster kidney cells were cultured in MEM supplemented with 7. 5% FBS and BHK/NZAP-Zeo cells [21] were maintained in the same medium with the addition of 200 µg/ml zeocin. Anti-Myc mouse monoclonal antibody 9E10 (ATCC CRL1792 hybridoma) was used in Western and immunofluorescence or immunoprecipitation at 2. 5 and 16 µg/ml, respectively. Anti-Flag antibody (Sigma M2 mouse monoclonal) was used in Western analysis at 1∶1000 dilution. Yellow fever NS3 rabbit polyclonal antiserum was previously described [22] and utilized at 1∶5000 dilution for Western analysis, and 1∶500 for immunofluorescence. Rabbit polyclonal anti-GFP antiserum was generated as described [23] and utilized at 1∶20,000 dilution in Western and 1∶1000 in immunoprecipitation. Mouse monoclonal anti-calnexin antibody (BD Biosciences, 610523) was used in Western and immunofluorescence at 1∶250 and 1∶50 dilution, respectively. Mouse monoclonal anti-actin (Sigma, A5441) antibodies were utilized in Western analyses at 1∶5000 dilution. Rabbit polyclonal anti-DNAJC14 antibody (Sigma, HPA017653) was used in Western and immunofluorescence at 1∶2000 and 1∶200 dilution, respectively. Mouse monoclonal anti-double stranded RNA (dsRNA) J2 antibody (English & Scientific Consulting, Bt. Szirák, Hungary), kindly provided by Dr. Elena Frolova (University of Alabama at Birmingham), was used at 1∶200 dilution. Alexa Fluor 488 donkey anti-mouse IgG (A-212020) and Alexa Fluor 594 goat anti-rabbit IgG (A-11012, Invitrogen) were utilized in immunofluorescence at 1∶000 dilution. YF 17D neutralizing mouse monoclonal antibody 8A3 [24] ascitic fluid was kindly provided by Jack Schlesinger (University of Rochester). A dilution of 1∶100 was found to efficiently neutralize up to 105 pfu of YFV (data not shown). HRP-conjugated secondary anti-mouse (Jackson ImmunoResearch, 115-035-146) and anti-rabbit (Pierce, 31462) IgG antibodies were utilized at 1∶20,000 dilution. Normal rabbit IgG used in immunoprecipitations was from Santa Cruz Biotechnology, Inc. YFV stocks were generated by electroporation of BHK-J cells as previously described [25] with in vitro transcribed YFV RNA. Plasmids pACNR-FLYF17Dx [12], pYF17D (5′C25Venus2AUbi), and pACNR-FL-YF-Asibi were used for generation of YF 17D, YFV-Venus or YF Asibi, respectively. All work with YF Asibi was conducted under Biosafety level 3 containment conditions. Virus stocks and samples were titered by infection of BHK-J cells with 10-fold serial dilutions in MEM with 2% FCS. Two hundred µl of diluted virus was added to each 35 mm well and after 1 h of infection the well was overlaid with 0. 6% agarose in MEM supplemented with 2% FBS. Plaques were enumerated by crystal violet staining after 72 h. For YFV infections, multiplicity of infection (moi) was based on titers obtained on BHK-J cells. HCVcc (Jc1FLAG2 (p7-nsGluc2A) ) a cell culture-derived HCV expressing Gaussia luciferase was prepared by electroporation of Huh7. 5 cells as described previously [26]. Stocks of Kunjin (derived from infectious clone FLSDX 250pro) were propagated on Vero cells as described [27], [28]. Langat (TP21 strain) was propagated on Vero cells as described [29]. Titrations of stocks and samples were performed by focus forming assay, as previously described [30], [31]. Briefly, Vero cells were infected with 10-fold serial dilutions and after the 1 h adsorption the wells were overlaid with 0. 8% methylcellulose in DMEM containing 2% FBS. After 4 d the monolayers were fixed with 100% methanol and plaques were visualized by incubation with polyclonal mouse antibody cross-reactive to Langat (hyperimmune mouse ascites fluid, clone Russian Spring Summer Encephalitis VR79; ATCC) or polyclonal mouse anti-West Nile virus E protein (obtained from Dr. Robert Tesh, World Reference Center for Emerging Viruses and Arboviruses) followed by secondary goat anti-mouse peroxidase-labelled polymer (DAKO Envision Systems) and application of peroxidase substrate containing 0. 4 mg/ml 3,3′ diaminobenzidine and 0. 0135% hydrogen peroxide in PBS. To bypass entry steps, SW13 cells were electroporated essentially as described [32] with in vitro transcribed RNA generated from plasmid pACNR-FLYF17Da, YF-R. luc2A-RP, or pYF-R. luc2A-RP-ΔDD. Stocks containing VSV-G pseudotyped lentiviral particles were generated essentially as described [7] by cotransfection using Fugene 6 (Roche) of 293T cells with plasmids encoding VSV-G, HIV gag-pol and the lentiviral provirus plasmid at a ratio of 1∶4∶4 µg. Medium overlaying the cells was harvested at 48–72 h after transfection, filtered through a 0. 45 µM filter, aliquoted and stored at −80°C. Transductions were performed by incubating cells with the pseudoparticles in the presence of 8 µg/ml polybrene. The tissue culture 50% infectious dose (TCID50) was determined essentially as described [33] by titration on the TZM HeLa cell derivative [34], which expresses β-galactosidase under the control of the HIV LTR. Comparison of the TCID50 of a VSV-G pseudotyped V1-GFP stock, as determined on TZM cells, to the number of GFP-positive cells obtained after transduction of the target cell line with the same V1-GFP stock allowed for calculation of the appropriate TCID50 to utilize to achieve the desired transduction efficiency. For protein expression, transduction efficiency was typically in the range of 70–95%. A cDNA library was generated as described [7] from mRNA isolated from a BHK-21 derivative cell line, designated BHK/NZAP-Zeo [21], expressing the amino terminal fragment of the rat zinc-finger antiviral protein, after treatment for 6 h with 100 U/ml Universal type I IFN (PBL Biomedical Laboratories). Briefly, total RNA was harvested with Trizol (Invitrogen) and mRNA isolated by oligo dT selection (Oligotex mRNA Maxi Kit, Qiagen) according to the manufacturer' s recommendations. The cDNA synthesis was carried using the method of the SMART cDNA Library Construction Kit (Clontech) with the outlined modifications [7] which utilized Superscript III (Invitrogen) for first strand synthesis and TaqPlus Long PCR System (Stratagene) for second strand synthesis and amplification, SfiI digestion and size fractionation with cDNA Size Fractionation Columns (Invitrogen). After ligation to the minimal HIV provirus V1 vector that had been SfiI-digested, and electroporation of DH10B cells (Invitrogen), the library was divided into two sub-libraries (L1 and L2), each with >3,000,000 clones, for amplification. Plasmid DNA was isolated from the amplified libraries using a Qiagen MaxiPrep Kit, and VSV-G pseudotyped particles expressing the library cDNAs were generated as described above. Library L1 contained insert sizes ranging from ∼400 to ∼2,400 nt and was utilized for these studies. The L1 cDNA library was screened for cDNAs able to confer resistance to YFV 17D-mediated cell death using a previously described iterative approach [7] and as outlined in Figure 1. SW13 cells (18 million) were transduced with the L1 library of lentiviral particles (0. 45 TCID50/cell) and two days later were challenged with YFV 17D (moi = 5). After maintenance in MEM with 2% FBS for 7 d, the surviving Round 1 (Rd 1) cells were pooled and expanded in growth medium. The cDNA clones present in the surviving Rd 1 cells were rescued by transfection of cells in two 10 cm dishes with 15 µg VSV-G- and 5 µg HIV gag-pol-encoding plasmids diluted in OptiMem containing 40 µl Lipofectamine 2000 (Invitrogen) -according to the manufacturer' s recommendations. The medium overlying the cells was collected 2 d later, pooled, filtered through a 0. 45 µM filter, aliquoted and stored at −80°C. For subsequent steps, the rescued lentiviral stocks were treated for 2 h at room temperature with a 1∶100 dilution of mouse monoclonal 8A3 YFV neutralizing antibody prior to transduction of naïve SW13 cells in order to prevent cell death mediated by residual YFV in the lentiviral stock. Each rescued stock was utilized undiluted for subsequent transductions. Two additional rounds of transduction and challenge were performed to generate Rd 2 and Rd 3 cells and rescued lentiviral stocks. Rd 2 was performed on ∼4×106 SW13 cells, (0. 0003 R1 TC1D50/cell), and a YFV 17D challenge (moi = 5) 2 d later. Rescue was performed on the surviving Rd 2 cells in three 35 mm dishes, using Lipofectamine 2000-mediated transfection as described above (13 µl reagent, with 3 µg VSV-G- and 1 µg HIV gag-pol-encoding plasmids per dish). Rd 3 was performed on ∼2×106 SW13 cells (0. 07 R2 TC1D50/cell) and a YFV 17D challenge (moi = 1) 2 days later. A large number of cells survived the challenge compared to cells transduced with V1-GFP and challenged in parallel (see Figure 2A) and these Rd 3 cells were expanded for further testing. Rescue was performed as in Rd 2. A repeat experiment using the same Rd 2 rescued lentiviral stock gave similar results and an additional selection round using particles rescued from the Rd 3 cells also yielded many surviving Rd 4 cells (not shown). The predominant cDNA present in the Rd 3 cells was isolated by PCR. Rd 3 cellular DNA was isolated using the DNeasy Blood and Tissue Kit (Qiagen) and amplified using the Expand High Fidelity PCR System (Roche) and primers flanking the cDNA insert in the V1 vector (5′-GATTGTAACGAGGATTGTGGAACTTCTGGG-3′ and 5′-GATCCACAGATCAAGGATATCTTGTCTTCTTTGGG-3′). The PCR product was digested with SfiI, recloned into the V1 vector and sequenced using the above primers, as well as with primers designed to bind within the DNAJC14 sequence (5-TTGAAGCCACAGCATCC-3′ and AAGTCTACAGCTGCTCGAG-3′). Blast analysis demonstrated high homology with murine and human DNAJC14 with the cDNA insert predicted to express an amino-terminally truncated form of the protein. The nucleotide sequence of the truncated hamster DNAJC14 cDNA was submitted to GenBank (BankIt1399336 DNAJC14 HQ415606). To demonstrate DNAJC14 self interaction, HEK293T cells were seeded 16 h before transfection onto 60 or 100 mm tissue culture dishes at a density of 1. 6×106 or 4×106 cells, respectively. The cells were co-transfected using Fugene 6 (Roche) with 2 (60 mm dish) or 4 (100 mm dish) µg each of pTrip-EGFP-hDNAJC14-NT5 and pV1-hDNAJC14-FL or mutants. Forty-eight hours post transfection, cells were scraped into ice-cold PBS and solublized with lysis buffer (10 mM HEPES, pH 7. 5,150 mM KCl, 3 mM MgCl2,0. 5% NP-40,1×Proteinase inhibitor cocktail (Roche) ), using 300 or 600 µl for 60 or 100 mm dishes, respectively. After disruption by passing through a 27G needle 5 times and clarification by centrifugation at 15,000×g for 10 min at 4°C, 300 µl of the soluble fraction was incubated overnight at 4°C with anti-myc, anti-GFP or control antibody. Pre-equilibrated protein A/G-agarose beads (Santa Cruz) were then added, and after 2 h of incubation, were collected by centrifugation and then washed four times with 600 µl washing buffer (10 mM HEPES, pH 7. 5,150 mM KCl, 3 mM MgCl2,0. 05% NP-40). The bound proteins were eluted by boiling in sodium dodecyl sulfate (SDS) sample buffer and were subjected to Western analysis. To demonstrate the NS3-DNAJC14 interaction, SW13 cells were transduced with lentivirus expressing the myc tagged CT1 hDNAJC14 mutant and 2 d later were infected with YFV (moi = 1). After 2 d the cells were harvested and immunoprecipitation performed as described above except that the lysis and wash buffer contained 1% NP-40. Cells were directly lysed with 2×SDS loading buffer (100 mM Tris-Cl pH 6. 8,20% Glycerol, 4% SDS, 3% β-mercaptoethanol, 0. 02% bromophenol blue) and boiled for 5 min. Proteins were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and transferred to a Hybond ECL Nitrocellulose Membrane (GE Healthcare Life Sciences). The membrane was incubated in blocking buffer (PBS, 0. 05% Tween 20,5% dried milk) for 2 h, and then incubated with primary antibody diluted in blocking buffer at 4°C overnight. The membrane was washed 3 times in PBS supplemented with 0. 05% Tween 20 and incubated for 2 h at room temperature with HRP-conjugated secondary antibody. After 3 washes, the membrane was visualized by ECL Supersignal West Pico (or Femto) Chemiluminescent substrate (Thermo scientific). Cells were fixed with 4% formaldehyde in PBS and permeablized with 0. 2% Triton X-100 in PBS for 5 min at room temperature. After being washed with PBS, samples were then blocked and incubated overnight with primary antibody in 3% BSA in PBS at 4°C After three washes with PBS, samples were incubated at 37°C for 1 h with Alex488- or Alex594-conjugated secondary antibody. Coverslips were finally mounted with Mowiol Mounting Media [0. 1 M Tris-HCl, pH 8. 5,25% glycerol, 10% Mowiol 4–88 (Calbiochem 475904) ] and observed by Leica LSM510 confocal laser with a 100×NA 1. 3 oil immersion objective. Images were captured using the LSM software and processed using ImageJ. Cells were harvested by trypsinization, resuspended in PBS with 1% BSA, and then fixed in 2% formaldehyde in PBS. Samples were analyzed for expression of RFP and Venus using a BD LSR II flow cytometer, analyzing 10,000 events per sample. Data were processed using the FlowJo software. For the luciferase activity assay, transduced and infected Huh7. 5 cells were washed twice with PBS and lysed with 1× Passive Lysis Buffer (Promega) according to the manufacture' s recommendations. Luciferase activity was measured using the Renilla Luciferase Assay system (Promega) using a Lumat LB9507 Luminometer (Berthold). Triplicate wells of SW13 cells transduced with V1-GFP control or DNAJC14-expressing lentiviruses were seeded in 24 well plates at 1×105 cells per well in the presence of 60 nM Stealth RNAi siRNA Negative Control Med GC (12935-300) or DNAJC14-targeting Stealth siRNA (CCGAGGAACUAUGUCAACUUGGACA) and Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer' s reverse transfection protocol. The siRNA transfection was repeated 2 d later, using forward transfection with 60 nM siRNA. After an additional 2 d incubation, the cells were infected with YFV (moi = 5) and 24 h later the medium was collected for virus titration. For each condition, cells from one of the triplicate wells were harvested for Western blot analysis, while the remaining 2 wells were pooled for RNA harvest. RNA was purified using the RNeasy minikit (Qiagen) and each sample was reverse transcribed in triplicate using random primers and the Superscript III first strand synthesis kit (Invitrogen). Quantitative PCR was performed using the QuantiTect SYBR Green PCR Kit (Qiagen) and a LightCycler 480 (Roche) for detection as previously described [35]. Qiagen QuantiTect primers (QT00197043) were used for DNAJC14 amplification; levels were normalized to those of GAPDH, using a GAPDH primer set (sense: CCCACTCCTCCACCTTTGAC, antisense: CATACCAGGAAATGAGCTTGACAA) as described [36]. To identify cellular factors with antiviral activity against flaviviruses, we initiated a screen for host proteins that could inhibit cell death caused by YFV infection. We reasoned that a cDNA expression library generated from cells treated with interferon (IFN) -α to increase expression of antiviral factors would represent both IFN induced and constitutively expressed factors, some of which might have a protective effect against YFV. For this study, we utilized a cDNA library that we had generated (for other unrelated studies) from a BHK-21 cell derivative previously shown to develop dramatic resistance to Sindbis virus infection upon treatment with IFN [21]. Although we had some concerns regarding possible species incompatibilities for the function of hamster proteins in cells of other species, we thought it likely that factors influencing YFV, which has conserved replication strategies in both vertebrate and invertebrate cells, would function in a broad range of cells. We transduced YFV-susceptible human SW13 cells with the expression library, challenged these cells with YFV (vaccine strain 17D), and identified the cDNA (s) expressed in cells that survived the infection. During initial screens, we encountered several challenges, including difficulty cloning out rare surviving cells, and the presence of multiple library integrants. In order to overcome these obstacles, we expressed the cDNA library using a lentiviral vector (V1) in cells that are amenable to repackaging, as has been previously described [7]. Transfection of cells surviving the YFV challenge with helper plasmids expressing HIV gag-pol and an envelope glycoprotein (VSV-G) allows packaging of the lentiviral genomes, generating a lentivirus stock enriched for genes that confer a selective advantage (Figure 1). This approach obviates the need to clone individual cells, and allows iterative cycles of library transduction, YFV challenge, and rescue of sequences conferring survival. A YFV neutralization step of the selected, rescued lentiviral particles was used to prevent cell death mediated by residual virus during transduction of the naïve SW13 cells. After multiple rounds, the pool of cDNA-expressing lentiviruses will have markedly reduced complexity and active cDNA clones will be highly enriched. After two rounds of selection, transduction of SW13 cells with the enriched library of lentiviral constructs resulted in extensive resistance to YFV-induced cell killing (Figure 2A). These “Round 3” (Rd 3) cells were expanded, retested for their susceptibility to YFV-induced cytopathicity and found to be resistant at several multiplicities of infection (moi, Figure 2B). DNA was harvested from the Rd 3 cells, and the cDNA inserts were amplified using primers specific for the V1 vector (Figure 2C). The single major PCR product (∼2. 5 kb) was cloned, sequenced and found by BLAST analysis to show high homology to a murine (as well as human) Hsp40 family member, DNAJC14. The cDNA was predicted to express an N-terminally truncated version of the protein, which, based on the human sequence, lacked the first 304 amino acids of the 702 amino acid protein. DNAJC14 and the truncated hamster clone are shown schematically in Figure 2D. To test whether the DNAJC14 sequence obtained by PCR could confer resistance to YFV-mediated cell death in naïve cells, the PCR product was cloned back into the V1 lentiviral vector. Five individual clones (designated 1-1,1-2,1-3,1-4, and 1-5) were packaged and the VSV-G pseudotyped lentiviral particles were used to transduce naïve SW13 cells, which were challenged with YFV. Clone 1-1 was unable to confer resistance to YFV-mediated cell death, while each of the remaining four clones resulted in protection (data not shown). Sequencing of clones 1-1 and 1-2 indicated that both encoded a 398 amino acid protein (equivalent to human DNAJC14 aa 305–702) but clone 1-1 encoded a leucine to proline mutation at position 466, likely introduced during PCR amplification. This leucine is within the highly conserved J domain and is conserved amongst human, chimp, dog, cow, mouse and rat sequences. We conclude from these studies that expression of amino acids 305–702 of hamster DNAJC14 is able to confer resistance to YFV-mediated cell death. Because our screen used cDNA from IFN-α-treated cells, we tested whether interferon treatment of SW13 cells results in upregulation of DNAJC14 mRNA levels. After treatment for 8 h with IFN-α, DNAJC14 RNA levels were quantified by real time RT-PCR. No significant differences in DNAJC14 RNA levels were found in cells treated with doses of IFN-α ranging from 0 to 1000 IU/ml (data not shown). Detection of eIF2α phosphorylation by Western blot demonstrated that IFN was active in these cells (data not shown). Although we cannot exclude that the gene may be upregulated in response to IFN-α in some cell types, DNAJC14 appears to be constitutively expressed in SW13 cells. Our screen utilized protection from viral-mediated cell death as an endpoint for the isolation of host proteins with activity against YFV. Survival could be due to inhibition of virus replication or prevention of activation and/or blocking of cell death pathways. To test whether expression of DNAJC14 resulted in inhibition of viral growth, we infected the Rd 3 cells with YFV and quantified virus production (Figure 3A). Compared to naïve cells, YFV propagation was markedly reduced in DNAJC14 expressing cells, with a greater than 2 log reduction in infectious titers at 48 h, and virus production continuing to decrease over time. In contrast, robust replication occurred at 48 h in the naïve cells, with infectious titers decreasing at 5 d, at which time the cells displayed massive cytopathic effect. To determine whether the decreased infectious titers were a result of decreased intracellular viral replication, we performed Western blot analysis (Figure 3B) on YFV-inoculated cells transduced with V1 vector containing the clone 1-2,1-3,1-4 and 1-5 inserts. We found reduced levels of the viral NS3 protein in the cells expressing hamster DNAJC14 compared to control cells transduced to express GFP. These results demonstrate that truncated hamster DNAJC14 blocks YFV infection and/or replication, which results in prolonged cell survival. The truncated hamster DNAJC14 isolated in our screen is highly homologous to murine and human DNAJC14 proteins (89% identical and 93% similar to the corresponding region of the proteins). We therefore determined if expression of the 702 amino acid human DNAJC14 protein, also designated dopamine receptor interacting protein (DRIP78), could also confer protection against YFV-mediated cell death. We generated V1 lentiviral expression constructs to for both the full-length hDNAJC14 (hDNAJC14-FL) as well as an amino terminal truncation mutant (designated NT1) expressing amino acids 305–702 of human DNAJC14 (hDNAJC14-NT1), which corresponds to the hamster protein identified in our screen. A C-terminal myc epitope tag was engineered in the constructs to allow detection of the proteins. After packaging, the lentiviral pseudoparticles were used to transduce SW13 cells, which were then challenged with YFV (Figure 3C). Both the full-length and truncated versions of DRIP78 inhibited intracellular YFV NS3 accumulation. These studies demonstrate that the human DNAJC14 homolog is able to inhibit YFV infection and/or replication, as well as show that the addition of a C-terminal epitope tag does not interfere with the inhibitory activity. Bovine DNAJC14 (also known as J-domain protein interacting with viral protein, or Jiv) has previously been implicated in regulation of pestivirus (BVDV) replication. Intriguingly, cytopathic strains of the virus can contain insertions of DNAJC14 within their genome [37], [38], [39]. A portion of the Jiv protein (Jiv90) is required for the substrate interaction and activity of the viral autoprotease (NS2) and the subsequent establishment of replication complexes [40], [41], [42]. We therefore wondered how expression of DNAJC14 would affect other members of the Flavivirus genus, as well as the Hepacivirus genus member HCV. We first compared the ability of DNAJC14 to inhibit YFV 17D (vaccine strain) and the prototype virulent Asibi strain isolated from a young Ghanaian patient in 1927 [43]. Measuring of infectious virus production by transduced cells indicated that both the vaccine and Asibi YFV strains were susceptible to inhibition by the truncated or full-length hDNAJC14 (Figure 4A and B). Kunjin, a more distantly related mosquito-borne Flavivirus genus member (from the Japanese encephalitis serocomplex group) was also found to be susceptible to hDNAJC14-mediated inhibition (Figure 4C). A representative of the tick-borne encephalitis group, Langat virus, was similarly susceptible to hDNAJC14' s inhibitory effects (Figure 4D). It therefore seems likely that DNAJC14 broadly affects members of the Flavivirus genus. To establish whether hDNAJC14 also has effects on the Hepacivirus genus, we used HCV Jc1FLAG2 (p7-nsGluc2A), a cell-culture infectious virus (HCVcc) expressing a luciferase reporter [26]. Again, both the full-length DNAJC14 and NT1 truncation mutant, corresponding to our isolated hamster clone, inhibited viral propagation. Taken all together, these results suggest that DNAJC14 modulates the replication of many or all members of the Flaviviridae family. Since DNAJC14 regulates dopamine D1 receptor transport [44], it is possible that inhibition of YFV is the result of disrupting the transport of a cell surface receptor (s) utilized by the virus. We introduced the YFV genomic RNA into cells by electroporation in order to bypass entry and to determine if downstream steps are affected by DNAJC14. We found that DNAJC14 was still able to mediate inhibition of YFV protein expression when entry steps are bypassed; infectious virion production was also reduced (Figure 5A and B). The reduced protein levels and virion production might be due to decreased translation, RNA replication, assembly and/or egress. Using a YFV replicon (Figure 5C) expressing luciferase in place of the structural proteins [17], we tested whether DNAJC14 results in reduced expression of viral protein. The results (Figure 5D, wildtype replicon) demonstrate that YFV translation levels are reduced at later time points in the DNAJC14-expressing cells. Since the replicon does not express the structural proteins and is incapable of spread, the results suggest that DNAJC14-mediated inhibition can occur at a step after entry and prior to assembly, egress and spread. Interestingly, luciferase expression at early time points after electroporation was similar in the control (V1-GFP) and DNAJC14-expressing cells, suggesting that genome translation is not inhibited by DNAJC14. At later times (after 8 h) increased luciferase activity was detected in the control cells, suggesting that new RNA had been synthesized for translation. Use of a replicon containing a mutation abolishing RNA-dependent RNA polymerase activity (ΔDD) demonstrates that translation of the genomic RNA is similar in DNAJC14 overexpressing and control cells. Taken together the results indicate that a step after entry and translation and before assembly and egress is affected by DNAJC14. To ascertain determinants of DNAJC14 inhibitory function, we generated deletion and point mutants and tested their ability to inhibit YFV infection. A schematic of the deletion mutants is shown in Figure 6A. Expression levels, as determined by Western blot detecting the C-terminal myc tag on each of the constructs, were variable (Figure 6B), although immunofluorescence analysis verified almost 100% percent transduction efficiency for each of the mutants (data not shown). DNAJC14 has been proposed to reside in the ER membrane with both its N and C termini located within the cytoplasm [44]. This predicted topology was based on the interaction of DNAJC14 with the C terminus of the dopamine D1 receptor, as well as on DNAJC14 hydrophobicity plots and the absence of a signal peptide. Topology prediction programs suggest three potential regions that may serve as transmembrane (TM) domains. The truncated hamster mutant identified in our screen (NT1) contains an amino terminal deletion and is predicted to have one TM domain. Of the N terminal deletion mutants, NT3, NT4 and NT5 (lacking one, two or all three potential TM domains) exhibited similar antiviral activity to the full-length protein, while NT1 (lacking two) and NT2 (containing all three TM domains) exhibited the most potent activity (Figure 6C). Thus while the most inhibitory mutants contained at least one putative TM domain, the presence of a TM domain is not strictly required for inhibition. The C terminal deletion series were uninformative with respect to the role of the TM domains, since deletion of the C terminal 77 amino acids of DNAJC14 (mutant CT1), and various further deletions (mutants CT2, CT3, CT4, and CT5, which lacks all 3 TM domains) all resulted in a protein devoid of antiviral activity (Figure 6C). This suggests the carboxyl terminal 77 amino acids of DNAJC14 are required for antiviral activity. Although mutants CT3, CT4, and CT5 all contained deletions of the J domain, they also were not informative as to the role of the J domain in antiviral activity, since they also lacked the important C terminal domain. We utilized the NT5 mutant, which has robust expression and inhibitory activity equivalent to wildtype hDNAJC14, as the backbone to test several point mutations for their affect on YFV inhibition (Figure 6B, C). Our initial hamster clone 1-1 construct contained a presumed PCR-induced mutation at leucine 466 (to proline) within the J domain and was unable to confer resistance to YFV (not shown). We tested the L466P mutation in the context of hDNAJC14-NT5 and found that it abrogated the antiviral activity against YFV, suggesting a role for the J domain in the inhibitory process. Within the J domain, the conserved HPD motif is important for accelerating the ATPase activity of Hsp70 [45] and mutation of this motif (mutant H471Q) resulted in a noninhibitory protein. Studies on the interaction of rat DNAJC14 with the dopamine receptor [44] implicate the zinc fingers within the Jiv90 domain as important to the dopamine receptor-DNAJC14 interaction; mutation of cysteine 536 (537 in human DNAJC14, Figure 6), to serine abolished the DNAJC14-dopamine receptor interaction. We therefore generated mutations in two of the conserved Jiv90 cysteine residues, predicted to be involved in zinc coordination. Interestingly, mutants C537S and C559S could still inhibit YFV infection. We also mutated two residues (Y617A, I619A) that are required for maximal bovine Jiv90-mediated stimulation of BVDV NS2-3 cleavage [40]. Interestingly, these two mutants also displayed potent anti-YFV activity. Taken all together, the results suggest that the J domain and C-terminal domain are important for DNAJC14' s inhibitory effects on YFV. Both the DNAJC14 determinants important for modulation of pestivirus and flavivirus replication, as well as the result of DNAJC14 overexpression differ for viruses in these two Flaviviridae genera. Hsp40 family members are categorized into three classes [45], [46], [47]. Type I proteins contain the J domain at the N terminus followed by a glycine/phenylalanine rich region, four zinc finger motifs and a peptide binding fragment, with a C-terminal dimerization domain. Type II proteins contain an amino terminal J domain, and C-terminal peptide binding fragment, but lack the zinc-finger motifs, while Type III proteins are variable, with the J domain localized anywhere in the protein. DNAJC14 would thus be categorized as a Type III Hsp40, although the presence of two zinc-finger motifs downstream of the J domain suggests some similarities to the Type I members. Structural and functional analyses of several Type I and Type II Hsp40 members [45], [46] demonstrate that the C-terminal domains mediate dimerization. We therefore investigated whether DNAJC14 was capable of self-interaction. Using the NT5 mutant, which contains the C-terminal region, we tested its ability to interact with itself and with full-length DNAJC14. SW13 cells were cotransfected with plasmids expressing GFP and myc tagged DNAJC14 proteins and immunoprecipitations were performed using anti-myc antibodies. GFP-tagged NT5 co-purified with myc-tagged DNAJC14 or NT5 during myc-mediated immunoprecipitation, demonstrating self-interaction (Figure 7A). The NT5 self-interaction was verified by the reciprocal immunoprecipitation using anti-GFP antibodies (Figure 7B, left panel). However, mutant NT5 lacking the C-terminal 77 amino acids (NT5CT1) failed to co-purify in the immunoprecipitation (Figure 7B, right panel). Thus, similar to the Type I Hsp40 members, DNAJC14 multimerizes, and the self-interaction is mediated by the C-terminal 77 amino acids. Since the CT1 mutant also fails to inhibit YFV, it is possible that multimerization (likely dimerization) is important for DNAJC14' s antiviral activity. DNAJC14 is a required cofactor for the BVDV NS2 protease, which mediates autoproteolytic cleavage of NS2-3 as a necessary prerequisite for RNA replication [40]. Overexpression of DNAJC14 enhances cleavage at the 2/3 site, RNA replication and cytopathogenicity, but results in reduced infectious virion production due to a requirement for uncleaved NS2-3 for late life cycle events [48]. In contrast, in the case of YFV, DNAJC14 overexpression inhibits RNA replication (Figure 5D). Based on this apparent opposite effect, we wondered if DNAJC14 might inhibit (rather than enhance) YFV NS2B-3 cleavage and result in reduced levels of subsequent RNA replication. It is of interest that for YFV, cleavage at the NS2B/3 site is mediated by the viral NS3 protease, while for HCV the cleavage of the NS2/3 site is mediated by NS2. Since the effects of DNAJC14 on cleavage at the NS2/3 site of BVDV was successfully determined by coexpression of DNAJC14 and viral fragments capable of self-cleavage [39], we took a similar approach to test whether DNAJC14 inhibited YFV NS2B-3 cleavage. We first generated a doxycycline-inducible cell line expressing hDNAJC14 mutant NT5 with a C-terminal myc tag (Figure 8A). As expected, YFV replication was reduced in this cell line when treated with doxycycline to induce hDNAJC14-NT5 expression (Figure 8B). Using transfection, we expressed Flag-tagged self-cleavage competent YFV NS2B-3181 as well as a form incapable of cleavage due to a S138A active site mutation within NS3 [49] and monitored cleavage in doxycycline treated (expressing hDNAJC14-NT5) or non-induced control cells by Western blot. Similarly, we expressed Flag-tagged self-cleavage competent HCV NS2-3 protease, as well as a form incapable of autocleavage due to a H143A active site mutation within NS2 [11]. A plasmid expressing GFP was cotransfected to monitor transfection efficiency. As shown in Figure 8C, wildtype NS2B-3 or NS2-3 was efficiently processed resulting in similar levels of NS2 in the presence or absence of DNAJC14-NT5. The low levels of cleavage incompetent HCV NS2B-3 are likely due to the previously described rapid degradation of uncleaved NS2-3 [50]. Thus, contrary to our prediction, DNAJC14 does not grossly inhibit YFV, or HCV polyprotein cleavage. However, due to the sensitivity of this assay, subtle inhibition of processing efficacy would likely not be detected. Moreover, given the efficiency of cleavage in cells not induced to express DNAJC14, we cannot exclude an enhancement effect, similar to that seen with BVDV NS2-3 processing, of DNAJC14 on YFV NS2B-3 or HCV NS2-3 cleavage. Since DNAJC14 does not inhibit YFV genome translation yet blocks RNA replication, we wondered if it might interfere with the formation of functional replication complexes, which assemble on ER-derived membranes. Studies to investigate whether hDNAJC14 colocalizes with replication complexes in YFV infected cells are complicated by the fact that DNAJC14 expression inhibits YFV replication. To determine whether hDNAJC14 colocalizes with YFV replication complexes, we made use of the non-inhibitory DNAJC14 mutants H471Q and CT1 and monitored their colocalization with YFV NS3. SW13 cells transduced with lentiviruses expressing hDNAJC14 mutants were infected with YFV and the localization of NS3 and hDNAJC14 was examined by confocal microscopy (Figure 9A). As a control, we examined the localization of calnexin and demonstrated that YFV infection results in a redistribution of this ER marker to colocalize with NS3 in infected cells (Figure 9A). Full-length DNAJC14 containing the J domain mutation H471Q (FL-H471Q) and mutant CT1 both colocalized with NS3 in infected cells. The results are consistent with the known ER reorganization that occurs during YFV replication complex formation and suggest that DNAJC14 proteins associated with the ER membrane redistribute to replication complexes during YFV infection. Expression levels of the CT1 mutant, which is recruited to sites containing NS3 (Figure 9A) without blocking replication (Figure 6C), are similar to expression levels of the inhibitory full-length protein (Figure 6B). This makes a nonspecific process, such as the induction of ER stress due to protein overexpression, unlikely for the inhibitory mechanism. To assess further whether DNAJC14 associates with the viral replication complexes, we looked for a physical interaction using coimmunoprecipitation. Cells transduced (or not) to express the myc epitope tagged noninhibitory CT1 mutant were infected with YFV. CT1 and associated proteins were isolated from lysates using anti-myc antibody. As can be seen in Figure 9B, NS3 was coimmunoisolated with CT1, while the ER marker calnexin was not. Thus while both calnexin and CT1 colocalize with NS3 in immunofluorescence assays (Figure 9A), NS3, but not calnexin, was found to be in a physical complex with CT1. We utilized antibodies directed against dsRNA as another means to identify replication complexes and assess whether endogenous DNAJC14 is present (Figure 9C). Using anti-DNAJC14 antibody, we found that in uninfected cells, endogenous DNAJC14 in the cytosol predominantly displayed a diffuse pattern with occasional punctate staining. In infected cells the endogenous DNAJC14 demonstrated a more punctate staining pattern and the dsRNA was found colocalized with these punctate sites of staining. These findings demonstrate that both endogenous DNAJC14 and overexpressed non-inhibitory DNAJC14 mutants are recruited to YFV replication complexes, which suggests that endogenous DNAJC14 may facilitate replication complex formation. To test whether DNAJC14 might be required for, or facilitate virus replication, we used siRNA-mediated silencing to reduce levels of endogenous DNAJC14 and tested the ability of YFV to replicate. To evaluate replication capacity across a range of DNAJC14 levels, we used cells transduced with vector as well as cells transduced with lentivirus expressing DNAJC14 and subjected them to silencing with a control irrelevant siRNA or siRNA targeting DNAJC14 mRNA within the protein coding region. It should be noted that the absolute level of DNAJC14 RNA in normal cells (vector-transduced cells treated with control siRNA) is low, with DNAJC14 RNA levels more than 1000 fold lower than GAPDH mRNA levels (data not shown). Reducing levels of DNAJC14 mRNA by ∼2 fold, as measured by quantitative RT-PCR (Figure 10A), resulted in a ∼4 fold statistically significant (p<0. 0001) reduction in YFV titer (Figure 10B, compare vector cells treated with the control and DNAJC14 siRNAs). Western blot analysis using anti-DNAJC14 antibody demonstrates a reduction upon silencing at the protein level as well (Figure 10C). Although no protein band is apparent in the vector cells treated with the DNAJC14-targeting siRNA, given that mRNA was still detectable, a low level of residual protein could account for the modest reduction in viral replication. Despite multiple attempts we were unable to reduce the DNAJC14 mRNA levels lower than ∼2 fold (data not shown). Interestingly, cells transduced with lentivirus expressing DNAJC14 had a >300 fold increase in the level of DNAJC14 mRNA (Figure 10A,) and a corresponding increase in DNAJC14 protein levels (Figure 10C), which resulted in a ∼15 fold inhibition of YFV virion production (Figure 10B, compare vector and DNAJC14 cells treated with the control siRNA). Silencing of DNAJC14 in the DNAJC14-overexpressing cells resulted in intermediate mRNA (Figure 10A) and protein (Figure 10C) levels, although the RNA levels remained ∼50–60 fold higher than endogenous levels (compare DNAJC14 siRNA-treated DNAJC14 cells to control siRNA-treated vector cells). This residual intermediate level of DNAJC14 was less inhibitory than the high levels present in DNAJC14 overexpressing cells treated with the control siRNA, but still resulted in a 4. 7 fold inhibition of YFV replication compared to vector cells treated with the control siRNA (Figure 10B). Thus decreasing DNAJC14 levels by ∼2 fold or increasing levels by ∼50 fold each had a similar (∼4 fold) inhibitory effect towards YFV. Thus maximal YFV replication requires an optimal DNAJC14 concentration; levels too low or too high result in inhibition. A requirement for an optimal level of DNAJC14 and the ability of overexpressed wildtype DNAJC14 to inhibit YFV replication could be explained by DNAJC14 facilitating YFV replication complex formation in a stoichiometric process such that increased levels might function in a dominant negative fashion to inhibit replication complex formation. There is precedent for this, since DNAJC14 modulates BVDV NS2-3 cleavage in a temporal manner due to a stoichiometric mechanism [40], [41]. After translation of the BVDV polyprotein, NS2-3 autoprocessing is mediated in cis by the cysteine protease residing in NS2 [41], which requires DNAJC14 in a 1∶1 ratio [40]. RNA replication is dependent on this cleavage, due to a requirement for free NS3 for the formation of functional replication complexes [41]. Limiting amounts of cellular DNAJC14 thus limit processing and result in downregulation of RNA replication at later time points, allowing viral persistence [40]. Overexpression of DNAJC14 results in increased cleavage at NS2/3, increased RNA replication, and increased cytopathogenicity [39]. We wondered if inhibition of YFV might exhibit similar properties, in which the ratio of DNAJC14 to viral substrate is critical for its antiviral activity. If so, then inhibition would be expected to be dose-dependent, and continued translation of the incoming genome over time might restore the appropriate stoichiometry and thus allow replication to begin. Consistent with this hypothesis, we noticed that the antiviral activity of DNAJC14 diminished at later times after infection or electroporation (Figures. 4A, 5B). To test this hypothesis, we monitored the antiviral activity of DNAJC14 at the single cell level by flow cytometry. Cells transduced with RFP-tagged DNAJC14 (full-length and NT1 mutant) were infected with a YFV variant expressing Venus, and both virus replication (Venus) and DNAJC14 expression (RFP) were monitored. As shown in Figure 11A, the YFV signal was dramatically reduced in RFP-DNAJC14-FL- and -NT1-expressing cells compared to levels seen in cells expressing ZAP, an anti-Sindbis virus protein with no effect on YFV [10], [51]. In addition, the NT1 mutant demonstrated more potent inhibitory activity than full-length DNAJC14, which may be due to higher expression levels of NT1 as reflected by the RFP signal. Interestingly, both full-length and the NT1 mutant inhibited YFV in a dose dependent manner (Figure 11A), with lower YFV (Venus) signal seen in cells expressing higher levels of DNAJC14 (RFP). We next investigated the antiviral activity of mutant NT1 at various time points after infection (Figure 10B). At late time points (4 d), substantial YFV (Venus) signal was detected in RFP-positive cells. Even at this late time, when substantial YFV replication was occurring, the inhibition mediated by NT1 was still dose-dependent. To verify that increased Venus expression was due to increased replication, in a separate experiment we monitored infectious virus production in cells expressing RFP-DNAJC14-NT1 compared to nontransduced cells. Virion production on day 4 was found to be equivalent (Figure 10C), despite the fact that the DNAJC14 overexpressing cells were more resistant to cell death as monitored by crystal violet staining (data not shown). To exclude the possibility that the increase in virus replication seen after 3 to 4 d is due to the generation and replication of escape mutants, we collected the culture medium from cells after 4 d of infection and re-infected new cells expressing RFP-DNAJC14-NT1. Infection with this virus resulted in a similar early inhibition with a time-dependent increase in virus replication (data not shown). These results demonstrate that not only is DNAJC14-mediated inhibition dependent on the level of DNAJC14, but with time, levels that initially blocked YFV replication no longer are inhibitory. The failure to observe YFV replication in the Rd 3 cells (Figure 3A) is likely due to the fact that these cells had undergone prior infection and selection, resulting in a population of cells with maximal inhibitory properties. DNAJC14 (also designated DRIP78, Jiv and HDJ3) is a member of the Hsp40 family of protein chaperones [45], [52]. Proteins in this family contain a 70 amino acid motif, designated the J-domain, which recruits Hsp70 family members and stimulates the ATP hydrolysis step of the chaperone process. J-domain containing proteins are involved in diverse cellular processes. The human DNAJC family has 23 members with the presence of the J domain being the single common feature. Although not extensively studied, involvement of these proteins in mitochondrial import, translation, endocytosis and exocytosis has been noted [45]. DNAJC14 has previously been implicated in the life cycle of a member of the Flaviviridae. The bovine homolog of this factor, Jiv, is essential for the polyprotein cleavage and replication of the pestivirus BVDV. Jiv acts as a required co-factor for the viral NS2 autoprotease, influencing its cleavage from NS3 and modulating RNA replication, virus production and cytopathogenicity of this pestivirus [38], [39], [42]. In contrast to our findings with YFV, increased expression of Jiv results in higher levels of BVDV RNA replication and virus-induced cell death. Interestingly, some cytopathic biotypes of BVDV are naturally occurring recombinant viruses, which have insertions of DNAJC14 in the NS2-3 coding region. A 90 amino acid domain common to all of the Jiv-containing cytopathic BVDV isolates is designated Jiv90 (see Figure 2D). This sequence is distinct from the J-domain and contains two conserved CXXCXXXH motifs. DNAJ proteins regulate the ATPase cycle of Hsp70 via their J domain, with the HPD motif critical in accelerating the Hsp70 ATPase activity, while the substrate-binding domain loads the substrate onto Hsp70 [53], [54]. Our studies with mutant H471Q suggest that the critical HPD motif within the J domain is required for DNAJC14 antiviral function, suggesting that ATP-driven Hsp70 chaperone activity may be involved in the process of RNA replication and its inhibition. Since Hsp70 chaperone activity occurs via a stoichiometric mechanism, with a single Hsp70 monomer per substrate [54] it seems likely that DNAJC14/Hsp70 chaperone activity is required for YFV replication complex assembly and that overexpression of DNAJC14 disrupts the chaperone/substrate complex. Dimerization of some Hsp40 family members is evolutionarily conserved and required for their function [45], [53]. The CT1 mutant lacks the ability to multimerize (Figure 8) and fails to inhibit YFV (Figure 6) as well as HCV (data not shown). Thus multimerization is likely critical for DNAJC14' s antiviral function. In our studies, we demonstrated that DNAJC14 noninhibitory mutants are found in YFV replication complexes, as measured by colocalization and coprecipitation with NS3 (Figure 10A and B). Moreover, endogenous DNAJC14 rearranges upon YFV infection and is found colocalized with active replication complexes, as determined by the presence of dsRNA (Figure 10C). This suggests that YFV replication complexes assemble at a specific ER membrane site where DNAJC14 is located, and that DNAJC14 (and likely Hsp70) is specifically recruited to facilitate formation of the viral replication complex. DNAJC14 overexpression would then result in disrupted chaperone/substrate stoichiometry and inhibit replication complex assembly. Alternatively, YFV may hijack DNAJC14-containing membranes for its replication complex assembly, and overexpression may inhibit the distribution and recruitment of other host factors localized to this membrane microdomain and required for replication complex formation. We realized that the inhibitory effect of DNAJC14 on YFV was diminished at later time points post infection and that inhibition was dose dependent, with higher levels of DNAJC14 resulting in lower levels of virus replication (Figure 11). One possible explanation is that at early time points, overexpressed DNAJC14 is in vast excess to its substrates (viral proteins) and this inappropriate stoichiometry results in inhibition of replication complex formation. Since DNAJC14 does not inhibit virus genome translation (Figure 5D), nor polyprotein processing (Figure 8C), viral protein would be predicted to accumulate with time. At some point, the level of viral protein (s) would result in an appropriate DNAJC14 to substrate ratio to allow the chaperone process to occur and thus overcome DNAJC14' s inhibitory effect. This is not dissimilar to the scenario occurring with BVDV, in which the DNAJC14 Jiv90 domain interacts with BVDV NS2-3 at a ratio of 1∶1. This stoichiometric mechanism might be a common requirement for normal DNAJC14 cellular function, as either overexpression or sequestration of DNAJC14 inhibits dopamine D1 receptor transport [44]. We found that multiple Flaviviridae were inhibited under conditions of DNAJC14 overexpression and wondered whether viruses from other families might be similarly affected. We tested DNAJC14' s effects on Sindbis virus, a positive strand RNA virus from the Alphavirus genus. In contrast to the flaviviruses, we found that DNAJC14 overexpression had no effect on viral replication, as measured by expression of a fluorescent reporter from the viral subgenomic RNA (data not shown). Thus replication complex formation for Sindbis virus is not likely affected by DNAJC14 overexpression. Interestingly, however, Sindbis virion production was reduced by DNAJC14 overexpression (data not shown). Thus a step in Sindbis virus assembly, such as glycoprotein maturation and transport from the ER-Golgi to the plasma membrane, where Sindbis budding occurs, may require DNAJC14-containing membrane microdomains and chaperone function. In addition, overexpression of DNAJC14 reduced VSV virion production (data not shown), suggesting an effect on VSV glycoprotein ER-golgi transport. Since it has been reported that DNAJC14 is involved in dopamine D1 receptor transport [44], it is likely that DNAJC14 facilitates specific membrane processes including vesicle transport and viral replication complex assembly. It is possible that many virus families have specific requirements for chaperone processes at various steps in their life cycle. Understanding these requirements and identifying the chaperones and proteins undergoing the chaperone process may lead to insights into similarities and differences between different virus families in these critical life cycle steps. DNAJC14 can both facilitate and inhibit YFV replication. Based on all of our findings, we propose the following model (Figure 12): Translation of the incoming YFV RNA and subsequent polyprotein processing generates the viral proteins necessary for the viral RNA replication process. DNAJC14 functions as a chaperone system, most likely with involvement of Hsp70, to facilitate a step in the YFV membrane-associated multiprotein complex assembly that is critical for the formation of replication complexes. The TM domains within DNAJC14 target the protein to a specific subcellular ER membrane location, wherein substrate selection and YFV replication complex formation occurs. Multimerization of DNAJC14 via its C-terminus is likely required for assembly of the chaperone/substrate complex. It is possible that each DNAJC14 monomer binds a substrate and together they promote the proper folding and interaction of the substrate pair, which might be different sites on the same protein, two different YFV proteins, or a viral protein and host protein necessary for viral replication. Newly generated viral RNA is produced, which after translation generates new substrate for the chaperone process, and the formation of additional replication complexes. Overexpression of DNAJC14 mutants that fail to multimerize (CT1, CT2, CT3, CT4 or CT5), or contain mutations in the critical J domain (L466P, H471Q, Figure 12A) has no effect on virus replication; these mutants fail to interact with and disrupt the normal chaperone components and therefore exhibit no antiviral activity. Expression of full-length (wildtype) DNAJC14 results in an excess of DNAJC14 relative to the substrate, and complexes with an inappropriate stoichiometric ratio are formed, disrupting the chaperone process (Figure 12 B). The N-terminal truncation mutants, which contain the C terminal multimerization motif and an intact J domain, also interact with the chaperone components, disrupting the proper chaperone/substrate stoichiometry. With time, continued translation of the incoming viral genome (or genome generated by very low levels of viral replication) results in the accumulation of viral proteins. Once the optimal substrate/chaperone ratio is established, the restored chaperone process results in replication complex formation and viral RNA replication. Further studies are required to address the viral and cellular substrate (s) for DNAJC14 and to determine if other host factors (for example, Hsp70) participate in this important chaperone process.

Viruses in the Flavivirus genus are transmitted by arthropods and cause significant disease burden across the globe. We undertook a screening approach to select for and identify host factors that provide resistance to death caused by infection with the mosquito-transmitted Flavivirus, yellow fever virus (YFV). We identified the host factor DNAJC14, an Hsp40 chaperone protein family member, as able to inhibit replication of the vaccine strain of YFV, and the virulent parental Asibi strain. We found that DNAJC14 also inhibits several other members of the Flavivirus genus, including Kunjin and the tick-borne Langat virus. Moreover, the Hepacivirus hepatitis C virus is also inhibited, suggesting a role for DNAJC14 in modulating the replication of all three genera of the Flaviviridae family. By probing the mechanism of the YFV inhibitory process, we determined that DNAJC14 inhibits at a post entry step, and most likely prevents the formation of functional replication complexes. We determined that DNAJC14 is required for YFV replication and that expression of inappropriately high levels of this protein results in a disruption of a process critical for viral RNA replication. Understanding how host factors inhibit or contribute to Flavivirus replication steps may identify new targets for antiviral drug development.

Abstract Introduction Materials and Methods Results Discussion

cell biology/membranes and sorting virology/viral replication and gene regulation virology/mechanisms of resistance and susceptibility, including host genetics

2011

Identification and Characterization of the Host Protein DNAJC14 as a Broadly Active Flavivirus Replication Modulator

The population genetic perspective is that the processes shaping genomic variation can be revealed only through simultaneous investigation of sequence polymorphism and divergence within and between closely related species. Here we present a population genetic analysis of Drosophila simulans based on whole-genome shotgun sequencing of multiple inbred lines and comparison of the resulting data to genome assemblies of the closely related species, D. melanogaster and D. yakuba. We discovered previously unknown, large-scale fluctuations of polymorphism and divergence along chromosome arms, and significantly less polymorphism and faster divergence on the X chromosome. We generated a comprehensive list of functional elements in the D. simulans genome influenced by adaptive evolution. Finally, we characterized genomic patterns of base composition for coding and noncoding sequence. These results suggest several new hypotheses regarding the genetic and biological mechanisms controlling polymorphism and divergence across the Drosophila genome, and provide a rich resource for the investigation of adaptive evolution and functional variation in D. simulans. Given the long history of Drosophila as a central model system in evolutionary genetics beginning with the origins of empirical population genetics in the 1930s, it is unsurprising that Drosophila data have inspired the development of methods to test population genetic theories using DNA variation within and between closely related species [1–4]. These methods rest on the supposition of the neutral theory of molecular evolution that polymorphism and divergence are manifestations of mutation and genetic drift of neutral variants at different time scales [5]. Under neutrality, polymorphism is a “snapshot” of variation, some of which ultimately contributes to species divergence as a result of fixation by genetic drift. Natural selection, however, may cause functionally important variants to rapidly increase or decrease in frequency, resulting in patterns of polymorphism and divergence that deviate from neutral expectations [1,2, 6]. A powerful aspect of inferring evolutionary mechanism in this population genetic context is that selection on sequence variants with miniscule fitness effects, which would be difficult or impossible to study in nature or in the laboratory but are evolutionarily important, may cause detectable deviations from neutral predictions. Another notable aspect of these population genetic approaches is that they facilitate inferences about recent selection—which may be manifest as reduced polymorphism or elevated linkage disequilibrium—or about selection that has occurred in the distant past—which may be manifest as unexpectedly high levels of divergence. The application of these conceptual advances to the study of variation in closely related species has resulted in several fundamental advances in our understanding of the relative contributions of mutation, genetic drift, recombination, and natural selection to sequence variation. However, it is also clear that our genomic understanding of population genetics has been hobbled by fragmentary and nonrandom population genetic sampling of genomes. Thus, the full value of genome annotation has not yet been applied to the study of population genetic mechanisms. Combining whole-genome studies of genetic variation within and between closely related species (i. e. , population genomics) with high-quality genome annotation offers several major advantages. For example, we have known for more than a decade that regions of the genome experiencing reduced crossing over in Drosophila tend to show reduced levels of polymorphism yet normal levels of divergence between species [7–10]. This pattern can only result from natural selection reducing levels of polymorphism at linked neutral sites, because it violates the neutral theory prediction of a strong positive correlation between polymorphism and divergence [5]. However, we have no general genomic description of the physical scale of variation in polymorphism and divergence in Drosophila and how such variation might be related to variation in mutation rates, recombination rates, gene density, natural selection, or other factors. Similarly, although several Drosophila genes have been targets of molecular population genetic analysis, in many cases, these genes were not randomly chosen but were targeted because of their putative association with phenotypes thought to have a history of adaptive evolution [11,12]. Such biased data make it difficult to estimate the proportion of proteins diverging under adaptive evolution. In a similar vein, the unique power of molecular population genetic analysis, when used in concert with genome annotation, could fundamentally alter our notions about phenotypic divergence due to natural selection. This is because our current understanding of phenotypic divergence and its causes is based on a small and necessarily highly biased description of phenotypic variation. Alternatively, a comprehensive genomic investigation of adaptive divergence could use genome annotations to reveal large numbers of new biological processes previously unsuspected of having diverged under selection. Here we present a population genomic analysis of D. simulans. D. simulans and D. melanogaster are closely related and split from the outgroup species, D. yakuba, several million years ago [13–15]. The vast majority of D. simulans and D. yakuba euchromatic DNA is readily aligned to D. melanogaster, which permits direct use of D. melanogaster annotation for investigation of polymorphism and divergence and allows reliable inference of D. simulans–D. melanogaster ancestral states over much of the genome. Our analysis uses a draft version of a D. yakuba genome assembly (aligned to the D. melanogaster reference sequence) and a set of light-coverage, whole-genome shotgun data from multiple inbred lines of D. simulans, which were syntenically aligned to the D. melanogaster reference sequence. Seven lines of D. simulans and one line of D. yakuba were sequenced at the Washington University Genome Sequencing Center (the white paper can be found at http: //www. genome. gov/11008080). The D. simulans lines were selected to capture variation in populations from putatively ancestral geographic regions [16], recent cosmopolitan populations, and strains encompassing the three highly diverged mitochondrial haplotypes previously described for the species [17]. These strains have been deposited at the Tucson Drosophila Stock Center (http: //stockcenter. arl. arizona. edu). A total of 2,424,141 D. simulans traces and 2,245,197 D. yakuba traces from this project have been deposited in the National Center for Biotechnology Information (NCBI) trace archive. D. simulans syntenic assemblies were created by aligning trimmed, uniquely mapped sequence traces from each D. simulans strain to the euchromatic D. melanogaster reference sequence (v4). Two strains from the same population, sim4 and sim6, were unintentionally mixed prior to library construction; reads from these strains were combined to generate a single, deeper, syntenic assembly (see Materials and Methods), which is referred to as SIM4/6. The other strains investigated are referred to as C167. 4, MD106TS, MD199S, NC48S, and w501. Thus, six (rather than seven) D. simulans syntenic assemblies are the objects of analysis. Details on the fly strains and procedures used to create these assemblies, including the use of sequence quality scores, can be found in Materials and Methods. The coverages (in Mbp) for C167. 4, MD106TS, MD199S, NC48S, SIM4/6, and w501, are 56. 9,56. 3,63. 4,42. 6,89. 8, and 84. 8, respectively. A D. yakuba strain Tai18E2 whole-genome shotgun assembly (v2. 0; http: //genome. wustl. edu/) generated by the Parallel Contig Assembly Program (PCAP) [18] was aligned to the D. melanogaster reference sequence (Materials and Methods). The main use of the D. yakuba assembly was to infer states of the D. simulans–D. melanogaster ancestor. For many analyses, we used divergence estimates for the D. simulans lineage or the D. melanogaster lineage (from the inferred D. simulans–D. melanogaster ancestor) rather than the pairwise (i. e. , unpolarized) divergence between these species. These lineage-specific estimates are often referred to as “D. simulans divergence, ” “D. melanogaster divergence, ” or “polarized divergence. ” A total of 393,951,345 D. simulans base pairs and 102,574,197 D. yakuba base pairs were syntenically aligned to the D. melanogaster reference sequence. Several tens of kilobases of repeat-rich sequences near the telomeres and centromeres of each chromosome arm were excluded from our analyses (Materials and Methods). D. simulans genes were conservatively filtered for analysis based on conserved physical organization and reading frame with respect to the D. melanogaster reference sequence gene models (Materials and Methods). We took this conservative approach so as to retain only the highest quality D. simulans data for most inferences. The number of D. simulans genes remaining after filtering was 11,466. Ninety-eight percent of coding sequence (CDS) nucleotides from this gene set are covered by at least one D. simulans allele. The average number of lines sequenced per aligned D. simulans base was 3. 90. For several analyses in which heterozygosity and divergence per site were estimated, we further filtered the data so as to retain only genes or functional elements (e. g. , untranslated regions [UTRs]) for which the total number of bases sequenced across all lines exceeded an arbitrary threshold (see Materials and Methods). The numbers of genes for which we estimated coding region expected heterozygosity, unpolarized divergence, and polarized divergence were 11,403,11,439, and 10,150, respectively. Coverage on the X chromosome was slightly lower than autosomal coverage, which is consistent with less X chromosome DNA than autosomal DNA in mixed-sex DNA preps. Variable coverage required analysis of individual coverage classes (n = 1–6) for a given region or feature, followed by estimation and inference weighted by coverage (Materials and Methods). The D. simulans syntenic alignments are available at http: //www. dpgp. org/. An alternative D. simulans “mosaic” assembly, which is available at http: //www. genome. wustl. edu/, was created independently of the D. melanogaster reference sequence. One of the main goals of large-scale investigations of sequence divergence is to characterize the many biological factors influencing variation in substitution rates throughout the genome. Most analyses of Drosophila data focus on variation in functional constraints or directional selection as the main cause of heterogeneity in substitution rates across genes or functional elements [20,21]. However, the available data have been too sparse to detect any patterns of increasing or decreasing divergence along chromosome arms. Centromere proximal regions tend to be more divergent than distal regions (Figure 1, Figure S4, and Table S5). This pattern is more consistent for D. simulans than for D. melanogaster. Proximal euchromatic regions tend to have lower inferred ancestral GC content compared with distal regions of chromosome arms (Figure S4 and Table S5), which is consistent with the observation that D. simulans divergence was negatively correlated with inferred ancestral GC content (Materials and Methods) (50-kb windows, Spearman' s ρ = −0. 23, p = 1. 4 × 10−26) [30]. The correlation between ancestral GC content and divergence was much weaker and only marginally significant for D. melanogaster (Spearman' s ρ = −0. 05, p = 0. 03). However, while chromosomal gradients of divergence were observed for most chromosome arms (Figure S4 and Table S5), inferred ancestral GC content tends to show a less-consistent pattern. For example, some arms showed a more U-shaped distribution, with euchromatic regions near centromeres and telomeres tending to have higher estimated ancestral GC content (Figure S5). More proximal and distal regions also tend to have reduced crossing-over [39], which is consistent with the observation that inferred ancestral GC content is negatively correlated with cM/kb (Materials and Methods) on the X chromosome (Spearman' s ρ = −0. 33, p = 0. 0002) [59], the only chromosome arm for which we investigated correlates of recombination rate variation (see below). The neutral model of evolution predicts that gradients of divergence along chromosome arms are explained by gradients of functional constraint or mutation rates. For example, higher divergence in regions near centromeres could be explained if such regions harbor a lower density of functional elements (e. g. , genes). However, with the exception of chromosome arm 2L (Spearman' s ρ = −0. 19, p = 6 × 10−5), variation in coding sequence density (CDS bases per 50-kb window) showed no significant chromosomal proximal–distal trend, suggesting that variation in constraint that is associated with coding density plays, at best, a small part in explaining chromosomal gradients of divergence. More generally, the expectation of a negative correlation between coding density and nucleotide divergence in D. simulans was not met. This seemingly counterintuitive result probably reflects the fact that exons constitute a relatively small fraction of the genome and were not dramatically less diverged (0. 016) compared with intergenic DNA (0. 027). If proximal–distal gradients of decreasing divergence along chromosome arms result from variation in mutation rates, then the neutral theory predicts that we should observe similar gradients of polymorphism. This is the case for some chromosome arms but not others (Figure 1 and Table S5), after regions of reduced πnt in the most distal/proximal regions are excluded (Materials and Methods; this result is robust to variation in the extent of proximal and distal chromosomal regions removed from the analysis). Thus, variable neutral mutation rates alone is an insufficient explanation for the overall genomic patterns of variation. Below we address the possibility that recombination rate variation contributes to variation in D. simulans πnt and divergence across chromosome arms. There was considerable variance of polymorphism and divergence across chromosome arms, even when regions of severely reduced heterozygosity near centromeres and telomeres were excluded. Figure 1 clearly shows that variance in polymorphism and divergence is not randomly arranged, but rather appears to be spatially structured on the scale of several tens of kilobases. These qualitative visual assessments were supported by significant statistical autocorrelations (Materials and Methods) for nucleotide heterozygosity and divergence across all chromosome arms (Table S6) [60]. Furthermore, the strength of this autocorrelation appeared to differ across arms, because X and 3L show evidence of stronger correlations over longer distances (Figure 1). The strength of autocorrelation is consistently higher for heterozygosity than for divergence. Under the neutral theory, fluctuations in polymorphism and divergence could be the result of variation in gene density, with windows that have more exons per kb showing lower polymorphism and divergence. This expectation was not met. Indeed, for 50-kb autosome windows (but not X-linked windows), divergence is positively correlated with coding density (Spearman' s ρ = 0. 12, p < 0. 0001). This is consistent with an important role of directional selection on coding sequence to genome divergence, a point we will revisit in several analyses below. In contrast to the positive correlation between coding density and divergence, we found a negative correlation between coding density and D. simulans πnt (autosome Spearman' s ρ = −0. 10, p < 0. 0001; X Spearman' s ρ = 0. 29, p < 0. 0001). Overall, the contrasting correlations between coding density and polymorphism versus divergence suggest that directional selection in exon-rich regions generates greater divergence and reduced polymorphism due to hitchhiking effects [3,6, 61]. The analyses presented above, especially for the X chromosome data, strongly suggest that hitchhiking effects contribute to shaping patterns of polymorphism across the D. simulans genome. To provide a more quantitative assessment of the physical extent, magnitude, and biological basis of these hitchhiking effects, we carried out a genomic analysis of polymorphism and divergence in the context of the Hudson-Kreitman-Aguade (HKA) test [2] (Materials and Methods). The analysis should be thought of as a method for identifying unusual genomic regions rather than as a formal test of a specific model, since our data violate the assumptions of the simple neutral model (neutral alleles sampled from a single, equilibrium, panmictic population). The results (Figure 1, Datasets S6, S16–S20) statistically support our earlier contention and previous reports [7,8, 10,34,36], that Drosophila chromosomes show greatly decreased polymorphism, relative to divergence, in both telomere- and centromere-proximal regions. The fact that corrected X chromosome heterozygosity was not significantly different from autosomal heterozygosity, although X chromosome divergence was significantly higher than autosomal divergence, supports a role for hitchhiking effects reducing nucleotide variation on the X chromosome. Our previously mentioned result, that coding density is positively correlated with divergence and negatively correlated with polymorphism, suggested that hitchhiking effects of directional selection are more common in exonic sequence. The HKA-like analysis supports this contention. We identified regions of the genome that had either two or more consecutive, nonoverlapping 10-kb windows with p < 1 × 10−6 or four such windows with p < 0. 01. The number of coding nucleotides per 10 kb in these “hitchhiking windows” (n = 378 windows, mean coding density = 2,980 bp) was much higher than coding density in other windows (n = 9,329, mean coding density = 1,860 bp) (Mann-Whitney U, p < 0. 0001). An alternative hypothesis for the strong correlation between recombination and polymorphism and the high density of coding sequence in regions showing reduced heterozygosity-to-divergence ratios is background selection, a phenomenon whereby the removal of deleterious mutations reduces polymorphism at linked sites [1]. To address this possibility, we calculated Fay and Wu' s H [56] for 10-kb windows across the genome using only sites with a coverage of five alleles and windows not located in extended regions of reduced heterozygosity near the distal and proximal ends of chromosome arms (Materials and Methods). Hitchhiking effects of beneficial mutations are expected to cause an excess of high-frequency derived alleles (and a more-negative H statistic) relative to neutral theory predictions, while background selection predicts no such excess [1,72]. We compared the average H statistic for regions of the genome showing four or more consecutive 10-kb windows with an HKA-like test of p < 0. 01 versus 10-kb windows from the rest of the genome. For each chromosome arm, the H statistic was significantly more negative in windows showing a reduced heterozyogsity-to-divergence ratio (Mann Whitney U, p < 10−4 for each arm), which strongly supports the proposition that hitchhiking effects of beneficial variants is a major cause of the fluctuations in heterozygosity across the genome. Note, however, that this analysis does not rule out a contribution of background selection [1]. Several factors can generate lineage differences in divergence. For example, higher divergence in a lineage (relative to the lineage of its sister species) could be due to higher mutation rates, shorter generation times, or stronger directional selection. Investigating which classes of mutations or functional elements tend to show different levels of divergence in two lineages can inform our understanding of the causes of rate variation. Previously collected data from coding regions suggest that D. melanogaster evolves faster than D. simulans [89,90]. We found a similar pattern in that dN and dS are greater in D. melanogaster (median = 0. 0045 and 0. 0688) than in D. simulans (median = 0. 0036 and 0. 0507) (Table 1 and S3). This pattern has been interpreted as reflecting the reduced efficacy of selection against slightly deleterious variants in D. melanogaster, supposedly resulting from its smaller effective population size relative to D. simulans [89]. However, a different pattern is observed on a genome-wide scale, as median D. simulans divergence (50-kb windows; 0. 025), though only slightly greater than D. melanogaster (50-kb windows; 0. 022), is consistently greater across a large proportion of windows (Wilcoxon sign rank test, p = 1. 8 × 10−275). We consider the genomic faster D. simulans finding as provisional due the potential biases associated with D. melanogaster-centric alignments. For example, genomic regions that are evolving quickly only in D. melanogaster may drop out of the D. melanogaster–D. yakuba alignment, whereas regions evolving quickly only in D. simulans may be retained because of the relatively short D. melanogaster–D. simulans branch. Analysis of rate variation across site types (Table 1 and Table S3) reveals a more complex pattern. For example, D. simulans shows greater divergence than D. melanogaster for intergenic, intron, and 3′ UTR sites, whereas D. melanogaster shows greater divergence than D. simulans for 5′ UTRs, nonsynonymous sites, and synonymous sites. A decades-old issue in population genetics is the extent to which directional selection determines protein divergence. Several analytic strategies for investigating the prevalence of adaptive protein divergence between closely related species have been proposed (reviewed in [91]). Here we focused on two approaches. First, we used comparisons of synonymous and nonsynonymous polymorphic and fixed variants in individual genes to test the neutral model. Second, we identified proteins that show very different divergence estimates in D. simulans versus D. melanogaster. The same logic originally proposed in the MK test using nonsynonymous and synonymous variation can be extended to any setting in which variant types can be categorized, a priori. We tested variation in individual noncoding elements (introns, UTRs, and intergenic sequences) relative to variation at tightly linked synonymous sites (Materials and Methods) using the same criteria described for the MK tests; we present only polarized analyses (Datasets S2–S5). The proportion of tests (Materials and Methods) that rejected (p < 0. 05) the null model for 5′ UTR, 3′ UTR, intron, and intergenic sites are 0. 13,0. 13,0. 12, and 0. 17, respectively. However, unlike the case for the nonsynonymous versus synonymous polarized MK tests, of which only 6% of the significant tests deviated in the direction of excess polymorphism (relative to synonymous sites), a much greater proportion of noncoding MK tests deviated in this direction—0. 13,0. 24,0. 28, and 0. 28 for 5′ UTR, 3′ UTR, intron, and intergenic regions, respectively. Thus, the proportion of noncoding elements showing evidence of adaptive evolution for 5′ UTR, 3′ UTR, intron, and intergenic sites is 0. 12,0. 10,0. 08, and 0. 12, respectively, which is similar to the proportion of coding sequences inferred (by polarized MK tests) to be under direction selection (0. 14). It would be tempting to conclude from this result that intergenic variants are as likely to be under directional selection as nonsynonymous variants. However, such an interpretation ignores the fact that the number of variants per element for each MK test is much greater for intergenic sequence (median = 87) compared to the numbers for coding regions (median = 42), 5′ UTRs (median = 34), 3′ UTRs (median = 35), or introns (median = 64). Thus, there is more power to reject the neutral model for intergenic sequence and introns than for exonic sequence. The fact that MK p-values are significantly negatively correlated with the total number of observations per test is consistent with this explanation. There was no evidence of different proportions of significant versus nonsignificant tests for X-linked versus autosomal elements. Tables S22–S24 report data from the ten most highly significant MK tests (average coverage > 2) indicative of directional selection on 5′ UTRs, 3′ UTRs, and intron sequences, respectively. Among the most unusual 5′UTRs are those associated with genes coding for proteins associated with the cytoskeleton or the chromosome, categories that also appeared as unusual in the MK tests on protein variation. Two of the top-ten 3′ UTRs are associated with the SAGA complex, a multi-subunit transcription factor involved in recruitment of RNA Pol II to the chromosome [111]. Among the extreme introns, two are from genes coding for components of the ABC transporter complex and two are from genes coding for centrosomal proteins, again pointing to the unusual evolution of genes associated with the cytoskeleton and chromosome structure and movement. As previously noted, a large number of significant UTRs deviate in the direction of excess polymorphism (relative to synonymous mutations). Given the potential importance of the UTRs in regulating transcript abundance and localization, translational control, and as targets of regulatory microRNAs [112], such UTRs could be attractive candidates for functional investigation. Contingency tests of significant versus nonsignificant MK test for amino acids versus each of the noncoding elements yielded p-values of 0. 65,0. 04, and 0. 07 for 5′ UTRs, 3′ UTRs, and introns, respectively. Thus, there is weak evidence that genes under directional selection on amino acid sequences tend to have 3′ UTRs and introns influenced by directional selection as well. Up to this point, our analyses have investigated various attributes of polymorphism and divergence based on windows or genes. An alternative approach for understanding the causes of variation and divergence is to analyze polymorphism and divergence across site types. Table 2 shows whole-genome counts of polymorphic and polarized fixed variants for UTRs, synonymous sites, nonsynonymous sites, introns, and intergenic sites. We also provide data for polarized, synonymous preferred or unpreferred variants. Almost all preferred versus unpreferred codons in D. melanogaster end in GC versus AT, respectively [113]; thus, preferred versus unpreferred codons can be thought of as GC-ending versus AT-ending codons. Nonsynonymous sites showed the smallest ratio of polymorphic-to-fixed variants, which is consistent with the MK tests and supports the idea that such sites are the most likely to be under directional selection. Nonsynonymous polymorphisms also occur at slightly lower frequency than do noncoding variants (Table S25). Synonymous sites have the highest ratio of polymorphic-to-fixed variants, which supports the previously documented elevated ratio of polymorphic-to-fixed unpreferred synonymous variants in D. simulans [89]. The confidence intervals of the ratio of polymorphic-to-fixed variants among site types are nonoverlapping with the exception of intron and intergenic sites. If preferred synonymous mutations are, on average, beneficial [89,114], then the smaller polymorphic-to-fixed ratio for nonsynonymous and UTR variants versus preferred variants implies that a large proportion of new nonsynonymous and UTR mutations are beneficial. Using similar reasoning, the data in Table 2 suggest that directional selection plays a larger role in nonsynonymous and UTR divergence compared to intron and intergenic divergence [20,115,116]. These conclusions are consistent with estimates of α [11,117], the proportion of sites fixing under directional selection (assuming that synonymous sites are neutral and at equilibrium) for different site types. Determining the relative contributions of various mutational and population genetic processes to base composition variation and inferring the biological basis of selection on base composition remain difficult problems. Much of the previously published data on base composition variation in D. simulans have been from synonymous sites [55,89,90,118]. Several lines of evidence [55,89,90,113,118] suggest that on average, preferred codons have higher fitness than unpreferred codons, with variation in codon usage being maintained by AT-biased mutation, weak selection against unpreferred codons, and genetic drift [23,114]. However, the possibilities of nonequilibrium mutational processes and/or natural selection favoring different base composition in different lineages have also been addressed [119]. The D. simulans population genomic data allow for a thorough investigation of the population genetics and evolution of base composition for both coding and noncoding DNA [59,120]. The analyses discussed below use parsimony to polarize polymorphic and fixed variants. Complete genomic and gene-based data are available as Datasets S9 and S10. The genomic analysis of polymorphism and divergence based on alignments to a reference sequence is poised to become a central component of biological research. Here we have demonstrated that such analysis can be based on high-quality whole-genome syntenic assemblies from light shotgun sequence data; accounting for variable coverage and data quality is fundamentally important. Several, noteworthy new results have been reported here. First, our genome-level investigation of adaptive protein evolution has revealed a large number of proteins and biological processes that have experienced directional selection, setting the stage for a general analysis of functional protein divergence under selection in Drosophila. Second, we identified several UTRs, introns, and intergenic sequences showing the signature of adaptive evolution. The functional biology of such noncoding elements and their connections to adaptive protein and gene expression evolution is open to investigation. Third, D. simulans populations exhibit large-scale chromosomal patterning of polymorphism and divergence that is poorly explained by current genome annotations. Variation in recombination rates across chromosomes may contribute to these patterns. Fourth, the population genetics of the X chromosome differs in several ways from that of the autosomes. It evolves faster, harbors less polymorphism, and shows a different spatial scale of variation of polymorphism and divergence compared to the autosomes. Finally, base composition is evolving in both coding and noncoding sequences, for reasons that are as of yet unclear. This project is, in many ways, a first step toward population genomics in general, and in the Drosophila model specifically. For example, the average number of alleles sampled per base is too small for investigating many interesting properties of variation. Some genomic regions have been excluded due to low coverage, their repetitive nature, or very high divergence from D. melanogaster. Many aspects of biological annotation have not been investigated here, and many new Drosophila annotations will be produced in the near future as comparative and functional annotations of the D. melanogaster genome move forward. Nevertheless, the data are a stunningly rich source of material for functional and population genetic investigation of D. simulans polymorphism and divergence. It will be interesting to compare the processes determining polymorphism and divergence in D. simulans to those controlling variation in D. melanogaster (http: //www. dpgp. org) and in other species, such as humans. Such comparisons are likely to result in new insights into the genetic, biological, and population genetic factors responsible for similarities and differences among species in the genomic distribution of sequence variation. D. simulans 4 (males and females). This strain was established by ten generations of sibling mating from a single, inseminated female collected by D. Begun in the Wolfskill orchard, Winters, California, USA, summer 1995. D. simulans 6 (males and females). This strain was established by ten generations of sibling mating from a single, inseminated female collected by D. Begun in the Wolfskill orchard, Winters, California, summer 1995. D. simulans w501 (males and females). This strain carries a white (eye color) mutation. It has been in culture since the mid 20th century, likely descended from a female collected in North America. The strain used for sequencing was sib-mated for nine generations by D. Barbash at UC Davis. Libraries for sequencing were prepared from DNA isolated from embryos. D. simulans MD106TS (males and females). This strain was descended from a single, inseminated female collected by J. W. O. Ballard in Ansirabe, Madagascar on 19 March 1998. It has the siII mitochondrial genotype, and was cured of Wolbachia by tetracycline. The strain was sib-mated for five generations in the Ballard lab, followed by an additional five generations of sib-mating by D. Begun. D. simulans MD199S (females). This strain was descended from a single, inseminated female collected by J. W. O. Ballard in Joffreville, Madagascar on 28 March 1998. It has the siIII mitochondrial genotype, and probably lost Wolbachia infection. The strain was sib-mated for five generations in the Ballard lab, followed by an additional five generations of sib-mating by D. Begun. All-female DNA was made to assist in assembly of the Y chromosome by comparison to mixed-sex libraries of other lines. D. simulans NC48S (males and females). This strain was descended from a collection by F. Baba-Aissa in Noumea, New Caledonia in 1991. It has the siI mitochondrial genotype, and was sib-mated for five generations in the Ballard lab, followed by an additional five generations of sib-mating by D. Begun. D. simulans C167. 4 (males and females). This strain was descended from a collection in Nanyuki, Kenya. It is unusual in that can produce fertile females when hybridized to D. melanogaster. The line used for genome project was obtained from the Ashburner laboratory via D. Barbash, and was subjected to a total of 13 generations of sib- mating. D. yakuba Tai18E2 (males and females). This strain derives from a single inseminated female captured in 1983 by D. Lachaise in the Taï rainforest, on the border of Liberia and Ivory Coast. This line was sib-mated for ten generations by A. Llopart and J. Coyne. Inspection of 21 salivary gland polytene chromosomes showed no chromosomal rearrangements segregating within the strain. Therefore, Tai18E2 appears homokaryotypic for the standard arrangement in all chromosome arms, save 2R, which is homokaryotypic for 2Rn. DNA preparations for sequencing all lines except w501 and Tai18E2 were made from adults. Drosophila nuclei were isolated following Bingham et al. [121]. For all lines except w501 and Tai18E2, DNA was isolated by phenol/chloroform extraction of nuclei followed by ethanol precipitation. For lines w501 and Tai18E2, embryos were collected using standard procedures [122] followed by DNA isolation on CsCl gradients [121]. Sequence data were obtained from paired-end plasmids and fosmids (Table S32) using standard Washington University Genome Sequencing Center laboratory protocols (http: //genome. wustl. edu). A highly automated production pipeline using a 384-well format ensured the integrity of the paired-end data. We determined the nucleotide-level accuracy by reviewing the quality values of the D. yakuba consensus assembly and by comparison to manually edited D. yakuba sequence. More than 97% of the D. yakuba genome sequence had quality scores of at least 40 (Q40), corresponding to an error rate of ≤10−4. Further, we extracted reads from two local fosmid-sized regions (68 kb, defined by fosmid-end sequence pairs, one on chromosome X and one on chromosome 3L) of the assembly and reassembled using Phrap (P. Green, unpublished data). The resulting “fosmids” were manually reviewed and edited. Comparison of the sequence to these manually edited regions revealed a high-quality discrepancy rate of 2 × 10−4 substitutions and 1 × 10−4 insertion/deletion errors, consistent with the above estimates based on consensus base quality. We also found no evidence of misassembly when comparing the WGS assembly to these projects. Repetitive content was estimated both in D. yakuba and D. melanogaster using RECON to generate the repeat families and RepeatMasker to then identify those repeats in the genomes. The D. yakuba genome was ∼27% repetitive overall (of which ∼2. 5% is simple sequence repeats/low complexity sequence) and 8% in the euchromatic portion of the genome. The D. melanogaster genome was ∼11% repetitive overall (of which 2. 3% is simple sequence repeats/low complexity sequence) and ∼7% in the euchromatic portion of the genome. The first step in creating D. yakuba chromosomal fasta files was to align the D. yakuba WGS assembly data against the D. melanogaster genome. D. yakuba supercontigs were artificially broken into 1,000-bp fragments and aligned against the D. melanogaster genome using BLAT [123]. An alignment was defined as “unique” if its best scoring match had a score of at least twice that of its next best scoring alignment. Of the 139. 5 Mb of D. yakuba supercontigs that uniquely aligned to the D. melanogaster genome (4. 2 Mb of which aligned uniquely to D. melanogaster unlocalized sequence, chrU), only 16 supercontigs totaling 15. 1 Mb contained unique assignments to more than one chromosome arm. Eleven of these involved alignments to heterochromatin where only less than ∼5% of the supercontig aligned uniquely to the D. melanogaster genome. These contigs were assigned to either chrU or the heterochromatic portion of the chromosome for cases where the contig aligned to both the heterochromatic and nonheterochromatic portion of the same chromosome. One 200-kb contig had only 6. 2 kb that uniquely mapped to the D. melanogaster genome, 3. 8 kb mapping to chr2R, and 2. 4 kb mapping to chrX. This supercontig was assigned to chrU. The remaining four supercontigs were alignments to chromosome arms 2L and 2R, the location of a known pericentric inversion between D. melanogaster and D. yakuba. The D. yakuba contigs were initially ordered by their position along the assigned D. melanogaster chromosomes. Because there are rearrangements in D. yakuba as compared to D. melanogaster, we allowed one portion of a D. yakuba supercontig to align to one region of a chromosome and the remaining portion to align elsewhere along that chromosome. For example, four supercontigs aligned to both chromosome arms 2L and 2R. However, these 2L/2R cross-overs and other interspecific nonlinearities are expected given the known chromosome inversions [124] between D. yakuba and D. melanogaster. This initial ordering for 2L, 2R, 3L, 3R, and X was used as the starting point for manually introducing inversions in the D. melanogaster-ordered D. yakuba supercontigs. The goal was to minimize the total number of inversions required to “rejoin” all D. yakuba supercontigs previously assigned to distant chromosomal regions based on D. melanogaster alignments (L. Hillier, unpublished data). Inversions were only introduced between contigs and not within contigs. Using this process, we created the final chromosomal D. yakuba sequence. Sequence data were obtained from paired-end plasmids from the various D. simulans strains using standard laboratory protocols (http: //genome. wustl. edu). A genomic assembly was also created. We began by generating an ∼4× WGS assembly of D. simulans w501 using PCAP [18]. The w501contigs were initially anchored, ordered, and oriented by alignment with the D. melanogaster genome in a manner similar to that described above for alignments between the D. yakuba and D. melanogaster genome. The assembly was then examined for places where the w501 assembly suggested inversions with respect to the D. melanogaster assembly. One major inversion was found, confirming the already-documented inversion found by [124]. Six other D. simulans lines (C167. 4, MD106TS, MD199S, NC48S, SIM4, and SIM6) were also assembled using PCAP with ∼1× coverage. Using the 4× WGS assembly of the D. simulans w501 genome as a scaffold, contigs and unplaced reads from the 1× assemblies of the other individual D. simulans lines were used to cover gaps in the w501 assembly where possible. Thus, the resulting assembly is a mosaic containing the w501 contigs as the primary scaffolding, with contigs and unplaced reads from the other lines filling gaps in the w501 assembly (L. Hillier, unpublished data). The D. simulans w501 whole-genome shotgun assembly can be accessed at GenBank. The goal was to align unique D. melanogaster reference sequence assembly v4 to orthologous D. simulans sequence. The D. melanogaster genome was preprocessed to soft mask all 24mers that were not unique, as such sequences were not expected to have a discriminating effect during mapping of D. simulans reads. Transposable elements in the reference sequence were also masked. The D. simulans WGS reads were quality trimmed prior to assembly based on their phred-score derived error probability. These error probabilities were used to trim the read back to the longest contiguous interval with an average probability of error less than 0. 005. Each end was then examined and trimmed until its terminal 10 bp had an average probability of error less than 0. 005. If the read was less than 50 bp after this process, it was discarded. These criteria resulted in 164,480 discarded reads from a total 2,424,141 reads. See Table S33 for read and trim statistics. A dynamic programming algorithm was used to create a maximum-likelihood description of the evolutionary path between sequences from the two species with respect to the standard alignment model, which was extended to incorporate the possibility of sequencing error. To improve the accuracy of the alignments, optimal parameters were estimated with respect to the overall expected evolutionary distance between the two species. This was done from a first-pass assembly using the method described in [129]. Because dynamic programming is not feasible on a genomic scale, we determined candidate locations for each read using the MegaBLAST (http: //www. ncbi. nlm. nih. gov/BLAST/docs/megablast. html) algorithm. A read was then realigned to each candidate location as a single contiguous alignment using a derivative of the Smith-Waterman algorithm, which was adapted to incorporate the expected divergence and the error probabilities provided by Phred quality scores. Alignments were ranked by score. Reads were considered unambiguously mapped if their alignment covered at least 90% of the sequence and showed more than two high-quality differences between the putative best orthologous location and a possible secondary candidate location. Reads were incorporated into the assembly on a clone-by-clone basis only if both mate-pairs were unambiguously mapped with the proper orientation and appropriate distance from each other. For each D. simulans line, the aligned reads were coalesced into syntenic contigs using their overlap with respect to the D. melanogaster genome. Note that “overhanging” or unaligned sequence that may represent transposable elements, other repetitive sequence, or highly diverged sequence, was not considered. This “master–slave” multiple alignment contains reads that are aligned “optimally” with respect to the D. melanogaster reference sequence. However, this does not ensure that the reads are optimally aligned with respect to each other. For instance, small, identical insertion or deletion variants may not be mapped to precisely identical locations in all D. simulans reads. To address this problem, the D. melanogaster reference sequence was set aside, and the method of Anson and Meyers [125] was used to optimize the alignment of each component read of each D. simulans line with respect to a D. simulans–only consensus sequence. This method, which minimizes the sum of differences between each of the reads and the consensus sequence, belongs to the class of expectation maximization algorithms [125]. The round robin, align-and-update algorithm converges on a consensus sequence and alignment that most parsimoniously describe the differences between each read and the consensus. This has the effect of coalescing deletions and aligning insertions. The end result of the assembly is a multi-tiered alignment with associated quality scores for (i) the trimmed reads, (ii) the assembled sequences within lines, and (iii) a species consensus sequence, all aligned to the D. melanogaster reference sequence. A reference sequence was produced for each D. simulans line by concatenating the syntenically assembled contigs that were padded with respect to the D. melanogaster reference sequence. The result is a set of D. simulans genomes onto which D. melanogaster annotation can be directly mapped. Nine regions, including coding and noncoding DNA, were chosen to cover a range of polymorphism levels as predicted by an early version of the syntenic assembly. These regions were amplified from lines C167. 4, MD106TS, NC48S, and w501, and sequenced at UNC-Chapel Hill High-Throughput Sequencing Facility. Sequences were assembled using Consed; a minimum quality score of 30 was required. Approximately 27,500 bp were sequenced per line. The per-base discrepancy between these sequences and the current syntenic assembly (insertions, deletions, and masked bases omitted) was estimated as 0. 00043. An orthology map (with respect to the D. melanogaster reference sequence) of D. yakuba assembly (v1. 0) was generated by the Mercator program (http: //rd. plos. org/pbio. 0050310a). The MAVID [126] aligner was run on each orthologous segment in the map. MAVID uses protein-coding hits reported by Mercator to anchor its alignment of each segment. It recursively finds additional anchors and then runs the Needleman-Wunsch algorithm in between the anchors to obtain a single, global alignment of the entire orthologous segment. These regions were filtered based on manual examination of the density of annotated repetitive sequence in the centromere and telomere proximal regions of the five large arms. The transition from the “typical” euchromatic density of large repeats to the typical “beta heterochromatic” pattern is obvious. The “euchromatic/heterochromatic boundaries” were drawn roughly at the edges of the first annotated gene within each euchromatic arm. The following regions were excluded from analysis: (i) X 1 to 171944 AND 19740624 to END; (ii) 2L 1 to 82455 AND 21183096 to END; (iii) 2R 1 to 2014072 AND 20572063 to END; (iv) 3L 1 to 158639 AND 22436835 to END; and (v) 3R 1 to 478547 AND 27670982 to END. The sequence for each line is derived from the multiple alignment of reads to the D. melanogaster reference assembly (v4). For each line and each column (nucleotide position) corresponding to a D. melanogaster base, a likelihood model was used to determine the quality score for each of the four bases. The quality score was calculated as −10log (1 – probability base is correct). To compute the probability a base call is correct, we assume that each read is an observation of a random variable with equal likelihoods for all four bases with some probability of error. From the definition of a phred score, the probability of error for a particular observed call is: 10 (phred score/–10). We assumed that a base in error is equally likely to be any one of the three other bases. Then, for a given position A, Bayes theorem implies the probability (Pr) that the call is correct is Where Pr[A] = 1/4, Pr[Observations|A is correct] = likelihood of A observations being correct and non-A observations being incorrect, and Pr[Observations] = likelihood of seeing observed values given frequency and error rates. Quality scores were truncated at 90. The sequences for each line were investigated for regions containing unusually high densities of high-quality discrepancies, which are due to residual heterozygosity, duplication, and erroneous sequence. These regions were filtered from subsequent analysis (see below). For each line, the support for each alternative (A, G, C, and T) at each aligned base was the sum of the qualities, with the highest quality base assigned as the base for that line/position. Implicit in this approach is that a base is called only if the highest quality base has a quality score that is 30 or more greater than that of the next highest quality base. The combined SIM4/SIM6 consensus was also treated in this manner. Residual heterozygosity within lines or duplications present in D. simulans but not D. melanogaster can lead to regions with excess high-quality discrepancies between reads within lines. We refer to these as single-nucleotide discrepancies. We derived a distribution of the number of discrepancies per site over each chromosome for each D. simulans line. We based the distributions on counts of within-line discrepancies per site in 500-bp windows that had 250-bp overlap. We took the conservative approach of filtering windows in all the lines that fell into the top 0. 5% of the distribution in any single line. In other words, a window with a high-quality discrepancy in one line was filtered from the entire dataset, even if the other lines had no discrepancy. Overall, 334,500 base pairs were filtered from the genome. The number of sites filtered for each chromosome arm were 39 kb for 2L, 86. 5 kb for 2R, 58 kb for 3L, 73 kb for 3R, and 78 kb for X. One large inversion on chromosome arm 3R distinguishes the two species. Phylogenetic analysis of the cytogenetic data suggested that the inversion fixed in the D. melanogaster lineage [39]. Thus, D. simulans 3R is rearranged with respect to the D. melanogaster reference sequence. We used D. melanogaster/D. simulans alignments provided by the UC Santa Cruz Genome Browser to locate the putative breakpoints of the inversion as Chr3R: 3874907 and 17560827. All features were defined in the D. melanogaster v4. 2 annotation (http: //flybase. org). For each gene, the longest isoform (i. e. , the isoform the with greatest number of codons) was chosen for analyses. Exons that were not part of the longest isoform were excluded from all feature-based analyses, but were included in window analyses. The analyzed introns correspond to these longest isoforms; all introns were included in window analyses. Intronic sequences within annotated UTRs or that overlapped any coding sequence were excluded. UTRs investigated for this paper were restricted to those inferred from “Gold Collection” genes with completely sequenced cDNAs (http: //www. fruitfly. org/EST/gold_collection. shtml). All annotated CDS sequences were used regardless of the associated empirical support. Intergenic regions were defined as noncoding segments between annotated genic regions (UTRs, coding sequence, and noncoding RNAs) regardless of strand. Defined intergenic regions from v4. 2 annotation were checked against all known coding and UTR coordinates; any nucleotides that overlapped a genic region were removed from the intergenic set before analysis. We established a conservative gene set for analyses (base composition analyses excepted) by including only genes for which the start codon (ATG or otherwise), splice junctions (canonical or otherwise), and termination codon position agreed with the D. melanogaster reference sequence. We took the conservative approach of excluding from the gene-based analysis any gene for which any of the six D. simulans gene models disagreed with the D. melanogaster gene model. Long insertions and deletions (indels) are difficult to identify using only aligned reads. As indel length increases, the likelihood that indels are missed increases because they are either too long or occur near the end of a read, which compromises alignment. Furthermore, indel error probabilities are difficult to estimate. These considerations led us to restrict our analysis to indels of 10 bp or less and to restrict our analysis of divergence to the D. simulans versus D. melanogaster comparison. Variants were classified as insertions or deletions relative to the D. melanogaster reference sequence. The quality score for an insertion was the average quality score of sequence in that insertion; the quality score for a deletion was the minimum of qualities of the two flanking nucleotides. Qualities were determined this way to provide a metric of overall sequence quality in the region of a putative indel, thereby allowing a quantitatively defined cutoff for inferring indel variants; only indels of high quality (over phred 40) were considered in the analysis. Light, variable coverage of each line requires that statistical estimation and inference account for coverage variation. When appropriate (e. g. , contingency tables of frequency variation), counts of variants within a coverage category were used. In other estimation and inference settings, the familiar estimators were applied to each coverage class and then averaged, weighting by the proportion of total covered base pairs in the window or other feature. Heterozygosity. The expected nucleotide, insertion, and deletion heterozyogsity was estimated as the average pairwise differences between D. simulans alleles as follows: πi is the coverage-weighted average expected heterozygosity of nucleotide variants (i = nt), deletions (i =Δ) or insertions (i = ▿) per base pair. “Expected heterozygosity” assumes the six sequenced genomes were drawn from a single, randomly mating population. Variable coverage over sites led us to extend the typical calculation of expected heterozygosity [127,128] to the following: where nc is the number of aligned base pairs in the genomic region (e. g. , gene feature or window) with sequencing coverage c. kcj is the number of sites in this region with coverage c at which the derived state (nt, ▵, or ▿) occurs in j out of the c sequences. This estimator was used for 10-kb windows, 50-kb windows, 30-kb sliding windows (10-kb increments), 150-kb sliding windows (10-kb increments), and 210-kb windows (10-kb increments), including all windows for which coverage was >200 bp. Expected heterozygosity was also estimated for genomic features (exons, introns, UTRs, and intergenic sequence) that had a minimum size and coverage [i. e. , n (n – 1) × s ≥ 100, where n = average number of alleles sampled and s = number of sites]. For coding regions, the numbers of silent and replacement sites were counted using the method of Nei and Gojobori [129]. The pathway between two codons was calculated as the average number of silent and replacement changes from all possible paths between the pair. The variance of pairwise differences in sliding windows (150-kb windows, 10-kb increments) was used as a method of summarizing the magnitude of linkage disequilibrium across the D. simulans genome. For each window, we calculated coverage weighted variance of the expected heterozygosity (see above) for all pairs of alleles. Divergence. Unpolarized (i. e. , pairwise) divergence between D. simulans and D. melanogaster was estimated for 10-kb windows, 50-kb windows, 30-kb sliding windows (10-kb increments), 150-kb sliding windows (10-kb increments), 210-kb windows (10-kb increments), and genomic feature that had a minimum number of nucleotides represented [i. e. , n × s > 100, with n and s as above in calculations of π. Unpolarized divergence was calculated as the average pairwise divergence at each site, which was then summed over sites and divided by the total number of sites. A Jukes-Cantor [130] correction was applied to account for multiple hits. For coding regions, the numbers of silent and replacement sites were counted using the method of Nei and Gojobori [129]. The pathway between two codons was calculated as the average number of silent and replacement changes from all possible paths between the pair. Estimates of unpolarized divergence over chromosome arms were calculated for each feature with averages weighted by the number of sites per feature. Lineage-specific divergence was estimated by maximum likelihood using PAML v3. 14 [131] and was reported as a weighted average over each line with greater than 50 aligned sites in the segment being analyzed. Maximum likelihood estimates of divergence were calculated over 10-kb windows, 50-kb windows, 30-kb sliding windows (10-kb increments), 150-kb sliding windows (10-kb increments), 210-kb windows (10-kb increments), and gene features (exons, introns, and UTRs). PAML was run in batch mode using a BioPerl wrapper [132]. For noncoding regions and windows, we used baseml with HKY as the model of evolution to account for transition/transversion bias and unequal base frequencies [133]; for coding regions, we used codeml with codon frequencies estimated from the data. Insertion and deletion divergence was calculated as divi, the coverage-weighted average divergence of deletions (i = ▵) or insertions (i = ▿) per base pair. where nc is the number of aligned base pairs in the genomic region (e. g. , gene feature or window) with sequencing coverage c. kcj is the number of sites in this region with coverage c at which the derived state with respect to the D. melanogaster reference sequence (▵ or ▿) occurs in j out of the c sequences. Unpolarized MK tests [4] used D. simulans polymorphism data and the D. melanogaster reference sequence for counting fixed differences. Polarized MK tests used D. yakuba to infer the D. simulans/D. melanogaster ancestral state. For both polarized and unpolarized analyses, we took the conservative approach of retaining for analysis only codons for which there were no more than two alternative states. For cases in which two alternative codons differed at more than one position, we used the pathway between codons that minimized the number of nonsynonymous substitutions. This is conservative with respect to the alternative hypothesis of adaptive evolution. Polymorphic codons at which one of the D. simulans codons was not identical to the D. melanogaster codon were not included. To reduce multiple testing problems, we filtered the data to retain for further analysis only genes that exceeded a minimum number of observations; we required that each row and column in the 2 × 2 table (two variant types and polymorphic versus fixed) sum to six or greater. Statistical significance of 2 × 2 contingency tables was determined by Fisher' s Exact test. MK tests were also carried out for introns and Gold Collection UTRs by comparing synonymous variants in the associated genes with variants in these functional elements. For intergenic MK tests, we used synonymous variants from genes within 5 kb of the 5′ and/or 3′boundary of the intergenic segment. For some analyses, we restricted our attention to MK tests that rejected the null in the direction of adaptive evolution. This categorization was determined following Rand and Kann [134]. Polarized 2 × 2 contingency tables were used to calculate α, which under some circumstances can be thought of as an estimate of the proportion of variants fixing under selection [11]. Bootstrap confidence intervals of α and of the ratio of polymorphic-to-fixed variants for each functional element (Table 2) were estimated in R using bias correction and acceleration [135]. Our approach takes overall rate variation among lineages into account when generating expected numbers of substitutions under the null model and allows for different rates of evolution among chromosome arms (e. g. , a faster-X effect). For example, the number of substitutions for all X-linked 50-kb windows was estimated using PAML (baseml), allowing different rates for each lineage. All D. simulans lines were used, with the estimated substitution D. simulans rate for each window being the coverage-weighted average. This generated an empirically determined branch length of the X chromosome for the average over each of the D. simulans lines (from all three way comparisons with D. melanogaster and D. yakuba) weighted by the number of bases covered. We carried out a relative rate test for windows or features in D. simulans and D. melanogaster by generating the expected number of substitutions for each window/feature/lineage based on the branch length of the entire chromosome in each lineage (PAML) and the coverage of the window/feature in question in each lineage. We then calculated the deviation from the expected number of substitutions as (observed – expected substitutions) 2/expected substitutions for any window/feature/lineage. For each GO term associated with at least five MK tests, we calculated the proportion of significant (p < 0. 05) tests. We then randomly selected n p-values from the set of all MK p-values, where n is the number of tests in the ontology category. We repeated this procedure 10,000 times to get the empirical distribution of the proportion significant p-values for each GO term. The relative rate χ2 for dN was calculated for each gene as described above. Genes showing a significant (p < 0. 05) acceleration of dN in the D. simulans lineage were identified as described in the previous section. The probabilities of observing as many, or more, significant relative rate χ2 tests for dN were determined by permutation as described in the previous section. We retrieved ontology terms associated with genes that fell under windows of interest in linked selection analyses. Then, for each term, we divided the number of instances that the term was represented in the windows of interest by the total number of genes in the genome that are associated with the ontology term. This gave us a proportional representation of each GO term in windows of interest. We compared this proportion for each GO term with the empirical distribution of proportions derived from permuted datasets. For each permuted dataset, we randomly picked a nonoverlapping set of windows that were the same size in numbers of base pairs as the observed windows. Each window was guaranteed to contain at least one gene, given that windows of interest have higher-than-average gene density. We then retrieved the ontology terms associated with the genes under the random set of windows. We next calculated the proportion of each term as described above for the observed windows. We repeated this procedure 1,000 times to obtain an empirical distribution of proportions of each term in random windows. The proportion of each GO term in the original observed windows of interest was compared to this empirical distribution to obtain a probability of observing that proportion of each term in windows of interest. We wanted to know whether ontology terms were clustered in the genome. We tested whether each ontology term was significantly clustered by calculating the coefficient of variation based on occurrence in 1-Mb, nonoverlapping windows and compared that to the coefficient of variation from permuted datasets in which we randomized the locations of genes on each chromosome arm. Genes were assigned to expression categories, with the goal of determining whether certain categories had a greater proportion of significant MK tests for adaptive protein divergence than expected by chance. Two types of data, expressed sequence tag (EST) collections and microarray experiments, were used. Genes associated with EST collections from D. melanogaster (head, ovary, and testis from Flybase and spermatheca from Swanson et al [136]) were assigned to that tissue expression category. Female-mating responsive genes were those defined by microarray experiments [137]. Male- and female-biased genes were defined based on microarray experiments of Parisi et al. [138] and Arbeitman et al. [139]. Male- and female-biased genes from Parisi et al. [138] were obtained directly from their Tables S41 and S42. Arbeitman et al. [139] measured expression over the D. melanogaster life cycle for adult males and females. We averaged expression for each gene over the time points taken for each stage. For example, there were 30 time point measurements during the egg stage; we used the average expression over those 30 time points. We repeated this for larvae, metamorph, adult female, and adult male stages. Each gene was provisionally designated as having biased expression for the stage with the maximum average expression, which we will call the biased stage. For each gene, we calculated the average difference between the biased stage expression value and the other stage expression values. This generated a distribution of differences for each comparison of stages. A gene was finally determined to have biased expression if the average difference between the biased stage and the other stages fell into top half for that stage distribution. This procedure resulted in 592,374,223,466, and 296 stage-biased genes for egg, larvae, metamorph, adult male, and adult female stages, respectively. We calculated the proportion of genes in a group (e. g. , male-biased) that had significant MK tests (p < 0. 05). We used permutation testing to determine whether the proportion of significant polarized MK tests deviating in the direction of adaptive protein evolution exceeded the 95% tail of the empirical distribution, based on 10,000 datasets of randomly selected MK tests, sampled without replacement. We tested whether pairs of proteins that interact with one another were more likely to show evidence of adaptive protein divergence than random pairs of proteins with no evidence of interaction. Data were from Giot et al. [140]. We considered pairs of genes to have a significant interaction if the probability of interaction was greater than 0. 5. We calculated the proportion of interacting pairs where both members had significant evidence of adaptive evolution (MK p-values < 0. 05). We compared this proportion to the distribution of proportion generated from permuted datasets generated by randomly drawing pairs of genes without replacement from the Giot et al. [140] dataset. Hudson, Kreitman, and Aquadé [2] proposed a test of the neutral theory of molecular evolution in which the numbers of polymorphic and (fixed) divergent sites are contrasted between two independent loci (genomic regions). The distribution of a χ2-like test statistic can be determined by simulation for any assumed values of recombination within each locus. However, given the small sample size here and the genomic scale of the data, we used an analogous statistic for polymorphisms and fixations on the D. simulans lineage in various sizes of sliding windows, combined over coverage classes. First, the average proportion of segregating sites in D. simulans and parsimony-inferred fixed differences for each chromosome arm in D. simulans was determined for each coverage class over a range of sliding window sizes (10 kbp to 510 kbp). The test statistic is a simple two-cell χ2, in which the observed numbers (summed over coverage classes) of segregating and fixed sites are contrasted with their expected numbers (summed over coverage classes, the chromosome arm average for each coverage class times the total numbers of segregating and fixed sites in that class). Only sites for which unambiguous, parsimony-inferred D. simulans/D. melanogaster ancestral states could be determined were included in the analysis. In a number of figures, χ[–log10 (p) ] is plotted; –log10 of p, critical value for this χ2, was given the sign of the difference, observed numbers of segregating site – expect number of segregating sites. As expected (Figure 1), there is a clear tendency for the level of polymorphism (both πnt and proportion of segregating sites) to decline proximal to the telomeres and centromeres. Therefore, the test statistics discussed in this section were determined by generating expected values as described above, but only including the “central euchromatic” regions. These were defined as the regions bounded by the first and last position on each chromosomes arm for which the proportion of segregating sites was greater than or equal to the chromosome arm average in a 510-kbp window. While this makes deviations in the centromere and telomere proximal regions appear greater, it removes the obvious bias toward positive deviations (i. e. , excess polymorphism) that would be created by including large genomic regions known to show reduced polymorphism when generating expectations. Minimum values for the expected numbers of segregating and fixed sites were one (unless otherwise indicated). Windows with coverage <200 bp were dropped (unless otherwise indicated). Expected nucleotide heterozygosity and polarized divergence were calculated for 10-kb and 50-kb nonoverlapping windows spanning each chromosome arm as described above. For each arm, autocorrelation between successive windows was calculated as: where there are n windows along an arm, and xt represents the value of nucleotide heterozyogsity or divergence for the t-th window. Significance of r for all arms for both polymorphism and divergence was calculated by permutation. All calculations were carried out in R (http: //www. r-project. org). We set out to find putative selective sweeps that occurred concomitantly with migration by D. simulans out of Africa/Madagascar. We expect the signature of these sweeps to be low variation in New World (NW) lines, defined here as w501 and SIM4/6, compared to Old World (OW) lines, defined here as C167. 4, MD199S, and MD106TS. The method described here addresses the issue of autocorrelated loci. Our approach was to simulate datasets with the same degree of autocorrelation as that of the observed data, and to determine whether there are longer runs of windows with disproportionately low NW π in the actual data than one would expect by chance. All data manipulation, calculations, and simulations were carried out in R using functions available within the “base” and “stats” packages. Mean nucleotide diversity (π) from 10-kb nonoverlapping windows throughout the genome from the two NW and three OW lines were used. Adjacent 10-kb windows were averaged (weighted by coverage) to obtain 20-kb windows. Remaining windows for which no estimate of π was available were conservatively estimated by interpolation. There were no gaps in the 20-kb window data longer than three consecutive windows in either population. For each window, the ratio of NW π: OW π (π NW: πOW) was measured. Maximum likelihood estimates of first-order coefficients of autocorrelation for each of the chromosome arms were found (all were significant). Monte Carlo simulations of the ratio πNW: πOW were performed according to the following procedure. We first randomly sampled ratios of π NW: πOW from the data with replacement for each arm separately; this ensures that our simulated data has the same mean and variance as the actual data. A first-order autoregressive filter was then applied to the randomly sampled data using the estimate of autocorrelation for the given chromosomal arm, according to the following relationship: where parameter μ is the mean of the sampled data, ρ is the autocorrelation, Xi – 1 is previous value in the series, and Xi is the original sampled measure for the ith window. This filter imposes the observed autocorrelation onto the sampled data to mimic the observed autocorrelation, resulting in a new value, Xi*, for each window. Variance and estimated first-order autocorrelation of the simulations were similar to those of the empirical data without altering this procedure. A lower threshold of π NW: πOW, below which 5% of the empirical data points reside, was determined. Significance of runs of windows below this threshold was determined by comparison to the distribution of the run lengths in 10,000 Monte Carlo simulation runs for each chromosome arm, performed as described above. P-values for each arm were corrected for multiple comparisons conservatively via Bonferroni correction (Dunn-Sidak corrections did not result in an increased number of significant sweeps). Parsimony was used to infer D. simulans/D. melanogaster ancestral states; D. yakuba was the outgroup. Only codons with one synonymous variant among the three species were included in these analyses. The preferred codon set was defined following Akashi [113]. For some analyses, preferred and unpreferred substitution rates were determined by dividing the number of substitutions of each type by the number of ancestral codons of the appropriate ancestral state (unpreferred ancestors for the preferred substitution rate and preferred ancestors for the unpreferred substitution rate), all inferred by parsimony. In principle, excess unpreferred polymorphisms at synonymous sites could erroneously lead one to infer directional selection on other sites. However, the ratio of preferred-to-unpreferred polymorphisms is not significantly different (pooled across genes or gene-by-gene) for UTRs that had significant versus nonsignificant MK tests in contrasts of synonymous and UTR sites. For introns that showed a significant MK test versus synonymous sites, there was a slightly larger ratio of unpreferred-to-preferred polymorphisms compared to the ratio for introns that were not significant. However, this was seen only in the pooled analysis and not in the gene-by-gene analysis. We found that significant intron and UTR MK tests had more similar coverages (e. g. , 5′ UTR versus synonymous) compared to tests that were not significant, suggesting that the large number of significant noncoding versus synonymous tests cannot be explained by relatively small coverage differences across site-types. Overall, these data suggest that most of the highly significant MK tests of noncoding DNA are not explained by excess unpreferred polymorphisms or coverage variation. Base composition analyses on noncoding DNA were carried out in a similar fashion, with parsimony being used to infer the D. simulans/D. melanogaster ancestor. Only unambiguous parsimony-inferred sites were used in these analyses. All X-linked genes for which Flybase reported genetic and physical locations (first nucleotide of the gene in Flybase annotation of D. melanogaster v4. 2) were used. Genetic and physical distances were determined for 12-gene intervals, sliding one gene at a time; estimates of cM/kb per interval were used as estimates of recombination rate per physical length. Mean physical and genetic distances per interval were 1. 55 Mb and 5 cM, respectively. Two intervals with negative estimates of cM/kb, indicative of discordant genetic and physical data were removed, leaving estimates of cM/kb for 150 intervals. The physical location of the interval was defined as the midpoint between physical locations of the first and last gene. For analyses investigating correlations of 50-kb windows of polymorphism and divergence with crossing-over, midpoints were rounded to the nearest 50,000. If multiple intervals were rounded to the same number, the distal interval was used in the analyses. Cloned elements. The “hanging ends” of well-mapped plasmid clones that were not fully aligned to D. melanogaster were examined by BLAST for extensive (100 bp or greater), high-quality (90% or greater) sequence similarity to known transposable elements of D. melanogaster (v 9. 2, http: //www. fruitfly. org/p_disrupt/TE. html). The coordinates are slightly rounded to facilitate finding duplicates slightly off in alignment. Clustered elements. This analysis used plasmid clones for which only one mate pair mapped uniquely and unambiguously to the genome according to the method described previously. The other mate pair was compared to the D. melanogaster transposable element database v9. 2. If the read mapped uniquely and unambiguously to a transposable element (90% coverage, 90% identity, at least two high quality differences to a secondary candidate), a transposable element was considered as mapped to the general genomic location of its mate pair. The estimated location begins at the end of the mate pair read and ends 10 kb away in the appropriate direction determined by the direction of the alignment. Transposable elements from the same family located within 5 kb of each other in the same D. simulans line were considered the same element, and therefore, were clustered. The GenBank (http: //www. ncbi. nlm. nih. gov/Genbank/) accession number for D. yakuba is AAEU01000000 (version 1) and for the D. simulans w501 whole-genome shotgun assembly is TBS-AAEU01000000 (version 1).

Population genomics, the study of genome-wide patterns of sequence variation within and between closely related species, can provide a comprehensive view of the relative importance of mutation, recombination, natural selection, and genetic drift in evolution. It can also provide fundamental insights into the biological attributes of organisms that are specifically shaped by adaptive evolution. One approach for generating population genomic datasets is to align DNA sequences from whole-genome shotgun projects to a standard reference sequence. We used this approach to carry out whole-genome analysis of polymorphism and divergence in Drosophila simulans, a close relative of the model system, D. melanogaster. We find that polymorphism and divergence fluctuate on a large scale across the genome and that these fluctuations are probably explained by natural selection rather than by variation in mutation rates. Our analysis suggests that adaptive protein evolution is common and is often related to biological processes that may be associated with gene expression, chromosome biology, and reproduction. The approaches presented here will have broad applicability to future analysis of population genomic variation in other systems, including humans.

Abstract Introduction Results/Discussion Materials and Methods Supporting Information

computational biology evolutionary biology genetics and genomics

2007

Population Genomics: Whole-Genome Analysis of Polymorphism and Divergence in Drosophila simulans

In both humans and Drosophila melanogaster, UDP-galactose 4′-epimerase (GALE) catalyzes two distinct reactions, interconverting UDP-galactose (UDP-gal) and UDP-glucose (UDP-glc) in the final step of the Leloir pathway of galactose metabolism, and also interconverting UDP-N-acetylgalactosamine (UDP-galNAc) and UDP-N-acetylglucosamine (UDP-glcNAc). All four of these UDP-sugars serve as vital substrates for glycosylation in metazoans. Partial loss of GALE in humans results in the spectrum disorder epimerase deficiency galactosemia; partial loss of GALE in Drosophila melanogaster also results in galactose-sensitivity, and complete loss in Drosophila is embryonic lethal. However, whether these outcomes in both humans and flies result from loss of one GALE activity, the other, or both has remained unknown. To address this question, we uncoupled the two activities in a Drosophila model, effectively replacing the endogenous dGALE with prokaryotic transgenes, one of which (Escherichia coli GALE) efficiently interconverts only UDP-gal/UDP-glc, and the other of which (Plesiomonas shigelloides wbgU) efficiently interconverts only UDP-galNAc/UDP-glcNAc. Our results demonstrate that both UDP-gal and UDP-galNAc activities of dGALE are required for Drosophila survival, although distinct roles for each activity can be seen in specific windows of developmental time or in response to a galactose challenge. By extension, these data also suggest that both activities might play distinct and essential roles in humans. Galactose is an essential component of glycoproteins and glycolipids in metazoans, and as a constituent monosaccharide of the milk sugar, lactose, also serves as a key nutrient for mammalian infants. Galactose is also found in notable quantities in some fruits, vegetables, and legumes. Galactose is both synthesized and catabolized in all species via the Leloir pathway, which is highly conserved across branches of the evolutionary tree [1]. The reactions of the Leloir pathway are catalyzed by the sequential activities of three enzymes: (1) galactokinase (GALK) which phosphorylates alpha-D-galactose to form galactose-1-phosphate (gal-1P), (2) galactose-1-phosphate uridylyltransferase (GALT), which transfers uridine monophosphate (UMP) from uridine diphosphoglucose (UDP-glc) to gal-1P, forming UDP-galactose (UDP-gal) and releasing glucose-1-phosphate (glc-1P), which can proceed to phosphoglucomutase and the glycolytic pathway, and (3) UDP-galactose 4′-epimerase (GALE) which interconverts UDP-gal and UDP-glc [1]. In addition to a role in the Leloir pathway, metazoan GALE enzymes also interconvert UDP-N-acetylgalactosamine (UDP-galNAc) and UDP-N-acetylglucosamine (UDP-glcNAc) (Figure 1). Because it catalyzes reversible reactions, GALE therefore not only contributes to the catabolism of dietary galactose, but also enables the endogenous biosynthesis of both UDP-gal and UDP-galNAc [2], [3] when exogenous sources are limited. Deficiency in any of the three Leloir enzymes in humans results in a form of the metabolic disorder galactosemia, although the symptoms and clinical severity differ according to which enzyme is impaired and the extent of the impairment. Profound loss of hGALE results in generalized epimerase-deficiency galactosemia, an autosomal recessive and potentially severe disorder. To date, however, no patient has been reported with complete loss of GALE, and even the most severely affected demonstrate at least 5% residual enzyme activity [4]. Previous studies have indicated that different patient mutations impair hGALE to different extents [5]–[9]. Further, while some mutations impair both GALE activities similarly, others do not. For example, the hGALE allele V94M, which leads to severe epimerase-deficiency galactosemia in the homozygous state, encodes an enzyme that retains ∼5% residual activity toward UDP-gal but ∼25% residual activity toward UDP-galNAc [8], [9]. Disparities such as this have raised the question of whether the pathophysiology of epimerase deficiency galactosemia results from the loss of GALE activity toward UDP-gal/UDP-glc, or toward UDP-galNAc/UDP-glcNAc, or both. To address this question, we applied a Drosophila melanogaster model of GALE deficiency [10]. Using this model, we have previously established that GALE is essential in Drosophila; animals completely lacking endogenous dGALE succumb as embryos, and conditional loss of dGALE in larvae results in death within two to four days of knockdown. Finally, partial loss of dGALE leads to galactose sensitivity in larvae, and transgenic expression of human GALE (hGALE) rescues each of these negative outcomes [7]. Here we have applied our transgenic Drosophila model to uncouple and examine the individual roles of GALE separately. Toward that end, we generated flies that lacked endogenous dGALE and expressed either of two prokaryotic transgenes, one encoding E. coli GALE (eGALE) which exhibits an approximately 8,000-fold substrate preference for UDP-gal/UDP-glc over UDP-galNAc/UDP-glcNAc [11], and the other encoding P. shigelloides wbgU, which exhibits an approximately 2,000-fold substrate preference for UDP-galNAc/UDP-glcNAc over UDP-gal/UDP-glc [12]. By expressing these prokaryotic transgenes individually or in combination in dGALE-deficient Drosophila we determined that both GALE activities are required for survival of embryos and larvae. We also found that restoration of one activity or the other in later development rescued some phenotypes. Combined, these results provide insight into the varied roles of dGALE in Drosophila development and homeostasis, and by extension, suggest that hGALE may play similarly complex and essential roles in humans. Human and other mammalian GALE enzymes efficiently interconvert both UDP-gal/UDP-glc and UDP-galNAc/UDP-glcNAc (e. g. [13]–[15]). Previously, we reported that Drosophila GALE interconverts the first of these substrate pairs (UDP-gal/UDP-glc) [10], but did not address whether dGALE could also interconvert the second. Here we demonstrate that dGALE from wild-type adult flies efficiently interconverts both substrate sets (left most bar, Figure 2). Of note, while purified human GALE [15] and dGALE each interconvert both UDP-gal/UDP-glc and UDP-galNAc/UDP-glcNAc, the apparent specific activity of both human and fly enzymes toward UDP-gal is significantly higher than toward UDP-galNAc. To generate flies with epimerase activity toward only UDP-gal/UDP-glc or only UDP-galNAc/UDP-glcNAc, we created transgenic lines expressing eGALE (UAS-eGALE) or wbgU (UAS-wbgU), respectively, each in a conditionally dGALE-impaired background. Each of these prokaryotic GALE genes has been demonstrated previously to encode epimerase activity toward only one of the two sets of epimer pairs (e. g. [11], [12]). To minimize background, activities of the encoded eGALE and WbgU enzymes toward UDP-gal and UDP-galNAc were assayed in flies knocked down for endogenous dGALE; results for the transgenes that demonstrated activities closest to those seen in wild-type Drosophila, eGALE62A and wbgU19A, are presented in Figure 2. As expected, lysates from dGALE knockdown flies expressing the eGALE transgene demonstrated strong activity toward UDP-gal, but not UDP-galNAc, and lysates from dGALE knockdown flies expressing the wbgU transgene demonstrated strong activity toward UDP-galNAc, but not UDP-gal. As a control we also tested lysates from dGALE knockdown flies expressing a human GALE transgene; as expected, those samples demonstrated very strong activity toward both substrates. Previously, we created and characterized two dGALE-deficient alleles, dGALEf00624. 4 and dGALEΔy, which allowed us to demonstrate that GALE is essential for survival in Drosophila [10]. To examine the requirement for the two different epimerase activities separately, we set up crosses which allowed for the expression of eGALE or wbgU, individually or in combination, driven by Act5C-GAL4 in an otherwise dGALE-deficient background (dGALEf00624. 4/dGALEΔy). Table 1 shows the observed to expected ratios of surviving transgenic offspring that eclosed from these crosses. As presented in Table 1, neither eGALE alone nor wbgU alone was sufficient to rescue survival of the dGALE-deficient animals; however, expression of both eGALE and wbgU, in combination, was sufficient. These results demonstrate that GALE activities toward both UDP-gal and UDP-galNAc are essential for survival of D. melanogaster. To rule out the possibility that rescue with eGALE plus wbgU in combination occurred not because both GALE activities are essential but rather because neither individual transgene expressed sufficient enzyme, we also tested additional eGALE and wbgU transgenes that individually demonstrated higher levels of expression; none was sufficient to rescue (data not shown). Of note, there also was no apparent over-expression phenotype; for example, animals expressing either eGALE or wbgU in addition to endogenous dGALE, and animals dramatically over-expressing human GALE, remained viable, fertile, and appeared morphologically normal (data not shown). Previously, we described an approach that achieves conditional knockdown of dGALE in Drosophila using a UAS-RNAidGALE transgene (12030-R2, National Institute of Genetics Fly Stock Center, Mishima, Shizuoka, Japan) in combination with a temperature sensitive allele of yeast GAL80 (GAL80ts) ([10] and Figure 3A). Using this system, we found that dGALE is required from embryogenesis through pupation, and that loss of dGALE during pupation leads to defects in fecundity and perhaps also a shortened life span [10]. Here we have expanded the GAL80tsconditional dGALE knockdown system to include different GAL4-dependent GALE transgenes and have applied this expanded system to test the ability of each transgene, or pair of transgenes, to compensate for the loss of endogenous dGALE. By using age-synchronized cohorts of animals and shifting from the permissive (18°C) to the restrictive temperature (28–29°C) at different times we also were able to test the ability of each GALE transgene, or pair of transgenes, to sustain survival and fecundity at different stages of development. At 18°C these animals expressed endogenous dGALE, but not their transgenes, and at 28–29°C these animals expressed their transgenes but not dGALE (Figure 3A). Specifically, we tested Drosophila that carried no GALE transgene, an eGALE transgene, a wbgU transgene, both eGALE and wbgU transgenes, or an hGALE transgene. As expected from prior results ([10] and Table 1), animals expressing no GALE transgene succumbed when shifted to the restrictive temperature as larvae, while animals expressing either human GALE or both eGALE plus wbgU remained viable and fertile (Figure 3C). Surprisingly, expression of either eGALE or wbgU alone was also sufficient to rescue survival, albeit to a lesser extent. The fact that the individual prokaryotic transgenes were sufficient to rescue dGALE knockdown animals, but not animals genetically null for dGALE (Table 1), suggests that trace residual dGALE expression in the knockdown animals lowered the threshold of transgene function required for rescue. Of note, while dGALE knockdown animals encoding either eGALE or wbgU remained viable following a shift to the restrictive temperature in early to mid-development (Figure 3C), these survivors were not entirely healthy. Specifically, these animals demonstrated either partial or complete loss of fecundity as adults. To test whether the degree of dGALE knockdown was comparable between males and females, and therefore not a confounding factor in differential outcome, we performed GALE and GALT enzyme assays on newly eclosed and three day old male and female knockdown adults that carried no GALE transgene and that had been switched to the restrictive temperature as early to mid-stage pupa. The degree of GALE knockdown in both males and females was profound and comparable (Figure 3B). As expected, the level of GALE activity was even lower in the older animals, presumably because any GALE synthesized prior to the temperature switch had three additional days to decay. Also as expected, GALT activity was normal and apparently unaffected by the dGALE knockdown in all samples tested (data not shown). To examine fecundity, we collected and sequestered newly eclosed virgin female and male flies from each surviving cohort, crossed them to an equal number of wild-type flies of the opposite sex, and counted the numbers of viable offspring resulting from each cross. Crosses resulting in large numbers of viable offspring (>50) were scored as “normal fecundity”. Crosses resulting in fewer than 10 viable offspring were scored as “reduced fecundity, ” and crosses resulting in no viable offspring were scored as “loss of fecundity” (Figure 3C). For example, when dGALE knockdown was initiated during early to mid-stage pupal development, animals of both sexes displayed diminished fecundity. Expression of eGALE alone, but not wbgU alone, rescued the male defect, whereas expression of both prokaryotic transgenes in combination, or hGALE alone, was required to rescue the female defect. These results indicate that GALE activity toward UDP-gal is both necessary and sufficient for male fecundity, but that GALE activities toward both UDP-gal and UDP-galNAc are required for female fecundity. We have previously demonstrated that Drosophila expressing a hypomorphic allele of dGALE are viable but sensitive to galactose exposure [10]. To assess the roles of the two GALE activities in coping with environmental galactose, we collected adult flies in which dGALE knockdown coupled with hGALE, eGALE, wbgU, or eGALE plus wbgU transgene expression was initiated using the GAL80ts system during late larval or early-to-mid-pupal development. These animals were allowed to develop on a standard molasses-based food, and were then transferred as newly eclosed adults to food containing either 555 mM glucose as the sole sugar, or 555 mM glucose plus 175 mM galactose. We assessed the lifespan of each cohort of animals on both foods; as a control, knockdown animals expressing no GALE transgene were also monitored (Figure 4). In the absence of galactose, all cohorts showed similar longevity profiles, although females (Figure 4C) showed greater variability than males (Figure 4A). In the presence of galactose, however, both males and females expressing either no GALE transgene, or only the wbgU transgene, demonstrated a dramatic reduction in life span (p<0. 0001, Figure 4B and 4D). Females expressing eGALE alone exhibited a slight decrease in life span that was independent of diet. Animals expressing hGALE or eGALE+wbgU had lifespans comparable to control animals expressing endogenous dGALE, regardless of diet. These data implicate loss of UDP-gal activity as responsible for the galactose-dependent early demise of adult dGALE-impaired Drosophila. As one approach to explore the pathophysiology underlying the different galactose-dependent outcomes observed in Drosophila deficient in GALE activity toward UDP-gal or UDP-galNAc we measured the levels of gal-1P, UDP-gal, and UDP-galNAc in lysates prepared from galactose-exposed third instar larvae expressing different GALE transgenes. As illustrated in Figure 5, galactose exposed animals deficient in both GALE activities (bars marked “KD” for knockdown) accumulated abnormally high levels of gal-1P (Figure 5A and 5D) and UDP-gal (Figure 5B and 5E). Animals deficient only in GALE activity toward UDP-gal (bars marked “wbgU” in Figure 5) also demonstrated elevated gal-1P (Figure 5A and 5D) and UDP-gal (Figure 5B and 5E). In contrast, galactose exposed larvae deficient only in GALE activity toward UDP-galNAc (bars marked “eGALE” in Figure 5C and 5F) demonstrated no extraordinary metabolic abnormalities, although, as expected, the absolute level of UDP-galNAc was diminished in these animals independent of diet relative to the “no knockdown” control (Figure 5C). Also as expected, animals expressing either hGALE or both eGALE plus wbgU demonstrated no clear metabolic abnormalities (Figure 5). The disparate metabolic profiles observed in GALE-impaired flies exposed to galactose provide a window of insight into potential mechanisms behind the outcomes observed. For example, gal-1P accumulates to abnormal levels in animals missing GALE activity toward UDP-gal but not UDP-galNAc, and only those animals demonstrate substantially reduced lifespan when exposed to galactose as adults. This metabolic result is expected, since only GALE activity toward UDP-gal should impact the Leloir pathway, and this outcome result implies that gal-1P might contribute to the early demise of these animals. However, the gal-1P result also implies that the negative outcomes observed in Drosophila deficient in GALE activity toward UDP-galNAc, e. g. compromised survival in embryos and compromised fecundity in adult females, do not result from gal-1P accumulation. This is an important point because it challenges the common supposition that gal-1P underlies pathophysiology in both classic and epimerase deficiency galactosemias. Clearly there must be another basis for the negative outcomes observed in these animals. It is also interesting to note that while loss of GALE activity toward UDP-galNAc in developing animals has phenotypic consequences, at least for female fecundity, it does not appear to negatively impact the “global” level of UDP-galNAc in animals exposed to galactose. The explanation for this apparent disparity might involve subtle or tissue-specific differences below the threshold of detection of our experimental approach. The implications of this work for patients with epimerase deficiency galactosemia are two-fold. First, these results demonstrate that both GALE activities are essential for health of flies, and possibly also people. To our knowledge clinical laboratories that test patient samples for GALE activity only test activity toward UDP-gal. While this practice is certainly understandable, given that mutations may impact the two GALE activities differently [18]–[20], the results presented here raise the possibility that rare patients with GALE deficiency limited to UDP-galNAc activity could be missed. Second, given the impact of GALE-loss on both male and female fecundity in flies, these results suggest that long-term studies of both male and female reproductive issues in epimerase-deficiency galactosemia patients might be warranted. The Drosophila stocks used in this study are listed in Table S1. Stocks were maintained at 25°C on a molasses-based food that contained 43. 5 g/l cornmeal, 17. 5 g/l yeast extract, 8. 75 g/l agar, 54. 7 ml/l molasses, 10 mls propionic acid and 14. 4 ml/l tegosept mold inhibitor (10% w/v in ethanol). For experiments in which the levels and types of sugar were to be varied, we used a glucose-based food [5. 5 g/l agar, 40 g/l yeast, 90 g/l cornmeal, 100 g/l glucose, 10 ml/l propionic acid and 14. 4 ml/l tegosept mold inhibitor (10% w/v in ethanol) ] [21] supplemented with galactose, as indicated. UAS-eGALE and UAS-wbgU transgenes were generated by subcloning the eGALE and wbgU coding sequences, respectively, as EcoRI/XhoI fragments, into pUAST [22] using the EcoRI and XhoI sites in the pUAST polylinker region. The wbgU sequence was amplified from a plasmid generously provided by Peng George Wang (Ohio State University). Resulting plasmids were confirmed by sequence analysis. UAS-eGALE stocks were generated using standard transgenic techniques following injection of the transgene into embryos by the fly core of the Massachusetts General Hospital, Charlestown, MA. UAS-wbgU stocks were generated using standard transgenic techniques following injection of the transgene into embryos by Genetic Services, Inc. , Cambridge, MA. Transformants were selected by the presence of the white gene within pUAST. Expression of functional eGALE or wbgU was confirmed by enzymatic assay of lysates from transformants. Lysates were prepared and assays for GALK, GALT and GALE with UDP-gal as the substrate were performed (n≥3) as described previously [10]. Activity was calculated from the conversion of UDP-galNAc to UDP-glcNAc. The initial reaction mixture concentrations were: 100 mM glycine pH 8. 7,1. 6 mM UDP-galNAc and 0. 5 mM NAD. Enzyme assays were performed as described in Sanders et al. [10] except for the following changes: To start each reaction, 7. 5 µl of diluted protein and 5 µl of a cocktail of substrates and cofactors were combined. Reaction mixtures were incubated at 25°C for 30 minutes and then quenched by the addition of 112. 5 µl of ice-cold high-performance liquid chromatography (HPLC) -grade water (Fisher). Lysates were diluted 1∶4, except for those prepared from animals with RNAi knockdown, which were undiluted, and those prepared from animals overexpressing hGALE or wbgU transgenes, which were diluted to a greater extent. Lysates from Act5C>hGALE22C animals were diluted 1∶60. Lysates from Act5C>wbgU19A animals were diluted 1∶20. Generation of animals in which GALE knockdown was initiated at 24-hour intervals throughout development was achieved as described previously [10]. A stock homozygous for both P{tubP-GAL80ts}10 and 12030R-2 was used in all crosses. These flies were then crossed to the appropriate genotypes to obtain offspring expressing various transgenes; for: no transgene, P{Act5C-GAL4}25FO1; +/T (2; 3) TSTL, Tb, Hu; eGALE only, P{Act5C-GAL4}25FO1, UAS-eGALE62A/CyO; wbgU only, P{Act5C-GAL4}25FO1/CyO; P{Act5C-GAL4}25FO1/CyO; UAS-wbgU19A/TM6B; eGALE plus wbgU, P{Act5C-GAL4}25FO1, UAS-eGALE62A/CyO; UAS-wbgU19A/TM6B; hGALE, P{Act5C-GAL4}25FO1/CyO; UAS-hGALE22C/TM6B. Adult flies eclosing from the vials were scored for the presence or absence of humeral and/or curly, as appropriate for each cross. Animals in which dGALE knockdown with concurrent transgene expression was achieved throughout development were obtained as described above. These animals were maintained on standard molasses medium until eclosion. Within 24 hours after eclosion, approximately 20 virgin male or female flies were placed in fresh vials of food containing 555 mM glucose only or 555 mM glucose plus 175 mM galactose. Flies were transferred to fresh food every 2–3 days, and the number of dead flies in each vial was recorded every other day. Log rank and Wilcoxon tests were used for statistical analysis using the program JMP (http: //www. jmp. com/). Cohorts of newly hatched larvae raised at 18°C were transferred to vials of food containing either 555 mM glucose only or 555 mM glucose plus 175 mM galactose. The larvae were maintained at 18°C for one additional day, then transferred to 28°–29°C and allowed to develop for another four days prior to harvest. Metabolites were extracted and quantified as described previously [10], and were separated and quantified using a Dionex HPLC, as described previously [23] with the following changes: UDP-gal and UDP-galNAc were separated using a high salt isocratic procedure with a flow rate of 0. 5 mL/min and buffer concentrations of 45% A and 55% B (0–61 min), followed by washing with a linear increase of B to 95% (61–80 min). For all samples, 20 µl were injected into a 25 µl injection loop. Ratios of the level of each metabolite on food containing galactose over the level on food containing glucose only were calculated, and 95% confidence intervals were determined using Fieller' s theorem.

In this manuscript we apply a fruit fly model to explore the relative contributions of each of two different activities attributed to a single enzyme—UDP-galactose 4′-epimerase (GALE); partial impairment of human GALE results in the potentially severe metabolic disorder epimerase deficiency galactosemia. One GALE activity involves interconverting UDP-galactose and UDP-glucose in the Leloir pathway of galactose metabolism; the other activity involves interconverting UDP-N-acetylgalactosamine and UDP-N-acetylglucosamine. We have previously demonstrated that complete loss of GALE is embryonic lethal in fruit flies, but it was unclear which GALE activity loss was responsible for the outcome. Using genetically modified fruit flies, we were able to remove or give back each GALE activity individually at different times in development and observe the consequences. Our results demonstrate that both GALE activities are essential, although they play different roles at different times in development. These results provide insight into the normal functions of GALE and also have implications for diagnosis and intervention in epimerase deficiency galactosemia.

Abstract Introduction Results Discussion Materials and Methods

developmental biology model organisms genetics biology molecular cell biology genetics and genomics

2012

UDP-Galactose 4′-Epimerase Activities toward UDP-Gal and UDP-GalNAc Play Different Roles in the Development of Drosophila melanogaster

The identification and characterization of antigens expressed in Trypanosoma cruzi stages that parasitize mammals are essential steps for the development of new vaccines and diagnostics. Genes that are preferentially expressed in trypomastigotes may be involved in key processes that define the biology of trypomastigotes, like cell invasion and immune system evasion. With the initial aim of identifying trypomastigote-specific expressed tags, we constructed and sequenced an epimastigote-subtracted trypomastigote cDNA library (library TcT-E). More than 45% of the sequenced clones of the library could not be mapped to previously annotated mRNAs or proteins. We validated the presence of these transcripts by reverse northern blot and northern blot experiments, therefore providing novel information about the mRNA expression of these genes in trypomastigotes. A 280-bp consensus element (TcT-E element, TcT-Eelem) located at the 3′ untranslated region (3′ UTR) of many different open reading frames (ORFs) was identified after clustering the TcT-E dataset. Using an RT-PCR approach, we were able to amplify different mature mRNAs containing the same TcT-Eelem in the 3′ UTR. The proteins encoded by these ORFs are members of a novel surface protein family in T. cruzi, (which we named TcTASV for T. cruzi Trypomastigote, Alanine, Serine and Valine rich proteins). All members of the TcTASV family have conserved coding amino- and carboxy-termini, and a central variable core that allows partitioning of TcTASV proteins into three subfamilies. Analysis of the T. cruzi genome database resulted in the identification of 38 genes/ORFs for the whole TcTASV family in the reference CL-Brener strain (lineage II). Because this protein family was not found in other trypanosomatids, we also looked for the presence of TcTASV genes in other evolutionary lineages of T. cruzi, sequencing 48 and 28 TcTASVs members from the RA (lineage II) and Dm28 (lineage I) T. cruzi strains respectively. Detailed phylogenetic analyses of TcTASV gene products show that this gene family is different from previously characterized mucin (TcMUCII), mucin-like, and MASP protein families. We identified TcTASV, a new gene family of surface proteins in T. cruzi. Trypanosoma cruzi, a kinetoplastid protozoan parasite, is the causative agent of the American trypanosomiasis, also known as Chagas' disease, a zoonotic disease that affects about 8 million individuals in Latin America [1]. The disease is a chronic illness, which symptoms appear 10 or more years after the beginning of the infection, being the most common clinical forms the digestive megas and heart failure, which can lead to death. Currently, there is no effective therapy nor vaccine for the treatment or prevention of the disease [1], [2]. The identification and characterization of proteins expressed in the mammalian stages of T. cruzi (amastigotes and trypomastigotes) are key to drug and vaccine development [3]. The genome of the CL-Brener clone of T. cruzi was already sequenced by 2005 [4], but its final assembly has only been partially completed recently, mainly because of the high number of repetitive sequences [5]. Although 90% of the genes were assembled in 41 chromosomes, the remaining 10%, the majority of which belong to multigene families, are still excluded from the assembly, as they have not been assigned to any chromosome. Moreover, 64% of the predicted genes have been annotated as hypothetical proteins –their function and/or expression is unknown-, and it is possible that other genes may not have been annotated as genes at all. Therefore, the generation of expressed sequence tags (ESTs, single pass reads obtained from randomly selected cDNA clones) is still a valuable approach to map the location of genes, to obtain experimental evidence about their expression, to identify stage-specific transcripts, and to identify their untranslated regions (UTRs). Previously, we reported the sequencing and analysis of two full-length cDNA libraries constructed from trypomastigotes and amastigotes [6]. Because those libraries were not normalized and were prepared under similar conditions, we were able to identify a number of EST clusters that showed a significantly biased composition in the number of sequences derived from either the trypomastigote and/or the amastigote cDNA libraries. However, only one cluster corresponded to a case of apparent increased expression in trypomastigotes. In the present work, we focused our attention on the identification of mRNA transcripts over-represented in the mammalian trypomastigote stage as compared to the vector-associated epimastigote stage, using a subtractive PCR approach [7]. Molecules that are differentially expressed in the trypomastigote stage may be involved in the extracellular survival, dissemination to different organs and cell invasion that are the hallmarks of of this parasite stage. Besides finding a large proportion of novel and differentially expressed mRNAs in trypomastigotes (most of them with an unknown function), we discovered a novel protein family, which we denominated TcTASV. The expression profile and the genetic mapping of TcTASVs in the CL-Brener, Dm28 and RA T. cruzi strains were also investigated in this work. All procedures requiring animals were performed in agreement with the guidelines of the Animal Ethics Comitee of our Institution. The CL-Brener clone (reference strain), RA (lineage II) and Dm28 (lineage I) strains of T. cruzi were used [8], [9], [10]. Trypomastigotes and amastigotes were obtained in vitro by infection of Vero cells grown in Minimum Essential Medium (MEM) -3% foetal bovine serum. For the library construction essentially pure CL-Brener trypomastigotes (with less than 3% amastigote forms) were used. Epimastigotes were obtained from axenic cultures, as previously described [11]. Total RNA was isolated from trypomastigotes and epimastigotes with TRIzol (Gibco-BRL) and mRNA purified with polyA-Tract mRNA isolation system (Promega). The PCR-Select cDNA Subtraction kit was used for library construction following the selective subtractive hybridization protocol provided by the manufacturers (CLONTECH). First strand cDNA synthesis was performed with 2 µg of polyA+ of each T. cruzi stage (trypomastigote and epimastigote), oligo dT primer with a 5′ RsaI site and Superscript II reverse transcriptase (Gibco-BRL). Second strand cDNA synthesis was performed with T4 DNA polymerase. After RsaI digestion of double stranded cDNA, two different sets of adaptors were ligated to the tester cDNA (trypomastigotes) but not to the driver cDNA (epimastigotes). Two rounds of subtractive hybridization in the presence of an excess of driver cDNA were performed, thus leading to the enrichment of differentially expressed sequences in the tester cDNA population that were the templates for further suppressive PCR amplification performed with adaptor-specific primers [7]. The subtraction efficiency was verified by monitoring the PCR amplification of the T. cruzi histone 2A transcript in subtracted and unsubtracted samples (H2_3′: tcttggacgccttcttcgct; H2_5′: gtgatgccgagcctgaacaa). PCR products enriched for tester differentially expressed sequences -higher than 100 bp- were cloned into the pGEM T-Easy vector (Promega). E. coli DH5α cells were transformed with ligations; white colonies were randomly picked and the TcT-E library plated in 384-well microplates. Template preparation of clones for sequencing was carried out as previously described [12]. Sequencing reactions were performed in a Perkin Elmer 9600 thermal cycler by using a Dye Terminator Cycle sequencing Ready Reaction Kit with AmpliTaq DNA polymerase according to the protocols supplied by the manufacturer (Applied Biosystems). Single-pass sequencing was performed on an ABI 377 automated sequencer. Bases were called by PHRED and an automated protocol for the analysis of the data was used to assess sequence quality and trim vector, adaptors and unreliable data from sequences [6]. Sequences longer than 100 bases were further analyzed. Sequence similarity searches against in-house databases were run locally using the BLAST suite of programs as distributed by the NCBI in a PC computer running Linux. Sequences were also compared against the NCBI non-redundant protein or nucleotide databases by using BLASTX or BLASTN programs respectively (cut off values: BLASTN p<10e-40; BLASTX p<10e-5) [12], [13]. For Northern blot, total RNA (20 µg/lane) from trypomastigotes and epimastigotes was electrophoresed on 1. 5% agarose formaldehyde gels and transferred to nylon membranes (Zeta-Probe, BioRad). All TcT-E clones used as probes were labeled with 32P by PCR using adaptor-specific primers (Nest_2R: agcgtggtcgcggccgaggt; Nest_1: tcgagcggccgcccgggcaggt). Hybridization and washing were performed at 63°C following standard procedures [14]. The complete ORF Tcruzi_1863-4-1211-93 (TriTrypDB. org) was amplified by PCR from the clone G53E20 (GenBank Acc AZ050960) from a random genomic library DNA [12], labeled by PCR and used as probe in northern and southern blot experiments. For reverse northern blots, clones of the TcT-E library were picked, grown in LB-ampicillin in 96-well plates and subjected to colony-PCR using 1 µl of culture and primers Nest_2R and Nest_1 [15]. The sizes of the inserts were checked on a 2% agarose gel and PCR products were then denatured and dotted in duplicate onto nitrocellulose membranes. Filters were hybridized with cDNA probes synthesized from total RNA of trypomastigotes and epimastigotes by reverse transcription using 32P-dCTP. Plasmids containing tubulin and SAPA (shed acute phase antigen) T. cruzi genes were dotted on membranes as positive controls, whereas a plasmid containing a non-related (mouse) gene was used as a negative control. For southern blots, DNA was prepared from epimastigotes of the CL-Brener strain by using the conventional Proteinase K phenol-chloroform method and digested with the indicated restriction enzymes. Electrophoresis, hybridization and washing were performed by standard procedures [14]. The complete TcT-E element (TcT-Eelem) was obtained from CL-Brener genomic DNA by PCR using Pfu DNA polymerase and the primers TcT-Ee_pp_Hind (taaagcttccgggcaggtacagtat) and TcT-Ee_pp_Xho (atctcgagtgagaatcccgcaggact). Mature mRNA transcripts containing both the TcT-Eelem and the different upstream open reading frames (ORFs) were identified by RT-PCR and sequencing in the CL-Brener strain. RNA was treated with RQ1 DNase (Promega Corp. , Madison, USA) and first strand cDNA synthesized using an oligo dT primer. PCR was performed using a 5′ primer specific for the T. cruzi miniexon containing an EcoRI site (cccgaattcaacgctattattgatacagtttctgt) and a 3′ antisense primer corresponding to the 3′ region of the TcT-Eelem (TcT-Ee_int_R: aagaaatgattcggcaggaa). PCR products were gel- excised, purified using QIAex II (Qiagen) and cloned. Alternatively, after first strand cDNA synthesis, PCR was performed with primers corresponding to the 5′ and 3′ conserved regions of the majority of the ORFs (CDS_desc_L: gtcgagcgactctacgacg; CDS_desc_R: acagcagcacagacaaggg) or with the 5′ CDS_desc_L and the 3′ T-Ee_int_R primers. Bands were also gel-excised, cloned and sequenced. Conceptually translated proteins corresponding to the cloned CDS were aligned by the Clustal method. To search for TcTASV in other T. cruzi strains, genomic DNA from Dm28 (lineage I, currently T. cruzi I) and RA (lineage II, currently T. cruzi VI) was amplified by PCR using primers CDS_desc_L and CDS_desc_R [8], [9], [10]. The bands obtained were gel-purified, cloned and sequenced on both strands on an ABI 3130. The sequences of each clone were assembled using the program DNAbaser. Phylogenetic trees were constructed from amino acid alignments using the Neighbour Joining method, and bootstrapped using 1000 permutations. The trees were rooted using 6 sequences as outgroups, and were visualized with the TreeView program (http: //taxonomy. zoology. gla. ac. uk/rod/treeview. html). The nucleotide sequence AF080220 (GenBank Accession number) was used to carry out a BLASTN search against the TcT-E database. A multiple sequence alignment was computed using the Clustal Method [16]. The consensus sequence of the TcT-E element (TcT-Eelem) was used as bait to search the unassembled whole genome shotgun sequences (GSSs) of T. cruzi at TIGR (http: //tigrblast. tigr. org/er-blast/index. cgi? project=tca1). GSSs identified in this way were assembled into contigs, that were then visualized and edited in Artemis to identify in silico additional TcTASVs [17]. Motif scanning for signal peptide, cleavage sites (SignalP) and Ser, Thr, and Tyr phosphorylation sites (NetPhos) was performed in the ExPASy proteomics server at http: //www. expasy. org/. The prediction of glycosylphosphatidylinositol (GPI) anchor addition sites, was performed using DGPI (run locally) and FragAnchor (http: //navet. ics. hawaii. edu/~fraganchor/NNHMM/NNHMM. html) [18]. The peptide RQ28 (GKLRWRFQGEKDWRKC) comprising amino acids 57 to 72 of TcTASV-A1 (GenBank AM492199) was purchased from Sigma-Genosys. This sequence was chosen because it is present in the conserved, noncleaved N-terminal region of the protein family which is also predicted not to be glycosylated or modified. The KLH coupled peptide was used to develop an anti-TcTASV-A specific serum in rabbit. Total IgG from anti-RQ serum was purified with protein G columns (HiTrap, GE Healthcare Life Sciences) and specific anti-TcTASV-A antibodies were purified by column affinity with SulfoLink Kits coupled with the RQ28 peptide (Thermo Scientific). Antibodies were used at 0. 1 µg/ml. Protein extracts of T. cruzi epimastigotes, trypomastigotes and amastigotes were resuspended in cracking buffer (60 mM Tris-HCl pH 6. 8; 2% SDS, 0. 1% glycerol, 5% â-mercaptoethanol) in the presence of protease inhibitors at a density of 1–2×106 parasites/µl. Conventional SDS-PAGE was performed on 12% polyacrylamide gels, and proteins were then transferred to nitrocellulose filters. Blots were incubated with anti-TcTASV-A antibodies, washed, incubated with an anti-rabbit secondary antibody labelled with horseradish-peroxidase (DAKO) and developed with chemiluminescence. Note: Nucleotide sequence data reported in this paper have been submitted to the EMBL/GenBank/DDBJ databases with the accession numbers AM492199–AM492211, GW883555–GW883875, HO052091–HO052172 and FN599093–FN599167. To gain information about the genes that are differentially expressed in the trypomastigote (circulating stage in mammals) but not in the epimastigote (replicative stage in the insect vector) of T. cruzi, we built a library of trypomastigote cDNA subtracted with epimastigote cDNA (TcT-E library). Partial sequencing of this library provided high-quality sequences of 403 clones (GenBank Acc GW883555–GW883875 and HO052091–HO052172). With this set of data sequence (BLAST) analyses were performed against various databases of trypanosomatid ESTs (T. cruzi trypomastigotes, T. cruzi epimastigotes, ESTs of all kinetoplastids) and against protein databases (nr at GenBank and SwissProt) (Fig. 1A). The BLAST reports for all searches can be accessed online at http: //genoma. unsam. edu. ar/projects/tct-e/tct-e. p. html (Table S1). Briefly, more than 46% of the TcT-E dataset do not have any known mRNA or protein homologue and only two clones give positive hits against all databases (Fig. 1A). The comparison of the TcT-E dataset against T. cruzi ESTs showed that (a) only 3. 5% of them corresponded to trypomastigote-specific tags that were identified prior to this work and (b) 38% of TcT-E clones have significant matches with epimastigote sequences (Fig. 1A). The latter was expected in part because of the high number of epimastigote ESTs available in public databases. This could also indicate that these transcripts, although present in both stages, might be expressed at a higher level in trypomastigotes than in epimastigotes, since our library had been subtracted with epimastigote cDNA. By Northern blot (Fig. 1B) and reverse Northern blot on 86 randomly picked TcT-E clones (data not shown), we were able to confirm that even in those cases were a TcT-E clone had identity with ESTs obtained from epimastigotes, the mRNA levels were consistently higher in trypomastigotes than in epimastigotes. Sixty-eight (16. 9%) clones had similarity to known proteins (nr and SwissProt databases) and 33 of them matched previously described trypomastigote antigens like the flagellum-associated surface protein FL-160 (gb|AAA30196) or sialidase homologues (AF051695 and AF051696) (Table S1). The top 50 hits against SwissProt and GenBank (nr) are provided in Table S2. The whole TcT-E dataset can be searched by blast at http: //genoma. unsam. edu. ar/projects/tct-e/. To compensate for sequencing errors and to obtain longer sequences, we next generated a non-redundant TcT-E EST set by clustering (using the blastclust tool from the NCBI C Toolkit), which was composed of 23 clusters containing 261 sequences and 142 singletons (ESTs with no similarity against any other EST) (Table S3). EST clones belonging to clusters with the largest number of sequences as well as other clones that showed significant similarity to SwissProt and/or GenBank nr databases were selected to analyze their expression by reverse northern blot (Fig. 1C). We observed that most of the TcT-E clones were actually overexpressed in trypomastigotes, which again confirms the correct subtraction of the library, and that the sequences generated provide information about the transcripts differentially expressed in the trypomastigote stage. The larger groups of sequences in the clustered TcT-E EST dataset contained sequences that had no similarity against sequences in other databases. To further characterize these sequences, we then attempted to find any motif or conserved sequence in these contigs. By lowering the cut off value for BLASTN (e≤10e-5), we found that some contigs in cluster 1 showed similarity with a 100-bp region in the 3′ untranslated region (UTR) of the flagellar T. cruzi FL-160-2 gene (GenBank Acc AF080220) [19], [20]. By computing a multiple sequence alignment of the TcT-E clones that presented identity in these 100 bases, we reconstructed a consensus sequence of 280 bp (Fig. 2A). We named this 280-bp element TcT-E element (TcT-Eelem) because of its high representation in the subtractive TcT-E library. The first 27 and last 17 bases of the TcT-Eelem (bold in Fig. 2A) are polypyrimidine tracts and bases 66 to 165 (in italic) correspond to those similar to the 3′ UTR of FL-160-2. Another feature of the TcT-Eelem is the presence of a variable number (between 3 and 5) of TTA repeats (bold underlined in Fig. 2A). By southern blot, we found a pattern indicative of multiple genomic copies of the TcT-Eelem since several bands were detected, even though none of the restriction enzymes used are predicted to cut into the probe (Fig. 2B). Because the similarity of the TcT-Eelem to the 3′ UTR of the FL-160 gene was limited to 100 bp out of the 280 bp of the TcT-Eelem, we reasoned that TcT-Eelem might be associated to other genes (i. e. be present in gene contexts other than FL-160 genes). The sequence of the TcT-Eelem was used to search the T. cruzi genome raw data (unassembled whole genome shotgun sequences, or contigs assembled by the genome project) and the position of the TcT-Eelem relative to upstream open reading frames (ORFs) was determined. Interestingly, the polypyrimidine tracts contained within the TcT-Eelem were always found 30–70 bases downstream of the stop codon of a coding region (CDS), in different contigs. The close proximity of the TcT-Eelem to the end of the upstream coding sequence strongly suggested that the TcT-Eelem was part of the 3′ UTR of the gene. By southern blot, using a probe corresponding to the complete ORF of a predicted protein associated with the TcT-Eelem (currently identified as ORF Tcruzi_1863-4-1211-93, TriTrypDB database [21]), we observed a hybridization pattern similar to that obtained using the TcT-Eelem as probe (Fig. 2B), thus reinforcing the genetic linkage between the TcT-Eelem and the identified ORFs (data not shown). By northern blot analysis we confirmed that, like most of the ESTs from the TcT-E library, this ORF was also differentially expressed in the trypomastigote stage (Fig. 2C). Interestingly, fragments of several of the CDSs found associated with the TcT-Eelem in this bioinformatic analysis were also represented in the TcT-E library (for example clones TcT-E01p24 and TcT-E01k23, corresponding to GenBank GW883736 and HO052122, respectively, Fig. 1C). A schematic diagram of the CDS - TcT-Eelem arrangement found by in silico analysis is shown in Figure 2D. Although different coding sequences were located upstream of the TcT-Eelem, we observed that the amino- and carboxy–termini of those conceptually translated proteins were conserved, suggesting that these ORFs are members of the same family. Besides, we detected three bands by northern blot when using the clone TcT-E01k23 (GenBank HO052122) as probe, which corresponds to the last 270 nucleotides of one TcT-Eelem-associated ORF (not shown). Thus, both in silico and experimental observations suggested the presence of a new protein family sharing conserved amino- and carboxy-termini and the 3′ UTR of the mRNA (TcT-Eelem). To further investigate this hypothesis we looked for the presence of full-length transcripts, by performing RT-PCR experiments using a reverse primer specific for the 3′ end of the TcT-Eelem (TcT-Ee-intR) and a forward primer specific for the T. cruzi miniexon (ME) that is added by trans-splicing to all RNA PolII transcripts in T. cruzi [22]. In parallel, other RT-PCR reactions were designed to amplify the CDSs codifying for this new protein family irrespectively of their untranslated region, using the primers indicated in Figure 2D. Bands of ∼1500 bp and ∼900 bp obtained with primers ME/ TcT-Ee-intR as well as the three bands obtained in trypomastigotes with CDS-L/CDS-R primers (Fig. 2E) were cut from the gel, cloned and sequenced. Thirteen different transcripts were identified and deposited at GenBank with the accession numbers AM492199–AM492211. The coding region of the transcripts was conceptually translated and aligned, confirming that they belong to a multigene family with conserved amino- and carboxy-termini and a variable central core (Fig. 3A). The proteins are enriched in Ala, Ser and Val residues, and therefore, the family was named TcTASV (for Trypomastigote Alanine Serine Valine rich protein). According to the length of the central region, we were able to define three subfamilies (A, B and C, see Fig. 3A), with a conserved Glu-Ala-Pro motif in the variable region (asterisks in Fig. 3A). A visualization of the alignment in Fig. 3A using the partial order multiple sequence alignment visualizer POAVIZ (Fig. 3B) [23] helps to define the overall structure of how sequences match and diverge in the alignment, and facilitates the identification of complex branching structures, such as domains or large-scale insertions/ deletions. Aligned regions are joined together in the partial order graph whereas regions that are unaligned are separated, clearly showing the shared and divergent regions of the 3 TcTASV subfamilies, schematically represented in light blue (TcTASV-A), green (TcTASV-B) and orange (TcTASV-C) (Fig. 3B). The predicted molecular weights of the subfamilies are 18 kDa, 27 kDa and 36 kDa for the A, B and C apoproteins respectively. All proteins had a predicted signal peptide (arrows above and below the alignment in Fig. 3A show the predicted cleavage site) and a consensus sequence for the addition of a GPI anchor (red box in alignment), suggesting that the proteins could be located at the parasite surface. A high proportion of Ser and Thr that could be glycosylated (as found for other surface proteins of T. cruzi) were also identified [24], [25]. Surprisingly, many TcTASVs genes were not annotated as genes in the T. cruzi genome (available at http: //TriTrypDB. org) [4], [5], [21]. For example, only the TcTASV-A 2,4 and 8 genes (GenBank AM492200, AM492209 and AM492210, respectively) were annotated as protein coding genes (hypothetical), while all other TcTASVs-A were found as unannotated ORFs in the data base (Table S4). Besides, most TcTASVs-A have been annotated starting in an ATG codon (Met residue) located ∼145 aa upstream of the one we identified here as the site of trans-splicing, based on the amplification of mature mRNAs using a primer specific for the 5′ spliced leader. The best hits found in TriTrypDB for each TcTASV gene and the corresponding additional information (assigned gene number, contig, identity and other observations about the annotation of these genes is presented in Table S4. TcTASV genes are only present in T. cruzi. Sequence similarity searches revealed no orthologues in the genomes of T. brucei and Leishmania spp, and in ESTs obtained from T. rangeli, the most closely-related trypanosomatids. Based on this observation it is possible to hypothesize that TcTASVs may be involved in T. cruzi-specific strategies of survival and/or immune evasion. Although we obtained experimental evidence supporting the presence of 13 TcTASVs in CL-Brener (nine TcTASVs-A, two TcTASVs-B and two TcTASVs-C), it is likely that the TcTASV family is composed by a higher number of members. Recently, Arner et al. developed a public database specifically designed for the identification of repeated genes in the T. cruzi genome [26], the assumption being that the genes present in high copy numbers were collapsed during the assembly. By using this resource, the estimated number of genes was predicted to be 14 for TcTASVs-A, 6 for TcTASVs-B and 22 for TcTASVs-C. Our own detailed inspection of the T. cruzi data base allowed us to identify 20 TcTASVs-A, 5 TcTASVs-B and 13 TcTASVs-C members, giving a total number of 38 genes for the TcTASV family (Table S5, additional material). In the case of TcTASV-A and B families, when predicted as genes, they were annotated as hypothetical proteins. However, 6 out of 7 TcTASV-C genes were annotated as mucin-like genes. The mucin-like family is another family of surface proteins in T. cruzi and, as currently annotated in TriTrypDB, is composed of 28 genes [21]. Although the overall structure of mucin-like genes (conserved amino- and carboxy-termini, predictions for signal peptide and GPI anchor addition) resembles the one for TcTASV genes, mucin-like and TcTASV have very different amino acid composition. The hypotheses that (a) TcTASV is a protein family different from the mucin-like gene family, and (b) the genes that we identified here as TcTASV-C (but were annotated as mucin-like) are indeed members of the TcTASV family and not of the mucin-like family, were tested by a phylogenetic analysis. Starting with an alignment that included all TcTASV and mucin-like genes, we computed a neighbor-joining phylogenetic tree (Figure S1). The tree clearly shows two major branches: one for mucin-like genes and another for TcTASVs genes (including these 6 incorrectly annotated mucin-like genes). On the other hand, the monophyletic origin of TcTASVs in relation to other structurally similar protein families (TcMUCII, mucin-like, and MASP), was also tested through a phylogenetic analysis of 15 sequences from each family (Figure S2) [4], [27], [28], [29]. The limited phylogenetic distribution of the TcTASV family (so far only detected in T. cruzi), prompted us to investigate the presence of TcTASV genes in T. cruzi strains from other evolutionary lineages (T. cruzi I and II). For this, we amplified the genes of the TcTASV family from two representative strains (Dm28 and RA, respectively) using primers specific for the 3′ and 5′ conserved regions. Each of the amplicons obtained for each strain and for each TcTASV subfamily, was cloned to build a mini-library, in order to identify as many members as possible. We obtained 73 clones from the RA strain and 41 from Dm28 strain, but, for further analysis, we selected only those who presented unique sequences for each strain (RA: 48; Dm28: 28; GenBank Acc FN599093–FN599167) (Fig. 4, Table). The 76 unique sequences obtained for RA and Dm28, together with 8 sequences of CL-Brener (TcTASV-A: 4, TcTASV-B: 2 and TcTASV-C: 2) were used to compute a phylogenetic tree, using sequences of other T. cruzi glycoprotein families (mucin-like, TcMUCII and MASP) as outgroups (Fig. 4). All three (TcTASV-A, B and C) subfamilies were identified in this dataset. We also identified a new subgroup composed of six Dm28 and one RA sequences with mixed characteristics that could constitute a new TcTASV subfamily with some characteristics shared with members of the A subfamily (amino acid sequence) and others shared with members of the C subfamily (length) (Fig. 4, gray box). As in the case of the CL-Brener strain, the subfamilies with most members were TcTASV-A and TcTASV-C and, interestingly, we noted the absence of the TcTASV-B subfamily in the Dm28 strain, which could be explained either by the absence of TcTASV-B genes in this strain or by the accumulation of mutations that prevented the amplification of members of this subfamily in our PCR experiments. Another distinguishing characteristic between the two evolutionary lineages of T. cruzi is that proteins of the subfamily C found in lineage II (RA and CL-Brener) are longer than the same proteins found in lineage I (Dm28) (Fig. 4, Table). To assess the expression of TcTASV family, we took advantage of proteomic data, available in TcruziDB/TriTrypDB, together with experimental data obtained in this work. Mass spectrometry data strongly suggest the differential expression in trypomastigotes of al least 1 out of 4 TcTASV-A genes -Tc00. 1047053506337. 80, Tc00. 1047053506337. 100, Tc00. 1047053510717. 10 and, Tc00. 1047053510717. 20- that share a peptide that was detected only in this parasite stage (KPGEYESVTDDCAR, 2 spectra) (Table S5) [28], [30]. On the other hand, proteomic evidence of the expression of the TcTASV-A8 gene (GenBank AM492210; Tc00. 1047053506573. 5) has been reported for trypomastigotes (five mass spectra) and amastigotes (one spectrum) [28] and can be accessed through TcruziDB. org [30]. The peptide identified is also 100% identical to amino acids 89–104 of other TcTASV gene products (A9: GenBank AM492211, A7: GenBank AM492202 and A5: GenBank AM492201). Moreover, the peptide is completely conserved (15/16 identical aa) in all the other TcTASV-A members. The expression pattern of members of the TcTASV-A subfamily was also analyzed using affinity-purified antibodies that had been generated against a peptide that is conserved throughout the subfamily (see Methods). We were able to find TcTASV-A proteins only in cell-derived trypomastigotes, detecting two bands of ∼18 kDa by western blot (Figure S3). Our first goal in this work was the identification of genes preferentially expressed in the trypomastigote stage of Trypanosoma cruzi, the etiological agent of Chagas' disease. To achieve this goal we followed an approach based on the sequencing of a subtractive cDNA library. Most of the clones of this TcT-E library represent mRNAs that are preferentially expressed in trypomastigotes, as confirmed by northern and reverse northern blots (Fig 1). The sequence information derived from the TcT-E library allowed us to identify genes that were not previously described in T. cruzi. For example, we found several clones with similarity to proteins that have been proposed to function in processes such as rRNA processing, ribosome assembly and the control of cell cycle in other eukaryotic organisms (Bop1, Nop56, BEM, Cwf17; see Tables S1 and S2) [31], [32], [33], [34]. Little is known in T. cruzi about these checkpoints in the cell cycle and it is interesting to note that the preferential expression of these mRNAs was detected in a non-replicative stage of the parasite. The lack of transcriptional control in trypanosomatids is well known, and, therefore, stage-specific differences in mRNA abundance are likely to be the result of selective mRNA stabilization and/or the absence of degradation mechanisms for those transcripts [22]. Therefore, one possibility is that these transcripts are being accumulated for the production of the corresponding proteins once the trypomastigote differentiates into the replicating amastigote within the host cell, or when the trypomastigote differentiates into epimastigotes upon entering the insect vector. After clustering the TcT-E dataset we identified a sequence that was found to be over-represented in trypomastigotes and that has a subregion of 100 bp with high similarity to the 3′ UTR of the T. cruzi flagellar antigen FL-160-2. This observation called our attention because the FL-160-2 gene is a member of a numerous family that is differentially expressed on the surface of trypomastigotes and is involved in parasite virulence [19], [20]. All these facts suggested that the 100 bp region could be part of a longer conserved region, and, indeed, we reconstructed a 280 bp element by multiple sequence alignment of TcT-E clones that matched the 100-bp motif that we named TcT-E element (TcT-Eelem), because of its high representation in the TcT-E library. Although we ended up associating the TcT-Eelem with the 3′ UTR of the new TcTASV family, we also observed that part of the TcT-Eelem (∼120–150 nt) is also found downstream of genes that do not belong to this family. For example, some hypothetical proteins, trans-sialidase genes or other coding sequences harbouring part of the TcT-Eelem were identified (Tc00. 1047053507875. 70, Tc00. 1047053504533. 40, Tc00. 1047053507491. 20). Interestingly, in all those cases, the 120–150-bp subregion of the TcT-Eelem is farther downstream from the stop codon than in the case of TcTASVs genes. Post-transcriptional cis-acting elements conserved among different genes and included into more extended 3′ UTRs have been previously identified. The existence of a regulatory region of 770 bp that is specific for amastin genes and that contains a 450-bp zone shared by amastin and other developmentally-regulated mRNAs has been reported in Leishmania [35]. This 450-bp sub-region (currently known to be part of the LmSIDER1 subfamily) mediates the translational regulation of mature transcripts in response to elevated temperature, the main environmental change that the parasite encounters upon its transmission from the vector to the mammalian host [36], [37]. Taking this into account, it could be hypothesized that a general stage-specific regulation of genes can be achieved in a similar way in T. cruzi. The 120–150-bp motif that is shared by different genes preferentially expressed in trypomastigotes, such as FL-160, TS and TcTASVs, could be involved in this stage-specific expression, probably forming part of a post-transcriptional regulon that allows the coordinated expression of these genes [38]. The new TcTASV gene family described in this work is composed of 38 members in the CL-Brener strain, none of which show significant similarity to other surface proteins in T. cruzi. A meticulous comparative analysis between sequences of TcTASV, MASPs, TcMUCII and mucin-like genes, shows that each of the families diverge in a diferent branch of the computed phylogenetic tree, thus reinforcing the idea that these are indeed different protein families. After the recent re-assembly of the genome of T. cruzi [5], previously annotated genes that we have now identified as TcTASVs could be found in 5 chromosomes, with a high proportion of TcTASVs-A on chromosome 16 and almost all annotated TcTASV-Cs on chromosome 24. TcTASV genes are apparently not arranged in tandem and most of them are surrounded by other hypothetical proteins (they are not TcTASVs). However, the majority of the TcTASV genes were not annotated by the genome-sequencing consortium and are still left out of the final genome assembly (they are only present as ORFs identified in unassembled or small partially assembled contigs). This is highly suggestive of assembly problems that occur frequently when highly similar genes are present in a moderate to high copy numbers. Therefore, it is possible that the copy number of TcTASVs genes in the CL-Brener genome could have been underestimated because of the collapse of repeated genes into fewer copies during assembly. However, the identification of a similar number of members of the TcTASV family in another type II strain (RA) of T. cruzi probably indicates that for lineage II the number of TcTASV genes is around 45, whereas those for lineage I is probably around 30, i. e. , the lowest number. These conclusions were derived from PCR experiments using oligonucleotides designed on highly conserved regions. Therefore, there is still the possibility that other TcTASV genes could not be amplified by these primers. Complex glycoproteins cover the surface of all the developmental stages of trypanosomatid human pathogens [2], [39]. Among the species-specific families, the best-studied ones are probably the mucins of T. cruzi and the proteophosphoglycans (PPGs) of Leishmania spp. Both protein families are rich in Ser, Thr and Pro residues, are retained in the membrane by GPI anchors and can be released from the parasite. In the case of TcTASVs, the amino acid composition is different, being enriched in Ala, Ser and Val. Regarding their expression, different groups of mucins and proteophosphoglycans are developmentally expressed, i. e. TcMUC I and II are expressed in the mammalian stages of T. cruzi, whereas TcSMUGs are only found in insect-derived stages ([40] and reviewed in [41], [42], [43]). In Leishmania, filamentous PPGs are secreted by promastigotes and have been implicated in protection from digestive enzymes in the insect midgut and in the formation of a plug in the sandfly digestive tract, which causes an increased frequency of feeding and correlates with parasite invasion and virulence [44], [45], [46]. On the other hand, membrane-bound PPGs have been implicated in parasite binding and invasion of macrophages [47], [48], [49]. For both mucins and PPGs several mechanisms leading to immune system evasion have also been demonstrated [41], [42], [49], [50], [51], [52], [53]. Therefore, it is clear that the parasite expresses different kind of mucins or PPGs, even with a differential cellular localization, in the different developmental stages in order to invade and persist in the parasitized host. In this work we demonstrate that TcTASV-A are expressed in trypomastigotes and could not be detected in other stages, suggesting that the TcTASV population could undergo developmental regulation. However, we cannot completely rule out the possible expression in other parasite stages, because we did not analyze the expression of the TcTASV-B nor TcTASV-C subfamilies and used only one peptide to obtain anti-TcTASV-A antibodies. Related to this, after following a proteomic approach Atwood et al. were able to identify a peptide in trypomastigote and amastigote extracts that is completely conserved (100% identity) in the TcTASV-A subfamily. The expression of TcTASV-A in amastigotes, though, probably occurs at very low levels since we were unable to detect TcTASV-A proteins in this parasite stage. Moreover, only one spectrum was detected by Atwood et al. in amastigotes (vs. 5 in trypomastigotes) [28]. Based on computational analyses, we predicted a signal peptide in the amino terminus and a potential site for the addition of a GPI anchor at the carboxy terminus of TcTASVs. However, at this moment, we cannot rule out the possibility that some members of TcTASV have a membrane-associated expression and others a cytosolic or secreted form. In summary, in the present work we have identified and partially characterized a new surface protein family in T. cruzi wich we named TcTASV. All TcTASV members have a conserved 3′ untranslated region (the TcT-Eelem, also identified for the first time here), conserved amino- and carboxy- termini, and could be grouped into three subfamilies according to the relative molecular mass of the predicted proteins. The presence of a high number of Ser and Thr susceptible to glycosylation as well as a signal peptide and a consensus sequence for the addition of a GPI anchor were predicted. The expression of the TcTASV-A subfamily in trypomastigotes was demonstrated. One other interesting characteristic of the TcTASV family is the lack of orthologues in other trypanosomatids. Finally, we would like to emphasize that TcTASV is a new gene family in T. cruzi, which so far had remained unnoticed (unannotated or missing from the assembled genome). We have worked closely with other groups to make sure that this is solved in future releases of T. cruzi genome databases. However, given the still draft nature of the T. cruzi genome, the possibility exists that this can happen for other genes. Moreover, by means of a genetic vaccination approach, one of the members of TcTASV (formerly TcYASP) has been found as part of a protective pool of antigens [3], which suggests that they are possible good vaccine candidates.

Chagas' disease, caused by the kinetoplastid protozoan parasite Trypanosoma cruzi, is endemic in Latin America. At present there are neither vaccines for prevention nor totally effective drugs for the treatment of the disease. T. cruzi has a complex life cycle alternating between a reduviid insect (the vector) and a mammalian host, where different parasite stages are found. Differentially expressed genes are the hallmark of the specialized biology of each life cycle stage. The aim of this work was to identify genes expressed in the trypomastigote stage (a blood-circulating stage that invades new cells and spreads the infection in different organs of the mammalian host) that could be used to develop new vaccines or diagnostics. An initial screening of trypomastigote transcripts was performed by sequencing of an epimastigote-subtracted trypomastigote cDNA library. Besides identifying a large proportion of differentially expressed mRNAs, we discovered a novel protein family, which we denominated TcTASV.

Abstract Introduction Methods Results Discussion

genetics and genomics/gene discovery genetics and genomics/gene expression infectious diseases/neglected tropical diseases microbiology/parasitology infectious diseases/protozoal infections genetics and genomics/bioinformatics

2010

TcTASV: A Novel Protein Family in Trypanosoma cruzi Identified from a Subtractive Trypomastigote cDNA Library

The interaction of Mycobacterium tuberculosis (Mtb) with host cell death signaling pathways is characterized by an initial anti-apoptotic phase followed by a pro-necrotic phase to allow for host cell exit of the bacteria. The bacterial modulators regulating necrosis induction are poorly understood. Here we describe the identification of a transcriptional repressor, Rv3167c responsible for regulating the escape of Mtb from the phagosome. Increased cytosolic localization of MtbΔRv3167c was accompanied by elevated levels of mitochondrial reactive oxygen species and reduced activation of the protein kinase Akt, and these events were critical for the induction of host cell necrosis and macroautophagy. The increase in necrosis led to an increase in bacterial virulence as reflected in higher bacterial burden and reduced survival of mice infected with MtbΔRv3167c. The regulon of Rv3167c thus contains the bacterial mediators involved in escape from the phagosome and host cell necrosis induction, both of which are crucial steps in the intracellular lifecycle and virulence of Mtb. Apoptosis is a major programmed cell death pathway but now it is well established that necrosis can also be induced via defined signal transduction pathways [1,2]. The importance of apoptosis in host defense against pathogens is well described [3,4]. In contrast, the function of programmed necrosis in host resistance or susceptibility to pathogens is still an open question in many cases and may depend upon the context of the infection and the pathogen [5]. For instance, the RIPK1/3 necrosis pathway acts as a back-up mechanism of death induction in cells infected with viruses that are able to inhibit host cell apoptosis [6]. Consequently, programmed necrosis is associated with increased host resistance against viral pathogens in the case of vaccinia virus, adenovirus and MCMV [5,6]. Nevertheless, for the influenza A virus, programmed necrosis leads to increased pathology and host susceptibility [7]. Limited results are available for interaction of bacterial pathogens with host cell necrosis pathways but similar to viral pathogens the role of programmed necrosis may vary depending upon the pathogen. Enteropathogenic Escherichia coli can inhibit RIPK3-dependent necrosis via the glycosyl transferase NleB and this activity is important for bacterial virulence [8,9]. In contrast, IRF-3-dependent necrosis induction by Listeria monocytogenes promotes pathogen dissemination and virulence [10]. The interaction of wild-type Mycobacterium tuberculosis (Mtb) with its host cell in regard to cell death signaling is complex [11–13]. According to one model, virulent strains of Mtb are capable of inhibiting host cell apoptosis during the early phase of the infection to allow for intracellular replication but the bacteria induce necrosis in order to exit the host cell at a later stage [14]. The discovery of Mtb genes that inhibit host cell apoptosis such as nuoG [15], pknE [16], secA2 [17], Rv3654c [18], and ndk [19] supports this model. Furthermore, the Mtb nuoG mutant is attenuated in the mouse model of tuberculosis, thus illustrating the importance of host cell apoptosis inhibition for Mtb virulence [15]. Consistently, mice with reduced host cell apoptosis induction upon Mtb infection are more susceptible [20]. The mechanisms leading to increased host resistance include an increase in efferocytosis of apoptotic host cells leading to killing of the bacteria [21,22]. In addition, there are various lines of evidence that increased host cell apoptosis will lead to a more rapid and increased cytolytic T-cell response [17,23,24]. In contrast to apoptosis, host cell necrosis induction is associated with increased host susceptibility and virulence of Mtb as well as Mycobacterium marinum (Mm) in mice and in zebrafish [20,25]. Several studies demonstrated the central role of host cell eicosanoids lipoxin A4 (LXA4) and prostaglandin E2 (PGE2) in the regulation of host cell apoptosis versus necrosis induction and their importance for bacterial virulence and host resistance [24,26,27]. The enzyme Leukotriene A4 hydrolase (LTA4H) regulates synthesis of the eicosanoids LXA4 and leukotriene B4 (LTB4); excessive production of either lipid mediator leads to macrophage necrosis [28]. Polymorphisms in LTA4H in humans are associated with hypersusceptibility to mycobacterial infections [29]. Lysosomal destabilization and macrophage necrosis was found to occur following accumulation of about 20 or more intracellular bacteria [30,31]. The escape of Mtb from the phagosome to the cytosol precedes necrosis and exit from the host cell [32–34]. The Mtb type VII secretion systems, ESX-1 and ESX-5, are implicated in host cell necrosis induction. The deletion of the Mtb ESX-1 secretion system leads to a reduced induction of host cell necrosis and dissemination of the mutant mycobacteria [35–37], this could be due the inability of mutant strains to escape from the phagosome [33,34,38]. The Mtb ESX-5 system is involved in mediating cell necrosis after the bacteria have escaped the phagosome [39]. The PE25/PP41 complex secreted via ESX-5 may be one of the effectors of ESX-5-mediated host cell necrosis as addition of the purified protein complex induced necrosis of macrophages [40]. Host cell necrosis induction by Mtb is important for cell exit and dissemination but the molecular mechanisms involved are still poorly understood. Here we describe the discovery of a tetracycline repressor family protein, Rv3167c, which negatively regulates the capacity of the bacteria to induce host cell necrosis. Infection of macrophage with the Rv3167c deletion strain (MtbΔRv3167c) led to a rapid increase in host cell necrosis via a novel host cell signaling pathway that involves the reduced activation of the protein kinase Akt leading to an increase in mitochondrial reactive oxygen species (mROS). Interestingly, we discovered that MtbΔRv3167c escape the phagosome in higher numbers than wild-type Mtb, which most likely triggers the host cell necrosis signaling. Finally, aerosol infection of mice demonstrated the increased virulence of MtbΔRv3167c. In conclusion, we find that Rv3167c regulates the escape of Mtb from the phagosome, which marks the beginning of the host cell exit program of the Mtb intracellular life cycle. We previously performed a gain-of-function genetic screen and identified a genomic region in Mtb H37Rv containing anti-apoptotic genes (S1A Fig) [15]. A series of deletion mutants spanning several genes within this region was generated and tested for loss of apoptosis inhibition. THP1 cells were infected with wild-type Mtb (Mtb) and the deletion mutants and stained for genomic DNA fragmentation using TUNEL assay. Two deletion mutants, one being the single gene nuoG mutant [15], and the other a five gene deletion mutant designated 7/10, induced higher levels of cell death compared to the Mtb control (S1B Fig). Screening of genes within the 7/10 region revealed that deletion of Rv3167c had the maximal effect on loss of cell death inhibition. Infection with the deletion mutant MtbΔRv3167c (MtbΔ) resulted in almost a 3-fold increase in TUNEL-positive THP1 cells compared to infection with the control Mtb strain (S1C Fig). Both southern blotting and RT-PCR confirmed deletion of Rv3167c (S2A–S2C Fig). Increased cell death induction by MtbΔRv3167c was also observed in primary human monocyte derived macrophages (hMDMs) by hypodiploid staining which measures loss of genomic DNA content following cell death (Fig 1B). Cell death induction by the complement strain MtbΔRv3167c-C (MtbΔC) was comparable to Mtb (Fig 1A and 1B), thus confirming that Rv3167c is required for Mtb-mediated host cell death inhibition. Replication of MtbΔ is similar to Mtb; both, in infected THP1 cells and in growth media (S2D and S2E Fig). Rv3167c is most likely a member of the tetracycline-like family of regulators (TFR) since 89% of the Rv3167c amino acid sequence can be modeled with 99. 9% confidence to the highest scoring template, the TFR SCO0332 of Streptomyces coelicolor, using Phyre2 software. Although TUNEL staining has been historically used for detection of apoptotic DNA fragmentation, recent studies have shown that necrotic cells can also be TUNEL positive [41–43]. A characteristic feature of apoptotic cells is preservation of cell membrane integrity [3]. To determine whether cell death induced by MtbΔRv3167c is accompanied by cell membrane damage, we tested for the presence of adenylate kinase, normally located within healthy cells, in the supernatants of THP1 cells using the Toxilight assay. At 24 and 48h, a 3-fold higher level of adenylate kinase activity was detected in supernatant from MtbΔRv3167c-infected cells compared to uninfected controls (Fig 1C). Mtb-infected cells also undergo necrosis albeit at lower levels compared to MtbΔRv3167c-infected cells. Mtb has been previously shown to induce necrosis in a dose and time dependent manner and our data supports this observation [44]. It is important to note that the Toxilight assay cannot differentiate between primary necrosis or secondary necrosis of apoptotic cells and consequently further analysis into the nature of the induced cell death was required. Execution of apoptotic cell death requires the cleavage and activation of the effector caspases-3, -6 and -7 [2]. We infected wild type (WT) and Casp3-/- bone marrow derived macrophages (BMDMs) and performed TUNEL staining to confirm that MtbΔRv3167c does not induce apoptotic cell death. No differences in TUNEL-positive cells were observed between MtbΔRv3167c-infected WT and Casp3-/- BMDMs (Fig 1D and 1E). To ensure that the lack of death inhibition in Casp3-/- BMDMs is not due to redundancy of caspase-3 with other effector caspases, the pan-caspase inhibitor zVAD-FMK was added to MtbΔRv3167c-infected cells. Inclusion of zVAD-FMK did not inhibit MtbΔRv3167c-induced cell death in both WT and Casp3-/- BMDMs (Fig 1E) although it did inhibit apoptosis induced by camptothecin (S3A Fig). We also did not observe cleavage of the DNA repair enzyme PARP, another feature of apoptosis, in Mtb or MtbΔRv3167c-infected THP1 cells (Fig 1F). Furthermore, the zVAD-FMK inhibitor was added to infected Ripk3-/- cells and had no effect on MtbΔRv3167c-induced cell death (S3D Fig). These results show that Rv3167c is required for inhibition of Mtb-induced necrotic host cell death. Next, we investigated the involvement of programmed necrosis pathways in cell death mediated by MtbΔRv3167c. In conditions where caspase-8 expression and activation are inhibited, the serine threonine protein kinases RIPK1 and RIPK3 induce necrosis downstream of TNF-receptor ligation via increased reactive oxygen species (ROS) generation, mitochondrial fission and formation of plasma membrane pores [45–48]. RIPK1 and RIPK3 have also been implicated in TNF-mediated necrosis in Mm-infected zebrafish [49]. We investigated the involvement of RIPK1 and RIPK3 in MtbΔRv3167c-induced cell death by using Ripk3-/- BMDM’s and the RIPK1 inhibitor necrostatin1 (Nec1). Similar levels of PI-positive cells were observed in MtbΔRv3167c-infected Ripk3-/- BMDMs and Nec1 treated cells compared to WT BMDMs and solvent control-treated cells respectively (Figs 2A and S3B). Nec1 efficacy was confirmed by its ability to inhibit LPS and zVAD FMK induced RIPKI dependent cell death (S3C Fig) [50]. Necrosis induction following TNF treatment has been reported to change to apoptosis in the absence of RIPK1 and consequently, the absence of an effect on cell death induction by MtbΔRv3167c could be due to an increase in apoptosis in cells deficient in RIPK1/3 signaling [1]. Addition of zVAD-FMK to Ripk3-/- cells did not affect MtbΔRv3167c-induced cell death thereby ruling out a switch between apoptosis and necrosis in MtbΔRv3167c-infected cells (S3D Fig). Infection of Tnfr1-/- BMDMs established that MtbΔRv3167c-induced necrosis was independent of TNF signaling (S4A Fig). The DNA repair enzyme PARP1 has been implicated in necrosis induction via ATP depletion and nuclear translocation of mitochondrial apoptosis inducing factor in response to DNA alkylating agents and infection with BCG and enterovirus71 [51–53]. Necrosis induction by MtbΔRv3167c is independent of PARP1 since similar levels of necrosis were observed in infected Parp1-/- BMDMs and WT control cells (Fig 2B). The pro-inflammatory caspases, caspase-1 and caspase-11 have been shown to be involved in necrosis induction in response to several bacterial pathogens [54,55]. The role of these caspases in MtbΔRv3167c-induced necrosis was excluded by PI staining of Casp1/11-/- BMDMs (Fig 2C). NLRP3-dependent but caspase-1-independent necrosis has been reported to occur in response to infection with Mtb and Shigella flexneri [56,57]. Using immortalized NLRP3-deficient BMDMs we ruled out involvement of NLRP3 in MtbΔRv3167c-induced necrosis (S4C Fig). Silencing of the inflammasome component ASC did not inhibit MtbΔRv3167c-induced necrosis as measured by the toxilight assay, rather necrosis was increased in MtbΔRv3167c-infected THP1shASC cells compared to control cells (S4D Fig). Involvement of IFNβ signaling (S4E and S4F Fig) and TLR signaling (S5A–S5C Fig) was also ruled out in necrosis induction by MtbΔRv3167c. These data indicate that MtbΔRv3167c does not engage the pathways of programmed necrosis currently described in the literature to induce host cell death. Lysosomal membrane permeabilization (LMP) and the subsequent release of lysosomal contents into the cell cytosol leads to cell death [58,59]. Moderate lysosomal permeabilization leads to apoptosis while more severe damage precedes necrosis [60]. Lysosomal permeabilization and cathepsin release into the cytosol has been previously observed in cells infected with Mtb at high MOI [61]. To investigate whether LMP contributes to MtbΔRv3167c-induced cell death, we used the dye acridine orange (AO) that accumulates within lysosomes. A two-fold increase in the number of cells with loss of AO staining was observed as early as 8h post infection in MtbΔRv3167c-infected THP1 cells compared to Mtb-infected controls (Fig 2D). After 20h of infection the difference was even more pronounced with only about 8% of Mtb-infected cells showing low AO-staining compared to about 25% in mutant infected cells (Fig 2D). This indicates that LMP precedes necrosis induction by MtbΔRv3167c and is not merely a consequence of the disintegration of the cell triggered via a different mechanism. Autophagy is a catabolic process that allows for cell survival via recycling of cellular contents and contributes to pathogen elimination [62]. However, autophagy induction can also lead to cell death [63]. Therefore, we investigated whether MtbΔRv3167c could induce autophagy in macrophages. First, we analyzed recruitment of LC3 into aggregates, an indicator of autophagy, by confocal microscopy [64]. THP1 cells expressing GFP-tagged LC3 (THP1 LC3GFP) were infected with AF647-NHS stained bacteria and at 8h, the percentage of cells showing aggregation of LC3 was estimated. A two-fold increase in autophagosome formation was observed in MtbΔRv3167c-infected cells compared to Mtb-infected control cells (Fig 3A). Previous studies have shown that Mtb induces xenophagy resulting in co-localization of bacteria with autophagosomes and bacterial killing [65,66] However, we observed minimal colocalization of both Mtb and MtbΔRv3167c (<1%) with autophagosomes (Fig 3A). This was confirmed by examination of infected THP1 cells by transmission electron microscopy (TEM) (Fig 3B). Next, we measured conversion of cytosolic LC3I to autophagosomal membrane bound LC3II, another hallmark of autophagy [64]. Uninfected and infected THP1 LC3GFP cells were washed with saponin-containing buffer leading to removal of cytosolic LC3I-GFP. Retention of autophagosomal membrane-bound LC3II-GFP was examined by flow cytometry [67]. A two-fold increase in autophagy induction was observed in MtbΔRv3167c-infected cells compared to those infected with Mtb and MtbΔRv3167c-C (Fig 3C). Increased conversion of both GFP tagged and endogenous LC3I to LC3II in MtbΔRv3167c-infected cells was also seen by immunoblotting (Fig 3D). Autophagy induction by MtbΔRv3167c was confirmed using 3-methyladenine (3-MA), a classical autophagy inhibitor [64]. Inclusion of 3-MA inhibited LC3II formation by MtbΔRv3167c in THP1 LC3GFP cells (Fig 3E). Increased accumulation of LC3II can be attributed either to an increase in autophagosome formation or to a decrease in LC3II degradation due to inhibition of autophagosome-lysosome fusion and maturation [64]. Addition of the vacuolar H+ ATPase inhibitor bafilomycin A1 (BafA1) that inhibits autophagosomal degradation to MtbΔRv3167c-infected THP1 LC3GFP cells led to a further increase in LC3II levels (S6A Fig). Autophagosome maturation leads to the degradation of LC3GFP to yield free GFP [64]. GFP was detected only in MtbΔRv3167c-infected cells by immunoblotting (S6B Fig). These data indicate that MtbΔRv3167c-induced autophagy but did not inhibit the maturation of the autophagosome. Finally to determine whether necrotic death of MtbΔRv3167c-infected cells was a consequence of autophagy induction, we measured adenylate kinase release from Atg5fl/fl LysM Cre+ (Atg5-/-) and Atg5fl/fl LysM Cre- (Atg5+/+) BMDMs. We observed no differences in necrosis induction by MtbΔRv3167c in autophagy-deficient Atg5-/- cells compared to the Atg5+/+ controls (Fig 3F). Additionally, the inclusion of 3-MA did not result in inhibition of MtbΔRv3167c-induced cell death in THP1 cells (S6C Fig). Therefore, while MtbΔRv3167c-infected cells undergo macroautophagy, this does not contribute to their death via necrosis. The concept that Mtb resides within phagosomal compartments at all times has been challenged by recent studies demonstrating bacillary escape to the cytosol both ex vivo and in vivo [32–34,38]. Necrosis induction by Mtb and Mm was shown to closely follow escape to the cytosol [33,68]. We examined cytosolic escape by MtbΔRv3167c using a fluorescence resonance energy transfer (FRET) based assay [33,34,69]. Uninfected and infected THP1 cells differentiated for three days with PMA were loaded with CCF4-AM. Intact CCF4-AM emits green fluorescence (535nm) due to FRET between the fluorescent moieties. Cleavage of CCF4-AM by β-lactamase expressed by cytosolic bacteria leads to FRET loss and a shift in the emission wavelength to 450nm that was measured by flow cytometry. Cells were co-stained with Live/Dead Fixable Red stain to restrict analysis to live cells only. While minor increases in fluorescence emission at 450nm were observed in Mtb-infected cells compared to uninfected cells at 48h, the largest shift in the CCF4 emission spectrum was seen in MtbΔRv3167c-infected cells (Fig 4A and 4B). A three-fold increase was observed in MFI450nm of cells infected with MtbΔRv3167c compared to Mtb-infected controls. Increased cytosolic escape of MtbΔRv3167c was reversed following complementation (Fig 4A–4C). The pro-necrotic phenotype of MtbΔRv3167c was preserved in these macrophages (Fig 4D). Bacterial β-lactamase activity was not affected by either deletion of Rv3167c or gene complementation (S7 Fig). To confirm the increased cytosolic escape by MtbΔRv3167c, we examined infected THP1 cells by TEM and quantified cytosolic bacteria in a double-blinded fashion by examining for absence of phagosomal membranes in healthy cells (Fig 4E, above). Increased presence of MtbΔRv3167c was observed in the cytosol at 8h compared to controls although statistical significance was not achieved. At 24h, 60% of MtbΔRv3167c were found to be cytosolic compared to 20% of Mtb corroborating the increased cytosolic escape by MtbΔRv3167c observed with the CCF4-AM assay (Fig 4E, below). A reduction of cytosolic escape was observed in cells infected with the complemented MtbΔRv3167c-C strain (Fig 4E, below). These data suggest that Rv3167c negatively regulates Mtb escape from the phagosome to the cytosol, an event that has been shown to be followed by induction of host cell necrosis [33,34]. Next we investigated the molecular mechanisms underlying autophagy induction by MtbΔRv3167c. The mitogen activated protein kinases (MAPKs) JNK and p38 have been implicated in autophagic responses of cells infected with Mtb following exposure to cytokines and vitamin D3 [70,71]. Mtb Eis has been shown to inhibit autophagy induction by suppressing JNK activation [72]. Hence we examined MAPK activation in response to MtbΔRv3167c infection by immunoblotting for phosphorylated forms in whole cell lysates prepared from infected THP1 LC3GFP cells at the indicated times. JNK activation was not observed at 0h, however increased JNK phosphorylation was detected in MtbΔRv3167c-infected cells compared to those infected with Mtb and MtbΔRv3167c-C at 18h (Fig 5A). In contrast to JNK, increased p38MAPK activation was observed in MtbΔRv3167c-infected cells at 0h. However by 18h, p38MAPK phosphorylation in MtbΔRv3167c-infected cells was similar to control cells infected with Mtb and MtbΔRv3167c-C (Fig 5A). Consistent results were observed in human monocyte-derived macrophages (hMDMs) as well; however, elevated JNK activation could be detected earlier at 0h in MtbΔRv3167c-infected cells (Fig 5B). We then determined whether JNK and p38MAPK contributed to autophagy induction by MtbΔRv3167c. THP1 LC3GFP cells were pre-treated and infected with JNK (SP600125) or p38MAPK (SB203580) inhibitors prior to infection with MtbΔRv3167c; the percentage of autophagic cells was measured by flow cytometry. Inclusion of the JNK inhibitor led to a partial, dose-dependent decrease in autophagy induction by MtbΔRv3167c while the p38MAPK inhibitor exerted no effects (Fig 5C). Neither of the inhibitors reversed the pro-necrotic phenotype of MtbΔRv3167c, instead a modest increase in PI-positive cells was observed in both cases (Fig 5D and 5E). The serine threonine protein kinase Akt functions as a critical negative regulator of autophagy at the initiation stage by activating mTOR and at the nucleation step by phosphorylating Beclin1 [73,74]. Inhibition of Akt activation has been implicated both in macroautophagy induction in response to nutritional stresses as well as selective autophagy-induction in response to pathogens such as Toxoplasma gondii and Salmonella typhimurium [73,75,76]. To determine role of Akt in autophagy-induction by MtbΔRv3167c, we examined Akt phosphorylation by immunoblotting. A complete loss of Akt activation was observed in MtbΔRv3167c-infected cells compared to the controls (Fig 5F). Consistently, the Akt activator, sc-79, inhibited autophagy induction by MtbΔRv3167c in a dose-dependent manner (Fig 5G) [77]. Additionally, Akt inhibition exerts effects on MtbΔRv3167c-mediated necrosis as well, since sc-79 significantly reduced MtbΔRv3167c-induced necrotic cell death (Fig 5H). Mtb-mediated suppression of reactive oxygen species (ROS) generated by the host phagocytic NADPH oxidase complex (NOX2) has been shown to inhibit host cell apoptosis [78]. Conversely, necrosis-induction by Mm has been shown to require mitochondrial ROS generation [49]. To determine whether necrosis-induction by MtbΔRv3167c involves ROS, we first measured ROS levels in cells infected by MtbΔRv3167c. Uninfected and infected BMDMs were stained with the either 2’7’-dichlorofluorescein diacetate (DCFDA) or MitoSOX Red for measurement of cytosolic and mitochondrial ROS respectively. At 0h, similar levels of ROS were observed in all infected cells compared to the uninfected controls (Fig 6A and 6B). However by 24h, approximately three-fold higher levels of both cytosolic and mitochondrial ROS were detected in MtbΔRv3167c-infected cells compared to those infected with Mtb and MtbΔRv3167c-C (Fig 6A and 6B). Next we examined whether increased ROS levels contribute to necrosis induction by MtbΔRv3167c using the flavoprotein inhibitor diphenylene iodonium (DPI) and the ROS scavengers glutathione and N-acetyl cysteine (NAC). Inclusion of these inhibitors and scavengers reversed necrosis induction by MtbΔRv3167c to levels observed in uninfected cells (S8A and S8B Fig). Thus, elevated ROS levels in MtbΔRv3167c-infected cells contribute to their necrotic cell death. Furthermore, addition of DPI to MtbΔRv3167c-infected THP1 LC3GFP cells completely abrogated autophagy induction compared to control cells (S8C Fig). As ROS in eukaryotic cells may be derived from the NOX2 complex or mitochondria, we sought to determine which of these sources is implicated in necrosis-induction by MtbΔRv3167c. Similar levels of necrosis induction by MtbΔRv3167c were detected in WT and Nox2-/- BMDMs by PI staining (Fig 6C). However, a complete inhibition of MtbΔRv3167c-mediated cell death was observed in mCAT BMDMs obtained from transgenic mice overexpressing mitochondrial targeted human catalase (Fig 6D). Increased mitochondrial ROS generation was also accompanied by a time dependent loss of mitochondrial membrane potential as measured by DIOC6 staining of MtbΔRv3167c-infected cells (S8D Fig). Increased mROS generation in MtbΔRv3167c-infected cells was found to be attributable to reduced Akt activation as inclusion of the Akt activator sc-79 inhibited mROS generation (Fig 6E). Taken together, our data reveal Akt and mitochondrial ROS to be critical regulators of MtbΔRv3167c-mediated necrosis and autophagy. We assessed the contribution of Rv3167c to Mtb virulence in vivo by performing a survival study of C57Bl/6 mice infected with approximately 100 CFU of Mtb, MtbΔRv3167c and MtbΔRv3167c-C via the aerosol route. Increased mortality was observed in MtbΔRv3167c-infected mice (median survival time—33 weeks) compared to those infected with Mtb or MtbΔRv3167c-C (median survival times– 59 and 60 weeks respectively) (Fig 7A). Decreased survival following MtbΔRv3167c infection was also observed in immunodeficient SCID mice (S9A Fig). Lung bacterial burden on day one after infection was similar for all three strains indicating comparable initial inoculum of infection in both C57Bl/6 and SCID mice (Figs 7B and S9B). Relative to control mice, the lung bacillary burden was 10-fold higher in MtbΔRv3167c-infected animals at14 and 28 days, and this difference was magnified by day 56 following aerosol infection (Fig 7B). Increased bacterial burdens also were observed in the liver and spleen of MtbΔRv3167c-infected mice relative to control mice (Fig 7C and 7D). Higher levels of pro-inflammatory cytokines (TNF, IL1α and IL6) (Fig 7E) and chemokines (CCL3, CCL5 and MMP9) (Fig 7F) were detected in the lung tissues of mice infected with MtbΔRv3167c. Comparison of lung histopathology revealed a two-fold increase in cellular infiltration in MtbΔRv3167c-infected animals (Fig 7G). Consistent with the findings in the mouse model of chronic TB infection, increased bacterial burdens were observed in the lungs of guinea pigs infected with MtbΔRv3167c relative to controls at 28 days following aerosol infection with similar bacterial loads (S9C and S9D Fig). MtbΔRv3167c was also found to induce cell death in ex vivo infection of guinea pig alveolar macrophages by TUNEL staining (S9E Fig). Collectively, these results show that Rv3167c negatively regulates Mtb virulence. The intracellular location of bacteria has important consequences for their recognition by the host and the generation of innate and adaptive immune responses. Mtb was thought to restrict itself to a modified host cell phagosomal compartment after infection [79,80]. However, electron microscopy studies performed on infected macrophages and dendritic cells provided evidence that Mtb and other mycobacterial species are present in the cytosol and that phagosomal escape is dependent upon the ESX-1 secretion system [32,38]. This was confirmed by a FRET-based method dependent on β-lactamase production by Mtb in ex vivo-infected cells as well as in pulmonary phagocytic cells obtained from infected mice [33,34]. We report in the current study that the mycobacterial gene Rv3167c negatively regulates the escape of Mtb from the phagosome to the cytosol (Fig 4). Our study is the first to suggest that Mtb can exert temporal control on phagosomal escape as MtbΔRv3167c was found in the cytosol as early as 24h after infection while Mtb has been reported to access the cytosol much later in the infection process (4–5 days) [33]. We hypothesize that Rv3167c represses cytosolic escape at early stages when the bacterial load is low, favoring Mtb replication and establishment of infection. Genes involved in cytosolic escape could be induced once bacterial numbers reach about 20 per cell which seems to be an important threshold to switch on the cell escape program of Mtb [30]. Gene deletions resulting in increased cytosolic translocation from vacuolar compartments have been reported for other bacteria; for example, the sdhA (a Dot/Icm-secreted effector) mutant of Legionella pneumophila and the sifA (an SPI2-secreted effector) mutant of Salmonella typhimurium [81,82]. Deletion of the secreted phospholipase A abrogated the early escape of the L. pneumophila sdhA mutant [81]. The Mtb genome encodes four phospholipases (plc A-D), which could potentially contribute to the early escape of MtbΔRv3167c. The mechanisms involved in Mtb escape from the phagosome and eventual induction of host cell necrosis to exit the cell are difficult to study because of the slow kinetic of the process. Consequently, the early induction of cytosolic escape and necrosis by MtbΔRv3167c make it a useful model to study the host cell escape mechanisms of Mtb. The escape of Mtb from the phagosome to the cytosol is closely followed by necrotic death of the host cells [32,33,83]. Consistently, we measured higher levels of cell death by necrosis in MtbΔRv3167c-infected cells (Figs 1A, 1B and S9E). Mtb may induce necrosis via the manipulation of host cell lipid mediators by favoring the production of the eicosanoid LXA4 [27]. In addition, the activation of the NLRP3-inflammasome was also shown to induce necrosis after Mtb infection [56]. We found that NLRP3 is dispensable for MtbΔRv3167c-mediated necrosis (S4C Fig). The regulation of cell death is complex and recently there have been major discoveries of signal transduction pathways for the regulation of programmed necrosis [1]. Using a combination of inhibitors and macrophages from knock-out mice, we screened for host factors required for MtbΔRv3167c-induced necrosis and ruled out the involvement of known programmed necrosis pathways (Figs 2, S4 and S5). Redundancy between the various necrosis-signaling modules may explain this result. For instance, both RIPK1-RIPK3 and caspase-1 activation are required for S. tymphimurium-induced necrotic cell death and blocking either one of the signaling pathways led only to a marginal inhibition of the death phenotype [43]. However, we found elevated mitochondrial ROS (mROS) to be required for the pro-necrotic phenotype of MtbΔRv3167c (Fig 6). Elevated mROS production could lead to increases in cytosolic ROS levels that may via lipid oxidation cause lysosomal permeabilization and cell death [1,84]. Lysosomal permeabilization has been implicated previously in necrosis induction by high bacillary loads of Mtb [61]. It is possible that MtbΔRv3167c may exploit a similar mechanism to kill host cells as increased lysosomal permeabilization was observed in cells infected with the mutant bacteria (Fig 2D). The Mm-mediated induction of necrosis after zebrafish infection also requires an increase in mROS [49]. Nevertheless, in contrast to our data (Figs 2A, S3B and S4A), Mm signals through the TNF/RIPK3 pathway to induce an increase of mROS and necrosis. The differences may reflect the variations in molecular pathogenesis pathways engaged by the human pathogen Mtb and the fish pathogen Mm. Our data indicates involvement of diminished Akt activation in mitochondrial ROS generation in MtbΔRv3167c–infected cells (Fig 6E). The augmented phagosomal escape observed in MtbΔRv3167c-infected cells may allow for previously sequestered Mtb proteins to target mitochondria and trigger an increase in mROS generation. The mycobacterial type VII secretion system, ESX-5, is involved in the secretion of proteins containing Pro-Pro-Glu (PPE), Pro-Glu (PE) and polymorphic GC-rich sequences (PGRS) and has been implicated in cell lysis after Mtb escapes from the phagosome [39,85]. Interestingly, the ESX-5 substrates PE25 and PPE41 form a complex and induce necrosis [39,40]. Furthermore, the Mtb PE_PGRS33 protein, when ectopically expressed in a eukaryotic cell, localizes to the mitochondria and induces apoptotic and necrotic cell death [86]. The importance of these proteins in the context of infection with live bacteria has not been demonstrated yet. Another compelling target could be the secreted Mtb toxin CpnT which induces RIPK1-independent necrosis in Mtb-infected macrophages [87] via its NAD+ glycohydrolase activity [88]. It is possible that both mitochondrial localization of mycobacterial proteins and inhibition of Akt activation via Mtb proteins may both contribute to elevated mitochondrial ROS levels seen observed in MtbΔRv3167c-infected cells. Autophagic clearance is a defense mechanism employed by host cells following detection of cytosolic pathogens. While macroautophagy (hereafter referred to as autophagy) is defined as the engulfment of cytosol by the autophagosome, selective autophagy describes the process in which autophagosome formation is directed towards a specific organelle, protein complex or microorganism by cargo receptor proteins (p62, NDP52, NBR1, Optineurin) [89–91]. Selective autophagy augments killing of intracellular mycobacteria, since reduced bacterial viability was seen following autophagy induction with IFNγ treatment of BCG-infected macrophages [92]. The relevance of selective autophagy for host defense against Mtb was demonstrated by the dramatically increased susceptibility of Atg5-/- mice when compared to wild-type mice [66,93]. It was thus unexpected that MtbΔRv3167c was hypervirulent in the mouse model (Fig 7A) even though increased autophagy induction was observed in MtbΔRv3167c-infected cells compared to Mtb-infected controls (Fig 3C and 3D). We found that while MtbΔRv3167c induces autophagy, there was no increase in selective autophagy, as very few mycobacteria (both wild-type and mutant) co-localized with autophagosomes (Fig 3A and 3B). Previous studies have shown that Mtb has evolved mechanisms to avoid recruitment into the autophagosome [66,72,94–96]. Our results support this observation and show that this immune evasion strategy remains intact in MtbΔRv3167c. The increased autophagy seen in MtbΔRv3167c-infected cells is most likely a host stress response to an increased number of cytosolic bacteria [66]. Autophagy may dampen inflammation by negative regulation of the inflammasome and via the degradation of danger associated molecular patterns during host cell necrosis [97,98]. Atg5-/- mice have higher basal level of inflammation when compared to wild-type mice [93] and Mtb-infected Atg5-/- mice had increased levels of pulmonary pro-inflammatory cytokines and exhibited increased lung tissue damage [66,99]. It is thus possible that in the absence of autophagy induction, the increased inflammatory response seen in MtbΔRv3167c-infected mice (Fig 7E and 7F) would have been even stronger. Unlike apoptosis, which benefits the host by reducing mycobacterial viability, Mtb-induced necrosis is beneficial to the pathogen allowing it to exit from infected cells and to disseminate [14,100]. Consistent with this concept was our finding that the necrosis-inducing MtbΔRv3167c strain was hypervirulent in mice and guinea pig. The hypervirulence of various clinical Mtb strains and Mtb deletion mutants has been reported previously [101]. For example, the Beijing strain HN878 was found to be more virulent than another member of the same family in immunocompetent mice [102]. Deletion of the mce1 operon, two component response regulators KdpDE, tcrXY and the serine threonine protein kinases pknH, pknE and pknI rendered Mtb hypervirulent in mouse studies [103–107]. The presence of multiple anti-virulence genes in Mtb gives rise to the question: why would Mtb encode genes that suppress its virulence? While virulence may be defined as the ability of a pathogen to cause disease, an important aspect of virulence is successful transmission between hosts [108]. As Mtb has probably co-evolved with humans for more than 50,000 years, moderation of its virulence would have prevented elimination of the early existent small host populations thus maximizing transmission opportunities and improving persistence of the pathogen [109,110]. THP1 monocytes were obtained from ATCC (TIB 202). GFP tagged LC3 expressing THP1 monocytes (THP1 LC3GFP) were provided by Dr. John Kehrl (NIH). THP1shASC and THP1shcontrol cells were obtained from Dr. Jenny Ting (University of North Carolina). C57Bl6, Nox2-/-, Casp3-/- and mCAT transgenic mice were obtained from Jackson Laboratories. Ripk3-/- mice were obtained from Genentech. Casp1/11-/- mice were provided by Dr. Denise Monack (Stanford School of Medicine). Parp1-/- mice were obtained from Dr. Ted Dawson (Johns Hopkins University). Tnfr1-/-, Il1r1-/-, Irf3-/- and Ifnβ-/- mice were provided by Dr. Alan Sher (NIH). Immortalized wildtype, Nlrp3-/ and Trif-/-MyD88-/- BMDMs were provided by Dr. Igor Brodsky (University of Pennsylvania). Atg5fl/fl LysM Cre+ (Atg5-/-) and Atg5fl/fl LysM Cre- (Atg5+/+) mice were obtained from Dr. Herbert Virgin IV (Washington University School of Medicine). zVAD FMK, Necrostatin 1, MAPK inhibitors (SP600125, SB203580), Akt activator (sc-79) and DPI were purchased from Calbiochem. BafilomycinA1, glutathione and N-acetyl cysteine were sourced from Sigma. 3-MA was purchased from Tocris Biosciences. Rv3167c was deleted in M. tuberculosis H37Rv using a specialized phage transduction strategy described previously [111]. Gene deletion was confirmed by RT-PCR as well as by southern blotting. The probes used were labeled with biotin using BrightStar Psoralen-Biotin Kit. Genomic DNA was digested with EcoRI. The DNA fragments were separated by agarose gel electrophoresis, transferred to charged nylon membrane, and denatured with 0. 4N NaOH. The probe was denatured at 90°C for 10 min in the presence of 10mM EDTA and hybridized to the membrane at 55°C for 16h in hybridization buffer (AlkPhos Direct hybridization buffer with 0. 5M NaCl). The membrane was washed and the probe was detected using a BrightStar BioDetect Nonisotopic Detection Kit. For generating the complement strain, Rv3167c gene sequence including 60bp upstream was cloned into the episomal plasmid pMV261, electroporated into the Rv3167c mutant strain and plated on 7H10 plates with 40μg/ml kanamycin. Bacterial strains were grown in 7H9 medium supplemented with 10% ADC, 0. 5% glycerol and 0. 05% Tween 80. Hygromycin (50μg/ml) and kanamycin (40μg/ml) were added to the mutant and complement cultures respectively. For infection, cultures with an OD600 between 0. 6–0. 8 (corresponding to the late log phase of growth) were pelleted and resuspended in 0. 05% PBS-Tween 80 prior to addition to cells. To measure in vitro bacterial growth, bacteria were added to 7H9 medium to obtain a starting OD600 of 0. 01. OD600 measurements were made at 24h intervals until 7 days. Ex vivo bacterial growth was determined by infecting THP1 macrophages and lysing them at the indicated timepoints with 0. 1% Triton X 100. Appropriate dilutions were plated on 7H11 medium in triplicate. Inoculated plates were incubated at 37°C and colonies were counted approximately 2 weeks after plating. THP1 monocytes were maintained in RPMI 1640 supplemented with 10% heat inactivated FCS. Cells were differentiated with 20ng/ml PMA for 20–24 hours, washed and infected in growth medium containing 5% human serum. Bacteria were added to cells at MOI 3 for 4 hours at 37°C, extracellular bacteria were removed by PBS washes and chase medium containing 100μg/ml gentamicin was added. BMDMs were prepared from cells obtained from femurs and tibia of various mouse strains and cultured in DMEM supplemented with 10% heat inactivated FCS, 25% L929 supernatant and 1% Penicillin-Streptomycin. Growth medium was replaced with DMEM containing 10% non-heat inactivated FCS for 4h and cells were infected at MOI 10 in same medium in the manner described above. Chase media contained 10% L929 supernatant in order to avoid cell death induction due to cytokine withdrawal. Immortalized BMDMs were maintained in DMEM containing 10% heat inactivated FCS and infected in media similar to that used for primary BMDMs. Human monocyte derived macrophages (hMDMs) were prepared from elutriated monocyte fractions obtained from NIH blood bank. Monocyte fractions were seeded in serum free RPMI for one hour. Non-adherent cells were removed and adherent cells were differentiated in RPMI medium containing 5% off-the-clot AB human serum (Gemini) and 10ng/ml human MCSF (Peprotech) for 7 days. Inhibitors (with the exception of 3-MA) were added to cells one hour prior to infection and included in chase medium. 3-MA was added only to chase medium. For all experiments, 0h time point refers to end of infection period when cells have been exposed to bacteria for 4 hours. Cells were stained with 1μg/ml propidium iodide (PI) (Sigma-Aldrich) for 10 minutes at room temperature and analyzed by flowcytometry. For TUNEL stain, cells were fixed in 4% paraformaldehyde overnight, stained as per manufacturer’s instructions (Roche) and examined by either flow cytometry or fluorescence microscopy. Hypodiploid stain was performed using PI/RNase staining buffer (BD Pharmingen) following overnight fixation in 70% ethanol as per manufacturer’s instructions. For all flow cytometry analyses, at least 10,000 cells were acquired (BD Accuri C6). Toxilight assay to measure adenylate kinase release from cells was performed as per manufacturers instructions. Autophagy induction in THP1 LC3GFP expressing cells was analyzed as described previously [67]. Briefly, cells were permeabilized with 0. 05% saponin for 5 minutes, washed and resuspended in PBS containing 5% FCS. Permeabilization resulted in loss of cytosolic LC3I while LC3II bound to autophagosome membranes were retained, which was measured by flow cytometry (50,000 cells acquired, BD Accuri C6). For immunofluorescence analysis, bacteria were stained with 0. 4mg/ml AF647-NHS ester (Molecular Probes) in 0. 1M sodium bicarbonate solution for 30 minutes at 37°C and used for infecting cells on slides. At specified time points, cells were fixed with 4% paraformaldehyde overnight, stained with Hoechst 33342 and analyzed by confocal microscopy (Zeiss LSM710). Cell lysates were obtained by lysing cells with RIPA buffer containing protease (Complete, Mini EDTA free, Roche) and phosphatase inhibitor cocktails (PhosStop, Roche) followed by centrifugation at 12,000g for 5 minutes. Pierce BCA protein assay kit (Thermo Scientific) was used to measure protein concentrations to ensure equivalent loading. Antibodies against phosphorylated and total Akt and MAPKs, PARP, tubulin and GFP were purchased from Cell Signaling and used at 1: 1000 dilution. Anti LC3 antibody was purchased from Epitomics and used at 1: 2500 dilution. Densitometric analysis was performed using ImageJ software. To detect mycobacterial escape from the phagosome, the CCF4 FRET assay was performed as described previously [34]. Briefly, cells were stained with 8μM of CCF4 (Invitrogen) in EM buffer (120mM NaCl, 7mM KCl, 1. 8mM CaCl2,0. 8mM MgCl2 5mM glucose, 25mM Hepes, pH7. 3) containing 2. 5μM of probenecid (Sigma-Aldrich) for 1. 5 hours at room temperature. Live populations were distinguished from dead ones by addition of Live/Dead Fixable Red stain (Invitrogen) for 30 minutes at room temperature. After staining cells were fixed with 4% PFA overnight and analyzed by flow cytometry (BD FACS CantoII). 40,000 cells were acquired and post acquisition analysis done using FlowJo software (Treestar, OR). For estimation of bacterial β-lactamase activity, bacteria were resuspended in PBS containing 50μg/ml porcine esterase liver extract and 100nM CCF4-AM and incubated at 37°C for 12h. Fluoresence measurements were made using Biotek Synergy 4 microplate reader. For measurement of ROS levels in BMDMs, cells were infected as described previously [78]. At indicated time points after infection, cells were harvested and stained with 10μM CM-H2DCFDA (Molecular Probes) or 1. 25μM MitoSOX Red (Molecular Probes) for 30 minutes at 37°C in HBSS. Cells were analyzed by flow cytometry (at least 10,000 cells acquired, BD Accuri C6) after HBSS wash. Cells were stained with 40 nmol of DIOC6 stain (Molecular Probes) at 37°C for 15 minutes, washed and analyzed by flow cytometry (10,000 cells acquired, BD Accuri C6). THP1 cells were fixed in 2% gluteraldehyde and 2% paraformaldehyde in 0. 1M sodium cacodylate buffer pH7. 4 for 1 hour. They were carefully pelleted and re-suspended in 2% paraformaldehyde for several hours, followed by rinsing in 0. 1M sodium cacodylate buffer and pelleted in 2% agar in the same buffer. The samples were post fixed in 1% osmium tetroxide in 0. 1M sodium cacodylate buffer, rinsed in distilled water and, en bloc stained in 2% aqueous uranyl acetate for a further hour. They were then rinsed and dehydrated in an ethanol series (50% to 100%) followed by resin infiltration Embed 812 (Electron Microscopy Sciences) and baked overnight at 60°C. Hardened blocks were cut using a Leica UltraCut UC7. 60nm sections were collected on formvar/carbon coated nickel grids and contrast stained using 2% uranyl acetate and lead citrate. Grids were all viewed in a FEI Tencai Biotwin TEM at 80Kv. Images were taken using Morada CCD and iTEM (Olympus) software. Embedding and sectioning was performed at the electron microscopy core facility at the Yale School of Medicine. C57Bl/6 mice were infected with 100 CFU of each of the various bacterial strains grown to late log phase via the aerosol route using a Glas-Col full body inhalation exposure system. At the indicated time points, 3 mice per group were sacrificed and bacterial load was determined by homogenizing the organs in PBS and plating serial dilutions on 7H11 plates. Lung homogenate supernatants were used for cytokine analysis using the Luminex MAGPIX platform (R&D Bioscience). Superior lobes of the lungs were fixed in 10% buffered formalin for histopathology. Paraffin embedding, sectioning and hematoxylin and eosin (H and E) staining were performed by AML Labs, Baltimore. Total lung area and areas of inflamed regions in H and E stained lung sections were quantified using ImageJ. Female outbred Hartley guinea pigs (250-300g) were purchased from Charles River Labs (Wilmington, MA). Animals were infected with each of the three Mtb strains via aerosol using a Madison chamber aerosol generation device (University of Wisconsin, Madison, WI) calibrated to deliver ~3 log10 CFU in the lungs. Four animals from each group were sacrificed on day 1 and day 28 post-infection. The lungs were homogenized, as previously described, and the lung homogenates were plated on 7H11 Middlebrook agar and incubated at 37°C for 4 weeks before final CFU counts were determined [112]. Alveolar macrophages were harvested by bronchoalveolar lavage (BAL) as described previously [113]. Briefly, cold PBS with 3% FCS was instilled into the lungs following insertion into the trachea of an 18-gauge cannula fixed to a 20-ml syringe. The cells were pelleted by centrifugation at 380g for 10 min, washed twice with RPMI-1640 supplemented with 10% FCS, and 10μM 2-mercaptoethanol (RPMI complete medium), and resuspended in 1 ml RPMI complete medium. Following transport from Johns Hopkins to University of Maryland on ice, viable cells were enumerated by the trypan blue exclusion method and seeded in RPMI complete medium overnight. Adherent cells were infected in RPMI medium containing 5% FCS at specified MOI for 4 hours at 37°C, extracellular bacteria were removed by PBS washes and chase medium containing 100μg/ml gentamicin was added. Statistical analysis was performed using GraphPad Prism version 6. 0 software. Data is presented as mean ± S. E. M. of three independent experiments and one-way ANOVA with Tukey post-test was used unless mentioned otherwise in the figure legends. p-value significance is as follows—*- ≤0. 05, ** - ≤0. 01, *** - ≤0. 001, **** - 0. 0001. All animals were handled in accordance with the NIH guidelines for housing and care of laboratory animals and the studies were approved by the Institutional Animal Care and Use Committees at the University of Maryland, College Park (protocol no—R-12-55) and Johns Hopkins University School of Medicine (protocol no—GP12M88).

Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis, is a highly successful human pathogen. Following entry into host phagocytic cells, Mtb resides within a modified phagosomal compartment and inhibits apoptotic host cell death. Recent studies have demonstrated that Mtb eventually translocates from the phagosomal compartment to the cytosol. This event is followed by the induction of necrotic host cell death allowing the bacteria to exit the host cell and infect naive cell populations. Our study adds to this relatively unexplored aspect of Mtb pathogenesis by revealing that the transcriptional repressor Rv3167c of Mtb negatively regulates phagosomal escape and host cell necrosis. We furthermore demonstrate that the increased necrosis induction by the Mtb mutant strain deficient in Rv3167c required elevated reactive oxygen species levels within host cell mitochondria and reduced activation of the protein kinase Akt. In addition, the increased virulence of the Mtb mutant strain observed after aerosol infection of mice strengthens the link between the ability of the bacteria to induce host cell necrosis and virulence. The Mtb genes negatively regulated by Rv3167c are thus potential virulence factors that can be targeted for drug and vaccine development.

Abstract Introduction Results Discussion Materials and Methods

flow cytometry cell death medicine and health sciences autophagic cell death pathology and laboratory medicine viral transmission and infection cell processes microbiology signs and symptoms bacteria research and analysis methods specimen preparation and treatment staining spectrum analysis techniques necrotic cell death actinobacteria spectrophotometry necrosis cytophotometry cell staining diagnostic medicine host cells cell biology mycobacterium tuberculosis virology apoptosis biology and life sciences organisms

2016

Identification of a Transcription Factor That Regulates Host Cell Exit and Virulence of Mycobacterium tuberculosis

The agr quorum-sensing system of Staphylococcus aureus modulates the expression of virulence factors in response to autoinducing peptides (AIPs). Recent studies have suggested a role for the agr system in S. aureus biofilm development, as agr mutants exhibit a high propensity to form biofilms, and cells dispersing from a biofilm have been observed displaying an active agr system. Here, we report that repression of agr is necessary to form a biofilm and that reactivation of agr in established biofilms through AIP addition or glucose depletion triggers detachment. Inhibitory AIP molecules did not induce detachment and an agr mutant was non-responsive, indicating a dependence on a functional, active agr system for dispersal. Biofilm detachment occurred in multiple S. aureus strains possessing divergent agr systems, suggesting it is a general S. aureus phenomenon. Importantly, detachment also restored sensitivity of the dispersed cells to the antibiotic rifampicin. Proteinase K inhibited biofilm formation and dispersed established biofilms, suggesting agr-mediated detachment occurred in an ica-independent manner. Consistent with a protease-mediated mechanism, increased levels of serine proteases were detected in detaching biofilm effluents, and the serine protease inhibitor PMSF reduced the degree of agr-mediated detachment. Through genetic analysis, a double mutant in the agr-regulated Aur metalloprotease and the SplABCDEF serine proteases displayed minimal extracellular protease activity, improved biofilm formation, and a strongly attenuated detachment phenotype. These findings indicate that induction of the agr system in established S. aureus biofilms detaches cells and demonstrate that the dispersal mechanism requires extracellular protease activity. Most bacteria have an inherent ability to form surface-attached communities of cells called biofilms [1]. The opportunistic pathogen Staphylococcus aureus can form biofilms on many host tissues and implanted medical devices often causing chronic infections [2]–[5]. The challenge presented by biofilm infections is the remarkable resistance to both host immune responses and available chemotherapies [6], [7], and estimates suggest that as many as 80% of chronic bacterial infections are biofilm associated [8]. In response to certain environmental cues, bacteria living in biofilms are capable of using active mechanisms to leave biofilms and return to the planktonic (free-living) state in which sensitivity to antimicrobials is regained [9]–[11]. Therefore an improved understanding of the molecular mechanism of biofilm detachment could facilitate the discovery of innovative treatment options. Studies on the opportunistic pathogen Pseudomonas aeruginosa indicate that cell-to-cell communication (often termed “quorum-sensing”) is required to make a robust biofilm under some growth conditions [12]. Surprisingly, the opposite is true in S. aureus, as the presence of an active quorum-sensing impedes attachment and development of a biofilm [13], [14]. The S. aureus quorum-sensing system is encoded by the accessory gene regulator (agr) locus and the communication molecule that it produces and senses is called an autoinducing peptide (AIP), which is an eight-residue peptide with the last five residues constrained in a cyclic thiolactone ring [15]. During growth, AIP is synthesized and secreted through a poorly understood mechanism that requires multiple peptidases [16], [17]. Once AIP reaches a critical concentration, it binds to a surface histidine kinase receptor, initiating a regulatory cascade that controls expression of a myriad of virulence factors, such as proteases, hemolysins, and toxins [18]. A recent study by Yarwood et. al. [19] raised the possibility that the agr quorum-sensing system is involved in biofilm detachment. This study demonstrated that bacteria dispersing from biofilms displayed high levels of agr activity, while cells in a biofilm had predominantly repressed agr systems. These findings correlate well with prior data indicating that agr deficient S. aureus strains form more robust biofilms compared to wild type strains [13], [14]. In the study presented here, we demonstrate that activation of the agr system in established biofilms is necessary for detachment. This activation could be accomplished with exogenous AIP addition or by changing nutrient availability to the biofilm. We also demonstrate that agr-mediated detachment requires the activity of extracellular proteases. Our findings suggest that agr quorum-sensing is an important regulatory switch between planktonic and biofilm lifestyles and may contribute to S. aureus dispersal and colonization of new sites. Mutations in the agr quorum-sensing system are known to improve biofilm development [13], [14]. Based on these studies, it seemed probable that there is a correlation between agr activity and biofilm formation. Regassa et al. reported that growth on rich media containing glucose represses the agr system through the nonmaintained generation of low pH [20]. Interestingly, in most published flow cell biofilm studies, one commonality is the use of growth media containing or supplemented with glucose [9], [19], [21]–[24]. In our own efforts to grow S. aureus flow cell biofilms, we found a strict dependence on glucose supplementation. For the experimental setup, a once-through, continuous culture system was employed as previously described [19], [25], and S. aureus SH1000 constitutively expressing red fluorescent protein (PsarA-RFP, plasmid pAH9) was used as the testing strain. Using 2% TSB as the growth media, SH1000 cells did not attach and develop a biofilm (Figure 1A), instead passing right through the flow cell to the effluent. However, in the presence of 0. 2% glucose (TSBg), cells attached and a formed a robust biofilm (10–20 microns thick) after two days of growth, which was visually evident and monitored with confocal laser scanning microscopy (CLSM, Figure 1B). As expected, glucose strongly inhibited expression from the P3 promoter using a GFP reporter (Figure 1E), suggesting that repression of RNAIII is essential for attachment and biofilm formation. In broth culture and biofilm effluents, we observed a glucose-dependent pH decrease to the 5. 5 range similar as previously reported [20], [26]. As a control, flow cell biofilms were prepared with an agr mutant strain (SH1001, Δagr: : TetM) containing plasmid pAH9 (Figure 1C & D), and this strain developed a biofilm even in the absence of media supplementations (Figure 1C). As anticipated, the P3 promoter did not activate in the agr mutant (Figure 1E). Overall, these observations indicate that environmental conditions favoring low agr activity are essential for attachment and biofilm formation. To investigate the role of the agr system in established biofilms, we developed strategies to modulate level of agr activity within a biofilm. Initially, media supplementation experiments were performed using purified AIP signal in order to place the agr system under external control. We recently developed a new method for AIP biosynthesis [27], enabling the production of sufficient signal levels for flow cell experiments. Through exogenous AIP addition, we could test wild-type strains and avoid any potential complications of constructed agr deletion mutants. For this approach, established flow cell biofilms were prepared using S. aureus SH1000 constitutively expressing RFP with plasmid pAH9. The flow cell media was supplemented with glucose to attenuate agr expression [20], allowing cell attachment and biofilm development. After two days, either 1 mL of buffer (100 mM phosphate [pH 7], 50 mM NaCl, 1 mM TCEP; Figure 2A) or 1 mL of 20 µM AIP-I in buffer (Figure 2B and Video S1) was diluted 1000-fold (50 nM final concentration) into the growth media. Using our synthesized AIP-I in dose-response curves [27], we estimate the amount of AIP-I in supernatants of TSB broth cultures (OD600 1. 0–1. 3) reaches approximately 400 nM (data not shown), indicating the 50 nM level used for the biofilm experiments is within a relevant concentration range. Examination with CLSM showed that the AIP-I treated biofilm sloughed off the flow cell over a period of 1–2 days (Figure 2B and Video S1), suggesting that AIP-I activated a detachment mechanism. To confirm that AIP-I caused detachment, we counted viable S. aureus cells in the effluent media (Figure 2C). The concentration of bacteria in the effluent increased markedly 24–36 hours after AIP-I addition. In contrast, the number of bacteria in the biofilm effluent without AIP-I addition remained relatively constant. Computational analysis of the detachment phenotype indicated that 91. 3±4. 3% of the biomass dispersed within 48 hrs of AIP-I addition. Among S. aureus strains, there are four types of agr quorum-sensing systems. Each of these agr systems, referred to as agr-I through agr-IV, recognizes a unique AIP structure (AIP-I through AIP-IV). Through an intriguing mechanism of chemical communication, these varying quorum-sensing systems can be subdivided into three cross-inhibitory groups: agr-I/IV, agr-II, and agr-III. The activating signals of each group cross-inhibits the alternative signal receptors with surprising potency, a phenomenon termed “bacterial interference” [15]. Since AIP-I and AIP-IV differ by only one amino acid and function interchangeably [28], they are grouped together in the classification scheme, although this assignment has been controversial [29], [30]. To determine the generality of the detachment mechanism, we examined the effect of AIP addition using S. aureus strains representing different agr groups. The strains tested were (i) FRI1169, agr-I, toxic shock syndrome isolate [31]; (ii) SA502a (ATCC27217), nasal isolate and prototype agr-II strain [15], [32]; and (iii) ATCC25923, clinical agr-III isolate [9]. When the correct AIP signal was added to 2-day old biofilms of each strain (FRI1169, AIP-1; SA502a, AIP-II; ATCC25923, AIP-III), signal addition resulted in robust detachment of each biofilm over a period of 48 hours (Figure 3). These findings indicate biofilm detachment is a general S. aureus phenomenon that occurs in laboratory strains and clinical isolates, and functions across diverse agr systems. If AIP was promoting biofilm detachment via the agr system, we predicted that agr expression would be induced prior to detachment and an agr deficient mutant would not detach in response to AIP. To determine whether the agr system is activated prior to biofilm detachment, a dual fluorescent-labeled SH1000 strain was constructed with a constitutive RFP (PsarA-RFP, pAH9) and an agr responsive GFP reporter (PagrP3-GFP, pDB59). After two days of biofilm growth, we added AIP-I to the biofilm flow medium and this resulted in strong induction of the GFP reporter (Figure 4A), indicating activation of the agr system. As shown, the GFP reporter was clearly activated before dispersal of the biofilm cells. By the fourth day, all cells with detectable GFP expression detached from the biofilm. These observations provide convincing evidence that AIP activates the agr system prior to biofilm dispersal. To further investigate the role of the agr system, we utilized a mutant strain with a complete deletion of the agr locus (SH1001). Unlike the wild type strain (Figure 4A), the agr mutant biofilm harboring the same dual reporters did not respond to AIP-I treatment, as evidenced by a lack of GFP induction, and the mutant biofilm did not disperse (Figure 4B). Similarly, addition of an inhibitory AIP (50 nM AIP-II) to the dual-labeled SH1000 biofilm failed to induce GFP expression, and again, the biofilm did not disperse (Figure 4C). Taken together, these data demonstrate that an active agr quorum-sensing system is necessary for AIP-mediated biofilm dispersal. We have demonstrated that low agr activity is important for biofilm formation and that activation of the agr system in established biofilms induces detachment. Considering changes to the physiochemical environment may occur in vivo, we investigated whether an alteration in nutrient availability could reproduce the detachment phenotype. Again, two day flow cell biofilms were prepared with the dual-labeled strain (AH596) in TSBg (Figure 5A). The glucose was removed and significant activation of the P3 promoter was apparent by monitoring GFP levels using CLSM (Figure 5A), supporting our previous result (Figure 1A). Once the agr system was activated, robust detachment from the flow cell was observed and monitored with CLSM (Figure 5A). An agr deletion mutant did not respond to glucose depletion (Figure 5B), indicating the detachment phenotype was dependent upon a functional agr system. These findings demonstrated that glucose depletion can disperse an S. aureus biofilm and again the detachment occurred through an agr-dependent mechanism. These experimental observations mirrored those with AIP addition and further support the apparent inverse correlation between agr activity and biofilm formation. Biofilm growth of S. aureus increases resistance to antimicrobials when compared to the planktonic growth mode [9], [19]. This biofilm mediated resistance hinders treatment of many chronic S. aureus biofilm related infections, including endocarditis, osteomyelitis, and indwelling medical device infections [3], [33]. Therefore, we asked whether AIP-dispersed bacteria regained sensitivity to a clinically relevant antibiotic, rifampicin. To test this, we collected detached cells from an AIP-treated biofilm effluent and compared resistance to intact biofilms exposed to different levels of rifampicin. Similar to previous antibiotic susceptibility results [19], even at the highest concentration tested (100 µg/ml), the level of rifampicin killing was <2 log units of the biofilm biomass (Figure 6). In contrast, the viability of detached cells displayed a different antibiotic response. At 10 µg/ml rifampicin, a 6 log decrease of viable cells was detected, and at 100 µg/ml, complete killing of the detached cells was observed (Figure 6). The AIP-detached cells were more resistant than broth culture to comparable levels of rifampicin, suggesting parts of the detached biofilm may remain in emboli that are known to possess elevated antibiotic resistance [9]. These observations demonstrated that S. aureus cells detached from a biofilm regain susceptibility to a clinical antibiotic. S. aureus possesses the icaRADBC locus that is required to synthesize and generate an exopolysaccharide, which is referred to as the polysaccharide intracellular adhesin or PIA (also called PNAG). S. aureus is known to form biofilms through both ica-dependent and ica-independent mechanisms [34], [35]. To gain insight on the biofilm detachment mechanism, we sought to distinguish whether our S. aureus biofilms were dependent on PIA. In strain SH1000, we constructed an Δica: : Tet deletion mutant (strain AH595) using generalized transduction and confirmed the mutation with PCR and sequencing. In microtiter biofilm assays, we were unable to identify a biofilm phenotype (Figure 7A and 7B). Similarly in flow cell biofilms, we did not observe a defect in the ability of strain AH595 to form a biofilm (Figure 7C). No difference was observed compared to flow cell biofilms of SH1000 grown in parallel (data not shown). While SH1000 is a derivative of 8325-4, and there are reports that the ica locus is required for 8325-4 derived strains to make a biofilm [36], the ica locus was not required for biofilm formation under our experimental conditions. Similar to our observations, an ica mutant of the clinical S. aureus isolate UAMS-1 displays no defect in microtiter and flow cell biofilm assays [22]. In contrast, when proteinase K was added to SH1000, biofilms were unable to develop in the microtiter plate format (data not shown), indicating the biofilms are forming through an ica-independent mechanism. These findings suggest that PIA is unlikely to have a role in biofilm detachment in the SH1000 strain background. Knowing the agr system is essential for biofilm detachment, what agr regulated products are responsible for the dispersal phenotype? In S. aureus strains that produce ica-independent biofilms, proteinase K eliminates adherence and biofilm formation [35], [37]–[39], perhaps through cleavage of surface structures. S. aureus is coated with cell wall attached proteins that mediate adherence to a variety of substrates [40], and some of these adhesins, such as biofilm associated protein (BAP) and SasG are important for biofilm formation [41], [42]. It is also known that some surface adhesins, such as protein A and fibronectin-binding protein, are cleaved by the native S. aureus secreted proteases [43], [44]. Considering the agr system regulates the secreted proteases [45], [46], we hypothesized that increased expression of extracellular proteases could be responsible for biofilm detachment. If S. aureus proteases have a role in detachment, proteinase K should be able to disperse an established biofilm. To test this proposal, proteinase K (2 µg/mL) was added to a SH1000 biofilm and resulted in rapid detachment over 12 hrs (Figure 8A). With this preliminary observation, we measured the levels of protease activity in effluents from biofilms with and without AIP-I addition using Azocoll (azo dye impregnated collagen) reagent. As shown in Figure 8B, we detected a baseline level of protease activity in biofilm effluents without AIP-I addition and referenced other measurements to this baseline. With the addition of activating AIP-I, the protease activity increased approximately five-fold compared to a biofilm with no AIP-I treatment. As anticipated, addition of inhibitory AIP-II reduced the level of proteolytic activity in the effluent. Similarly, an agr mutant biofilm supplemented with activating AIP-I displayed very low levels of extracellular proteases (Figure 8B). There are 10 known extracellular proteases produced by most S. aureus strains and expression of all these enzymes is controlled by the agr system [18], [45], [46]. These 10 proteases include the metalloprotease aureolysin (aur), two cysteine proteases (scpA and sspB), and seven serine proteases (sspA (V8) and splABCDEF) [47]. To elucidate what class (es) of proteases are prevalent in AIP-treated biofilms, the effluent from a detaching biofilm was assayed for protease activity in the presence of protease inhibitors or activating agents. The addition of EGTA, an inhibitor of the metalloprotease aureolysin [16], had a negligible effect on overall protease activity (Figure 8C). The addition of PMSF, a potent serine protease inhibitor, however reduced overall protease activity to almost undetectable levels. Lastly, the addition of DTT, a reducing agent used to activate thiol proteases [48], did not significantly change protease activity in the effluents. These results suggest that serine proteases are the dominant, detectable secreted protease in AIP-treated biofilms. With our observation that serine proteases are abundant in detaching biofilms, we examined the effect of a serine protease inhibitor on AIP-mediated detachment. The addition of 10 µM PMSF in combination with AIP-I to an S. aureus biofilm significantly reduced the level of detachment compared with AIP-I alone (Figure 9A vs. B). However, 48. 8% (±5. 2) of the biomass still detached indicating that serine proteases are necessary but not sufficient for complete detachment. To further examine the mechanism, knock-out mutations were constructed in the genes encoding the V8 (SspA) and SplABCDEF serine proteases. Surprisingly, sspA: : Tet and Δspl: : Erm single mutants, and an sspA: : Tet Δspl: : Erm double mutant, all increased extracellular protease levels (Figure 10A) and eliminated biofilm formation under microtiter plate conditions (Figure 10B & 10C). To block other extracellular proteases, a mutation was constructed in the gene encoding aureolysin (Aur). Aur is a metalloprotease that is required to initiate a zymogen activation cascade [49], [50], starting with the V8 protease [51], which in turn activates the SspB cysteine protease [52]. The activation mechanism of the ScpA cysteine protease remains unresolved [49]. In contrast to the serine protease mutations, introduction of the Δaur deletion into S. aureus reduced extracellular protease levels (Figure 10A) and did not affect biofilm formation (Figure 10B). Interestingly, under conditions of high agr activity, the Δaur deletion displayed improved biofilm formation versus wild-type (Figure 10C). In biofilm detachment tests, the Δaur mutant reduced AIP-mediated detachment, but 54. 6% (±8. 1) of the biomass still detached (Figure 9C). Considering the Spl proteases are not zymogens [53], we examined the combined effects of the Aur cascade and the Spl proteases by constructing an Δaur Δspl: : Erm double mutant. The Δaur Δspl strain possessed very low levels of extracellular protease activity (Figure 10A) and had a minor enhancement in biofilm formation (Figure 10B). Similar to the Δaur mutant, the Δaur Δspl double mutant also displayed improved biofilm formation versus wild-type under conditions of high agr activity (Figure 10C). After AIP-I addition, only 21. 7% (±6. 6) of the Δaur Δspl mutant biomass detached in comparison to 91. 3 (±4. 3) of the wild-type strain (Figure 9D). These experiments indicate that the extracellular proteases have anti-biofilm properties and they demonstrate that agr-mediated biofilm detachment requires the activity of these proteases. The majority of studies on biofilm detachment have focused on factors capable of initiating the process, such as nutrient availability [54], [55], nitric oxide exposure [56], oxygen tension [57], iron salts [58], chelators [59], and signaling molecules [60]–[63]. Alternatively, detachment studies have addressed effector gene products that contribute to the dissolution of the biofilm, including surfactants [10], [13], [64], [65], hydrolases [66], [67], proteases [37]–[39], and DNase [68]. Here were do both, by demonstrating that the increasing AIP levels or lowering available glucose can function as a S. aureus biofilm detachment signal by activating the agr quorum-sensing system, resulting in increased levels of extracellular proteases needed for the detachment mechanism. Importantly, agr-mediated detachment also restores antibiotic sensitivity to the released bacteria, suggesting the mechanism could be a target for treating biofilm infections. These results are in accord with previous studies showing that agr mutants have a propensity to form biofilms [13], [14] and that cells actively expressing agr leave biofilms at a high frequency [19]. Our findings also explain why S. aureus biofilm formation requires glucose supplementation to growth media. Unless the agr system is repressed or inactivated, or the enzymes mediating detachment are inhibited, S. aureus will remain in a planktonic state. The presence of glucose is known to represses RNAIII through a nonmaintained pH decrease to ∼5. 5 [20], resulting from the secretion of acidic metabolites. The RNAIII repression is not due to glucose itself, but results from the mild acid conditions [26] and can be mimicked with other carbon sources, such as galactose [20], that also lower the media pH. In microtiter biofilm experiments, we found these alternative pH-lowering carbon sources could substitute for glucose in facilitating biofilm formation (data not shown). The molecular mechanism through which low pH inhibits RNAIII expression remains to be determined. In the host, many niches colonized by S. aureus are maintained in lower pH ranges, such as the skin and vaginal tract [26], colonization sites that repress agr function could promote biofilm formation. Based on our findings, we propose that the S. aureus agr quorum-sensing system controls the switch between planktonic and biofilm lifestyles. When the agr system is repressed, cells have a propensity to attach to surfaces and form biofilms as detachment factors are produced at low levels. In our detachment model, dispersal of cells from an established biofilm requires reactivation of the agr system and occurs through a protease-mediated, ica-independent mechanism. Yarwood et al. demonstrated through time-course, flow cell studies that reactivation of agr does occur in a biofilm [19], presumably through autonomous AIP production that reaches local concentrations high enough to activate agr. Under these fixed conditions, the agr system may function primarily as a mechanism to detach clumps (also called emboli) that seed new colonization sites. In the experiments presented herein, we have employed growth conditions that tip the balance of the agr system, allowing an investigation into full agr reactivation within an established biofilm. This delicate balance can be offset with an increase in local AIP concentration or through changing environmental conditions, both situations that induce agr and result in massive dispersion of the cells. Biofilms are dynamic and dispersal is always operating [11], but accelerated detachment has been observed in response to changing environmental conditions, such as oxygen levels [57], [69], nutrient depletion [54], changing nutrient composition [55], or increased concentration of quorum-sensing signals [61]. An S. aureus biofilm growing in vivo is likely to encounter a changing physiochemical environment, which could serve as a cue to induce accelerated detachment through an agr-mediated mechanism. S. aureus has been reported to form biofilms through an ica-dependent mechanism suggesting that PIA could have a role in detachment [34], [36]. We observed no defect in microtiter or flow cell biofilm formation using an ica mutant of SH1000 (Figure 7). Our findings support the growing evidence that PIA is not a major matrix component of S. aureus biofilms, as exogenous addition of dispersin B, an N-acetyl-glucosaminidase capable of degrading PIA, has little effect on established biofilms of SH1000 and other S. aureus strains [70]. In contrast, dispersin B does detach S. epidermidis biofilms indicating a more significant role for PIA in the S. epidermidis matrix structure [70]. Our experiments with proteinase K and the S. aureus proteases indicate that some proteinaceous material is important for SH1000 biofilm integrity, and this result supports a number of recent studies demonstrating that proteases can inhibit biofilm formation or detach established biofilms from many S. aureus strains [35], [37]–[39]. It is not clear whether agr-mediated detachment will function in S. aureus strains that produce an ica-dependent biofilm. In this study, we document a role for the Aur and Spl proteases in biofilm detachment. Global expression analysis has shown that activation of the agr quorum-sensing system results in up-regulation of extracellular proteases (Aur, SplABCDEF, ScpA, SspAB) and down-regulation of many surface proteins [45], [46]. However, the target of these agr controlled proteases is not clear. One potential target is the surface adhesins, and possible candidates include the surface proteins Atl, Bap, and SasG, all of which have reported roles in biofilm formation [41], [42], [71]–[73]. Atl is additionally known to require proteolytic processing for activation, and this processing is PMSF inhibited [74]. Other possibilities include microbial surface components recognizing adhesive matrix molecules (MSCRAMMs), which are important for adherence to the extracellular matrices of mammalian cells [40]. Also, the S. aureus secreted proteases are known to activate lipase (Sal-1 and Sal-2) precursors [75] and process other secreted enzymes, such as staphylococcal nuclease [76], [77]. In addition to proteases, there may be other agr regulated factors that contribute to biofilm detachment. Surfactant-like molecules, such as δ-toxin, are induced by the agr system and may exert dispersal effects on biofilms [13], [78]. There is growing evidence that extracellular DNA (eDNA) is an important S. aureus biofilm matrix component [24], [70], and expression of staphylococcal nuclease is reported to be under control of the agr system [18]. Thus, while agr induced proteases are required for the detachment phenotype, the agr controlled expression of an array of factors (proteases, nuclease, surfactants) may also contribute to the biofilm detachment mechanism. There is increasing interest in understanding how bacteria detach from biofilms and initiate colonization of new surfaces. The regulation of quorum-sensing systems may be one mechanism by which many bacteria control biofilm formation and dispersal. Quorum-sensing has been implicated in dispersal of biofilms formed by Yersinia pseudotuberculosis [79], Rhodobacter sphaeroides [80], Pseudomonas aureofaciens [81], Xanthomonas capmestris [62], and Serratia marceascens [61]. However, homoserine lactone signals play a divergent role in Pseudomonas aeuruginosa [12], Pseudomonas fluorescens [82], and Burkholderia cepacia [83], where the active versions of these quorum-sensing systems are necessary for biofilm formation and robustness under some growth conditions. In both cases, it appears quorum-sensing plays a significant role in biofilm development and determining the environmental stimuli that modulate quorum-sensing activity will provide insight on bacterial colonization, detachment, and dispersal to new sites. The bacterial strains and plasmids used in this study are described in Table 1. S. aureus or Escherichia coli were grown in tryptic soy broth (TSB) or on tryptic soy agar (TSA) with the appropriate antibiotics for plasmid selection or maintenance (erythromycin 10 µg/ml; chloramphenicol 10 ug/ml; tetracycline 5 ug/ml) and incubated at 37°C. Plasmid DNA was prepared from E. coli and transformed by electroporation into S. aureus RN4220 as described [84]. Plasmids were moved from RN4220 into other S. aureus strains by transduction with bacteriophage α80 [85] or by purifying the plasmid DNA and transformed by electroporation into appropriate strains. To move sspA and splABCDEF mutations into appropriate genetic backgrounds, phage transduction with α80 was used as described [85]. To construct the Δaur mutation, the pKOR1-aur plasmid was used as described [16]. Fluorescence measurements with S. aureus strains containing pDB59 were performed as previously described [27]. The sarA P1 promoter region was PCR amplified from SH1000 genomic DNA with oligonucleotides (for 5′-GTTGTTAAGCTTCTGATATTTTTGACTAAACCAAATGC-3′, rev 5′-GTTGGATCCGATGCATCTTGCTCGATACATTTG-3′), digested with HindIII and BamHI, and cloned into the erythromycin shuttle plasmid pCE107 [19]. The mCherry (RFP) gene was PCR amplified from pRSET-mCherry [86] with oligonucleotides incorporating a 5′ ribosome binding site and KpnI site and a 3′ EcoRI site (for 5′-GTTGGTACCTAGGGAGGTTTTAAACATGGTGAGCAAGGGCGAGGAGG-3′, rev 5′-GTTGAATTCTTACTTGTACAGCTCGTCCATGCC-3′). The mCherry fragment was cut with KpnI and EcoRI and cloned downstream of the sarA promoter to generate a constitutive RFP expressing plasmid called pAH9. Milk agar plates for detection of protease activity consisted of 3 g/L Tryptic Soy broth, 20 g/L non-fat dry milk, and 15 g/L agar. To determine relative protease activities of strains, assays were performed as described previously using the Azocoll (Calbiochem) reagent [48]. For measuring protease levels in biofilm effluents, 100 mL of effluent was collected on ice (∼12 hours) after AIP addition to the biofilm medium. Cells were removed from the effluents through centrifugation and filtering, and ammonium sulfate was added to 60% over one hour at 4°C to concentrate proteins. The precipitated proteins were pelleted by centrifugation at 19,000 rpm for 30 min, and the pellet was washed and resuspended in 1 ml with 10 mM Tris pH 7. 5. For the protease assay, the reaction mixture was supplemented with either 1 mM EGTA, 200 µM PMSF, or 1 mM DTT to gauge relative levels of protease classes. Microtiter plate biofilms were performed as described [87] except that the plates were incubated at 37°C with shaking at 200 rpm for 12 hours. For flow cell experiments, AIPs were generated using the DnaB intein method, and the AIP concentrations were determined as previously described [27]. AIPs stocks (∼20 µM) were stored in 100 mM phosphate [pH 7], 50 mM NaCl, 1 mM tris (2-carboxyethyl) phosphine (TCEP) and were diluted into the biofilm flow medium to a final concentration of 50 nM. When required, 5 µg/ml of erythromycin and/or chloramphenicol were added to the flow cell media to maintain plasmids. The growth medium for flow cell biofilms consisted of 2% TSB plus 0. 2% glucose unless otherwise indicated. Flow cell biofilm experiments and confocal microscopy were performed as previously described [19]. Flow cells were inoculated with overnight cultures diluted 1: 100 in sterile water and laminar flow (170 µl/min) was initiated after one hour incubation. Confocal microscopy was performed using a Radiance 2100 system (Biorad) with a Nikon Eclipse E600 microscope. Confocal images were processed using Velocity software (Improvision, Lexington, Mass.). Biofilm biomass was quantified with the COMSTAT program [88] and percent biomass detached was calculated by subtracting biomass present at day 4 from day 2. To quantitate the number of bacteria detaching from a biofilm, 1 ml of flow cell effluent was collected on ice at indicated time points. The collected effluent was vortexed and sonicated in a water bath for 10 minutes to break up clumps, and serial dilutions were plated on TSA plates to determine colony forming units (CFUs). For the Proteinase K detachment experiments, the enzyme (Sigma-Aldrich) was suspended in water and added to the media reservoir at a final concentration of 2 µg/ml. S. aureus biofilms were grown for two days in a flow chamber lined with removable polycarbonate coupons (Flow Cell FC271, Biosurface Technologies, Bozeman MT). Biofilm effluents were collected on ice ∼24 hours after AIP-I addition. In parallel, coupons with biofilm growth were removed from flow cells not exposed to AIP-I. Both detached bacteria and the biofilms were exposed to the indicated levels of rifampicin for six hours. Subsequently, cells were vortexed, and the coupons were sonicated in a water bath to break up the biofilm or cell clumps. Serial dilutions were plated on TSA to determine surviving CFU' s.

A biofilm is a surface-attached community of cells bound together by an extracellular matrix. In a bacterial infection, biofilm-encased cells are protected from antibiotic therapy and host immune response, and these encased cells can develop into a chronic infection. Staphylococcus aureus is a prominent bacterial pathogen known to form biofilms on many medical implants and host tissues. In this report, we demonstrate that repression of the S. aureus quorum-sensing system is required to form a biofilm, and quorum-sensing reactivation in established biofilms disperses the cells. Genetic and molecular analysis demonstrates that quorum-sensing is activated before and required for the detachment mechanism. Detachment is protease-mediated, as established biofilms are sensitive to a non-specific protease and quorum-sensing activation increases the production of extracellular proteases. Using mutations in the protease genes, we show that these secreted enzymes are required for the detachment mechanism. These findings denote that S. aureus quorum-sensing can function as a dispersal mechanism to colonize new sites, and our results suggest this mechanism could be modulated to treat recalcitrant biofilms.

Abstract Introduction Results Discussion Materials and Methods

infectious diseases/bacterial infections microbiology/microbial growth and development microbiology/cellular microbiology and pathogenesis microbiology/medical microbiology infectious diseases/antimicrobials and drug resistance

2008

agr-Mediated Dispersal of Staphylococcus aureus Biofilms

Although malaria has been the leading cause of fever for many years, with improved control regimes malaria transmission, morbidity and mortality have decreased. Recent studies have increasingly demonstrated the importance of non-malaria fevers, which have significantly improved our understanding of etiologies of febrile illnesses. A number of non-malaria febrile illnesses including Rift Valley Fever, dengue fever, Chikungunya virus infection, leptospirosis, tick-borne relapsing fever and Q-fever have been reported in Tanzania. This study aimed at assessing the awareness of communities and practices of health workers on non-malaria febrile illnesses. Twelve focus group discussions with members of communities and 14 in-depth interviews with health workers were conducted in Kilosa district, Tanzania. Transcripts were coded into different groups using MaxQDA software and analyzed through thematic content analysis. The study revealed that the awareness of the study participants on non-malaria febrile illnesses was low and many community members believed that most instances of fever are due to malaria. In addition, the majority had inappropriate beliefs about the possible causes of fever. In most cases, non-malaria febrile illnesses were considered following a negative Malaria Rapid Diagnostic Test (mRDT) result or persistent fevers after completion of anti-malaria dosage. Therefore, in the absence of mRDTs, there is over diagnosis of malaria and under diagnosis of non-malaria illnesses. Shortages of diagnostic facilities for febrile illnesses including mRDTs were repeatedly reported as a major barrier to proper diagnosis and treatment of febrile patients. Our results emphasize the need for creating community awareness on other causes of fever apart from malaria. Based on our study, appropriate treatment of febrile patients will require inputs geared towards strengthening of diagnostic facilities, drugs availability and optimal staffing of health facilities. Febrile illnesses due to different etiological agents are the common causes of morbidity and mortality in developing countries [1]. Malaria has been the leading cause of fever in sub- Saharan Africa for many years [2]. For instance, in Tanzania, malaria was contributing to about 42% of hospital diagnoses and 32% of hospital deaths in the last decade [3]. Accordingly, presumptive treatment of all febrile illnesses in children under five years with anti-malarial drugs was adopted as policy in many countries of sub-Saharan Africa [4]. However, in recent years, there has been gain in malaria control strategies which has led to decreased malaria prevalence particularly in endemic countries [5]–[7]. The decrease in malaria burden has also been indicated by the 2012 World Malaria Report where there is a good achievement in worldwide reduction of malaria transmission, morbidity and mortality [8]. The decline in malaria transmission is mainly a result of increased coverage of different malaria control strategies that have been implemented for several years. This includes the use of long lasting insecticide treated bed nets, indoor residual spraying, intermittent presumptive treatment of malaria during pregnancy or intermittent presumptive treatment of malaria to infants and treatment with effective anti-malaria drugs such as Artemisinin-based Combination Therapies [5], [9]. The decrease in malaria transmission led World Health Organization (WHO) to change its policy in 2010 where anti-malarial treatment is initiated after parasitological confirmation [10]. However, the decline in trend of malaria transmission in many malaria-endemic countries corresponds to an increasing proportion of febrile patients who are diagnosed as not having malaria [11], [12]. Recent studies have increasingly demonstrated the importance of non-malaria fevers, which have significantly improved our understanding of etiologies of febrile illnesses [13], [14]. In this regard, a reasonable proportion of febrile illnesses are now ascribed to be non-malaria febrile illnesses [13] and episodes of such diseases are reported to increase [12]. In Tanzania, diseases such as respiratory tract infections, urinary tract infections, typhoid fever and rotavirus infection are among non-malaria febrile illnesses that have been commonly affecting people particularly children [15]–[18]. A study conducted in Dar es Salaam and Ifakara had shown that among 1005 children, 498 (50%) had acute respiratory infection, while 54 (5. 4%) had urinary tract infections and 33 (3. 3) had typhoid fever [18]. Diseases such as Rift Valley Fever (RVF), dengue fever, Chikungunya virus infection, leptospirosis, tick-borne relapsing fever, Q-fever, rotavirus infection and brucellosis have also been reported in Tanzania [13], [14], [19]–[24]. A recent study conducted in northern Tanzania has reported the occurrence of 55 (7. 9%) cases of Chikungunya virus infection, 40 (33. 9%) cases of leptospirosis, 24 (20. 3%) cases of Q-fever and 16 (13. 6%) cases of brucellosis among 870 admitted febrile patients [13]. Some non-malaria febrile illnesses may contribute to high morbidity and mortality in humans. For instance, rotavirus takes the lives of more than 8,100 Tanzanian children under five each year [25]. Furthermore, the most recent outbreak of RVF in 2006/2007 which occurred in 10 regions of Tanzania mainland [26] and in other countries such as Kenya and Somalia were associated with widespread morbidity and mortality in humans [27]. The diagnosis of non-malaria febrile illnesses poses a challenge since many of these illnesses may have similar symptoms with malaria and thus making their clinical diagnosis difficult [28]. Also, non-malaria febrile illnesses could have common overlapping manifestations and therefore, this absence of specific symptoms make it difficult to distinguish several non-malaria febrile conditions that often occur in the same area [29]. Clinical overlap between diseases may result in inappropriate antimicrobial therapy and therefore, laboratory tests for differential diagnosis of causative agent are essential. Following a long tradition of regarding malaria as the leading cause of fever, it is important for the community to understand the other causes of fever apart from malaria particularly during this period when the episodes of malaria related fevers are reported to decrease [8], [11], [30]. Understanding the awareness of the community on non-malaria febrile illnesses is critical and relevant particularly in management and control of such illnesses. Despite their importance, only few studies aiming at assessing the awareness of the communities regarding non-malaria febrile illnesses have been conducted in Tanzania [31], [32]. Therefore, this study intended to contribute in filling the information gap by assessing the knowledge and attitude of the communities regarding non-malaria febrile illnesses. In addition, the study explored treatment seeking behaviors for febrile illnesses among community members. Following the longstanding practice of treating most fevers as malaria, health workers may still treat febrile patients with anti-malarial drugs even if the patients had a negative test results. Studies from Tanzania and other countries like Zambia, Uganda and Burkina Faso have indicated that febrile patients were prescribed anti-malarial drugs following negative mRDT/microscopy result [33]–[36]. Therefore, there is a need to know the management of febrile patients following the decline in the incidence of malaria. The current study also assessed health workers' practices related to diagnosis and treatment of febrile patients. The study was conducted in Kilosa district which is one of the six districts in Morogoro region, located in eastern Tanzania. The district borders with Tanga and Manyara regions to the north and Mvomero district and Mikumi National Park to the east. On the western border are Dodoma and Iringa regions whereas to the south it borders with Kilombero district. The district lies between latitudes 6° south and 8° south and longitudes 36°30′ east and 38° west. The area has semi humid climate with an average rainfall of 800 mm annually. The early rains start in November and end in January followed by heavy rainfall between March and May. The district experiences a long dry season from June to October and the average annual temperature is 24. 6°C. The district has an area of 14,245 square kilometers and has a population of 438,175 people [37]. It consists of a mixture of different ethnic groups predominantly Kaguru, Sagara and Vidunda. The main economic activities are crop production and livestock keeping. More than 77% of people are subsistence farmers and major crops cultivated include maize, cassava, rice, paddy and sorghum whereas the major cash crops are sisal, sugarcane, cotton and oilseeds. Kilosa was selected due to its possession of intensive human activities with livestock as well as its proximity to wildlife from the Mikumi National Park (figure 1), what was expected to be a good interface for zoonotic diseases such as RVF and Brucellosis [38]. Administratively, Kilosa district is divided into 9 divisions, 37 wards and 164 villages [39]. In terms of health care services, Kilosa district has 71 health facilities and among these, there are 3 hospitals, 7 health centers and 61 dispensaries [40]. However, the number of villages exceeds the number of health facilities and hence most health facilities serve more than one village. Kilosa district is an area with holoendemic malaria transmission with seasonal peaks following the long and short rainy seasons [41]. According to Tanzania HIV and Malaria Indicator Survey, in 2007–2008 malaria prevalence was estimated to be 15. 7% in Morogoro region [42] and decreased to 13% in the year 2011–2012 [43]. The common non-malaria febrile illnesses that have been reported in Kilosa district include acute respiratory diseases, UTIs and typhoid fever [24]. Data from a platform for health monitoring and evaluation in Tanzania (Sentinel Panel of Districts) have shown that in the year 2011, acute respiratory diseases and UTIs comprised of 20% and 2. 5% respectively of total recorded illnesses (77,862) in outpatient department in children aged less than 5 years [44]. This study was specifically conducted in 6 divisions, namely Kimamba, Kilosa town, Magole, Masanze, Rudewa and Ulaya. Within these divisions, 12 wards namely Dumila, Chanzuru, Magomeni, Kilosa, Msowero, Zombo, Ulaya, Kimamba, Mkwatani, Kasiki, Masanze and Rudewa were purposively selected based on (i) geographical representation within the district e. g. Zombo is in the south western part of the district, whereas Dumila ward is in north-eastern part (figure 1), (ii) the presence of government health facilities (iii) connectivity of the wards and ease of accessibility by road. We conducted a cross-sectional study in which qualitative data collection methods were used. Focus group discussions (FGDs) with members of the communities were conducted to assess their knowledge, attitude and perception of community members on non-malarial febrile illnesses. In-depth interviews (IDIs) with health workers were conducted in order to obtain their views about practices related to diagnosis and management of non-malarial febrile illnesses. Parents, guardians or caregivers aged between 18–59 years for children of less than 10 years were eligible to participate in FGDs. This group of participants was targeted because febrile illnesses have been shown to be common in children and contribute to high proportion of hospital admissions globally, with significant morbidity and mortality [13], [18], [45]. The participants were recruited from different hamlets within the study wards with the assistance of local government and villages leaders. In total 12 FGDs were conducted in urban, peri-urban and rural areas in the selected wards of which 5 FGDs were with men and 7 involved women. Each FGD comprised 6–8 people, but women and men were separately interviewed to give the participants freedom to talk during the discussions. Two FGDs were conducted per day and each FGD took about 60 to 90 minutes. In each of Dumila, Rudewa, Chanzuru and Ulaya wards 2 FGDs were separately conducted for men and women. In Masanze and Zombo wards 2 FGDs were conducted with women and the remaining 2 FGDs (1with men and 1with women) involved participants selected from Kilosa, Kasiki and Magomeni wards. For IDIs, only health workers who were on duty and attended patients (prescribers) in the health facilities during the study period were eligible to participate into the study. Health workers from 12 health facilities located in Dumila, Chanzuru, Ulaya, Zombo, Kilosa, Magomeni, Msowero, Kimamba, and Mkwatani wards were interviewed. Two health workers were interviewed from each health facility and only one health worker was interviewed at a time. In-depth interviews with health workers on average lasted for one hour and 2 IDIs were conducted per day. Focus group discussions and IDIs were conducted by a skilled and experienced social scientist who was assisted by an observer and note taker. All discussions and interviews were conducted based on a prepared semi-structured interview guide that consisted of questions corresponding to the research topic. To ensure accuracy of the information, the data collection tool was translated from English to Kiswahili and then back-translated. The interview guide for FGDs consisted of questions about their knowledge on non-malaria febrile illnesses, health care seeking behaviors and their recommendations on non-malaria febrile illnesses (supporting information S1). Health workers were asked questions on the awareness of the communities on non-malaria febrile illnesses, communities' health seeking behaviors, how they perform diagnosis and treatment of febrile patients and their recommendations on proper management of non-malaria febrile illnesses (supporting information S1). During the interviews and discussions, notes were taken and conversations were digitally recorded. Field notes were expanded on the same day of the interview/discussion. All FGDs and IDIs were held in Swahili language which is the most widely spoken language by the community (national language). FGDs and interviews were transcribed verbatim and translated from Swahili to English. Thereafter, the transcripts were converted into rich text format and imported into MaxQDA, a software for qualitative data analysis [46]. Text files were independently reviewed by the two researchers (IM and CM) before agreeing on the different themes and categories. In case of differing interpretations, the discussion between the researchers took place until the final agreement was reached. The findings were also validated by the interviewing researcher (BC). The agreed themes and categories were then coded. The retrieved segments were analyzed using thematic content analysis and their respective codes were exported to Excel for quantitative analysis. Ethical approval was obtained from Institutional Review Board of Ifakara Health Institute (IHI/IRB/No: 01-2013) and Medical Research Coordinating Committee of Tanzania' s National Institute for Medical Research (NIMR/HQ/R. 8a/Vol. 1X/1472). A written informed consent was obtained from each respondent and participant prior to IDIs or FGDs. To protect identification of the respondents and FGD participants, all personal information that could identify the study participants were only used during the analysis and omitted from the final reports. The participants were assured of anonymity in the presentation and publishing of the data. The study participants were asked to explain what they know about the term “fever” (“homa” in Swahili language). There were different levels of understanding among the participants. The responses provided by majority of the participants did not associate fever with high body temperature. They described fever as an illness condition such as malaria, colic, rheumatism and sleeping sickness or associated it with symptoms such as headache, coughing, rashes and body pain. Others reported that they did not know the exact meaning of the term fever. There were few participants who described fever as a raise of body temperature (hot body). When the participants were asked to mention the causes of fever in children, things such as change of weather (cold, high temperature) and sunlight were listed by many participants. The participants believed that exposure to a cold environment or prolonged stay under the sun by itself can lead to fever. Only a few participants mentioned the right cause of fever which included illnesses like measles, tuberculosis (TB), typhoid fever and UTIs and other participants associated fever with symptoms/clinical signs such as flu, coughing and diarrhea. Moreover, inappropriate beliefs were perceived as causes of fever by the participants. For example, the presence of false teeth (meno ya plastiki in Swahili language), breastfeeding after long-term sunlight exposure of the mother as well as cessation of pulsation in the fontanel were mentioned by some participants. When the FGD participants were asked to mention the causes of fever other than malaria, several of them listed diseases or symptoms which were neither associated with fever nor non-malaria febrile illnesses. The commonly reported diseases/symptoms were headache, colic, hernia and abdominal pain. However, some participants admitted to be unaware of such illnesses. Only a small number of the participants mentioned the correct non-malaria febrile illnesses such as typhoid fever, UTIs (dirty urine), pneumonia, measles and tuberculosis (TB) and sleeping sickness. Both men and women participants had similar level of knowledge, but participants older than 30 years were more knowledgeable than those who were younger. Moreover, the participants from FGDs conducted in rural areas had limited knowledge in comparison with participants from urban and semi-urban areas. It was also revealed that despite the decrease of malaria, the participants believed that most instances of fever were due to malaria. This was noted when several participants mentioned only malaria as the cause of fever. The participants explained that when their children get fever, they mostly associated it with malaria. This was also acknowledged when the participants were asked to describe the meaning of the term fever (theme 1). A considerable number of the participants explained fever is malaria: During the interviews, health workers were asked to give their views about the knowledge of the community members on non-malaria febrile illnesses. All health workers reported that majority of community members were not aware or had little knowledge on these illnesses. Health workers explained that fevers were perceived to be caused by malaria by several members of communities. This situation was reported to be a challenge to health workers particularly when they want to obtain a comprehensive history of illness from patients since patients explain to health workers that they are suffering from malaria. Health workers identified this wrong perception as an impediment to proper management of patients and emphasized the need for change of attitude. In our study the most common reason for unawareness of the community on non-malaria febrile illnesses reported by the majority of health workers was lack of health education. Health workers pointed out that health education is offered at health facilities but it has not reached a wider community. Several community members live in remote villages and they rarely visit health centres for health care. Health workers emphasized that health education on diseases associated with non malaria fevers will help to create awareness to members of the community. Findings from the FGDs revealed that community members from the study population sought treatment from both the health facilities and traditional healers. When queried on their health seeking behavior, the majority of the study participants reported sending their febrile children to the nearby health facility. However, when asked for any alternative treatment, almost half of the participants reported to seek treatment from the traditional healers. Therefore, this indicates that health facilities and traditional healers were both utilized by the participants. Even though several participants reported seeking treatment from health facilities, but interviewed health workers explained that community members rarely visit health facilities. They said the majority of community members live in remote areas and thus long distances pose as an impediment for visiting health facilities. In addition, health workers stated that the habit of seeking treatment from traditional healers was practiced by some members of the community. They further pointed out that sometimes children were sent to health facilities when they were terminally ill. Likewise, some parents/guardians admitted to health workers that delays to send their children to health facilities was due to prior consultations made from traditional healers. Our findings have also shown that seeking treatment from the traditional healers was sought when there was a persistent fever following treatment from health facilities. If children still had fever after completion of anti-malarial dosage, majority of the participants opted to consult traditional healers for further treatment. Only a few mentioned taking their children back to the health facilities. This study found that self-medication was commonly practiced by several participants. Many participants reported purchasing drugs from pharmacy/drug shops without prior medical prescription. They explained that they prefer using anti-malarial drugs for the treatment of fever in children. The practice of self-medication was reported by FGD participants from wards located near the health facilities as well as wards which were far off from health facilities. During the interviews with health workers, all of them acknowledged that self-medication was commonly practiced by members of the communities. Heath workers further pointed out that some members of the communities would still visit health facilities if they had found no improvement following self-medication. Health workers explained that the habit of self-medication delays provision of prompt and proper treatment and in most cases results into death. The common reported reasons which influenced many members of the community to opt for self-medication were poor health services from health facilities, shortages of drugs, lack of diagnostic facilities, long distance to the nearby health facility and inability to afford health care charges. In our study, we found that the diagnosis of febrile patients was mostly done by mRDTs or clinical symptoms/signs as presented by patients with the assistance of Integrated Management of Childhood Illness guidelines. Interviewed health workers explained that mRDTs were used to distinguish malaria from non-malaria fevers and the majority (10/14) prescribed anti-malarial drugs only to patients with mRDT positive result. When patients had negative mRDT result, health workers reported looking for other causes of fever based on clinical signs and history of the fever. Only a few health workers (4/14) stated initiating anti-malarial drugs even if patients had negative mRDT, as one health worker said: Moreover, when mRDTs were not available, majority of health workers relied only on clinical manifestations of the patient. When asked to describe symptoms or signs which guide them to make a conclusive diagnosis of febrile illnesses such as malaria, typhoid fever or urinary tract infections, most health workers (12/14) mentioned presence of fever, vomiting, headache, loss of appetite and diarrhea as typical symptoms of malaria. With regards to UTIs, pain during urination was mentioned by majority of health workers as a definitive symptom whereas fewer health workers (2/14) considered urine coloration (from yellowish to milky colour) and small urine volume as symptoms of UTIs. For typhoid fever, symptoms such as abdominal pain and diarrhoea were commonly listed by the majority of health workers. It was also revealed that when mRDTs were available many febrile patients tested negative and hence other causes of fever were very likely to be considered. However, in the absence of mRDTs, health workers said several febrile patients were suspected to have malaria and were treated with anti-malarial drugs. They considered that reliance on clinical signs and symptoms only, is prone to lead to misdiagnosis and over-prescription of anti-malarial drugs. Opinions of health workers towards management of persistent fevers following completion of anti-malarial dosage were quite divergent. While majority of health workers (10/14) reported opting for symptomatic diagnosis of non-malaria febrile illnesses and appropriate prescription of antibiotics following failed malaria treatment, however fewer health workers (4/14) reported to switch from first line anti-malarials (artemisinin-based combination therapy) to second line anti-malarial treatment (quinine). Proper management of non-malaria febrile illnesses is largely dependent on the capacity of the health facility to perform accurate diagnosis and treatment of these illnesses. Health workers in this study repeatedly reported lack of diagnostic facilities, shortage of trained health workers, and stock-out of medications as major barriers to proper management of non-malaria febrile illnesses. Our study revealed that from the 12 health facilities, only 2 had diagnostic facilities for a few febrile illnesses; the dispensary had Widal test for detection of typhoid and a microscope used in diagnosis of UTIs and mRDTs, while the health center only had a microscope. The remaining health facilities (10/12) had no diagnostic tests except mRDTs in few health facilities. Health workers explained that they experienced challenges in managing febrile patients without tools for laboratory investigation. Even though mRDT was named as the only diagnostic test which was used to rule out malaria from febrile patients, stock-out of mRDTs was mentioned as an ongoing problem. Health workers said that the supply of mRDTs from the government to the health facilities was normally done on a quarterly basis. However, all health workers repeatedly reported receiving inadequate mRDTs and there were frequent delays in supply of mRDTs. It was clearly stated by health workers that in most cases half way through a quarter, they experienced lack of mRDTs. Among 12 health facilities, only 4 had mRDTs available at the time of our visit. Health workers insisted need for diagnostic tools for malaria and non-malaria febrile illnesses. During the interview with health workers, stock-out of medication was mentioned as a common problem. Drug shortages were reported in most (9/12) health facilities although a few (3/12) had a few boxes of basic drugs such as ALU, antibiotics (amoxicillin, septrin and metronidazole) and pain killer (paracetamol). With regards to staffing, our study revealed significant shortage of health workers particularly in health facilities located in rural areas. Among visited health facilities, four (a hospital, two health centers and one dispensary) had more than two health workers at the level of clinical officers. The remaining eight health facilities each had only one clinical officer assisted by a nurse/midwife or medical attendant. Some of the interviewed health workers (medical attendants/nurses) explained that although they prescribe drugs to patients, they had inadequate skills to manage febrile patients. Moreover, health workers reported to work beyond normal working hours and thus attend more patients beyond the standard average number of patients per physician. FGD participants were selected from different divisions, wards and hamlets, hence they were from diverse demographic backgrounds and their views were a representation of the general population in the district. Interviewed health workers were selected from different health care levels, dispensary to hospital levels. This was purposively done to grasp a wide scope of attitudes and practices by health care staff from the few health facilities which were visited. During focused group discussion with communities more discussions were done with women as compared to men, but this was purposively done because women are responsible for child care in the family hence their views were expected to give a balanced representation of the household health. However, the results of this study showed that both men and women had similar level of knowledge when it comes to perceptions of non-malaria febrile illnesses. It is also possible that some information was lost during the translation of the transcripts before analysis. This study has demonstrated that the awareness and level of knowledge of communities on non-malaria febrile illnesses was low. Knowledge from this and other similar studies will provide insights into better and practicable methods for improving the management of febrile patients. The wrong perception among communities, whereas fever is understood as being synonymous with malaria, as encountered in this study pose a challenge to the health sector and thus we emphasizes the need of creating public awareness regarding causes of fever other than malaria. Community misconceptions on fever and its causes must be addressed since such beliefs often distract or delay treatment seeking from health care facilities. It is also crucial that relevant authorities intervene against existing habits of self-medication and seeking treatment from traditional healers. Appropriate treatment of febrile patients will require inputs geared towards strengthening of diagnostic facilities, drugs availability and optimal staffing of health facilities. Therefore, it is advisable that the government and other stakeholders should take appropriate measures to improve health care services delivery.

Understanding the awareness of the community on non-malaria febrile illnesses is crucial, especially during the recent decline of malaria episodes of malaria. This study conducted focus group discussions with communities to assess their awareness of non-malaria febrile illnesses. In addition, in-depth interviews with health workers were conducted to explore their views and practices related to diagnosis and management of these illnesses. We identified that the awareness of the study participants on non-malaria febrile illnesses was low and the majority believed that most instances of fever are due to malaria. Moreover, the participants could not mention the right causes of fever and many had inappropriate beliefs about possible causes of fever. Health workers from our study looked for non-malaria febrile illnesses when a febrile patient had negative mRDT result or there was persistence of fever following completion of anti-malarial dosage. Shortages of diagnostic facilities were identified as one of the impediments to proper diagnosis of febrile illnesses. These findings indicate the need for creation of public awareness on causes of fever other than malaria. We recommend appropriate measures be taken by the government and other stake holders to improve health care services delivery particularly at primary health care facilities.

Abstract Introduction Methods Results Discussion

biology and life sciences veterinary science medicine and health sciences social sciences

2014

Community Knowledge and Attitudes and Health Workers' Practices regarding Non-malaria Febrile Illnesses in Eastern Tanzania

The simultaneous targeting of host and pathogen processes represents an untapped approach for the treatment of intracellular infections. Hypoxia-inducible factor-1 (HIF-1) is a host cell transcription factor that is activated by and required for the growth of the intracellular protozoan parasite Toxoplasma gondii at physiological oxygen levels. Parasite activation of HIF-1 is blocked by inhibiting the family of closely related Activin-Like Kinase (ALK) host cell receptors ALK4, ALK5, and ALK7, which was determined in part by use of an ALK4,5, 7 inhibitor named SB505124. Besides inhibiting HIF-1 activation, SB505124 also potently blocks parasite replication under normoxic conditions. To determine whether SB505124 inhibition of parasite growth was exclusively due to inhibition of ALK4,5, 7 or because the drug inhibited a second kinase, SB505124-resistant parasites were isolated by chemical mutagenesis. Whole-genome sequencing of these mutants revealed mutations in the Toxoplasma MAP kinase, TgMAPK1. Allelic replacement of mutant TgMAPK1 alleles into wild-type parasites was sufficient to confer SB505124 resistance. SB505124 independently impacts TgMAPK1 and ALK4,5, 7 signaling since drug resistant parasites could not activate HIF-1 in the presence of SB505124 or grow in HIF-1 deficient cells. In addition, TgMAPK1 kinase activity is inhibited by SB505124. Finally, mice treated with SB505124 had significantly lower tissue burdens following Toxoplasma infection. These data therefore identify SB505124 as a novel small molecule inhibitor that acts by inhibiting two distinct targets, host HIF-1 and TgMAPK1. Toxoplasma gondii infects approximately one-third of the world' s population and causes disease primarily in developing fetuses or immunocompromised individuals [1]. Humans and other intermediate hosts are infected with Toxoplasma by digesting either sporozoite-containing oocysts that are shed in feline fecal matter or bradyzoite-laden tissue cysts in undercooked meat [2]. In the intestine, the parasites infect intestinal epithelial cells and then convert into the replicative form of the parasite called tachyzoites [3]. As tachyzoites disseminate through the host, they encounter various host defenses or pharmacological agents that in most cases kill the tachyzoites. However, some escape killing and transform into bradyzoites that go on to develop into tissue cysts. These tissue cysts cannot be detected by the host' s immune system and are impervious to most, if not all, currently prescribed drugs [3]–[8]. Thus, Toxoplasma is highly prevalent in humans, in large part, because tachyzoites have evolved to grow within its host until it is challenged and then respond by forming a life-long chronic infection. For tachyzoites to be able to grow, they must establish a replicative niche within their host cells and do so by inducing a number of changes to host cell signaling, transcription, and organellar/cytoskeletal organization [9]–[17]. One of these changes includes activation of the host cell transcription factor hypoxia-inducible factor 1 (HIF-1), which is important for parasite growth [18]. Toxoplasma activates HIF-1 by stabilizing the abundance of the HIF-1α subunit. HIF-1α stabilization is accomplished by the parasite down regulating the abundance and activity of the Prolyl Hydroxylase Domain 2 (PHD2) protein whose primary function is to target HIF-1α for proteasomal-dependent degradation [19], [20]. To down regulate PHD2, Toxoplasma requires signaling via a family of host serine/threonine kinase receptors named the activin-like kinases (ALK4, ALK5, and ALK7) [21]. SB505124 [2- (5-benzo[1,3]dioxol-5-yl-2-tert-butyl-3H-imidazol-4-yl) -6-methylpyridine hydrochloride], which is a highly selective ALK4,5, 7 competitive inhibitor, was one reagent used to demonstrate a role for ALK4,5, 7 signaling in HIF-1 activation [21], [22]. But besides inhibiting ALK4,5, 7/HIF-1, SB505124 also potently blocked parasite replication. Although supporting a role for ALK4,5, 7 signaling in Toxoplasma growth, these data were enigmatic because the drug' s effect on parasite growth was more severe than the loss of HIF-1α. For example, the drug significantly blocked parasite replication at 21% O2 (Figure 6 in [21]) whereas parasite growth was more modestly attenuated in HIF-101629226835445KO cells at this O2 tension (Figure 5 in [18]). Two plausible explanations exist to address this: first, ALK4,5, 7 signaling regulates not only PHD2/HIF-1 but other host pathways that are important for parasite growth. Second, the drug has a second target that may be either host- or parasite-encoded. These possibilities were addressed using a forward genetic screen to isolate and characterize SB505124 resistant parasites. This screen revealed that SB505124 inhibits Toxoplasma growth by not only inhibiting ALK4,5, 7/HIF-1 signaling but by targeting a parasite MAP kinase named TgMAPK1. To define how SB505124 inhibited parasite growth, we developed a forward genetic screen to isolate SB505124-resistant (SBR) parasites. Wild-type RHΔHXGPRT (RHΔ) parasites were chemically mutagenized with N-Ethyl-N-nitrosourea (ENU) and SBR mutants were isolated by growth in the presence of SB505124. Three SBR mutants (SBR1, SBR2, and SBR3) were isolated from three independent mutagenesis screens and relative resistance of each mutant to SB505124 was determined by plaquing assays. SBR mutants displayed similar IC50 values ranging from 4. 9 µM to 5. 4 µM, which relative to the parental RHΔ strain represented ∼5-fold increases in resistance to SB505124 (Figure 1A). Unless stated otherwise, the remaining assays in this report were performed with SBR1. To define how SB505124 affected parasite growth, mock- or drug-treated parasites were stained with DAPI and anti-SAG1 antisera to assess DNA content and identify individual parasites by immunofluorescence. SAG1 staining revealed that, as we previously reported [21], 3 µM SB505124 caused parasites to arrest growth as single SAG1+ parasites (Figures 1B&C). However, DAPI staining indicated that most vacuoles, including the seemingly single SAG1+ parasites, contained irregular numbers (non-2n) of parasites/vacuole (Figure 1C “Irregular Vacuole”). Numbers of nuclei/parasite bodies of those parasites growing within the “Irregular Vacuoles” were quantified (Figure 1D). Approximately 65% of vacuoles from the drug-treated RHΔ parasites contained ≥2 nuclei/parasite body. We also noted that approximately 25% of the SB505124-treated vacuoles contained parasites that were SAG1+/DAPI−, which are denoted as <1 nucleus/parasite body. In contrast, SBR1 replication was neither affected by SB505124 nor did the drug trigger an accumulation of multinucleated parasites or irregular vacuoles (Figure 1C&D). These data suggest that SB505124 disrupts tachyzoite cell cycle progression. SB505124 is a substituted imidazole compound that was developed from the same structural scaffold as the p38 MAP kinase inhibitor SB203580 [22]. SB203580 inhibits Toxoplasma replication and does so presumably by targeting a SB203580-sensitive, parasite-encoded kinase [23], [24]. Even though SB505124 more potently inhibits ALK4,5, 7 than human p38 MAP kinases [22], it was possible that SB505124 reduced parasite growth by targeting the same kinase that SB203580 inhibits. A plaquing growth assay, however, revealed that SB203580 similarly inhibited RHΔ and SBR1 growth (IC50∼15 µM) (Supplemental Figure S1), which is consistent with earlier reports [23]. Thus, these two structurally related kinase inhibitors appear to inhibit Toxoplasma through distinct mechanisms and target proteins. The preceding assays were performed with RH strain parasites because it is the strain best suited for genetic manipulation. But one limitation of the RH strain is that it grows exclusively as tachyzoites and does not readily undergo bradyzoite differentiation either in vitro or in vivo as other parasite strains can. Before testing the effect of SB505124 on the ability of the other parasite strains to undergo bradyzoite development, we first assessed whether SB505124 inhibited their proliferation as tachyzoites. We therefore grew Pru (a Type II strain), CTG (a Type III strain), and GT-1 (a Type I strain that can form bradyzoites) tachyzoites in the presence of SB505124 and determined the drug' s IC50 towards each strain (Figure 2A). SB505124 blocked lytic replication of all 3 strains with IC50 values lower than RH' s IC50. To assess the impact of SB505124 on bradyzoite development, GT-1, Pru, and CTG tachyzoites were grown on coverslips in the presence of SB505124. After 72 hours, parasites were fixed and stained with Dolichos-FITC to detect bradyzoite-containing cysts. We found that 3 µM SB505124 induced the formation of Dolichos+ cysts (Figure 2B and Supplemental Figure S2). But unlike high pH-induced bradyzoites, SAG1 staining could still be detected in the encysted parasites. To further assess bradyzoite development, Pru parasites were grown for 72 h in human foreskin fibroblasts (HFFs) in the presence of 3 µM SB505124 and bradyzoite specific gene expression was assessed by real time PCR. The data indicated that SB505124 induced a several hundred-fold increase of the bradyzoite gene transcripts, BAG1, LDH2, and ENO1 (Figure 2C). While a down-regulated trend was observed for the tachyzoite-specific gene ENO2, decreases in its abundance were not statistically significant (Figure 2C). Thus, SB505124 treatment induces some, but not all, features of bradyzoite development in cystogenic strains of Toxoplasma. To identify the genetic basis for SB505124 resistance, genomic DNA purified from each SBR mutant and the parental RHΔ parasites were subjected to Illumina whole genome sequencing. The sequenced genomes were aligned to the GT-1 Toxoplasma gondii reference genome and compared to identify ENU-induced single nucleotide variants (insertions or deletions were not detected) (Supplementary Table S1). To prioritize testing of candidate SBR mutations, we searched for genes that were mutated in more than one SBR mutant. From this filter, only one gene, Toxoplasma MAP Kinase 1 (TgMAPK1; TGGT1_312570 @ www. toxodb. org), was mutated in all three SBR mutants (Figures 3A&B and Supplemental Table S1). These three mutations affected the translation of two amino acids (Leu162→Gln and Ile171→Thr or Asn) in exon 2 that are part of the kinase' s predicted ATP binding pocket within its catalytic domain. To determine whether these TgMAPK1 mutations were bona fide SBR alleles, we used an allelic replacement strategy to test whether replacing the wild-type TgMAPK1 allele with the mutant one conferred SB505124 resistance (Figure 3C). Thus, RHΔku80 parasites were transfected with TgMAPK1WT or TgMAPK1SBR1 allelic replacement constructs consisting of 900 bp fragments of the TgMAPK1 gene containing either WT or candidate SBR alleles. The transfected parasites were grown in the presence of 3 µM SB505124 and parasite growth was monitored visually. While we could not detect growth of those parasites that were transfected with TgMAPK1WT DNA, we found that parasites grew when they received any of the candidate SBR alleles (Figure 3D). To verify that the mutant TgMAPK1 alleles had properly integrated into the genome, we confirmed the presence of each SBR allele by sequencing a PCR fragment containing TgMAPK1 mutations (data not shown). Thus, single nucleotide mutations in exon 2 of TgMAPK1 were sufficient to rescue growth in SB505124. We hypothesized that if the function of TgMAPK1 was epistatic to ALK4,5, 7/HIF-1 then HIF-1 activation would be unimpaired in the presence of SB505124 and that growth of the SBR mutants would be restored in HIF-1 knockout cells. First, HIF-1 luciferase reporter-transfected host cells were infected for 18 h with wild-type RHΔ, SBR1, SBR2, or SBR3 parasites in the absence or presence of SB505124. The data indicated that HIF-1 was activated by all 4 strains and this activation was similarly inhibited by SB505124 (Figure 4A). Next, we used [3H]-uracil incorporation to measure SBR mutant growth in HIF-1α knockout cells at 3% O2 and found that growth of the SBR mutants were as reduced as wild-type parasites growing in the HIF-1αKO cells (Figure 4B). HIF-1 is activated by a diffusible factor released from extracellular tachyzoites [18]. We, therefore, needed to eliminate the possibility that SB505124 blocked HIF-1 activation due to a global inhibition of host cell responses to extracellular parasites. Thus, we tested whether SB505124 inhibited STAT3 activation since activation of this host cell transcription factor is due to ROP16, which is a polymorphic rhoptry-encoded kinase that is injected by Type I and III strain parasites into host cells independently of invasion [15]. We found that RH strain-induced STAT3 phosphorylation and nuclear localization was unaffected by SB505124 (Figure 4C). Together, these data indicate that SB505124 has two distinct targets in Toxoplasma-infected cells – TgMAPK1 and host HIF-1. SB505124 was identified as an ATP competitive inhibitor of ALK4,5, 7 kinase activity [22]. Although SB505124 does not inhibit host MAPKs [22], we tested whether TgMAPK1 was identified as an SBR gene due to the drug inhibiting the parasite kinase. Our (and others [25]) initial attempts to express in E. coli either full-length TgMAPK1 or the kinase' s N-terminal half containing the kinase domain resulted in a protein preparation of low purity and limited specific activity (not shown). Moreover, we were unable to eliminate the possibility that any activity we did detect was due to a bacterial kinase contaminating our preparations. We therefore developed a Toxoplasma strain in which the C-terminus of TgMAPK1 is epitope tagged at its endogenous chromosomal locus (Figure 5A). Western blotting whole cell lysates with anti-HA antibodies revealed a single immunoreactive band at ∼150 kDa, which is consistent with the predicted molecular weight of TgMAPK1-HAx3 of 141 kDa (Figure 5B). Unlike Brumlik et al [26], we failed to note the presence of smaller immunoreactive bands. Because TgMAPK1 target substrates have yet to be identified, we assessed its activity using an autokinase assay. Thus, HA-tagged TgMAPK1 was immunoprecipitated from Toxoplasma lysates and in vitro kinase assays were performed in the presence of γ32P-ATP. We found that TgMAPK1 was autophosphorylated (Figure 5B), indicative of an active kinase, and that this autokinase activity was not detected in immunoprecipitates from wild type non-transgenic controls (Figure 5C). Dose response curves revealed that SB505124 inhibited TgMAPK1 autokinase activity with an apparent IC50 of 125 nM (Figure 5D&E). To address the possibility that the TgMAPK1 immunoprecipitates contained a contaminating kinase whose activity was SB505124 sensitive, we attempted to ectopically express a kinase-dead TgMAPK1 mutant cDNA by mutating the conserved lysine140 in the catalytic domain to an arginine. Repeated attempts to either transiently or stably express this mutant were unsuccessful as were attempts to express SBR mutants that were epitope tagged as either transgenes or at the endogenous TgMAPK1 locus. We therefore turned our attention to a TgMAPK1 temperature sensitive mutant that was isolated independently of this study in a temperature sensitive growth screen (LE and MW; manuscript in preparation). To minimize potential misfolding of the TgMAPK1ts protein, autokinase activity was compared between parasites grown at 34°C (the permissive growth temperature for the ts mutant) that harbored either HA-tagged TgMAPK1WT or TgMAPK1ts alleles. Unlike TgMAPK1 kinase-dead and SBR mutants, TgMAPK1ts-HA could be stably expressed and similar amounts of TgMAPK1WT-HA or TgMAPK1ts-HA could be immunoprecipitated using anti-HA beads (Figure 5F). In contrast, significantly lower levels of autokinase activity could be detected in the TgMAPK1ts immunoprecipitates suggesting that the kinase activity in the immunoprecipitates was most likely dependent on TgMAPK1 and not a contaminating kinase (Figure 5G). Even though the wild-type and ts mutants grew at similar rates at 34°C we cannot fully rule out the possibility that potential misfolding of TgMAPK1ts at the permissive temperature may affecting binding of a contaminating kinase. The differences in autokinase activity between the wild-type and ts kinases and the fact that the SBR mutations are in the region of the ATP binding pocket where the SB class of kinase inhibitors interact with their target kinases [27]–[29] strongly suggests that TgMAPK1 is an active kinase that can be inhibited by SB505124. Our in vitro data suggests that SB505124 blocks Toxoplasma growth in an unusual manner by which both host and parasite pathways are simultaneously targeted. To assess whether SB505124 can limit parasite growth in vivo, C57Bl/6 mice were intraperitoneally infected with 103 GFP-expressing RH tachyzoites and then treated with daily intraperitoneal injections of either 10 mg/kg SB505124 or vehicle (DMSO). On Day 5 post-infection, mice were sacrificed and peritoneal exudate cells were stained with anti-CD45 (to identify infiltrating leukocytes) and analyzed by flow cytometry. Mice treated with SB505124 had significantly fewer numbers of infected cells (12. 8%) compared to the vehicle controls (25. 7%) as measured as GFP+/CD45+ events (p = 0. 0252) (Figures 6A, B). To determine whether the infected cells contained similar numbers of parasites, we measured the mean fluorescence intensity (MFI) of GFP in CD45+/GFP+ cells and found that infected cells from SB505124-treated mice displayed a 26. 8% reduction in relative GFP MFI compared to vehicle-treated control mice (p = 0. 0988) (Figure 6C). Though this decrease was not statistically significant, it is likely an underestimation as SB505124-treated parasites display enlarged, translationally active parasite bodies that could still produce high levels of GFP. The polyploidy phenotype induced by SB505124 precluded us from further assessing parasite burdens by qPCR analysis of genomic DNA. Since SB505124 inhibits TGFβ signaling, a known antagonist of IFNγ production [22], [30], it is possible that the effect of SB505124 on Toxoplasma in mice is due to enhanced IFNγ-driven immunity. However, IFNγ serum levels from the mice 5 days post infection were identical between SB505124 and vehicle-control treated mice (Figure 6D). Our initial interest in SB505124 stemmed from finding that this compound blocked the parasite' s ability to activate HIF-1 via ALK4,5, 7 signaling [21]. This conclusion was supported by additional data showing that overexpression of SMAD7, which is an endogenous ALK4,5, 7 inhibitor, also blocked parasite activation of HIF-1 [21]. Additional unpublished work demonstrated that HIF-1 activity is increased in cells transfected with ALK4,5, 7 expression plasmids. But since parasite replication was more severely reduced in SB505124-treated cells than in HIF-1α knockout cells, these data suggested that in addition to the ALK4,5, 7/HIF-1 pathway the drug had at least one additional target that could be in either the host or parasite. Our forward genetic screen resolved this issue and identified SB505124 as the first compound that we are aware of that inhibits growth of an intracellular pathogen by acting on seemingly unrelated targets in both the host and parasite. We believe that collectively our data demonstrate that TgMAPK1 and ALK4,5, 7/HIF-1 are the two relevant SB505124 targets in parasite-infected cells. While we cannot rule out that the drug may have additional targets in Toxoplasma-infected cells, these targets would not be important for parasite growth. Although the catalytic domain of TgMAPK1 is homologous to other MAPKs, it has two unusual features that may facilitate TgMAPK1-specific agonist development. First, TgMAPK1 has 3 unique insertions within its catalytic domain with the most predominant one being a 93 amino acid insertion between the DFGLAR-motif that coordinates divalent cations. How the catalytic domain of the kinase can properly fold with this large insertion is currently unknown and cannot be easily predicted since the insertion prevents protein-modeling programs from using known MAPK structures as scaffolds (Brown and Blader, unpublished results). Second, TgMAPK1 has an ∼800 amino acid C-terminal extension that lacks conserved domains and has no homology to any known protein either in Toxoplasma or its host. Given that inhibition of TgMAPK1 induces multinucleated parasites (suggesting a cell cycle defect) and that the C-terminal extension facilitates its localization to mitotic structures (E. Suvorovo et al, manuscript in preparation) we hypothesize that TgMAPK1 functions in cell cycle regulation and our future work will define if and how TgMAPK1 regulates the cell cycle. TgMAPK1 was originally discovered based on its homology to human MAPKs [24]. It was assigned then as a p38 MAPK homolog, in part, because its catalytic activity was reported to be sensitive to SB203580. In our hands, TgMAPK1 activity was largely unaffected by this compound (Supplemental Figure S1B), although a general decrease in autokinase activity by low SB203580 levels was noted that may reflect reduced non-specific binding of ATP to the kinase. Regardless, higher amounts of SB203580 did not significantly decrease TgMAPK1 autokinase activity and SBR mutants are similarly sensitive to SB203580 (Supplemental Figure S1A&B). We believe that there are two primary reasons for differences in between our data and those of [24]. First, our experiments used endogenously epitope-tagged TgMAPK1 that was immunoprecipitated from tachyzoites. In contrast, Brumlik and colleagues used a bacterially expressed construct that only contained TgMAPK1' s catalytic domain and most of this protein was expressed in bacterial inclusion bodies (our unpublished observations) [24]. Even though we could purify the protein from the inclusion bodies under denaturing conditions, very weak kinase activity was detected after refolding and we could not verify whether this protein was properly folded. Attempts to purify the soluble protein resulted in a preparation of poor purity (as noted by [24]) whose activity appeared to be due to a contaminating bacterial kinase since recombinant expression of a catalytically inactive kinase mutant had the same rate of activity as the wild-type kinase (not shown). Recently, Sugi et al demonstrated that the SBR2 (Leu162→Gln) and SBR3 (Ile171→Asn) alleles conferred resistance to the effect of the bumped kinase inhibitor 1NM-PP1 on parasite replication [25]. TgMAPK1 resistance to 1NM-PP1 is independent of its well-described inhibition of Toxoplasma Calcium-Dependent Protein Kinase 1, whose activity is required for parasite invasion and egress [31]. xxx But unlike our work, Sugi et al. did not test whether 1NM-PP1 affects parasite replication by directly inhibiting TgMAPK1 or if resistance emerged because TgMAPK1 functions downstream of the protein (s) that 1NM-PP1 interacts with [25], [32]. It is also noteworthy that we were unable to successfully express a kinase-dead mutant of TgMAPK1 as a transgene either in transiently or stably transfected parasites. The most likely explanation for this is that expression of the kinase dead TgMAPK1 mutant has a dominant negative effect either by binding its upstream activating kinase or its downstream targets. The continuing emergence of antimicrobial resistance requires novel approaches to the design of new drugs and treatments. Although simultaneous inhibition of both host and parasite targets is an untested approach, one benefit would be that resistance to a compound that impacts a host cell target is less likely to develop [33]. As a proof of concept, we showed that SB505124 reduced tachyzoite burdens 5 days after mice were intraperitoneally infected. We did not examine later time points for three reasons. First, we demonstrated that SB505124 will induce bradyzoite development in vitro and therefore it is possible that long-term treatments with this drug would lead to increased cyst burden. Second, germline loss of functions mutations in HIF-1α causes an embryonic lethal phenotype in mice [34]. Thus, HIF-1α can only be deleted in specific cell types using cre-recombinase expressing mice. But, Toxoplasma infects and forms bradyzoites in multiple types of cells. Since tissues consist of heterogeneous populations of cells it would, therefore, be difficult to assess how a cell-specific loss of HIF-1α would impact bradyzoite development. Third, ALK5 is the TGFβ receptor and long-term inhibition of this key immunosuppressive cytokine could have a deleterious effect on mice independent of its role in regulating parasite growth. Parasites treated with SB505124 developed into bradyzoites under in vitro conditions. Given that the drug induces RH parasites to become multi-nucleated (suggesting a role for TgMAPK1 in cell cycle), these data are consistent with earlier observations that bradyzoite development represents an exit from the cell cycle during the transition between S to M phase (Toxoplasma lacks a G2 phase) [35], [36]. One implication of bradyzoite development being potentially triggered by TgMAPK1 inhibition is that it may open the door to two novel but not necessarily mutually exclusive approaches for treating toxoplasmosis. First, we hypothesize that compounds that either act as TgMAPK1 agonists or mimic the activities of TgMAPK1 substrates would be predicted to act as inhibitors of bradyzoite development. Because bradyzoites are impervious to both immune surveillance mechanisms and anti-parasitic compounds [3]–[8], these small molecules would maintain the parasite as tachyzoites, which is a growth state that would prolong their susceptibility to currently prescribed drugs. Second we will focus on identifying a compound (s) that specifically blocks parasites from activating ALK4,5, 7 or prevent these receptors from triggering HIF-1 activity. We believe that either of these approaches would be valid even if SB505124 induces bradyzoite development as a consequence of the drug activating a more generalized stress response. Our future analysis of TgMAPK1 regulation and substrate identification will resolve how TgMAPK1 influences bradyzoite differentiation. In summary, we used a forward genetic screen to demonstrate that a serine/threonine kinase inhibitor blocks Toxoplasma growth through two distinct targets and our future work will focus on this issue. Kinase inhibitors are particularly useful for dual-target screens because even though they are designed to inhibit human kinases their off target effects cannot be predicted [37]. For example, we showed that SB203580 does not inhibit TgMAPK1, which bears homology to p38 MAPKs. Rather TgMAPK1 is inhibited by SB505124 even though this compound does not appear to significantly affect p38 MAPK activity [22]. Our work, therefore, also highlights the utility of repurposing drugs and investigative compounds originally developed to target cancer and other diseases for the study and/or treatment of microbial infections [38]–[41]. Animal protocols (IACUC Protocol #11-075I) were approved by the University of Oklahoma Health Sciences Center IACUC. This study was carried out in strict accordance with the Public Health Service Policy on Humane Care and Use of Laboratory Animals and AAALAC accreditation guidelines. Toxoplasma RH, RHΔHXGPRT, RHΔ-GFP (kindly provided by Dr. Gustavo Arrizabalaga of Indiana University), RHΔ RHΔHXGPRTΔKu80 (kindly provided by Dr. David Bzik of Dartmouth University), GT-1, Pru, and CTG parasites were maintained in human foreskin fibroblasts in Dulbecco' s Modification of Eagle' s Medium (DMEM) (Mediatech; Manassas, VA) supplemented with 10% fetal bovine serum (Mediatech), 2 mM L-glutamine (Mediatech), and 100 IU/ml penicillin – 100 µg/ml streptomycin (Mediatech). The TgMAPK1ts-HA temperature sensitive (ts) TgMAPK1 mutant will be described in detail elsewhere (manuscript in preparation). Other host cell lines were maintained in the HFF growth medium. All host cell lines and parasites were routinely tested for Mycoplasma contamination with the MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland) and found to be negative. Intracellular RHΔHXGPRT tachyzoites grown in T-75 flasks were mutagenized with 300 µg/ml ENU (Sigma; St. Louis, MO) in complete DMEM for 2 hours at 37°C as previously described [42], [43]. After ENU treatment, media was removed and the monolayer was washed four times with 1× phosphate-buffered saline. Mutagenized parasites were then released from host cells by scraping and syringe-lysis (27 g needle) and washed with 30 ml DMEM. Parasites were allowed to recover in fresh HFFs for 72 h and then grown in the presence of SB505124 (TGF-βRI Kinase Inhibitor III (EMD Millipore; Billerica, MA) ). After several rounds of selection in SB505124, individual clones were obtained by limiting dilution in 96 well plates of HFFs. SBR1 was isolated from a population after 4 passages in 3 µM SB505124 and 11 passages in 5 µM SB505124. SBR2 was isolated after 4 passages in 5 µM SB505124. SBR3 was isolated after 5 passages in 5 µM SB505124. Plaque assays were performed by adding 200 parasites to each well of a 24 well plate containing confluent HFFs. After 5–7 days, the monolayers were methanol-fixed and stained with 0. 1% crystal violet. Plaques were counted using 4× magnification on an inverted dissecting microscope. Uracil incorporation assays were performed by growing parasites in 24 well plates. After 66 h, 5 µCi 3H-uracil (MP Biomedical; Santa Ana, CA) was added to the wells for an additional 6 h of the assay. Wells were washed, precipitated with 10% trichloroacetic acid, and 3H-uracil counted by liquid scintillation. Parasites were grown in confluent HFFs until the host monolayer had completely lysed. The media containing the extracellular parasites were collected without further scraping, washed with serum-free DMEM, passed three times through a 27 g needle, and then filtered through a 3 µm pore size filter membrane. Genomic DNA was isolated using the Qiamp DNA mini kit (Qiagen; Valencia, CA) using the manufacturer' s protocol for cultured cells. qRT-PCR (not shown) determined that host DNA contamination was less than 0. 5% and this result was confirmed by the whole genome sequencing data. Sequencing libraries were prepared from genomic DNA (1 µg) using the Truseq DNA LT Sample Prep Kit v2 as per the manufacturer' s protocols (Illumina, San Diego, CA) and the libraries sequenced on an Illumina HiSeq 2000 instrument with 100 bp, paired-end reads. Between 30–37-fold coverage was achieved for all genomes. Genome sequences were aligned with the MOSAIK program [44], [45] using the type I parasite genome of GT-1 as a reference and single nucleotide variants identified with variant caller program FreeBayes [46]. For MOSAIK, the default parameters we used had a hash size of 14. For FreeBayes, we used the program' s default parameters with the exception that: ploidy (-p) was set to 1, a P value cutoff of 0. 9 (-P) was used, complex events were ignored (-u), and population priors were turned off (-no-population-priors) as they are not applicable to the Toxoplasma genome. A 944 bp fragment of TgMAPK1 was PCR amplified using Platinum Pfx polymerase (Invitrogen; Carlsbad, CA) from wild-type and SBR mutant parasites with forward (5′-TGCATGGCGATGAGTTTCTGAACG-3′) and reverse (5′- TCGTGTCGACGTTTCTTCTGTGGA-3′) primers. PCR reactions were incubated with Taq polymerase (Invitrogen) for 10′ to add 3′ A-overhangs. PCR products were gel purified and cloned into pCR2. 1 (Invitrogen) by TA ligation. Inserts were sequenced verified by conventional sequencing. Parasites were transfected with 50 µg of linearized plasmid resuspended in cytomix buffer (2 mM ATP, 5 mM glutathione). For each transfection, 2×107 RHΔHXGPRTΔKu80 tachyzoites were washed in cytomix buffer and resuspended in 0. 5 ml complete cytomix buffer. The plasmid DNA was added to the parasites in a BTX electroporation cuvette (0. 4 mm gap), and electroporated into the parasites at 2000 V, 50 ohm, 25 µF with a BTX ECM 360 (Holliston, MA). Parasites were then transferred to 75 cm2 flasks containing confluent HFF monolayers and after 72 h 3 µM SB505124 was added. Parasites receiving pCR2. 1: TgMAPK1WT failed to grow whereas those that received pCR2. 1: TgMAPK1SBR1-3 were able to be passaged indefinitely in 3 µM SB505124. Individual clones were isolated by limiting dilution as described above and single clones were named RH: MAPK1SBR1, RH: MAPK1SBR2, and RH: MAPK1SBR3. To ensure that the SBR mutation was incorporated into the endogenous MAPK1 locus, PCR fragments were generated from genomic DNA using primers that flanked the region of the amplicon. The amplicon was then gel purified and cloned into PCR2. 1. At least 5 independent transformants were sequenced via conventional Sanger sequencing using M13 forward and reverse universal primers. RHΔ or SBR1 tachyzoites were added to 24 well plates containing confluent HFFs on glass coverslips in the presence or absence of 3 µM SB505124. After 24 h, coverslips were methanol-fixed, blocked with 3% bovine serum albumin, and labeled with rabbit anti-SAG1 (1∶100000; kindly provided by Dr. John Boothroyd, Stanford University) followed by Alexa Fluor 488-conjugated goat anti-rabbit (Invitrogen). Coverslips were mounted to slides with Vectashield containing DAPI (Vector Lab; Burlingame, CA). One hundred randomly chosen vacuoles/coverslip were counted. Phospho-STAT3 was detected in HFFs infected with RH-GFP (MOI of 10) tachyzoites for 6 h in the absence or presence of 5 µM SB505124. Cells were fixed with 4% formaldehyde, permeabilized with ice-cold methanol for 5 minutes, blocked at room temperature for 2 h with 3% BSA, and incubated with rabbit anti-phosphoSTAT3 (Cell Signaling; Danvers, MA) overnight at 4°C. After washing, cells were stained with Alexa Fluor 594 conjugated goat anti-rabbit (Invitrogen) and mounted to slides with Vectashield containing DAPI. Bradyzoite cysts were detected in HFFs 72 h after adding either 3 µM SB505124 or pH 8. 1 media [47]. Monolayers were fixed with ice-cold methanol for 5′ and then stained to detect SAG1 as described above. Bradyzoite cyst wall was detected by staining with 5 µg/ml FITC-conjugated Dolichos (Vector Labs). The frequency of Dolichos+ vacuoles was quantified by examining a minimum of one thousand vacuoles per strain from three independent experiments using a Cytation 3 (Biotek Instruments, Inc. , Winooski, VT) cell imaging multi-mode reader at 20× magnification. To generate strains in which TgMAPK1 is epitope tagged at its C-terminus with three HA repeats (RHΔHXGPRTΔKu80: TgMAPK1-3XHA), a 1513 bp fragment was amplified from the 3′ end of TgMAPK1 using primers TgMAPK1-3XHAF and TgMAPK1-3XHAR and cloned into p3xHA-LIC-HXGPRT by ligation independent cloning to create pTgMAPK1-3XHA [48]. RHΔHXGPRTΔKu80 parasites were transfected with PacI-linearized TgMAPK1 plasmids and transfectants selected with mycophenolic acid/xanthine. Proper integration and expression was by PCR (not shown), Western blotting (Figure 5B), and immunofluorescence (E. Suvorova et al. , Manuscript in Preparation). RHΔHXGPRTΔKu80: TgMAPK1WT-HA, RHΔHXGPRTΔKu80: TgMAPK1ts-HA, or wild-type RHΔHXGPRTΔKu80 were grown overnight in confluent 75 cm2 flasks. Extracellular parasites were washed away and intracellular parasites were released by scraping and syringe lysis. The parasites were washed and lysed on ice with 200 µl modified RIPA buffer (50 mM Tris pH 7. 4,1% NP-40; 0. 1% SDS, 500 mM NaCl) supplemented with 1× EDTA-free Protease Inhibitor Cocktail (Roche; Indianapolis, IN) per 107 parasites. Lysates were centrifuged at 20,000×g for 15′ at 4°C to remove insoluble material and the supernatant was precleared with rabbit IgG-conjugated sepharose beads (Cell Signaling). Precleared lysates were added to 10 µl anti-HA-sepharose beads (Cell Signaling) and incubated overnight at 4°C, washed 3 times in modified RIPA buffer, and 1 time in kinase assay buffer (20 mM HEPES pH 7. 48,25 mM glycerophosphate, 25 mM MgCl2,0. 5 mM DTT, and 0. 1 mM Na3VO4). Immune complexes were then evenly distributed between sample tubes (the equivalent of 200 µg of precleared lysate was used for each sample) and incubated with 10 µCi γ32P-ATP (MP Biomedicals) in kinase assay buffer at 34°C for 1 hour. Kinase reactions were stopped by adding 5 µl 6× sample buffer and boiling the samples for 5′. Reactions were separated by SDS-PAGE (10% acrylamide) and gels were fixed for 20′ in acetic acid∶methanol∶H20 (1∶5∶4) and dried. The gels were exposed to film and analyzed using ImageJ software. We attempted to express kinase-dead TgMAPK1 (K140R) and SBR mutants as a transgenes by synthesizing full length wild-type and mutant constructs in frame with C-terminal 3× HA-tags (Genescript; Piscataway, NJ). The cDNA constructs were cloned into pSAGCAT by replacing the CAT gene with the TgMAPK1 gene [49]. The constructs were then transfected into RH strain parasites and 24 h later the parasites were harvested and transgenic TgMAPK1 expression assessed by Western blotting and activity by in vitro kinase reaction. We were only able to detect low levels of mutant TgMAPK1 protein even when 175 cm2 flasks were used for the transfection and this amount of kinase was not sufficient for the autokinase reactions (not shown). In contrast, 25 cm2 flasks were sufficient to detect and assay the wild-type kinase. We next cloned the TgMAPK1 alleles into ptubYFPYFP/sagCAT [48] by replacing the YFP genes with the kinase alleles. Using chloramphenicol selection, stable transformants could be isolated for those parasites receiving wild-type TgMAPK1 but we could not obtain transformants for parasites receiving mutant alleles. We also attempted to epitope tag the SBR mutants at their endogenous locus by cloning the 1. 2 kb 3′ most region (up to the stop codon) of the genomic region of the TgMAPK1 into pSF-TAP-LIC-HXG in frame with the SF-TAP tag [48]. This single construct is compatible with both the wild type and SBR strains since the 3′ region of homology is downstream of the SBR mutations. We linearized the vector with NcoI, a unique restriction site within the TgMAPK1 sequence giving 677 bp of homology to the endogenous TgMAPK1 allele, and transfected the DNA into RHΔKu80ΔHXGPRT: TgMAPK1SBR1 and the parental wild type strain. But, we were unable to recover viable SBR parasites from selection with MPA/xanthine even though we were able to do so for the wild-type parasites. Proteins were separated by SDS-PAGE and transferred to PVDF membranes. Membranes were blocked with 5% BSA-TBST. Rat anti-HA (Roche) was used at a concentration of 1∶500. Goat anti-rat HRP (Cell Signaling) was used at a concentration of 1∶2000 and detected using Supersignal West Pico ECL reagent (Thermo, Waltham, MA) with a gel doc system. RNA was isolated with the RNeasy Midi kit (Qiagen) using the manufacturer' s instructions for cultured cells. RNA quality and yield was assessed by UV spectrophotometry and horizontal electrophoresis. Total RNA was treated with DNAse I (Invitrogen) and converted to cDNA using Superscript III (Invitrogen) using random hexamer primers. Relative abundances of bradyzoite gene transcripts BAG1, LDH2, ENO1 and tachyzoite transcript ENO2 were calculated with the 2−ΔΔCt method [50] using Toxoplasma β-actin as an internal control (see Supplemental Table S2 for primer sequences). HIF-1 luciferase assays were performed as previously described [18]. Briefly, murine embryonic fibroblasts were transfected pHRE-luc and pTK-Rel (as a transfection efficiency control) and 24 h later 2×106 parasites were added to the wells in the presence or absence of SB505124. 18 hours later, luciferase activity was measured using the Dual-Glo luciferase assay kit (Promega, Madison, WI). C57Bl/6 female mice were infected intraperitoneally with 103 freshly purified RH-GFP parasites in a volume of 200 µl phenol-red-free DMEM. Beginning 1 h post infection, mice received 10 mg/kg SB505124 or DMSO alone intraperitoneally once daily. The drug had no apparent effect on the health of the animals during this time course. On day 5 post-infection, mice were sacrificed by CO2 asphyxiation and PECs collected in a 5 ml PBS-lavage. Cells were washed, stained with anti-mouse APC-conjugated CD45, and analyzed by flow cytometry using a FACSCalibur cytometer (BD Biosciences, San Jose, CA). 105 events were counted and subsequent analysis was performed using Summit (Dako, Carpinteria, CA). IFNγ was measured using Mouse IFNγ ‘Femto-HS’ High Sensitivity ELISA (eBioscience; San Diego, CA).

Understanding how a compound blocks growth of an intracellular pathogen is important not only for developing these compounds into drugs that can be prescribed to patients, but also because these data will likely provide novel insight into the biology of these pathogens. Forward genetic screens are one established approach towards defining these mechanisms. But performing these screens with intracellular parasites has been limited not only because of technical limitations but also because the compounds may have off-target effects in either the host or parasite. Here, we report the first compound that kills a pathogen by simultaneously inhibiting distinct host- and parasite-encoded targets. Because developing drug resistance simultaneously to two targets is less likely, this work may highlight a new approach to antimicrobial drug discovery.

Abstract Introduction Results Discussion Materials and Methods

signal transduction medicine and health sciences molecular cell biology pathology and laboratory medicine cell biology host-pathogen interactions genetic screens medical microbiology gene identification and analysis gene expression genetics biology and life sciences microbiology pathogenesis parasitology

2014

Forward Genetic Screening Identifies a Small Molecule That Blocks Toxoplasma gondii Growth by Inhibiting Both Host- and Parasite-Encoded Kinases

We introduce a series of experimental procedures enabling sensitive calcium monitoring in T cell populations by confocal video-microscopy. Tracking and post-acquisition analysis was performed using Methods for Automated and Accurate Analysis of Cell Signals (MAAACS), a fully customized program that associates a high throughput tracking algorithm, an intuitive reconnection routine and a statistical platform to provide, at a glance, the calcium barcode of a population of individual T-cells. Combined with a sensitive calcium probe, this method allowed us to unravel the heterogeneity in shape and intensity of the calcium response in T cell populations and especially in naive T cells, which display intracellular calcium oscillations upon stimulation by antigen presenting cells. Calcium ion acts as a universal second messenger in response to most cellular stimuli [1]. The pattern of the calcium response is biphasic, and primarily results from the production of inositol-3 phosphate (IP3) which triggers the release of calcium from the endoplasmic reticulum (ER store release) into the cytoplasm. This decrease is sensed by the stromal interaction molecules (STIM) that secondarily trigger the capacitative entry of extracellular calcium via the calcium release activated channels (CRAC) of the ORAI family [2]–[4]. Measuring the intracellular concentration of calcium is therefore of primary interest when analyzing transduction processes in living cells. Currently, this is achieved by methods which combine flow cytometry with intracellular diffusive fluorescent calcium-sensitive dyes in immunological relevant cells such as macrophages, NK cells, T or B cells. As an example, the calcium response is routinely monitored in T cells [5]–[15] as a functional read-out of the outside-in signal transduction subsequent to T-cell receptor (TCR) engagement by major histocompatibility complex (MHC) molecules with agonist peptide. However, when naive T cells encounter antigen-presenting cells (APC) and more generally when signaling is induced by intimate signaling-to-target cell-cell contact, flow cytometry approaches cannot fully recapitulate the physiological conditions of stimulation. In addition, recent works have demonstrated that TCR triggering by the MHC molecules follows unusual physico-kinetic parameters of serial engagement-disengagement [16], [17], which could be the molecular basis for the broad selectivity, high specificity, extreme sensitivity [18] and the capacity to induce a rapid intracellular response that characterize TCR triggering [19]. While soluble anti TCR or anti CD3 antibodies [20], antibody coated beads [21], [22], and phorbol myristate acetate/ionomycin [23] can all induce a productive calcium signal in T cells that ultimately leads to their activation, proliferation and cytokine production, the calcium elevation triggered by these strong irreversible stimuli is usually sustained. It may not therefore be representative of the response to physiological stimulations, which is more likely to consist in calcium spikes and oscillations [9], [24]–[26]. In order to capture the true calcium responses triggered during cell-cell contacts such as those occurring during T-cell and APC stimulation, video-imaging is compulsory in that it provides informative parameters on individual cell behavior (i. e. displacements, shape and intensity fluctuation) [27]. Obtaining such imaging data requires a complex custom-built experimental set-up usually dedicated to the detection of UV-excitable calcium probes and to the maintenance of physiological parameters for long-term recordings [9], [25], [28]. In any case cell tracking is mandatory and is often performed by manual approaches [28]–[30]; however, in addition to being time-consuming, manual analysis is prone to systematic errors due to subjective choice. Such pre-selection is an unavoidable step in any manual analysis. Automating the simultaneous tracking of hundreds of cells over hundreds of time frames would overcome these issues. Nevertheless, simultaneously tracking moving cells at high density represents a considerable challenge, particularly considering the need to correctly resolve interlaced tracks of stretching cells while providing valuable statistical confidence and robustness. While many software packages do incorporate a cell tracking module or plugins, the normalization of the calcium signal for each cell as well as the classification of calcium responses and any quantification generally have to be performed manually involving tedious excel datasheets [31]. Here we have developed a complete approach named Methods for Automated and Accurate Analysis of Cell Signals (MAAACS) which enables the simultaneous tracking of a population of individual moving cells (multiple target tracking, MTT) [32] and the automatic extraction of robust statistics on pertinent observables. The MAAACS program has been conditioned to synchronize, normalize and assemble the recorded cell traces to provide an at a glance calcium barcoding of a heterogeneous cell population and facilitate the a posteriori data mining and interpretation. We used MAAACS to examine the calcium responses induced in T cells upon interaction with APCs and with it were able to reveal the oscillatory calcium responses in mouse naive CD4+ T cells upon antigen recognition. Aiming to establish an easy, robust, sensitive and reliable way of evaluating calcium fluctuations in T cells, we assessed many visible calcium indicators such as Fluo-4 AM, Fluo-3 AM, and Fluo-8. All displayed short term leakiness of the loading without membrane extruder blockers [33] (such as probenecid) [34] and subsequent intracellular compartmentalization, incompatible with unbiased calcium measurements [35]. In contrast, T-cell hybridomas (3A9) loaded with the calcium indicator BD PBX appeared to overcome most of these problems [36]. Compared to standard loading conditions, this procedure provided a stable loading of the fluorescent indicator (emission spectrum fully stackable with Fluo-4 AM; Figure 1A), without affecting cell viability (Figure S1A). We also documented that BD PBX fluorescence was photostable upon repetitive confocal illumination for 30 min, unlike Fura-red the fluorescence of which rapidly decreased (Figure S1B). This precluded us from performing BD PBX/Fura red ratiometric cytosolic calcium measurements [37], [38]. In addition, we investigated whether cytosolic calcium gradients could be visualized under such experimental conditions since few discrete hotspots were detectable among the homogenous fluorescence. BD PBX and mitotracker red loaded cells were imaged to decipher whether mitochondria would accumulate calcium indicator. In 3A9 T cells the two signals were not mutually exclusive (Figure S1. C, D) unlike in Jurkat cells or primary human T cells [35], [39]. Part of the signals was correlated under stimulation, most presumably caused by FRET between the two dyes (Figure S1. C, D). We detected no significant sequestration of the BD PBX calcium dye, unlike Fura-2 in Jurkat cells although this phenomenon had a limited impact on whole cell calcium measurement [35]. These inconsistencies with previous reports [34] motivated us to consider BD PBX as a close relative of Fluo-4 AM but harboring subtle differences, that required the dedicated loading buffer to avoid the dye leakage displayed by Fluo-4 AM (Figure S1A). Due to the lack of information about the BD PBX from the manufacturer, we thus determined the in vitro Kd of the BD PBX as described previously [40] (Figure 1B), which gave a consistent value for Kd of 312 nM±33, comparable to the Kd for Fluo-4 AM of 327±49 nM and that in the literature (Kd = 345 nM) [41] (see Materials and Methods). The great similarity between BD PBX and Fluo-4 AM led us to assume the in situ Kd of BD PBX to equal the reported in situ Kd value (1 µM) of Fluo-4 AM [34], deemed acceptable when accurate determination by electrophysiology is either not feasible or not available [28]. Based on this Kd value, we estimated that the intracellular calcium concentration in resting 3A9 T cell hybridomas would be around 90 nM and consistent with previously published values for these hybridomas [42] as well as leukemic cell lines Jurkat [39], mouse thymocytes [43] and human peripheral blood lymphocytes [35]. We used flow cytometry to compare the detection sensitivity of BD PBX with the ratiometric Indo-1 AM (routinely used in calcium assays, UV excitable). In terms of response pattern, T cell hybridomas loaded with BD PBX do not strictly speaking respond the same as those loaded with Indo-1 AM, as the non-ratiometric calcium indicator BD PBX does not allow the evaluation of intracellular free calcium concentration (Figure 1C). We therefore sought to estimate the intracellular calcium concentrations in BD PBX loaded T cell hybridomas. Upon various concentrations of ionomycin, we compared intracellular calcium elevation in BD PBX versus Indo-1 AM loaded 3A9 T cell hybridomas. As previously mentioned, the kinetics were not fully stackable and the calculated intracellular calcium concentration differed between the two methods due to Kd discrepancy and loss of linearity in the relationship linking calcium concentrations and high fluorescence amplitudes [40]. Indeed at such high fluorescence values under strong ionomycin concentrations, the calcium concentration was overestimated and is the reason for us reporting fluorescence amplitude instead of erroneous calcium concentrations throughout this manuscript. Nevertheless to our surprise, lower concentrations of ionomycin rapidly abrogated fluorescence elevations of indo-1 AM whereas BD PBX fluorescence elevation remained detectable even at subnanomolar ionomycin concentrations. This indicated that BD PBX is sensitive to low intracellular calcium elevation. Thapsigargin (Figure 1D) or the cross-linking of the TCR/CD3 complex by anti-CD3ε (2C11 biotin/streptavidin) induced in cells loaded with BD PBX responses similar to those induced by ionomycin. This method was extendable to naive primary CD4+ T cells labeled or not with an anti CD4 monoclonal antibody (Figure 1E). While surface receptor crosslinking with antibodies is a convenient way of stimulating calcium responses in T cells, it cannot physiologically reproduce the dynamics of TCR/MHC-antigen interactions in the context of T-cell/APC contacts. Our goal was to decipher calcium signals arising from cellular contacts and requiring imaging approaches. We chose to perform all recordings on a conventional argon laser equipped-confocal scanning microscope, widely found in laboratories. Considering the lack of available methods [44] able to combine automatic tracking of cells and calcium signal processing, we set up our own procedure to automatically track moving cells at high densities with a minimum of input parameters. It is a customized version of our previously developed MTT algorithm [32], originally dedicated to tracking single fluorophores coupled to plasma membrane molecules at high density. We built a plugin that converts cellular images into cell position images that are comparable with the single molecule images supported by MTT (Figure 2A & Figure S2). Raw fluorescence images were first passed through a median filter to eliminate the electronic noise emanating from detection and an appropriate mask, consisting of a disk of adequate radius (see Materials and Methods for details on mask size), was applied to identify cells as single objects. Each cell was thus defined by the xy coordinates of its centroid which then served to reconstruct the cell' s trajectory over the stack of images (Figure 2A & Videos S1, S2, S3, S4, S5, S6, S7). Noteworthy, more complex cellular shapes could be handled if using other appropriate detection schemes. Next, we generated synthetic images containing Gaussian peaks at the corresponding positions, with a radius optimized for the tracking performed by MTT and an intensity equal to the integrated cell signal, itself proportional to the intracellular calcium concentration. The resulting sequence of single molecule like images could then be treated by MTT (Videos S5, S6). Replacing each cell by a Gaussian peak of smaller size prevented the occurrence of two targets crossing over each other, initially a major concern for MTT, thus rendering the reconnection of traces during the MTT procedure far more efficient. Overlapping was then handled at the detection stage, where crossing cells presenting as a “peanut” shape were detected as two targets. However, strongly overlapping cells resulting in more of a spherical shape were detected as a single target and thus required z-stack acquisitions and appropriate analysis to recognize the occurrence of such crossing trajectories. MAAACS generates traces which are defined by the position of the detected cells over time and intrinsically calculates their instantaneous velocity and their fluorescence intensity. Given that the basal level varies from cell to cell due to differences in intracellular calcium concentrations, or due to heterogeneous efficiency in calcium indicator loading, we needed to accurately set a baseline of fluorescence that we defined as the median of fluorescence calculated until the maximum fluorescence value had been reached for each cell (Figure 2B). To establish the best mode of normalization, we analyzed for each cell in non-stimulating conditions how the mean signal amplitude was correlated with signal fluctuations. We found that this relationship was proportional, thus implying that the normalization can be performed by division. Calcium responses are highly diverse, both in terms of magnitude and oscillation (Figure S3). The amplitude of calcium mobilization varies according to the type of stimulus and the addition of inhibitors, but also within a cell population for a given stimulus/treatment. Moreover, the shape of these signals, their maintenance and their oscillations are also varied. We therefore defined analytical parameters to describe and characterize these response diversities (Figure 2C). For each cell, the response magnitude is described as the fluorescence amplitude (FA) of calcium mobilization, corresponding to the time-average of normalized intensities on the whole trace. The temporal fluctuations are deciphered by analyzing the persistence and the oscillations of the calcium signals. We defined as the response fraction (RF) the ratio of two phases: the time when the normalized intensity is greater than the threshold over the total time during which the intensity is detected. We also calculated the number of bursts/min (BPM) defining the number of peaks detected above the threshold divided by the duration of the detected trace (Table S1). In order to provide a global, comprehensive view of the calcium response in a substantial number of cells for any given condition, we color coded the normalized calcium intensities with a gradient of blue and orange for values below or above the threshold (see below), respectively. The resulting values, for each cell at each time-point, could then be pooled to generate a heat map, the dimensions of which hence corresponded to the cell number and time (Figure 2C left panel). Non-relevant pixels, either before or after detection of a given cell in the time-lapse movie, were left in black. This representation simultaneously depicts the global tendency, together with the intra-population variability of response [45]. Collectively, all calcium signal parameters are summarized on scatter dot plots where responding cells are represented by a single dot (Figure 2C right panel). By integrating these different parameters, the heterogeneous behavior of activated cells (maintained, oscillatory and unique) can be determined and inactive cells identified, the proportions of which are then represented in a pie chart diagram (see Materials and Methods for details on the classification). An endemic problem in automatic tracking approaches is reaching a level of completeness that manual tracking only can guarantee. We analyzed several videos in parallel by the automatic and by manual methods and determined the percentage of cells tracked by MAAACS compared to that by manual tracking in the observation field (detection percentage). 3A9 T cell hybridomas or naive CD4+ T-cells loaded with BD PBX were seeded onto a monolayer of COS-7 cells stably expressing the molecules of the major histocompatibility complex [46]. We chose experimental situations with high cell densities on a rough and irregular surface. We only considered tracked cells detected by MAAACS over more than 5 images. The superimposition of the trajectories obtained manually or with MAAACS (Figure 3A) illustrated the efficiency of our algorithm, the detection percentage of which was greater than 96% (N = 125). Surprisingly, more traces were generated by MAAACS than obtained manually. Consequently, the MAAACS cell traces were fragmented into several parts as shown in Figure 3B illustrated by the number of fragments needed to reconstitute the full length trajectories (1. 3 for primary T cell and 1. 9 for hybridomas, Figure 3B). The lower efficiency for the latter can be explained by significant cellular shape changes over time preventing their detection by the circular mask and thus their reconnection. To fully document the cell response over time and to improve the quantitative analysis of cell signaling, we implemented a program allowing the reconnection of fragments of trajectories (Figure 3C). First, the method selected among the set of trajectories terminating before the end of the acquisition (plotted in gray in Figure 3C) to be reconnected to traces starting after the breaking off (plotted in red and blue in Figure 3C). Then the algorithm classified the candidates for reconnection by minimizing the interval between the stop and start times (Δt), the distance between the final and initial positions (Δr) and the difference of the mean fluorescence amplitude of the fragments (ΔI). The user is free to decide whether the trajectories should be reconnected by consulting the original video. In this way, the tracked time percentage was clearly improved (95% for primary T cells and 83% for hybridomas, Figure 3D). Within the course of this study, we noticed that without specific stimulation, cytoplasmic calcium concentrations displayed spontaneous oscillations, the amplitude of which was almost negligible for T cell hybridomas but not for primary naive T cells [13]. To account for this, we performed a median filtering with a sliding window of 7 frame-size on each fluorescence amplitude to remove aspecific oscillations in the absence of any stimulus. It was then important to carefully define the threshold of peptide-specific activation. As mentioned by many authors [26], [42], [47] and in our observations (Figure 3), calcium signaling exhibits high diversity even within the same cell line and depends on the applied perturbations (stimulation, drug treatment or mutation). It should be noted that defining any biological threshold for activation could be misleading since it is highly dependent upon the experimental conditions. Moreover, the definition of a criterion for specific activation should respect the response heterogeneity without favoring a subset of responding cells. Accordingly, we set up a detector which compares the fluorescence amplitude to a threshold of activation in order to identify activated cells in our experiments (Figure S4). If the fluorescence amplitude is greater than the threshold, then the cell is declared as activated. To determine this threshold, we investigated the statistical properties of the fluorescence amplitude of cells in the absence of any stimulation compared to that in cells that have been activated as a result of stimulation. For a given probability of false alarm (PFA), an activation threshold could be deduced from the cell responses in the absence of any stimulation. The probability of detection (PD) could then be calculated as the percentage of activated cells revealed by the detector. We then aimed to identify threshold values that minimized false detections (low PFA) without decreasing the identification of activating cell (high PD). A robust method to objectively determine the activation threshold is to perform a receiver operating characteristic analysis (ROC curves in Figure S4A, B) to explore the values of activation threshold that maximize the overall score PD x (1-PFA). All activation thresholds are reported in Figure S4C. Surprisingly, the activation threshold was quite stable whatever the cell type (hybridoma or naive T cells) or stimulation process (antibody or APC). For T-cell hybridomas, this method exhibited very high PD (>0. 99) and low PFA (<0. 02). PD was also high in primary T cells (>0. 98) and the PFA was reasonable (<0. 08) though slightly higher due to a higher diversity in the cell signaling. To test our MAAACS algorithm, we analyzed a population of T hybridomas 3A9 loaded with BD PBX and seeded at the bottom of a well of a Lab-Tek before their stimulation with thapsigargin after 1 minute. We generated a sequence with a frequency of 1 confocal image every 7 seconds for 30 minutes. Fluorescence intensity was not affected by repetitive illuminations as previously mentioned (Figure S1B). No a priori assumption was made and raw recordings were subjected to MAAACS. To compare the calcium signals, we normalized the fluorescence intensities as a fold of the basal fluorescence. A mean curve of variations in fluorescence over time was obtained (Figure 4C) and compared to flow cytometry measurements (Figure 4B). Fluorescence rose immediately after thapsigargin addition, reaching a plateau 3 minutes after induction [48] before slowly decreasing (Figure 4C). This response was fully stackable with the kinetics monitored by flow cytometry (Figure 4B). We noted that the response/baseline ratio was higher according to imaging recordings, presumably due to a better sensitivity of detection on confocal photomultipliers. In this case the benefit of MAAACS is limited since all individual cells responded homogenously by a strong, sustained non-oscillating response (fluorescence amplitude, FA = 7. 93±0. 50; response fraction, RF = 0. 91±0. 02; bursts per minute, BPM = 0. 04±0. 001). In the presence of the CRAC channel blocker 2-aminoethoxydiphenyl borate (2-APB) [49], thapsigargin induced a weak elevation of fluorescence that was similar to Ca2+/Mg2+ deficient incubation conditions [48] (Figure 4A) and consistent between the two methods (Figure 4B, C). The analysis of the global tendency clearly masked the singularity of each individual cell. Through MAAACS analysis, while 50% of the cells displayed a unique rise of fluorescence as the average tendency would have suggested, the other half was equally composed of cells displaying sustained or oscillatory behavior. This implies that within a cell population and for any given stimulus, the observed differences in response are not limited to intensity dispersion but also to the mode of response. We sought to document this point using different experimental stimuli each producing their own calcium response in terms of shape, intensity and heterogeneity (Figure 5) [50]. Indeed, when the cells were seeded onto an activating surface (anti TCR antibody coated pits), while most responses were sustained (FA = 4. 4±0. 43, RF = 0. 62±0. 03, BPM = 0. 09±0. 01), heterogeneous responses were also observed. These heterogeneities were even more obvious when the 3A9 cells were stimulated by I-AK-HEL expressing COS-7 APCs. The asynchronous landing of the T cells and the heterogeneous MHC II agonist peptide expression levels are parameters that affect the calcium response in addition to irregular crawling and scanning activities of the T cells on an APC monolayer. Indeed, during the first 30 minutes, 50% of the cells displayed a specific calcium rise [15]. Most of these exhibited a maintained fluorescence amplification but which was weaker in term of intensity and response fraction as compared to that in response to the stimulating antibody (Figure 5C), supporting the notion that abundant, immobile, highly affine ligands are not strictly recapitulating the stimulation by the natural membrane ligand of the TCR. Supporting this view, we analyzed the cell motility by MAAACS, as an integral parameter of T cell activation [29], [42], [51]. MAAACS analysis of cell velocity showed that inducers of strong and sustained calcium responses (thapsigargin or anti TCR coated slides and to a lesser extent anti CD45 unstimulating surfaces) negatively impacted the motility of cells [12], since instant speed measurements did not exceed 2 µm/min in the few minutes after landing on the slide, indicating that the cells were almost immobile. In contrast, T-cells migrating on COS-7 I-AK-HEL displayed higher velocities (unactivated: 5. 1±2. 4 µm/min; activated: 3. 7±1. 2 µm/min) with a high mobile fraction (unactivated: 0. 48±0. 17; activated: 0. 56±0. 16). These velocities are fully consistent with 2-photon-microscopy measurements (Figure S5), where migratory T cells in lymph nodes, or thymus slices display mean velocities around 4 µm/min [9], [28], [52]–[54] or stimulating lipid bilayers [55]. Although experimental, COS-7 I-AK-HEL have been shown to efficiently simulate physiological situations of T cell activation by inducing productive calcium signals, in a context of immunological kinapses [56], [57] rather than stable immunological synapses leading to specific cytokine secretion such as Interleukin-2 (data not shown). Additionally, as previously demonstrated, a clear correlation exists between activation and motility since unactivated T cells appear to move faster than activated ones within the same population [58] and T cell hybridoma mobility appears to decrease rapidly after calcium rise and rounding of the cells [10], [12] (Figure S6A). MAAACS was conditioned to automatically analyze the velocity and shape of the cells in addition to fluorescence signals, although no link between these parameters was found that was as tight as previous reports in cell systems expressing co-stimulatory or specific adhesion molecules. In the presence of 2-APB, the fluorescence amplitudes upon either anti TCR (FA = 3. 26±0. 24) or COS-7 I-AK-HEL stimulation (FA = 1. 56±0. 06) were strongly reduced compared to CRAC active control conditions. However, although the response fraction and the number of bursts per minute were left unaffected (RF = 0. 56±0. 03, BPM = 0. 09±0. 02) (Figure 5C) upon anti-TCR T cell activation, short and low calcium oscillations [26] were predominant in most T-cells (RF = 0. 26±0. 01, BPM = 0. 14±0. 01) (Figure 5C and Figure S3B) seeded onto COS-7 I-AK-HEL. In this case, the lack of co-stimulatory or specific adhesion molecules that usually sustain T cell/APC interactions and signaling [59] suggested that following TCR engagement by MHC-peptide, signaling events would occur through waves such as displayed by the calcium oscillations [60]. We therefore titrated the TCR-dependent calcium signaling in the presence of 2-APB in T-cells as a function of the amount of peptide loaded onto COS-7 I-AK. As shown in Figure 6, the peptide-specific, 2-APB sensitive calcium response was dependent upon the amount of peptide presented by COS-7 I-AK. In the absence of peptide, about 5% of the cells displayed a weak but significant calcium response above threshold. The percentage of responding cells increased proportionally to the peptide concentration to reach a plateau at a HEL 46–61 peptide concentration of 50 nM (Figure 6A), while being constantly oscillatory (Figure 6E). The fluorescence amplitude of the responding cells was significantly higher than that observed in the absence of antigenic peptide, even at low antigen doses (down to 0. 5 nM). (Figure 6B) Surprisingly, the fluorescence amplitude, response fraction, and burst frequency were independent of the antigen concentration (except at higher antigen concentrations) (Figure 6B–D). These data show that, at least in 3A9 T cell hybridomas, the TCR-mediated antigen-dependent ER-store-operated calcium response is digitally triggered irrespective of the antigenic peptide concentration. This is consistent with results from an earlier study [42] focusing however on the global calcium responses in 3A9 T cells. The antigen-dependent calcium response of mouse primary T cells has previously been investigated in vivo by microscopy on explants or sections of lymphoid organs [9], [28], [53] or ex vivo on artificial activating surfaces [55]. However, most of these studies were performed on lymphoblasts obtained by continuous activation in the presence of IL-2 for several hours and which therefore differ from naive T cells [18], [47], [61]. Data from studies that have examined the calcium response of naive T cells suggest that the calcium homeostasis of naive CD4+ T cells ex vivo is complex and at least in part antigen-independent [13], [62]. We evaluated these calcium responses of T-cells with MAAACS (Figure 7). Naive 3A9 transgenic CD4+ T cells [63], [64] seeded onto a surface coated with anti TCR antibodies showed a strong increase in fluorescence (FA = 4. 61±0. 25) that was maintained over time (RF = 0. 71±0. 03, BPM = 0. 13±0. 01) similar to that observed with 3A9 hybridomas (Figure 7A, B, 5C). However, we also observed that when seeded onto non-stimulating I-AK expressing COS-7 cells, around 20% of naive T cells responded spontaneously with short weak calcium pulses (FA = 1. 23±0. 03, RF = 0. 13±0. 01; BPM = 0. 14±0. 01) reminiscent of those previously reported [13] and [9]. Peptide specific calcium signals in the presence of COS-7 I-AK-HEL were mostly oscillatory in more than 60% of the cells (FA = 2. 02±0. 09, RF = 0. 32±0. 02, BPM = 0. 22±0. 01), in contrast to those observed in hybridomas (Figure 7B) which were mainly sustained. Another fundamental difference with hybridomas is that we found no clear correlation between calcium fluxes and cell velocity (Figure S6B), which nevertheless was expected considering in vivo reports [65]. We then wondered whether these calcium responses in CD4+ naive T cells were sensitive to 2-APB. CRAC channel activity in T cells is characterized upon stimulation by thapsigargin or soluble anti-CD3 antibody (2C11), generating sustained calcium responses that are absent in patients suffering from an inherited form of severe combined immune deficiency (SCID) syndrome or upon 2-APB treatment [3] Indeed, upon 2-APB treatment, the thapsigargin-induced calcium response was drastically reduced. Equivalent kinetics (evaluated by flow cytometry) was obtained with 2C11 stimulation in the presence of 2-APB or EDTA (Figure S7A–B). No additive or competitive effect was detected under these experimental conditions (Figure 7C). In fact, 2-APB-treated naive T cells seeded onto anti TCR coated surfaces did show a calcium response (FA = 2. 00±0. 14, RF = 0. 32±0. 03, BPM = 0. 13±0. 01) that was greatly reduced compared to the native conditions without CRAC inhibitor (Figure 7A). More unexpectedly, 2-APB treatment did not induce a unique calcium peak suggesting that other calcium channels than SOCE mediate calcium entry in naive mouse T cells, since calcium oscillations are dependent upon calcium influx [66]–[68]. Similarly, we analyzed the calcium response to COS-7 I-AK-HEL in the presence of 2-APB. There was a moderate but significant decrease in the amplitude of the calcium response (FA = 1. 66±0. 07, RF = 0. 29±0. 01, BPM = 0. 16±0. 01) compared to conditions in absence of 2-APB, together with a slight decrease of the oscillation frequency (Figure 7A–C). In addition 2-APB on naive T-cells seeded onto COS-7 I-AK did not show any significant impact on calcium fluxes (FA = 1. 28±0. 05, RF = 0. 13±0. 02, BPM = 0. 15±0. 02), remaining lower to the calcium response in presence of antigenic peptides. Altogether upon blockade of the CRAC channel activity by 2-APB, we evidence that the SOCE dependent calcium entry plays a limited role in mouse naive T-cells upon TCR triggering by-MHC-peptide. The first aim of this study was to significantly increase the sensitivity, accuracy, completeness and statistical reliability of video-microscopy approaches to record calcium fluxes, by combining a strong calcium probe with a robust algorithm for high density cell tracking coupled to an automated interface for rigorous post-acquisition analysis. The second objective was to use this method to describe the characteristic parameters of intracellular calcium fluxes within a population of T cells in response to different stimuli. We highlighted the heterogeneous nature and dynamics of these fluxes after TCR engagement by its natural ligand in a cell/cell context, which cannot be documented by flow cytometry. The TCR-MHC-peptide is a paradigm for unconventional intercellular receptor-ligand interaction [69] based upon successive cycles of engagement/release [16], [17]. However, more functional data supporting this current view are needed, taking care to account for the free motility of cells prospecting for cognate antigens supported by MHC molecules in a 2D cell membrane environment. Our goal was to develop experimental tools to contribute to the understanding of these mechanisms. The first challenge was finding a bright and stable fluorescent calcium probe in the visible range of the spectrum and that was easy to monitor both by cytometry and on conventional confocal microscopes. T cells dedicated to calcium imaging are usually loaded with calcium indicators the emitted fluorescence of which is UV-shifted upon calcium elevation thus allowing ratiometric measurements (such as Fura-2). The sensitivity of these probes can however be impacted by their intracellular compartmentalization (adsorption by proteins, interaction with membranes or sequestration by organelles e. g. mitochondria), or their extrusion by organic cation transporters. To overcome these technical issues, loading can be performed with diluents (pluronic acid) or transporter blockers (probenecid), although these compounds can be noxious to T-cells and thereby affect their response [38]. Despite these potential drawbacks, Fura-2 loaded T cells are routinely used for long experimental procedure followed by transcriptomics without limitations by reduced cell viability [25]. In our study, cells loaded with BD PBX were used over extended periods of time without any evidence of mortality, compartmentalization, or photobleaching which have been reported to affect Fluo-4 AM use [34]. Although they could be considered as anecdotic or trivial, such properties enable more reproducibility and the use of BD PBX in a greater number of experiments compared to other fluorescent visible probes. Our here proposed MAAACS method incorporates our previously reported MTT algorithm dedicated to single particle tracking [32] and nanoscopy [70], [71] set up to enable the detection, monitoring and reconnection of trajectories of moving T cells acquired by conventional confocal microscopy. The ability to simultaneously track a great number of targets is in itself a challenge but in particular encounters difficulties when tracks are interlaced or crossing over. The performance of MTT was found to be slightly superior compared to existing algorithms however the implementation of a program of assistance proposing candidate traces to be reconnected to aborted traces was a major breakthrough in terms of improving the accuracy and completeness. In addition, during the analysis process, MAAACS enabled the rejection if necessary of dead or dividing cells. Consequently, while MAAACS is not yet a fully unsupervised method, we speculate that 3D time lapse video-acquisition methods (on a spinning disk confocal microscope, for example) would greatly reduce the number of aborted traces due to focus loss that occurs on a 2D+time acquisition scheme such as that in this study (in particular for primary T cells). Completeness is a major issue in this kind of study, since the baseline calculation could be under-estimated or incorrect when the first time points after cell landing are missing. Here, the automated normalization of calcium signals facilitates their comparison among a population of cells. The MAAACS analysis makes simple the analysis and, more importantly, the quantification of signaling. MAAACS deciphers a video sequence in about 10 minutes, where manual tracking and analysis would take at least 2 hours (depending on the duration of the time lapse and the number of cells). Video microscopy records the behavior of individual cells over time and not just part of a population of anonymous cells. This allowed us to demonstrate that calcium oscillations are highly diverse among cells [15] both in terms of intensity and frequency; they are mostly transient oscillations in primary T cells in contact with antigen loaded APCs. This diversity in cell responses supports the notion that T-cell triggering is stochastically linked to heterogeneity in the T cell population [50]. This is conceivable for T-cell clones due to genomic drift, but may seem more surprising for primary CD4+ T cells. Literature reports that oscillatory calcium fluctuations are associated to effector function of T cells and proliferation [21] in contrast to memory T cells which display unique increases in calcium [72]. In addition, sustained calcium responses are observed mainly in apoptotic T cells [73]. Altogether, our approach would be able to reveal in a seemingly homogenous population, T cell diversity in terms of function or fate, based upon antigen dependent calcium response mode. Another interesting finding is that CRAC channel dependent activity does not support a sustained calcium response in naive T cells encountering APCs, and that the predominant calcium response modes in T cells are oscillations, in agreement with literature [5], [9], [25], [74], at least in part sensitive to 2-APB blockade. This result supports recent works showing an intriguing role of voltage dependent Ca2+ channels (Cav1. 4) [66]–[68] in the calcium influx into naive T cells. Our results also suggest that membrane calcium channel openings are tightly correlated to ER-calcium waves upon TCR triggering [75], [76], and that sustained calcium fluxes such as those triggered by stimulating antibodies and revealed by flow cytometry or video imaging are not strictly physiologic, at least not in naive T-cells. It could be valuable to consider our results in the light of recent evidence suggesting a role for cytoplasmic calcium sustained elevation in the orientation of the cytoplasmic domains of the CD3 chains of the TCR/CD3 complex upon activation [77]. As a major conclusion, the introduction of MAAACS emphasizes the urgent need to record the effects of cell-to-cell stimuli using real-time videos. We believe that MAAACS holds huge scope that could be easily adapted to study various kinds of targets (such as Qdots, vesicles, cells, animals) based on various types of emitted signal, however one immediate application would be to compare our in vitro results to 2-photon imaging of calcium indicator-loaded T cells migrating in lymph nodes [53], [65]. 2-Aminoethoxydiphenylborate (2-APB) (10 µM final concentration used for hybridomas, 20 µM for naive T cells) and thapsigargin (1 µM final concentration) were purchased from Calbiochem, and Ionomycin (0. 1 µg/mL, final) from Sigma. The PBX calcium assay kit, the antibody against CD3ε (clone 145-2C11) (6. 5 µg/ml, final), 2C11 biot (10 µg/mL final) and the F23. 1 anti-TCR Vβ 8 1-3 antibody (10 µg/mL, final) were supplied by Becton Dickinson. The mitochondrion label, Mitotracker red CMX-Ros, and the calcium indicators Indo-1 AM, Fluo-4 AM and Fura Red AM were supplied by Life technologies (Molecular probes). Streptavidin (5 µg/mL) was supplied by Jackson Immunoresarch. C4H3 (anti I-Ak-HEL), GK1. 5 (anti CD4), and H193. 16. 3 (anti CD45) (10 µg/mL, final) monoclonal antibodies were produced and purified in the lab from hybridoma supernatants according to standard protocols. 3A9 hybridoma T CD4+ cells are specific for hen egg lysozyme peptide (HEL) bound to MHC II I-Ak molecules [78]. These cells were cultured in RPMI medium supplemented with 5% FCS, 1 mM sodium pyruvate and 10 mM Hepes. COS-7 cells were cultured in DMEM medium with 5% FCS, 1 mM sodium pyruvate and 10 mM Hepes. Experimental antigen-presenting cells (APCs) were generated by stably transfecting COS-7 cells (Amaxa, V solution, A024) with plasmid cDNAs coding for the α chain of MHC II and the β chain of MHC II I-AK alone or covalently fused to a peptide derived from HEL (provided by D. A. Vignali) [46]. Cells were sorted according to their positivity to surface labeling by C4H3 antibodies (Facsvantage, Becton Dickinson). APC monolayers were generated by seeding 5. 5 104 cells into poly-L-Lysine-coated 8-well Lab-Tek chamber (Nunc). Spleens and lymph nodes were recovered from CBA/J x C3H non-transgenic mice and 3A9 TCR transgenic mice [63], [64]. After the extraction of cells onto nylon membrane in DMEM F12 medium (Lonza), splenic erythrocytes were removed via NH4Cl lysis. CD4+ T cells were isolated by depleting the CD4 negative cells according to manufacturer instructions (Dynal Mouse CD4 Negative Isolation Kit, Invitrogen). The day before experiments, cells were overnight serum starved in DMEM-F12 medium supplemented with 1% of Nutridoma-SP (Roche). For COS-7 transfected by plasmid cDNAs coding for the α and β chains of MHC II (I-AK), HEL peptide was added to the culture medium the day before the experiment. Coated surfaces were obtained by incubation with the appropriate concentration of antibody 24 h before the experiment at 37°C. Cells were analyzed on a LSR I flow cytometer (Becton Dickinson) with Cell Quest software or LSR II (for Fura Red/BD PBX and Indo-1 AM acquisitions) using the FACSDiva software. PBX calcium indicator was observed over time on the FL1 channel with an excitation by an Argon laser 488 nm and a 530/30 nm emission filter at 37°C, maintained using a water bath. Data analysis was performed with FlowJo software and the median intensity of fluorescence was plotted vs. time after exclusion of dead cells and cell debris. Movies were made on a Zeiss LSM 510 Meta confocal microscope equipped with a 30 mW argon laser (25% output, 1% AOTF). Pictures were taken with a C-Apochromat 40×/1. 2 water immersion objective, using the 488 nm line of the argon laser, HFT UV/488 dichroic mirror and a 505 nm long pass filter at 37°C, maintained using a hot plate. Time-lapse movies were composed of 300 images (512×512 pixels; 8 bit; 225 µm×225 µm; pinhole set to 3 airy units) taken every 7 seconds. Additional observations were performed on an Ultraview VoX Perkin Elmer spinning disk confocal microscope. All scripts, including multiple target tracking (MTT) [32], were developed under Matlab (The Mathworks). The source code of MTT, deposited at the Agence pour la Protection des Programmes, n° IDDN. FR. 001. 270021. 000 S. P. 2008. 000. 31230, is freely available for research purposes at http: //www. ciml. univ-mrs. fr/lab/he-marguet. htm. Cell tracking and automated analysis of cell signals with MAAACS can be done either in command line (directly in Matlab) or using a graphical user interface (GUI). While the GUI is more intuitive it is limited to the analysis of a single acquisition whereas the command line solution permits the sequential analysis of several video-acquisitions. Spectra of BD PBX and Fluo-4 AM were performed on the Cary Eclipse spectrofluorimeter (Varian). In vitro Kd determinations were performed using the calcium calibration kit (Life technologies) according to the manufacturer' s instructions. All statistical analyses and normality tests were performed using GraphPad Prism 5. 00. To determine the normality of the data, we performed a D' Agostino-Pearson normality test. Since not all our data were normally distributed, we used a non-parametric statistical test (two-tailed Mann-Whitney test with an alpha level of 5%).

The adaptive immune response to pathogen invasion requires the stimulation of lymphocytes by antigen-presenting cells. We hypothesized that investigating the dynamics of the T lymphocyte activation by monitoring intracellular calcium fluctuations might help explain the high specificity and selectivity of this phenomenon. However, the quantitative and exhaustive analysis of calcium fluctuations by video microscopy in the context of cell-to-cell contact is a tough challenge. To tackle this, we developed a complete solution named MAAACS (Methods for Automated and Accurate Analysis of Cell Signals), in order to automate the detection, cell tracking, raw data ordering and analysis of calcium signals. Our algorithm revealed that, when in contact with antigen-presenting cells, T lymphocytes generate oscillating calcium signals and not a massive and sustained calcium response as was originally thought. We anticipate our approach providing many more new insights into the molecular mechanisms triggering adaptive immunity.

Abstract Introduction Results Discussion Materials and Methods

2013

Barcoding T Cell Calcium Response Diversity with Methods for Automated and Accurate Analysis of Cell Signals (MAAACS)

There is an ultimate need for efficacious vaccines against human cytomegalovirus (HCMV), which causes severe morbidity and mortality among neonates and immunocompromised individuals. In this study we explored synthetic long peptide (SLP) vaccination as a platform modality to protect against mouse CMV (MCMV) infection in preclinical mouse models. In both C57BL/6 and BALB/c mouse strains, prime-booster vaccination with SLPs containing MHC class I restricted epitopes of MCMV resulted in the induction of strong and polyfunctional (i. e. , IFN-γ+, TNF+, IL-2+) CD8+ T cell responses, equivalent in magnitude to those induced by the virus itself. SLP vaccination initially led to the formation of effector CD8+ T cells (KLRG1hi, CD44hi, CD127lo, CD62Llo), which eventually converted to a mixed central and effector-memory T cell phenotype. Markedly, the magnitude of the SLP vaccine-induced CD8+ T cell response was unrelated to the T cell functional avidity but correlated to the naive CD8+ T cell precursor frequency of each epitope. Vaccination with single SLPs displayed various levels of long-term protection against acute MCMV infection, but superior protection occurred after vaccination with a combination of SLPs. This finding underlines the importance of the breadth of the vaccine-induced CD8+ T cell response. Thus, SLP-based vaccines could be a potential strategy to prevent CMV-associated disease. Human cytomegalovirus (HCMV) contributes substantially to morbidity in immunocompromised individuals. Organ or hematopoietic stem cell transplant recipients, people infected with HIV and patients with lymphocytic leukaemia are particularly vulnerable to HCMV-associated disease [1]. Moreover, congenital HCMV infection of unborn and new born children can lead to severe and permanent neurological symptoms [2]. Although currently available antivirals for HCMV are able to decelerate viral progression, thereby reducing the odds for major side effects, they require prolonged treatment periods and are accompanied with significant toxicity. Adoptive transfer of HCMV-specific T cells is an alternative treatment modality but is costly and laborious. The apparent burden of HCMV-associated disease and the paucity of cost-effective measures without side-effects have led to major efforts to develop effective HCMV vaccines but unfortunately no licensed vaccines are currently available [3,4]. There is accumulating evidence that effective control of persistent viral infections requires the induction of a balanced composition of polyfunctional T cell responses [5]. T cell immunity against CMV plays a critical role in controlling the primary viral infection and latency [6]. Whereas CMV-specific CD4+ T cells are important during the primary infection phase, CD8+ T cells are associated with greater benefits at the persistent infection phase and confer superior protection during reactivation and re-exposure [7–9]. Upon CMV infection, extra-ordinary large CD8+ T cell responses of diverge phenotype arise. CD8+ T cell response kinetics specific to most antigens follow the traditional course comprised by expansion after antigen encounter, rapid contraction, long-term maintenance at low levels and acquisition of a central-memory phenotype. Interestingly, CD8+ T cell responses to certain CMV antigens do not dwindle post-infection but inflate and exhibit a polyfunctional effector-memory phenotype [10–13]. In immunocompromised hosts, the balance between CMV and cellular immunity is apparently underdeveloped or lost, and therefore instigating the development and/or restoration of the T cell compartment specific for CMV would be particularly informative. The overarching aim of this study was to test a potential prophylactic vaccine platform against CMV based on synthetic long peptides (SLPs) containing immunodominant T cell epitopes. Previously, we reported that in therapeutic settings SLP-based vaccines can be successfully designed to stimulate effector and memory T cells against human papilloma virus-associated disease in mice and human [14–16]. As the efficacy of SLP-based vaccines is directly linked to the phenotypical and functional characteristics of the vaccine-induced CD8+ T cell response, we rigorously evaluated SLP-induced T cell responses. MCMV-specific SLP vaccines, assessed in two different mouse strains (C57BL/6 and BALB/c mice), lead to strong polyfunctional T cell responses, and combined SLP vaccines targeting different antigens provide a successful vaccine modality to control MCMV infection. To assess the potential of SLP-based vaccines in eliciting protecting CD8+ T cell responses against MCMV infection, we designed SLPs containing immunodominant MHC class I T cell epitopes from MCMV encoded proteins, and evaluated this vaccine platform in two different immunocompetent mouse strains with different susceptibility to MCMV; the C57BL/6 strain (MHC haplotype H-2b) and the more MCMV-susceptible mouse strain BALB/c (MHC haplotype H-2d) (S1 Table). C57BL/6 mice are less susceptible to MCMV infection compared to BALB/c mice because C57BL/6 mice express the NK cell-activating receptor Ly49H, which recognizes the MCMV protein m157 at the surface of infected cells [17–20]. Mice were vaccinated subcutaneously with SLPs along with the TLR9 ligand CpG as adjuvant. The SLP vaccine administration was well tolerated without adverse events. At day 7 after SLP immunization, epitope-specific CD8+ T cell responses were detected in the blood but a booster vaccination was required for induction of vigorous CD8+ T cell responses (Fig 1A and 1B). Prime-boosting with SLP vaccines induced very high frequencies of circulating CD8+ T cells against the noninflationary epitopes M45985-993 and M57816-824 in C57BL/6 mice, and were even higher than the percentages of the circulating MCMV-induced CD8+ T cells at the peak of infection (day 7). Also the response against m139419-426, known to be non-inflationary during the early phase after MCMV and at later time points as inflationary, is strong. The response against the non-inflationary M45507-515 epitope in BALB/c mice was even much higher in the SLP-vaccinated group as compared to the MCMV infected mice. The frequencies of the circulating CD8+ T cells against the inflationary M38316-323 and IE3416-423 epitopes in C57BL/6 mice and the inflationary m164257-265 and IE1168-176 epitopes in BALB/c mice were comparable (Fig 1A and 1B). SLP vaccines containing MHC class I epitopes may comprise unidentified class II epitopes and linear B cell epitopes leading to CD4+ T cell and antibody responses. To exclude this possibility, we performed polychromatic intracellular cytokine staining with the SLPs and performed SLP-specific antibody ELISAs, respectively (S1 and S2 Figs). Neither MCMV-specific CD4+ T cells nor peptide specific Abs were detected in these assays, indicating that the designed SLPs lead exclusively to antigen-specific CD8+ T cell responses and that epitope-specific responses induced by SLP or MCMV can only be compared for CD8+ T cells. Longitudinal analysis of the antigen-specific CD8+ T cell responses revealed that all SLP-induced T cell responses in both mice strains contracted gradually over time after the booster immunization (Fig 1B). Two months after the booster vaccination, the SLP-induced responses to most epitopes were still clearly detectable in blood. During MCMV infection, the epitope-specific CD8+ T cell responses followed a different course, consistent with previous reports [10,11]. T cell responses to the non-inflationary epitopes M45985-993, M57816-824, and M45507-515 rapidly contracted after the peak response and were stably maintained in time while T cell responses to the epitopes M38316-323, m139419-426, m164257-265, IE1168-176 and IE3416-423 inflated (Fig 1B). These data indicate that the context of epitope expression determines the kinetics of the T cell responses, which is uniform for diverse epitopes after SLP vaccination but in the case of MCMV infection this results in a dichotomy of responses related to the chronic nature of this infection. At the peak after the booster SLP immunization (day 7–8), high frequencies of epitope-specific CD8+ T cells, analogous to the responses elicited by MCMV virus were observed in the spleen (Fig 1C). However, in absolute numbers, MCMV infection led to a higher T cell magnitude compared to SLP vaccination, which can be attributed to virus-associated inflammation leading to splenomegaly. At the memory phase (day 60), MCMV-specific T cell responses to the non-inflationary epitopes were significantly lower than the equivalent SLP vaccine-induced responses (Fig 1C). The MCMV-induced CD8+ T cell responses to the inflationary epitopes were of higher magnitude compared to those induced by SLP vaccination. Taken together, these results show that prime-boost vaccination with SLP vaccines containing MHC class I MCMV epitopes elicit in mouse strains with different susceptibility to MCMV high percentages of effector and memory CD8+ T cells that contract gradually in time. Next, we aimed to dissect the underlying mechanisms of the relatively low responses to some of the SLPs (i. e. M38 and IE3 in C57BL/6; M45 in BALB/c) compared to the other. First, we endeavoured to alter the SLP sequences by altering the C-terminal cleavage, which may improve the immunogenicity (S3 Fig). However, the altered M38316-323 SLP did not exhibit a significant improvement in the SLP-induced T cell response whilst the altered SLP containing the IE3416-423 epitope elicited responses that were actually reduced. Then we questioned if the differences in the magnitude of the T cell responses triggered by the various single SLP vaccines might be related to the functional avidity of the T cells, which is determined by the affinity of the peptide for MHC and the TCR affinity for the peptide-MHC complex (Fig 2A and 2B). The SLPs elicited T cells with different levels of functional avidity but no correlation was found with the strength of the CD8+ T cell response. Moreover, in both C57BL/6 and BALB/c mice the functional avidity of the T cells, elicited either by SLP vaccines or MCMV infection, were remarkably similar and remained stable in time as they were similar during the acute and memory phase of response. Thus, differences in TCR affinity are not involved in the observed difference in the magnitude of the T cell responses. The data presented above illustrated that factors other than peptide-MHC/TCR affinity are implicated in shaping the strength of SLP-induced T cell responses. Recently, it was shown that the precursor frequency of naive T cell populations can predict the immunodominance hierarchy of viral epitope specific CD8+ T cell responses [21]. To test whether the precursor frequency is predictive for the magnitude of SLP-induced T cell responses we determined the precursor frequency of all the epitopes included in this study in naive C57BL/6 and BALB/c mice (Fig 2C). In C57BL/6 mice, the precursor frequencies for the M45985-993 and M57816-824 epitopes were among the highest followed by the precursor frequencies to the m139419-426 epitope. The lowest precursor frequencies were detected to the M38316-323 and IE3416-423 epitopes, confirming a previous report [22]. In BALB/c mice, the highest precursor frequencies were observed for the m164257-265 and IE1168-176 epitopes whereas the frequency of M45507-515 specific T cells was lower (Fig 2C). Markedly, the average level of the precursor frequency of each epitope-specific CD8+ T cell population was proportional to the expansion of the antigen-specific populations found in mice following either SLP immunization or MCMV infection. Together, these results indicate that naive precursor frequencies rather than TCR avidity determine the magnitude of SLP vaccine-mediated CD8+ T cell responses. To assess the phenotypical and functional quality of MCMV-specific CD8+ T cells induced by either the SLPs or the virus, we determined the formation of the diverse T cell subsets that develop after antigenic challenge. Early after the booster, SLP vaccination resulted in the induction of a highly activated CD8+ T cell subset exhibiting an effector-like phenotype (CD62Llo, CD44hi, CD127lo, KLRG1hi), which completely resembled the MCMV-specific T cell phenotype during the acute phase of the infection (Fig 3A and 3B). In the memory phase, both SLP- and MCMV-induced T cell phenotypic traits diverged (Fig 3C and 3D). All SLP-induced CD8+ T cells exhibited a fairly mixed phenotype sharing features of both central-memory T cells (CD62Lhi, CD44lo, CD127hi, KLRG1lo), effector-memory T cells (KLRG1hi, CD44hi, CD127lo, CD62Llo) but also an intermediate phenotype (i. e. KLRG1hi, CD127hi). As expected, during MCMV infection, the non-inflationary M45985-993, M45507-515 and M57816-824-specific CD8+ T cells gained a predominant central memory-like phenotype while the inflationary M38316-323, m139419-426, IE3416-423, m164257-265 and IE1168-176-specific T cells appeared mostly effector-memory like. To assess the cytokine profiles of the SLP-induced CD8+ T cells, we performed intracellular cytokine staining for IFN-γ, TNF and IL-2 and compared these to MCMV-induced T cells. At the peak response after booster vaccination, SLP-induced T cells consisted mainly of single IFN-γ and double IFN-γ/TNF producing populations (Figs 4A and S4). The cytokine producing traits of the MCMV-induced effector CD8+ T cells matched in general with the SLP-elicited T cells. Except relatively more single IFN-γ producing CD8+ T cells after MCMV infection compared to SLP vaccination were found in the T cell populations reactive to the epitopes IE3416-423, IE1168-176, M45507-515 and m164257-265. At the memory phase, the SLP-specific CD8+ T cells gained the ability to co-produce the three cytokines, at the expense of single cytokine producing cells (Figs 4B and S4). This gain in triple cytokine production (IFN-γ/TNF/IL-2) during MCMV infection is mainly observed in the non-inflationary CD8+ T cells. Both during the acute and memory phase, the percentage of the total CD8+ T cell population producing IFN-γ, either in case of SLP vaccination or MCMV infection, corresponded to the percentage of MHC class I tetramers, indicating full differentiation of the elicited T cells. A hallmark of memory T cells is the ability to undergo secondary expansion upon antigenic challenge [23]. To assess this property of vaccine-induced memory T cells, we performed adoptive transfer experiments in which congenically marked (CD45. 1+) memory M45985-993 and m139419-426-specific CD8+ T cells from SLP vaccinated and MCMV infected mice were isolated and transferred into naive recipient mice, which were subsequently challenged with MCMV (Fig 5). SLP-induced M45985-993 and m139419-426-specific T cells expanded; albeit to a lesser extend as compared to the MCMV-induced (Fig 5). The MCMV-elicited M45985-993-specific T cells exhibited, corresponding to their central-memory phenotype, a superior capacity in expansion as compared to the MCMV-elicited m139419-426-specific T cells with an effector-memory phenotype. Of note, the expansion of the SLP-induced M45-specific T cells was comparable to the m139-specific T cells induced by MCMV, although the phenotype of SLP-induced cells were more central-memory like. This indicates that the instruction that T cells receive in different settings can result in cells with a different expansion potential despite a seemingly similar phenotype based on markers for central/effector memory cells. All together we conclude that SLP-based vaccines induce a heterogeneous pool of memory T cells with a secondary expansion potential that is somewhat lower as compared to memory T cells elicited by virulent virus. The various SLP vaccine formulations were evaluated for their capacity to confer protection against MCMV challenge (at day 60 after booster vaccination). In C57BL/6 mice, the viral load of unvaccinated (naive) mice challenged with MCMV was found to be significantly higher in spleen, liver and lungs, when compared to the viral load of MCMV re-challenged mice that successfully controlled a previous MCMV infection, indicating that pre-existing immunity to MCMV can clearly reduce the viral load upon re-infection (Fig 6A). All the different SLP vaccines resulted in a reduction in viral load in the spleen compared to unvaccinated mice, albeit less effective when compared to MCMV infected mice. Mice vaccinated with the SLPs containing the M38316-323 and m139419-426 epitopes display a significant reduction in viral titres in the liver and lungs. Also, the M57816-824 and IE3416-423 epitope containing SLPs were capable in reducing the viral replication in the liver after MCMV challenge, albeit to a lesser extent (Fig 6A). Re-challenge of MCMV infected BALB/c mice resulted in substantial protection of the m164257-265 epitope containing SLP vaccine in spleen and liver (Fig 6B). The M45507-515 and IE1168-176 epitope containing SLPs however did not induce protective immunity in vaccinated mice. These results indicate that certain SLPs but not all have the potency to elicit protective immunity against virus challenge, and that this protection is not necessarily correlating to the size of the SLP-induced CD8+ T cell response. Since vaccination with the m139419-426 and M38316-323 epitope containing SLPs was accompanied with some reduction of the viral load, we examined in C57BL/6 mice whether vaccination with these two, or even more, SLPs combined is able to exceed the protection efficacy of single SLP immunization. Strong and long-lived peptide-specific CD8+ T cell responses were measured in mice vaccinated with the mixture of the m139 SLP plus the M38 SLP and with a mixture of all 5 SLP vaccines (Fig 7A). Notably, the T cell response against each peptide epitope with the combined SLP vaccines was lower as compared to single SLP vaccination (except for the m139-specific response), indicating that competition among antigen-specific CD8+ T cell populations can occur in multivalent vaccines. Especially, altered were the responses to the epitopes in M57 and IE3 because these were not boosted (Fig 7A). Such competition among T cells during boosting has also been observed after viral infection [24]. The kinetics of the combined SLP vaccine-induced T cell responses was found to be similar to single SLP vaccines, and the phenotype (Fig 7B) and cytokine polyfunctionality of the T cells as well (S5 Fig). At day 60 post booster vaccination, mice were challenged with MCMV and 5 days later viral titres were measured in different organ tissues. The efficacy of the combined SLPs to protect upon acute MCMV challenge was remarkably improved compared to the single SLP vaccines, as all mice that received a mixed SLP vaccine exhibited significant reduction in the viral load, especially in the liver (Fig 7C), suggesting that the breadth of the response or the magnitude of the total anti-viral response is important. Remarkably, the combination of the m139 SLP with the M38 SLP was as efficacious as the combination with all 5 SLPs. To assess if superior viral control was related to the breadth of the response, we adoptively transferred 1 × 104 m139 SLP-induced CD8+ T cells, 1 × 104 M38 SLP-induced CD8+ T cells, or an equal total number of a pool of both m139 (0. 5 × 104) and M38 (0. 5 × 104) SLP-induced CD8+ T cells in naive recipient mice (Fig 7D). The transfer of SLP-induced CD8+ T cell populations with a dual specificity resulted in a significant reduction in viral titres, while the transfers of equal amounts of T cells with single specificity did not. Thus, combinations of at least two distinct SLP vaccines have an increased potency to protect compared to single SLP vaccines, indicating that the breadth of the vaccine-induced CD8+ T cell responses plays a crucial role in anti-viral immunity. We conclude that vaccination with single SLPs can be applied as a prophylactic vaccine strategy against CMV infection, but vaccination with combinations of different SLPs serve as a superior vaccine technology platform against viral challenge. In this study we report that SLP-based vaccines are an effective modality against CMV infection. In a prime-boost vaccine regimen, SLPs containing single MCMV epitopes are highly immunogenic in both C57BL/6 and BALB/c mice, and generate long-lasting polyfunctional CD8+ T cell responses. Our study revealed three key findings. First, the magnitude and phenotype of the SLP-induced T cell responses initially resemble those evoked by a real viral infection. Second, the magnitude of the SLP-induced T cell response strongly correlated to the naive T cell precursor frequency, and third the protection against viral infection by SLP-induced memory CD8+ T cells was most pronounced when vaccination was performed with combinations of distinct SLPs leading to an increased breadth of the antigen-specific T cell response. In the last decades many vaccine strategies such as attenuated virus, DNA constructs, protein, and virally vectored vaccines targeting HCMV have been developed [3,4]. The focus of most of these vaccines was to generate protective antibodies. Our finding that SLP-based vaccines that solely provoke CD8+ T cell responses are efficacious suggests that the design of more efficient vaccines against CMV should incorporate the induction of CD8+ T cell immunity. Although we observed some epitope competition among SLP vaccine-induced CD8+ T cell responses, we anticipate that inclusion of CD4+ T cell and B cell epitopes will further improve the vaccine efficacy given that CD4+ T cells and antibodies have also antiviral actions against CMV. Moreover, SLP-based vaccines allow further refinement by different prime-booster regimens and by combinations with adjuvants, immunomodulatory antibodies or other vaccine platforms [25]. Conceivably, this will positively impact the phenotype and effectivity of the vaccine-induced T cells. As to date, the high CD8+ T cells responses elicited with the SLP vaccines encoding MCMV epitopes have not been observed before with other SLPs including those containing epitopes of human papilloma virus (HPV) [14], lymphocytic choriomeningitis virus (LCMV) [26], influenza [27] or model antigens [28]. This may be explained by the relatively high precursor frequency of T cells responding to some of the MCMV epitopes. Our study indicates that it is of interest for T cell-based vaccines to determine the antigen-specific T cell precursor frequencies as these correlate to the magnitude of the vaccine-induced antigen-specific response, allowing the selection of epitopes generating the most robust responses. This knowledge can be very useful for development of vaccines that are based on selection of epitopes. Nevertheless, the magnitude of the vaccine-induced T cell response appears not necessarily to correlate to protective immunity but seems to depend also on the specificity. For example, in C57BL/6 mice, the large vaccine-elicited responses to the M45985–993 and M57816-824 epitopes do not provide as good protection as the seemingly lower response to the M38316–323 epitope. Similarly, in BALB/c mice the m164 SLP confers immunity in liver and spleen whereas the IE1168-176 epitope containing SLP, which is analogous in magnitude, does not show protective effects. Previous studies using short peptide or DNA vaccination also reported that the strength of the vaccine-induced IE1-specific CD8+ T cell response does not necessarily correlate to protection [29–31], suggesting that the quality of the vaccine-induced T cell is more decisive. Dissimilarities in transcription of viral genes [32], which may even vary in different tissues, as well as the efficiency of peptide processing and presentation at the cell surface, may also be implicated in the differential efficacy of the T cell response to each particular epitope to confer resistance to MCMV. In this respect it is of interest to note that SLP vaccines containing “inflationary” epitopes (i. e. , M38 and m139) elicit better protection as compared to the non-inflationary epitopes. This may relate to differences in the presentation of the inflationary epitopes as compared to the non-inflationary epitopes by infected cells and/or by (cross-presenting) APCs. An important requirement for memory inflation is chronic antigenic exposure [12]. The fact that SLPs do not elicit inflation suggests that SLPs are broken down in such a manner that epitopes are not presented over a long period of time as occurs during persistent CMV infection. Other factors important for memory inflation during CMV infection, such as dependence on certain T cell costimulatory interactions (e. g. , CD27-CD70 [33]), are likely also not in place at late time points post SLP vaccination. In addition, a characteristic feature of inflationary T cells is their predominant effector-memory like phenotype. The SLP vaccine-induced T cells are not mostly effector-memory like, as may be expected because of the apparent absence of memory inflation. Although the expansion of the SLP-induced CD8+ T cells seems to be somewhat negatively influenced as compared to virus-induced T cells, it remains to be determined whether protection on a per-cell basis is influenced as well. Nevertheless, the SLP-induced T cells were well capable to reduce the viral load upon viral challenge, especially when a mixture of distinct SLPs was used for vaccination. The somewhat lesser expansion potential of the SLP-induced T cells might relate to some of the differences in the phenotype of the SLP and MCMV-elicited T cells. Although the effector T cells induced by either SLP boost vaccination or MCMV infection had an analogous phenotype (KLRG1hi, CD44hi, CD127low, CD62Llow, IL2+/-) and cytokine profile, the memory T cells elicited by SLPs displayed a mixed profile of effector-memory (KLRG1hi, CD127lo), central-memory (KLRG1lo, CD127hi) and double-positive T cells (KLRG1hi, CD127hi). In contrast, MCMV infection induces a more polarized phenotype: either a central-memory phenotype (mainly non-inflationary responses) or an effector-memory phenotype (mainly inflationary responses). Whether a lack of CD4+ T cell helper signals [34] or a lack of virus-associated inflammatory signals [26] is responsible for the observed SLP vaccine-associated phenotype and secondary expansion potential remains to be examined in future studies. We showed that the efficacy of SLP vaccines to protect against MCMV is primarily driven by the breadth of the CD8+ T cell responses rather than the magnitude of the individual SLP vaccine-induced T cell responses. A possible explanation is that viral infected cells are to a certain degree resistant to CD8+ T cell mediated killing due to sophisticated immune evasion mechanisms including downmodulation of MHC class I molecules and prevention of apoptosis [35–37]. Accordingly, it has been estimated that one effector CD8+ T cell kills only 2–16 MCMV-infected cells per day and the probability of death of infected cells increases for those contacted by more than two CTLs, which is indicative of CTL cooperation [38]. Our study suggests that multiple encounters with cytotoxic CD8+ T cells with different specificity result in more effective killing of infected cells. Overall, this study provided evidence that SLP-based vaccines eliciting memory CD8+ T cell responses have protective effects against acute MCMV infection with respect to lowering the viral load in tissues. These promising results highlight the need for additional studies to elucidate the role of vaccine-induced T cells against CMV and other persistent viral infections. C57BL/6 mice and BALB/c mice were purchased from Charles River Laboratories (L' Arbresle, France). CD45. 1 (Ly5. 1) congenic mice on a C57BL/6 background were obtained from The Jackson Laboratory. Mice were maintained under specific-pathogen-free conditions at the Central Animal Facility of Leiden University Medical Center (LUMC), and were aged 8–10 weeks at the beginning of each experiment. The mice did not undergo any immunosuppressive treatments and were fully immunocompetent. All animal experimental protocols were approved by the LUMC Animal Experiments Ethical Committee in accordance with the Dutch Experiments on Animals Act and the Council of Europe (numbers 13156 and 14187). MCMV virus stocks were prepared from salivary glands of BALB/c mice infected with MCMV-Smith (American Type Culture Collection (ATCC) ). The viral titres of the produced virus stocks were determined by viral plaque assays with 3T3 mouse embryonic fibroblasts (MEFs) (ATCC). Age- and gender-matched C57BL/6 mice were infected with 5 × 104 PFU MCMV, and age- and gender-matched BALB/c mice with 5 × 103 PFU MCMV. Viruses were administered intraperitoneally (i. p) in a total volume of 400 μl in PBS. At 65 days post-booster vaccination or infection, SLP vaccinated or MCMV infected mice were (re) -challenged with 5 × 104 PFU MCMV. Determination of viral load was performed by real-time PCR as described previously [39]. Short (9–10 aa) and long (20–21 aa) peptides containing MHC class I-restricted T cell epitopes from MCMV encoded proteins in C57BL/6 and BALB/c mice were produced at the peptide facility of the LUMC (peptide sequences are described in S1 Table). The purity of the synthesized peptides (75–90%) was determined by HPLC and the molecular weight by mass spectrometry. Synthetic long peptide (SLP) vaccinations were administered subcutaneously (s. c.) at the tail base by delivery of 50 μg SLP and 20 μg CpG (ODN 1826, InvivoGen) dissolved in PBS in a total volume of 50 μl. Booster SLP vaccinations were provided after 2 weeks. Vaccination with a mixture of SLPs was done with 50 μg of each SLP and 20 μg CpG. Cell surface and intracellular cytokine stainings of splenocytes and blood lymphocytes were performed as described [40]. For examination of intracellular cytokine production, single cell suspensions were stimulated with short peptides for 5 h in the presence of brefeldin A or with long peptides for 8 h of which the last 6 h in presence of brefeldin A (Golgiplug; BD Pharmingen). MHC class I tetramers specific for the following MCMV epitopes: M45985–993, M57816–824, m139419-426, M38316–323, and IE3416–423 in C57BL/6 mice and M45507–515, m164257-265 and IE1168-176 in BALB/c mice were produced as reported [41]. Fluorochrome-conjugated mAbs were purchased from BD Biosciences, Biolegend or eBioscience. Flow cytometry gating strategies are shown in S6 Fig. Samples were acquired with the LSRFortessa cytometer (BD Biosciences) and analysed with FlowJo-V10 software (Tree star). A peptide dose-response titration was performed to determine and compare the TCR avidity of the CD8+ T cells induced after SLP vaccination and MCMV infection at the acute and memory phase. In brief, splenocytes were stimulated with various concentrations of short peptide in presence of 2 μg/ml brefeldin A for 5 h at 37°C. Subsequently, cell surface staining and an intracellular IFN-γ staining were performed. Responses were analysed using the same approach as described above. Blood was collected from the retro-orbital plexus and after brief centrifugation, sera were obtained and stored at −20°C. Specific immunoglobulin levels in serum were measured by ELISA as described [39]. Briefly, Nunc-Immuno Maxisorp plates (Fisher Scientific) were coated either with 2 μg/ml SLPs or with MCMV-Smith in bicarbonate buffer, and after blocking with skim milk powder (Fluka BioChemika) diluted sera were added. Next, plates were incubated with HRP-conjugated antibodies (SouthernBiotech) to detect different Ab isotypes. Plates were developed with TMB substrate (Sigma Aldrich) and the colour reaction was stopped by addition of 1M H2SO4. To serve as a positive control, a peptide from the M2 protein (eM2) of influenza A virus with identified ability to induce antibodies and corresponding serum was used. Optical density was read at 450 nm (OD450) using a Microplate reader (Model 680, Bio-Rad). To determine the endogenous naive precursor frequency of MCMV-specific CD8+ T cell populations in C57BL/6 and BALB/c mice, enrichment assays of antigen-specific CD8+ T cells were performed as described [42]. In short, single cell suspensions were generated from pooled spleen and lymph nodes (mesenteric, inguinal, cervical, axillary, and brachial) of individual mice. Cells were stained with PE and APC-labelled MHC class I tetramers for 0. 5 h at RT, then washed, labelled with anti-PE and anti-APC microbeads (Miltenyi Biotec), and passed over a magnetized LS column (Miltenyi Biotec). The tetramer-enriched fractions were stained with fluorochrome labelled Abs against CD3 (clone 500A2), CD4 (clone L3T4), CD8 (clone 53–6. 7) for 30 min at 4°C, and subsequently analysed. Samples were acquired with the LSRFortessa cytometer (BD Biosciences). The expansion capacity and vaccine efficacy of SLP vaccine and/or MCMV-induced antigen-specific CD8+ T cells was determined by adoptive transfers. Splenic CD8+ T cells from chronically (day 60) infected and SLP vaccinated CD45. 1+ mice were enriched with magnetic sorting using the CD8+ T cell isolation kit in accordance with the manufacturer’s protocol (Miltenyi Biotec). Next, cells were stained with MHC class I tetramers and with fluorochrome labelled antibodies against CD3 and CD8. Tetramer positive CD8+ T cells were sorted using a FACSAria II Cell Sorter (BD Biosciences) and 1 × 104 tetramer+ CD8+ T cells were transferred (retro-orbital in a total volume of 200μl in PBS) into naive CD45. 2+ C57BL/6 recipients. Recipients were subsequently (2 h later) infected with 5 × 104 PFU MCMV. At day 5 post viral challenge the viral titres were determined by qPCR and the number of donor-specific CD8+ T cells by flow cytometry. Statistical significance was assessed with Student’s t-test or ANOVA using GraphPad Prism software (GraphPad Software Inc. , USA). The level of statistical significance was set at P<0. 05.

The majority of infections with the betaherpesvirus human cytomegalovirus (HCMV) are clinically unnoticed, but in immunocompromised hosts HCMV infections can be severe and even fatal. Here we investigated in preclinical mouse models the efficacy and mechanisms of synthetic long peptide (SLP) -based vaccines eliciting mouse CMV (MCMV) -specific CD8+ T cells as a platform modality to protect against CMV infection. The percentages of MCMV-specific T cells in the circulation elicited by prime-booster SLP vaccination were equivalent or higher compared to those induced by the virus itself. We further show that the naive T cell precursor frequency rather than the functional avidity of T cells predicts the magnitude of SLP-induced CD8+ T cell responses. Superior protection against MCMV infection depends strongly on the combined use of distinct SLP vaccines leading to broader viral-specific responses. This finding highlights the importance of the breadth of vaccine-induced CD8+ T cell responses.

Abstract Introduction Results Discussion Materials and Methods

blood cells innate immune system medicine and health sciences immune cells immune physiology cytokines immunology cytomegalovirus infection vaccines preventive medicine developmental biology clinical medicine molecular development cytotoxic t cells vaccination and immunization public and occupational health infectious diseases white blood cells major histocompatibility complex memory t cells animal cells t cells immune system cell biology clinical immunology physiology biology and life sciences cellular types viral diseases

2016

The Breadth of Synthetic Long Peptide Vaccine-Induced CD8+ T Cell Responses Determines the Efficacy against Mouse Cytomegalovirus Infection

Pervasive natural selection can strongly influence observed patterns of genetic variation, but these effects remain poorly understood when multiple selected variants segregate in nearby regions of the genome. Classical population genetics fails to account for interference between linked mutations, which grows increasingly severe as the density of selected polymorphisms increases. Here, we describe a simple limit that emerges when interference is common, in which the fitness effects of individual mutations play a relatively minor role. Instead, similar to models of quantitative genetics, molecular evolution is determined by the variance in fitness within the population, defined over an effectively asexual segment of the genome (a “linkage block”). We exploit this insensitivity in a new “coarse-grained” coalescent framework, which approximates the effects of many weakly selected mutations with a smaller number of strongly selected mutations that create the same variance in fitness. This approximation generates accurate and efficient predictions for silent site variability when interference is common. However, these results suggest that there is reduced power to resolve individual selection pressures when interference is sufficiently widespread, since a broad range of parameters possess nearly identical patterns of silent site variability. Natural selection maintains existing function and drives adaptation, altering patterns of diversity at the genetic level. Evidence from microbial evolution experiments [1], [2] and natural populations of nematodes [3], fruit flies [4], [5], and humans [6], [7] suggests that selection is common and that it can impact diversity on genome-wide scales. Understanding these patterns is crucial, not only for studying selection itself, but also for inference of confounded factors such as demography or population structure. However, existing theory struggles to predict genetic diversity when many sites experience selection at the same time, which limits our ability to interpret variation in DNA sequence data. Selection on individual nucleotides can be modeled very precisely, provided that the sites evolve in isolation. But as soon as they are linked together on a chromosome, selection creates correlations between nucleotides that are difficult to disentangle from each other. This gives rise to a complicated many-body problem, where even putatively neutral sites feel the effects of selection on nearby regions. Many authors neglect these correlations, or assume that they are equivalent to a reduction in the effective population size, so that individual sites evolve independently. This assumption underlies several popular methods for inferring selective pressures and demographic history directly from genetic diversity data [8]–[12]. Yet there is also extensive literature (recently reviewed in Ref. [13]) which shows how the independent sites assumption breaks down when the chromosome is densely populated with selected sites. When this occurs, the fitness effects and demographic changes inferred by these earlier methods become increasingly inaccurate [14], [15]. Linkage plays a more prominent role in models of background selection [16] and genetic hitchhiking [17], which explicitly model how strong negative and strong positive selection distort patterns of diversity at linked sites. Although initially formulated for a two-site chromosome, both can be extended to larger genomes as long as the selected sites are sufficiently rare that they can still be treated independently. Simple analytical formulae can be derived in this limit, motivating extensive efforts to distinguish signatures of background selection and hitchhiking from sequence variability in natural populations (see Ref. [18] for a recent review). However, this data has uncovered many instances where selection is neither as rare nor as strong as these simple models require [7], [19]–[24]. Instead, substantial numbers of selected polymorphisms segregate in the population at the same time, and these mutations interfere with each other as they travel towards fixation or loss. The genetic diversity in this weak Hill-Robertson interference [25] or interference selection [26] regime is poorly understood, especially in comparison to background selection or genetic hitchhiking. The qualitative behavior has been extensively studied in simulation [22], [25]–[29], and this has led to a complex picture in which both genetic drift and chance associations between linked mutations (genetic draft) combine to generate large fluctuations in the frequencies of selected alleles, and the occasional fixation of deleterious mutations due to Muller' s ratchet. In principle, these forward simulations can also be used for inference or model comparison using approximate likelihood methods [7], [30], but in practice, performance concerns severely limit both the size of the parameter space and the properties of the data that can be analyzed in this way. Here, we will show that in spite of the complexity observed in earlier studies, simple behaviors do emerge when interference is sufficiently common. When fitness differences are composed of many individual mutations, we obtain a type of central limit theorem, in which diversity at putatively neutral sites is determined primarily by the variance in fitness within the population over a local, effectively asexual segment of the genome. This limit is analogous to the situation in quantitative genetics, where the evolution of any trait depends only on the genetic variance for the trait, rather than the details of the dynamics of individual loci. We exploit this simplification to establish a coalescent framework for generating predictions under interference selection, which is based on a coarse-grained, effective selection strength and effective mutation rate. This leads to accurate and efficient predictions for a regime that is often implicated in empirical data, but has so far been difficult to model more rigorously. Our method also has important qualitative implications for the interpretation of sequence data in the interference selection regime, which we address in the Discussion. We investigate the effects of widespread selection in the context of a simple and well-studied model of molecular evolution. Specifically, we consider a population of N haploid individuals, each of which contains a single linear chromosome that accumulates mutations at a total rate U and undergoes crossover recombination at a total rate R. We assume that the genome is sufficiently large, and epistasis is sufficiently weak, that the fitness contribution from each mutation is drawn from some distribution of fitness effects ρ (s) which remains constant over the relevant time interval. For the sake of concreteness and connection with previous literature, we will focus on the special case where all mutations confer the same deleterious fitness effect, which approximates a potentially common scenario where a well-adapted population is subject to purifying selection at a large number of sites. However, our results will hold for more general distributions of fitness effects, both beneficial and deleterious, provided that individual mutations are sufficiently weak or the overall mutation rate is sufficiently large. Since the effects of linked selection are most pronounced in regions of low recombination, we devote the bulk of our analysis to the asexual limit where R≈0. Later, we will show that recombining genomes can be treated as an extension of this limit by means of an appropriately defined linkage block, within which recombination can be neglected. These assumptions define a simple “null-model” of sequence evolution with a straightforward computational implementation (see Methods). In the present work, we focus on the genetic diversity at an unconstrained locus (e. g. , a silent or synonymous site) embedded near the center of the chromosome. We focus in particular on the site frequency spectrum, , which counts the number of mutations at this locus that are shared by i individuals in a sample of size n. The pairwise diversity π is equal to in this notation. We note that on average, , so we can summarize the average site frequency spectrum using a combination of π and the relative values, . In this parameterization, π measures of the overall levels of diversity, while measures the shape of the site frequency spectrum. Expectations of other commonly used diversity statistics (e. g. , Tajima' s D [31] or the average minor allele frequency) can be directly computed from. Although our model is simple, the expected patterns of silent-site variability remain poorly characterized for many biologically relevant parameters. Previous theoretical work has focused on combinations of N, U, s, and R that result in relatively few selected polymorphisms per unit map length. In the limit that, these populations converge to the background selection limit, where interference between deleterious mutations can be neglected and each selected site evolves independently. Traditionally, the term “background selection” is used to refer both to the general effects of purifying selection on linked neutral diversity as well as to the limiting behavior that emerges when. Here we use the term only in the latter sense, and we have opted for the slightly more precise label “background selection limit” in order to minimize confusion. This limit arises for arbitrary levels of recombination, but is easiest to visualize in the asexual case (R≈0). The expected fraction of individuals with k deleterious mutations (“fitness class k”) follows a Poisson distribution, (1) where parameterizes the relative strength of mutation and selection [32]. Patterns of silent site variability are equivalent to a demographically structured neutral population, where the fitness classes are treated as fixed subpopulations and mutation events are recast as migration between them (see Figure 1). This is a special case of the structured coalescent [33], which traces the ancestry of a sample as it moves through the population fitness distribution. The structured coalescent can be used to derive approximate analytical expressions for several simple diversity statistics [16], [34]–[38]. Previous work has shown that to lowest order in, silent site diversity resembles an unstructured neutral population with an effective population size. The overall level of diversity is therefore reduced from its neutral expectation () by the fraction (2) while the shape of the site frequency spectrum is unchanged. Higher-order corrections, which become increasingly relevant for larger sample sizes [39], can be efficiently calculated from backward-in-time simulations of the structured coalescent (Methods) [40]–[42]. For example, in Text S1 we show that the predicted reduction in diversity is well-approximated by (3) provided that Ns is not too small. In practice, structured coalescent methods provide reasonable accuracy for a range of parameters that we collectively term the background selection regime. Figure 1 shows that this constitutes a “strong-selection” region of parameter space (), although the precise meaning of strong is somewhat different from colloquial usage. In particular, this depends on more than just the magnitude of Ns alone, since mutations can have selective effects that are considered strong in a single-site setting (Ns∼100) but nevertheless have if the mutation rate is sufficiently high. Nor is this simply a statement about the magnitude of U/s. Somewhat confusingly, background selection is sometimes regarded as a “weak selection” effect, since is significantly reduced only when. We will avoid such terminology here. Instead, we find it more productive to think of the background selection regime as a “rare interference” limit, since the distribution of fitnesses within the population coincides with the independent-sites prediction in Eq. (1). In the present work, we focus on the opposite limit, the so-called interference selection regime, where mutation rates are sufficiently high or fitness effects sufficiently weak that many selected polymorphisms segregate in the population at once. In this regime, the frequencies of nearby deleterious mutations become correlated, and the distribution of fitnesses within the population fluctuates and eventually diverges from the independent-sites prediction in Eq. (1). As a result, structured coalescent methods based on this distribution start to break down (Figure S1) [36], [41], [43]. In order to predict silent site diversity in the interference selection regime, we must therefore devise an alternate method. In the interference selection regime, the twin forces of genetic drift and genetic draft generate massive deviations from the predictions described above. Yet despite the complexity of these forces, the patterns of silent-site variability display a number of striking regularities in this regime, which we now demonstrate through simulations of our evolutionary model (see Methods). This approach is similar to earlier simulation studies [22], [25]–[29], but we focus on identifying patterns that can be exploited for prediction, rather than simply describing the behavior observed in the presence of interference. We later generalize these patterns and use them to establish a new coalescent framework for predicting genetic diversity when interference is common. First, we measured the average site frequency spectrum, , and the average fitness variance, , in 280 asexual populations evolving under our simple purifying selection model, where all mutations share the same deleterious fitness effect. These populations were arranged on a grid, with mutation rates (NU) ranging from 10 to 104 and selection strengths (Ns) ranging from 10−3 to 103. We distinguish between populations that fall in the background selection regime or the interference selection regime, which loosely coincide with the red and blue regions in Figure 2 (see Methods). Figure 3 shows the observed reduction in diversity, as measured by the pairwise heterozygosity π relative to its neutral expectation, . As expected, the reduction in diversity is well-approximated by Eq. (2) in the background selection regime (triangle symbols) [27], but it breaks down for populations in the interference selection regime (circles) [37]. In addition, the traditional measure of the deleterious load ceases to be a good predictor of diversity in the interference selection regime, with more than an order of magnitude variation in for the same value of λ. However, when the same populations are reorganized according to their variance in fitness (Figure 2 B), the pattern essentially flips. The variance in fitness within the population is a strikingly accurate predictor for in the interference selection regime (circles), but it is a poor predictor in the background selection regime (triangles). The distortions in the site frequency spectrum are illustrated in Figure 4. The top left panel depicts a typical site frequency spectrum in the interference selection regime, using parameters consistent with the fourth (dot) chromosome of Drosophila melanogaster (see Methods). Apart from an overall reduction in polymorphism, the most prominent features of this frequency spectrum include an excess of rare alleles [22], [29], and a non-monotonic (or “U-shaped”) dependence at high frequencies [44]. Since we only include silent mutations in Figure 4, the distortions in the site frequency spectrum are entirely determined by distortions in the genealogy of the sample (Figure 4 B). The excess of rare alleles is due to an increase in the relative length of recent branches, compared to more ancient ones, and the non-monotonic behavior arises from imbalance in the branching structure of the tree [22]. In the right three panels of Figure 4, we show how these distortions vary over the broad range of parameters depicted in Figure 3. For clarity, we only include populations in the interference selection regime, and we focus on the two particular features of the site frequency spectrum discussed above (the full site frequency spectra for all of the populations in Figure 3 are shown in Figure S2). Figures 2C and 2D show the excess of rare alleles as measured by the reduction in average minor allele frequency and Tajima' s D respectively. These distortions cannot be explained by any constant, including the background selection limit. Similarly, Figure 4 E shows a measure of the non-monotonic or “U-shaped” dependence at high frequencies, using the statistic. In this case, deviations from neutrality () reflect topological properties of the genealogy, which cannot be explained even by a time-dependent. Ref. [45] showed that a “U-shaped” frequency spectrum cannot arise in any exchangeable coalescent model [e. g. , [37], [46], [47]] unless it also allows for multiple mergers. Together, the simulations in Figure 4 show that even simple models of purifying selection can generate strong distortions in the silent site frequency spectrum, and that these distortions can persist even when individual mutations are only weakly deleterious (Ns∼1). Yet the most striking feature of these distortions is not simply that they exist, but rather that they are extremely well-predicted by the reduction in pairwise diversity in these populations — which is itself well-predicted by the variance in fitness. This strong correlation is a nontrivial feature of interference selection, and it disappears for the populations that were classified into the background selection regime (Figure S3). Figure 4 also shows that correlations persist when we repeat our simulations with nonzero rates of recombination. As long as there is a sufficient density of selected mutations per unit map length, recombination seems to modify only the degree of the distortions from neutrality, while the qualitative nature of the distortions remains the same. Together, Figures 3 and 4 suggest an approximate “collapse” or reduction in dimensionality from our original four-parameter model to a single-parameter curve. The evidence so far is merely suggestive, so we will revisit the generality of this result in the following sections. Yet if such a collapse exists, it carries a number of practical benefits for predicting genetic diversity in the interference selection regime: if we can predict, we can in principle predict all of the relevant patterns of silent site variability (e. g. , the site frequency spectrum) even when these quantities significantly deviate from the neutral expectation. We will exploit this idea to our advantage below. However, this increased predictive capacity places fundamental limits on our ability to resolve individual selection pressures from patterns of silent site variability, even in this highly idealized setting. Our simulations suggest that in the interference selection regime, two asexual populations with the same variance in fitness will display nearly identical patterns of silent site variability, regardless of the fitness effects of the nonsynonymous mutations. The patterns that emerge from the simulations in Figures 3 and 4 reflect a fundamental limit of our evolutionary model, similar to the familiar background selection limit. To demonstrate this, we restrict our attention to nonrecombining genomes (R = 0), which leads to a key simplification: different genotypes with the same fitness are completely equivalent, both in terms of their reproductive capacity and their potential for future mutations. The evolutionary dynamics are completely determined by the proportion, f (X), of individuals in each fitness class X. The frequency of a mutant allele at some particular site can be modeled in a similar way, by partitioning f (X) into the contributions and from the ancestral and derived alleles. These fitness classes evolve according to the Langevin dynamics (4) where is the mean fitness of the population and is a Brownian noise term [48]–[52]. Equation (4) decomposes the change in the frequency of the derived allele into the deterministic action of selection and mutation, and the random effects of genetic drift. It represents a natural extension of the standard diffusion limit for genomes with a large number of selected sites. Crucially, Eq. (4) tracks only the fitnesses of the mutant offspring as they accumulate additional mutations. The advantage of this description is that it can be analyzed with standard perturbative techniques. For example, while the background selection limit is not always motivated in this fashion, Eq. (2) arises as a formal limit of the dynamics in Eq. (4) when (Text S1). To avoid the trivial behavior, where selection can be entirely neglected, we must also take so that the deleterious load λ (and therefore) remains constant. In this limit, molecular evolution is completely determined by λ, or equivalently by, which represents the fraction of mutation-free individuals in the population. The collapse observed in the left panel of Figure 3 indicates that populations quickly converge to this limit when is large but finite. Inverting this line of reasoning, a similar collapse in the right panel of Figure 3 suggests convergence to a second, infinitesimal limit when. Of course, if Ns vanishes on its own we simply recover the neutral result, . To maintain nontrivial behavior, Figure 2 B shows that we must take as well, so that the variance in fitness (and therefore) remains constant. In this way, the infinitesimal limit resembles a linked version of the infinitesimal trait models from quantitative genetics, where phenotypic variation (in this case, for the fitness “trait”) arises from a large number of small-effect alleles. The evidence from Figure 3B is merely suggestive, but we can establish the infinitesimal limit more rigorously using Eq. (4), where it corresponds to the limit that and with the product held constant. In Text S2 we demonstrate this by rescaling the moment generating function for Eq. (4); it can also be shown term-by-term using the perturbation expansion from Ref. [52]. This latter approach provides some intuition for the origin of the control parameter. Specifically, in a nearly neutral population (), the variance in fitness is equal to (5) which is the average mutational spread that accumulates during the coalescent timescale. The only way that this quantity can remain finite as is if the product is held fixed. This argument also shows that extension of the infinitesimal limit to a distribution of fitness effects is straightforward, provided that we replace with. In this infinitesimal limit, the distribution of fitnesses within the population and the patterns of molecular evolution depend only on the product and not any other properties of ρ (s). The effects of beneficial and deleterious mutations are symmetric [44], so our analysis also applies to the long-term balance between beneficial and deleterious substitutions in finite genomes [53]. In the infinitesimal limit, selected mutations are negligible on their own, and are virtually indistinguishable from neutral mutations, but the population as a whole is far from neutral. Rather, infinitesimal mutations arise so frequently that the population maintains substantial variation in fitness, and this leads to correspondingly large distortions at the sequence level. The distribution of fitnesses within these populations is well-characterized by “traveling wave” models of fitness evolution [49], [54]–[57], which provide explicit formulae for the variance in fitness (Nσ) as a function of the control parameter (Text S2). These formulae show that Nσ increases monotonically with, so either quantity can be used to index populations in the infinitesimal limit. We will use Nσ for the remainder of the paper in order to maintain consistency with Figure 3. Note that because of the pervasive interference between selected mutations, σ2 is typically much smaller than the deterministic prediction from Eq. (1), , and for large Nσ it grows less than linearly with the number of loci under selection. Unfortunately, patterns of molecular evolution are less well-characterized in this limit, which makes it difficult to predict the correlations observed in Figures 3 and 4. A complete description has been obtained only in the special cases where or. The former corresponds to a neutral population, with small corrections calculated in Ref. [52]. The latter case was recently analyzed in Ref. [44], which showed that the genealogy of the population approaches that of the Bolthausen-Sznitmann coalescent [58]. In this limit, silent site diversity decays as, while the shape of the site-frequency spectrum, , becomes independent of all underlying parameters. However, Figure 4 shows that many biologically relevant parameters fall somewhat far from these extreme limits, so we require an alternate method to predict genetic diversity for the moderate values of Nσ that are likely to arise in practice. In the absence of an exact solution of the infinitesimal limit, we employ an alternate strategy inspired by the simulations in Figures 3 and 4. Convergence to the infinitesimal limit is extremely rapid in these figures — so rapid that we can effectively neglect any corrections to this limit all the way up to the boundary of the background selection regime. In other words, the structured coalescent and the infinitesimal limit are both approximately valid along this boundary. Thus, instead of using the infinitesimal limit to approximate a population with a given Nσ, this rapid convergence suggests that we could also use a population on the boundary of the background selection regime with the same Nσ. Intuitively, this resembles a “coarse-graining” of the fitness distribution, since it approximates several weakly selected mutations in the original population with a smaller number of strongly selected mutations in the background selection regime. On a formal level, this is nothing but a patching method [59] that connects the asymptotic behavior in the infinitesimal () and background selection () limits. This intuition suggests a simple algorithm for predicting genetic diversity in the interference selection regime: (i) calculate Nσ as a function of Ns and NU as described in Text S2, (ii) find a corresponding point on the boundary of the background selection regime with the same Nσ, and (iii) evaluate the structured coalescent at this corresponding point. Step (ii) requires a precise definition of the boundary between the interference and background selection regimes, which we have not yet specified. Like many patching methods, this boundary is somewhat arbitrary, since the transition between the interference and background selection regimes is not infinitely sharp. Previous studies have identified several candidates (see Text S3), but in general this definition must balance two competing goals. The boundary should be close enough to the background selection limit to minimize errors in the structured coalescent. But at the same time, it must be close enough to the infinitesimal limit so that the populations rapidly converge. Our definition here is based on a specific feature of the structured coalescent, which is already evident from the first-order correction in Eq. (3). For each Nσ, the structured coalescent starts to break down near the point of maximum reduction in, which is also close to the crossover point where Muller' s ratchet starts to click more frequently [41], [50]. Together, these maxima define a “critical line” in the plane (Figure 5 A), which serves as the boundary between the interference and background selection regimes. Populations above or to the left of this line are classified into the interference selection regime, and the silent site variability in these populations can be predicted from the coarse-graining algorithm above. The remaining populations belong to the background selection regime, where the structured coalescent is already valid. We have implemented this coarse-graining procedure in a freely available Python library (see Methods), which rapidly generates predictions for the site frequency spectrum for arbitrary combinations of Ns and NU, and implements the linkage block approximation for recombining genomes described below. Other common diversity statistics (e. g. , MAF or Tajima' s D) can be computed from this site frequency spectrum as desired. Concrete examples of these predictions for the reduction in pairwise diversity are shown in Figure 5. We see that the coarse-grained predictions accurately recover the transition to the neutral limit when (Figure 5 B), and they also reproduce the power-law decay in heterozygosity when (Figure 5 C). We note that similar predictions in Figure 4 C–E (red lines) reproduce the observed distortions in the frequency spectrum statistics, while Figure 6 illustrates the predictions for the full shape of the frequency spectrum for the specific parameter combination in Figure 4 A. As is apparent from the figures, there is a broad range of parameters where the coarse-grained predictions are significantly more accurate than either the neutral expectation or the limit studied in Ref. [44]. In order to illustrate the transition between the interference and background selection regimes, we have focused on the simplest case where all selected mutations confer the same deleterious fitness effect. However, many of our results extend to more realistic scenarios where mutations are drawn from a distribution of fitness effects (DFE). In this case, it is useful to partition the fitness effects into a weakly selected category () and a strongly selected category (), with an intermediate zone separating these two regimes (Figure 7). If the DFE is entirely contained in the weakly selected region, then our previous analysis can be easily extended. Recall that the infinitesimal limit exists for arbitrary DFEs, provided that we replace s with the root mean squared effect in each of the expressions above. In other words, the patterns of diversity in the infinitesimal limit are equivalent to a single-s DFE with an effective selection coefficient. We can therefore obtain predictions for arbitrary by computing and applying our coarse-graining procedure to this corresponding single-s population, and we expect similar accuracy as long as the original population is sufficiently close to the infinitesimal limit. As an example, we use this procedure in Figure 6 to calculate the shape of the site frequency spectrum for a few representative DFEs consistent with the Drosophila dot chromosome parameters in Figure 4 A. We plot overall levels of diversity for a broader range of parameters in Figure S4. These figures illustrate the accuracy of our coarse-graining method for several different DFE shapes. While this single-s mapping applies when all the mutations are sufficiently weak, there are other possible scenarios where a single effective selection strength is clearly inappropriate. For example, deleterious mutations in natural populations often span several orders of magnitude [60], which could lead to scenarios where the DFE contains a mixture of weakly and strongly selected mutations. A full treatment of this case is beyond the scope of the present paper, but we can illustrate the basic features with the help of a simple example. Suppose that the DFE contains two deleterious fitness effects: (i) a weakly deleterious mutation which occurs at rate and (ii) a strongly deleterious mutation which occurs at rate. Taken individually, these mutations belong to the interference and background selection regimes, respectively. Yet the combined DFE does not belong to either regime, since it is fundamentally a mixture of the two. On the one hand, this population must fall outside of the background selection regime because the two-effect generalization of the structured coalescent [41], [61] breaks down (Figure S5). At the same time, this population cannot belong to the interference selection regime because the patterns of diversity differ from a more weakly selected population (e. g. , , , ,) with similar variance in fitness (Figure S5). Nevertheless, our coarse-graining procedure provides a way out of this impasse by transforming the weakly selected mutations into a form that can be handled by the structured coalescent. In this case, we note that the strongly selected mutations primarily influence the weakly selected mutations through a reduction in the effective population size, . At this smaller population size, the weakly selected mutations generate a smaller variance in fitness than they would in the absence of the strongly selected mutations. Given this smaller fitness variance, we can use our single-s coarse graining procedure above to map the weakly selected mutations to a population on the critical line (as defined in the single-s case) with effective parameters and. Then we can predict the patterns of diversity using the two-effect generalization of the structured coalescent, where the two effects are the strongly deleterious mutation, , and the coarse-grained weakly deleterious mutation, (Figure S5). Of course, this simple two-effect example is almost as artificial as the single-s limit above. Ideally, we would like to generate predictions for arbitrary distributions of fitness effects. In general, we expect the qualitative behavior of these distributions to resemble the two-effect model above. Imagine for example that the DFE contains several weakly selected deleterious fitness effects and a single strongly selected effect. In this case, the weakly selected mutations can be combined into a single root-mean-square effect, , and the two-effect example above then applies. If on the other hand there are several strongly selected effects, we can account for them using a higher-dimensional structured coalescent. However, in the most general case where there is a continuous distribution of fitness effects, some additional complications arise. In this case, weakly selected mutations can still be coarse-grained to the infinitesimal limit, while those mutations that are sufficiently far into the strong selection regime () influence the evolutionary dynamics primarily through a reduction in the effective population size, . For the weakly selected mutations, this will tend to produce a smaller fitness variance and therefore a smaller deviation from neutrality than one would expect in the absence of the strongly selected mutations. However, a smaller also pushes more of the strongly selected mutations into the weak selection regime, which will tend to increase the fitness variance and the corresponding deviations from neutrality. Due to these competing factors, the division between “weak” and “strong” mutations will strongly depend on the population size, the mutation rate, and the precise shape of the DFE. In addition, there may also be mutations in the intermediate region that are too strong for the infinitesimal limit to apply, but still weak enough to bias allele frequencies. For a discrete DFE, the effects of these mutations can be predicted using the structured coalescent in the appropriate number of dimensions. However, no analogous structured coalescent framework presently exists for a continuous DFE. This remains an important avenue for future work. We note that our discussion has also ignored the effects of strongly beneficial mutations, which have been analyzed in several related studies [51], [62]–[66]. Unlike in the strongly deleterious case, where larger fitness effects have a smaller influence on diversity, strongly beneficial mutations tend to dominate the evolutionary dynamics if they are sufficiently common [51], [62], [64]. In this case, larger population sizes generate increased fitness variation with larger number of selected polymorphisms, and the patterns of silent site variability rapidly approach those attained in version of the infinitesimal limit [65], [66]. So far, our analysis has focused on nonrecombining genomes, but our simulations in Figure 4 show that similar behavior arises when R>0 as well. A formal analysis is more difficult in this case, since recombination requires explicit haplotype information and cannot be recast in terms of the evolution of fitness alone. Thus, while the structured coalescent has been extended to recombining genomes [42], [61], and an analogous version of Eq. (2) has been derived [34], [35], (6) there is no simple analogue of Eq. (4) that we can use to formally extend the infinitesimal limit. Nevertheless, we can gain considerable insight with a simple heuristic argument, which leverages our previous analysis in nonrecombining genomes. Neighboring regions of a linear chromosome recombine much less than the genome as a whole. Sites separated by a map length will typically not recombine at all in the history of the sample, so the ancestral process should predominantly resemble an asexual population on these length scales. On the opposite extreme, sites with will recombine many times in the history of the sample, and will effectively act as if they were unlinked [67]. To the extent that this transition is sharp, the evolution of a recombining genome can be viewed as a collection of independent, freely recombining linkage blocks, each of which evolves asexually. This simple heuristic has a long history in the population genetics literature [68], [69], and it underlies many of the “sliding window” techniques used to analyze polymorphism in long genomes [70]. If each block comprises a fraction of the genome, then the distribution of fitness and the patterns of molecular evolution within each block are by definition the same as an asexual population with an effective mutation rate (7) Strictly speaking, the unlinked blocks also contribute to a reduction in the effective population size [46], [67], [71], [72], but we follow Ref. [73] and neglect these effects here. Given the weak population size dependence in the interference selection regime, this is often a good approximation in practice. But in principle, the logarithmic corrections from unlinked blocks can become important in extremely large genomes with a large proportion of selected sites (see Text S4 or Ref. [73] for additional discussion). The block size itself must satisfy the condition that there are few recombination events within a block in a typical coalescence time, or (8) Here, is the pairwise coalescence time for the linkage block, which is itself a function of and can be calculated from Eq. (7) and the asexual methods above. Together, Eqs. (7) and (8) uniquely determine the block size in a given population. In practice, we use a generalized version of Eq. (8), , which accounts for constant factors and the saturation of the block size when. Using our coarse-grained predictions for, we can solve for and obtain explicit predictions for the molecular evolution in recombining genomes (see Methods). Ref. [73] has recently employed a similar argument to analyze an infinitesimal model analogous to the one studied here. They initially treat the maintenance of phenotypic (i. e. , fitness) diversity as a “black box, ” utilizing a top-down approach to calculate the decay of linked fitness variation caused by successive recombination events. Based on this analysis, they obtain predictions for the genetic diversity in the limit that the number of selected loci per block and the fitness variance per block become large, which, for an infinitely long genome, requires that (Text S4). For recombining genomes, this plays the role of the asexual limit analyzed in Ref. [44]. Similar to the asexual case, our present analysis extends the asymptotic results of Ref. [73] to more moderate parameter values where. Evidence from fine-scale recombination maps [74] suggests that these parameters may be relevant for regions of reduced recombination in the autosomes of obligate sexual organisms (e. g. , in humans, see Figure S6), in addition to nonrecombining sex chromosomes [29], [30] and highly selfing species such as C. elegans [75] where linked selection is already thought to play a large role. As an example, we utilize this linkage block approximation to calculate the relationship between diversity and local recombination rate in Figure 8 (predictions for other quantities, e. g. the rate of Muller' s ratchet, are discussed in Text S4). The reduction in minor allele frequency in particular provides a clear signature of natural selection that can be observed in human autosomal DNA (Figure S6) [7]. Interference clearly plays a large role for the populations in Figure 8, since the observed genetic diversity significantly deviates from the recombining structured coalescent [42] and the background selection limit in Eq. (2). In contrast, the crude approximation above is surprisingly accurate for these populations, even when U/R is of order one. This accuracy is especially surprising given that the predictions are obtained from an asexual population with a coarse-grained selection strength and mutation rate. Evidently, interference on a linear chromosome more closely resembles an asexual genome (with an appropriately defined length) rather than the freely recombining, single-site models that are more commonly employed. A more thorough investigation of the linkage block concept and its implications for other aspects of sequence diversity (e. g. , linkage disequilibria, variation in recombination rate, etc.) remain an important avenue for future work. Interfering mutations display complex dynamics that have been difficult to model with traditional methods. Here, we have shown that simple behavior emerges in the limit of widespread interference. When fitness variation is composed of many individual mutations, the magnitudes and signs of their fitness effects are relatively unimportant. Instead, molecular evolution is controlled by the variance in fitness within the population over some effectively asexual segment of the genome. This implies a corresponding symmetry, in which many weakly selected mutations combine to mimic the effects of a few strongly deleterious mutations with the same variance in fitness. We have exploited this symmetry in our “coarse-grained” coalescent framework, which generates efficient predictions across a much broader range of selection pressures than was previously possible. Our results are consistent with previous studies that have investigated interference selection in silico [22], [25]–[29], [44], but our coarse-grained model offers a different perspective on the relevant processes that contribute to molecular evolution in this regime. By using the term interference selection, we have tried to emphasize that interference (i. e. , correlations in the frequencies of selected alleles) is the distinguishing feature that separates these populations from the traditional background selection regime. Previous work, on the other hand, has argued that virtually all of the deviations from the background selection limit can be attributed to fluctuations in the fitness distribution and the effects of Muller' s ratchet [22], [41], [43]. Yet our coarse-grained framework includes neither of these complications directly, and the quantitative behavior is unchanged even when beneficial compensatory mutations balance the loss of fitness due to Muller' s ratchet. Moreover, fitness class fluctuations and the ratchet are arguably maximized in neutral populations [52], which are well-characterized by the neutral coalescent. Instead, our results show that we can capture many aspects of silent site diversity simply by correcting for the average bias in the fitness distribution away from the prediction in Eq. (1), similar to the findings of Ref. [47]. In order to predict this bias from first principles, it is crucial to account for correlations in the frequencies of selected mutations, similar to rapidly adapting populations [44], [65]. Of course, the degree of interference in any particular organism is ultimately an empirical question — one that hinges on the relative strengths of mutation, selection, and recombination. Although interference is often observed in microbes and viruses [76]–[79], its prevelance in higher sexual organisms is still controversial because it is difficult to estimate these parameters in the wild. Mutation and recombination rates can be measured directly (at least in principle), but population sizes and selection strengths can only be inferred from a population genetic model, and these have historically struggled to include the effects of selection on linked sites. Many estimates of “” ignore linkage by fiat (e. g. [80]) under the assumption that sites evolve independently. But these estimates become unreliable precisely when small- and intermediate-effect mutations are most common, and the reasons for this are apparent from Figure 4. All of the distortions in Figure 4 C and Figure 4 D would be mistakenly ascribed to demography (or in the case of Figure 4 E, population substructure), thereby biasing the estimates of selection at nonsynonymous sites. At best, these estimates of “” represent measurements of, which carry little information about the true strength of selection (Ns) or even the potential severity of interference. For example, all of the populations in Figure 8 have Ns = 10 and, even though they fall in the interference selection regime, and show a strong distortion in minor allele frequency that cannot be explained by Eq. (2). In other words, we cannot conclude that interference is negligible just because “”, as inferred from data, is larger than one. More sophisticated analyses avoid these issues with simulations of the underlying genomic model [7], [22], [29], [30]. In principle, this approach can provide robust estimates of the underlying parameter combinations that best describe the data. But in practice, simulation-based methods suffer from two major shortcomings which are highlighted by the symmetry above. We have seen that strongly-interfering populations with the same variance in fitness possess nearly identical patterns of genetic diversity. This suggests a degree of “sloppiness” [81] in the underlying model, which can lead to large intrinsic uncertainties in the parameter estimates and a strong sensitivity to measurement noise. A more fundamental problem is identifying the nearly equivalent populations in the first place. Even in our simplified model, large genomes are computationally expensive to simulate, and this obviously limits both the number of dependent variables and the various parameter combinations that can be explored in a single study. We have shown that sets of equivalent populations lie along a single line (namely, the line of constant Nσ) in the larger parameter space, which can easily be missed in a small survey unless the parameters are chosen with this degeneracy in mind. In this way, our theoretical predictions can aid existing simulation methods by identifying equivalent sets of parameters that also describe the data. As an example, we consider the D. melanogaster dot chromosome that inspired the parameter combination in Figure 4 A. Earlier, we showed that the reduction in silent site diversity on this chromosome () is consistent with the parameters Ns≈30, NU≈300, and NR≈0, which fall in the middle of the interference selection regime (Ref. [29], see Methods). Our calculations allow us to predict other parameter combinations with the same patterns of diversity, and we plot the simulated frequency spectrum for three of these alternatives in Figure 6. We see that even with highly resolved frequency spectra (unavailable in the original dataset), there is little power to distinguish between these predicted alternatives despite rather large differences in the underlying parameters. However, this “resolution limit” suggests that individual fitness effects are not the most interesting quantity to measure when interference is common. Individual fitness effects may play a central role in single-site models, but we have shown that global properties like the variance in fitness and the corresponding linkage scale are more relevant for predicting evolution in interfering populations. Estimating these quantities directly may therefore be preferable in practice. Our coarse-grained predictions provide a promising new framework for inferring these quantities based on allele frequency data or genealogical reconstruction. A concrete implementation presents a number of additional challenges, mostly to ensure a proper exploration of the high-dimensional parameter space, but this remains an important avenue for future work. Finally, our findings suggest a qualitative shift in the interpretations gleaned from previous empirical studies. We have provided further evidence that even weak purifying selection, when aggregated over a sufficiently large number of sites, can generate strong deviations from neutrality. Moreover, these signals can resemble more “biologically interesting” scenarios like recurrent sweeps, large-scale demographic change, or selection on the silent sites themselves. Here we refer not only to the well-known reduction in diversity and skew towards rare alleles, but also to the topological imbalance in the genealogy (or the “U-shaped” frequency spectrum), and the strong correlations in these quantities with the rate of recombination. Since weakly deleterious mutations are already expected to be common [60], they may constitute a more parsimonious explanation for observed patterns of diversity unless they can be rejected by a careful, quantitative comparison of the type advocated above. At the very least, these signals should not be interpreted as prima facie evidence for anything more complicated than weak but widespread purifying selection. Forward-time simulations were implemented in a custom C++ program using a discrete-generation Wright-Fisher algorithm. Each simulation started with a clonal population of N = 104 individuals with initial fitness W = 1, and subsequent generations were obtained by performing a reproduction step, a recombination step, and a mutation step. In the reproduction step, the new generation was formed by sampling individuals with replacement from the previous generation, weighted by the relative fitnesses. In the recombination step, we drew Poisson (NR) recombination events, and for each of these, we drew two individuals from the population and replaced the first individual with the recombinant offspring formed from a single randomly chosen crossover of the two chromosomes. Finally, in the mutation step, we drew Poisson (NU) nonsynonymous mutations, and for each of these, we drew an individual from the population and placed the mutation at a random location on the chromosome. The fitness effect of each mutation was drawn from the distribution of fitness effects, ρ (s), so that the fitness of the mutated individual was given by. Mutations at the neutral locus were handled similarly, except that these occurred with rate and were always placed at the exact center of the chromosome so that they could not recombine with each other. Starting at generation t = 0, each population was allowed to “burn-in” for Δt generations until the neutral locus developed a most recent common ancestor. After equilibration, we drew 100 independent samples of n individuals every Δt generations, and the site frequency spectrum was computed at the neutral locus. We also measured the average fitness of the population and computed the variance in fitness using Fisher' s fundamental theorem, , where v is the rate of fitness change (e. g. , due to Muller' s ratchet) which is estimated by. This process was continued for a total of 20N generations per population, and for 300 independent populations per parameter combination. Backward-in-time simulations of the asexual structured coalescent, the recombining structured coalescent, and the Bolthauzen-Sznitman coalescent were implemented as a set of custom C++ programs similar to Hudson' s ms [82]. To improve performance, neutral mutations were omitted, and the time to the next event was replaced with its expected value when calculating the average site frequency spectrum. Asexual coalescent simulations were evaluated 105 times for each parameter value, while the more computationally-demanding recombinant version was evaluated 104 times per parameter value. The boundary of the background selection regime was obtained by minimizing Eq. (3) as a function of Ns with held fixed. Numerical solutions were obtained by analytically differentiating Eq. (3) and inverting the stationarity condition using the Newton-Raphson algorithm in the SciPy library. See Text S3 for additional discussion. The coarse-grained parameters were obtained by calculating Nσ (as described in Text S2) and identifying the corresponding point on the boundary of the interference selection regime with the same value of Nσ (as described above). Coarse-grained predictions were obtained from structured coalescent simulations of the coarse-grained parameters, except for, which was approximated by numerical evaluation of Eq. (3). The effective linkage scale, , was obtained by inverting the condition (9) where denotes the coarse-grained prediction for in Eq. (3). Numerical solutions were obtained using the Brent algorithm in the SciPy library. See Text S4 for additional discussion. We have implemented the methods described above as a Python library, coarse_coal, which can be used to calculate coarse-grained parameters and frequency spectrum predictions for arbitrary combinations of Ns, NU, and NR in the interference selection regime. Our source code is available for download at https: //github. com/benjaminhgood/coarse_coal. Possible parameter combinations for the fourth (dot) chromosome of Drosophila melanogaster were obtained by reapplying the method of Ref. [29] for our simple purifying selection model. These authors estimated the reduction in diversity on the dot chromosome to be, based on sequence data containing approximately L∼5 kb of silent sites sequenced in each of n≈24 lines [83], [84]. The per-site heterozygosity is of order, which implies a silent mutation rate of. Based on these estimates for the sample size and, forward-time simulations of the parameters Ns = 30, NU = 300, and NR = 0 yield (mean ± s. d.), which is consistent with the observed reduction. Local recombination rates in Figure S6 were estimated from deCODE' s fine-scale genetic map [74], assuming an equal sex ratio and averaging over 1 Mb windows. The local mutation rate was approximated using a uniform point-mutation rate of per base pair per generation [85]. Average minor allele frequencies were estimated using the African SNPs identified in the low-coverage portion of the 1,000 Genomes Project [86]. We only included autosomal SNPs that fell within one of the 1 Mb windows identified above, and we excluded repetitive elements (RepeatMasker), RefSeq exons, and all SNPs that were absent or fixed within the African subpopulation or did not have a high-confidence ancestral allele.

A central goal of evolutionary genetics is to understand how natural selection influences DNA sequence variability. Yet while empirical studies have uncovered significant evidence for selection in many natural populations, a rigorous characterization of these selection pressures has so far been difficult to achieve. The problem is that when selection acts on linked loci, it introduces correlations along the genome that are difficult to disentangle. These “interference” effects have been extensively studied in simulation, but theory still struggles to account for interference in predicted patterns of sequence variability, which limits the quantitative conclusions that can be drawn from modern sequence data. Here, we show that in spite of this complexity, simple behavior emerges in the limit that interference is common. Patterns of molecular evolution depend on the variance in fitness within the population, and are only weakly influenced by the fitness effects of individual mutations. We leverage this “emergent simplicity” to establish a new framework for predicting genetic diversity in these populations. Our results have important practical implications for the interpretation of natural sequence variability, particularly in regions of low recombination, and suggest an inherent “resolution limit” for the quantitative inference of selection pressures from sequence polymorphism data.

Abstract Introduction Results Discussion Methods

ecological metrics mutation population size genetic polymorphism ecology effective population size natural selection genetics biology and life sciences population genetics evolutionary biology evolutionary processes genetic drift

2014

Genetic Diversity in the Interference Selection Limit

We analyzed the transcriptional signatures of mouse bone marrow-derived macrophages at different times after infection with promastigotes of the protozoan parasite Leishmania major. Ingenuity Pathway Analysis revealed that the macrophage metabolic pathways including carbohydrate and lipid metabolisms were among the most altered pathways at later time points of infection. Indeed, L. major promastiogtes induced increased mRNA levels of the glucose transporter and almost all of the genes associated with glycolysis and lactate dehydrogenase, suggesting a shift to anaerobic glycolysis. On the other hand, L. major promastigotes enhanced the expression of scavenger receptors involved in the uptake of Low-Density Lipoprotein (LDL), inhibited the expression of genes coding for proteins regulating cholesterol efflux, and induced the synthesis of triacylglycerides. These data suggested that Leishmania infection disturbs cholesterol and triglycerides homeostasis and may lead to cholesterol accumulation and foam cell formation. Using Filipin and Bodipy staining, we showed cholesterol and triglycerides accumulation in infected macrophages. Moreover, Bodipy-positive lipid droplets accumulated in close proximity to parasitophorous vacuoles, suggesting that intracellular L. major may take advantage of these organelles as high-energy substrate sources. While the effect of infection on cholesterol accumulation and lipid droplet formation was independent on parasite development, our data indicate that anaerobic glycolysis is actively induced by L. major during the establishment of infection. Leishmania, the causative agent of vector-borne diseases, known as leishmaniases, lives as an obligate intracellular parasite within mammalian hosts. The outcome of infection depends largely on the activation status of macrophages, the first line of mammalian defense and the major target cells for parasite replication. Understanding the strategies developed by the parasite to circumvent the macrophage defense mechanisms and to survive within these cells may help defining novel therapeutic approaches for leishmaniases. High-throughput techniques have allowed the simultaneous identification and analysis of thousands of genes or proteins modulated in response to host-pathogen interaction. Different previous studies have used microarray technology to investigate the responses of macrophages from human and mouse origins to Leishmania infection [1], [2], [3], [4]. Most of these studies have dealt essentially with established infection, and limited responses to various species of Leishmania were observed. To obtain a dynamic and informative picture of macrophage behaviour in response to Leishmania promastigotes, we investigated the mouse macrophage response to initial invasion of L. major over a time course that extended from one to 24 hours post-infection. As controls, we used heat-killed promastigotes infected macrophages to determine the genes and pathways actively regulated by Leishmania parasites. Transcripts significantly modulated by Leishmania infection over time were identified and a subset of these genes confirmed by reverse- transcription quantitative real-time PCR (RT-qPCR). Hierarchical Clustering was performed to identify gross gene expression features and Ingenuity Pathway Analysis (IPA) was used to flag the mouse biological pathways, networks, and functions significantly altered by Leishmania infection during the first 24 hours post-infection. Analysis of the microarray data presented here revealed that in addition to oxidative stress, immune responses, and inflammatory genes that have been widely described in previous works, the lipids and carbohydrates metabolic pathways are among the most relevant biological networks fitting our data set, modulated by L. major infection. Among those, anaerobic glycolysis was identified as one of the major pathway actively regulated by the parasite. Promastigotes of the L. major tunisian strain GLC94 (MHOM/TN/95/GLC94 zymodeme MON25) were grown at 26°C in RPMI 1640, supplemented with 5 mM L-glutamine, 10% heat inactivated foetal calf serum (Perbio science, Brebières, France), penicillin (100 U/ml) and streptomycin (100 µg/ml). Metacyclic rich fraction obtained using Ficoll gradient was used in all experiments. Briefly, stationary phase cultures of Leishmania were centrifuged at 5 000 g for 10 min at room temperature and resuspended in 2 ml of PBS. The cell suspensions were then loaded onto a Ficoll gradient composed, from the bottom of 2 ml of 20%, 5 ml of 10% and 5 ml of 5% Ficoll diluted in PBS. The gradient was next centrifuged at 1 300 g for 10 min at room temperature. The metacyclic promastigotes were recovered on the top of 10% Ficoll layer. BALB/c mice (Elevage Janvier) were killed and hind legs removed for bone marrow derived macrophages (BMDM) isolation. Briefly, femurs and tibias were flushed with RPMI 1640 using a 25-gauge needle. Contaminating erythrocytes were lysed through the addition of Gey lysis solution (ammonium chloride 1. 5M, EDTA 0. 1 mM, pH 7. 3). All cells were incubated in T75 culture flasks at 1. 5 106 cell per ml in RPMI 1640 media supplemented with 5 mM L-glutamine, 10% heat inactivated foetal calf serum, penicillin (100 U/ml) and streptomycin (100 µg/ml) and 80 ng/ml M-CSF (Peprotech, Neuilly sur Seine, France) overnight for stromal cell elimination. Non-adherent, immature macrophages were transferred to fresh culture-treated Petri dishes (Nunc, USA) and grown for 7 days, with re-feeding on day 3, to induce macrophage differentiation. Generated macrophages were assessed by flow cytometry for expression of F4/80 (around 90% were positive). All mouse work was done according to the directive 86/609/EEC of the European parliament and of the council on the protection of animals used for scientific purposes. Approval for mice experiments was obtained from the ethic committee of Institute Pasteur of Tunis with ethics approval number 1204. BMdM were incubated at a parasite to cell ratio of approximately 10∶1 with Ficoll purified metacyclic promastigotes of L. major. After the desired time of incubation at 37°C in 5% CO2, non-ingested parasites were removed and the cells were harvested to prepare samples. Standard Giemsa staining was used to determine the percentage of infected cells and to insure for homogenate cell infection under the different conditions. Macrophages were lysed directly in 1 ml Trizol reagent (Invitrogen). Total RNA from uninfected and infected macrophages were prepared using the RNeasy mini kit (Qiagen) and treated with DNase according to the manufacturer' s protocol. Extracted RNAs were stored at −80°C. RNAs were quantified using NanoDrop ND-1000 micro-spectrophotometer and RNA quality was monitored on Agilent RNA Pico LabChips (Agilent Technologies, Palo Alto, CA). 100 ng of RNA from each biological condition were amplified and labelled with biotin according to the GeneChip whole transcript sense target labeling assay manual and using the GeneChip WT cDNA Synthesis and amplification Kit and WT terminal labeling Kit. The fragmented ssDNA was checked on Agilent RNA Pico LabChips. The fragmented and labeled ssDNA was hybridized to the GeneChip Mouse Gene 1. 0 ST array (Affymetrix, Santa Clara, CA), washed with the Fluidics station 450 and scanned using the Affymetrix Scanner 30007G. QC analysis was performed before and after normalization using BoxPlot of total intensities, MAPlots for all replicates and PCAplots. All microarrays of this study passed the quality control. Cell intensity files were generated with GeneChipOperating Software (GCOS). Each infection and control time points were performed on three different samples, using different preparations of BMdMs, and processed independently to give three biological replicates. The data preprocessing step included intrachip and interchip normalisation and summarisation. The intrachip normalisation step corrects for the GC content of the probes, the interchip normalisation step reduces non-biological differences between chips and the summarisation step combines the probe intensities into single gene expression values. The data from BALB/c mouse are different to classical 3′-type Affymetrix chips as mismatch probe sets are not available. Since the annotation of the new “exon-like” mouse Affymetrix chip (Affymetrix/MoGene-1_0-st-v1) was not present in all databases, IDs had to be re-mapped on their Ensembl gene. The process is implemented completely in R (Version R-2. 7. 0) with the use of several BioConductor (BioC 2. 2) packages [5]. Probe sets are defined each by one Entrez gene. If the probe sequence matches uniquely to the gene and no SNP hits the probe segment of the DNA the probe is assigned to the Entrez gene. We used a probe-Entrez gene assignment from version 11. 0. 0 of Dai et al. [6]. (Entrez database was downloaded for human from 11. 03. 2008 and for mouse from 28. 06. 2008). After summarisation the Entrez identifer were directly mapped to Ensembl gene. For this mapping BioMart/EnsMart was used via the biomaRt package [7]–[9] in Ensembl v50. The processing relies on Ensembl genes and it is straightforward to use a probe-Ensembl gene assignment. At the time of processing, no assignment for HuGene and MoGene arrays to Ensembl genes was available from [6]. For the intrachip normalisation the ‘Model-based Analysis of Tiling-arrays’ (MAT) was implemented, similar to [10], since this method provides the most advanced GC correction for whole-transcript prepared samples. MAT is a probe affinity model which combines content and position dependency of probe sequences in a unified linear model. The parameters of the model are estimated from the control probes and subsequently the probe affinities are calculated for the perfect matches. On linear intensity scale, the probe intensities are divided by an estimated probe affinity. As a second step, we applied an interchip normalisation in form of a quantile normalisation to adjust the intensity distributions over the arrays, as this kind of normalisation appeared successful to reduce unwanted effects between 3′ expression arrays [11]. As a last step in the preprocessing of the data we applied a summarisation of probe intensities to a probe set expression. In this process, a median is computed over the intensities in two replicate dimensions: a) The different probe intensities within the probe set; b) The arrays with the same biological condition. This provides very robust summarised expression values also for low-replicate settings. Detection call p-values were computed for each probeset with a paired Wilcoxon signed rank test that compares probe intensity to control probes of similar GC content. More precisely, each probe is compared to the 75% quantile of the set of control probes with similar GC content. A gene probeset was called present when the corresponding FDR corrected p-value was below 5%. The sole threshold in this approach is the height of the quantile (75%) in the GC bin. The same probe-Entrez gene assignment and subsequent mapping was used as in the summarisation process. Expression analysis used the R Bioconductor package Limma to identify genes that met statistical (P<0. 05 after adjustment according to the method of Benjamini and Hochberg and fold-change criteria (at least a 1. 5-fold change) for differential expression using the following contrasts: macrophages infected with live parasites at a given time point versus non infected macrophages incubated with vehicule (media) for the same time. The same contrast was used for heat-killed Leishmania-infected macrophages. Macrophage genes modulated during the kinetics were first detected. In accordance with MIAME (Minimum Information About a Microarray Experiments) regulations, all data were deposited into GEO (Gene Expression Omnibus) database at www. ncbi. nlm. nih. gov/geo/ under the accession number GSE31995. Ingenuity Pathways Analysis (IPA; Ingenuity, CA; Systems 2008) is derived from known functions and interactions of genes published in the literature as well a set of canonical pathways and cellular Toxic molecules markers in the context of several studied diseases. The Ingenuity IPA Tool was used to identify the most significant macrophagic biological networks, cellular functions and canonical pathways altered by Leishmania parasite infection, based on a Fischer' s exact test to calculate a p-value for each biological function founded (at least P<0. 01). A list of the statistically significant differential genes expression in L. major-infected BALB/c macrophage for each time points of the kinetic was generated and mapped to their functional networks in the IPA database and ranked by score. The relationships between the generated networks and known pathways were then investigated using the canonical pathway analysis function. In addition, we applied 2 IPA analysis methods: an analysis of dataset corresponding to each time point of infection and a global analysis along the kinetic of infection that allowed us to identify the Top canonical and cellular pathways altered across all time points of the infection. The Affymetrix 1. 07 array data were analyzed using dChip software (http: //www. biostat. harvard. edu/~cli/dchip_2010_01. exe) in order to identify the marker genes cluters regulated later or earlier throughout the kinetic of Leishmania infection. Moreover, the dchip software was used to check the signature of live Leismania major parasite infection comparing with the Killed parasite one. RNA quantity, was controlled using NanoDrop ND-1000 micro-spectrophotometer and RNA quality and integrity (RNA Integrity Number, RIN>9) was monitored on Agilent RNA Pico LabChips (Agilent Technologies, Palo Alto, CA). Reverse transcriptions were performed for each of 96 mice samples in 20 µl final reaction volume with 273 ng of total RNA using 200 Units of SuperScript III enzyme (M-MLV RT, Invitrogen) and 250 ng of random primers according to manufacturer' s instructions (25°C 10 min, 42°C 50 min, 70°C 15 min). All RT reactions were performed the same day with same pipetor set and same manipulator. A negative control was included by performing a RT with no template. qPCR experiments were carried out using EVA Green chemistry on BioMark qPCR apparatus (Fluidigm) following manufacturer' s instructions. For each cDNA sample, a Specific Target Amplifications (STA) was performed with a pool of primers targeting all selected genes (Pre-Amplification of 14 cycles using TaqMan PreAmp Master Mix (Applied Biosystems) and following manufacturer' s instructions): Each qPCR was performed with 1/20 STA dilution, in duplicate. Relative gene expression kinetics was created by a first normalization with 4 reference genes followed by a second normalization with Non Infected macrophage cells (NI). Values are expressed in fold changes (2−Delta Delta Ct Method) [12] compared to NI macrophage cells. Cells extracts were obtained by adding 25 µl of lysis buffer containing 10 mM Tris-HCL pH 7,5, 50 mM NaCl, 50 mM sodium fluoride (NaF), 2 mM EDTA, 1 mM ethylene glycol bis (β-aminoethyl ether) -N. N; N′. N′-tetraacetic acid (EGTA), 2% Nonidet-P40 (NP-40), 0,75% sodium deoxycholate (DOC), 1 mM orthovanadate, 1 µg/ml aprotinine, 1 mM PMSF, 1 mM DTT. After 30 min incubation on ice, the extracts were centrifuged at 15000 rpm for 20 min. 25 µg of whole cell lysates were resolved by electrophoresis in a 12% SDS-polyacrylamide gel. Resolved proteins were electrophoretically transferred onto PVDF sheets (Hybond-P; Amersham) and membranes were blocked by incubation in 3% non-fat milk and 0,1%Tween in PBS for 1 h at room temperature followed by incubation with COX-2 antibody (BD Biosciences, France) in 0,1% tween-PBS. Bound antibody was detected by incubation with horseradish peroxidase-coupled secondary antibody (Amersham Pharmacia Biotech. , Buckinghamshire, U. K.). To ensure for equal loading, the blots were then stripped (62. 5 mM Tris (pH 6,8), 0,1M β-mercaptoethanol, 2%SDS) and re-probed with ERK1/2, antibody. All incubations and washes were done in 1× PBS. Macrophages were fixed in 4% paraformaldhehyde for 10 min at room temperature. For Lipid Droplet Staining, Bodipy 493/503 (Molecular Probes) was used as previously described with some modifications [13]. Briefly, cells were stained for 30 min at room temperature using a solution of Bodipy and DRAQ5 (Biostatus, Leicestershire, UK) that allows the staining of macrophage and parasite nuclei. All coverslips were mounted on slides with Fluoromount-G (Southern Biotechnology Associates). Otherwise, for free cholesterol accumulation, Filipin staining was used as previously described with some modifications [14], [15]. Coverslips were incubated in Glycine to quench paraformaldehyde and cells were then stained with Filipin and DRAQ5 for 2 h at room temperature. All coverslips were washed and mounted on slides with Fluoromount- G. Detailed analysis of lipid droplet accumulation was performed using an oil immersion Nikon Plan Apo 100 (N. A. 1. 4) objective mounted on a Nikon Eclipse E800 microscope equipped with a Bio-Rad Radiance 2000 confocal imaging system (Bio-Rad, Zeiss). Analysis of Filipin staining was visualized with the Eclipse TE 2000-U epifluorescence microscope equipped with Lambda DG-4 illumination. Intensity differences in Filipin staining were quantified using linescan analysis (Metamorph software). Our data show that the transcription of different cytokines and chemokines involved in the inflammatory response and cellular recruitment to the site of inflammation was significantly altered. Indeed, transcription of CXCL1 (up to 3 fold increase), CXCL2 (up to 3. 9 fold) and CXCL3 (up to 2 fold) (alias respectively GROα, GROβ, GROγ) was rapidly induced and persisted during the first 12 hours of infection. Consistent with previous studies [2], [16] and associated with the classical activation M1 phenotype, transcription of TNFα (up to 5), CXCL10 (IP10), CCL2 (MIP1α), and other chemotactic cytokines such as CCL3 (up to 6. 8) and CCL4 was upregulated following infection. By contrast, transcription of CCR2, the CCL2 receptor, was down-regulated while the mRNA expression of CCRL2 was strongly up-regulated. Leishmania infection also led to mRNA downregulation of other cytokines such as IL-1β. Moreover, M2 polarization-associated anti-inflammatory cytokines including IL-1Ra, and receptors such as CD36 were up-regulated in infected cells. Finally, transcription of both NOS2 (which characterizes M1 macrophage phenotype) (up to 5 fold) and arginase1 (ARG1) (which characterize M2 macrophage phenotype) (up to 4 fold) was increased in response to L. major promastigotes at different time points. Importantly, both NOS2 and ARG1 were actively induced by live parasite. Indeed, as assessed by qPCR, both genes were not transcribed in heat-killed promastigote-infected cells (data not shown). Infection also induced the expression of co-stimulatory molecules such CD40, CD83, CD86, adhesion molecules (CD38, Itga5, and ICAM1), and tissue invasion molecules such as MMP12 and MMP14. Among the most relevant pathways modulated by L. major infection, we found different metabolic pathways (Figure 2) including glycolysis, gluconeogenesis, tricarboxylic cycle, oxidative phosphorylation and pentose phosphate shunt. To identify host cell pathways modulated by Leishmania, DNA microarrays were used in several studies to monitor transcriptional changes in host cells following infection [1], [2], [4], [17]. As previously reported, we show here that L. major promastigotes induce the expression of genes encoding chemokines such as MIP-1alpha/CCL3, MIP-1beta/CCL4, potent chemotractants for monocytes/macrophages [2], and inflammatory mediators including TNFα, CCL5, CXCL1-3 [1]. We also observed an upregulation of CXCL10 gene expression, a Th1-mobilizing chemokine shown to be produced by lesion cells from self-healing cutaneous leishmaniasis (CL) -patients during active leishmaniasis [18], and the induction of the proinflammatory monocyte chemotractant protein 1 (MCP-1) encoded by the CCL2 gene, which has been associated to cutaneous leishmaniasis. Besides inflammatory mediators, L. major promastigotes also induced the transcription of genes normally associated to an M2 response such as arginase1. A similar hybrid macrophage activation profile that does not strictly fall into one of the two categories (classical and alternative) has been previously observed in macrophages infected with L. chagasi [3]. This inflammatory response of infected macrophages is rapidly induced and is likely triggered by the stimulation of the receptors implicated in the recognition of Leishmania parasites, as heat-killed promastigotes displayed similar effect on macrophage mRNA profile. Once the parasites get into the cells and start to multiply, the host cells seem to adapt their metabolism to face the infection. The microarray data indicated that Leishmania infection result in an enhanced rate of glycolysis but with reduced glucose flux through tricarboxylic acid cycle. The requirement to regenerate NAD+ to maintain glycolysis (i. e. , conversion of glyceraldehyde-3-phosphate to 1,3-bisphosphoglycerate) is accomplished by lactate dehydrogenase (LDH) -catalyzed reduction of pyruvate to lactate. Thus, infected macrophages would tend to convert glucose into lactate even in the presence of sufficient oxygen to support mitochondrial oxidative phosphorylation. Moreover, the down-regulation of a number of genes implicated in the TCA cycle and oxidative phosphorylation suggests that increased glycolysis may be the mechanism L. major relies on for energy production. Interestingly, expression of the glycolytic enzymes encoding genes, including LDH and PDK, was not induced when heat-killed promastigotes were used to infect BMdM, suggesting that L. major promastigotes actively induce macrophages to shift their metabolism to anaerobic glycolysis. Up-regulation of host glycolytic transcripts has been reported for macrophages infected by Toxoplasma gondii [19] and was shown to correlate with the activation of the HIF-1α transcription factor [20]. On the other hand, expression of HIF-1α has been observed in the cutaneous lesions of L. amazonensis infected BALB/c mice [21]. Stabilization of HIF1α in response to Leishmania infection may explain this increasing flux through the glycolytic pathway, the conversion of pyruvate to lactate and the suppression of the TCA cycle, and the oxidative phosphorylation observed in infected BMdM. Whether the effect of Leishmania on this metabolic pathway relies on HIF-1α activation is currently under investigation. Almost all the scavenger receptors such as CD36, expressed on several cell types including macrophages, deliver associated lipids to lysosomes. Then, essential enzymes for the cleavage of cholesteryl esters and triglycerides generate free cholesterol and fatty acids that are next released from lysosomes into the cytosol. The efflux of excess cholesterol is promoted by different transporters including the ATP binding cassette (ABC) gene family. Among those, ABCA1 plays a major role in this process and decreased expression of this transporter, as found in our microarray data, could lead to the inhibition of the cholesterol efflux. Moreover, our data indicate a down-regulation of CYP27 mRNA levels in Leishmania-infected cells, suggesting that infection results in an alteration of the host cell oxysterol content. Decreased levels of this natural Liver X Receptor (LXR) endogenous ligand in infected cells may explain the inhibition of LXR target genes expression such ABCA1 [22], [23]. Moreover, stearoyl-CoA desaturase, whose expression was induced by Leishmania, inhibits ABCA1-mediated cholesterol efflux [24]. Epifluorescence microscopy confirmed the microarray data, indicating that L. major infection modulates macrophage cholesterol content and induces its accumulation into the cytoplasm. Cholesterol plays a key role during the infection process by several intracellular pathogens, including adhesion and internalization [25], as well as for survival within host cells as reported for the protozoan parasite Toxoplasma gondii [26], [27]. In Leishmania-infected cells, cholesterol depletion [28] or sequestration [29] reduces the extent of infection by promastigotes. The role of cholesterol accumulation into L. major-infected cells remains to be defined and whether this accumulation is induced by other Leishmania species remains to be investigated. L. major infection also promotes macrophage-derived foam cell formation. This was suggested by the analysis of microarray data and was validated by the staining of infected cells with Bodipy that clearly demonstrated the accumulation of lipids droplets into the cytoplasm. Interestingly, these lipid droplets are mainly localized in the proximity of the parasitophorous vacuoles. The induction of foam macrophages has been observed during infection by different bacteria such as Mycobacterium, Chlamydia [30] or parasites including Trypanosoma cruzi [31], Toxoplasma gondii [32], [33], and Plasmodium falciparum [34]. Moreover, a recent study revealed that cytoplasmic LDs are translocated across the inclusion membrane into the lumen of the parasitophorous vacuole (PV) in Chlamydia infected cells [35]. A similar mechanism of lipid acquisition by Leishmania may exist that allows the parasite to take advantage of this high-energy source. From another point of view, accumulation of LD may be implicated in the cross-presentation of antigens derived from PVs containing live parasites. Indeed, accumulation of lipid bodies has been reported to be required for normal and efficient cross-presentation of OVA-coated latex beads by dendritic cells [36]. Interestingly, expression of a member of the IRG gene family (Irgm3) that has recently been implicated in this pathway was induced in Leishmania-infected BMdM. Both live and heat-inactivated Leishmania promastigotes induce the transcription of the same set of genes involved in the intracellular cholesterol accumulation and foam cell formation in bone marrow-derived macrophages suggesting that the development of the parasite is not required. Accumulation of cholesterol and formation of lipids bodies may be achieved through the stimulation of different receptors such as Toll-like receptors (TLRs) that have been implicated in the recognition and control of Leishmania parasites [37], [38], [39]. Indeed, a recent study revealed that TLR stimulation impairs macrophage cellular cholesterol efflux in vivo [40]. Moreover, in vitro activation of TLR-3 and TLR-4 by microbial ligand blocks the induction of LXR target genes including ABCA1 [41]. Similarly, the formation of lipid bodies in response to bacteria is dependent on TLR (and particularly TLR2) signalling [42] and more generally the activation of TLRs by pathogen-derived agonists promotes lipid accumulation [43], [44]. TLR may thus provide an important link between lipid metabolism, infection and the innate immune response. In inflammatory cells, lipid bodies have been associated to arachidonic acid metabolism and eicosanoid-forming enzymes have been localized in lipid bodies that are sites for 5-LO- and COX-derived eicosanoid synthesis [45]. Moreover, lipid bodies have been reported as intracellular domains for eicosanoid synthesis in vivo. Consistent with previous studies [17], [46], we show here that despite the induction of significant COX-2 mRNA levels the COX-2 protein was not detected in L. major infected macrophages. The control of gene expression in eucaryotes is subjected to dynamic regulation in the cell. This control is a multi-step process that includes transcription, splicing, translation and post-translational regulation. We have thus to take in mind that besides transcription, different other levels of control, may take place in L. major-infected macrophages and influence the level of biologically active protein. Collectively, our results suggest that L. major promastigotes push the macrophages to shift toward anaerobic glycolysis and induce the accumulation of cholesterol and the formation of foam cells. These metabolic changes occurring in host cells appear to be induced by a large number of pathogens, and are likely to play an important role in pathogenesis.

Leishmania are obligated intracellular pathogens that develop almost exclusively in macrophages. Experimental leishmaniasis in mice is one of the most extensively studied models of intracellular infections both at the level of the parasite and host immune responses. We took advantage of Balb/c mice model to investigate gene expression profile through Affymetrix oligonucleotide arrays. In order to have a general and dynamic picture of the complex biological events that are acting in the context of Leishmania intracellular parasitism, we investigated the mouse macrophage response to initial invasion of L. major over a time course that extended from one to 24 hours post-infection. Our results reveal the alteration of several biological processes and metabolic changes. Indeed, similarly to different other pathogens, Leishmania induces cholesterol accumulation and foam cell formation that have been confirmed by confocal microscopy experiments. Whether Leishmania parasites take advantage of this high-energy source is now under investigation. Our findings provided further understandings in host responses to Leishmania infection.

Abstract Introduction Materials and Methods Results Discussion

microarrays genomics molecular cell biology gene expression biology computational biology microbiology host-pathogen interaction molecular biology genetics and genomics

2012

Transcriptomic Signature of Leishmania Infected Mice Macrophages: A Metabolic Point of View

Mycobacterium tuberculosis (Mtb), possesses at least three type VII secretion systems, ESX-1, -3 and -5 that are actively involved in pathogenesis and host-pathogen interaction. We recently showed that an attenuated Mtb vaccine candidate (Mtb Δppe25-pe19), which lacks the characteristic ESX-5-associated pe/ppe genes, but harbors all other components of the ESX-5 system, induces CD4+ T-cell immune responses against non-esx-5-associated PE/PPE protein homologs. These T cells strongly cross-recognize the missing esx-5-associated PE/PPE proteins. Here, we characterized the fine composition of the functional cross-reactive Th1 effector subsets specific to the shared PE/PPE epitopes in mice immunized with the Mtb Δppe25-pe19 vaccine candidate. We provide evidence that the Mtb Δppe25-pe19 strain, despite its significant attenuation, is comparable to the WT Mtb strain with regard to: (i) its antigenic repertoire related to the different ESX systems, (ii) the induced Th1 effector subset composition, (iii) the differentiation status of the Th1 cells induced, and (iv) its particular features at stimulating the innate immune response. Indeed, we found significant contribution of PE/PPE-specific Th1 effector cells in the protective immunity against pulmonary Mtb infection. These results offer detailed insights into the immune mechanisms underlying the remarkable protective efficacy of the live attenuated Mtb Δppe25-pe19 vaccine candidate, as well as the specific potential of PE/PPE proteins as protective immunogens. It is estimated that two billion people are latently infected with Mycobacterium tuberculosis (Mtb), and this huge reservoir is sustaining the pan/epidemic spread of the bacterium. Due to the relative inefficiency of the Mycobacterium bovis BCG (Bacille Calmette-Guerin) vaccine in preventing these latent infections becoming active tuberculosis (TB) disease cases in adolescents and adults, new improved TB vaccines are warranted [1]. Mtb harbors five chromosomal esx clusters of highly conserved genes, which code for specialized type VII secretion systems (T7SSs), some of which are also conserved in other mycobacteria [2]. Some of these systems are dedicated to the export/secretion of key mycobacterial factors and play a determinant role in host-pathogen interaction [3]. Several lines of evidence indicate that the construction of genetically modified mycobacterial strains expressing WT and/or mutated variants of these specialized T7SSs is a promising strategy to set up new live attenuated TB vaccines [4–7]. We have previously shown that BCG complemented with the esx-1 gene cluster (BCG: : ESX-1), produces and secretes the 6 kDa-Early Secreted Antigenic Target (ESAT-6, EsxA) and its partner, the 10 kDa-Culture Filtrate Protein (CFP-10, EsxB) thereby inducing specific host immune responses (S1A and S1B Fig) that ultimately confer improved protection against an Mtb challenge in animal models, relative to the parental BCG strain [4,8, 9]. Sweeny and colleagues generated a recombinant Mycobacterium smegmatis strain harboring the orthologous Mtb esx-3 region, which displays an improved protective efficacy compared to BCG [6]. More recently, we developed an attenuated Mtb esx-5 mutant, i. e. , Mtb Δppe25-pe19, lacking the five esx-5-coded pe/ppe genes, as a promising vaccine candidate [5,10]. Most of the esx loci contain clusters of genes coding for members of the PE/PPE protein families. These proteins are named after their characteristic N-terminal Pro-Glu (PE) or Pro-Pro-Glu (PPE) motifs and are unique to the mycobacterial species. The genome of the Mtb H37Rv strain contains 99 pe and 69 ppe genes, which most probably evolved from ancestral esx-associated pe/ppe genes [11]. Although the function of most PE/PPE proteins remains to be unraveled [12–14], some of them have been demonstrated to play a role in mycobacterial virulence, being involved in mycobacterial growth in macrophages and/or in the mouse infection model, or in modulation of mycobacteria-mediated inhibition of phagosome maturation [10,15–21]. PE/PPE proteins display numerous repetitive sequences and possess abundant immunogenic regions, representing a rich source of B and T cell epitopes [22]. The esx-5 region of Mtb (rv1782-rv1798) harbors 2 pe (pe18,19) and 3 ppe (ppe25,26,27) genes (S1A Fig). The corresponding PE18,19, and PPE25,26,27 proteins are exported/secreted through the transmembrane channel of the ESX-5 secretion apparatus, involving the ESX-Conserved Component EccD5 [10] (S1B Fig, right). In addition, many other non-esx-5-associated PE/PPE proteins with various degrees of sequence similarity with their esx-5-coded homologs, are also exported/secreted via the ESX-5 system [3,23,24]. Our recent observation that the Mtb Δppe25-pe19 strain, devoid of the five esx-5-coded pe/ppe genes, is attenuated for growth in immune-competent and SCID mice, indicates a role in virulence of these PE/PPE proteins [5,10]. Importantly, as the transmembrane channel EccD5 is unaffected/intact, the Mtb Δppe25-pe19 strain continues to be able to export PE/PPE proteins, which are encoded outside the esx-5 locus. As a results, IFN-γ+ CD4+ T-cell responses are induced against a plethora of non-esx-5-coded PE/PPE homologs in the immunized host. The involved T cells, via their high cross-reactivity, recognize esx-5-coded PE/PPE virulence-related factors (S1B Fig, right). Therefore, due to the expression of a functional EccD5-associated transmembrane channel and an intact ESX-5 T7SS, the Mtb Δppe25-pe19 strain shows the unique property to induce cross-reactive T-cell immunity against the esx-5-associated PE/PPE virulence-related factors, despite their absence in this strain [3,5]. Epitope mapping of the PE/PPE proteins in bovines also revealed that the highly immunogenic nature of PE/PPE immunogens is essentially driven by a substantial degree of cross-reactivities in the elicited T cells, which results from the sequence homologies among the PE/PPE proteins [25,26]. An Mtb eccD5 KO strain, largely deficient in PE/PPE protein secretion, does not phenocopy the Mtb Δppe25-pe19 strain and is markedly less protective in vaccination assays performed in the mouse model [5]. This observation strongly suggests that immunity to PE/PPE proteins is a relevant requisite for an efficient protection against TB. The distribution of the functional CD4+ T-cell subsets defines the quality of the adaptive immune response in infectious diseases including TB [27] and several reports indicate that, at least in animal preclinical models, poly-functional CD4+ T cells mediate protection [28]. Therefore, we here characterize at the single-cell level the functionality and some aspects of differentiation status of the cross-reactive PE/PPE-specific Th1 cells induced by Mtb Δppe25-pe19 immunization and evaluated the contribution of PE/PPE-specific T cells in the protective immunity against pulmonary Mtb infection in mice. These experiments provided new insights on the potential of PE/PPE proteins as protective immunogens. Moreover, the Mtb Δppe25-pe19 mutant is able to secrete ESX-1 substrates and thereby elicits CD4+ T-cell responses against these protective immunogens. In addition to its particular T-cell antigenicity, the Mtb Δppe25-pe19 exhibits unique properties to trigger the host innate immunity. Unlike BCG, the expression of a functionally active ESX-1 system enables the Mtb Δppe25-pe19 vaccine candidate to induce phagosomal membrane rupture and thereby establishing a phagosome-cytosol communication inside phagocytes, a phenomenon which has instrumental consequences on the activation of innate immunity [29–32]. These results elucidate part of the immune properties of the remarkable protective capacity of the live attenuated Mtb Δppe25-pe19 vaccine candidate. We previously identified two groups of PPE25- and PE19-derived MHC-II (I-Ab) -restricted T-cell epitopes. One group is highly specific to Mtb esx-5-encoded PE/PPE proteins and show no homologies with other PE/PPE (S1A Table), while the second group contains epitopes which are shared with PE/PPE homologs coded outside of esx-5 (S1B Table) [5]. Immunization of C57BL/6 (H-2b) mice with the attenuated Mtb Δppe25-pe19 strain confirmed and extended our previous finding that this strain is unable to induce Th1 immunity, i. e. , IL-2, TNF-α, and IFN-γ responses, against the esx-5-specific PE/PPE epitopes (S2 Fig). However, this attenuated vaccine candidate preserved its capacity to induce Th1 immunity against the PE/PPE homologs coded outside of esx-5, due to the expression of a functional ESX-5 transmembrane channel associated to EccD5 [33,34]. Such cross-reactive Th1 cells specific to the shared PE/PPE epitopes recognize the ESX-5-associated and virulence-related PE/PPE proteins, which are absent in the Mtb Δppe25-pe19 strain. To delineate the effector mechanisms of PE/PPE-specific T-cell immunity, we next subjected such T-cell responses to a fine analysis of the functional CD3+ CD4+ Th1 subsets by IL-2-, TNF-α-, and IFN-γ-specific IntraCellular Staining (ICS). We first set up the strategy for the PPE25: 1-20-specific responses in the spleen of Mtb Δppe25-pe19-immunized mice (Figs 1A and S3A and S3B). This shared epitope is representative of the identified PE/PPE peptides listed in the S1 Table. This approach allowed the determination of the frequencies of total antigen specific Th1 cytokine-producing cells (Fig 1B), as well as the definition of seven functional subsets, which are single, double, or triple positive for the expression of these key Th1 cytokines and their percentages compared to total CD4+ T cells (Fig 1C). Moreover, as (i) human Mtb-specific memory CD4+ T cells are enriched in the T-cell population expressing the chemokine receptors CCR6 and CXCR3 [35,36], (ii) the Programmed Cell Death-1 (PD-1) marker is associated with proliferative potential, self maintenance, IFN-γ production and protection in the context of anti-mycobacterial immunity [37,38], and (iii) CD27 expression is a pertinent marker to distinguish different Th1 effector subsets [39], we also performed, together with ICS, simultaneous surface staining with these markers in order to characterize the differentiation status of the antigen-specific, functional Th1 subsets whose the numbers were high enough to allow such analysis (Fig 1D). For instance, in this framework, most of the TNF-α+ single positive cells were CCR6- CXCR3- and PD1-, only a few percentages of IFN-γ+ single positive or TNF-α+ IFN-γ+ double positive cells were CCR6+ CXCR3+ and CD27- PD1+, while the triple positive Th1 cells contained the highest percentages of CCR6+ CXCR3+ and CD27- PD-1+ cells. To get mechanistic insights towards the fine composition of functional Th1 cells generated subsequent to vaccination with Mtb Δppe25-pe19, we first performed such detailed comparative analyses in the groups of Mtb Δppe25-pe19- or Mtb WT-immunized mice (Fig 2A and 2B). We observed that the profile of the Th1 responses specific to the different shared PE/PPE epitopes studied was overall similar. In the Mtb Δppe25-pe19-immunized mice, different degrees of PE/PPE-specific functional T subsets were present, which ranged from very small percentages of IL-2+ TNF-α- IFN-γ- (green), predominant amounts of IL-2- TNF-α+ IFN-γ- (blue) to intermediate percentages of IL-2- TNF-α- IFN-γ+ (yellow) single positive Th1 cells. Moreover, IL-2+ TNF-α+ IFN-γ- (dark blue) and IL-2+ TNF-α- IFN-γ+ (white) double positive Th1 cells were barely found, while intermediary levels of IL-2- TNF-α+ IFN-γ+ (purple) double positive and IL-2+ TNF-α+ IFN-γ+ (red) triple positive Th1 subsets were detected. In these mice, the cumulated numbers of IL-2- TNF-α+ IFN-γ+ or IL-2+ TNF-α+ IFN-γ+ cells specific to the totality of the shared PE/PPE epitopes can be estimated at 7. 6 x 105 and 5. 4 x 105 splenocytes per mouse, respectively. Comparison with the Mtb WT-immunized mice (Fig 2B) showed that the composition of the functional effector Th1 subsets specific to different shared PE/PPE epitopes were globally similar to those found in the Mtb Δppe25-pe19-immunized group, except for the frequencies of the terminally differentiated IL-2- TNF-α+ IFN-γ- (blue) cells, which decreased for the benefit of the terminally differentiated IL-2- TNF-α- IFN-γ+ (yellow) population (Fig 2C). This finding suggests that the virulence/persistence of WT Mtb might fine-tune such functional switches. As expected, no such Th1 subsets against the esx-5-specific PE/PPE epitopes were detected in the Mtb Δppe25-pe19-immunized mice, in contrast to the responses found in the WT Mtb-immunized mice (S4 Fig), which displayed characteristics that were similar between esx-5-associated and non-esx-5-associated (Fig 2B) epitopes. In the Mtb Δppe25-pe19- or Mtb WT-immunized groups, we detected comparable frequencies of CCR6+ CXCR3+ or CD27- PD-1+ cells in the TNF-α+ and IFN-γ+ single positive, TNF-α+ IFN-γ+ double positive and IL-2+ TNF-α+ IFN-γ+ triple positive functional Th1 subsets, specific to the representative PPE25: 1–20 shared epitope (Fig 3A and 3B). These results showed that the differentiation status of the functional Th1 subsets were very similar subsequent to immunization with the Mtb WT- or Mtb Δppe25-pe19. Therefore, compared to the WT Mtb, the Mtb Δppe25-pe19 strain induces a similar range of differentiated cross-reactive Th1 effectors specific to the shared PE/PPE epitopes, which also recognize the ESX-5-associated PE/PPE virulence-related factors, with very slight differences in the proportions of TNF-α+ or IFN-γ+ single positive cells. One of the most relevant properties of the Mtb Δppe25-pe19 candidate vaccine is its capacity to secrete ESX-1 virulence determinants ESAT-6 and CFP-10 [40], while displaying a strongly attenuated phenotype relative to parental H37Rv Mtb [5,10]. Virulence comparison test in SCID mice showed that Mtb Δppe25-pe19 was slightly more virulent compared to BCG Danish (S5 Fig). The Mtb Δppe25-pe19 attenuation profile resembles that of BCG strains belonging to the DU2 IV group (BCG Phipps, BCG Frappier, BCG Pasteur, BCG Tice), which also showed somewhat elevated virulence relative to BCG Danish in a recent comparative study of 13 BCG strains [41,42]. We further characterized the fine-tuned Th1 immunity specific to ESAT-6 in Mtb Δppe25-pe19- or WT Mtb-immunized mice. Compared to the PE/PPE-specific responses, the levels of ESAT-6-specific Th1 cytokine released by splenocytes were generally stronger in both groups (S2 Fig). The distribution of ESAT-6-specific Th1 subsets (Fig 4A–4C) was distinct from that of PE/PPE-specific Th1 subsets (Fig 2A–2C). Comparatively, the ESAT-6-specific response was characterized by decreased percentages of IL-2- TNF-α+ IFN-γ- (blue) single positive cells for the benefit of IL-2- TNF-α+ IFN-γ+ (purple) double positive cells. This suggests that the distribution of Th1 subsets can vary as a function of the antigen specificity following vaccination, probably linked—among others—to the different level of expression and secretion patterns of these different antigens. In addition to ESAT-6- and CFP-10-specific responses, Δppe25-pe19-immunized mice mounted strong T-cell responses against EspC (Rv3615c), another ESX-1 substrate (Fig 4D), which is also considered as a protective immunogen [43]. In addition to Th1 cells specific to ESX-1/ESX-5-related antigens, other properties of the Mtb Δppe25-pe19 strain, including ESX-1-mediated triggering of innate immunity [29–32,44,45], may also take part in the improved protective capacity of this strain. A major characteristic feature of ESX-1-proficient mycobacteria is their capacity to induce phagosomal rupture in infected host cells [46,47], which is followed by the activation of numerous pathways of innate immune responses. These include the cytosolic translocation of mycobacterial DNA, detected by the host cyclic GMP-AMP synthase (cGas) and activation of the STING/TBK/IRF3 pathway. This process leads to the production of IFN-β, as well as the activation of Absent In Melanoma 2 (AIM2) inflammasome/caspase-1 pathway, which contributes to the release of active IL-1β [29–32,44,45] and IL-18-mediated noncognate IFN-γ production [48]. Using a FRET method, based on the accessibility of the intrinsic β-lactamase activity of the phagocytosed mycobacteria to the host cytosol [46,47], we demonstrated that the Mtb Δppe25-pe19 strain, in contrast to BCG, is able to induce such phagosomal rupture (Fig 4E). Unlike BCG, the Mtb Δppe25-pe19 strain induced secretion of IFN-β by the infected macrophages, albeit at a lesser extent than the virulent Mtb WT strain (Fig 4F). Similarly, the Mtb Δppe25-pe19 strain induced significantly more IL-1β release than the BCG strain (Fig 4G). These important properties may also take part in the previously reported improved protective capacity of the Mtb Δppe25-pe19 strain in comparison to BCG [5]. By comparative immunological investigation of the Mtb Δppe25-pe19 and Mtb eccD5 KO strains, we previously showed that the former induces robust cross-reactive CD4+ T cells against ESX-5-associated PE/PPE and also against a plethora of other PE/PPE antigens, while the latter, which is largely deficient in PE/PPE export/secretion, induces no T-cell response to the panel of PE/PPE epitopes that we selected (S1 Table) [5]. Consistent with our previous observations [5], the Mtb Δppe25-pe19 strain displayed a better protective potential than the Mtb eccD5 KO strain (Fig 5A). Based on this observation, we hypothesized that Th1 immunity to PE/PPE antigens may contribute to the cellular mechanisms of TB protection. To experimentally test this hypothesis and to directly evaluate the contribution of PE/PPE-specific Th1 cells in the protection, we established an immunization protocol to induce PE/PPE-specific Th1 responses, not with the live attenuated Mtb Δppe25-pe19 vaccine, with complex multifaceted immunological properties (Fig 4), but by use of PE/PPE-derived synthetic peptides (S1 Table). C57BL/6 mice (n = 3 per group) were immunized s. c. twice at a 10-day interval with each of the individual PE/PPE-derived peptides. As adjuvant TLR9 agonist CpG oligodeoxynucleotide, associated with the liposomal transfection reagent DOTAP (N-[1- (2,3-DioleOyloxy) ]-N, N, N-TrimethylAmmonium Propane methylsulfate) was used. At day 10 after the second injection, antigen-specific production of IL-2, TNF-α, and IFN-γ by CD4+ T splenocytes was readily detected (Fig 5B). However, for all of the epitopes, the amounts of these cytokines produced by the splenocytes of the peptide-immunized mice were weaker than the levels produced by their mycobacteria-immunized counterparts (S2 Fig). Results from an ICS assay performed in mice immunized with each peptide (Fig 6A–6C) showed increased frequencies of IL-2+ TNF-α- IFN-γ- (green) single positive, IL-2+ TNF-α+ IFN-γ- (dark blue) double positive and IL-2+ TNF-α+ IFN-γ+ (red) triple positive Th1 cells for each epitope, compared to the frequencies observed in their Mtb Δppe25-pe19-immunized counterparts (Fig 2C). Moreover, the terminally differentiated TNF-α+ single positive cells constituted the major Th1 cell subset in these peptide-immunized mice (Fig 6A–6C). The PE/PPE-specific Th1 cells induced by peptide or Mtb Δppe25-pe19 immunization notably displayed the following functional and phenotypic features: (i) Fluorescence Intensities (MFI) of the ICS staining for each Th1 cytokine, which are proportional to the amounts of cytokine production per cell (Fig 6D), and (ii) the differentiation status of the TNF-α+ and IFN-γ+ single positive, TNF-α+ IFN-γ+ double positive and IL-2+ TNF-α+ IFN-γ+ triple positive functional Th1 subsets, in terms of CCR6, CXCR3, CD27 and PD-1 surface expression (S6 Fig and Fig 3). We further evaluated directly the contribution of the anti-PE/PPE poly-specific Th1 cells, systematically and locally induced by use of the PE/PPE-derived synthetic peptides, in the protection against virulent Mtb. For a better understanding of the protective adaptive immunity in terms of fine specificity of T cells, we compared the protective potential of Th1 cells either specific to the esx-5-associated PE/PPE epitopes or to the shared epitopes. C57BL/6 (n = 6 per group) mice were vaccinated according to the protocol schematized in the Fig 7A with individual PE/PPE peptides (for prior immune response study) or mixtures of such peptides (for protection studies). PE/PPE peptides which harbor esx-5-specific or shared epitopes were formulated in CpG (DOTAP). Moreover, since previous works demonstrated that mucosal local immunity and the homing of effector T cells from the lung vasculature to the parenchyma is crucial for the pulmonary TB protection [49–52], the mice were also boosted i. n. with homologous PE/PPE peptides 10 days before the challenge. In the pulmonary CD4+ T-cell compartment of these mice, we detected in ex vivo tests increased percentages of CD27- CD62L-, CCR6+ CXCR3+, CD27- PD-1+ (Fig 7B), as well as CD44hi (S7A Fig) cells, representing a hallmark of migratory antigen-specific poly-functional effector T cells of the peripheral tissues. The protective potential of immunization with the peptide mixtures was compared to that of vaccination with BCG 1173P2 Pasteur strain. At day 40, vaccinated mice or untreated controls were aerosol infected with the virulent Mtb H37Rv strain, delivered at dose of ≈ 200 CFU/lungs. At day 70, determination of the mycobacterial loads in the lungs (Fig 7C) and spleen (Fig 7D) showed that immunization with the mixtures of PE/PPE peptides, either specific to the esx-5 region or shared with other homologs, induced a significant protection, which was only partially due to the effect of the adjuvant alone. BCG has an intact ESX-5 secretion system and induces T-cell immunity against all the selected PE/PPE epitopes (S7 Fig). We thus evaluated the effect of BCG priming and PE/PPE boosting on immune responses and TB protection following immunization with PE/PPE epitopes, as detailed in the Fig 8A In the lungs of these mice, as determined ex vivo, we observed increased percentages of CD27- CD62L-, CCR6+ CXCR3+, CD27- PD-1+ (Fig 8B), and CD44hi (S7B Fig) cells, within the CD4+ T-cell compartment, as well as increased total numbers of CD4+ T cells (S8C Fig). BCG priming followed by PE/PPE boosting significantly improved the control of mycobacterial growth in the lungs (Fig 8C) and limited the mycobacterial dissemination to the spleen (Fig 8D). Altogether, our results highlight the protective capacity of these PE/PPE proteins as immunogens and unravel part of the immune mechanisms of the remarkable protective property of the Mtb Δppe25-pe19 vaccine candidate. In the present study, we demonstrated that PE/PPE-specific Th1 cells contribute to the cellular protective immune mechanisms developed by the live attenuated Mtb Δppe25-pe19 TB vaccine candidate that we recently generated [5,10]. Secretion or export to the bacterial cell-envelop is a prerequisite for most mycobacterial antigens to access the antigen presentation machinery inside the host innate immune cells and for specific detection by effector CD4+ T cells [3]. A large number of PE/PPE proteins are exported/secreted via the ESX-5 T7SS [24], although a few PE/PPE proteins might also be handled via the SecA general secretory pathway [53]. The biological activities of PE/PPE proteins are thus likely linked to their cell surface-associated or extracellular localization, which may also explain their notable immunogenicity [14,54–58]. It has been proposed that the duplication and random insertion of the pe/ppe genes throughout the Mtb genome may have led to their transcriptional control by a random assortment of unrelated promoters and regulators, which could result in substantial degrees of variability in their expression profiles during different phases of infection [59]. Besides, there exist compelling sequence homologies among the PE/PPE members resulting from gene duplication. This context may lead to the consecutive availability of groups of PE/PPE epitopes during various phases of infection, despite the variability in the expression profiles of the PE/PPE members from which they derive. Such properties may contribute to the interest of these proteins in the protective immunity against chronic mycobacterial infections. The attenuated Mtb Δppe25-pe19 strain is deficient only in five PE/PPE proteins, i. e. , PPE25-27 and PE18-19, which are coded inside the esx-5 region [5,10]. However, due to the intact secretion machinery of the ESX-5 system, the Mtb Δppe25-pe19 strain has preserved the capacity to export/secrete substantial numbers of other PE/PPE proteins encoded elsewhere in the Mtb genome (S1B Fig, right) [24]. Immunization with the Mtb Δppe25-pe19 strain thus induces T-cell immunity against these PE/PPE proteins, including the non-esx-5-associated members, which display compelling sequence homologies with the missing esx-5-coded PE/PPE antigens [5]. Therefore, the antigenic repertoire of the Mtb Δppe25-pe19 strain remains virtually comparable to that of the WT Mtb strain despite its strongly attenuated virulence phenotype. Immune correlates of TB protection remain elusive. So far, in human, there is no consensus whether the induction of poly-functional Th1 cells and the distribution of the various Th1 subsets are markers of either active TB disease or of protective immunity in latent TB infection [60–62]. Single positive IL-2+ Th1 cells are usually central memory T cells, able to proliferate and differentiate to effector memory and/or effector cells, while single positive IFN-γ+ or TNF-α+ Th1 cells are terminally differentiated, not proliferative and short-lived populations [28]. The accumulation of TNF-α+ single positive cells is considered as predictor of diagnosis of active TB [63]. Consistently, it is admitted that in chronic diseases like TB, the continuous antigenic stimulation of T cells leads to the loss of both memory potential and poly-functionality, which results in terminally differentiated T cells that only produced IFN-γ or TNF-α. In mice, poly-functional IL-2+ TNF-α+ IFN-γ+ Th1 responses against prominent mycobacterial immunogens cells display a positive correlation with proliferative capacity, indicative of their effector capacity. In the mouse model, these cells are considered the most reliable parameter able to control the growth and dissemination of Mtb in vivo [27]. Here, we showed that immunization of mice with the Mtb Δppe25-pe19 strain induces marked IL-2- TNF-α+ IFN-γ+ double positive and IL-2+ TNF-α+ IFN-γ+ triple positive poly-functional Th1 effector cells specific to a panel of PE/PPE epitopes. Notably, most PE/PPE-specific triple positive Th1 splencoytes in the Mtb Δppe25-pe19-immunized mice exhibited a CXCR3+ CCR6+ PD-1+ phenotype, as a hallmark of effector memory and protective T-cell population. [35–38]. Expression of the PD-1 inhibitory receptor by Th1 cells has been recently shown to be of utmost important in the TB protection via the negative regulation of IFN-γ-over-expressing CD4+ T cells [38]. We further compared, in mice immunized with the Mtb WT or the Mtb Δppe25-pe19 strain, the profiles of functional Th1 subsets specific to PE/PPE epitopes, which are either highly specific to the esx-5 region or shared with PE/PPE homologs coded outside of esx-5 [5]. As expected, the Mtb Δppe25-pe19 strain does not induce T-cell responses against the first group of epitopes. However, the responses were comparable in the Mtb Δppe25-pe19- or Mtb WT-immunized groups against the shared PE/PPE epitopes in terms of their fine composition of Th1 effector subsets and their differentiation status. Therefore, despite its attenuation, the Mtb Δppe25-pe19 strain generates bi- and poly-functional Th1 cells, which recognize the PE/PPE antigens that it lost, with the diverse Th1 subsets distributed comparably to Mtb WT. We further demonstrated that PE/PPE-specific Th1 responses contribute actively to the anti-TB immunity. This is shown by the induced Th1 cells, as well as the recruitment and activation of effector T cells in the lungs, following systemic and local immunization of mice with selected PE/PPE epitopes, formulated in CpG (DOTAP) adjuvant. Importantly, the mycobacterial PE/PPE epitopes, either esx-5-sepcific, with no homologies with other PE/PPE proteins, or shared with PE/PPE homologs coded outside of esx-5, induce similar levels of protection. Therefore, it can be proposed that the shared surrogate PE/PPE homologs in the Mtb Δppe25-pe19 strain compensate for the loss of the esx5-specific epitopes. In addition, booster immunization with such PE/PPE epitopes in BCG-primed individuals significantly improved the protection. These results thus show that such PE/PPE proteins represent potent immunogens to be included in TB subunit vaccines or as boosters. Despite its attenuated phenotype, the Mtb Δppe25-pe19 strain remains able to induce Th1 responses against ESX-1-associated virulence factors, including at least ESAT-6, CFP-10 and EspC, known as protective immunogens [40,43,64–69]. Moreover, the presence of a functional ESX-1 system preserves the capacity of the Mtb Δppe25-pe19 strain at inducing ruptures in the phagosomal membrane inside the host phagocytes. The phagosomal rupture results in a phagosome-cytosol communication, leading to the release of mycobacterial compounds, including the extracellular mycobacterial DNA, to the host cytosol. Mycobacterial DNA is then sensed by cGas and ultimately activates IFN-β gene transcription [29–32]. We showed that Mtb Δppe25-pe19- (but not BCG-) infected macrophages secrete IFN-β. On the other hand, detection of mycobacterial DNA by the cytosolic AIM2 inflammasome increases caspase-1 activation and contributes significantly to the release of mature IL-1β [70]. It has been shown that following the ESX-1-dependent phagosomal rupture, the ESX-5 T7SS, via still unknown mechanisms, activates inflammasome and caspase-1, which results in IL-1β release [71]. It is noticeable that the ESX-5 mutant Mtb Δppe25-pe19 strain is only deficient in five PE/PPE esx-5-associated proteins, and still harbors a functional ESX-5 system. This property seemingly confers to this strain an increased capacity to activate inflammasome and IL-1β release. Even though the role of the type-I IFN in the induction of protective immune responses remains elusive, that of IL-1β is instrumental in the anti-mycobacterial host defense [72]. Collectively, the immunological fine characterization presented in this study emphasizes the unique properties of Mtb Δppe25-pe19 strain to stimulate host immunity in terms of both antigenic repertoire and innate immune responses. While the safety profile for the Mtb Δppe25-pe19 strain in SCID or immune-competent mice [5,10] is within the range of BCG strains [41,42], work is in progress to introduce a second attenuating gene deletion in order to satisfy the Geneva Consensus recommendations for novel live TB vaccines [73]. This process shall provide an Mtb Δppe25-pe19 derivative with increased security and safety, but unaffected advantageous immunological profile, to be used as a new veterinary or human TB vaccine. Mtb Δppe25-pe19 [10] and Mtb WT H37Rv strains were grown in Dubos broth, complemented with Albumine, Dextrose and Catalase (ADC, Difco, Becton Dickinson, Le Pont-de-Claix, France). The bacterial contents were determined by OD measurement at 600 nm. CFU were counted on Middlebrook 7H11 solid Agar medium after 18 days of incubation at 37°C. All experiments with pathogenic mycobacteria were performed in an L3 protection level laboratory, in accordance with the hygiene and security recommendations of Institut Pasteur. The synthetic peptides which contain MHC-II-restricted antigenic epitopes were synthesized by PolyPeptide Group (Strasbourg, France), reconstituted in H2O containing 5% dimethyl sulfoxide (DMSO) (Sigma-Aldrich, France) and stored at -20°C. Six-to-eight week-old female C57BL/6 (H-2b) mice (Janvier, Le Genest-Saint-Isle, France) were immunized by s. c. injection, at the base of the tail, of 1 x 106 CFU/mouse of Mtb Δppe25-pe19 or Mtb WT strains in 200 μ1 volume. Immunizations with peptides were performed by two s. c. injections at a 10-day interval, with 100 μg/mouse of individual peptide, formulated with 30 μg of CpG 1826 oligodeoxynucleotides as adjuvant (Sigma-Aldrich, France), 60 μl of liposomal transfection reagent DOTAP (N-[1- (2,3-DioleOyloxy) ]-N, N, N-TrimethylAmmonium Propane methylsulfate, Roche, France) and 10 μl of Opti-MEM (Life Technologies, France) in a final volume of 200 μl. The liposomal transfection reagent DOTAP optimizes the adjuvant effect of CpG by conducting it to the endosomal sites where the intracellular TLR-9 receptor is localized. Studies in immunocompetent mice were performed in agreement with guidelines of the European and French guidelines (Directive 86/609/CEE and Decree 87–848 of 19 October 1987), after approval by the Institut Pasteur Safety, Animal Care and Use Committee and under local ethical committee protocol agreement # CETEA 2013–0036 and CETEA 2012–0005. Analysis of virulence in SCID mice was approved by the UK Home Office (HO) regulations for animal experimentation which requires a HO-approved licence and approval from local ethical committees of Public Health England, Porton Down (Licence number PPL30/2704) and London School of Hygiene and Tropical Medicine (LSHTM) Animal Welfare and Ethical Review Board (Authorization # 70/6934). Splenocytes from immunized mice were cultured in flat-bottom 96-well plates (TPP, Denmark) at 1 x 106 cells per well in HL-1 medium (Biowhittaker, Lonza, France), complemented with 2 mM GlutaMax (Invitrogen, Life Technologies, France), 5 x 10−5 M β-mercaptoethanol, 100 U/ml penicillin and 100 μg/ml streptomycin (Sigma-Aldrich, France) in the presence of 10 μg/ml of individual peptides. After 12,48 and 72 hours of incubation at 37°C and 5% CO2, IL-2, TNF-α and IFN-γ were respectively quantified in the culture supernatants by ELISA as previously described [8]. Monoclonal antibodies (mAbs) specific to IL-2 (clone JES6-1A12 for coating and clone JES6-5H4 for detection) or IFN-γ (clone AN-18 for coating and clone R4-6A2 for detection) were from BD Pharmingen (Le pont-de-Claix, France). Anti-TNF-α mAbs (clone 1F3F3D4 for coating and clone XT3/XT22 for detection) were from eBioscience. Single-cell suspensions from spleen of immunized mice were obtained by tissue dissociation, homogenization and passage through 100 μm-pore filter. Cells were cultured at 7. 5 x 106 cells/well in the presence of 1 μg/ml anti-CD28 (clone 37. 51) and 1 μg/ml of anti-CD49d (clone 9C10-MFR4. B) mAbs (BD Pharmingen) together with 1 x 106 cells/well syngenic bone-marrow dendritic cells, loaded with 10 μg/ml of homologous or control peptide during 1h, followed by 5h of incubation with Golgi Plug (BD Pharmingen), according to manufacture’s instructions. Cells were then harvested, washed twice with PBS containing 3% Fetal Bovine Serum (Invitrogen, Life Technologies, France) and 0. 1% NaN3 (FACS buffer) and incubated for 15 minutes at 4°C with FcγII/III receptor blocking anti-CD16/CD32 (clone 2. 4G2) mAb. Cells were then incubated for 25 minutes with appropriate dilutions of AlloPhycoCyanin (APC) -eFluor780-anti-CD3ε and PB-anti-CD4 mAbs (BD Pharmingen) at 4°C and sheltered from light. The stained cells were washed twice in FACS buffer, permeabilized by use of Cytofix/Cytoperm kit (BD Pharmingen). Cells were then washed twice with PermWash 1X buffer from the Cytofix/Cytoperm kit and incubated with appropriate dilutions of PerCP-Cyanine5. 5-anti-IL-2 (clone JES6-5H4, eBioscience), PE-anti-TNF-α (clone 554419, BD Pharmingen), and Alexa Fluor647-anti-IFN-γ (clone XMG1. 2, eBioscience) mAbs during 30 minutes at 4°C. Appropriate staining with control Ig isotypes was performed in parallel. Cells were subsequently washed twice in PermWash buffer, once in FACS buffer and then fixed with 4% paraformaldehyde overnight at 4°C. When indicated, cells were stained at the surface, either ex vivo or after in vitro simulation before ICS, with a cocktail of (APC) -eFluor780-anti-CD3ε, PB-anti-CD4, PE-Cy7-anti-CCR6 (Sony Biotechnology), FITC-anti-CXCR3 (eBioscience), FITC-anti-PD-1 (Biolegend) and PE-Cy7-anti-CD27 (BD Pharmingen) mAbs. We preliminarily checked that the expression of CCR6, CXCR3, CD27 and PD-1 markers did not change during the short in vitro stimulation required for ICS. To study the phenotype of the pulmonary T cells, lungs were first disaggregated by treatment with 400 U/ml type IV collagenase and DNase I (Roche). Following a 45-min incubation at 37°C, single-cell suspensions were prepared by use of GentleMacs (Miltenyi) and passage through 100-μm nylon filters (Cell Strainer; BD Falcon). Cell suspensions were then enriched in lymphocytes by 20-min centrifugation at 3000 rpm at RT on Ficoll gradient medium (Lympholyte M, Cedarlane Laboratories). The cells were then washed twice and stained with a cocktail of (APC) -eFluor780-anti-CD3ε, PB-anti-CD4, PE-anti-CD27, PE-Cy7-anti-CD62L, PE-anti-CD44 (eBioscience) and FITC-anti-CD45RB (eBioscience) mAbs in the presence of FcγII/III receptor blocking mAb. The stained cells were then fixed with 4% paraformaldehyde. The cells were acquired in an LSR Fortessa flow cytometer system by use of BD FACSDiva software (BD Bioscience). Data were analyzed using FlowJo software (Treestar, OR, USA). PMA-differentiated THP-1 cells were infected with Mtb WT or Mtb Δppe-25-pe19 strains at MOI of 1. At day 3 p. i. , the phagosomal rupture was assessed by Fluorescence Resonance Energy Transfer (FRET) assay as previously described [47]. Briefly, cells were stained with 8 μM CCF-4 (Cephalosporin core linking a 7-hydroxyCoumarin to a Fluorescein) (Invitrogen) in EM buffer (120 mM NaCl, 7 mM KCl, 1. 8 mM, CaCl2,0. 8 mM MgCl2,5 mM glucose and 25 mM Hepes, pH 7. 3) complemented with 2. 5 μM probenecid, during 1h at room temperature. Cells were then washed in PBS and stained with APC-anti-CD11b (BD Pharmingen) mAb in FACS buffer. Cells were then fixed with 4% paraformaldehyde overnight at 4°C and were analyzed in a CyAn cytometer (Beckman Coulter, France). Human IL-1β and IFN-β were quantified in the culture supernatants of these infected THP-1 cells at 24h p. i. by use of (DY201-05, R&D Systems) and (41410, PBL Assay Science) kits, respectively. Six-to-eigth week-old female C57BL/6 mice were left untreated or were immunized s. c. with 1 x 105 CFU/mouse of BCG (1173P2 Pasteur strain) at day 0 or immunized s. c. twice at days 10 and 20, with 50 μl of each PE/PPE-derived peptide of interest, 30 μg of CpG, 60 μl of DOTAP and 10 μl Opti-MEM contained in 200 μl/mouse. At day 30, peptide-immunized mice received via intra-nasal route under anesthesia 20 μg of each PE/PPE-derived peptide of interest, 20 μg of CpG, 10 μl of DOTAP and 3 μl Opti-MEM contained in 20 μl/mouse. For anesthesia, mice received i. p. 100 μl/mouse of suspension containing weight-adapted quantities of Imalgène1000 (Kétamine, i. e. , 100 mg/kg, Merial, France) and Rompun 2% (Xylazine solution, 10 mg/kg, Bayer, Germany), prepared in physiological solution. Mice were challenged 10 days after the last immunization by use a homemade nebulizer via aerosol. Five ml of a suspension containing 2. 5 x 106 CFU/ml of Mtb H37Rv WT strain were aerosolized to reach an inhaled dose of ≈ 200 CFU/mouse, as determined by day 1 p. i. CFU count in the lungs of the challenged mice. The infected mice were placed and manipulated in isolator in A3 protection-level facilities at Institut Pasteur. One month later, the lungs and spleen of the infected mice were individually homogenized by using a MillMixer organ homogenizer (Qiagen, Courtaboeuf, France). Serial 5-fold dilutions were plated on 7H11 Agar medium supplemented with ADC (Difco, Becton Dickinson). The CFU were counted after 18–21 days of incubation at 37°C. The statistical analyses were performed by use of GraphPad Prism software (GraphPad Software, La Jolla, CA, USA) and Mann-Whitney test for simple comparison or One Way ANOVA test with Tukey’s correction for multiple comparisons in order to determine the statistical significance of obtained data. ID numbers for proteins mentioned in the text according to Tuberculist (http: //genolist. pasteur. fr/TubercuList/index. html) ESAT-6, EsxA: Rv3875 CFP-10, EsxB: Rv3875 EspC: Rv3615c PE18: Rv1788 PE19: Rv1791 PPE25: Rv1787 PPE26: Rv1789 PPE27: Rv1790 EccD5: Rv1795

Mycobacterium tuberculosis (Mtb), the causative agent of human tuberculosis, is one of the most widely spread human pathogens, responsible for more than 9. 6 million of new tuberculosis cases and 1. 5 million deaths, annually. The resurgence of pulmonary tuberculosis in immuno-compromised patients, including HIV-co-infected populations, and increasing spread of drug-resistant Mtb strains are worrying. Given the estimated 2 billion cases of latent Mtb infections and the only partial efficacy of the unique, currently available tuberculosis-vaccine Mycobacterium bovis BCG (Bacille Calmette-Guerin) it is necessary to develop improved vaccines. Here, we demonstrate that the host cellular immunity, mediated by CD4+ T lymphocytes, specific to the “PE/PPE” families of mycobacterial antigens, contribute to the protection against Mtb-induced disease. We revealed the fine composition of the PE/PPE-specific T cells by characterizing their effector functions and differentiation status. We previously described a live attenuated mycobacterial strain as a vaccine candidate that is able to induce such CD4+ T cells and which displays particular properties at stimulating the cells of the innate immune system. These responses play a central role in the initiation of the host defense and in the protection against tuberculosis. Our results pave the way for further development of candidates in preclinical models of anti-tuberculosis vaccination.

Abstract Introduction Results Discussion Materials and Methods

blood cells t helper cells medicine and health sciences immune cells immune physiology immunology cell-mediated immunity cell differentiation vaccines preventive medicine developmental biology vaccination and immunization bacteria public and occupational health immune system proteins white blood cells animal cells proteins antigens t cells actinobacteria immune response biochemistry cell biology attenuated vaccines mycobacterium tuberculosis immunity physiology biology and life sciences cellular types organisms

2016

CD4+ T Cells Recognizing PE/PPE Antigens Directly or via Cross Reactivity Are Protective against Pulmonary Mycobacterium tuberculosis Infection

The emergence and rapid global spread of the swine-origin H1N1/09 pandemic influenza A virus in humans underscores the importance of swine populations as reservoirs for genetically diverse influenza viruses with the potential to infect humans. However, despite their significance for animal and human health, relatively little is known about the phylogeography of swine influenza viruses in the United States. This study utilizes an expansive data set of hemagglutinin (HA1) sequences (n = 1516) from swine influenza viruses collected in North America during the period 2003–2010. With these data we investigate the spatial dissemination of a novel influenza virus of the H1 subtype that was introduced into the North American swine population via two separate human-to-swine transmission events around 2003. Bayesian phylogeographic analysis reveals that the spatial dissemination of this influenza virus in the US swine population follows long-distance swine movements from the Southern US to the Midwest, a corn-rich commercial center that imports millions of swine annually. Hence, multiple genetically diverse influenza viruses are introduced and co-circulate in the Midwest, providing the opportunity for genomic reassortment. Overall, the Midwest serves primarily as an ecological sink for swine influenza in the US, with sources of virus genetic diversity instead located in the Southeast (mainly North Carolina) and South-central (mainly Oklahoma) regions. Understanding the importance of long-distance pig transportation in the evolution and spatial dissemination of the influenza virus in swine may inform future strategies for the surveillance and control of influenza, and perhaps other swine pathogens. Swine influenza A viruses cause severe respiratory disease in pigs, similar to that which presents in humans, and constitute an important economic concern for the US swine industry and threat to public health. Influenza was first clinically recognized in pigs in the Midwestern US in conjunction with the severe 1918 ‘Spanish flu’ H1N1 pandemic in humans [1], although whether the pandemic originated in humans or pigs remains unresolved [2]. Periodic transmission of influenza viruses between humans and swine occurs in both directions, including such notable cases as the 1976 outbreak of swine A/H1N1 influenza virus in humans in Fort Dix, New Jersey [3] and the 2009 swine-origin A/H1N1 pandemic virus in humans [4], [5]. The 1918-origin ‘classical’ H1N1 swine influenza virus circulated in US swine for 80 years with relatively few antigenic changes [6], but in the last decade the antigenic diversity of swine influenza viruses in the US has multiplied, stimulating research, development, and uptake of influenza vaccines in the US swine industry. Currently, influenza A viruses of the H1N1, H1N2, and H3N2 subtypes all co-circulate in US swine. In 1998–1999, a triple reassortant H3N2 influenza virus emerged in US swine that possessed HA (H3), NA (N2), and PB1 segments of human H3N2 virus origin, PB2 and PA segments of avian virus origin, and NP, M1/2, and NS1/2 segments of classical swine virus origin [7] (Fig. 1). Over the next decade these H3N2 triple reassortant swine viruses further reassorted with human H3N2 viruses [8], [9], as well as with the co-circulating H1N1 classical swine viruses [10], [11]. Mainly these reassortment events involved the HA and NA segments, preserving what has been termed the ‘triple reassortant internal genes’ (TRIG) constellation (avian-origin PB2 and PA, human H3N2-origin PB1, and classical swine-origin NP, M1/2, and NS1/2). In 2003 influenza A virus of entirely human H1N2 origin was identified in Canadian swine [12], and in 2005 H1N1 viruses with human-origin H1 and N1 segments were identified in the United States, representing two separate introductions of human H1 virus into swine that were referred to as ‘δ-1’ (H1N2) and ‘δ -2’ (H1N1) lineages based on the order of identification [13]. These human-H1 origin swine viruses also acquired novel genome segments via reassortment with other swine and human influenza viruses [12], [13]. Globally, the swine influenza virus population is spatially separated into the North American and Eurasian lineages, although both lineages co-circulate in Asia, which imports swine from North America and Europe. In the US the traditional center of swine production is located in the ‘Corn Belt’ of the Midwest, including Iowa, Illinois, Indiana, and Minnesota [14]. Beginning in the 1970' s, swine production expanded into large new facilities located in the Southeastern US, mainly North Carolina, and more recently into Oklahoma in the South-central US [15]. Due to the lower cost of transporting swine versus the required amount of feed, the majority of swine born in the South-central and Southeastern regions are transported by road to the Midwestern Corn Belt to be fattened and slaughtered, resulting in continuous large-scale movements of swine (‘swine-flows’) into the Midwest [14]. However, the role of local, regional, and global swine-flows in the ecology and evolution of swine influenza viruses remains unclear. The aim of our study was to investigate the role of inter-regional swine-flows in the spatial dissemination of newly introduced swine viruses in the US, using the human-origin A/H1 influenza virus as a case study. We utilize HA1 sequence data from a large data set of swine influenza virus isolates (n = 1,516 sequences) collected from 23 US states during 2003–2010 and apply recently developed methods of Bayesian phylogeography. The strength of the Bayesian approach is that the diffusion process among discrete location states is integrated with time-scaled phylogenies that incorporate phylogenetic uncertainty. This approach provides a formal framework to test hypotheses about viral diffusion processes driven by known population distributions and movements. Of the 1,516 HA1 (H1) influenza virus sequences collected from swine in the United States and Canada from 2003–2010 that were included in this study, 41 were related to the human pandemic H1N1/09 virus, all of which were collected in 2009–2010 and appear to result from multiple human-to-swine transmission events. These pandemic viruses have been described previously and thus are not the focus of the present study [16]. Of the remaining 1,475 swine viruses, 327 were phylogenetically related to seasonal human H1 viruses (Fig. S1), which constitute two phylogenetically distinct clusters, representing two contemporaneous, but independent introductions of different human influenza viruses into swine (Fig. 2), consistent with previous findings [13]. Both of these clusters are phylogenetically most closely related to human H1 influenza viruses collected in early 2003. One cluster (n = 138 sequences) is related to widespread human seasonal A/H1N1 virus, while the other cluster (n = 187 sequences) is related to a less common human reassortant A/H1N2 virus that circulated globally in humans from 2001–2003. The A/H1N2 reassortant virus contains an HA derived from human seasonal H1N1 viruses and 7 segments of human H3N2 influenza virus origin [17]. We estimated the Time to the Most Recent Common Ancestor (TMRCA) for the nodes adjoining the branch that represents the human-to-swine transmission events of the H1N1 and H1N2 viruses. Accordingly, the cross-species transmission of H1N1 from humans into swine is estimated to have occurred during the period October 2002–March 2003, which coincides with the timing of the A/H1N1-dominant 2002–2003 winter influenza epidemic in humans in North America [18] (Fig. 2, Table S1). Similarly, the timeframe for the cross-species transmission of the H1N2 virus into swine is estimated to be August 2002–February 2003, which overlaps with the time period when A/H1N2 viruses circulated in humans in North America (Table S1). To explore the whole-genome evolution of these human-origin swine influenza viruses, maximum likelihood trees were inferred for the subset (n = 31) of the human-origin swine influenza virus HA1 sequences for which the NA and internal gene sequences were publicly available at GenBank [19]. Major reassortment events are summarized in Table 1 and Fig. 1, including the H1N1 and 2003–2004 H1N2 reassortment events (#1 and #2/3 respectively, Table 1) that have been described previously [12], [13]. The PB2 phylogeny is depicted in Fig. 3, the NA (N2) phylogeny is depicted in Fig. 4, and the phylogenies of other 5 segments and N1 are available in the Supporting Information (Figs. S2, S3, S4, S5, S6, and S7). Notably, all H1N1 and H1N2 isolates collected after 2004 have acquired the triple reassortant internal genes (TRIG) cassette, which were originally derived in 1998 from avian influenza viruses (PB2 and PA), human influenza viruses (PB1), and classical swine influenza viruses (NP, M, and NS). The topology of these trees suggests that the human H1N2-origin lineage may have acquired components of the TRIG cassette approximately 3–4 times over the course of 2007–2008 via multiple reassortment events (Fig. 3, Fig. S2, S3, S4, S5, S6, and S). The largest clade (n = 21) of 2008 human H1N2-origin swine isolates (#7, Table 1) contains the TRIG, but also has acquired via reassortment a human H3N2-origin NA (N2) segment that had circulated in swine at least since 2003, when human H3N2 viruses appear to have reassorted with a lineage of swine A/H3N2 triple reassortant swine viruses that is referred to ‘clade IV’ in the nomenclature for the HA segment [9] (Fig. 4). To investigate the spatial dissemination of these novel viruses within the US swine population, we inferred separate Bayesian phylogenies for the H1N1 and H1N2 data sets, considering the three discrete US regions that are well sampled in our data: the Midwest (IL, IN, IA, KS, MI, MN, MO, NE, OH, SD, WI), South-central (OK, TX), and Southeast (NC, SC), which are delineated broadly according to the US farm production regions defined by the USDA [20]. Distinct spatial patterns are clearly evident for both the H1N1 and H1N2 lineages that are depicted in the phylogeny presented in Fig. 2, as all of the H1N1 viruses are from the Southeast (83/138 isolates), mainly representing North Carolina, or the Midwest (55/138 isolates), whereas the H1N2 isolates are predominantly collected in the Midwest (97/169 isolates) and South-central (70/169 isolates) regions (Fig. 2). Both phylogenetic trees exhibit strong spatial structuring, and we observe a statistically significant correlation between phylogeny and location state for the Midwest (p<0. 01), South-central (p<0. 01), and Southeast (p<0. 05) regions on both the H1N1 and H1N2 trees using the parsimony score (PS) and association index (AI) statistics [21]. The maximum clade credibility (MCC) trees annotated with most probable nodal locations indicate multiple introductions of both H1N1 and H1N2 viruses into the Midwest, with the H1N1 virus disseminating Southeast-to-Midwest, and the H1N2 virus disseminating South-central-to-Midwest. In contrast, there is little evidence of viral migration in the opposite directions, or between the South-central and Southeast regions (Fig. 2). ‘Markov jump’ counts [22] of the expected number of location state transitions along the phylogenetic branches provide a quantitative measure of gene flow between regions, representing successful viral introductions from one region to another (Fig. S8). Across the posterior distribution of trees inferred for both subtypes, the vast majority of inter-regional introductions occur in the directions of Southeast-to-Midwest (mean, 13. 1) and South-central-to-Midwest (mean, 9. 4), with less frequent viral migration also detected from Midwest-to-Southeast (mean, 3. 3) (Table 2). Based on the number of swine transported from one region to another over the years of high sampling (2005–2008) (Table S2), we estimate that an introduction of a human-origin H1 swine influenza virus occurs roughly per million swine transported from one region to another (Table 2), although this provides only a lower boundary as the introductions are estimated based on our limited sampling, and we can only detect introductions with substantial onward transmission. To quantitatively estimate the importance of known geographical swine population distributions and movements in the spatial dynamics of the virus, we encoded four potential predictors of viral dissemination between pairwise regions as phylogeographic models [23] and fitted these models individually to the sequence data: (i) the number of swine transported annually from one region to another (with directionality), (ii) the swine population size in the region of origin, (iii) the swine population size in the region of destination, and (iv) the product of the swine population sizes in the region of origin and the region of destination (Tables S2 and S3). Given that the South-central, Southeast, and Midwest regions are approximately equidistant from each other by road and geodesic distance, we did not consider geographical distances to be a potential predictor of viral movements in our inter-regional analysis. Bayes factor comparisons [24] via marginal likelihood estimates of the model fit for each potential predictor indicates that the spatial dynamics of the human-origin H1 virus in swine are best described by the number of swine transported annually from one region to another (Table 3). Fixing the rates relative to the swine population size of the region of destination also improved the marginal likelihood, reflecting the directionality of swine-flows from regions of relatively lower swine population size in the South-central and Southeast regions to the largest swine population found in the Midwest. The poorest marginal likelihood was obtained when rates were fixed relative to the swine population in the region of origin, indicating low rates of viral dissemination out of the large swine populations in the Midwest. Finally, to ensure that the observed geographical patterns were not an artifact of sampling (Fig. S9), we repeated the phylogeographic analysis using a balanced data set that was randomly subsampled from the original data to obtain equal numbers of sequences from each region (n = 70). Using this balanced data set we find very similar patterns as those derived from the full data set, with substantial viral movement from South-central to Midwest and Southeast to Midwest and strongest support for the ‘swine-flows’ model (Tables S4 and S5). The numbers of viral introductions are somewhat lower than in the original analysis (Table S4) and there is weaker support for the ‘swine-flows’ model (Table S5), but this is expected given the smaller number of sequences used in the sensitivity analysis. To capture the early spatial patterns of a newly emergent virus in swine populations prior to extensive geographical mixing, this study focused on an H1 influenza virus that was introduced twice from humans into swine around 2003. The fact that this human H1 virus was introduced into swine on two separate occasions (H1N1 and H1N2) allows, uniquely, a side-by-side comparison of the spatial dynamics of two similar emergent viruses. In our statistical analysis, we also take advantage of the independent nature of these two introductions through a model that simultaneously draws information from the H1N1 and H1N2 evolutionary histories to inform the rates of movement in an asymmetric diffusion model. The latter allows us to fully characterize the bidirectional movement between the three major sampling regions despite the fact that the independent lineages provide very different numbers of samples from these regions. We find that the key source population of the human-origin H1N1 virus is likely to be swine in the Southeastern US, particularly North Carolina, whereas the source population of the H1N2 virus appears to be swine in the South-central US, including Oklahoma. Subsequently, both the H1N1 and H1N2 virus rapidly disseminated to the Midwestern US, apparently following the main swine transportation routes (‘swine-ways’) to the Midwest, the traditional center of American pig farming, to be fattened on the feed corn produced in the region prior to slaughter. Although the Midwest swine population is >4-fold larger than the Southeast swine population and >12-fold greater than the South-central population, the Midwest effectively serves as an ecological sink for the virus due to its commercial function as a final marketing destination and net importer of pigs. These results appear to be robust to sampling bias, as we found similar patterns of viral migration using a subsampled data set comprising 70 isolates that were randomly sampled from each of the three US regions (Tables S4 and S5). It is certainly possible for novel lineages of influenza virus to begin their spread in the Midwest, and we have not considered farm density, climatic conditions, husbandry practices, biosecurity, vaccination status, or any other factors that would favor viral emergence in the South-central or Southeast versus the Midwest. The role of newer high-density swine production facilities in Oklahoma and North Carolina in viral evolution, in tandem with other immunological or environmental factors, clearly requires study at a finer spatial scale. Rather, our findings suggest that any viral lineage that originates in the Midwest would be less likely to spread to other US regions due to lower rates of regional exportation of Midwestern swine, whereas viruses that originate in the South-central or Southeast are likely to rapidly disseminate to the Midwest. Although the Midwest does not appear to be a source population for swine influenza viruses, the region is likely to provide a reservoir for multiple genetically distinct variants to co-circulate and exchange segments via reassortment due to the continual importation of swine influenza viruses from other regions. Even a limited sampling (31 whole-genome sequences) revealed extensive reassortment between the human-origin swine viruses and other swine and human influenza viruses over a 7-year period. Both the human H1N1- and H1N2-origin swine viral genomes exhibit a pattern of HA and NA segments that are closely related to human viruses, but internal segments related to triple reassortant swine viruses (TRIG), suggesting that such genomic arrangements may be selectively favored (although this clearly requires further study). Overall, our study captures the effects of at least a decade of large-scale structural changes in the US commercial swine industry on the evolution and spread of one of the most economically important pathogens in US swine. Further understanding of the role of long-distance pig transport in the ecology and evolution of swine influenza viruses may inform targeted surveillance and mitigation strategies in the future, including intensified surveillance in the less sampled Southern regions. While increased genetic and antigenic diversity observed in swine influenza viruses in recent years has stimulated ongoing research into the development of new influenza vaccines for swine, including live-virus and DNA-based approaches [25], identifying key geographical sources of the virus and reservoirs of genetic diversity may direct vaccination strategies in pigs of different age groups and specified localities. Although the patterns of viral dissemination we identify using the human-origin H1 influenza virus as a case study are striking, these findings invite further study into the phylogeography of swine influenza viruses at more precise spatial scales, including within our broadly defined Midwest region, as well as globally. For this study we newly generated a total of 1,412 HA1 sequences (889 nt) from H1 influenza A viruses collected from swine in the United States and Canada that exhibited respiratory disease during the period 2003–2008 [26] (Table S6). Two of the isolates were swine viruses that were isolated from turkeys: A/turkey/North Carolina/00533/2005 and A/turkey/North Carolina/00536/2005, but these were triple reassortant viruses and not included in the phylogeographic analysis. HA1 gene sequences were obtained either from virus isolates or directly from the originally submitted nasal swab or lung tissue material. To isolate viruses, the swab or tissue supernatant (in 400-µl amounts) was inoculated on monolayers of MDCK cells grown in 25-cm2 flasks with 5 ml of MEM+ media [27]. All cultures were incubated at 37°C under a 5% CO2 atmosphere. All flasks were examined daily for 7 days under an inverted light microscope to observe virus-induced cytopathic effects (CPE). Viral RNA was extracted from 50 µl of swab supernatant using a magnetic bead procedure (Ambion MagMAX AM1835 and AM1836, Applied Biosystems, Foster City, CA). Segment specific PCR fragments were obtained with One-Step RT-PCR (Qiagen, CA) using influenza A specific primers for HA as described previously [28]. These data were supplemented with 104 additional HA1 sequences from H1 North American swine influenza viruses sampled during 2003–2010 that were downloaded from the National Center for Biotechnology Information (NCBI) Influenza Virus Resource (http: //www. ncbi. nlm. nih. gov/genomes/FLU/FLU. html) available at GenBank [19]. This overall total of 1,516 sequences were collected from 23 US states and Canada: Arkansas (AR), Colorado (CO), Georgia (GA), Illinois (IL), Indiana (IN), Iowa (IA), Kansas (KS), Kentucky (KY), Michigan (MI), Minnesota (MN), Missouri (MO), Nebraska (NE), North Carolina (NC), Ohio (OH), Oklahoma (OK), Oregon (OR), Pennsylvania (PA), South Carolina (SC), South Dakota (SD), Tennessee (TN), Texas (TX), Virginia (VA), and Wisconsin (WI). The majority of isolates were collected from the Midwest (n = 921), followed by Southeast (n = 426) and South-central (n = 139) regions (Table S6, Fig. S9). We excluded the possibility that the spatial patterns detected were simply an artifact of uneven sampling during early emergence of the human-like H1 influenza virus in swine (2003–2005) by observing no statistical difference between the number of isolates collected in each region during 2003–2005 compared to 2006 when the virus was widespread in the US (p-value = 0. 9055, Pearson' s Chi-square test). Nucleotide alignments were manually constructed for the HA1 region (889 nt) using the Se-Al program [29]. To infer the evolutionary relationships for the complete data set of 1,516 HA1 sequences, we employed maximum likelihood (ML) methods available through the PhyML program, incorporating a GTR model of nucleotide substitution with gamma-distributed rate variation among sites, and a heuristic SPR branch-swapping search [30]. This phylogenetic analysis identified a cluster of 327 sequences that were separated by a very high number of expected substitutions from the remaining 1,193 swine sequences. To explore the evolutionary origins of these highly divergent sequences in greater detail, a second tree was inferred for the 325 divergent swine sequences (two were excluded due to poor sequence quality) and 92 randomly selected human H1 (HA1) sequences: 3 H1N1 sequences selected from each of the following years: 2000,2001,2004,2005,2006,2007,2008, and 2009; 3 H1N2 sequences selected from 2001; plus an additional 33 H1N1 and 32 H1N2 sequences for the years 2002–2003 during which human-to-swine transmission occurred (the XML file is available in Supplemental Information, Text S1). For this data set, posterior distributions were estimated under a phylogenetic model using a Bayesian Markov chain Monte Carlo (MCMC) method implemented in the BEAST package (v1. 6), incorporating the date of sampling [31]. Given the time span of our data set, sequences for which only the year of sampling was known were included and assigned a mid-year sampling date of June 1st. Only 30 of 325 isolates did not have an exact date of collection, mainly because collection dates were not available on GenBank [19]; the majority of isolates without exact dates were collected in 2008 in Oklahoma (Table S6). We employed a strict molecular clock, a flexible Bayesian skyline plot (BSP) prior (10 piece-wise constant groups), HKY85 +Γ4 model of nucleotide substitution, and the SRD06 codon position model with two partitions for codon positions (1st+2nd positions, 3rd position), with substitution model, rate heterogeneity model, and base frequencies unlinked across all codon positions. The MCMC chain was run for 100 million iterations, with sub-sampling every 50,000 iterations. All parameters reached convergence, as assessed visually using Tracer (v. 1. 5). The initial 10% of the chain was removed as burn-in, and maximum clade credibility (MCC) trees were summarized using TreeAnnotator (v. 1. 5. 4). A phylogenetic analysis also was conducted upon the 31 human-origin swine influenza viruses (3 H1N1,28 H1N2) for which whole-genome sequences were available at the NCBI Influenza Virus Resource [19] at GenBank (http: //www. ncbi. nlm. nih. gov/genomes/FLU/FLU. html) (Table S6). As the evolutionary relationships of the H1 already had been extensively analyzed (Fig. S1), we downloaded only the remaining 7 genome sequences from GenBank. Due to the divergence of the NA (N1) and NA (N2) sequences, two separate alignments were constructed. In each alignment, 15 representative human influenza viruses collected during 2001–2003 were included, representing the H3N2 (n = 3), H1N2 (n = 5), and H1N1 (n = 7) subtypes. Given the complexity of phylogenetic relationships on the NA (N2) tree arising from frequent reassortment, 99 additional human H3N2 NA sequences were included. Twenty-three swine triple reassortant H3N2 viruses collected during 1998–2009 were included as background. Varying numbers of swine H1N1 influenza virus sequences were available on GenBank for each segment as background: PB2 (n = 38), PB1 (n = 47), PA (n = 36), NP (n = 31), N1 (n = 35), N2 (n = 60), M1/2 (n = 47), NS1/2 (n = 67). Sequence alignments were manually constructed for the major coding regions of PB2 (2,277 nt), PB1 (2,271 nt), PA (2,148 nt), NP (1,494 nt), NA (1,407 nt), M1/2 (979 nt), and NS1/2 (835 nt). Regions of overlapping reading frame were deleted in the case of M1/2 and NS1/2. Here, phylogenetic trees were inferred using the maximum likelihood (ML) method under a GTR+I+Γ4 model available in PAUP* [32] for each of these 8 alignments. In all cases TBR branch-swapping was employed to determine the globally optimal tree. To assess the robustness of each node, a bootstrap re-sampling process (1,000 replications) using the neighbor-joining (NJ) method was used, incorporating the ML substitution model. Clades of related isolates were identified by high bootstrap values (>70%) and exceptionally long branch length estimates. Due to high sampling heterogeneity among US states, we categorized each isolate into three US regions: Midwestern (IL, IN, IA, KS, MI, MN, MO, NE, OH, SD, WI), South-central (OK, TX), and Southeastern (NC, SC). These regions generally correspond to the US farm production regions defined by the US Department of Agriculture (USDA) [20], with the Midwest region including the Corn Belt (IL, IN, IA, MO, OH), Lake States (MI, MN, and WI), and Northern Plains (KS, NE, ND, SD); the Southeast region including Appalachia (KY, TN, NC, VA, WV) and the Southeast (AL, FL, GA, SC); and the South-central region corresponding to the Southern Plains region (OK, TX). Sequences from the other geographic regions that were sampled at relatively low levels were excluded, as were highly phylogenetically divergent sequences that might represent possible sequencing error. This resulted in a final data set of 127 H1N1 and 169 H1N2 isolates that could be used in our detailed spatial analysis. Although we considered separate evolutionary histories for our 127 H1N1 and 169 H1N2 human-like swine HA1 sequences, we jointly inferred the asymmetric rates of movement under a single model of discrete diffusion among the three regions to perform spatial model testing (see below). Moreover, estimating the rates of a single diffusion matrix applied to independent phylogenies may also improve statistical efficiency [23]. Posterior distributions under the Bayesian phylogeographic model [23] were estimated using a MCMC method implemented in BEAST using BEAGLE [33] to improve computational performance. The model incorporated the date of sampling and used a strict molecular clock, BSP prior, and the SRD06 model of nucleotide substitution described. The MCMC chain was run for 100 million iterations, with sub-sampling every 10,000 iterations. All parameters reached convergence, as assessed visually using Tracer (v. 1. 5). The initial 10% of the chain was removed as burn-in, and MCC trees were summarized using TreeAnnotator (v. 1. 5. 4). The expected number of location state transitions conditional on the observed data was obtained using Markov jump counts [22], [34] again implemented in BEAGLE [33], and summarized per branch and for the complete evolutionary history. Ad hoc measures of the extent of geographic structure in the MCC trees were determined for the H1N1 and H1N2 data sets using the parsimony score (PS) and association index (AI) tests as available in the Bayesian Tip-association Significance testing (BaTS) program [21]. To test the importance of swine population sizes and movements in the US in the spatial patterns that were observed, we parameterized the discrete phylogeographic diffusion model in terms of four sources of state-level information on swine populations, aggregated to the regional level and normalized (mean of 1) (Tables S2 and S3). First, we used the number of swine transported annually between states in a pairwise manner for the year 2001, available through the United States Department of Agriculture (USDA) Economic Research Service (http: //www. ers. usda. gov/Data/InterstateLivestockMovements/view. asp) (XML file available in the Supplemental Information, Text S2). Second, we obtained data from the USDA 2007 Census of Agriculture [35] to integrate as instantaneous diffusion rates (i) the swine population size of the region of origin (XML file, Text S3), (ii) the swine population size of the region of destination (XML file, Text S4), and (iii) the product of the swine population sizes from the region of origin and the region of destination (XML file, Text S5). Each of these predictors was incorporated into an asymmetric transition matrix that allows for separate directional rates between each pair of locations. A Bayes factor comparison [36] via the relative marginal model likelihoods was used to select the most appropriate model for the data, compared to equal migration rates (XML file, Text S6). Finally, the phylogeographic analysis was repeated using a balanced data set that was randomly subsampled from the original data to obtain equal numbers of sequences from each region (n = 70) (XML file, Text S7) and using independent rate matrices (XML file, Text S8). All sequences were submitted to GenBank and given accession numbers CY040460 – CY082963 (Table S6).

Since 1998, genetically and antigenically diverse influenza A viruses have circulated in North American swine due to continuous cross-species transmission and reassortment with avian and human influenza viruses, presenting a pandemic threat to humans. Millions of swine are transported year-round from the southern United States into the corn-rich Midwest, but the importance of these movements in the spatial dissemination and evolution of the influenza virus in swine is unknown. Using a large data set of influenza virus sequences collected in North American swine during 2003–2010, we investigated the spatial dynamics of two influenza viruses of the H1 subtype that were introduced into swine from humans around 2003. Employing recently developed Bayesian phylogeography methods, we find that the spread of this influenza virus follows the large-scale transport of swine from the South to the Midwest. Based on this pattern of viral migration, we suggest that the genetic diversity of swine influenza viruses in the Midwest is continually augmented by the importation of viruses from source populations located in the South. Understanding the importance of long-distance pig movements in the evolution and spatial dissemination of influenza virus in swine may inform future strategies for the surveillance and control of influenza, and perhaps other swine pathogens.

Abstract Introduction Results Discussion Materials and Methods

medicine infectious diseases biology computational biology evolutionary biology

2011

Spatial Dynamics of Human-Origin H1 Influenza A Virus in North American Swine

Recent experiments showed that engineered Escherichia coli colonies grow and self-organize into periodic stripes with high and low cell densities in semi-solid agar. The stripes develop sequentially behind a radially propagating colony front, similar to the formation of many other periodic patterns in nature. These bacteria were created by genetically coupling the intracellular chemotaxis pathway of wild-type cells with a quorum sensing module through the protein CheZ. In this paper, we develop multiscale models to investigate how this intracellular pathway affects stripe formation. We first develop a detailed hybrid model that treats each cell as an individual particle and incorporates intracellular signaling via an internal ODE system. To overcome the computational cost of the hybrid model caused by the large number of cells involved, we next derive a mean-field PDE model from the hybrid model using asymptotic analysis. We show that this analysis is justified by the tight agreement between the PDE model and the hybrid model in 1D simulations. Numerical simulations of the PDE model in 2D with radial symmetry agree with experimental data semi-quantitatively. Finally, we use the PDE model to make a number of testable predictions on how the stripe patterns depend on cell-level parameters, including cell speed, cell doubling time and the turnover rate of intracellular CheZ. Understanding the formation of regularly spaced structures, such as vertebrate segments, hair follicles, fish pigmentation and animal coats, is a fundamental problem in developmental biology [1–7]. These patterns involve the complex interaction of intracellular signaling, cell-cell communication, cell growth and cell migration. The overwhelmingly complex physiological context usually makes it difficult to uncover the interplay of these mechanisms. Synthetic biology has recently been used to extract essential components of complex biological systems and examine potential strategies for pattern formation [8–11]. One of these problems relate to the bacterium Escherichia coli. Recently in [12], the chemotaxis signaling pathway of E. coli has been engineered and coupled with a quorum sensing module, leading to cell-density suppressed cell motility. When a suspension of the engineered cells is inoculated at the center of a petri dish with semi-solid agar and rich nutrient, the colony grows, moves outward and sequentially establishes rings or “stripes” with a high density of cells behind the colony front (Fig 1A). These spatial patterns form in a strikingly similar way as many periodic patterns in other biological systems. When the maximum density of the motile front reaches a threshold, an immotile zone is nucleated. The immotile zone then absorbs bacteria from its neighborhood to expand, forming alternating high and low density zones. These patterns do not form when using wild-type E. coli; instead, the colony simply expands outward and forms a uniform lawn. The goal of this paper is to use mathematical models to elucidate the underlying mechanisms for this pattern formation, with a special focus on the roles of intracellular signaling. E. coli is an enteric gram-negative bacterium that moves by alternating forward-moving “runs” and reorienting “tumbles”. It has 6-8 flagella on its surface that can rotate either clockwise (CW) or counterclockwise (CCW) (Fig 1B). If the majority of its flagella rotate CCW they form a bundle and push the cell to run forward with a speed ∼ 10 − 30μm/s. If some flagella rotate CW they fly apart and the cell tumbles in place. E. coli can bias its movement in response to external chemical signals, e. g, towards locations with higher concentration of chemoattractant or lower concentration of repellent, which is called chemotaxis. The molecular mechanism of E. coli chemotaxis is summarized in Fig 1C. The transmembrane chemoreceptors (denoted as MCP) form stable ternary complexes with the intracellular signaling proteins CheA and CheW. CheA is an auto-kinase and also a kinase for the response regulators CheY and CheB. The activity of CheA depends on the ligand-binding state of the receptor complex as well as its methylation level: attractant-binding reduces CheA activity and methylation increases it. The phosphorylated form CheYp binds to the flagella motor and increases the probability of clockwise (CW) rotation. On the other hand, CheBp and CheR change the methylation state of the receptor at a slower rate: CheR methylates it and CheBp demethylates it. Upon attractant binding, CheA activity is reduced immediately, leading to lower CheYp and CheBp. Then a shift of the methylation-demethylation cycle gradually restores CheA activity on a slower time scale. In [12], the quorum-sensing module of bacterium Vibrio fischeri was embedded into E. coli and used to control the transcription of cheZ (Fig 1D). The engineered cell synthesizes and secretes acyl-homoserine lactone (AHL), a small molecule that is freely diffusible across the cell membrane and degrades rapidly. At high concentrations, AHL suppresses the transcription of cheZ in an ultra-sensitive manner. If cheZ is suppressed, CheZ protein becomes diluted as the cell grows and divides. Because CheZ is a dephosphorylation kinase of CheYp, a reduction of CheZ protein can immediately lead to higher CheYp concentration and thus more persistent tumbles of the cell. This, in turn, causes changes to the chemoreceptors as well as to other proteins involved in chemotaxis, and triggers a non-classic chemotactic cellular response. To quantify the effect of AHL in single cell movement, one must take into account the whole chemotaxis pathway as well as CheZ turnover. A phenomenological PDE model was used to explain the pattern formation process in [12] and a simplified version was analyzed in [13]. The model consists of a system of reaction-diffusion equations for the cell density, AHL and nutrient concentrations. The diffusion rate of the cell population is assumed to be a switch-like function of the local AHL concentration. Since the whole chemotaxis pathway is involved in the pattern formation process, it is unclear how cell movement can be reduced to an anisotropic (or cross) diffusion process. Moreover, the model does not address the role of intracellular signaling in stripe formation and cannot be used to understand how the spatial structure of the high-density and low-density regions depends on cell-level parameters. To address these questions, we first developed a hybrid model for the stripe formation that accounts for the behavior of individual cells. The model starts with a detailed description of intracellular signaling, single cell movement and cell division. This individual-based component is then coupled with reaction-diffusion equations for AHL and nutrient concentrations. The multiscale nature of this model allows us to explore the relations between cellular processes on a time scale of seconds to minutes and population dynamics on a time scale of hours. Simulations of our hybrid model showed the same stripe patterns as observed in experiments, but they are very time-consuming due to the large number of cells involved in the pattern formation process. If cells double every 30 minutes, then during a typical time period for pattern formation, e. g. 10 hours, the population size can grow 220 ≈ 106 times. To overcome this computational challenge, we then derived a macroscopic PDE for the cell density from the hybrid model, using asymptotic analysis and moment closure methods. Parameters of the PDE model are fully determined using parameters of the hybrid model. Numerical comparisons of the hybrid model and the PDE model showed quantitative agreement in 1D under biologically-relevant parameter regimes. This justifies using the PDE model as a quantitative and predictive tool to understand the relation between population patterning and cellular dynamics. We then used our PDE model to investigate how concentric stripe patterns change when cells are subject to other chemicals or mutations as discussed in [12]. Numerical simulations of our PDE model in 2D with radial symmetry agree with experimental data semi-quantitatively. Finally, we used our PDE model to make a number of predictions on how stripe formation depends on cell-level parameters. Specifically, we investigated how the colony front speed, the wavelength of the spatial pattern and the structure within a single spatial element depend on the individual cell speed, cell doubling time as well as the rate of CheZ turnover. Our simulations suggested that the individual cell speed and the cell doubling time primarily affect the colony front speed and the pattern wavelength, while the the turnover rate of CheZ mainly affects the spatial structure of each stripe. Each cell is described as an individual particle with location xi, velocity vi, and internal state yi. The superscript i is the index for the cell. Cell signaling is modeled by an internal ODE system for yi. Cell movement is modeled by a velocity jump process in which transition rates are functions of yi. Cell growth is implemented by random creation of new daughter cells from mother cells. The cell dynamics is then coupled with reaction-diffusion equations for h (x, t) and n (x, t). A similar type of model was used to model pattern formation in the slime mold Dictyostelium discoideum in [14]. Details of each component are given below. For simplicity of notation, we omitted the superindex i below. To reduce computational cost, we derived a PDE model from the hybrid model using moment closure methods and asymptotic analysis. Let p (x, v, m, z, t) be the density of cells at position x, with velocity v, internal states m and z, and at time t. Let p0 (x, m, z, t) be the density of cells resting at position x with internal states m and z. According to the hybrid model we have ∂ t p + v · ∇ x p + ∂ z (g (z, h) p) + ∂ m (f (m, z) p) = Q (p, p 0), ∂ t p 0 + ∂ z (g (z, h) p 0) + ∂ m (f (m, z) p 0) = Q 0 (p, p 0). (15) Here g (z, h) and f (m, z) are the right-hand sides of Eqs (1) and (2), and Q (p, p 0) = - λ (m, z) p + μ (m, z) p 0 / | V | + r n p, Q 0 (p, p 0) = λ (m, z) ∫ V p d v - μ (m, z) p 0 + r n p 0, (16) where V = s 0 ∂ B 0 1, λ (m, z), μ (m, z) are given by Eqs (11) and (12), and n = n (x, t) is the local nutrient concentration. The first two terms in Q (p, p0) and Q0 (p, p0) represent the density change due to velocity jumps and the third terms are due to cell growth. Let ρz (x, z, t) be the density of cells at position x with internal state z, then ρ z = ∫ R (p 0 + ∫ V p d v) d m. We derived the following approximating equation for ρz (x, z, t) from (15) and (16) (see S2 Text), ∂ t ρ z = ∇ x · (D (z) ∇ x ρ z) - ∂ z (g (z, h (x, t) ) ρ z) + r n ρ z. (17) Here h is the AHL concentration and D (z) = s 0 2 μ 0 (z) d λ 0 (z) [ μ 0 (z) + λ 0 (z) ], (18) where d is the space dimension, and λ0 (z) and μ0 (z) are the switching frequencies when m equals its quasi-steady state. We note that the intracellular chemotactic signaling enters into Eq (17) through the quasi-steady state of m only. This is because the methylation time scale is much smaller than the time scale for the change of z. The derivation was based on time scale separation of the intrinsic biological processes: the time scale for chemotactic signaling is seconds to minutes, the time scale for CheZ dynamics is tens of minutes, and the time scale for the stripe formation is several hours. The derivation involves moment closure methods and asymptotic analysis, similar to our previous works [15,17,18,24]. The PDE model is formed by coupling Eq (17) with the continuous version of (14), namely, ∂ t h (x, t) = D h Δ h (x, t) + α ρ (x, t) - β h (x, t), ∂ t n (x, t) = D n Δ n (x, t) - γ ρ (x, t) n (x, t), (19) where ρ (x, t) = ∫ ρ z (x, z, t) d z. The parameters of the PDE model are fully determined by those of the hybrid model. In our simulations we choose the cell density scale to be ρs = 1000cells · mm−1. As a consequence, α and γ can be calculated as α = αd ρs and γ = γd ρs. We first investigated how intracellular signaling and cell movement depend on the total concentration of CheZ protein (denoted as Zt in this section). If cells are initially seeded on a horizontal line in an agar plate, they will grow, spread out laterally and form straight stripes of equal spacing (see Fig. S4 of [12]). Motivated by these experiments, we first investigated the population pattern formation on a 1D domain [−L, L], representing a cross-section of the stripe patterns, using the hybrid model and the PDE model. Simulations suggest that both models predict the same spatial-temporal population dynamics for the engineered stripe-forming mutants as well as wild-type cells as in experiments. Moreover, the derived PDE model agrees with the hybrid model quantitatively in biologically-relevant parameter regimes. To mimic the experimental setup, we assumed that all cells initially cluster near the center (x = 0) with internal states at equilibrium, i. e. , z = Zw and m = m0. Specifically, for the hybrid model, we randomly put 500 cells in the domain according to the distribution P (x) = 1 σ 2 π exp (− x 2 2 σ 2) with σ = 2mm at t = 0. Correspondingly, for the PDE model, we took ρ z (x, z, 0) = ρ 0 P (x) δ (z - Z w), where ρ0 = 500cells · mm−1/ρs = 0. 5 (ρs = 1000cells · mm−1 is the cell density scale). For both models, we took the initial nutrient concentration to be a constant everywhere and assumed that there was no AHL added in the domain, i. e. , n (x, 0) = 1, h (x, 0) = 0. We used no-flux boundary conditions throughout the paper. For AHL and nutrient concentrations, we imposed ∇h ⋅ n = ∇n ⋅ n = 0 at the boundary of the spatial domain, where n is the outward normal vector. For individual cell movement, we assumed that once a cell reaches the boundary, it bounces back with its velocity reflected by the boundary. In 1D, the cell direction simply reverses. For Eq (17), we chose the computational range z ∈ [zmin, zmax] to be large enough to include all possible CheZ concentrations such that ρz (x, zmin, t) = ρz (x, zmax, t) = 0. In the x direction, we imposed that ∇ρz ⋅ n = 0 for all z. We first simulated the cell population dynamics for the stripe-forming mutants with parameters specified in Tables 1–3. Fig 5 presents the time course data of the cell density as well as the distribution of the internal variable Zt. Panels A and B are the heat maps of the cell density as a function of space and time. The normalized cell density for the PDE model was obtained by integrating ρz over z. The normalized cell density for the hybrid model was calculated using histograms of the cell positions. Panels C-F present the detailed comparisons of the normalized cell density in space (top) as well as the Zt distribution (bottom) given by the two approaches at different time points. The Zt distribution was obtained by normalizing the cell number in each rectangular grid with size 0. 1mm × 0. 03μM by 100 cells. In these simulations, the AHL concentration h also shows the same stripe pattern as the cell density ρ, with peaks and valleys coinsiding with those of ρ; while the nutrient forms a wave front at the colony front, increasing from 0 to the initial normalized state. Fig 5 shows that the total cell number grows significantly as cells divide. As the colony grows, it propagates outward continuously with a more or less constant front speed. Meanwhile cells produce AHL continuously. As extracellular AHL concentration becomes high and reaches the threshold h0 locally, the intracellular Zt at these locations start to drop. As a result, cells at these locations spend more time in the tumbling stage and become less mobile. In contrast, cells in nearby regions with low AHL move more persistently until they migrate into a high AHL region. The existence of high and low mobility regions leads to the sequential establishment of high-density stripes behind the colony front, similar to experiments. As a comparison, we then simulated the population dynamics for wild-type cells that do not secrete AHL, i. e. , αd = α = 0 (Fig 6). In this case, cells grow, consume nutrients, and the colony propagates outward with a constant wave front speed. However, stripes do not appear behind the colony front. Figs 5 and 6 suggest quantitative agreement between the hybrid model and the PDE model. This justifies using the PDE model for further investigations to save computational cost. We also note that the colony front expansion speed for both the engineered mutant and wild-type are identical. This is because the front speed is primarily determined by the growth and motility of cells at the colony front, where AHL does not reach the threshold h0 required for quorum-sensing. Hence, cells therein have the wild-type phenotype for both cases. We use three important features to characterize the spatial-temporal pattern: the colony front propagation speed, the wavelength of the spatial stripes and the internal structure within a spatial period. We investigated how these features depend on intracellular dynamics, cell movement and cell growth. We calculated the front speed as the average speed between t = 10 hr and 20 hr and the wavelength as the average distance between the maximum densities of two successive high density stripes (Fig 10A). To characterize the internal structure of the stripes, we defined the height ratio and the density ratio of the stripes (Fig 9B and 9C). The height ratio is the minimum density (h2) divided by the maximum density (h1) within a stripe. The density ratio is the volume of the shaded region over the region defined by the rectangle ABCD, factoring in the radially symmetric profile of the solution, i. e. , ∫ B C ξ ρ (ξ, t) / h 1 d ξ. The height ratio measures the fluctuations of the cell density in the spatial pattern; while the density ratio quantifies the area fraction of the high-density regions when h2 is small. Synthetic biology has been used to design relatively simple systems to help understand how regularly-spaced structures form in nature. In [12], E. coli was engineered to couple chemotaxis and quorum sensing and these cells establish sequential stripe patterns when grown in semi-solid agar. In this paper, we developed multiscale models to help explain how these population patterns arise and predict their dependence on cell-level parameters. We first developed a hybrid model that takes into account great details of intracellular signaling and movement of each individual cell. This model provides a method to connect cell-level dynamics and population-level behavior in a quantitative manner, but simulating it is very time-consuming as the cell number becomes large. To overcome this challenge, we mathematically derived a PDE model from our hybrid model. All the parameters of the PDE model can be calculated from measurable cell-level parameters used in the hybrid model. The PDE model matches the hybrid model quantitatively and is much more efficient in terms of computation. Our benchmark comparisons showed that the computation of the PDE model was over 100 times faster than that of the hybrid model. This justifies using the PDE model as a quantitative and predictive tool to explore the relation between population patterning and individual behavior. Simulations of our models showed that the stripes arise sequentially due to suppression of CheZ in cells near the front of the expanding colony. At first, the self-secreted AHL reaches the threshold concentration for quorum sensing at these regions. This turns off the production of CheZ proteins in cells locally. The gradual drop of total CheZ inside these cells causes them to tumble excessively. As more and more cells move into these regions and get trapped, a high-density stripe develops. In the meantime, the colony grows and expands outward, and after some time, another high-density stripe establishes at a larger radius for the same reason. The self-trapping is due to the density-dependent suppression of motility, which has been studied before in [13,25]. The model in [13] eliminates CheZ level by enslaving it to the AHL level, while the model in [25] directly links motility to the cell density, however both models are qualitative. The main contribution of our model is that it can not only reproduce the pattern, but also predict how the patterns varies when the individual cell signaling or movement changes. The spatial-temporal dynamics predicted by our simulations match experimental data semi-quantitatively. We also made a number of predictions on the relation between the population patterns and cell level dynamics. Our simulations showed that the individual cell speed and the cell doubling time primarily affect the colony front speed and the wavelength of the stripe pattern (Fig 11). As the cell speed increases, the front speed and the pattern wavelength increases linearly. As the cell doubling time increases, the front speed decreases while the pattern wavelength increases. Moreover, the turnover rate of CheZ protein does not alter the colony front speed and pattern wavelength, but changes the spatial structure of each stripe characterized by the height ratio and density ratio (Fig 12). These predictions can be tested by further experiments. Our PDE model gives a detailed characterization of the anisotropic movement of the whole cell population in response to AHL. Cells with different intracellular CheZ concentration z have different mobility coefficient, given by D (z) (Eq (18) ). As a cell moves around, its internal state evolves with the extracellular environment. The change of z in each cell leads to the average mobility change of the whole population. We note that if z can be approximated by its steady state, which equals Zw if h < h0 and 0 otherwise, then Eq (17) can be “formally” reduced to the anisotropic diffusion model used in [12] ∂ t ρ = ∇ x · (∇ x (D ¯ (h) ρ) ) + r n ρ. (24) where D ¯ is a step function of h. Specifically, we have ρz (x, t, z) = ρ (x, t) Q (x, t, z) with Q (x, t, z) = { δ (z - Z w) h (x, t) < h 0, δ (z) h (x, t) ≥ h 0. Integrating (17) with respect to z, one obtains Eq (24) with D ¯ (h) = ∫ D (z) Q (x, t, z) d z = { D (Z w) h < h 0 D (0) h ≥ h 0. However, during the stripe formation, CheZ turnover correlates with cell growth, which is much slower than single cell movement and intracellular signal adaptation. As a result, CheZ has a broad distribution among all cells and so it is far from its steady state (Fig 5). This suggest that it is important for models to take into account the internal state of cells individually rather than averaging it out. We note that in this paper we used a multiscale modeling approach: start with a detailed, individual-based model for cell dynamics, then derive a PDE model and justify it using numerical simulations, and finally use the PDE model to make predictions on relations of phenomena at different scales. This multiscale approach allowed the macroscopic model to go beyond qualitative and can be used as a predictive tool. This type of multiscale modeling approach has also been used for classical bacterial chemotaxis [17,21].

One of the central problems in biology is to understand the underlying mechanisms responsible for spatial pattern formation in complex systems. This is a difficult task because the essential mechanisms for pattern formation often involve multiple space and time scales and are often buried in overwhelmingly complex physiological details. Recently, synthetic biology has made it possible to investigate strategies of pattern formation in relatively simpler, but still complex, systems. Here we develop multiscale models to help explain the role of intracellular signaling in the formation of stripe patterns in engineered E. coli colonies.

Abstract Introduction Methods Results Discussion

bacteriology cell physiology cell motility medicine and health sciences pathology and laboratory medicine engineering and technology pathogens signal processing microbiology biological locomotion simulation and modeling developmental biology methylation morphogenesis cellular structures and organelles pattern formation research and analysis methods flagellar rotation cell movement microbial physiology chemistry chemotaxis pathogen motility bacterial physiology cell biology virulence factors physiology biology and life sciences chemical reactions physical sciences flagella

2018

The role of intracellular signaling in the stripe formation in engineered Escherichia coli populations

Cancer is a heterogeneous disease often requiring a complexity of alterations to drive a normal cell to a malignancy and ultimately to a metastatic state. Certain genetic perturbations have been implicated for initiation and progression. However, to a great extent, underlying mechanisms often remain elusive. These genetic perturbations are most likely reflected by the altered expression of sets of genes or pathways, rather than individual genes, thus creating a need for models of deregulation of pathways to help provide an understanding of the mechanisms of tumorigenesis. We introduce an integrative hierarchical analysis of tumor progression that discovers which a priori defined pathways are relevant either throughout or in particular steps of progression. Pathway interaction networks are inferred for these relevant pathways over the steps in progression. This is followed by the refinement of the relevant pathways to those genes most differentially expressed in particular disease stages. The final analysis infers a gene interaction network for these refined pathways. We apply this approach to model progression in prostate cancer and melanoma, resulting in a deeper understanding of the mechanisms of tumorigenesis. Our analysis supports previous findings for the deregulation of several pathways involved in cell cycle control and proliferation in both cancer types. A novel finding of our analysis is a connection between ErbB4 and primary prostate cancer. In the past several decades, many genes have been discovered that govern important functions in the development of a variety of different cancers. However, biological insight from the list of genes is still limited and the underlying mechanisms that occur in the cell during tumorigenesis have not been well established. Numerous DNA microarray expression datasets have been collected to profile genetic changes throughout tumor progression [1–6]. Traditionally, gene expression profiling has been used to identify individual genes that become deregulated at distinct stages of tumorigenesis. Such analyses have shown that tumor cells have a great deal of heterogeneity as they progress through the stages of cancer development [7]. The multitude of differentially expressed genes can then be grouped together by shared biological function to uncover mechanistic alterations that may give rise to certain cancer states. This approach has resulted in the understanding of some of the genetic changes that occur during progression. However, single gene based methods do not always provide clear and accurate insight about the underlying biological processes governing tumor development since these processes involve sets of genes. Recently, gene set based methods have been developed to investigate phenotypic changes at the pathway level [8–11]. These methods provide an assessment of the enrichment of a group of genes defined a priori by some biological commonality for certain phenotypes. The main advantage of such methods over single gene based methods is that they begin with biological knowledge and therefore provide better functional or mechanistic insight into the cause of the phenotypic differences. In this paper we provide an integrative hierarchical analysis of tumor progression which discovers a priori defined pathways that are relevant either throughout progression or in particular steps in progression. Pathway interaction networks are inferred for these relevant pathways over the steps in progression. This is followed by the refinement of the relevant pathways to those genes most differentially expressed over progression. The final analysis step is a gene interaction network inferred for these refined sets of genes. This analysis pipeline is applied to model progression in prostate cancer and melanoma. The machine learning and statistical ideas used in the pipeline are regularized multi-task learning (RML) [12] and the ideas of learning gradients [13–16] and inverse regression [17,18]. The network inferences are based upon properties of discrete Gauss-Markov graphs [19]. The database of gene sets we use in this paper is the Molecular Signatures Database (MSigDB) [8]. This is a collection of curated gene sets from online pathway databases, publications in PubMed, and expert knowledge. Table S1 contains the MSigDB of gene sets used in the following analyses. A key constraint in using a priori defined gene sets and pathways is the quality of the database of gene sets and the accuracy of the annotation. Since the enrichment of gene sets is fundamental to our models, we need to validate the accuracy of these gene sets. Two points will be addressed in our validation studies: the accuracy of gene sets annotated according to known perturbations and a comparison of gene sets defined by experimental studies versus gene sets defined by expert knowledge. An affirmative answer to the first point provides confidence in the annotation and interpretations made based on the annotations. A study of the second point highlights the importance of the context of the gene set and again provides information about interpreting results. An affirmative answer to this question allows for a uniform analysis over both types of gene sets. The validation of gene sets requires the knowledge of which pathway gene sets are deregulated in which samples according to known biology. This requirement is satisfied by studies where a model system is genetically perturbed and a gene set is defined as genes that most differentially express under the perturbation. Expression studies where the pathways driving the phenotypic distinction are known also satisfy the above requirement. Due to the difficulty in finding data satisfying the above requirements for all the gene sets in our database [8], we focus on a few gene sets which we can validate: the P53 pathway, the hypoxia pathway, and the RAS pathway. The conclusion of our analysis will be that both experimental and expert defined gene sets seem to be accurately annotated and gene sets defined by expert knowledge may be applicable to more general conditions. We first state a summary of the framework we use to model tumor progression. A detailed description of the steps in the analysis and the methods from machine learning and statistics used is provided in the methods section. The analysis can be divided into three objectives that fall into a natural hierarchical framework—an analysis at the pathway level to identify important pathways and build pathway networks, followed by a gene level analysis to refine relevant pathways and then infer a gene network for those relevant pathways. The first two objectives are at the pathway level. The first objective is to determine which pathways are most relevant to progression over both transitions, normal tissues to primary tumors {n ↦ p} and primary tumors to metastasizing tumors {p ↦ m}, and those relevant to one transition or the other. The second objective is to estimate the interdependence of pathways relevant to each step of tumor progression. This provides a pathway network for each step of tumor progression. The third objective, the refinement of relevant pathways, is at the gene level. The refinement procedure removes genes in relevant pathway gene sets that are not relevant to progression, resulting in a “refined” gene set. A gene network for each refined gene set can be inferred by estimating the interdependence between genes. In the following two subsections we apply this framework to prostate cancer and melanoma. The prostate cancer data [1] is a collection of cDNA microarray expression measurements from 22 samples of benign epithelium (b), 32 samples of primary prostate cancer (p), and 17 samples of metastatic prostate cancer (m). The progression is benign to prostate cancer (PCA) to metastasis, {b ↦ p ↦ m}. We will follow the framework outlined in the previous section for the analysis. At each level of the analysis we use the six hallmarks of cancer [29] to organize the analysis results. These hallmarks, thought to be necessary for the development of cancer, were defined as self-sufficiency in growth signals, insensitivity to anti-growth signals, evasion of apoptosis, limitless replicative potential, sustained angiogenesis, and tissue invasion and metastasis. It was hypothesized that tumors must acquire alterations in each of these categories in order to evade the multitude of anticancer defense mechanisms in the cell. We use these hallmarks as categories to group gene sets found to be relevant in our analysis. The gene sets are assigned to the biological category most fitting, although there are cases where a gene set may fall into more than one category. In addition, pathway gene set dependence graphs inferred will illustrate the extent to which the hallmarks are interwoven. Pathways gene sets corresponding to all six hallmarks were found to be relevant. We will further test our analysis on a melanoma tumor progression expression dataset. Genome-wide expression at different stages of melanoma development is available in [2]. Samples were categorized as normal (n), primary (p), or metastatic (m), with 4 individuals per group. The progression is normal to primary to metastasis, {n ↦ p ↦ m}. We follow the same hierarchical framework, analyzing relevant pathways, pathway networks, and finally gene networks. The major innovation presented in this paper is the use of gene sets in modeling tumor progression rather than single genes. Previous research in tumor progression has studied single genes that can be used as markers of certain stages [2,3]. While it is important to identify these individual genes, a broader understanding of the biological processes occurring during progression has been missing. Recent research in prostate cancer progression has taken a step closer to gaining insight into biologically related gene sets implicated throughout disease using the Molecular Concept Map [1]. This method identifies cellular functions that may be relevant based on common characteristics in the individual genes found to be differentially expressed. An analysis of breast and colorectal cancer was performed by [53], grouping genes using similar gene set databases, such as Gene Ontology and the KEGG Pathway Database. Unlike these analyses, we start in the space of gene sets rather than individual genes. We have introduced a novel analysis pipeline that discovers a priori defined gene sets relevant at different stages of the disease. In addition, an interaction network of these relevant gene sets is inferred. This is followed by refinement of the relevant pathways and gene sets to include only the genes most relevant to progression. A gene interaction network is inferred for these refined gene sets. This approach provides a more accurate and descriptive understanding of pathway deregulation by identifying specific pathway gene sets whose expression becomes altered along with phenotypic changes. Since the method requires accurately defined gene sets, we first sought to validate the gene sets used in the analyses. P53, hypoxia, and RAS pathways have been defined by multiple sources, both in literature and experimental settings. For each, we took data with a specific perturbation for that pathway and calculated pathway enrichment. Results suggested that the gene sets are appropriate for use with careful attention to context. Context of both the dataset and gene set is important in enrichment analyses. We find this in the HRAS/KRAS example. Gene sets defined on one dataset may not be applicable for enrichment analyses when the context of the dataset changes. As such, careful attention should be paid for application of gene sets defined on one data set to a different context. Our results largely agreed with findings in previous studies but also provide some novel biological insights into tumor progression. We discovered gene sets which become deregulated at certain stages and throughout progression. Common themes in the results presented in this paper and in [1] include an increased activity in the cell cycle, an increase of energy requirements, and an initial increase followed by a decrease in hormonal levels. The technique applied in [1] gave concepts which were relevant at certain stages of prostate cancer progression such as the concept of “cell cycle, ” shown to have increased activity throughout progression. Our analysis takes this result a step further by discovering the specific pathways responsible for the increased activity, for example the ERK pathway in early progression. This provides better mechanistic insight into the process of proliferation during tumorigenesis. In [2], individual genes were identified that were specific to create genetic profiles for different stages of melanoma. Although it is important to identify such markers that can differentiate stages, they do not provide an understanding of the underlying mechanisms that drive progression. [2] finds that genes involved in cell cycle regulation and proliferation are of utmost importance during melanoma development. Our method discovers several mechanisms underlying cell cycle regulation and proliferation that become deregulated such as the AKT and P53 pathways. Studies have also been performed using comparative genomic hybridization (CGH) to associate DNA copy number aberrations with genetic progression [54,55]. We do not present such an analysis at this time; however, our analysis can be applied in this setting by using chromosomal regions as gene sets. By transforming expression datasets into the space of enrichment scores of gene sets, we have extended previous research to gain insight into disease processes at the pathway level. We are able to study simultaneously the progression over multiple steps in tumor progression and provide pathway interaction networks of pathways relevant to these multiple steps. The same data analysis algorithm or pipeline was followed for both the prostate cancer and melanoma examples. The following are the analysis steps: 1. Stratify data: The expression data is stratified into T datasets corresponding to stages of progression. For example, if the progression is {b ↦ p ↦ m} then there are two steps, T = 2. The first dataset D1 consists of samples of class b and p and the second dataset D2 consists of samples of class p and m; 2. Map to summary statistics: The stratified data Dt is mapped into a representation with respect to sets of genes or pathways, Γ, defined a priori. Pathways in this setting are genes putatively thought to co-express. Given the stratified data Dt and a pathway database Γ, the summary statistic provides a measurement of the enrichment of each sample in Dt with respect to each pathway in the database. If the dataset Dt has p genes and n samples and there are m pathways in the database then the result of the summary statistic is a new dataset, St, of the enrichment of the m pathways over the n samples. 3. Find pathways relevant to progression: The RML algorithm was applied to the mapped data. The output of the RML algorithm are a set of vectors where the elements of w0 correspond to the relevance of a pathway over all stages of progression and the elements of wt correspond to the relevance of a pathway with respect to the t-th step in progression. A permutation procedure was performed to obtain p-values for each gene set in respect to each step in progression; 4. Leave-one-out cross-validation: Given data set of enrichment scores, apply RML to training data in a leave-one-out setting. This results in an unbiased estimate of the error rate on the prediction of class labels. 5. Construct pathway association graphs: For the pathways found to be relevant over the T stages of progression construct a pathway association graph. Each ij-th element of At indicates the dependence between pathways i and j conditioned on all other pathways and the relevance in modeling transition t. 6. Refine relevant gene sets: The pathways Gt found to be relevant for the t-th step in progression are refined since not all genes in the pathway are relevant in modeling transition t. This results in a set of refined gene sets ΓR and g = | ΓR | is the number of refined gene sets. 7. Construct gene association graphs for refined pathways: For each refined gene set in ΓR = construct a gene association graph Ãk where the ij-th element of Ãk indicates the dependence between genes i and j conditioned on all other genes γk and the relevance in modeling transition t. The first two steps in the analysis pipeline are stratifying the data and mapping the data into a representation based on pathways. The data can be represented as a set of pairs D = with xi ∈ p the expression over p genes and yi is the stage of the patient. Assume that there are three stages y ∈ {b, p, m} with n1, n2, n3 samples in each stage and the progression is {b ↦ p ↦ m}. There are two steps in this progression, T = 2. We first stratify the data with respect to these two steps. The first dataset D1 = consists of the n1 samples corresponding to stage b followed by the n2 samples corresponding to stage p with the label of the first n1 samples labeled as 0 (less serious) and the remaining n2 labeled as 1 (more serious). The second dataset D2 is constructed similarly with samples corresponding to stages p and m. The first n2 samples are labeled as 0 (less serious) and the remaining n3 labeled as 1 (more serious) in this dataset. Each dataset Dt is then mapped into a representation with respect to sets of genes or pathways. This is done using the pathway annotation tool ASSESS [9] that assays pathway variation in individuals. Given phenotypic or label data Yn = {y1, …, yn}, expression data Xn = {x1, …, xn} and gene sets Γ = {γ1, …, γm} defined a priori ASSESS [9] provides the summary statistic Sn = S (Xn, Yn, Γ}. The summary statistic Sn is a matrix with n columns corresponding to samples and m rows corresponding to gene sets with each element Sij as the enrichment of gene expression differences in the j-th sample with respect to phenotype for genes in the i-th gene set. The application of ASSESS to the stratified datasets D1, D2 results in two datasets S1, S2. The gene sets used in our analysis were those annotated in the MSigDB [8]. The central statistical idea used in finding pathways relevant to progression is called multi-task learning [12] in the machine learning literature and hierarchical modeling with mixed effects in the statistics literature. We restrict ourselves in this paper to linear models and classification. The basic idea is that we have T classification problems in our case assigning a sample xi to labels 0 (less serious) or 1 (more serious). We assume that the classification tasks are related so the conditional distributions of the phenotype given the summary statistics μt (Y | S) are also related. The tasks in our case are the different steps in tumor progression and the data over all tasks is S = S1, . . . , ST where Sj = { (y1j, s1j), . . . , (ynj, Snj) } and nj is the number of samples in the j-th task. We assume the generalized linear model where g is a link function which for the SVM case is (yit − sit · wt − b) +, wt = w0 + vt, yit is the i-th sample in task t, sit are the summary statistics of the i-th sample in task t, w0 is the baseline term over all tasks, vt are the task specific corrections, and b is an offset. The vectors wt correspond to the linear model for each task. We used the RML framework developed in multitask to estimate the model parameters w0, vt, and b where (u) + = min (u, 0) is the hinge loss, f (sit) = sit · (wo + vt) + b, λ1 and λ2 are positive regularization parameters that trade-off between fitting the data and the smoothness or robustness of the estimates. In this paper we set λ1 = 1 and λ2 = 2,000, as to not over assume dependency between tasks. Given the vectors wt we select gene sets corresponding to coordinates of the vectors with p ≤ 0. 05 to find pathways relevant to the t-th step in progression, see [56] for details. RML was repeated with 1,000 class label permutations to obtain a null distribution of each vector: . The p-value for each gene set in wt was obtained by finding the percentile of wt in. Gene sets relevant to the t-th step in progression, Gt, are those corresponding to elements of wt with p ≤ 0. 05. We applied the leave-one-out procedure for classification. The dataset is split into si (the i-th data sample) and S\i (the data without the i-th sample). RML is applied to the training set, S\i to build a classifier based on which is applied to si to obtain a prediction. Prediction accuracy is computed by applying the leave-one-out procedure to all samples in the dataset. The central statistical idea used in constructing association graphs as well as refining gene sets is learning gradients [13–16] and inverse regression [17,18]. These ideas apply to linear and nonlinear models but we restrict ourselves to the linear setting since we only use linear models in this paper. We first formulate the statistical ideas and then describe how this is applied in our application. Learning gradients, inverse regression, and graphical models: The idea of inverse regression [17,18] is given the explanatory or input variables X and the output or response variable Y to study X|Y and more specifically ΩX|Y = cov (X | Y): 1. The i-th diagonal element of this covariance matrix is a measure of the relevance of the i-th variable with respect to changes in the label or output variable; 2. The j-th off-diagonal element is a measure of the covariation between variable i and j with respect to changes in the label. Estimating the inverse regression can be technically problematic if the covariance matrix is degenerate. The idea of learning gradients addresses this technical problem. Given a regression or classification function f its gradient is and the gradient outer product matrix (GOP) Γ is defined by its elements Γij =. In [16] the following relation between the GOP matrix and the covariance of the inverse regression ΩX|Y was derived for linear functions where ΣX = cov (X) for the inputs, and = var (Y) for the outputs. This states that Γ and ΩX|Y are equivalent modulo a scale parameter (the variance of the output variable) and a rotation (the precision matrix of the input variables). In [14] an efficient algorithm to estimate the gradient and GOP matrix, Γ̃ given data for the classification setting was developed. We use this method in this paper. The estimate of the GOP matrix Γ̃ is the covariance matrix of a multivariate Gaussian random vector by construction. The inverse of this matrix J = Γ̃−1 is by the theory of Gauss-Markov random fields [19] the conditional independence matrix (the pseudo-inverse is used when Γ̃ is degenerate), Jij = dependence between variables i and j | all other input variables and the output variable. Construction of pathway association graph: Given the list Gt of pathways relevant to the t-th step in progression and the enrichment summary statistic dataset St, this dataset is reduced by removing the enrichment scores of all pathways not in Gt. The GOP estimate, Γ̃t, given this data is computed. This matrix is d × d where d is the number of relevant pathways in Gt. The pseudo-inverse of this matrix At = is the pathway association graph for the t-th progression step. Gene set refinement and gene association graph: Not all genes in the relevant gene sets, Gt, are differentially expressed between the two stages of progression in step t. We reduce or refine each of the pathway gene sets to those genes most relevant in progression step t. The following procedure is iterated for each of the relevant gene set in Gt. Given the k-th pathway gene set the stratified data Dt is reduced to the genes in this gene set. The GOP estimate, Γ̃t, given this data is computed. This matrix is d × d where d is the number of genes in the k-th pathway in Gt. Genes corresponding to large diagonal elements of the GOP matrix, > τ, are those most relevant to the prediction and are the refined set R. In this paper, the threshold τ is selected such that we obtain a specific number of genes. The GOP estimate, Γ̃t, is then reduced to only the genes in the refined gene set R. The pseudo-inverse of this matrix is the gene association matrix that provides the dependence of the refined gene set. Input: data D, thresholds (τ1, τ2), gene sets Γ, RML algorithm M, graph algorithm A, refinement algorithm R. Return: relevant gene sets, pathway association graphs, refined gene sets ΓR, gene association graphs T = number of steps in progression; for t ← to T do / / Stratify data by taking subset relevant in step t Dt ← D; / / Map Dt to enrichment summary statistics over gene sets St = S (Dt, Γ); ; / / Apply RML to the summary statistics / / Select relevant gene sets over stages of progression for t ←1to T do Gt = Ø; Initialize relevant gene sets for stage t for all elements of wt do if |wti|> τ1 then add gene set i to Gt; / / Construct pathway association graphs for t ← 1 to T do At ← A (Gt, Dt); / / Refine relevant gene sets / / Construct gene association graphs for refined gene sets for i = 1 to |ΓR| do ← A (Dt, ΓR);

Cancer is a complex disease with many subtypes that differ substantially with respect to their onset, progression, and response to treatment. Better understanding of the etiology and mechanism of cancer should help improve the diagnosis, prognosis, and treatment of cancer that will kill more than half a million Americans this year alone. Our study illustrates how integration of data over multiple stages and modeling tumorigenesis at the level of regulatory pathways or sets of genes provide robust and interpretable novel hypotheses concerning root genetic causes responsible for cancer initiation, progression, and invasion. Our modeling approach is one of the first approaches that combines multiple microarray datasets in a truly integrative framework that promotes the interpretability of important factors or pathways in one or more datasets. We apply this analysis of tumor progression to both prostate cancer and melanoma to provide information that can lead to the identification of novel biomarkers and give a basis for how genetic disruptions serve to alter actions in specific cell types.

Abstract Introduction Results Discussion Methods

computer science mathematics computational biology homo (human) genetics and genomics

2008

Modeling Cancer Progression via Pathway Dependencies

The firing rate of single neurons in the mammalian hippocampus has been demonstrated to encode for a range of spatial and non-spatial stimuli. It has also been demonstrated that phase of firing, with respect to the theta oscillation that dominates the hippocampal EEG during stereotype learning behaviour, correlates with an animal' s spatial location. These findings have led to the hypothesis that the hippocampus operates using a dual (rate and temporal) coding system. To investigate the phenomenon of dual coding in the hippocampus, we examine a spiking recurrent network model with theta coded neural dynamics and an STDP rule that mediates rate-coded Hebbian learning when pre- and post-synaptic firing is stochastic. We demonstrate that this plasticity rule can generate both symmetric and asymmetric connections between neurons that fire at concurrent or successive theta phase, respectively, and subsequently produce both pattern completion and sequence prediction from partial cues. This unifies previously disparate auto- and hetero-associative network models of hippocampal function and provides them with a firmer basis in modern neurobiology. Furthermore, the encoding and reactivation of activity in mutually exciting Hebbian cell assemblies demonstrated here is believed to represent a fundamental mechanism of cognitive processing in the brain. The hippocampus and surrounding medial temporal lobe are implicated in declarative memory function in humans and other mammals [1]. Electrophysiology studies in a range of species have demonstrated that the activity of single pyramidal cells within this region can encode for the presence of both spatial and non-spatial stimuli [2]. The majority of empirical investigation has focussed on place cells – neurons whose firing rate is directly correlated with an animal' s spatial location within the corresponding place field [3]. Subsequent research has identified similar single cell responses to a variety of non-spatial cues including odour [4], complex visual images [5], [6], [7], running speed [8] and the concept of a bed or nest [9]. It has also been demonstrated that the exact timing of place cell discharge, relative to the theta oscillation which dominates the hippocampal EEG during learning, correlates with distance travelled through a place field [2], [7], [10]–[12]. This phase precession mechanism creates a compressed ‘theta coded’ firing pattern in place cells which corresponds to the sequence of place fields being traversed [13]. These findings have led to the hypothesis that the hippocampus operates using a dual rate and temporal coding system [14], [15]. Here we present a spiking neural network model which utilises a dual coding system in order to encode and recall both symmetric (auto-associative) and asymmetric (hetero-associative) connections between neurons that exhibit repeated synchronous and asynchronous firing patterns respectively. The postulated mnemonic function of the hippocampus has been extensively modelled using recurrent neural networks, and this approach is supported by empirical data [16]–[19]. The biological correlate of these models is widely believed to be the CA3 region, which exhibits dense recurrent connectivity and wherein synaptic plasticity can be easily and reliably induced. Pharmacological and genetic knockout studies have demonstrated that NMDAr-dependent synaptic plasticity in CA3 is critical for the rapid encoding of novel information, and synaptic output from CA3 critical for its retrieval [20], [21]. Recurrent neural network models of hippocampal mnemonic function have generally utilised rate-coded Hebbian learning rules to generate reciprocal associations between neurons with concurrently elevated firing rates [22], [23]. Hypothetically, this corresponds to the presence of either multiple stimuli or multiple overlapping place fields encountered at a single location [24]–[27]. The hippocampus is also implicated in sequence learning, and temporally asymmetric plasticity rules have subsequently been employed in recurrent network models to generate hetero-associative connections between neurons that fire with repeated temporal correlation [28]–[38]. Hypothetically, this corresponds to a sequence of place fields being traversed or stimuli being encountered on a behavioural timescale [13]. Importantly, previous computational models of hetero-associative learning have typically encoded each successive stage of a learned sequence with the activity of a single neuron, while empirical studies estimate that place fields are typically encoded by an ensemble of several hundred place cells [2], [39]–[45]. No computational model has thus far integrated auto- and hetero- associative learning in order to simultaneously generate both bi-directional and asymmetric connections between neurons that are active at the same and successive theta phases respectively using a single temporally asymmetric synaptic plasticity rule. Empirical data indicates that changes in the strength of synapses within the hippocampus can depend upon temporal correlations in pre- and post- synaptic firing according to a spike-timing dependent plasticity (STDP) rule [46]–[49]. It is not yet clear if rate-coded auto-associative network models of hippocampal mnemonic function are compatible with STDP or theta coded neural dynamics. Here, we examine the synaptic dynamics generated by several different STDP rules in a spiking recurrent neural network model of CA3 during the encoding of temporal, rate and dual coded activity patterns created by a phenomenological model of phase precession. We demonstrate that – under certain conditions - the STDP rule can generate both bi-directional connections between neurons which burst at concurrent theta phase and asymmetric connections between neurons which fire at consecutive theta phase. Subsequent superthreshold stimulation of a small number of simulated neurons generates putative recall activity, driven by recurrent excitation, that corresponds to pattern completion and/or sequence prediction in auto- and/or hetero- associative connections respectively. Interestingly, these neural dynamics are reminiscent of sharp wave ripple activity observed in vivo [50]–[54]. These findings demonstrate that STDP and theta coded neural dynamics are compatible with rate-coded auto-associative network models of hippocampal function. Furthermore, the encoding and reactivation of dual coded Hebbian phase sequences of activity in mutually exciting neuronal ensembles demonstrated here has been proposed as a general neural coding mechanism for cognitive processing [50], [55]–[60]. The neural network consists of 100 simulated excitatory neurons which, in the majority of simulations, are fully recurrently connected by single synapses except for self connections. Although the level of recurrent connectivity present in the CA3 region is estimated as 5–15% (and is non-random), full recurrent connectivity has most often been employed in previous computational models of auto- associative learning [16]–[19], [39]. However, all simulations described here were also performed using networks with more realistic levels of recurrent connectivity (15 separate pre-synaptic connections per simulated neuron, chosen from a random uniform distribution that excludes self-connections) and no significant differences were observed (data not shown). Simulated pyramidal cells operate according to the Izhikevich spiking model [61], which can replicate the firing patterns of all known types of cortical neurons with minimal computational complexity. The membrane potential (v) and a membrane recovery variable (u) are dynamically calculated based on the values of four dimensionless constants (a, b, c and d) and a dimensionless current input (I) according to Equations 1. 1–1. 3. (1. 1) (1. 2) (1. 3) The parameter values used to replicate firing of a standard excitatory neuron are [a = 0. 02, b = 0. 2, c = −65, d = 6]. Under these conditions, simulated neurons fire single spikes at low levels of stimulation, but produce complex bursts that are representative of hippocampal pyramidal cells (i. e. several action potentials at a spontaneous rate of ∼150Hz) at higher levels of stimulation [2], [62]. Further details of the dynamics produced by single simulated neurons in response to various forms of applied current can be found in Izhikevich (2004). Each simulated neuron has an axonal delay (Di) randomly assigned from a uniform distribution in the range [1ms: Dms] with D = 5 in the majority of simulations (this being realistic of the CA3 region [63]). At the beginning of each millisecond time step, before the parameters v and u are updated, any membrane potential values that exceed threshold are reset according to Equation 1. 3. The corresponding neuron (s) are considered to have fired in that time step (t*), and the corresponding spikes arrive at their post-synaptic targets at time t*+Di. The hippocampal EEG is dominated by both theta and gamma oscillations during stereotype learning behaviour [39], [43], [64], [65]. Here, we include only a minimal model of theta frequency inhibition. A variable θ, which oscillates sinusoidally in the range [0: 1] at a rate of 8Hz throughout all learning simulations, is used to dynamically represent the theoretical local field potential (LFP). Inhibitory input to every simulated neuron at each millisecond time step is randomly sampled from a Gaussian distribution with mean Iinh = −15θ and standard deviation σinh = 2. Neural noise at a rate of ∼0. 1Hz (this being realistic of the CA3 region) is generated in the network by the constant application of excitatory current, randomly sampled from a uniform distribution in the range [0: Inoise] where Inoise = 0. 8 in all simulations [66]. The interplay between afferent inhibitory and excitatory currents means that the majority of firing due to neural noise tends to occur around the peak of the LFP, as defined by the value of θ. Place cells are most often studied in the dorsal CA1 region of the hippocampus, although some data is available from CA3 and, importantly, significant differences can be observed [14], [67]–[69]. Approximately 30% of CA3 pyramidal cells are active in a typical environment, each of which can encode for several (occasionally overlapping) place fields of ∼30cm in size (although this varies along the septotemporal axis) [10], [12], [70]. The phase precession of place cell firing can cover a full theta cycle, but typically changes by 180° between entry and exit, and is correlated with both the relative distance travelled and time spent within a place field (i. e. the rate of phase precession is positively correlated with running speed) [10], [12], [71]. The firing rate of active place cells follows a Gaussian distribution, such that maximum firing rate occurs around the centre of the place field [10]. In CA3, the mean in-field firing rate of place cells is ∼15Hz, although this is strongly modulated by various non-spatial cues [14], [66], [68]. There is considerable debate regarding the mechanisms of phase precession in place cells, fuelled by apparently contradictory empirical findings [72]. Here, we are more directly concerned with the manner in which theta coded neural dynamics interact with local plasticity rules in order to mediate the learning and recall of auto- and hetero- associative connections between active neurons. Hence, the phase at which simulated neurons in our network model fire is primarily dictated by external excitatory input, although it is important to note that phase precession has been empirically observed in both the dentate gyrus and entorhinal cortex, which constitute the two principal synaptic inputs to CA3 [12], [73]. Furthermore, detailed biophysical simulations suggest that input from these afferent structures plays a significant role in dictating the neural dynamics observed in CA3 [74], [75]. During learning simulations, each place field is arbitrarily divided into eight equally sized sub-sections, and theta oscillations in the LFP (as defined by the value of θ) are similarly divided into subsections of π/4 (between π/8 and 15π/8). At each millisecond time step, the theoretical position within a place field dictates the theta phase window at which the corresponding place cell receives external excitation, randomly sampled from a normal distribution with mean Iext and standard deviation σext (Figure 1a, b). This phenomenological model dictates that the mean phase of (stochastic) activity in place cells recedes in a step-wise fashion as the corresponding place field is traversed. In the majority of simulations, values of Iext = 5 and σext = 22. 5 are used to generate a mean in-field firing rate of ∼15Hz, with active place cells tending to fire bursts at the peak of the LFP (as defined by the value of θ) and single spikes on the ascending and descending slope [10]. In other simulations, values of σext = [12. 5; 32. 5; 42. 5; 52. 5] are used to generate a range of mean in-field firing rates. In these simulations, hypothetical place fields are generally 80cm in diameter and traversed at a rate of 10cms−1. Although this place field size is larger than that typically observed in vivo [10], [12], these values are chosen for computational convenience such that active place cells fire stochastically in each theta phase window for a period of 1s before receding. Simulations were also performed using place field diameters of [10cm; 20cm; 40cm] – which effectively reduces the duration of time for which each theta coded stage of the learned sequence is applied to the network – and the only significant effect observed was a decrease in the rate of synaptic weight change (data not shown). The phase precession of place cells in the hippocampus produces a compressed, theta coded, sequence of firing within each oscillatory cycle that corresponds to the sequence of overlapping place fields being traversed on a behavioural timescale Figure 1c; [12], [13], [34]. These firing patterns are ideally suited to induce the long-term potentiation (LTP) and depression (LTD) of synapses by spike-timing dependent plasticity (STDP), and there is evidence that synaptic connections between overlapping place cells in rat hippocampus are potentiated during exploration [47]–[49], [76]. Mathematically, with s = tpost−tpre being the time difference between pre- and post- synaptic spiking, the change in the weight of a synapse (Δw) according to a standard STDP rule can be calculated using Equations 2. 1–2. 5 [77]–[82]. (2. 1) (2. 2) (2. 3) (2. 4) (2. 5) The parameters A+ and A− correspond to the maximum possible change in synaptic weight per isolated spike pair, while τ+ and τ− denote the time constants that approximate an exponential decay of potentiation and depression increments respectively. The co-efficient ε determines the contribution of an additional potentiation process, which is equal to a trace of the most recent weight decrease at a synapse (with s++ = tΔw+−tΔw−) decaying exponentially with a time constant τ++. This term accounts for experimental observations of STDP in the hippocampus obtained using triplets of pre- and post- synaptic spikes, which suggest that depression is suppressed by potentiation within a short temporal window [48]. In accordance with empirical data, coincident pre- and post- synaptic firing elicits maximal depression from all STDP implementations examined here [49]. Previous auto-associative network models of hippocampal mnemonic function have most frequently utilised rate-coded Hebbian learning rules [16], [17], [22], [23], [83] – which typically dictate that changes in synaptic strength are proportional to the product of pre- and post- synaptic firing rates (rj, i) scaled by a learning rate k (Equation 3. 1). This form of synaptic plasticity generates no competition between inputs or outputs of a single neuron, as any increase in synaptic weight produces an increase in post-synaptic firing rate in a positive feedback loop [84]. The BCM model (Equation 3. 2) was proposed to address this issue, and postulates the existence of a theoretical modification threshold (θm) that distinguishes between depression (at lower firing rates) and potentiation (at higher firing rates). The value of θm is itself a function of pre- or post- synaptic activity, generating competition between synaptic inputs by making potentiation more difficult to achieve as the long-term average firing rates increases [85]. (3. 1) (3. 2) Interestingly, it has been demonstrated that STDP can provide inherent competition using only local synaptic variables, and thus stabilise Hebbian learning processes [80], [82]. However, these properties rely on synaptic weights being either depressed or unchanged following an increase in pre-synaptic stimulation, which directly contradicts empirical data and the requirements of rate-coded associative learning. Conversely, several computational studies have described conditions under which STDP can be reconciled with the BCM formulation [77]–[79], [81]. This requires that the plasticity rule exhibit an increasing dominance of potentiation processes as inter-spike intervals (ISIs) are reduced [77]. Pair-based STDP rules, which assume a linear integration of potentiation and depression processes, require constraints to be placed on the nature of spike pair interactions and parameters that define the asymmetric learning window [77], [78], [81]. Triplet-based STDP rules, which explicitly account for the observed non-linear integration of potentiation and depression processes, dictate that mean synaptic weight increases with mean stochastic firing rate irrespective of the finer details of the STDP rule [77], [79]. We examine three different additive STDP implementations here, in order to draw a comparison between the emergent synaptic dynamics produced by each. The ‘BCM type’ pair- and triplet- based STDP rules have parameter values described in previous studies as allowing a reconciliation with rate-coded Hebbian learning (A+>A− and τ+<τ−), which also concur with empirical measurements made in the hippocampus [47], [49], [78], [81]. Conversely, the ‘non-BCM type’ pair-based STDP rule has parameter values noted in previous modelling studies for the generation of synaptic competition (A+<A− and τ+ = τ−) [80], [82]. For each of these STDP rules, a lax nearest neighbour spike pairing scheme – which dictates that values of P± are reset to the value of A± upon afferent or efferent firing – is employed. Values of A± are also scaled by the value of wmax such that ∼60 spike pairings are sufficient to traverse the range of possible synaptic weight values, in accordance with empirical data [47]–[49]. The full details of each plasticity model examined are given in Table 1. Empirical evidence indicates that the degree of synaptic plasticity incurred by consistent stimulation protocols differs across the theta cycle, with potentiation incurred by burst pairings at the peak and depression (or de-potentiation) incurred by burst pairings at the trough of the LFP [86]–[89]. Here, we examine the effects of three different forms of theta modulated plasticity, for comparison, the details of which are shown in Equation 4. (4) In all simulations, hard limits are placed on the achievable strength of synapses, such that synaptic weights are maintained continuously in the range [0: wmax]. While there is little clear biological basis to inform the relative scale of synaptic weights, it is known that recurrent synapses in the CA3 region are generally incapable of solely provoking post-synaptic activity [90]. In order to generate an action potential using the neural dynamics employed here, a single synaptic current of I = 16. 5 is required, and therefore the value generally assigned to the maximum weight limit in these simulations is wmax = 1. In each simulation, all synaptic connections in the network are initialised with a weight of 0. 01wmax. The neuromodulatory effects of Acetylcholine (ACh) have been hypothesised to separate periods of learning and recall in the hippocampus in order to avoid issues of interference [35], [91], [92]. Cholinergic input from the septum, terminating on local interneurons, can induce theta frequency oscillations in the CA3 region, facilitate LTP and enhance afferent input from the dentate gyrus and entorhinal cortex while suppressing recurrent excitation from intrinsic connections – thereby creating the ideal conditions for learning external associations via theta coding [92]. In the absence of cholinergic input pyramidal neurons in CA3 are disinhibited, synaptic plasticity is suppressed, and neural dynamics enter a state of large-amplitude irregular activity (LIA). During this period, postulated recall activity is observed in the form of sharp wave ripples (SWR) – short periods of high frequency firing in large populations of neurons with fine temporal structure that last ∼100ms and originate in CA3 [50]–[53], [69]. In our model, an abstract, global ACh signal modulates the scale of recurrent excitation and synaptic plasticity in the network throughout all simulations. The hypothetical concentration of ACh maintains a dimensionless value of Φ = 1 during periods of theta coded learning and falls to a lower value during periods of recall. In both cases, the relative magnitude of recurrent synaptic weights in the network is scaled by a factor of 1/Φ, while the magnitude of synaptic weight change is scaled by a factor of Φ. During periods of recall, theta frequency inhibitory input to the network is ceased and superthreshold excitation of magnitude Icue = 30 is provided to a small number of randomly selected neurons for a single millisecond time step. Subsequent activity - dictated by recurrent excitation alone - can then be compared to the auto- and hetero- associative correlations present in external input during learning and SWR activity observed in vivo. In these simulations, the effective speed of recall is strongly dependent on the size and overlap of place fields, as a reduction in place field size and offset implies a reduction in the total length of the learned route, such that the same temporal compression of recall firing equates to a slower traversal of that route. We use several different measures to assess the fidelity of putative recall activity in this model. For hetero-associative and dual coded activity patterns, we examine the timing of the first action potential fired by each simulated place cell: firing before the first action potential in any place cell encoding for the following place field on the learned route is considered to be accurate, firing at the same time as the first action potential in any place cell encoding for the following place field on the learned route is treated indifferently, and failure to fire or firing after the first action potential in any place cell encoding for the following place field on the learned route is considered to be erroneous. For auto-associative patterns, we examine firing in all simulated neurons for a period of 20ms following the external stimulation of a subset of ‘cued’ neurons from one of the learned activity patterns. Firing in any of the neurons from that learned pattern which are not externally stimulated (‘uncued’) during this period is considered to be accurate recall, while activity in any neuron that is not part of that pattern is considered to be erroneous. The Mann-Whitney U test is used to assess the significance of differences in the strengths of disparate populations of synaptic connections throughout this paper. Lesions of the hippocampus have been demonstrated to disrupt the temporal ordering of information in memory, impairing recall of a sequence of locations visited [93], [94], olfactory cues presented [4], [95], [96], and trace eyeblink conditioning performance [97], [98]. This has led to the theory that the hippocampus – which exhibits sparse connectivity, temporally asymmetric synaptic plasticity, and theta coded neural dynamics - is critical for sequence learning and predictive recall [13], [28]–[37]. Hence, in the first set of simulations, we examine the learning of theta coded activity patterns in single neurons. In these simulations, place cell firing corresponds hypothetically to ten traversals of a route of one hundred equidistant and overlapping place fields of 80cm diameter, each encoded by a single neuron, at a constant speed of 10cms−1 (Figure 2a). However, this form of activity could just as easily correspond to a temporal sequence of non-spatial stimuli encountered on a behavioural timescale [29], [40], [64]. The spike raster shown in Figure 2b is representative of the neural dynamics generated by the phenomenological theta coding model, which replicates the gross features of phase precession observed in the hippocampus in vivo. Figure 2c illustrates the typical asymmetric weight matrix that develops – with connections between each place cell and those that follow it on the theoretical route being selectively and significantly potentiated to create a bi-modal distribution of synaptic strengths (inset). Figure 2d illustrates the asymmetric expansion of place fields that proceeds over the course of these simulations, a phenomenon that has been observed experimentally [99]. This results from an increase in excitatory input to each place cell from those preceding it on the route as recurrent connections are potentiated, and the magnitude of place field expansion is therefore correlated with the value assigned to the maximum excitatory synaptic weight (wmax). It is important to note that the particular details of the STDP rule utilised here makes little difference to the efficient learning of hetero-associative sequences (Figure 2e). Furthermore, the strength of asymmetric connections saturates at the upper bounds regardless of whether neurons fire bursts or single spikes throughout each theta cycle – although mean in-field firing rate is correlated with the rate of synaptic weight change (Figure 2f). These results demonstrate that the combination of theta coding and STDP in a spiking recurrent network is sufficient to mediate rapid and robust sequence learning, irrespective of the finer details of the plasticity rule, in accordance with several previous models [12], [13], [28]–[37]. Although the majority of electrophysiology studies have focused on spatial memory, there is a growing body of evidence to suggest that non-spatial stimuli are also encoded in the activity of single neurons in the hippocampus and can significantly modulate the firing rate of established place cells [4]–[7], [9], [14], [57], [68], [100]. Computational theories of episodic memory function generally posit that discrete patterns of rate-coded activity, corresponding to the conjunctively coded sensory elements that constitute an experience, are auto-associated in the recurrent connections of CA3. This cortical activity can subsequently be fully recreated from partial sensory cues via a process of pattern completion [16]–[23], [68], [101]. However, auto-associative network models of the CA3 region have often been criticised on the grounds of biological realism for failing to include realistic neural and synaptic dynamics [39], [102]. Furthermore, it has been suggested that the inherently asymmetric nature of STDP directly contradicts rate-coded associative learning, which explicitly depend on the development of strong bi-directional connections [13], [82], [ but see 41], [42], [44], [45], [77]. Here, we examine whether auto-associative learning can be achieved in a network model that incorporates the main features of neural and synaptic dynamics observed in CA3 – namely, phase precession and STDP. In these simulations, input to the network effectively corresponds to ten presentations of ten binary and orthogonal activity patterns, in accordance with previous auto-associative network models [22], [23]. However, this form of input could also correspond to ten traversals of a route of ten non-overlapping place fields of 80cm diameter, each encoded by the activity of multiple place cells, at a constant speed of 10cms−1 (Figure 3a). The phenomenological phase precession model implemented dictates that neurons which are active in the same pattern (i. e. place cells that encode for the same place field) fire stochastically with equal mean phase, the value of which decreases in a step-wise fashion over the course of a single presentation (Figure 3b). Our results demonstrate that successful auto-associative learning depends on a plasticity rule that produces net potentiation at the high instantaneous firing rates (i. e. short ISIs) present during near-synchronous firing in bi-directionally connected neurons [77]. Accordingly, both the pair- and triplet- based BCM type STDP rules selectively and significantly potentiate synaptic connections between place cells that encode for the same place field (Figure 3c, Mann-Whitney U-test, p<0. 01). Conversely the non-BCM type STDP rule produces net depression of synaptic connections between concurrently active neurons (Figure 3d). This demonstrates that efficient auto-associative learning can be achieved in a spiking recurrent neural network when an STDP rule that can be reconciled with rate-coded Hebbian learning is employed, and that this function is fully compatible with theta coded neural dynamics created by the phase precession of principal cells in vivo. However, in contrast to the hetero-associative learning simulations described above, the mean weight of auto-associative connections in simulations with BCM type STDP rules generally reaches an asymptote well below the upper bounds - exhibiting a bi-modal distribution (Figure 3c, inset) except where potentiation and depression processes are inversely modulated (Figure 3e). This is a consequence of the persistently alternating temporal order of spike pairings at these synapses, which produces an equilibrium between potentiation and depression processes. The position of this equilibrium is significantly affected by several features of the neural dynamics and synaptic plasticity rule employed. For example, the asymptotic mean weight of auto-associative connections increases with mean in-field firing rate for the BCM type STDP rules, but decreases with mean in-field firing rate for the non-BCM type STDP rule (Figure 3f). In both cases, the rate of synaptic weight change (whether positive or negative) correlates with the mean in-field firing rate. In these simulations, spike pairing events that dictate changes in synaptic strength do not take place immediately following firing, but rather once an action potential reaches the pre-synaptic terminal. Hence, the range of axonal delays (D) can also have a significant impact on the relative strength of auto-associative connections. In bi-directionally connected neurons exhibiting near-synchronous bursting, longer axonal delays imply that the arrival of spikes at pre-synaptic terminals is more likely to occur after post-synaptic firing and therefore generate depression; while shorter axonal delays imply that afferent spikes are more likely to precede post-synaptic activity and therefore generate potentiation (Figures 4a, b). Accordingly, for the BCM type STDP rules examined here, both the rate of potentiation and asymptotic mean weight of auto-associative connections increase as the scale of axonal delays is decreased (Figure 4c). This also explains why, in Figure 3c, the strength of auto-associative post-synaptic connections formed by some place cells is uniformly weak, as the axonal delay of that neuron is higher than others encoding for the same place field. Similarly, for BCM type STDP rules, parameters that dictate the profile of the asymmetric learning window (A± and τ±) effectively define the position of the theoretical modification threshold (θm, Equation 3. 2) that marks the transition between net synaptic depression (at low stochastic firing rates) and potentiation (at high stochastic firing rates) [41]–[43], [64]. Hence, lowering the theoretical modification threshold – by increasing the value assigned to A+, for example – produces a greater degree of potentiation at set in-field firing rate, and therefore increases the asymptotic mean weight of auto-associative connections in these simulations (Figure 4d). It has been proposed that the sequential co-activation of groups of neurons during behaviour can be encoded via Hebbian plasticity [55], [56]. Subsequently, transient activity patterns in the same cell assembly can be initiated by internal cognitive processes and maintained via mutual excitation. Phase precession in ensembles of place cells encoding for overlapping place fields represents a prominent empirical model of cell assembly dynamics in the brain [50], [57], [59]. However, previous models of hippocampal mnemonic function have generally focussed on the learning and recall of either discrete rate-coded or sequential temporally-coded activity patterns, while few studies have attempted to integrate these computational models within a single framework [27], [39], [42]–[44]. Here, we demonstrate that both auto- and hetero- associative learning can proceed simultaneously in our network model, such that repeatedly synchronous firing with weak sequence bias produces bi-directional connections while repeatedly asynchronous firing produces asymmetric connections. Input to the network during these simulations corresponds to a route of twenty overlapping place fields of 80cm in diameter, each encoded for by five place cells, being traversed at a constant speed of 10cms−1 (Figure 5a). This form of input is equivalent to the repeated presentation of a sequence of binary orthogonal activity patterns [39]. Figure 5b illustrates a representative spike raster observed during these simulations, demonstrating how the phenomenological phase precession mechanism dictates that place cells encoding for the same place field fire stochastically within the same theta phase window while place cells encoding for successive place fields fire in successive theta phase windows. Figure 5c illustrates the typical synaptic weight matrix that develops during simulations with BCM type STDP rules, where synapses connecting each hypothetical place cell to those that encode for the same or successive place fields are rapidly, selectively and significantly potentiated (Figure 5d). Conversely, when the non-BCM type STDP rule is utilised (and potentiation and depression are not inversely modulated), then hetero-associative sequence learning proceeds robustly but strong, bi-directional auto-associative connections are not generated (Figure 5e). In fact, there is no significant difference in the asymptotic mean weight of auto- or hetero- associative connections generated in any of these dual coding simulations and those described above with equal parameter values (Figures 2e, 3e, f, 4c, d; 5f). The experimentally observed asymmetric expansion of place fields against the direction of motion during spatial learning (Figure 2d) also proceeds during dual coded learning simulations (data not shown). These results again demonstrate that an STDP rule which exhibits a dominance of potentiation at short ISIs – and can therefore mediate rate-coded Hebbian learning – is essential for efficient auto-associative learning to proceed. However, each of the BCM type STDP rules examined here exhibits a significant functional weakness: background connections (i. e. synapses between neurons which are not in the same or immediately successive patterns) undergo slight but continual potentiation throughout all simulations, indicating a lack of inherent synaptic competition (Figure 5d). Effectively, a positive feedback loop arises between the potentiation of a synapse and a reduction in the latency of post-synaptic firing following an identical pre-synaptic input. This lack of competition may be necessary to allow the development of strong, bi-directional connections using the asymmetric STDP rule, as the mean weight of background connections correlates with that of auto-associative connections in all simulations (Figure 5f), but is also reminiscent of the global stability issues commonly encountered by rate-coded Hebbian learning [84]. Electrophysiology studies have demonstrated that learned routes – corresponding to the theta coded activity patterns observed in place cells during exploration – are pre-played in sharp wave ripples (SWR) at the beginning of (and during) a journey, replayed in reverse order at the end of a journey, and replayed in the original order during sleep [51]–[53]. The temporal order and relative latency of firing observed during exploration is preserved during this rehearsal and replay activity, which suggests a Hebbian learning mechanism on the timescale of STDP [50], [54]. Here, we examine the recall activity generated by recurrent excitation in our network under similar conditions - when theta frequency inhibitory input is ceased, the hypothetical concentration of ACh is reduced (to modulate the magnitude of recurrent excitation and synaptic plasticity), and superthreshold external excitation is applied to small numbers of neurons. This activity can then be compared to both the auto- and hetero- associations created during learning in the simulations described above and SWR activity observed in vivo. Firstly, we examine sequence prediction following hetero-associative learning. As illustrated by Figure 6a, superthreshold stimulation of a single, randomly selected neuron typically produces accurate sequential firing in all neurons that constitute the original learned pattern over a period of ∼400ms. Over one thousand separate recall epochs, the fidelity of recall activity produced is typically ∼90% for every STDP rule and plasticity modulation scheme examined (Figure 6b). The sequential firing patterns observed in these recall simulations continue indefinitely in the absence of inhibitory input to suppress the effects of recurrent excitation. This is a product of the fact that each neuron has few strong post-synaptic connections, and hence the concentration of ACh must be reduced to a level whereby the relative scale of recurrent synaptic weights allows single synapses to produce post-synaptic firing (Φ = 0. 05 in Figures 6a, b for example). Secondly, we examine pattern completion following auto-associative learning by providing superthreshold excitation to random partial cues consisting of five out of ten simulated neurons from each learned pattern. As illustrated in Figure 6c, the uncued neurons in each pattern are typically activated by recurrent excitation shortly after externally cued activity while other neurons in the network remain silent. The fidelity of recall activity produced in these simulations reflects the relative strength of auto-associative connections generated during learning (Figures 3e; 6f). However, pattern completion does not rely on an ‘idealised’ weight matrix: >90% accurate recall activity is produced following learning with the triplet-based BCM type STDP rule, which produce a mean auto-associative weight of ∼0. 7wmax. Furthermore, no erroneous activity (i. e. firing in neurons that are not part of the cued pattern) is produced following learning with any of the STDP rules over a thousand separate recall simulations (Figure 6d). Finally, we examine recall activity following dual coded learning by applying superthreshold stimulation to a randomly selected subset of simulated neurons (three out of five) that encode for a single theoretical place field on the learned route. As illustrated in Figure 6e, this generates sequential recall activity in all neurons encoding for each consecutive place field on the route over a period of ∼33ms. This activity is self-terminating and on approximately the same timescale as sharp wave ripples observed in vivo. Interestingly, strong auto-associative connections are not necessary to generate these sequential activity patterns in encoded place cell assemblies. Consistently high recall fidelity is produced following learning with the non-BCM type STDP rule, when only strong hetero-associative connections are generated (Figure 6f). In fact, the fidelity of recall activity is generally inversely correlated with the relative strength of auto-associative synaptic weights, regardless of the concentration of ACh employed. However, further simulations demonstrate that the relative scale of background synaptic connections contributes more significantly to erroneous recall activity than that of auto-associative connections – as arbitrarily setting the weight of all background connections to zero following dual coded learning generally eliminates all incorrect firing activity during subsequent recall (Figure 7a). Furthermore, the temporal error in recall activity following dual coded learning with BCM type STDP rules is generally low (Figure 7b), such that correct sequence prediction might be produced if one considers only the mean time of firing in all neurons that encode for a single place field. It is also interesting to note that recall fidelity consistently decreases over time, with the vast majority of erroneous recall activity occurring in the final ∼15ms of each putative sharp wave ripple event (Figure 7c). Intuitively, the effective speed of putative SWR activity – calculated using the time taken for sequential activity to progress through place cells encoding for the entire length of the 2m track – is significantly affected by the concentration of ACh present in the network (Figure 7d), which dictates the magnitude of recurrent synaptic currents. The effective speed of recall following hetero-associative learning simulations is significantly slower (∼25ms−1), due to the fact that fewer strong pre-synaptic connections (and therefore weaker recurrent synaptic currents) exist for each simulated place cell. Recurrent neural networks have an established history in computational neuroscience as prototypical models of declarative memory function [16], [17], [22], [23]. It is widely accepted that the CA3 region of the hippocampus – which contains the densest recurrent connectivity in the brain, and wherein synaptic plasticity can be rapidly and reliably induced – represents their biological correlate [18]–[21]. Despite their success in replicating key features of spatial and declarative mnemonic function, these models have often been criticised for their lack of biological realism in failing to integrate neural and synaptic dynamics which correspond to those observed in the hippocampus [39], [102]. In contrast, we have presented a spiking recurrent neural network that utilises theta coded neural dynamics and STDP to encode and recall both rate and temporally coded input patterns. This integrates previous auto- and hetero- associative network models of the hippocampus within a single framework using a single plasticity rule and provides them with a firmer basis in modern neurobiology. The encoding and reactivation of dual coded cell assemblies – putative phase sequences of activity in mutually exciting ensembles of cells – is believed to represent a fundamental mechanism for cognitive processing [55], [56], [58]. Our findings demonstrate that, under certain biologically feasible constraints, the temporally asymmetric STDP rule can replicate rate-coded Hebbian learning by generating strong bi-directional connections between neurons firing at an elevated rate with no repeated sequence bias [77]–[79], [81]. This implies that STDP can support rate-coded auto-associative network function and mediate cognitive map formation during open field exploration [3], [16], [17], [20], [22], [23]. The critical condition upon which this dual rate- and temporally- coded learning relies is that the magnitude of potentiation exceeds the magnitude of depression incurred by spike pair interactions at shorter ISIs. For pair-based STDP rules, this requires temporal restrictions on spike pairing and constraints on the profile of the asymmetric learning window, which concur with empirical measurements in the hippocampus [47], [77], [78]. For triplet-based STDP rules, it is implicitly generated by the short-term dominance of potentiation which, interestingly, is on a similar timescale to the duration of a single theta cycle [48], [79]. Conversely, STDP rules which do not dictate a dominance of potentiation at short ISIs prevent the development of strong bi-directional connections, except where synaptic plasticity is modulated such that only potentiation can proceed at the peak of the LFP. Under these conditions, however, synaptic weights undergo net depression as mean in-field firing rate increases [80], [82]. Despite replicating the gross phenomenological features of rate- and temporally- coded synaptic plasticity data, the BCM type STDP rules examined here exhibit several emergent features that contradict empirical observations. Firstly, the additive nature of these plasticity rules generates bimodal weight distributions that are at odds with experimental measurements [103]. However, an additive STDP rule might better approximate the known bi-stability of synaptic strengths, and a unimodal distribution of maximum weight limits could account for their observed heterogeneity [104]. Previous computational modelling has also demonstrated that the synaptic dynamics produced by additive STDP rules can, under certain conditions, be qualitatively replicated by a multiplicative plasticity rule [77]. Secondly, empirical studies suggest that no depression is incurred at connections between place cells encoding for overlapping place fields in vivo [76]. In our model, a synaptic plasticity rule that accounts for this data would more fully potentiate auto-associative connections, although our results indicate that this is not necessary for efficient pattern completion. Furthermore, it is interesting to note that connections between place cells that encode for place fields with higher degrees of overlap appear to be more modestly potentiated in vivo see Figure 5G in [76]. Empirical studies of synaptic plasticity in the hippocampus have also demonstrated that the potentiation of asymmetric connections by STDP depends on post-synaptic bursting [46]. A plasticity rule that accounted for this data might therefore generate hetero-associative synaptic weights that rely explicitly on mean in-field firing rate, as observed for auto-associative connections in this study. This should allow the implications of rate re-mapping in pyramidal cells within CA3 – whereby the manipulation of non-spatial cues within an environment significantly modulates the firing rate of active place cells – to be examined [14], [68]. In this context, connections between place cells that exhibit high in-field firing rates during learning – indicating the current configuration of non-spatial stimuli within the corresponding environment – would be preferentially potentiated. During subsequent SWR activity, more complex transient dynamics within the global place cell assembly encoding for that environment might therefore be produced, according to the particular stimulus applied to the network and its relationship to previously encoded configurations. From a functional standpoint, our findings suggest that the synaptic competition described in several previous theoretical studies as a putative homeostatic mechanism is absent for BCM type STDP rules [77], [80], [82]. This is not surprising, considering the wealth of literature regarding the global instability of rate-coded Hebbian learning mediated by purely local variables [84], [85]. However, it does imply that the encoding of multiple, overlapping cell assemblies – as opposed to the single episodes examined here - could rapidly lead to the saturation of synaptic weights and interference during recall. Some additional mechanism – such as synaptic scaling, weight normalisation or metaplasticity – is therefore required to guarantee the long-term efficiency of network models that use BCM type STDP rules by preventing the slow potentiation of all connections, particularly since the strength of background connections in our network model has been shown to correlate with erroneous recall activity [84], [85]. It is interesting to note that empirical data suggests a broad dissociation between net synaptic potentiation during waking and net depression during sleep [105]. Within the context of modelling mnemonic function, any mechanism of synaptic competition is likely to affect the emergent dynamics of learning and recall in terms of the long-term stability of previously encoded associations. It is also useful to appraise the results presented here in terms of more general theories of hippocampal mnemonic function. The plasticity model implemented prevents the potentiation of synaptic connections between place cells corresponding to trajectories against the direction of motion, and therefore omits the possibility of reverse replay in encoded cell assemblies during sharp wave ripples [51], [54]. Interestingly, putative SWR activity in our simulations also proceeds an order of magnitude more quickly than that observed in vivo – with effective recall speeds of ∼80ms−1 (Figures 6e, 7d) compared to the ∼8ms−1 observed experimentally in CA1 [106]. Of course, the speed of SWR activity is strongly affected by estimates of place field size and overlap, which may differ significantly from the values used here. However, one critical abstraction in our network model may contribute to both the accelerated pace of SWR activity and the generation of erroneous recall activity following efficient dual coded learning, and that is the relative timescales of recurrent auto- and hetero- associative connections. Previous theoretical research has suggested that processes of pattern completion and sequence prediction in CA3 must operate on different timescales in order to effectively differentiate between neural activity corresponding to different stages of a putative phase sequence, and it is not clear how this could be achieved in a single network with a fixed range of axonal delays. It has therefore been suggested that different regions of the hippocampus may mediate auto- and hetero- associative learning at distinct sets of synapses using a single plasticity rule, such as that presented here [39]. Our model suggests that CA3 can feasibly implement auto- and/or hetero- associative learning and recall. However, we have also demonstrated that auto-associative connections are not necessary for the reactivation of dual coded cell assemblies, and it seems plausible that purely hetero-associative dynamics could account for the putative function of CA3 in the rapid encoding of novel information and subsequent pattern completion [20], [21]. Conversely, it is possible that auto-associative connections exist within CA3, where relatively short axonal delays (which we have demonstrated to be necessary for auto-associative learning) are observed; while hetero-associative connections may be located in polysynaptic feedback connections between CA3 and the dentate gyrus. Activity corresponding to sharp wave ripples, which are believed to originate in CA3, have been documented in the dentate gyrus during sleep [107]. Identifying the loci of auto- and hetero- associative synaptic connections in the hippocampus remains an open problem for empirical neuroscience. It seems feasible that simultaneous recordings from these two regions and/or or the pharmacologically induced inhibition of firing in granule cells during sleep could elucidate the relative contribution of each region to the replay of previously learned associations. The segregation of auto- and hetero- associative connections may also allow the reactivation of cell assemblies to proceed during encoding, rather than these processes being arbitrarily separated between different network states. Several converging strands of empirical research - as well as simple intuition - suggest that some element of prediction, based on prior experience, is present during periods of theta coded learning, including changes in place field geometry and predictive theta modulated activity in place cells at decision points on a maze task [65], [99], [108]–[110]. Indeed, it has been suggested that the phenomena of phase precession itself may be generated by self-propagating ‘recall’ activity in cell assemblies within the hippocampus [67]. In our model, hetero-associative connection delays are on the same timescale as those measured in inter-connected CA3 pyramidal cells (i. e. <5ms), and thus sequence prediction via recurrent excitation proceeds more quickly than theta coded activity corresponding to external input. It is possible that inhibition from different classes of interneuron, creating gamma oscillations within each theta cycle, and/or the modulated efficacy of recurrent excitation at different theta phases could selectively manipulate the timing of pyramidal cell firing [37], [39], [40], [43], [65]. Similarly, if the loci of hetero-associative connections are poly-synaptic feedback loops from the dentate gyrus, as discussed above, then the replay of sequences will be explicitly staggered and could therefore proceed between different (gamma) sub-cycles of the theta oscillation [39]. In summary, this research provides a synaptic plasticity rule that can mediate both rate and temporal coded learning within a spiking recurrent neural network. Furthermore, it provides an associative memory model that utilises this dual code in order to integrate the encoding and reactivation of both dynamic (spatial) and static (non-spatial) activity patterns. This allows manipulations of the plasticity rule, neuronal dynamics and neural network to be directly related to systems level function. Hebbian phase sequences of activity in mutually exciting cell ensembles, such as those examined here, have been postulated as a general mechanism of neural coding for cognitive processing [55], [58]. Support for this theory comes from recent empirical evidence from the hippocampus and pre-frontal cortex [57]–[60], [111], [112]. Furthermore, theoretical considerations are making it increasingly clear that cortical function cannot be characterised by fixed point attractor dynamics, and neural network models must therefore account for the transient dynamics observed in vivo [56]. This research provides a framework for an examination of how dual coded activity patterns could be encoded in recurrent synaptic connections and subsequently reactivated by ongoing internal or external dynamics.

Changes in the strength of synaptic connections between neurons are believed to mediate processes of learning and memory in the brain. A computational theory of this synaptic plasticity was first provided by Donald Hebb within the context of a more general neural coding mechanism, whereby phase sequences of activity directed by ongoing external and internal dynamics propagate in mutually exciting ensembles of neurons. Empirical evidence for this cell assembly model has been obtained in the hippocampus, where neuronal ensembles encoding for spatial location repeatedly fire in sequence at different phases of the ongoing theta oscillation. To investigate the encoding and reactivation of these dual coded activity patterns, we examine a biologically inspired spiking neural network model of the hippocampus with a novel synaptic plasticity rule. We demonstrate that this allows the rapid development of both symmetric and asymmetric connections between neurons that fire at concurrent or consecutive theta phase respectively. Recall activity, corresponding to both pattern completion and sequence prediction, can subsequently be produced by partial external cues. This allows the reconciliation of two previously disparate classes of hippocampal model and provides a framework for further examination of cell assembly dynamics in spiking neural networks.

Abstract Introduction Methods Results Discussion

neuroscience/cognitive neuroscience neuroscience/theoretical neuroscience computational biology/computational neuroscience

2010

Dual Coding with STDP in a Spiking Recurrent Neural Network Model of the Hippocampus

Intrathecal antibody synthesis is a well-documented phenomenon in infectious neurological diseases as well as in demyelinating diseases, but little is known about the role of B cells in the central nervous systems. We examined B cell and T cell immunophenotypes in CSF of patients with HTLV-1-associated myelopathy/tropical spastic paraparesis (HAM/TSP) compared to healthy normal donors and subjects with the other chronic virus infection and/or neuroinflammatory diseases including HIV infection, multiple sclerosis (MS) and progressive multifocal leukoencephalopathy. Antibody secreting B cells (ASCs) were elevated in HAM/TSP patients, which was significantly correlated with intrathecal HTLV-1-specific antibody responses. High frequency of ASCs was also detected in patients with relapsing-remitting multiple sclerosis (RRMS). While RRMS patients showed significant correlations between ASCs and memory follicular helper CD4+ T cells, CD4+CD25+ T cells were elevated in HAM/TSP patients, which were significantly correlated with ASCs and HTLV-1 proviral load. These results highlight the importance of the B cell compartment and the associated inflammatory milieu in HAM/TSP patients where virus-specific antibody production may be required to control viral persistence and/or may be associated with disease development. Various inflammatory neurologic diseases are associated with viral infections. These agents may cause direct cellular damage of infected cells associated with immunological alterations such as chronic activation, immunodeficiency and infiltration of inflammatory cells into the central nervous system (CNS) that underlie the pathogenesis of inflammatory neurologic disorders. Intrathecal antibody synthesis is a well-documented phenomenon in infectious and demyelinating neurologic diseases. Various viral infections of the CNS including polio, rabies, mumps, herpes simplex virus and Japanese encephalitis virus are characterized by intrathecal antibody production in cerebrospinal fluid (CSF) and/or presence of local antibody-secreting B cells (ASCs) [1,2]. While virus-specific antibodies play an important role in the control of viral infections in the CNS, intrathecal antibody synthesis has been associated with both protective and pathogenic functions in chronic infection and immune-mediated disorders of the CNS. Human T cell lymphotropic virus 1 (HTLV-1) is a human retrovirus that infects over 20 million people worldwide. Only a small proportion of infected people develop either adult T cell leukemia/lymphoma (ATL) [3] or HTLV-1-associated myelopathy/tropical spastic paraparesis (HAM/TSP) [4,5]. HAM/TSP is a chronic, progressive neurological disease characterized by perivascular inflammatory infiltrates in the brain and spinal cord [6]. High frequencies of effector T cells have been demonstrated in peripheral blood with even higher frequencies in CSF of patients with HAM/TSP [7–9]. As definitive laboratory diagnosis of HAM/TSP is based on the presence of anti-HTLV-1 antibodies in the blood and CSF, robust humoral immune responses against HTLV-1 antigens have been reported [5,10,11]. Thus, chronically activated immune responses and infiltration of inflammatory cells into the CNS have been suggested to underlie the pathogenesis of HAM/TSP. Intrathecal antibody synthesis against HTLV-1 has been also reported, as evidenced by the presence of HTLV-1-specific antibodies and oligoclonal IgG bands (OCB) in CSF of HAM/TSP patients [12–15]. Intrathecal antibody response to HTLV-1 inversely correlates with higher HTLV-1 proviral loads (PVL) and a worse prognostic outcome [16]. Moreover, antibodies against two HTLV-1 viral products, Tax and Gag p24, have been reported to cross-react with host antigens, heterogeneous ribonucleoprotein A1 (hnRNP A1) and peroxiredoxin-1 (PrX-1), respectively, suggesting that molecular mimicry may play a role in the pathogenesis of HAM/TSP [17,18]. Since little is known about the role of B cells in the CNS of HAM/TSP patients, it is of interest to characterize and compare local B cell immune responses associated with the inflammatory milieu in the other chronic virus infection or neuroinflammatory diseases, such as multiple sclerosis (MS) which has clinical features that resemble HAM/TSP [19]. MS is a chronic, neurodegenerative inflammatory disease of the CNS, which leads to demyelination and progressive neurological disability. Based on the disease course, there are three main forms of MS. The more common course, relapsing-remitting MS (RRMS) is characterized by clinical episodes interspersed by periods of stability, affects twice as many women than men and in 40% of patients later develops a secondary progressive MS (SPMS) within ten years. Approximately 10% of patients experience a primary progressive MS (PPMS), which is characterized by gradual neurological dysfunction with or without exacerbations [20]. Although the etiology of MS is still unknown, viruses, such as Epstein-Bar virus (EBV) and Human herpes virus type 6 (HHV-6), are considered to be leading candidates associated with the pathogenesis of MS [21]. A hallmark of MS is the detection of OCB in the CSF that are associated with long term B cell survival in this compartment. Interestingly, recent studies have also demonstrated that in part, these CSF OCB in MS are specific for infectious pathogens and host antigens [22]. Currently, B cell depletion, with medications such as rituximab and ocrelizumab, is a promising MS treatment strategy [23]. However, such therapies have been also linked to progressive multifocal leukoencephalopathy (PML) which is a rare, often fatal, demyelinating disease caused by reactivation of the ubiquitous JC virus [24]. These observations suggest that CSF B cells associated with antibody synthesis may function to control viral replication in the CNS. It is therefore important to define the CNS microenvironment involved in recruitment and retention of B cell and ASCs in patients with virus-associated neuroinflammatory diseases. Antigen-specific antibodies in CSF are either derived from the blood (leakage through the blood-brain-barrier) or are synthesized locally within the CNS. In patients with MS, there is a persistence of clonally expanded B cells and non-dividing plasma cells in the CSF, as well as increase of chemokines and cytokines involved in B cell migration, differentiation, and long-term survival in the CNS [25–27]. These results suggest that the presence of ASCs in the CNS and the associated environment are critical aspects of the immune response. Interaction between CXCL13 (B cell-attracting chemokine-1) and its receptor CXCR5 is responsible for the migration of B cells and a subset of T cells in the follicular areas of lymphoid tissues called follicular helper CD4+ T cells (Tfh cells) [28–31]. Tfh cells are generally characterized by their expression of the chemokine receptor CXCR5, the transcription factor BCL6, and the inhibitory molecule PD-1, expression of high levels of IL-21, and their promotion of B cell help [32]. Recently, it has been shown that after specific activation, human blood CXCR5+CD4+ T cells might correspond to a circulating pool of memory Tfh cells [33]. This pool is distinct from the Th1, Th2, and Th17 subsets and can prompt naive and memory B cells to differentiate into ASCs, mainly through IL-21- and ICOS-induced signals [33,34]. Although it remains unclear how circulating memory Tfh cells relate to tissue Tfh cells, recent studies suggested that a subset of blood-circulating memory CXCR5+CD4+ T cells that are characterized by stable and moderate expression of the Tfh cell marker PD-1 most resemble tissue Tfh cells among resting memory CD4+ T cells in terms of B cell help functionality and transcriptional signature [33,35]. It has been reported that IL-21 mRNA was elevated in peripheral blood CD4+ T cells of MS patients and IL-21 expressing CD4+ T cells were detected in MS lesions [36,37]. In addition, increased circulating memory Tfh cells and plasma IL-21 level as well as CSF IL-21 level have been reported to be significantly elevated in MS patients than in controls with non-inflammatory neuronal diseases [38]. These studies suggested that memory Tfh cells may be involved in B cell help through IL-21. IL-21 is a member of the common γ chain family of cytokine that also includes IL-2, IL-7, IL-9 and IL-15, and promotes B-cell growth, differentiation, and class-switching [39]. Viral genes (such as HTLV-1 Tax) have been shown to transactivate some common γ chain family of cytokines including IL-21 and its receptor (IL-21R) in human T cells [40–42], and therefore it is of interest to understand the molecular cues of T cell/B cell interaction in the CNS microenvironment in patients with viral mediated neuroinflammatory disease. In this study, we analyzed B cell and T cell immunophenotypes in CSF of subjects with the chronic virus infection and/or neuroinflammatory diseases including HAM/TSP patients, HTLV-1-infected asymptomatic carriers (ACs), HIV-infected subjects treated with antiretroviral drugs, MS patients (RRMS and PPMS) and PML patients, compared to healthy normal donors (NDs). Comparison of CSF B cell subsets revealed that ASCs are increased in the CSF of all or a subset of these patients suggesting that B cell-mediated immune activation might be a critical aspect of the regulation and/or the pathogenesis of neuroinflammatory diseases associated with (or suspected of being associated with) viruses. Moreover, we demonstrate that increased ASCs are correlated with CD4+CD25+ T cells in the CSF of HAM/TSP patients whereas it is correlated with memory Tfh cells in MS patients. These results highlight the importance of the B cell compartment and the associated inflammatory milieu where production of antigen-specific antibody may be required to control viral persistence and/or may be associated with disease development in neuroinflammatory diseases. In situ histopathological studies in spinal cord of HAM/TSP patients demonstrated that T cells including both CD4+ and CD8+ T cells were detected depending on the duration of illness whereas B cells were only rarely observed [6]. However, since elevated intrathecal antibody synthesis have been demonstrated in HAM/TSP patients, we hypothesized that B cell recruitment and/or differentiation may also be present in CSF of patients. To confirm the presence of B cells in the CSF of subjects with chronic virus infection and/or neuroinflammatory diseases, we examined a large collection of CSF lymphocytes obtained from HAM/TSP patients, ACs and the other chronic virus infection and/or neuroinflammatory diseases including HIV-infected subjects, MS patients and PML patients (Table 1). We also had the unique opportunity to collect CSF from eighteen NDs as controls (Table 1). T cells were the predominant population in CSF lymphocytes of NDs (about 60 to 80% of lymphocytes) and CD4+ T cells are more prevalent than CD8+ T cells in our study (the average of CD4/CD8 ratio; 3. 9). Low levels of B cells were detected in lymphocytes from ND CSF, but were significantly elevated in patients with RRMS (Fig 1A, left graph). B cell/monocyte ratio which has been previously shown to be an indicator of rapid progression in MS patients [43] was higher in CSF of HAM/TSP patients as well as RRMS patients compared to NDs (Fig 1A, right graph). Importantly, B cell frequency and B cell/monocyte ratio were low in the CSF of ACs, which was comparable to those in NDs (Fig 1A). Flow cytometric analysis was able to differentiate B cell subpopulation into five subsets, including naïve (IgD+CD27-), unswitched memory (IgD+CD27+), double negative (IgD-CD27-), switched memory (IgD-CD27+) B cells and ASCs (IgD-CD27++) in CSF of a ND and a HAM/TSP patient (Fig 1B). Representative dot plots demonstrate that switched memory B cells were predominantly detected in the CSF of both subjects (Fig 1B). Interestingly, ASCs were elevated in the CSF of HAM/TSP patient compared to a ND (red rectangles in Fig 1B). ASCs were not detected in the CSF of NDs, but the presence of elevated ASCs in CSF was also higher in patients with RRMS, HAM/TSP and PML (Table 2). Group analysis of B cell subset also revealed that high frequency and absolute number of ASCs was detected in the CSF of HAM/TSP patients as well as RRMS patients (Fig 1C). ASCs were not or rarely detected in the CSF of subjects without neurologic diseases including ACs and HIV-infected subjects, although B cell phenotyping was able to be analyzed in only two ACs due to limited CSF cell number (Table 2). It is also of interest that there was undetectable or a low frequency of ASCs in the CSF of PPMS patients although only a small number of subjects was analyzed (Fig 1C). These results demonstrated that B cell recruitment and/or differentiation may be present in CSF of subjects with chronic virus infection and/or neuroinflammatory diseases. Increased ASCs may be involved in intrathecal antibody synthesis in CSF of subjects with chronic virus infection and/or neuroinflammatory diseases. Since HAM/TSP patients had higher antibody responses for HTLV-1 virus proteins, Gag, Env and Tax, in serum [11], we next analyzed antibody responses for HTLV-1 antigens in CSF and serum of HAM/TSP patients and ACs to examine whether HTLV-1-specific antibody synthesis is associated with ASC in CSF of HAM/TSP patients. Robust antibody responses for Gag, Env and Tax were observed in CSF of HAM/TSP patients compared to ACs (Fig 2A). Immunoreactivities against Gag and Tax were detected in the CSF of all HAM/TSP patients and immunoreactivity against Env was detected in 93. 2% of the CSF. In the CSF, the mean anti-Gag level was elevated comparable to that in the serum, but other anti-HTLV-1 antibodies, anti-Env and anti-Tax antibody level, were significantly lower in the CSF than in serum (p<0. 0001 and p = 0. 0027, respectively; Fig 2A). When the data were analyzed as an index of CSF immunoreactivity to serum immunoreactivity against each HTLV-1 antigen, the anti-Gag antibody index was higher than anti-Tax antibody index and significantly more elevated than anti-Env antibody index in HAM/TSP patients (Fig 2B). In contrast, ACs showed low levels of CSF/serum antibody indexes to all three antigens (Fig 2B). These results demonstrate that intrathecal anti-Gag and anti-Tax antibody synthesis are significantly elevated in HAM/TSP patients compared to those in ACs. Moreover, increased ASCs in CSF of HAM/TSP patients were significantly correlated with anti-Gag antibody index, but not with anti-Env or anti-Tax antibody indexes (Fig 2C). These results suggested that ASCs were involved in the intrathecal antibody synthesis, especially anti-HTLV-1 Gag, in the CSF of HAM/TSP patients. In the CSF of MS patients, it is well documented that clonally expanded B cells and non-dividing plasma cells persist with increased levels of chemokines and cytokines associated with B cell migration and differentiation [22], suggesting that CXCR5+ Tfh cells can also migrate and interact with B cells in the CNS. As memory Tfh cells have been reported to promote B cell growth, differentiation and class switching and to resemble tissue Tfh cells [32,33,35], we examined whether CXCR5-expressing memory Tfh cells are present in the CSF and peripheral blood of HAM/TSP patients compared to NDs, ACs and patients with the other chronic virus infection and/or neuroinflammatory diseases. Fig 3A shows representative results of memory Tfh cells (CXCR5+ CD45RA-; red rectangles) in CD4+ T cells of peripheral blood and CSF of a ND and a HAM/TSP patient. As Tfh cells are also characterized by the expression of the inhibitory molecule PD-1 [32], high frequency of PD-1 was detected in CSF, much higher than in peripheral blood, in memory Tfh cells of both ND and HAM/TSP patient (Fig 3A). Memory Tfh cells were detected in both CSF and peripheral blood of all the subjects (Fig 3B and 3C). Comparison of memory Tfh cells in CD4+ T cells demonstrated that the frequency of memory Tfh cell subset was slightly decreased in CSF of HAM/TSP patients compared to those in NDs while RRMS patients showed an increase of memory Tfh cells in the CSF (Fig 3B). In subjects with HIV, PPMS and PML patients, the frequency of memory Tfh cells varied in the CSF (Fig 3B). Intriguingly, memory Tfh cells were also slightly decreased in the CSF of ACs compared to that in NDs (Fig 3B). In CD4+ T cells of peripheral blood, there was no significant differences of memory Tfh cell frequencies in all the groups (Fig 3C). Dynamic changes of memory Tfh cells were detected in CSF of each group despite the increased levels of ASCs detected in the CSF of each of these cohorts of virus-associated neurologic diseases, suggesting that CSF B cells might be regulated by a different mechanism in each chronic virus infection and/or neuroinflammatory disorder. In HAM/TSP patients, CD4+CD25+ T cells are the predominant reservoir for HTLV-1 and induce various cytokines including IFN-γ [44]. Higher HTLV-1 PVL was detected in CSF compared to PBMC in HAM/TSP patients [45,46], suggesting that HTLV-1-infected CD4+ T cells can be recruited into the CNS and may alter the inflammatory milieu in the CNS of HTLV-1-infected subjects. We next examined whether CD4+CD25+ T cells are present in the CSF and peripheral blood of HAM/TSP patients compared to that of NDs, ACs and subjects with chronic virus infection and/or neuroinflammatory diseases. The frequency of CD4+CD25+ T cells were significantly higher in both CSF and peripheral blood of HAM/TSP patients compared to NDs and subjects with HIV and MS (Fig 4A and 4B, respectively). CD4+CD25+ T cells were also significantly higher in the CSF of patients with PML compared to NDs, but not in the peripheral blood (Fig 4A and 4B). In addition, HAM/TSP patients have been demonstrated that the forkhead box P3 (FoxP3), which is critical for the function of regulatory T cells, was decreased in CD4+CD25+ T cells and regulatory function of the CD4+CD25+ T cells was also inhibited [47]. Based on these observations, we examined the expressions of FoxP3 and CTLA-4 in peripheral blood CD4+CD25+ T cells of HAM/TSP patients compared to NDs, ACs and patients with the other chronic virus infection and/or neuroinflammatory diseases. As previously reported [47], HAM/TSP patients showed decreased expressions of FoxP3 and CTLA-4 in CD4+CD25+ T cells of the peripheral blood compared to NDs and subjects with MS and PML (Fig 4C and 4D, respectively). These results strongly supported the previous finding that CD4+CD25+ T cells would be functionally dysregulated in HAM/TSP patients and also suggested that CD4+CD25+ T cells of HAM/TSP patients might be functionally different from those of subjects with the other neuroinflammatory diseases. Our results demonstrated that CD4+CD25+ T cells were highly elevated in the CSF of HAM/TSP patients while memory Tfh cells were decreased. Moreover, comparison of CD4+ T cell subsets in neuroinflammatory diseases demonstrated that balances of memory Tfh cells and CD4+CD25+ T cells were different in each group although there were increases of B cells and ASCs in CSF across all groups. Given the characteristic features of CD4+ T cell subsets in CSF of HAM/TSP patients, we asked whether these CD4+ T cell subsets contribute to B cell regulation. To clarify the involvement of CD4+ T cells with B cell help in the CSF, the correlation of ASCs with memory Tfh cells and CD4+CD25+ T cells was analyzed in each group of HAM/TSP, RRMS, HIV and PML. While the frequency of memory Tfh cells was were significantly correlated with that of ASCs in the CSF of RRMS patients, surprisingly, there was no correlation between memory Tfh cells and ASCs in the CSF of subjects with HAM/TSP, HIV and PML (Fig 5A). By contrast, HAM/TSP patients showed significant correlation of CD4+CD25+ T cells with ASCs in the CSF (Fig 5B). Although PML patients had an increase of CD4+CD25+ T cells in the CSF compared to NDs (Fig 4A), PML patients did not show any correlation of CD4+CD25+ T cells and ASCs in the CSF (Fig 5B). These results strongly demonstrate that different subsets of CD4+ T cells, CD4+CD25+ T cells and memory Tfh cells, are involved in the increase of ASCs in CSF of HAM/TSP and RRMS patients, respectively. Moreover, we compared CD4+CD25+ T cells and ASCs in the CSF with disease activity of HAM/TSP patients. When HAM/TSP patients were subdivided into two types, non-progressive and progressive types, characterized by disease course, progressive type of HAM/TSP patients showed significantly higher frequency of CD4+CD25+ T cells in the CSF compared to non-progressors (Fig 5C). However, both frequency and absolute number of ASCs in CSF B cells did not show any significant differences between these types of HAM/TSP patients (Fig 5C). In addition, there was no significant correlation of ASCs in CSF B cells with disease duration in HAM/TSP patients (Fig 5D). These results suggested that ASCs would not be directly involved in the disease activity of HAM/TSP patients but present for long periods in the CSF of HAM/TSP patients. Of the groups of patients with neuroinflammatory diseases assessed, only HAM/TSP patients showed an involvement of CD4+CD25+ T cells with an increase of ASCs in the CSF. To determine the role of HTLV-1 infection within these CD4+ T cell subsets in the CSF of HAM/TSP patients, we examined HTLV-1 PVL in the CSF lymphocytes of HTLV-1-infected subjects. HTLV-1 was detected in the CSF lymphocytes of all HTLV-1 subjects, and HTLV-1 PVL was significantly higher in HAM/TSP patients compared to ACs (Fig 6A). Further analysis revealed that CSF PVL significantly correlated with CD4+CD25+ T cells in the CSF of HTLV-1-infected subjects, but negatively correlated with memory Tfh cells in the CSF (Fig 6B). These results demonstrated that an increase of HTLV-1-infected cells is associated with an increase of CD4+CD25+ T cells and a decrease of memory Tfh cells in the CSF of HTLV-1 infected subjects, suggesting that CD4+CD25+ T cells with HTLV-1 infection could be involved with B cell help in HAM/TSP. Tfh cells have been reported to express high levels of IL-21 which is a member of the common γ chain family of cytokines and promotes B-cell growth, differentiation, and class-switching [32,39]. Since HTLV-1 Tax has been reported to transactivate IL-21 in human T cells [41], it is of interest whether or not IL-21 is involved with B cell help in HAM/TSP patients. We examined IL-21 in CSF of HAM/TSP patients and ACs. While IL-21 was undetectable in the CSF of ACs, high levels of IL-21 were detected in the CSF of HAM/TSP patients (Fig 6C). This suggests that the increase of IL-21 might be related to the increase of CD4+CD25+ T cells in the CSF of HAM/TSP patients as memory Tfh cells were decreased. To confirm IL-21 production in CD4+CD25+ T cells of HAM/TSP patients, we examined IL-21 expression in combination with HTLV-1 Tax expression in PBMCs of NDs and HAM/TSP patients. Fig 6D shows representative dot plots of HTLV-1 Tax and IL-21 expression in CD4+CD25+ T cells of a ND and a HAM/TSP patient after PBMC culture for 24 hours without any exogenous stimulation. IL-21 was detected in Tax-expressing CD4+CD25+ T cells of a HAM/TSP patient whereas ND did not show any IL-21 and Tax expression after culture (Fig 6D). Group analysis of five HAM/TSP patients demonstrated that Tax-expressing CD4+CD25+ T cells expressed IL-21 (Fig 6E). In addition, when B cells isolated from PBMC of NDs and HAM/TSP patients and stimulated with rhIL-21 for 7days, the frequency of ASCs, as well as IgG concentration, was significantly increased in rhIL-21-stimulated B cell cultures from both NDs and HAM/TSP patients (Fig 6F and 6G). Moreover, HTLV-1 Gag- and Tax-specific antibodies were also detected in the rhIL-21-stimulated B cell culture supernatants from HAM/TSP patients (Fig 6H). These results suggest that the increased IL-21 expression might be partly derived by HTLV-1 Tax expression in CD4+CD25+ T cells, and that it may serve to accelerate B cell function in HAM/TSP. Regulation of the local immune response is crucial in protecting the CNS from viral infection and immunopathologically mediated tissue damage. Although robust humoral immune responses, including OCBs specific to viral antigens, have been reported in the CSF of patients with virus-associated neuroinflammatory diseases, little is known about the CNS microenvironment related to this increased humoral immune response in disease and healthy controls. Multicolor flow cytometric analysis demonstrates that B cell/monocyte ratio and ASCs were increased in the CSF of HAM/TSP patients as well as RRMS patients, while ASCs were undetectable in CSF of NDs. Consistent with previous study [48], B cells were rarely detected in ND CSF, and when present, were in very low numbers in our study. Our results also strongly support recent studies in MS that in the B cell lineage, memory B cells and plasma blasts are predominantly detected in the CSF of MS patients [49]. Memory B cells can proliferate and rapidly differentiate into ASCs at much lower thresholds for activation than naïve B cells. Ig production is maintained by ASCs including proliferating “plasmablasts” and “plasma cells” that can be short or long-lived [50]. Therefore, it has been suggested that increased ASCs may be a marker for immune activation against a target, such as virus or host antigen, which is involved in disease development in neuroinflammatory disorders. In situ histopathological studies in spinal cord of HAM/TSP patients demonstrated that T cells including both CD4+ and CD8+ T cells were detected depending on the duration of illness whereas B cells were only rarely observed [6]. We here demonstrate an accumulation of B cells and elevated ASCs that may contribute to antibody production in the CSF of HAM/TSP patients. If CSF reflects the intrathecal inflammatory process in the CNS, increases of B cells and ASCs in the CSF might be an indicator of disease development in HAM/TSP. Importantly, ASCs were not detected in the CSF of ACs and a majority of HIV-infected subjects, although B cell phenotyping was able to be analyzed in only a small number of subjects with ACs. These results suggested that accumulation and differentiation of B cells might be well regulated in subjects with chronic virus infection but no neurologic diseases. Moreover, it is of interest that ASCs were also not detected in a subset of PPMS patients, while RRMS patients showed high frequency of ASCs in the CSF B cells. However, B cell is also important immune regulator for disease development of PPMS, since B cell depleting monoclonal antibody, Ocrelizumab, is the first drug ever to show efficacy in slowing the disease progression in a phase 3 clinical trial with PPMS patients [51]. Comparison of larger studies from MS patients would be required for confirmation of B cell phenotyping in CSF. Antibody responses against HTLV-1 antigens including Gag and Tax were detected in CSF of all HAM/TSP patients tested. Interestingly, CSF/serum antibody index of anti-Gag and anti-Tax were significantly elevated in HAM/TSP patients while ACs had a significantly lower CSF/serum antibody index of all HTLV-1 antigens; this suggests that antibody responses for HTLV-1 Gag and Tax might be generated in HAM/TSP patients according to increases of viral expression or immune activation while antibody responses against these HTLV-1 antigens might be well controlled in the CSF of ACs. Moreover, increased ASCs in CSF of HAM/TSP patients significantly correlated with CSF HTLV-1 Gag-specific antibody production which was not observed in ACs. It is of interest that HAM/TSP patients have also been reported to develop autoantibodies to neurons that cross-reacted with HTLV-1 Gag and Tax [17,18], suggesting that increased humoral immune responses including cross- or self-reactive antibodies to CNS antigens for HTLV-1 might alter the risk of CNS inflammation or autoimmune disease. Since HTLV-1-specific antibody responses and ASCs were stably detected in the CSF of HAM/TSP patients, persistent localization of ASCs may be associated with long-term stability of Ig production and OCB detection in the CSF of HAM/TSP patients. Intriguingly, antibody response for HTLV-1 Env was lower in the CSF of HAM/TSP patients compared to antibody responses for HTLV-1 Gag and Tax. Since neutralizing and antibody-dependent cellular cytotoxicity-inducing activity of antibodies against HTLV-1 Env gp46 have been reported to prevent viral infection in vitro [52], less robust antibody response for HTLV-1 Env may fail to control viral infection in the CNS of HAM/TSP patients. Therefore, it is important to identify target antigens of immune cells and antibodies for understanding disease development and therapeutic approach in chronic viral infection and neuroinflammatory diseases. To maintain memory B cells and generate ASCs, a subset of CD4+ T cells called Tfh cells is required. Recently, it has been addressed that CXCR5+CD4+ T cells are detected in organs that are affected by autoimmune disorders, such as systemic lupus erythematosus and Sjogren’s syndrome, suggesting that aberrant Tfh cells may induce autoimmunity [53]. Recent reports in chronic HIV infection demonstrated that Tfh cells are expanded, but impair B cell help and harbor high amounts of viral DNA [53–55]. It has been demonstrated that precise control of Tfh cell number is important to produce optimally affinity-matured antibody responses that are devoid of self-reactivity [56]. In the current study, we demonstrate that subjects with chronic virus infection and/or neuroinflammatory diseases lost the balance of memory Tfh cell frequencies in CSF compared to CSF of NDs in which a certain frequency of memory Tfh cells were stably maintained. HAM/TSP patients showed a decrease of memory Tfh cells in the CSF. Importantly, memory Tfh cells were decreased in the CSF of ACs which did not show any intrathecal antibody synthesis for HTLV-1 or accumulation of CD4+CD25+ T cells in the CSF. Therefore, memory Tfh cell might be inhibited in the CSF of subjects with chronic HTLV-1 infection to prevent excess B cell responses but excessive accumulation and/or activation of CD4+ T cells might promote B cell development in HAM/TSP patients. Tfh cell responses have been reported to be regulated by various mechanisms, such as follicular regulatory CD4+ T cells in a Bcl6-dependent manner and CD8+ regulatory T cells in IL-15 dependent manner [56]. Further studies will be required to understand the regulation of memory Tfh cells associated with chronic virus infection in the CNS. Tfh cells have also been reported to express high levels of IL-21 and promote B cell growth, differentiation and class switching [32]. It has been reported that IL-21 mRNA was elevated in peripheral blood CD4+ T cells of MS patients and IL-21 expressing CD4+ T cells were detected in MS lesions [36,37], suggesting that a CD4+ T cell subset, memory Tfh cells, might be involved in B cell help through IL-21 in MS patients. While MS patients showed a significant correlation of ASCs with memory Tfh cells in the CSF, there was no association between ASCs and memory Tfh cells in the CSF of HAM/TSP patients. Interestingly, CD4+CD25+ T cells were significantly correlated with the frequency of ASCs as well as HTLV-1 PVL in the CSF of HAM/TSP patients, suggesting that chronic viral activation could induce continuous differentiation of memory B cells into ASCs and Ig production. Moreover, HAM/TSP patients showed an increase of IL-21 level in the CSF even though there was the decreased memory Tfh cells. Since ACs did not show any accumulation of CD4+CD25+ T cells and increased IL-21 in the CSF, this suggests that the increased IL-21 in HAM/TSP patients might be derived from CD4+CD25+ T cells in the CNS. IL-21 is a member of the common γ chain family of cytokine that also includes IL-2, IL-7, IL-9 and IL-15, and promotes B-cell growth, differentiation, and class-switching [39,57]. After culture for 24 hours without any stimulation, IL-21 expression was detected in Tax-expressing CD4+CD25+ T cells of HAM/TSP patients. HTLV-1 Tax has been shown in vitro to induce the expression of IL-2 and IL-15 [40,42]. Increased expression of these cytokines has been shown to dysregulate T-cell activation and proliferation that may contribute to CNS inflammation in HAM/TSP patients [58]. Although Tax has been reported to trans-activate IL-21 and its receptor (IL-21R) genes in human T cells [41], little is known about involvement of IL-21 in HAM/TSP patients. Since IL-2 and IL-15 have been also shown to be associated with B cell function, such as proliferation and Ig secretion [59], increased expression of these cytokines might accelerate B cell function in HAM/TSP patients. Alternatively, Tfh cell independent induction of B cell function might cause impaired B cell responses and generation of antigen-specific antibodies with low specificity and function. Therefore, adequate and appropriate Tfh cells for B cell help would be required for control of viral infection in the CNS. Larger systematic studies of virus-associated neurologic diseases including the functional differences of CD4+ T cell subsets will further improve our knowledge of the B cell/T cell immune regulation in the CNS associated with chronic viral infections. Lastly, comparison of CSF immune phenotyping highlights that B cell/T cell interactions may be involved in the development and maturation of B cells in the CNS of neuroinflammatory diseases. Although ASCs were detected in high frequencies of patients with MS, HAM/TSP and PML, balances of CD4+ T cell subsets, memory Tfh cells and CD4+CD25+ T cells, were different in each group. Therefore, characterization of CSF immune responses that are associated with a neuroinflammatory milieu may provide evidence for a pathogenic “signature” of an immunopathogenic process in virus-associated neurologic diseases. A total of 71 HAM/TSP patients and 12 HTLV-1-positive ACs were evaluated in this report according to established criteria [60]. To characterize patient’s disease onset based on clinical and motor outcomes, we used the Osame Motor Disability Score. Onset of disease was defined as rapidly progressive if the OMDS score increased by >3 grades since clinical onset of HAM/TSP [61]. HAM/TSP patients were further categorized as either progressors or non-progressors based on clinical status at the time of the lumbar puncture analysis if their symptoms were changing or stable, respectively. Subsets of HAM/TSP patients were used for specific studies. For detection of HTLV-1-specific antibodies, serum and CSF samples were obtained from a total of fifty subjects, including HAM/TSP patients (n = 44) and HTLV-1-positive ACs (n = 4). For flow cytometric analysis, whole blood was obtained from a total 170 subjects, including ACs, patients with HAM/TSP, MS including RRMS and PPMS [62], PML [63], HIV-infected subjects adequately treated with antiretroviral drugs and without neurological disease and NDs (Table 1). Of 170 subjects, CSF samples were obtained from a total 149 subjects (Table 1). PBMCs were isolated by Ficoll-Hypaque (Lonza) centrifugation, and were cryopreserved in liquid nitrogen until use. CSF samples were obtained by lumber puncture and the cells were collected by centrifugation of CSF samples. The study was reviewed and approved by the National Institute of Neurological Disorders and Stroke Institutional Review Board. All samples used in this study were collected from the subject followed at the National Institute of Neurologic Disorders and Stroke under protocols # 98-N-0047,89-N-0045,13-N-0017,13-N-0149. Prior to study inclusion, written informed consent was obtained from the subject in accordance with the Declaration of Helsinki. The LIPS assay was performed as previously described [11]. Each mammalian expression vector with the HTLV-1 gene (HTLV-1 Gag, Env and Tax/pRen2) was transfected into 293T/17 cell line (ATCC) using X-tremeGENE 9 DNA transfection reagent (Roche Diagnostics) [11]. Serum, CSF samples or B cell culture supernatants were diluted to 1: 100. Each HTLV-1-specific antibody index was calculated as ratio of CSF immunoreactivity (LU) /serum immunoreactivity (LU). For analysis of peripheral blood lymphocyte and CSF lymphocyte populations, EDTA-treated whole blood or CSF cells were stained with CD3, CD4, CD8, CD14, CD19, CD25, CD27, CD45, CD45RA, CXCR5, IgD (all from BD Biosciences) and PD-1 (BioLegend). Since B cells (CD45+CD3-CD19+) were rarely detected in CSF, CSF samples from a total 93 subjects were used for B cell subset analysis (Table 2). For staining of FoxP3 and CTLA-4, EDTA-treated whole blood were stained with antibodies for surface markers. After fixed and permeabilized with Fixation/Permeabilization buffer (eBiosciences), the cells were stained with antibodies for FoxP3 (eBiosciences) and CD152 (CTLA-4; BD Biosciences). For detection of IL-21 production, PBMCs of NDs or HAM/TSP patients were suspended in RPMI media (RPMI1640 supplemented with 10% heat-inactivated fetal bovine serum, 100U/ml of penicillin, 100μg/ml of streptomycin sulfate and 2mM L-glutamine) for 24 hours and incubated with GoldiPlug (BD Biosciences) for the last 5 hours in 5% CO2 incubator at 37°C. After the culture without any stimulator, PBMCs were surface-stained with specific antibodies. After fixation and permeabilization with Fixation/Permeabilization solution (BD Biosciences), the cells were intracellularly stained with anti-human IL-21 (BD Biosciences) and anti-Tax (Lt-4) antibodies. All flow cytometric analysis was performed using a LSR II (BD Biosciences). The data were analyzed using FlowJo 10. 2 software (FlowJo LLC). HTLV-1 PVL was measured using droplet digital PCR (Bio-Rad) as previously described [64]. DNA was extracted from the PBMC and CSF cell pellets using a DNeasy Blood and Tissue kit (Qiagen) according to the manufacturer’s instructions. Primers and probes specific for HTLV-1 tax and human ribonuclease P protein subunit 30 (RPP30) was used. The duplex PCR amplification was performed in this sealed 96-well plate using a GeneAmp 9700 thermocycler (Applied Biosystems). Following PCR amplification, the 96-well plate was transferred to a QX100 droplet reader (Bio-Rad). For PVL calculation, QuantaSoft software version 1. 3. 2. 0 (Bio-Rad) was used to quantify the copies/μl of each queried target per well. All samples were tested in duplicate, unless otherwise specified, and PVL is reported as the average of the two measurements. IL-21 were detected in serum and CSF samples of HAM/TSP patients and ACs using Human Legend Max Human IL-21 ELISA kit (BioLegend) according to the manufacturer’s instructions. B cells were isolated from PBMCs of NDs and HAM/TSP patients using B cell isolation kit II (Miltenyi Biotec). The isolated B cells were cultured at 2x104 cells/well in 96 U-bottom microplates in RPMI media with or without 10ng/ml of recombinant human IL-21 (Cell Signaling Technology). After the culture for 7 days, the cells were surface-stained with specific antibodies and analyzed using a LSR II (BD Biosciences). Human IgG and HTLV-1-specific antibodies were measured in the culture supernatants using Human IgG ELISA Quantitation Set (Bethyl Laboratories) and LIPS assay, respectively. The Mann-Whitney Test was used to compare: anti-HTLV-1 antibodies, PVL and IL-21 in CSF between ACs and HAM/TSP patients, CSF lymphocytes in HAM/STP patients by disease activity. Paired T Test was used to compare: anti-HTLV-1 antibody indexes in each HTLV-1-infected subject. The Kruskal-Wallis test with Dunn’s test for multiple testing was used to compare: frequency or ratio of CSF lymphocytes between the different patient groups. Fisher’s exact test was used to compare frequency of CSF ASCs detection in each patient’s group with NDs. Spearman’s rank correlation test was used to compare: the CSF/serum anti-HTLV-1 antibody index, HAM/TSP disease duration, CD4+ T cell subsets with ASCs in CSF, and between CD4+ T cell subsets and PVL in CSF. All statistical analysis was performed using Prism (GraphPad software).

Regulation of the local immune response is crucial in protecting the central nervous system (CNS) from viral infection and immunopathologically mediated tissue damage. Intrathecal antibody synthesis is a well-documented phenomenon in infectious and demyelinating neurological diseases, but little is known about the CNS microenvironment related to this increased humoral immune response in disease and healthy controls. Comparison of CSF immune phenotyping highlights that B cell/T cell interactions may be involved in the development and maturation of B cells in the CNS of virus-associated neuroinflammatory diseases. Characterization of CSF immune responses that are associated with a neuroinflammatory milieu may provide evidence for a pathogenic “signature” of an immunopathogenic process in virus-associated neurologic diseases.

Abstract Introduction Results Discussion Materials and methods

blood cells medicine and health sciences body fluids immune cells pathology and laboratory medicine immune physiology multiple sclerosis nervous system neurodegenerative diseases pathogens immunology microbiology retroviruses viruses demyelinating disorders clinical medicine rna viruses antibodies memory b cells immune system proteins white blood cells animal cells proteins medical microbiology htlv-1 t cells microbial pathogens antibody-producing cells biochemistry anatomy cell biology b cells central nervous system clinical immunology physiology viral pathogens autoimmune diseases neurology biology and life sciences cellular types cerebrospinal fluid organisms

2018

Immunophenotypic characterization of CSF B cells in virus-associated neuroinflammatory diseases

Meiosis is a complex type of cell division that involves homologous chromosome pairing, synapsis, recombination, and segregation. When any of these processes is altered, cellular checkpoints arrest meiosis progression and induce cell elimination. Meiotic impairment is particularly frequent in organisms bearing chromosomal translocations. When chromosomal translocations appear in heterozygosis, the chromosomes involved may not correctly complete synapsis, recombination, and/or segregation, thus promoting the activation of checkpoints that lead to the death of the meiocytes. In mammals and other organisms, the unsynapsed chromosomal regions are subject to a process called meiotic silencing of unsynapsed chromatin (MSUC). Different degrees of asynapsis could contribute to disturb the normal loading of MSUC proteins, interfering with autosome and sex chromosome gene expression and triggering a massive pachytene cell death. We report that in mice that are heterozygous for eight multiple simple Robertsonian translocations, most pachytene spermatocytes bear trivalents with unsynapsed regions that incorporate, in a stage-dependent manner, proteins involved in MSUC (e. g. , γH2AX, ATR, ubiquitinated-H2A, SUMO-1, and XMR). These spermatocytes have a correct MSUC response and are not eliminated during pachytene and most of them proceed into diplotene. However, we found a high incidence of apoptotic spermatocytes at the metaphase stage. These results suggest that in Robertsonian heterozygous mice synapsis defects on most pachytene cells do not trigger a prophase-I checkpoint. Instead, meiotic impairment seems to mainly rely on the action of a checkpoint acting at the metaphase stage. We propose that a low stringency of the pachytene checkpoint could help to increase the chances that spermatocytes with synaptic defects will complete meiotic divisions and differentiate into viable gametes. This scenario, despite a reduction of fertility, allows the spreading of Robertsonian translocations, explaining the multitude of natural Robertsonian populations described in the mouse. A series of complex processes takes place during the first meiotic division, including pairing, synapsis, recombination and segregation of homologous chromosomes. Defects in any of these processes can affect the normal progression of meiosis, causing severe fertility reduction or even sterility [1]–[3]. This is a consequence of the existence of surveillance mechanisms that monitor the accurate progression of meiotic events and promote the removal of defective cells. Two main checkpoints have been proposed to act during the first meiotic division: the pachytene checkpoint, responsible for ensuring the correct occurrence of recombination and synapsis [2], [4], [5], and the metaphase-I or spindle checkpoint, which controls the precise segregation of homologous chromosomes [6], [7]. Although the process that eliminates meiocytes in metaphase-I and II might be similar to that acting during mitosis [6], [7], a clear understanding of the mechanisms that trigger the pachytene checkpoint is still lacking. Given the interdependence between meiotic recombination and synapsis, it has been difficult to ascertain the existence of separate checkpoints for these processes in mammals. Thus, many recombination-defective mutants exhibit a delay in synapsis and/or synaptic aberrations, and meiosis is aborted during the zygotene-pachytene transition [8]–[10]. Likewise, most mutants defective for synaptonemal complex (SC) components abort meiosis at pachytene with unresolved recombination processes [11]–[15]. In addition to the accumulation of unresolved recombination intermediates, unsynapsed chromosomal regions undergo a process of transcriptional inactivation called meiotic silencing of unsynapsed chromatin (MSUC) [1], [16]–[18]. The mechanisms involved in transcriptional inactivation are particularly well characterized in mammalian male meiosis, in which sex chromosomes undergo a special case of MSUC called meiotic inactivation of sex chromosomes (MSCI) [19], [20]. This process is initiated with the accumulation of BRCA1 protein on the unsynapsed axial elements (AEs). BRCA1 is a protein involved in DNA damage repair that allows the recruitment of other factors such as ATR, promoting the phosphorylation of H2AX at serine 139 on the surrounding chromatin [21], [22]. The inactivation of sex chromosomes, which affects the unsynapsed regions of both the X and Y chromosomes, comprises an additional plethora of chromatin modifications that includes: 1) histone modification [18], [23], [24]; 2) incorporation of specific histone variants [25], [26]; 3) specific incorporation of non-histone proteins [27]–[30]; and 4) accumulation of XIST RNA [31] and other families of non-coding RNAs [32]. The initiation of MSUC seems to also operate by the action of BRCA1 and ATR [17]. Furthermore, it has been reported that many chromatin modifications detected during MSCI are also involved in the inactivation of unsynapsed autosomes. This is the case of H2AX phosphorylation [17], histone H2A ubiquitination [18], methylation of histone H3 and H4, incorporation of histone H3. 3 [26] and Maelstrom protein [29]. However, the role of other chromatin modifications in MSUC remains to be demonstrated. On these grounds, it has been proposed that MSUC may interfere with the expression of genes necessary for the completion of meiosis and this would contribute to arrest the meiotic progression of pachytene spermatocytes with synapsis defects [17]. More recently, it has been suggested that extensive asynapsis and MSUC could also interfere with MSCI [33]. Indeed, activation of some sex chromosome-linked genes that should remain inactive during meiosis has been claimed as one of the causes of meiotic failure in some mouse models [1], [17], [34], [35]. Mahadevaiah and co-workers [33] have proposed that MSCI initiation could be impeded by the sequestration of MSUC triggering proteins like BRCA1 and ATR on extensively unsynapsed autosomes, a circumstance that would preclude these proteins to relocate to the unsynapsed AEs of the sex chromosomes. MSCI abrogation has thus been proposed as the primary cause of spermatocyte death in mouse models that typically arrest meiosis at the zygotene-pachytene transition, including many recombination-defective mutants [33]. Sequestration of BRCA1 has been proposed to occur also in female meiosis [36]. However, in both cases cells seem to tolerate a certain degree of asynapsis, since both spermatocytes and oocytes with a reduced number of asynapsed chromosomes are able to progress through first meiotic prophase without interfering with of MSUC or MSCI processes [33], [36]. In the house mouse (Mus musculus domesticus), individuals that are heterozygous for Robertsonian (Rb) translocations (the fusion of two acrocentric chromosomes) show reduced fertility. This reduction is strongly correlated with impairment of spermatogenesis and loss of meiotic cells [37]–[45]. Depending on the number and complexity of Rb heterozygosity (i. e. formation of trivalents, chains or rings), meiocytes may be eliminated during prophase-I [41], [46]–[48] or during metaphase-I and II [39], [47]–[50]. The synaptic behaviour of trivalents in Rb heterozygotes has been extensively analyzed by means of electron microscopy in a wide range of mammalian species, including mouse and humans [42], [44], [45], [47], [51]–[54]. During meiosis, heterozygous mice display a high frequency of pairing abnormalities including: 1) delay in synapsis completion of trivalents; 2) existence of a variety of heterelogous synaptic situations, both within and between trivalents and between trivalents and the sex chromosomes; and 3) persistence of unsynapsed regions in the trivalents throughout pachytene. Furthermore, a reduction of the recombination frequency and a decrease of chiasma interference in these hybrids have been demonstrated [40], [55]–[58]. However, little is known about the chromatin modifications associated with these synaptic disturbances. The aim of this study is to ascertain the extent of MSUC during meiosis in Rb heterozygous mice and to evaluate the consequences of this cellular response on the meiotic progression of spermatocytes. We used males generated by crossing individuals of a standard karyotype (2n = 40) with homozygous individuals bearing eight Rb translocations (2n = 24), collected from natural populations in Northern Italy. The resulting hybrids (2n = 32) bear eight trivalents that exhibit different degrees of asynapsis during meiosis. We have combined the analysis of synapsis and recombination progression during male meiosis with the localization of some proteins involved in MSUC, i. e. , γH2AX, ATR, ubiquitinated H2A, SUMO-1 and XMR, the latter two having only been reported to act in MSCI. Our results describe the kinetics of MSUC in Rb heterozygotes and highlight the capacity of spermatocytes with synaptic defects to pass through pachytene and progress to the metaphase stage. To characterize the progression of the first meiotic prophase, we used three main criteria: 1) the localization of SYCP3, the main component of the synaptonemal complex (SC) axial/lateral element (AE/LE), and that of RAD51, a protein related to early meiotic recombination and repair (Figure 1) that is abundantly incorporated along the chromosomes at zygotene, and then gradually disappears during pachytene and is absent at mid/late pachytene [59]; 2) the length of the pairing region between X and Y chromosomes, which extends up to 100% of the Y chromosome at early pachytene and becomes shorter as pachytene proceeds [60]; and 3) the reduction of the pairing region of sex chromosomes to the very distal end, the appearance of excrescences on the AEs of sex chromosomes, and the widening of SC attachment plates on the autosomes that identify the late pachytene stage. These criteria are comparable with those reported in recent studies carried out using RPA and MLH1 as markers of pachytene progression [61]. We found that synapsis was initiated at early zygotene in both bivalents and trivalents, but proceeded more quickly in the bivalents (Figure 1A). In fact, most trivalents were still undergoing synapsis when bivalents (b in Figure 1) were almost completely synapsed. This may be due to the fact that in trivalents, synapsis was initiated only at the distal ends of the chromosomes (Figure 1B–1B' ). At this meiotic stage, the X and Y chromosomes usually lay apart from each other. At early pachytene, all bivalents and some trivalents had completed synapsis (closed configuration), although in many trivalents the chromosomal regions close to the centromeres were still unsynapsed (open configuration) (Figure 1C–1E' and Table 1). The pattern of RAD51 localization at the early stages of prophase-I was similar to that exhibited by mice with the standard acrocentric karyotype. During zygotene, a large number of RAD51 foci appeared on both synapsed and unsynapsed AEs of bivalents and trivalents and on the X chromosome (Figure 1A–1B' ). Then, during early pachytene, the number of foci started to drop, although foci remained more abundant in both trivalents and sex chromosomes than in bivalents (Figure 1C–1E' ). RAD51 foci were associated with the trivalents in either the open or closed configuration and did not preferentially accumulate on the unsynapsed regions of the open trivalents (Figure 1C–1D' ). At mid-pachytene, many trivalents had completed synapsis and appeared in a closed configuration, but one to four trivalents remained in an open configuration (Figure 1F–1H' ). At this stage RAD51 was only present on the sex chromosomes and on some trivalents. At late pachytene, most cells exhibited up to four trivalents with an open configuration (Figure 1I–1K' ). At diplotene, when desynapsis starts and homologues initiate their separation, the proximal ends of the acrocentric chromosomes remained associated in some trivalents while they appeared clearly separated in others (Figure 1L–1M' ). RAD51 was not detectable at late pachytene (Figure 1I–1K' ) or diplotene (Figure 1L–1M' ). These results indicate that the repair of DNA might be delayed in some trivalents as it is in the sex chromosomes, but this process seems to culminate successfully in mid-late pachytene, when the signal of RAD51 disappeared, even though many unsynapsed chromosome regions are present. Trivalents commonly engage in ectopic heterologous associations with other trivalents and/or the sex chromosomes (Figure 1C). Thus, we wondered whether these associations would involve the assembly of the SC as a tripartite structure. For this purpose we analyzed the localization of SYCP1 protein, one of the main components of the SC transverse filaments and central element (Figure 2). At zygotene, we found that trivalents could establish an end-to-end connection that did not usually involve SYCP1 (Figure 2A–2A' ). However, the association of unsynapsed proximal ends of trivalents with the sex chromosomes frequently involved the formation of a short SC with either the distal region of the X chromosome, the proximal region or both (Figure 2B–2B' ; see also Figure 3 and Figure S1). Furthermore, the Y chromosome was sometimes found in a self-synapsed configuration (Figure 2B–2B' and Figure S1). These situations usually occurred at early pachytene and were more rarely detectable from mid-pachytene onwards. Heterologous synapsis was also found within each trivalent. Although it was expected that the two acrocentrics could synapse with the corresponding homologous segment of the metacentric (Figure 2C–2C' ), synapsis between the heterologous proximal chromosomal regions of the acrocentrics was the most frequent configuration. Heterologous synapsis could involve either a short (Figure 2D–2D' ) or a long segment of both chromosomes (Figure 2E-2E' ) and could be maintained from pachytene until late diplotene (Figure 2G–2G' ). Furthermore, we found that some unsynapsed chromosomal regions incorporated SYCP1 (Figure 2F–2F' ), perhaps representing either unsynapsed regions that were about to synapse or regions of self-synapsis. Alternatively, they may reveal only a non-specific binding of SYCP1 to unsynapsed AEs, a feature that is frequently observed in the sex chromosomes [62]. To evaluate the incorporation of MSUC markers on unsynapsed Rb trivalents, we first examined the temporal localization of γH2AX (Figure 3 and Figure S1; Video S1). This protein localizes in foci at DNA double-strand breaks during DNA repair and it is also associated with the inactivation of unsynapsed chromatin in autosomes and sex chromosomes [17], [63], [64]. At leptotene, the localization of γH2AX was dispersed throughout the nucleus (Figure 3A; Video S1); then, at zygotene, γH2AX began to disappear from the synapsed chromosome regions of the bivalents and from the synapsed distal regions of some trivalents (Figure 3B). The X chromosome appears intensely labeled while the Y chromosome is usually devoid of γH2AX labeling. When spermatocytes entered pachytene, γH2AX became restricted to the chromatin located close to the LEs of the synapsed regions of both bivalents and trivalents (see insets in Figure 3C and 3D) [65]. In the sex chromosomes, γH2AX was extended over the chromatin (Figure 3C and 3D). Interestingly, we observed that the Y chromosome is intensely labeled even when it occasionally appears self-synapsed (Figure S1). In the unsynapsed regions of the trivalents, γH2AX showed two labeling patterns: 1) occupying a wide chromatin area surrounding the unsynapsed segments, as in the sex chromosomes, and 2) occupying a more restricted chromatin area, very close to the unsynapsed AEs, as in the synapsed regions (see inset in Figure 3C). It is especially striking that in some trivalents one of the unsynapsed acrocentric chromosomes showed one of these labeling patterns while the other acrocentric showed the alternative pattern (see inset in Figure 3D). From mid to late pachytene, γH2AX labeling appeared as a bright signal on the entire chromatin surrounding the X and Y chromosomes and the AEs of the unsynapsed regions of open trivalents (Figure 3E). It is important to stress that open trivalents showed γH2AX labeling regardless of whether they were close to the X and Y chromosomes or far from them, indicating that this labeling was not a consequence of their association with the sex chromosomes (see Video S2). At diplotene the localization of γH2AX in the sex chromosomes remained visible, and it was also detectable in the pericentromeric regions of some trivalents (Figure 3F). These regions most likely represent chromosomal segments that have remained unsynapsed during pachytene, since those that began desynapsis during diplotene, in either bivalents or trivalents, were devoid of γH2AX labeling. These results suggest that MSUC is a mechanism triggered during the early stages of prophase-I. Furthermore, it indicates that most spermatocytes carrying unsynapsed trivalents would proceed normally into diplotene. Next, we investigated the presence of ATR in the unsynapsed regions of trivalents (Figure 4, Figure S2, and Figure S3). During zygotene, ATR labeling appeared as small foci located on the AEs/LEs in both synapsed and unsynapsed autosomes and in the X chromosome, but it was rarely observed in the Y chromosome (Figure 4A). At the zygotene/pachytene transition, ATR began to disappear from the chromosomes that had completed their synapsis (Figure 4B), although it remained as numerous and intense foci on the unsynapsed AEs. At this stage, a single ATR focus was always detected on the Y chromosome. During early pachytene, ATR labeling appeared as a continuous line along the unsynapsed AEs of trivalents and sex chromosomes (Figure 4C). ATR localization contrasted with that of γH2AX, the latter including the whole unsynapsed chromatin (Figure 3C). This result indicates absence of colocalization of the two proteins during late zygotene and early pachytene in the unsynapsed regions (Figure S2). On the other hand, we observed some unsynapsed trivalent regions without the ATR signal (Figure 4B–4D). This observation is consistent with the absence of γH2AX in some unsynapsed trivalent regions and strongly suggests the existence of two classes of unpaired chromosome segments during early pachytene: one class that shows neither γH2AX nor ATR protein, and another class that shows the presence of both proteins. At mid-pachytene, ATR labeling was still apparent along unsynapsed AEs and also appeared to extend to the surrounding chromatin of the unsynapsed regions of trivalents and of the sex chromosomes (Figure 4D); then, it became brighter as spermatocytes progressed to late pachytene (Figure 4E). During diplotene, ATR remained visible on the surrounding chromatin of the sex chromosomes and of the open trivalents (Figure 4F); its signal progressively faded and completely disappeared by the late diplotene stage. In view of these results, we next analyzed the pattern of appearance and localization of three other MSUC/MSCI-related proteins: 1) monoubiquitinated H2A histone (ubiH2A) (Figure 5), known to be associated with transcriptional silencing of unsynapsed autosomes and sex chromosomes in mouse male meiosis [18]; 2) SUMO-1 (Figure 6), which is involved in SC assembly as well as in the formation of the sex body [30], [66]–[68], and 3) XMR (Figure 7), a member of the XLR gene superfamily [69], known to localize in the sex body [27]. We found that during zygotene (Figure 5A) and early pachytene (Figure 5B), ubiH2A appeared as a faint signal on the chromosome ends on both synapsed and unsynapsed LEs, as previously observed [18]. On the contrary, at mid-pachytene, an intense labeling appeared on the chromatin of unsynapsed segments of trivalents and of the sex chromosomes (Figure 5C), persisting until mid diplotene (Figure 5D), when it started to disappear. SUMO-1 was not detected during zygotene (data not shown) and very early pachytene spermatocytes (Figure 6A). It appeared on the unsynapsed chromatin of sex chromosomes and trivalents during a temporal window between the early to mid-pachytene transition (Figure 6B–6D), indicating that its appearance was delayed compared to mice with standard karyotype [30], [70], and it remained detectable until the end of diplotene. XMR started to accumulate on the unsynapsed chromatin of trivalents and of the sex chromosomes at early to mid-pachytene transition and it disappeared at late diplotene (Figure 7). Interestingly, the intensity of the XMR signal seemed to be lower on the unsynapsed chromatin of the trivalents that were far from the sex chromosomes compared to that on the unsynapsed chromatin of the trivalents that were close to the sex body. These results suggest that the location of XMR in the open trivalents could be influenced by their association with the sex chromosomes. In summary, our results show that the proteins γH2AX and ATR started to appear at the beginning of prophase I (leptotene), but intense labeling of ubiH2A, SUMO-1 and XMR was detected at a later stage (early to mid-pachytene). The appearance of ubiH2A and SUMO-1, which are known to be involved in DNA repair [71], [72] slightly preceded the spread of ATR from the chromosome axes to the unpaired chromatin (see Figure S3 for a comparison between the timing of the appearance of SUMO-1 and ATR on unsynapsed chromatin). Therefore, ATR was the last protein to appear on the unsynapsed chromatin at mid-pachytene (Figure 8). All these proteins remained localized on the unsynapsed regions of trivalents and of the sex chromosomes until late diplotene, indicating an active repair of DNA on the unsynapsed chromatin of trivalents. Also, at mid-pachytene, ubiH2A and SUMO-1 might be involved in determining those chromatin modifications that would lead to the transcriptional inactivation of unsynapsed chromatin [72], [73]. The presence of trivalents with unsynapsed proximal regions throughout the first meiotic prophase raises the question of how many of these trivalents achieve a complete synapsis. To this end, we analyzed the number of completely synapsed (closed) and partially unsynapsed (open) trivalents during early, mid- and late pachytene and during early and mid-late diplotene (Table 1) in two individuals. No statistical differences were found between them. At the beginning of pachytene, only 2. 22% of spermatocytes had completed synapsis in all trivalents, whereas most spermatocytes showed one to eight open trivalents, with spermatocytes having four open trivalents occurring at the highest frequency (27. 41%). During prophase progression, the frequency of spermatocytes with a high number of open trivalents tended to decrease, even though, at late pachytene, the great majority (87. 55%) of spermatocytes possessed open trivalents (Table 1). At diplotene, the frequency of spermatocytes with closed trivalents or with one open trivalent (recognized by the presence of γH2AX) increased slightly (18. 36% and 51. 20%, respectively), although not significantly when compared to that of late pachytene. On the contrary, the frequency of spermatocytes with two, three and four open trivalents slightly decreased (Table 1). These data show that most spermatocytes maintained partially unsynapsed trivalents throughout pachytene, although their number decreased towards the end of pachytene along with an increase of spermatocytes with completely synapsed trivalents. To estimate germ cell death, we made a quantitative evaluation of the TUNEL-positive cells present in seminiferous tubule cross-sections. Confirming previous results [39], [49], TUNEL positive cells were almost exclusively present in the meiotic compartment of stage XII of the seminiferous epithelium, which contains spermatocytes at the zygotene-pachytene transition, metaphase I and II. An average of 19. 44% (±4. 37) of spermatocytes were TUNEL-positive, most of which were at the metaphase stage (Figure 9). When we specifically evaluated metaphase cells at stage XII, we found that 63% of them were TUNEL positive, as shown in an our previous study [39]. This suggests that 37% of metaphase cells are able to pass the spindle checkpoint and progress to further stages of differentiation. In this regard, we previously reported a mean ratio between round spermatids and pachytene spermatocytes of 1. 43, corresponding to 36% of germ cell survival following meiosis in the same type of Rb heterozygous mice, although in the homozygous parentals germ cell survival is 84% and 86% for 2n = 40 and 2n = 24 karyotypes, respectively [40], [43]. Moreover, the absence of extensive cell death in other stages of the spermatogenetic cycle suggests that pachytene and diplotene spermatocytes are able to progress to meiotic divisions despite the presence of unsynapsed trivalents. It has been repeatedly reported that synapsis is delayed in heterozygotes for Rb translocations [44], [45], [47], [51]–[54], [57] and other chromosomal rearrangements [46], [74], [75]. The results presented here are in agreement with these previous reports. During zygotene, while synapsis progresses rapidly in the bivalents, in the trivalents it is initiated at the distal ends and then slowly progresses to the proximal ends of the acrocentrics. Previous reports have suggested that a delay in synapsis might be influenced by architectural constraints [43], [52], [58], [76]. In fact, the centromeres and proximal telomeres of acrocentric chromosomes are located at the nuclear periphery, while centromeres of metacentric chromosomes are located more internally in the nucleus (unpublished observations). This distinct localization of centromeres is defined by the different trajectory of the metacentric chromosomes' AEs within the nuclear space compared to that of acrocentrics' AEs [76], [77]. As synapsis of trivalents progresses from their distal telomeres, metacentric centromeres tend to approach to the nuclear envelope, where acrocentric centromeres and proximal telomeres are bound. These circumstances would explain why at the beginning of pachytene, while bivalents and sex chromosomes have achieved their respective synapsis, trivalents still appear with an open configuration. The presence of many unsynapsed proximal regions in acrocentric chromosomes located at the nuclear periphery would promote their association, causing the appearance of an ectopic heterologous synapsis between them or with the sex chromosomes at early pachytene. Most of these associations tend to disappear as trivalents complete their synapsis during mid and late pachytene. In agreement with previous reports [45], [52], our results show a decrease in the number of open trivalents throughout pachytene. These results suggest that trivalents can complete synapsis during the mid and late pachytene stages, as previously reported by Moses and coworkers [52]. However, contrary to the results found in lemur Rb heterozygotes [52], in which all trivalents finally achieve complete synapsis, in mouse there is a striking persistence of trivalents in open configuration throughout pachytene and at later stages [45]. Compared to autosomal bivalents, we found that trivalents retain RAD51 until later stages; however, RAD51 foci are not specifically enriched in the unsynapsed segments of the trivalents, a finding that differs from previous studies that have reported the maintenance of this protein on asynaptic autosomal segments [78], [79]. RAD51 finally disappears from the trivalents during mid-pachytene, despite the presence of unsynapsed segments. This circumstance has two interesting implications. First, our results confirm previous data that cells can accomplish prophase-I with unsynapsed autosomes [1], [33], [36], [64], [80], [81]. Since completion of the recombination/repair process is considered necessary to bypass the pachytene checkpoint [2], [4], [5], [82] it is likely that unsynapsed segments are repaired by the end of pachytene. This behavior parallels the situation found in the sex chromosomes. Second, some trivalents probably complete synapsis after RAD51 has disappeared, indicating the existence of a mechanism that is able to complete synapsis independently of the usual recombination/repair pathway [83]. Interestingly, many of these late synapsis events culminate with the heterologous synapsis of acrocentric chromosomes within each trivalent. This process, called synaptic adjustment, has been previously reported for these and other chromosomal rearrangements [44], [45], [52], [78], [84]. An additional consequence of both the persistence of unsynapsed and the presence of non-homologous synapsed chromosome regions is the reduction of chromosome segments where reciprocal homologous recombination could take place. This could account, at least partially, for the displacement of chiasma from the centromeric regions and the overall decrease of recombination frequency observed in Rb heterozygotes [57], [85]. The results presented here show that the unsynapsed regions of trivalents incorporate many of the proteins related to MSUC, such as γH2AX, ATR and ubiH2A [17], [18] and some markers that have been previously reported only in association with MSCI, such as SUMO-1 and XMR [27], [30], supporting the idea that MSCI could be a particular case of MSUC [17], [18]. Our study on Rb heterozygotes reveals further interesting features of the MSUC process. We found that during early pachytene, some unsynapsed regions do not exhibit either γH2AX or ATR signals. This labeling is especially striking in those trivalents in which one of the open acrocentrics incorporates these markers while the other does not (see Figure 3D). This absence of either γH2AX or ATR signals might be due to a limited availability in the meiocytes of factors triggering MSUC and MSCI, like BRCA1 and ATR, as recently suggested [33], [36]. However, alternative explanations could be formulated taking into account that: 1), unlabeled unsynased chromosome segments are found in cells with either a high or a low number of open trivalents; 2), we never observed MSCI to be hampered in the sex chromosomes. Since the absence of either γH2AX or ATR labeling on some unsynapsed regions is mainly found at early pachytene, we favor the interpretation that unlabeled chromatin could represent chromosomal regions that are about to synapse and/or are asynaptic but MSUC is not initiated yet. In our model, asynapsis could not be extensive enough to exhaust MSUC/MSCI triggering factors; asynapsis in each trivalent affects just a short chromosome length, thus the total amount of unsynapsed chromatin in Rb heterozygous mice is lower than in other mouse models [33], [36]. However, given the physiological interdependence of spermatocytes in the seminiferous epithelium provided by the presence of intercellular bridges [86]–[88], it is also likely that cytoplasmic flux could compensate the mRNA/protein levels of MSUC components among different cells, buffering the effect of extensive asynapsis in some spermatocytes. These facts could determine the success of spermatocytes to have a normal MSUC/MSCI performance during the first prophase and will serve to avoid stage IV pachytene apoptosis. Our study also adds new clues to the understanding of the sequence of initiation and spreading of chromatin modifications involved in MSUC. H2AX phosphorylation detected at late zygotene was the first modification found in unsynapsed chromatin. This was followed by the accumulation of ubiH2A, SUMO-1, XMR and finally ATR on these regions during the early-mid pachytene transition. Thus, we suggest that the modifications of the chromatin involved in MSUC occur in at least two phases (Figure 8). The first phase initiates with the phosphorylation of H2AX, resulting in chromatin silencing at leptotene/zygotene. The second phase starts at early-mid pachytene with a second round of chromatin modifications, probably driven by the persistence of ATR at unsynapsed AEs, and it involves the incorporation of ubiH2A, SUMO-1, XMR, and finally ATR into unsynapsed chromatin. Whether it also involves other histone replacements and/or modifications, such as histone H3. 1 and H3. 2 replacement by H3. 3 and H3, and H4 methylation [26], or the incorporation of other specific proteins or RNAs, remains to be determined. Our analysis of the temporal appearance and localization of the proteins involved in MSUC has shown that ATR starts to spread over the chromatin of unsynapsed trivalents only at mid-pachytene, after the massive accumulation of γH2AX, while ubiH2A, SUMO-1, and XMR accumulate throughout early pachytene. Previous studies have suggested that ATR is involved in phosphorylating H2AX on the surrounding chromatin at late zygotene [17], [21] and that XMR and SUMO-1 accumulate on the sex body during early pachytene [27], [30], [70]. Although the pattern of appearance of some of these proteins is not completely established and discrepancies have been reported by different authors [17], [21], [30], [70], [89], [90], the comparison of these studies with our results suggests that: 1) the incorporation of many MSUC-related factors is delayed in Rb heterozygotes compared to homozygotes; and 2) our cytological approach, and previous studies [89], [91], are not completely congruent with the role of ATR in phosphorylating H2AX at late zygotene. Since we cannot rule out that undetectable amounts of ATR are present in the unsynapsed chromatin at late zygotene, other methodological approaches would be necessary to confirm this issue. Finally, our results indicate that in mouse MSUC is triggered during zygotene-early pachytene and that desynapsing LEs at diplotene do not incorporate MSUC markers, even if they are adjacent to regions that have remained unsynapsed during pachytene. This differs from the recently reported dynamics of sex chromosome inactivation in chicken females, in which two waves of H2AX phosphorylation, one at zygotene and other one at late pachytene, have been detected [92]. These differences in MSUC dynamics open interesting questions in an evolutionary context. Meiotic failure has been postulated as one of the main causes of infertility in organisms bearing chromosomal rearrangements. Several models have been proposed to explain this phenomenon, including the alteration of transcriptional activity of autosomes and sex chromosomes [1], [17], [34], [93]–[95], the impairment of synapsis and recombination progression [8], [42], [45], [75], [82], [96], the alteration of nuclear architecture during prophase-I [43], and the incorrect orientation and segregation of chromosomes during meiotic divisions [39], [49], [50]. Current models postulate the existence of a pachytene checkpoint that monitors synapsis and/or recombination progression [2], [4], [5]. Pachytene arrest resulting from asynapsis has been proposed to occur as a consequence of MSUC through the inactivation of genes that are crucial to meiotic progression [17]. Additionally, it has been suggested that sequestration MSUC-related proteins like BRCA1 and ATR resulting from an excess of asynaptic chromosomes might prevent their relocation to the sex chromosomes, hampering MSCI initiation in males [1], [33] and an extensive MSUC response in females [36]. The subsequent inability to inactivate the sex chromosomes has been proposed as a primary cause of spermatocyte apoptosis in a variety of mouse models [1], [33]. The presence of many open trivalents in our model does not result in sequestration of repair factors such as ATR on unsynapsed autosomal regions, allowing the correct progression of MSCI. These results indicate that in our model asynapsis per se could not be sufficient to trigger pachytene arrest. This agrees with recent reports on human [64], [80], [81] and mouse meiosis [1], [33], [36] indicating that cells can “tolerate” a limited degree of asynapsis. Therefore, it seems likely that there is not an stringent synapsis-specific checkpoint acting during pachytene in mouse and that MSUC involvement in triggering a checkpoint during prophase-I through MSCI hampering could be limited to extreme asynaptic situations. Nevertheless, we consider important to stress that the impairment of the meiotic progression of spermatocytes with synaptic defects could still rely on the deregulation of gene expression caused by MSUC. In this sense, MSUC effects would greatly depend on the number and/or nature of genes that are transcriptionally inactivated [17]. In Rb heterozygotes, the unsynapsed segments comprise the pericentromeric heterochromatin-rich regions and euchromatic regions meager in genes, most of which might not be critical for meiosis progression and subsequent spermiogenesis. However, while MSUC has little effect in determining pachytene arrest in this model, it is likely that the effect could be much more relevant in other models. We found that in Rb heterozygotes meiotic failure occurs mainly during meiotic divisions, as we recorded a high proportion of apoptotic cells at stage XII of the seminiferous epithelium and very few TUNEL-positive pachytene spermatocytes. We are aware that apoptotic pachytene cells are very rapidly removed and difficult to document by TUNEL [50]. On the other hand, metaphase apoptotic cells may be difficult to eliminate from the seminiferous epithelium, causing and overestimation of cell dead at these stages [47]. However, our result are in agreement with previous reports showing that in Rb heterozygotes bearing trivalents or complex rings cell death is mainly found during meiotic divisions [47]–[49] while cell death mainly occurs during prophase-I in Rb heterozygotes bearing chromosome chains [41], [46]–[48]. Furthermore, the absence of massive cell death at the pachytene stage is also supported by our previous studies [39], [40], which showed only a slight reduction of the number of this type of spermatocytes from stage I to XI of the cycle of the seminiferous epithelium. This could account for the elimination of those spermatocytes with a high number of open trivalents, whereas those that have one to four open trivalents might be able to bypass pachytene arrest and proceed to further stages. Therefore, meiotic failure in our Rb heterozygotes seems to rely mainly on the action of checkpoints during metaphase I and II [39], [40], [45], [49]. Trivalents may have difficulties in achieving a correct orientation on the meiotic spindle, determining a delay of anaphase initiation that would lead to cell degeneration [7], [49], [97] and subsequent reduction of fertility. Paradoxically, despite the reduced fertility of heterozygous mice, Rb translocations are very frequent in wild populations [41], [98], spread rapidly [99], [100] and represent one of the main causes of karyotype evolution in mammals [101]. We propose that the circumvention of pachytene arrest even in the presence of chromosome regions subjected to MSUC, as demonstrated in the present study, could contribute to increasing the chances of many spermatocytes to reach meiotic divisions and to differentiate into viable sperm. Although substantial cell death is produced at the metaphase stage (up to 63%), the chances of producing viable gametes are still much higher than if a more stringent pachytene checkpoint were able to eliminate up to 87% (Table 1) of pachytene spermatocytes bearing unsynapsed chromosomes. In an evolutionary context, it must be stressed that when a chromosomal rearrangement arises in a natural population, the rearranged chromosomes must still pair, synapse, recombine and segregate from their cognate homologues. Therefore, the possibility that a chromosomal rearrangement will spread into a population would greatly depend on the meiotic defects it may cause in the heterozygotes. Thus, while Rb rearrangements may have a relatively mild effect on mouse pachytene progression, for other chromosomal rearrangements and organisms, this model cannot be applied [102]. Heterozygous Robertsonian mice (2n = 32, eight Robertsonian chromosomes in a heterozygous state) were generated by mating females of the laboratory strain CD1 (2n = 40, all acrocentric chromosomes) and males of the Milano II race (2n = 24, eight pairs of Robertsonian metacentrics in a homozygous state, Rb (2. 12), Rb (3. 4), Rb (5. 15), Rb (6. 7), Rb (8. 11), Rb (9. 14), Rb (10. 13), Rb (16. 17). Six three-month old male mice were analyzed. Mice were maintained at 22°C with a light/dark cycle of 12/12 hours and fed ad libitum. Procedures involving the use of the mice were approved by the animal ethics committees of the Faculty of Medicine, University of Chile, and the University of Pavia (Italy). Spermatocyte spreads and squashes were obtained following the procedures described by Peters et al. [103] and Page et al. [104]. The slides were placed in PBS and incubated with the following primary antibodies: mouse anti-SYCP3 1∶100 (Abcam, Ab12452); rabbit anti-SYCP3 1∶100 (Abcam, Ab15093); rabbit anti-SYCP1 1∶100 (Abcam, Ab15087); rabbit anti RAD51 1∶50 (Calbiochem, PC130); mouse anti-phospho-histone H2AX (Ser139) 1∶1000, clone JBW301 (Upstate, 05–636); goat anti ATR 1∶80 (Santa Cruz Biotechnology, sc-1887); mouse anti ubiquityl-histone H2A 1∶15, clone E6C5 (Upstate, 05–678); mouse anti GMP-1 (SUMO-1) 1∶50 (Zymed, 33–2400); mouse RIK2D3 1∶100 that recognizes the XMR protein in the testis [27], kindly provided by Denise Escalier (Université Paris 5, Paris, France). After rinsing in PBS, the slides were incubated with appropriate secondary antibodies diluted 1∶100 in PBS: FITC-conjugated donkey anti-rabbit IgG, FITC-conjugated donkey anti-mouse IgG, TR-conjugated donkey anti-mouse IgG and FITC-conjugated donkey anti-goat IgG. Slides where then stained with 1 µg/ml DAPI. After a final rinse in PBS, the slides were mounted with Vectashield. Observations were made in a Nikon (Tokyo, Japan) Optiphot or an Olympus BX61 microscope equipped with epifluorescence optics and the images were photographed on DS camera control unit DS-L1 Nikon or captured with an Olympus DP70 digital camera. All images were processed with Adobe Photoshop CS software. Immunolabeled spermatocytes were observed in an Olympus BX61 microscope equipped with a motorized Z-axis, epifluorescence and an Olympus DP70 digital camera. A collection of optical sections were captured using the analiSYS software (Soft Imaging System, Olympus). Images were subsequently analyzed and processed using the public domain software ImageJ (National Institutes of Health, United States; http: //rsb. info. nih. gov/ij), and the output video files were edited with VirtualDub (VirtualDub, http: //www. virtualdub. com). The right testis of three mice were fixed in Bouin' s fluid and embedded in paraffin wax. Five-micrometer serial transverse cross-sections were made and at least four serial sections per testis were mounted on each glass slide. One slide was stained by the periodic-acid-Schiff (PAS) reaction and counterstained with haematoxylin to identify the stages of seminiferous epithelium according to Oakberg [105]; the other slide was processed with the terminal deoxynucleotidyl transferase-mediated dUTP nick end-labelling (TUNEL) method, using an ApopTag Plus Peroxidase In Situ Apoptosis Kit (Chemicon-Millipore, Billerica, USA), according to the manufacturer' s instructions. Positive and negative controls were also set up. The positive controls were established using the slides contained in the same kit and following the manufacturer' s instructions. For the negative controls, sections were processed without TdT enzyme in the labelling reaction mix. The sections were counterstained with 0. 5% (w/v) methyl green for 10 min at room temperature. For each animal testis, 100 cross-sectioned tubules were scored to evaluate the frequency of apoptotic tubules. A cross-section of a tubule was considered apoptotic when three or more TUNEL-positive spermatocytes were present within the seminiferous epithelium [39], [49]. The percentage of TUNEL positive cells was calculated taking into account the total number of spermatocytes per tubule section. Abercrombie' s correction was applied to all cell counts [106]. We analyzed 724 and 415 spermatocytes from two three month-old heterozygous Robertsonian mice. The synapsed condition of heterologous region of Robertsonian trivalents was determined by morphological analysis identifying chromosomes with SYCP3 and the presence or absence of γH2AX positive signal in the chromatin. The data obtained from each mouse in each prophase I stage were summarized. Statistical significance between mice was assessed by the one way analysis of variance (ANOVA), followed by Tuckey post test. A Z test for two proportions was used to compare the number of spermatocytes between late pachytene, early diplotene and middle/late diplotene. In both statistical analyses a p value<0. 05 was considered statistically significant with a confidence interval of 95%.

Cells have different mechanisms to assess the proper occurrence of cellular events. These mechanisms are called checkpoints and are involved in the surveillance of processes such as DNA replication and cell division. A checkpoint at the pachytene stage arrests meiosis when defects in the process of homologous chromosome synapsis and recombination are detected. In mammals, both transcriptional inactivation of chromosomal regions that are not correctly synapsed at pachytene and activation of sex chromosome genes that are normally silent during this stage could contribute to meiotic arrest. We found that when Robertsonian translocations appear in heterozygosis, many synapsis defects occur, and mechanisms that trigger transcriptional silencing of the unsynapsed chromatin are activated. However, meiotic prophase-I progression is not greatly compromised. This questions the ability of the meiotic checkpoints to halt meiosis progression when synapsis is not completed, allowing cells with synapsis defects to reach the first meiotic division. The fertility reduction of Robertsonian heterozygous mice seems to be mainly caused by errors detected by the metaphase-I checkpoint, when most of the spermatocytes die, rather than by synapsis defects. In an evolutionary context, a permissive pachytene checkpoint could contribute to increasing the chances of Robertsonian translocations to spread into natural populations.

Abstract Introduction Results Discussion Materials and Methods

developmental biology/germ cells cell biology/cell growth and division genetics and genomics/chromosome biology evolutionary biology/nuclear structure and function

2009

A High Incidence of Meiotic Silencing of Unsynapsed Chromatin Is Not Associated with Substantial Pachytene Loss in Heterozygous Male Mice Carrying Multiple Simple Robertsonian Translocations

When early canonical Wnt is experimentally inhibited, sea urchin embryos embody the concept of a Default Model in vivo because most of the ectodermal cell fates are specified as anterior neuroectoderm. Using this model, we describe here how the combination of orthogonally functioning anteroposterior Wnt and dorsoventral Nodal signals and their targeting transcription factors, FoxQ2 and Homeobrain, regulates the precise patterning of normal neuroectoderm, of which serotonergic neurons are differentiated only at the dorsal/lateral edge. Loss-of-function experiments revealed that ventral Nodal is required for suppressing the serotonergic neural fate in the ventral side of the neuroectoderm through the maintenance of foxQ2 and the repression of homeobrain expression. In addition, non-canonical Wnt suppressed homeobrain in the anterior end of the neuroectoderm, where serotonergic neurons are not differentiated. Canonical Wnt, however, suppresses foxQ2 to promote neural differentiation. Therefore, the three-dimensionally complex patterning of the neuroectoderm is created by cooperative signals, which are essential for the formation of primary and secondary body axes during embryogenesis. Embryonic cells of some animals tend to be differentiated into neuroectoderm cells/neural progenitors unless they receive an extrinsic signal, so-called default model [1,2]. This characteristic is also applicable to mammalian embryonic stem cells and induced pluripotent cells (e. g. , [3,4]). Therefore, normal development in such organisms can be rephrased as molecular mechanisms that repress the initial neuroectodermal fate and drive them to be differentiated into different cell types. Transforming growth factor-ß (TGF-ß) family members are one group of well-described signaling molecules that play essential roles in determining non-neuroectodermal cell fates. Among these, Chordin and Noggin, which were initially reported as neural inducers, function in protecting the initial neuroectodermal fate at the dorsal side in vertebrates from invading bone morphogenetic protein (BMP) signals that are expressed at the ventral side and that specify a non-neuroectodermal fate [2]. Wnts, another type of secreted signaling molecule, also have functions in repressing the initial anterior neuroectodermal fate. In vertebrates, posteriorly functioning Wnt inhibits anterior neuroectoderm specification genes, such as otx2, and leads to the specification of posterior neuroectoderm [5]. Together, these secreted signaling molecules that regulate body axis formation act to suppress the initial neuroectodermal fate during early embryogenesis. However, despite a large number of these non-neuroectodermal signals, embryos still maintain the neurogenic region in its proper size and location. In addition, within the remaining initial neurogenic ectoderm, each terminal cell differentiation is precisely controlled to organize the complicated neural network, i. e. , the patterning of the neurogenic ectoderm is highly sophisticated in the restricted neuroectoderm of normal embryos. It has been suggested that the pre-signaling cell fate in sea urchin embryos is also anterior neuroectoderm, called the animal plate (AP). This is shown by an experiment in which the earliest canonical Wnt (cWnt) signal is inhibited by injecting the intracellular domain of cadherin (Δcad) to interfere with the nuclear localization of ß-catenin, resulting in most of the ectoderm of the injected embryos becoming specified as AP and differentiating into serotonergic neurons as well as other types of neurons and non-neural cells (Fig 1A: [6,7]). The expanded AP in the early cWnt-deficient embryos lacks patterning, and the serotonergic neurons are therefore dispersed throughout the AP. In contrast, the restricted AP in normal embryos differentiates into serotonergic neurons only at the dorsal/lateral edge, i. e. , there are no serotonergic neurons observed at the ventral edge and central part (anterior end) of the AP (Fig 1A), even though most of the cells in the anteriorly restricted neurogenic region have the potential to be serotonergic neurons [8]. Nodal-BMP2/4, via Smad2/3-1/5/8 signaling along the dorsoventral axis, is one of the signaling networks that regulates the specification of cell fate and patterning in this region (reviewed in [9]), but their target transcription factors remain unclear. In summary, the developmental features of the AP of the sea urchin embryo are the following: 1) the serotonergic neural fate is executed only at the dorsal/lateral edge of the neuroectoderm, 2) the anterior end (i. e. , the central part) of the AP does not differentiate serotonergic neurons, and 3) no serotonergic neurons appear at the ventral edge of the AP. Although information regarding the morphological and phenomenological characteristics of the development of the AP in sea urchin embryos has accumulated, the details of the molecular mechanisms that perform the intrinsic system of serotonergic neural fate specification at the dorsal/lateral edge of the neuroectoderm and that suppress the neural fate in other regions must still be defined. Thus, we have focused on the functional regulation between the signaling molecules and the transcription factors that control the patterning of the AP in the sea urchin embryo. Serotonergic neurons in the sea urchin embryo are differentiated within the AP by nearly bilateral patterning (Fig 1A: [9,10]). Among the transcription factors that are zygotically expressed in the AP, the earliest is foxQ2. Based on its expression pattern and previous experimental data, FoxQ2 is present in all AP cells during early embryogenesis (Fig 1B–1E: [11,12]), and it is essential for the specification of most of the cell types in the AP region, including the serotonergic neurons and apical tuft of Hemicentrotus pulcherrimus (Fig 1G and 1H: [8,13]) and Strongylocentrotus purpuratus [12]. However, FoxQ2 mRNA disappears from the dorsal/lateral edge of the neuroectoderm, where the serotonergic neurons are differentiated (Fig 1F, arrowhead: [13]), and the protein cannot be detected in differentiating serotonergic neurons (Fig 1G–1J). In addition, FoxQ2 plays an essential role in the formation of the apical tuft cilia through the maintenance of the ankAT-1 gene in later stages [14]. Because apical tuft cells are not serotonergic, these data suggest that FoxQ2 is first required for the specification of the most of the AP cells [12] but that the expression is subsequently suppressed in the cells at the dorsal/lateral edge of the AP, in which the serotonergic neural fate is executed. Thus, identifying the regulatory mechanisms of FoxQ2/foxQ2 patterning along the dorsoventral axis must be one of the keys to understanding how the initial neurogenic ectoderm is patterned during sea urchin embryogenesis. Homeobrain (Hbn: LC064116 for Hemicentrotus pulcherrimus Hbn) is a paired-like homeobox gene that is classified into the homeobrain-like (hbnl) family [15]. The gene expression patterns of hbnl family members have been reported in the fruit fly [16], sandworm [15], sea urchin [17,18] and sea anemone [19], but the hbn gene has not been identified in chordate genomes. The hbn expression pattern was first investigated in Drosophila melanogaster, where it initially appears in the anterior dorsal head primordium, which forms portions of the brain, and then in the ventral nerve cord during later stages. In sandworms (Capitella sp. I), hbn expression was detected in the developing brain, as in fruit flies, and in its larval eyes. In the sea anemone Nematostella vectensis, hbn is expressed throughout the blastoderm except for around the blastopore, and its expression is excluded from the aboral pole, where the apical tuft and the subsequent neurogenic region are formed. The expression pattern of hbn in sea urchin embryos was reported during the genome sequencing of Strongylocentrotus purpuratus [17,18,20,21]. In those studies, hbn was initially expressed in the animal pole region during the early blastula stage, and, at later stages, it appeared outside of and then disappeared from the AP, where foxQ2 was expressed. Despite reports on the existence of the gene in some species, the control of its expression and its molecular function has not been investigated in all animals. Here, we focus on the function of Hbn because it is expressed in the same region as the foxQ2 gene during the early specification stage of the AP. Then, we describe the roles of Hbn in the specification of serotonergic neurons and report that the regulation of hbn and foxQ2 expression by Wnt and TGF-ß signals are essential for the precise patterning of the embryonic AP in the sea urchin H. pulcherrimus. In adding to FoxQ2, we focused on the function of Hbn, another AP-specific factor. hbn is initially expressed throughout the AP (Fig 2A), as previously described in different species [18]. During subsequent developmental stages, the expression of hbn progressively fades from the ventral half of the AP and appears at the dorsal/lateral ectoderm, adjacent to the AP, in the early gastrula stage (24 h) (Fig 2B and 2C). At the late gastrula stage (30 h), its expression is restricted to the dorsal/lateral ectoderm and it completely disappears from the anterior end of the AP (Fig 2D). In addition to its dorsal/lateral ectodermal expression, hbn appears at the upper lip region in the prism stage (36 h, Fig 2E, arrow), where it remains, at least until the pluteus stage (48 h Fig 2F, arrow). To compare the expression pattern of hbn with that of foxQ2 or tryptophan 5-hydroxylase (tph), a serotonin synthase gene, we employed two-color fluorescence in situ hybridization. foxQ2 and hbn expression nearly overlapped in the AP region of the unhatched blastula (Fig 2G), but hbn gradually faded from the region and a portion of the dorsal/lateral ectoderm that was adjacent to the AP began to express hbn, which resulted in the expression pattern of hbn being ‘shifted’ toward the dorsal side away from the AP (Fig 2H and 2I). By the late gastrula stage, hbn expression had completely disappeared from the foxQ2 area (Fig 2J). Double staining of hbn and tph in the pluteus stage showed that the serotonergic neurons were differentiated at the edge of the AP, which was adjacent to the hbn-expressing region (Fig 2K and 2K’). Hbn morphants developed into pluteus larvae without detectable defects in morphology or developmental timing, except for a defect in the elongation of the anterolateral arms at 72 h (cf. Fig 2N with 2L) and at 96 h (cf. S1C with S1A Fig). In addition, Hbn morphants have significantly fewer serotonergic neurons than normal embryos while non-serotonergic neurons at the AP and ciliary band are almost normal (Fig 2M, 2O, S1A–S1D and S1G Fig). Because the development of serotonergic neurons is affected by several signals from outside of the AP [9], we employed a Δcad-injected embryo to accentuate Hbn function under conditions where all other known signals were eliminated [22,23]. A Δcad-injected embryo, in which the initially fated AP contains a greatly increased number of serotonergic neurons (Fig 2P), is an appropriate system to analyze the intrinsic function of genes that are expressed within it, as was previously reported [6]. When Hbn was knocked down in Δcad-injected embryos, the development of serotonergic neurons was strongly inhibited, as was observed in normal morphants (Fig 2P and 2Q). These phenotypes were specific because they were also obtained when using a second morpholino (Hbn-MO2) that targeted a non-overlapping sequence in the mRNA (S1E and S1F Fig), and the microinjection of an mRNA encoding Hbn protein partially rescued the morpholino knockdown effect under the early cWnt-deficient condition (S2A–S2J Fig). These results indicate that Hbn is required for the development of serotonergic neurons in the AP. To identify the step in which Hbn is involved during the development of serotonergic neurons, we examined Hbn morphants for the expression of tph, which is a terminal differentiation marker, and zinc finger homeobox 1 (zfhx1; [8]) and forebrain embryonic zinc finger (fez), which are early neural markers [14]. In Hbn morphants, tph was not expressed in the neuroectoderm (Fig 2R and 2S, arrowheads show tph-cells in the control), which indicated that the transcription of tph required Hbn function. zfhx1 and fez are downstream of FoxQ2 but are independent of each other. Hbn morphants expressed neither of these genes in the AP (Fig 2T–2W, arrowheads show the expression patterns of each gene in the control embryos). As expected, zfhx1 expression in the lateral ganglion was not affected in Hbn morphants (Fig 2U, arrows). In adding to the previous report, which showed that the entire AP region had the potential to produce serotonergic neurons [8], these data indicated that Hbn is required for the specification of serotonergic neurons. The change in the foxQ2 expression pattern along the dorsoventral axis suggested that its expression may be regulated by or depend upon TGF-ß signals because cell fate specification along the secondary, dorsal-ventral axis of sea urchin embryos was determined by TGF-ß family members such as Nodal and BMP2/4 [23–25]. Therefore, we examined whether the Nodal pathway is involved in the regulation of foxQ2 expression throughout development. In Nodal morphants, the size of the foxQ2 region was smaller than that of control embryos at the hatched blastula stage (18 h) (cf. Fig 3G with 3B, quantification of foxQ2 region was shown in P, Q), but they were invariant in unhatched blastulae (12 h) (Fig 3A, 3F and 3Q). The protein localization of FoxQ2 in hatched blastula also showed the same size as its mRNA pattern, and the immunochemical signal in Nodal morphants was weak (cf. Fig 3H–3J with 3C–3E, between arrowheads), which indicated that Nodal is required for maintaining foxQ2 expression during the blastula stages. In contrast, in Lefty morphants, in which Nodal proteins are located throughout the ectoderm [26], the size of the AP in the hatched blastula stage was significantly wider than that in control embryos (Fig 3L–3O and 3Q), which indicated that misexpressed Nodal interferes with the restriction of the neuroectoderm during blastula stages. The difference in the foxQ2-mRNA positive region in controls, Nodal morphants and Lefty morphants measured with the angle from posterior pole was supported by the data, in which we counted the number of FoxQ2-protein positive cells in 18 h stages (Fig 3R). Based on these data, Nodal maintains the expression of foxQ2 during blastula stages, and this mimics the process that occurs on the ventral side of the AP during normal development. Next, to investigate what controls hbn expression along the secondary axis, we observed its pattern in the embryos, in which the TGF-ß signals responsible for secondary axis formation were disturbed [23,25]. In Nodal morphants, hbn expression was shifted uniformly to the AP-adjacent region by the early gastrula stage (24 h) (cf. Fig 4H, 4I with 4C, 4D), which suggests that Nodal suppresses the expression of hbn on the ventral side of normal embryos. However, quantitative PCR (qPCR) data indicated that the amount of hbn mRNA was not significantly changed in the morphants or even in Nodal-misexpressed embryos (S3 Fig), suggesting that the function of a strong hbn inducer was missing in both Nodal morphants and misexpressed embryos. As a result of the uniform shifting of hbn, zfhx1-positive cells were distributed around the foxQ2 region in Nodal morphants (Fig 4I and 4J, asterisks), whereas in normal embryos, the precursor cells of serotonergic neurons were present at the dorsal edge of the AP (Fig 4D and 4E). In Lefty or BMP2/4 morphants, in which the dorsal ectoderm is missing and the ventral ectoderm surrounds the AP [24,26], but in which hbn expression at the unhatched blastula stage (12 h) is not changed (cf. Fig 4K, 4P with 4A), the expression patterns became obscure after hatching and never showed clear shifting towards the edge of the AP region, unlike in control or Nodal morphants (Fig 4K–4N and 4P–4S). These results suggest that Lefty and BMP2/4 are required for maintaining the strong expression of hbn after the hatched blastula stage. In fact, because the significant decrease of hbn mRNA was observed only in BMP2/4 morphants by qPCR (S3D Fig), we concluded that BMP2/4 is essential for the maintenance of hbn on the dorsal side. Furthermore, when the clear shifting of hbn expression towards the edge of the AP was almost entirely missing in these morphants, no zfhx1 expression was observed at the AP region (Fig 4N, 4O, 4S and 4T), resulting in the loss of all serotonergic neurons, as previously reported [27]. Taken together, the findings show that Hbn plays a role in an intrinsic system that determines the initial neural fate at the dorsal/lateral edge of the AP and that its expression patterns are highly regulated by TGF-ß signals along the dorsoventral axis. The next question was what further restricts the foxQ2 region further to the anterior end without Nodal signaling (Fig 3G and 3H). To investigate this question, we focused on the cWnt pathway because the inhibition of the early cWnt pathway interfered not only with AP restriction but also with AP patterning, resulting in the serotonergic neurons being differentiated in a dispersed manner (Fig 1A). Because the effect of an exogenous cadherin fragment (Δcad) on blocking cWnt [22] might be short-lived due to the short half-life of the injected mRNA, we blocked the function of low-density lipoprotein receptor-related protein 6 (LRP6: LC064120 for H. pulcherrimus LRP6), a co-receptor that acts with the Wnt receptor Frizzled (Fzl) to mediate the cWnt pathway. Based on previous reports, LRP6 mRNA is expressed maternally, lasts throughout embryogenesis [28] and is present in all cells during the early stages [29]. The same observations were made in H. pulcherrimus embryos (S4A–S4F Fig). Although the localization of LRP6 mRNA was uniform in embryos, the protein was missing in the ingressed and future mesodermal region at the mesenchyme blastula stage (S4G and S4H Fig). In LRP6 morphants, the protein-detection level was significantly decreased (S4I and S4J Fig; each image was captured in the same microscopic condition), and mesenchyme ingression was normal but no endoderm invagination was observed (Fig 5A and 5B). This result occurred because the mRNA and likely, protein of LRP6 were present maternally, and LRP6-MO could not block the early cWnt, unlike Δcad injection. In LRP6 morphants, the foxQ2 expressing region was significantly wider than that of control embryos at 24 h (Fig 5A–5C), and the dorsal-ventral polarity in the ectoderm was normal based on nodal and hnf6 expression patterns (S4K–S4O Fig; [23]). The morphant retained the apical tuft, which should disappear by 72 h during normal development (Fig 5D and 5E; [13]), and, intriguingly, its essential regulatory gene, foxQ2, was also still detected at 96 h (Fig 5F and 5G). This result indicates that LRP6-mediated cWnt signaling is required for suppressing foxQ2 expression in the AP. Despite the wider foxQ2 region, LRP6 morphants had no differentiated serotonergic neurons at 48 h (Fig 5H and 5I). This result was likely derived from the maintenance of FoxQ2 in each cell in the expanded AP. This was confirmed in 48 h Δcad-embryos and Δcad-LRP morphants because a number of serotonergic neurons and no FoxQ2 protein were observed in the former, while strong FoxQ2 signal and no serotonergic neurons were observed in the latter (Fig 5J and 5K). Taken together, these results indicate that an LRP6-mediated signal seems to be involved in the disappearance of FoxQ2 from the AP and for the precise differentiation timing of serotonergic neurons. Among the Frizzled receptors for the Wnt pathway in sea urchin embryos, it was reported that Fzl1/2/7 and Fzl5/8 are expressed in the ectoderm [30], and Fzl5/8 is likely the only Fzl receptor whose function we can analyze during the modification of the restricted AP region because Fzl1/2/7 morphants lose the entire AP from the very beginning of its formation [31]. Thus, we investigated whether Fzl5/8 mediates the LRP6-based cWnt pathway that controls the disappearance of FoxQ2 from AP cells. In Fzl5/8 morphants at 56 h, the expression of foxQ2 was maintained (Fig 5M), whereas control embryos lost the foxQ2 message at this stage (Fig 5L). This result suggested that Fzl5/8 functions in controlling the cWnt at the AP region. The disappearance of hbn from the anterior end occurred normally in Nodal morphants, and the hbn-expressing region surrounded the central part of the AP (Fig 5N), which suggests that the disappearance of hbn expression is independent of the dorsoventral axis formation by TGF-ß signals (Fig 4). To confirm this finding and to investigate the involvement of the cWnt pathway as an upstream factor of TGF-ß signals during hbn clearance, we employed Δcad-embryos, which lack all known zygotic signals including early cWnt and TGF-ß signals [22,23]. In these embryos, hbn is expressed throughout the entire region during the early stages (Fig 5O and 5P). Then, hbn disappears from the central portion of the AP by 30 h (Fig 5Q; [18]). The hbn-negative region is progressively expanded, and the hbn-expressing region is observed only in the squamous epithelia in the posterior half of 48 h Δcad-embryos (Fig 5R). This result supports previous data from a different species, S. purpuratus [18]. This disappearance pattern of hbn in Δcad-injected embryos was spatially similar to that observed in the anterior end area of normal embryos, which suggests that the spatial control of the disappearance of hbn expression from the anterior end of the AP is independent of cWnt and TGF-ß signals. Focusing on the hbn expression pattern in Δcad-embryos in detail, we found that the disappearance of the gene from the anterior end was delayed. In normal embryos, hbn began to be diminished from the anterior end of the AP by 18 h (Fig 2H), but it did not disappear until 30 h in Δcad-embryos (Fig 5Q). In addition, because the cWnt pathway likely regulates the disappearance of foxQ2 as mentioned above, we investigated the function of LRP6 on hbn regulation. hbn gene expression remained in the entire anterior half in LRP6 morphants at 24 h, at when the clearance of the gene was quite obvious in the controls (Fig 5S and 5T). However, hbn had disappeared from the AP by 50 h, as in normal embryos (Fig 5U and 5V). This was confirmed by the result from double fluorescent in situ hybridization for foxQ2 (Fig 5W). In addition, the disappearance of hbn from the AP region is independent of the absence of LRP6 function (S4P Fig). Based on these observations, the LRP-mediated cWnt signal is not required for the disappearance of hbn from the AP, but it is required for the control of the timing of its clearance. To find the ligands for cWnt signaling in suppressing foxQ2, we focused on later-expressed Wnts in this study because the early Wnts that function in endomesoderm formation might have indirect effects on AP regulation. Based on the temporal expression profile previously reported, Wnt3 (LC064118 for H. pulcherrimus Wnt3), Wnt6 (LC0641198 for H. pulcherrimus Wnt6) and Wnt7 (LC064119 for H. pulcherrimus Wnt7) are expressed relatively late [28,32]. In H. pulcherrimus, wnt3 is expressed during the cleavage stage but not after the blastula stage, according to qPCR, whereas wnt6 and wnt7 were expressed after hatching (Fig 6A). Based on perturbation experiments, it is suggested that Wnt7 functions as a ligand for the cWnt pathway, and Wnt6 for non-cWnt pathways, and those reasons will be explained in this section for Wnt7 and in the next section for Wnt6. wnt7 was expressed broadly at 20 h and was abundantly expressed in the AP. The broad expression of wnt7 and its strong expression in the AP were invariable until 30 h (Fig 6B–6E). In Wnt7 morphants, foxQ2 mRNA was still expressed in the AP region even at 96 h (Fig 6F and 6H) and FoxQ2 protein remained to be detected in AP cell nuclei at the same stage (Fig 6G and 6I). Because FoxQ2 persisted, the differentiation of serotonergic neurons was extremely delayed. The number of serotonergic neurons at 72 h was significantly smaller than that in controls (Fig 6J). Because FoxQ2 persistence and missing serotonergic neurons were similar characteristics observed in LRP6 morphants, these results suggest that Wnt7 functions as a ligand of the cWnt pathway in mediating the differentiation of serotonergic neurons through the suppression of FoxQ2 expression. Although our data so far have suggested that the cWnt pathway is involved in AP patterning through suppressing foxQ2 expression, nuclear ß-catenin was not observed in the region until at least the 8th cleavage stage [22]. Because the antibody that recognizes nuclear ß-catenin in H. pulcherrimus is not available, we performed a TCF-luciferase reporter system (Top-Flash) assay to measure the level of cWnt signal [33,34]. Δcad-injected embryos have only approximately 30% Top-Flash activity compared to controls at 24 h (Fig 6K). This decreased activity tends to recover as the embryos grow due to the degradation of exogenous Δcad-mRNA and/or protein. However, without LRP6 or Wnt7 functions, Top-Flash activity remains low, and the scores of the activity are significantly lower than those in controls at both 24 h and 42 h (Fig 6K). These data support that in the AP region cWnt functions through the Wnt7-LRP6 pathway in suppressing foxQ2 expression. In contrast, Wnt7 morphants had normal clearance of hbn expression from the anterior end (Fig 6L and 6M), supporting the idea that cWnt is not involved in controlling hbn expression patterns. Because one of the interesting questions in this study is that of which signaling pathway crosstalks with Nodal signaling during the regulation of AP patterning (Fig 3), a cWnt pathway regulated by Wnt7 might be the candidate. In fact, when the Wnt7 function was blocked, the foxQ2 region was wider than in normal embryos at the blastula stage (S5A, S5B and S5D Fig). The excessive restriction of the foxQ2 region that was observed in Nodal morphants (S5A and S5C Fig) was rescued in Nodal-Wnt7 double morphants (S5A and S5E Fig), suggesting that Wnt7 is the factor that restricts foxQ2 expression to the anterior end and that the Nodal signal inhibits a Wnt7-mediated signaling pathway during the blastula stages. We next focused on the function of non-cWnt signals on AP patterning. As it is downstream of early cWnt signals from the posterior side, a c-Jun N-terminal kinase (JNK) signal functions in the restriction of the AP to the anterior end [31]. To examine whether a JNK signal also plays a role in AP patterning, we applied a JNK inhibitor from 2–4 cell stages and analyzed the expression patterns of foxQ2 at the desired stages. As was previously reported, the restriction of foxQ2 to the anterior end was inhibited in the absence of JNK function at 24 h (Fig 7A and 7C). foxQ2 disappearance from the AP was delayed in JNK-inhibited embryos, but the remaining signal was weak, and its area was very small at 60 h (Fig 7B and 7D). Unlike the cWnt pathway, the JNK pathway seem to be weakly involved in the maintenance of foxQ2 expression because the timing of the initial differentiation of the serotonergic neurons was slightly delayed (48 h, Fig 7E and 7F), but a number of serotonergic neurons were differentiated in the expanded AP one day later [31]. We next focused on the function of non-cWnt signals on hbn expression, and used a JNK inhibitor and analyzed the expression patterns of hbn. hbn clearance from the anterior end at 24 h was not observed in JNK-inhibited embryos (Fig 7G and 7I, arrowheads). In addition, the clearance was intriguingly not observed in JNK-inhibited embryos, even in the later stages (cf. Fig 7J with 7H; arrowheads). Together, these results suggest that a JNK signal acts as a part of non-cWnt signaling and that it mainly plays a role in the clearance of hbn from the anterior end of the AP. As mentioned above, Fzl5/8 is likely the only Fzl receptor whose function we can analyze during the modification of the restricted AP region [31]. Here, we investigated whether Fzl5/8 mediates the non-cWnt pathway that controls the clearance of hbn from the central part of the AP. In Fzl5/8 morphants at 56 h, the expression of hbn was maintained (Fig 7L) whereas all control embryos had no hbn mRNA (Fig 7K). This result suggested that Fzl5/8 mediates the non-cWnt signal at the AP region. To investigate the ligands for non-cWnt signaling in hbn patterning, we focused on Wnt6 because Wnt7 was not involved in the regulation of hbn expression (Fig 6). wnt6 is expressed in the veg2 endoderm region and is not obvious at the ectoderm (Fig 7M–7P). The morphology of Wnt6 morphants did not resemble a normal pluteus stage even at 56 h and had a straight archenteron and no pluteus arms (Fig 7Q and 7S). Focusing on the hbn expression pattern, we found that it did not disappear from the central part of the AP region in Wnt6 morphants (Fig 7S). This result was confirmed by experiments in embryos that were doubly injected with Δcad and Wnt6-MO, in which hbn was broadly expressed in the expanded AP (S6I and S6K Fig). To clarify this finding, we employed Nodal morphants; without Nodal function, hbn “shifting” to the periphery of the AP is more obvious (Fig 7R). The morphants, in which Nodal and Wnt6 were simultaneously knocked down, showed no hbn disappearance from the central part of the AP (Fig 7T). This result clearly indicated that Wnt6 is required for hbn suppression in the AP. Because hbn was maintained in the AP in Wnt6 morphants, serotonergic neurons were differentiated not at the edge of the region but at the anterior end of the AP (cf. Fig 7V, asterisk with U, arrow). These data are supported by the serotonergic neural patterns in later stage, in which the differentiated serotonergic neurons gather at the anterior end in Wnt6 morphants even if there is no Nodal inhibition (Fig 7W and 7X). In addition, the nuclear localization of FoxQ2 had already disappeared at 60 h, as in normal embryos (Fig 7U and 7V), indicating that Wnt6 is not strongly involved in the regulation of the FoxQ2 expression pattern, which is mediated by LRP6/cWnt signaling. Although it is difficult to distinguish, these results suggest that Wnt6 functions as one of the players in the non-cWnt pathways that regulate AP patterning, especially in the control of hbn expression. Because the expression patterns of foxQ2 and hbn are complementary after the gastrula stage in normal embryos, FoxQ2 is another candidate for suppressing hbn expression. To examine this possibility, we investigated hbn expression patterns in FoxQ2 morphants. The disappearance of hbn occurred normally in the morphants (S7A–S7F Fig), which suggests that FoxQ2 and its downstream genes do not regulate the suppression of hbn expression at the AP in later embryos. Furthermore, in Hbn morphants, the expression pattern of foxQ2 was the same as that in normal embryos (S7G–S7J Fig), indicating that FoxQ2 and Hbn are mutually independent. It was previously reported that serotonergic neurons in the neurogenic AP of the sea urchin embryo are formed at the dorsal/lateral edges of the region [9,10] and that the differentiation of serotonergic neurons at the ventral side is suppressed by Nodal, which is expressed in the ventral ectoderm [6]. In this study, we revealed that hbn expression is suppressed by Nodal on the ventral side but maintained by BMP2/4 on the dorsal side. hbn expression is eliminated from the animal pole, likely by the non-cWnt pathway mediated by Wnt6/JNK after the blastula stage, which will be discussed below, and its pattern forms a horseshoe-like shape (Fig 8A). This pattern was not reported in another sea urchin, S. purpuratus [18], but, in H. pulcherrimus, it is clear that the expression is missing at the ventral side of the normal AP. The loss of Nodal function supports this observation because Nodal morphants have a ring-like shape of hbn expression around the neurogenic ectoderm (Fig 4). Because the expression of foxQ2 is also under the control of secondary axis formation by the Nodal and BMP2/4 pathways (this study; [14]), we need to know whether Nodal and/BMP2/4 regulation is direct or indirect by further experiments, including chromatin immunoprecipitation analysis aimed at uncovering the cis-regulatory modules of foxQ2 and hbn. In vertebrates, TGF-ß signaling also functions in the neural plate patterning along the dorsoventral body axis [36]. For example, bmp2 and bmp7 expressing outside of the neural plate are necessary for the development of noradrenergic neurons through the induction of the homeodomain protein, phox2a, in zebrafish embryos [37]. Nodal, on the other hand, is required for suppressing the precocious acquisition of forebrain characteristics in mouse embryos [38]. Our data indicated that sea urchin embryos use the similar mechanisms to pattern the neuroectoderm, controlling the timing and location of the differentiation of serotonergic neurons. In addition, because there are other types of neurons present on the ventral side of the AP in sea urchin embryos [7], future investigations regarding the relationship between in-cell factors characterizing those neurons and TGF-ß signaling coming from the outside of the neurogenic AP region will lead us to understand the conserved mechanisms of neural patterning throughout the animal kingdom. Persistent FoxQ2 and apical tufts in the AP region in LRP6 morphants strongly indicate that the cWnt pathway is required to suppress FoxQ2 and exert a neural fate, although it has been reported that early cWnt, visualized with the nuclear localization of ß-catenin, was observed only at the posterior half of the embryo until the gastrula stage [22]. In addition, our results suggest that Wnt7 works as a ligand in the LRP-cWnt pathway in the AP that suppresses FoxQ2 with precise timing (Fig 6). Although we could not rule out the possibility that Wnt7 functions indirectly from outside of the AP of normal embryos, based on its expression patterns, the strong expression of wnt7 at the thickened AP of normal (Fig 6) and Δcad embryos (S6E and S6H Fig) suggested that it plays an intrinsic role in FoxQ2 suppression within the AP. The difference between LRP6 morphants (abundant FoxQ2 and no serotonergic neurons) and Δcad embryos (less FoxQ2 and a number of serotonergic neurons) might be attributed to the lifetime length of exogenous Δcad mRNA and/or protein. Because the injected mRNA can last approximately 24 hr (e. g. , S8A and S8B Fig), only early cWnt but not later cWnt is suppressed in Δcad embryos. This idea was well supported by Top-Flash assay (Fig 6). Of course, we cannot completely rule out the possibility that Δcad alone is not sufficient to block all cWnt, even during the early stages. The future TOP-Flash assay during the early stages can answer this question. Because foxQ2 expression is at a gradient from anterior tip to periphery (Fig 1; [11]), the area of the biggest effect of foxQ2 removal by Wnt7/cWnt might be the edge of the AP region, resulting in serotonergic neurons starting the differentiation process at the position. A previous report [31] showed that foxQ2 was expressed at the posterior half, where this gene is never detected by in situ hybridization in normal embryos, when Axin was misexpressed, and they implicated that foxQ2 is originally expressed throughout the embryo and early cWnt at the posterior half suppressed it. Thus, it is expected that the mechanisms suppressing foxQ2 expression were also applied in the AP region after it is restricted to the anterior end. The results from JNK inhibition led us to consider that non-cWnt is involved in neurogenesis in the AP of the sea urchin embryos (Fig 7). Additionally, data regarding its ligand, Wnt6, supported our JNK result (Fig 7). However, it is still not clear how Wnt6 mediates non-cWnt signaling in the AP region. Because it has been reported that wnt6 is zygotically expressed at the vegetal plate and functions in endomesoderm formation [30,39], it is possible that the signal is indirect. However, our data using Δcad and Wnt6-MO indicated that Wnt6 could function within the AP region even though the expression of mRNA at that location is faint (S6 Fig). More detailed studies of its distribution and function with protein level will be necessary to understand the complete mechanisms of Wnt6 function. In contrast to the Wnt7 data, the normal disappearance of FoxQ2 in Wnt6 morphants indicated that Wnt6 did not function as a ligand of the cWnt pathway in the AP. The non-cWnt includes pathways other than JNK, planar cell polarity (PCP) and Ca2+ pathways [40], suggesting that those pathways are also involved in AP patterning in the sea urchin embryo. In fact, as a ligand for the non-cWnt pathway, Wnt6 might not be sufficient because JNK inhibition led to some foxQ2 remaining in the AP, indicating that the JNK pathway weakly acts in suppressing FoxQ2 and affects the precise control of the timing and the location of serotonergic neurons (Fig 7). It was reported that the JNK pathway functions in restricting the AP region, represented by foxQ2 and hbn expression, to the anterior end during blastula stages [31], but the Wnt6 morphant had no expanded AP and tended to have a more restricted hbn region, supporting the idea that other ligands for non-cWnt signaling function during anterior neuroectoderm formation. Our results suggest that cWnt and non-cWnt signaling function in repressing FoxQ2 and Hbn, respectively, but we cannot completely rule out the possibility that each pathway affects both types of repression. This is because LRP6 morphants showed slightly delayed hbn clearance and JNK inhibition allowed some foxQ2 to remain in the AP region. Because cWnt and non-cWnt antagonize each other in some biological processes [41], this cross-reaction might be normal in AP formation in sea urchin embryos. We must also consider the functional combination of frizzled receptors, secreted frizzled-related protein, and Dickkopfs (Dkks) to determine the complete involvement of the Wnt pathways in AP patterning. In fact, the restriction of the AP to the anterior end is managed through their combination with other Wnt ligands, such as Wnt1 and Wnt8 [31], and the patterning of anterior structure, including neuroectoderm, is also regulated those factors in other deuterostomes [42–44]. In contrast, both LRP6 and Wnt7 morphants failed to finish the restriction of the AP region by the blastula stages (Figs 5 and 6), similar to Wnt1 and Wnt8 morphants [31], suggesting that those factors affect the early cWnt events that restrict AP to the anterior end. Our data using Fzl5/8 morphants suggested that this frizzled functions as a receptor for both cWnt and non-cWnt signals as was observed in other systems [45]. This relationship strongly supports the idea that cWnt and non-cWnt cross-react with each other through sharing the frizzled receptor during AP patterning, although we cannot rule out that other types of frizzled, which we did not analyze in this study, are more essential for each pathway. Adding to our knowledge of the involvement of these molecules, biochemical analyses to reveal ligand-receptor associations will be conducted in the future to understand the complete pathway regulating AP formation in the sea urchin embryo. As mentioned in the Introduction, the anterior neural fate in vertebrates is restricted by Wnt signaling from the posterior side [5]. Posterior Wnt signals are also reported in invertebrates, such as sea urchins [31,46,47] and amphioxus [48,49]. These species commonly use Wnt signals to establish posterior identities during early development. In this study, we found that wnt7 is expressed in the AP region and required for the differentiation of serotonergic neurons as a ligand for the cWnt pathway (Fig 6), indicating that the sea urchin embryos utilize cWnt signaling at both the posterior and anterior ends. In addition, anterior and posterior cWnt share a function, repressing foxQ2 expression (in this study, [12,31]). The simple mechanism that cWnt suppresses foxQ2 with shifting the functional timing, early at the posterior and later at the anterior ends, enables embryos to have a complicated body plan along the anterior-posterior body axis: FoxQ2 initially specifies the AP region only at the anterior end and later it disappears from the AP to let the serotonergic neurons differentiate within it. We accidentally found this both-end cWnt signaling because we blocked the early cWnt at the posterior end using exogenous mRNA encoding Δcad, which has a short life-time. If we permanently and completely inhibit some of the components of the cWnt pathway, it might be difficult to recognize the later functioning anterior cWnt. It is possible that this type of anterior cWnt commonly functions in other systems during early development because in vertebrates it was reported that the wnt7 family was expressed in the developing anterior neuroectoderm region [50]. One of the most interesting findings in this study is that the dorsoventral Nodal pathway might interfere with the anteroposterior Wnt pathway during the embryogenesis of the sea urchin. Because a number of studies previously revealed the molecular mechanisms of cell fate specification along these embryonic axes in many species, we have now accumulated the information to determine, for example, how the anteroposterior axis is formed by the bicoid gradient in the fly [51–53] or how left-right asymmetry is created by Nodal flow in mice [54,55]. However, because embryos must control cell fate specification along all three body-axes in our three-dimensional world with precise timing, the formation of the body axes should not be independent of each other. Thus, the information from processes that occur along each axis should be integrated with a high degree of sophistication and affect cell-fate specification during each step of embryogenesis. We have previously reported that a single transcription factor links anteroposterior-dorsoventral axis formation in the sea urchin embryo and that it regulates the timing of the onset of specification of the secondary axis downstream of primary axis formation [12]. By combining our results with previous reports, we propose the following five combinational signaling steps that regulate serotonergic neuron formation at the dorsal/lateral edge of the AP in the sea urchin embryo (Fig 8A and 8B): 1) posterior cWnt/non-cWnt signaling restricts the AP, which is specified by early FoxQ2, to the anterior end [6,31], 2) Wnt7/cWnt suppresses late FoxQ2, which induces the apical tuft cilia and represses neural fate, at the edge of the restricted AP along the anterior-posterior axis, 3) dorsoventral Nodal suppresses Wnt7/cWnt to maintain late FoxQ2 at the ventral side, 4) Wnt6/non-cWnt suppresses the neural specifier Hbn, preventing its expression in the anterior end, and 5) BMP2/4 strongly maintains the expression of neural specifier Hbn at the dorsal side whereas Nodal suppresses it at the ventral side. After the AP is restricted, the regulation of the expression of two opposing functional transcription factors, the neural specifier Hbn and late neural suppressor FoxQ2, is accomplished by the molecular mechanisms of neural patterning in the AP. Crosstalk between Wnts along the primary axis and Nodal along the secondary axis is carried out during the process of suppressing the serotonergic neural fate at the ventral side of the AP. Although suppressing the specifier of neurons, Hbn, at the ventral side seems to be sufficient, maintaining the suppressor, FoxQ2, within the same region is a great supporting system for embryos to ensure the removal of the neural fate. Our data suggested that the effect of Nodal suppressing cWnt at the AP region seems to reach to the dorsal edge (Fig 3). However, because it is reported that Nodal can diffuse to short range in AP region [56], we do not know how the Nodal pathway controls the sizing of the entire AP through interactions with cWnt signaling in the AP. As wnt7 is expressed abundantly in the AP (Fig 6), the Nodal pathway might affect its expression regulation even though Nodal itself can bind the receptor in a few rows at the ventral edge of the AP. In addition, the downstream factors of Nodal signaling, e. g. , Lefty and BMP2/4, can diffuse to the AP region [12,14,23], and they might interact with cWnt pathway directly or indirectly to regulate the size control of the AP. In S. purpuratus, it was indirectly implicated that Nodal regulates the expression of foxQ2 by controlling transcription factors, Not, and Emx. Their relationships seem to be complicated along the spatiotemporal patterning [57,58], and none of them have yet been analyzed in H. pulcherrimus. However, our data on Nodal loss-of-function and gain-of-function are quite reproducible during those stages (Fig 3), and even in S. purpuratus Nodal morphants had a smaller AP, judging from the distribution of serotonergic neurons [18], supporting the idea that Nodal maintains the foxQ2 expression in the sea urchin embryos. The precise cis-regulatory analysis of the foxQ2 expression pattern in the future will lead us to understand the detailed molecular mechanisms of how TGF-ß signals control AP patterning. Because we have not yet succeeded in completely dissociating single cell-cultures, we cannot, strictly speaking, conclude that the default fate of sea urchin cells is the neuroectoderm or neurons. However, if the earliest known signal, cWnt, which functions in the posterior half at the beginning of the 8–16 cell-stage [59], is blocked, almost the entire region develops into neurogenic AP [31], suggesting that the initial or pre-signaling fate of sea urchin cells is anterior neuroectoderm. Within this expanded pre-signaling AP, embryos differentiate a number of serotonergic neurons that are scattered throughout the AP, in which other cells are produced with long, immotile, apical tuft cilia [12,13]. Removing Notch signaling from the AP promotes an increased number and the clustering of serotonergic neurons, indicating that lateral inhibition in the AP is another signal that inhibits the serotonergic neural fate in the sea urchin embryo [8]. FoxQ2 is required to exert the serotonergic neural fate initially [12]. However, FoxQ2 is a bifunctional transcription factor that is required early for the specification of most of the cell types in the anterior neuroectoderm, and then it must be removed from the cell late, which subsequently takes on a serotonergic neural fate [8,12–14]. As an initial specifier, the function of FoxQ2 might be similar to that of Zfp521 in mouse embryos. Zfp521 is zygotically expressed by a cell-intrinsic mechanism to exert the initial neural fate [60]. In sea urchins, however, the initial specifier is substituted with a second one, Hbn, because the function of FoxQ2 becomes another one that is against serotonergic neural differentiation after the blastula stage. Taken together, the unknown mechanisms that initially induce foxQ2 and/or hbn at the anterior end of the sea urchin embryo are the substances that determine the initial cell fate, which will be clarified in the future through the analysis of cis-elements of the foxQ2 and hbn genes. So far, none of functional data of Hbn in other systems have been published, but understanding these mechanisms will lead us to answer the question of what is truly the default cell fate in the sea urchin embryo as well as in other organisms. Embryos of Hemicentrotus pulcherrimus were collected around Shimoda Marine Research Center, University of Tsukuba, and around the Marine and Coastal Research Center, Ochanomizu University. The divergence time between H. pulcherrimus used in this study and S. purpuratus used in most previously described studies was estimated to be 7. 2–14 million years ago [61], and the developmental time-course, gene expression patterns, and reported phenotypes in gene-knockdown and/or misexpressed experiments are almost the same. The gametes were collected by the intrablastocoelar injection of 0. 5 M KCl, and embryos were cultured in glass beakers or plastic dishes that contained filtered natural seawater (FSW) at 15°C. Cell-permeable JNK inhibitor I, (L) -Form (Merck Millipore, Billerica, MA, USA), was used at 50 μM from the two-cell stage to desired stages [31]. For the control experiment, we added same volume of dimethyl sulfoxide (DMSO), which is used for dissolving the JNK inhibitor. In whole-mount in situ hybridization, embryos were fixed with 3. 7% formaldehyde-sea water (SW) overnight at 4°C. After 7 x 7 min washes in MOPS buffer (0. 1 M MOPS, pH 7. 0,0. 5 M NaCl, 0. 1% Tween-20), MOPS buffer was substituted with hybridization buffer (HB: 70% formamide, 0. 1 M MOPS, pH 7. 0,0. 5 M NaCl, 0. 1% Tween-20,1% BSA), and specimens were pre-hybridized at 50°C for 1 h. Subsequently, pre-hybridization HB was substituted with fresh HB containing Dig-labeled RNA probes (0. 4 ng/μl final concentration), and samples were incubated at 50°C for 5–7 days. After washing in MOPS buffer for 7 min x 7 times at room temperature (RT), for 1 h x 3 times at 50°C, and for 7 min x 2 times at RT, samples were blocked with 1–5% skim milk (Nacalai Tesque, Tokyo, Japan) in MOPS buffer for 1 h at RT and thereafter incubated with anti-Dig antibody conjugated with alkaline phosphatase (Roche, Basel, Switzerland; 1: 1,500 dilution) overnight at 4°C. Tissue was washed with MOPS buffer for a half day with several buffer exchanges. Dig signal was detected with NBT/BCIP (Promega, Madison, WI, USA). For two-color fluorescent in situ hybridization, Dig-labeled and FITC-labeled probes were simultaneously applied to HB and detected with anti-Dig and anti-FITC POD-conjugated antibodies, respectively (Roche), followed by the Tyramide signal amplification plus system (TSA-plus; Perkin Elmer, Waltham, MA, USA). After blocking in 1–5% skim milk, specimens were incubated with 1: 1,000 diluted anti-Dig POD-conjugated antibody for 1 h at RT, washed with MOPS buffer for 7 min x 7 times at RT, and treated with tetramethylrhodamine TSA-plus for 10 min at RT. Then, samples were washed three times with MOPS buffer, and the remaining POD function was quenched by 0. 5% sodium azide in MOPS buffer for 30 min at RT. After washing, we repeated treatment of the samples with anti-FITC antibody and the FITC TSA-plus system. The size of the foxQ2-expressing region is quantified with the angle from the posterior end (Fig 3P). The angle was measured using ImageJ and Student’s t-test was applied to each quantification to judge whether their differences were significantly meaningful. The graph was drawn with software R [62]. In whole-mount immunohistochemistry, embryos were fixed with 3. 7% formaldehyde-SW for 10 min at RT. After washing with PBST (137 mM NaCl, 2. 7 mM KCl, 10 mM Na2HPO4,1. 76 mM KH2PO4, pH 7. 4,0. 1% Tween-20) for 7 min x 7 times, samples were blocked with 1–5% lamb serum in PBST for 1 h at RT and incubated with primary antibodies (dilutions: serotonin (Sigma-Aldrich, St. Louis, MO, USA) 1: 2,000, synB [7] 1: 100, LRP6 (Sigma) 1: 1,000, FoxQ2 [14] 1: 100 and c-myc (Sigma) 1: 1,000) overnight at 4°C. Antibodies were washed off with PBST for 7 min x 7 times, and the samples were incubated with the secondary antibodies (1: 2,000 diluted anti-rabbit IgG conjugated with Alexa 488 and/or 1: 2,000 diluted anti-mouse IgG conjugated with Alexa 568 (Thermo Fisher Scientific, Waltham, MA, USA) ) for 2 h at RT. The specimens were observed using a Zeiss Axio Imager. Z1 that was equipped with an Apotome system (Zeiss, Oberkochen, Germany) and an Olympus FV10i confocal laser scanning microscope (Olympus, Tokyo, Japan). The optical sections were stacked and analyzed using ImageJ and Adobe Photoshop. Panels and drawings for the figures were made using Microsoft PowerPoint. The number of FoxQ2-positive cells was counted under the fluorescent microscope (IX70, Olympus). Student t-test was applied on each quantification to judge whether their differences were significantly meaningful. The morpholino (Gene Tools, Philomath, OR, USA) sequences and the in-needle concentration with 24% glycerol were as follows: Hbn-MO1 (0. 7 mM): 5’- AAAATGAACGGAACAAGTCCAGTGT -3’, Hbn-MO2 (2. 0 mM): 5’- TAGGAGAACCAACGACCGCCGTCAT -3’, Nodal-MO (0. 2 mM): 5’- AGATCCGATGAACGATGCATGGTTA -3’, Lefty-MO (0. 4 mM): 5’- AGCACCGAGTGATAATTCCATATTG -3’, FoxQ2-MO (0. 2 mM): 5’- TCATGATGAAATGTTGGAACGAGAG -3’, BMP2/4-MO (0. 4 mM): 5’- GACCCCAATGTGAGGTGGTAACCAT -3’, LRP6-MO1 (1. 9 mM): 5’- GAAAGGTTTCAAGGCAGCCCATTTC -3’, LRP6-MO2 (1. 5 mM): 5’- TGCCGTTGACTAAATATCATCTACA -3’, Wnt6-MO1 (3. 8 mM): 5’- ACGTGTCCACTCCATCTTGTAATAC -3’, Wnt6-MO2 (1. 9 mM): 5’- TCGTCCAGCGATTTAATAAAGAGCT -3’, Wnt7-MO1 (3. 8 mM): 5’- ATAACCACACCAAgTTgggCCgCAT -3’, and Wnt7-MO2 (1. 9 mM): 5’- GCTCAGCGATGCCCGATGGATAAAA -3’. Two non-overlapping morpholinos that blocked the translation of Hbn, LRP6, Wnt6 and Wnt7 were used to confirm the specificity of their function. For negative control experiments, we injected 24% glycerol into eggs. mRNAs were synthesized from linearized plasmids using the mMessage mMachine kit (Thermo Fisher Scientific) and injected at the indicated concentrations in 24% glycerol in needles: hbn-mRNA (0. 1 μg/μl), Δ-cadherin (0. 3–0. 6 μg/μl; [22]), and myc-mRNA (0. 1 μg/μl). Microinjections into fertilized eggs and into one blastomere at the two-cell stage were performed as previously described [13]. Quantitative PCR (qPCR) was performed as previously described [13,63] with some modifications. The total RNA from 100 embryos of H. pulcherrimus was isolated, and reverse transcription was performed using the Realtime Ready Cell Lysis kit and Transcriptor Universal cDNA Master (Roche). GoTaq qPCR Master Mix (Promega) was used for PCR carried out with a Thermal Cycler Dice Real Time system (Takara, Shiga, Japan). Primer pairs used for qPCR were the following: Wnt3-qF1; 5’- TATATCCGGCAAACAGGTCC -3’, Wnt3-qR1; 5’- TCTTCTCCCTCGGAACTGAA -3’, Wnt6-qF1; 5’- GACCTGCTGGAAGAAAATGC -3’, Wnt6-qR1; 5’- GGGCTGTTTGACCGTATCAT -3’, Wnt7-qF1; 5’- CATGGTGTTTCAGGTTCGTG -3’, Wnt7-qR1; 5’- TCCTAGTTCGTTTGGCCTTG -3’, COI-qF1; 5’- CCGCATTCTTGCTCCTTCTT -3’, and COI-qR1; 5’- TGCTGGGTCGAAGAAAGTTG -3’. The relative concentrations of each mRNA were normalized with mitochondrial COI Ct values. Top-Flash plasmid M50 Super 8xTOPFlash (Addgene plasmid # 12456) and M51 Super 8xFOPFlash (TOPFlash mutant) (Addgene plasmid # 12457) were gifts from Dr. Randall Moon. DNA fragments containing TCF/LEF-binding sites with Firefly Luciferase gene were amplified by KOD-Fx DNA polymerase (TOYOBO, Tokyo, Japan) with RVprimer3 and EBV_rev_primer set and injected at 20 ng/μl in a needle into the fertilized eggs with a carrier EcoRV-digested H. pulcherrimus genomic DNA at 10 ng/μl. The signal was obtained from 20–40 embryos for each experiment (three independent batches) using the Bright-Glo Luciferase Assay System (Promega). The luminescence was detected with the LB941 Multimode Reader TriStar (Berthold Technologies GmbH & Co. KG, Bad Wildbad, Germany) for 60 sec. The Top-Flash signal was normalized to the Fop-Flash level for each experiment.

The sea urchin embryo is similar to vertebrate embryos in that the default cell fate is potentially neurogenic, and normal development restricts the neural fate to the narrow area that locates at the anterior/dorsal region of the embryo. Because maintaining the default neural fate to the anterior/dorsal region is required for embryos to precisely integrate information from both the primary anterior-posterior and secondary dorsal-ventral body axes, these axes must be mutually linked by some mechanisms. In this study, we describe how the combination of orthogonally functioning signaling pathways regulates their targeting transcription factors expressing at the anterior neuroectoderm to restrict and pattern the default neurogenic region. By loss-of-function experiments using sea urchin embryos, we revealed that canonical and non-canonical Wnt pathways regulate the anterior neuroectoderm patterning along the primary axis, and TGF-ß signals control the patterning of the neuroectoderm along the secondary axis. In addition, we showed that the crosstalk between the Wnt and TGF-ß pathways was of importance in regulating the neuroectoderm patterning. As the default cell fate in some deuterostome embryos, including embryonic stem cells, is neurogenic, our findings could be widespread mechanisms to coordinate the remaining and/or suppressing developmental programs along different embryonic axes because Nodal and Wnt signals are critical in establishing early developmental polarities in many embryos.

Abstract Introduction Results Discussion Materials and Methods

invertebrates neuronal differentiation neuroscience animals cell differentiation blastulas animal models developmental biology model organisms embryos echinoderms research and analysis methods ectoderm embryology animal cells signal transduction cellular neuroscience neuronal morphology cell biology neurons biology and life sciences cellular types wnt signaling cascade sea urchins cell signaling organisms signaling cascades

2016

Cooperative Wnt-Nodal Signals Regulate the Patterning of Anterior Neuroectoderm

The conundrum of cooperation has received increasing attention during the last decade. In this quest, the role of altruistic punishment has been identified as a mechanism promoting cooperation. Here we investigate the role of altruistic punishment on the emergence and maintenance of cooperation in structured populations exhibiting connectivity patterns recently identified as key elements of social networks. We do so in the framework of Evolutionary Game Theory, employing the Prisoner' s Dilemma and the Stag-Hunt metaphors to model the conflict between individual and collective interests regarding cooperation. We find that the impact of altruistic punishment strongly depends on the ratio q/p between the cost of punishing a defecting partner (q) and the actual punishment incurred by the partner (p). We show that whenever q/p<1, altruistic punishment turns out to be detrimental for cooperation for a wide range of payoff parameters, when compared to the scenario without punishment. The results imply that while locally, the introduction of peer punishment may seem to reduce the chances of free-riding, realistic population structure may drive the population towards the opposite scenario. Hence, structured populations effectively reduce the expected beneficial contribution of punishment to the emergence of cooperation which, if not carefully dosed, may in fact hinder the chances of widespread cooperation. Cooperation, understood as an action which incurs a cost c to the individual that performs it, inducing a benefit b>c to the recipient of that action, is ubiquitous at all levels of biological complexity (i. e. from bacteria to primates) [1]–[3]. However, cooperation requires the existence of an additional mechanism which, at par with it, leads to its evolutionary viability. Up to now, the different mechanisms which were found to pave the way for the emergence of cooperation are inherently “additive”, in the sense that two mechanisms, when acting together, enhance the viability of cooperation to emerge, compared to the effect accruing to each mechanism alone [4], [5]. In all cases, what is at stake is the paradoxical collision between individual and population goals. Evolutionary Game Theory (EGT) [6]–[8] provides an excellent mathematical framework to deal with this challenge and study the evolution of different behaviors in populations. Two popular metaphors to investigate the emergence and maintenance of cooperation under this framework are the Prisoner' s Dilemma (PD, widely employed in biology, and applied to many non-human species) and the Stag-Hunt Dilemma (SH, very popular in connection with the social contract and other human affairs) [9]–[16]. In particular, the PD constitutes the de facto prototype metaphor for studies of cooperation. From a game theoretical point of view a rational individual in a two-person one-shot PD engagement is always better off by not cooperating (defecting), while in real life one often observes the opposite, to a significant extent. Popular mechanisms that aim at solving this evolutionary conundrum such as kin selection [17], direct reciprocity [13], [18], voluntary participation [19], [20], reputation [21]–[24], social structure [25]–[29], peer and pool punishment [30]–[40], etc, are able to promote cooperation by transforming a PD into a SH [4], [16], [41], [42]. From a sociological perspective, the SH portrays a milder dilemma when compared to the PD, since it strips temptation from the latter, leaving only fear in the way between individual and collective interest [43], [44]. Recently, altruistic punishment (which occurs when one individual accepts to pay a cost to impose a higher loss to a peer) was proposed as an efficient mechanism promoting cooperation, based on laboratory experiments showing also that individuals embedded in different contexts punish quantitatively in different ways [34], [45]. Whenever Humans are at stake, one often observes that several mechanisms found to promote, each on its own, the emergence of cooperation, are active simultaneously. Indeed, kin often favor each-other, even in situations in which encounters are repeated, reputation is important and individuals interact and change their minds embedded in population structures well-described by complex social networks. In this context, punishment is no exception. It is thus important to investigate the impact of altruistic punishment in population environments which are structurally more realistic [46]. Here we explore the evolutionary consequences of altruistic punishment in heterogeneously structured populations for a wide range of the PD and SH game parameters (see Model section). We adopt the scale-free paradigm [47]–[49] to describe population structure, as it incorporates features which have been found recurrently in many network structures: heterogeneity, in the sense that different individuals, here associated with different nodes of the complex network, may have different number of neighbors, defined by the bi-directional links emerging from them; moreover, the degree distribution, that is, the probability that a given individual has k neighbors, follows a power-law dependence. These structures generate, in the population, an asymmetric distribution of wealth and influence [50]–[52] both of which greatly enhance the evolutionary chances for cooperation [28], [46], [53]–[56]. Indeed, in such structures, a few individuals (the hubs) are able to interact with a larger number individuals than the vast majority of the population, somewhat embedded in the spirit of the Pareto principle [57]. In the following we shall study the evolutionary dynamics of structured populations, assessing the role of altruistic punishment in comparison with the corresponding results of the model in which punishment is absent. Individuals engage in one-shot games with their first neighbors along the links of a scale-free network (see below) and acquire a fitness associated with the payoff accumulated from all their interactions. Each individual plays unconditionally either as cooperator or a defector. Hence, depending on the strategy pairs, there are four possible outcomes: mutual cooperation yields the reward (R), whereas mutual defection results in the punishment (P) for both individuals. A cooperative player facing a defector gets the sucker' s payoff (S<P) whereas the defector earns the temptation (T). Following usual practice [25], [28], [44], we set R = 1 and P = 0 reducing the number of free game payoff parameters to two. Hence, whenever T>R = 1 we obtain a PD (T>R>P>S), whereas T<1 gives rise to a SH (R>T>P>S). Hence under the PD rational players are driven into defection both by the temptation to cheat (T>R) and by the fear from being cheated (P>S), despite the fact that mutual cooperation (R = 1) offers a better collective outcome compared to mutual defection (P = 0) [11], [43]. Under the SH, the tendency to defect derives solely from the fear of being cheated [16], [43], [58]. In our model setup, cooperators have the option to punish defectors by means of peer punishment, that is, a ‘punisher’ pays the cost q to induce the punishment p on the opposing defector. To keep the analysis simple, we only consider two strategies, punishing cooperators (C) and defectors (D). In this case, the payoff matrix takes the following form: During the evolutionary process, players can adopt the strategy of their neighbors with a probability depending on the payoff difference. In each elementary step, a player x is chosen randomly from the population; a second individual y is selected at random from the neighborhood of x; player x adopts player y' s strategy according to the pairwise comparison rule [59]–[61], which ascribes the probability to this process, where Px and Py are the accumulated payoffs of player x and y, and β represents the intensity of selection (or alternatively, it measures the errors in decision making and the uncertainty of the strategy adoption process): For high β (strong selection) strategies with higher payoff are most likely imitated, whereas for lower β values (weak selection), the influence of payoff decreases. No mutations are considered. Scale-free networks are built according to the Barabási-Albert model of growth and preferential attachment. We generated 102 scale-free networks [47] with 103 nodes each and average degree of 4. We computed the average final fraction of cooperators (xffc) by averaging the final fraction of cooperators (1 or 0 as the evolution already reached fixation) over a total of 2. 5×104 simulations, each starting from an equal fraction of Cs and Ds randomly distributed in the network. We took the value β = 0. 25 for the intensity of selection, a value that optimizes the cooperation levels in scale-free networks in the absence of punishment [62]. This value does not correspond to the weak selection limit which we discuss in the following section. We first examine what happens in the absence of punishment (p = q = 0), which leaves the network structure as the only mechanism promoting the emergence of cooperation. Figure 1 shows the average final fraction of cooperators on the T-S plane in the region associated with the SH and PD domains (0<T<2, −1<S<0). We quantify the overall impact of each mechanism in the evolutionary dynamics of cooperation by defining an area-wide cooperation-index Ω as the fraction of the area of the T-S plane in which xffc>0. 5. As the decline of the distribution function describing the level of cooperation (displayed in Figure 1) is sharp and the function peaks at 1, the index gives a good measure for the scale of cooperation on average for the payoff parameter region under study. With this definition we obtain Ω = 1. 0 (Ω = 0. 0) for overall cooperator (defector) dominance on the whole T-S plane, while for the classical result of evolutionary game theory in well-mixed populations we obtain Ω = 0. 25 (corresponding to half of the SH area in the T-S plane). Figure 1 shows the evolutionary outcome on heterogeneous scale-free networks which lead to Ω = 0. 49, a significant increase of overall cooperation, corroborating previous works [44], [46]. The dashed line shows the threshold where cooperation crosses the 50% mark. In this setting the evolutionary dynamics is mainly hub-driven, given the feasibility of hubs to accumulate a very large fitness. In particular, defector-hubs, which may initially accumulate a high fitness, see their own income decrease in time as they become frequently imitated by their neighbors, leading to a rapid increase of mutual defections in their neighborhood. This dynamics is very different from the reinforcing dynamics induced by a successful cooperator located in a hub, who converts the neighbors to cooperators thereby forming a supporting cooperative cluster [28], [63]. The introduction of altruistic punishment induces a shift in the non-diagonal entries of the payoff matrix. This means that the outcome of evolutionary scenarios with punishment can be mapped onto scenarios without punishment for different values of T and S. Given that the entries are transformed as: T→T – p, S→S – q, punishment amounts to introduce a translation in the T-S plane defined by the vector with coordinates (p, q). The analysis of the slope σ of the edge-curve λ defined in Figure 1 can give us information about the non-trivial correspondence between the translation and the change of Ω. The slope function σ is bounded both from above and from below (0. 31 = σ1<σ<σ2 = 0. 77, see Figure 2), which means there are (p, q) values (q/p<σ1) for which punishment acts advantageously for cooperation in the whole T-S plane, but at the same time, still within the altruistic punishment region, there exist (p, q) values (σ2<q/p<1) for which cooperation is clearly set back. As the slope changes along the line, as shown explicitly in Figure 2, for intermediate punishment values the translation can influence the measure of cooperation differently at distinct points of the T-S plane. The slope provides, at any point, information only about the direction of the translation vector; however, its length is also relevant, in particular in the intermediate region referred to above. Indeed, in this region altruistic punishment can tip the balance and change the winning strategy depending on the location in the T-S plane. Figure 3 depicts the change in the evolutionary outcome of cooperation for the three different scenarios identified above, showing that the additional costs of punishment can do more harm (blue areas in Figure 3) than good (red areas in Figure 3) to overall cooperation while in some cases the outcome is mixed. Although punishment contributes to reduce sizably the fitness of defectors, at the same time cooperators are burdened by the cost of inflicting this effect on their defective partners. This is especially true for hub players as they can be overburdened by the cost of punishing a huge number of defecting neighbors, which may result in a less cooperative outcome than without punishment. Eventually, the joint effect of two mechanisms that, each alone, help softening the social dilemma and promote cooperation, can interfere destructively and inhibit cooperation. This said, it is clear that for any fixed punishment value, there will exist a cost for which cooperation is enhanced. That is, if the cost of punishing the defecting individuals can be decreased, then the introduction of punishment may be a viable way to promote cooperation in network structured populations. Given the analysis above, however, not all combinations of cost-punishment will lead to a positive outcome. The principle can be summed up as: Punish, but not at all costs. Figure 4 shows Ω for a wide range of p and q values. It can be seen that the regions with enhanced and diminished cooperation are clearly separated. The separation curve can be approximated very well by a straight line with slope q/p = 0. 54. Qualitatively, this value can be considered as the average of the σ function displayed in Figure 2. Comparing the area of enhanced cooperation to that of diminished cooperation (in the parameter range p>q) of altruistic punishment), we observe that the introduction of altruistic punishment can decrease the overall cooperation level in a wide parameter region. It is worth noting that, in the limit of weak selection (), the network structure plays a minor role [62] in the overall evolutionary dynamics. As a result, the separation curve pictured as a solid line in Figure 4 becomes unambiguously straight (with a slope of 1), that is, for any p>q, Ω increases, in agreement with the analytical results obtained in ref. [64]. Naturally, the simple model proposed here does not provide an exhaustive analysis of the fate of altruistic punishment in structured populations. Important issues such as the role of anti-social punishment [65], [66] or the central issue of second-order free-riding [11] remain absent from our 2-strategy analysis. Concerning the latter, however, we have checked the evolutionary dynamics whenever individuals are allowed to choose between three strategies – cooperator (but not punisher), defector and punishing cooperator. As expected, the evolutionary dynamics becomes more complex in this case but the main results remain valid. The simulations are started from an initial state where all three strategies are equally represented in the population. Evolution always ends in a monomorphic state. Cooperators and punishers are neutral towards each other but even after the extinction of defectors, evolution is not governed by random drift; Hubs dictate the most likely evolutionary outcome of the population [67]. Cooperators and punishers can be considered as cooperative strategies in this more general setting. They both contribute to cooperation dominance in essentially 50% of the simulations. The identity of the winning strategy depends sensitively on the initial conditions, more specifically on the initial strategy of the hub-players. Regarding the average strategy distribution at the end of the evolutionary process, one can interpret it as the “superposition” of scenarios with and without punishment. In other words, the shift of the edge-curve λ in the case of 3 strategies is about half of what would be obtained in a scenario of defectors and punishers only, for the same parameter values. Overall, the three-strategy scenario exhibits qualitatively the same features as the two-strategy case analyzed in greater detail here. To conclude, we study the impact of altruistic punishment in a population of individuals engaging in social dilemmas of cooperation where individuals can interact with each other alongside a structure described by a scale-free network. We find that depending on the q/p ratio between the cost to induce punishment and the actual extent of punishment, altruistic punishment can either enhance or inhibit cooperation. Mechanisms – such as structured populations and altruistic punishment – which separately promote cooperation, can have overall detrimental effects when applied together. This means that the introduction of punishment is not an easy question. The key to the success of punishment is to minimize the costs to be inflicted on those who engage in punishment. Indeed, only for low values of the q/p ratio will punishment effectively promote cooperation in networked populations. While from a well-mixed perspective punishment may seem a viable route towards cooperation [34], [35], [38], heterogeneous structured populations often narrow such pathway. In fact, and similar to what has been shown in the context of indirect reciprocity [68], the viability of punishment may be limited, such that it can be even easier to achieve cooperation in the absence of punishers whenever individuals interact in a realistic interaction setting.

Altruistic punishment — when a cooperative individual pays a cost to punish her defective partner — has been described as one of the mechanisms that help to explain cooperation' s ubiquity in nature. Here, we investigate a model population where individuals interact with each other along the links of a network. The network is built so that it contains the relevant features of real social and biological interaction webs. Individuals engage in cooperation dilemmas with each other and have the possibility to punish defective partners in order to enforce higher cooperation levels. However, it turns out that the introduction of altruistic punishment not always promotes cooperation – in fact, it can actually hinder the spread of cooperation in a variety of cases that we are able to characterize. Effects acting at “micro”, individual level, such as softening the dilemma and reducing the pressure originating from the fear from being cheated and/or the temptation to cheat, can result in lower overall cooperation at a “macro”, population-wide level, due to the complex interference of the social dilemma and the heterogeneous interaction network.

Abstract Introduction Model Results/Discussion

computer science stochastic processes computer modeling mathematics evolutionary biology population modeling evolutionary modeling biology computational biology evolutionary processes probability theory

2013

Reward from Punishment Does Not Emerge at All Costs

Fascioliasis has been sporadically associated with chronic liver disease on previous studies. In order to describe the current evidence, we carried out a systematic review to assess the association between fascioliasis with liver fibrosis, cirrhosis and cancer. A systematic search of electronic databases (PubMed, LILACS, Scopus, Embase, Cochrane, and Scielo) was conducted from June to July 2015 and yielded 1,557 published studies. Among 21 studies that met inclusion and exclusion criteria, 12 studies explored the association of F. hepatica with liver fibrosis, 4 with liver cirrhosis, and 5 with cancer. Globally these studies suggested the ability of F. hepatica to promote liver fibrosis and cirrhosis. The role of F. hepatica in cancer is unknown. Given the heterogeneity of the studies, a meta-analysis could not be performed. Future high-quality studies are needed to determine the role of F. hepatica on the development of liver fibrosis, liver cirrhosis, and cancer in humans. Food-borne trematodiases are an emerging public health problem in Southeast Asia and Latin America, and are caused by the following flukes: Clonorchis sinensis, Fasciola gigantica, Fasciola hepatica, Opisthorchis felineus, Opisthorchis viverrini, and Paragonimus spp [1]. Globally, it has been estimated that approximately 56 million people are infected by these parasites [2]. According to the International Agency for Research on Cancer, two of these parasites (O. viverrini and C. sinensis) have been recognized as definitive causes of cancer [3]. However, fascioliasis caused by F. hepatica or F. gigantica has not been clearly associated with cancer to date. Fascioliasis, as a neglected tropical disease, commonly affects poor people from developing countries [4]. It has been estimated that at least 2. 6 million people are infected with fascioliasis worldwide [2]. When a combination of serological and parasitological high-sensitive tools are performed in endemic areas, almost one-third of the population have been reported to be affected by this liver fluke [5,6]. Even though most patients are asymptomatic, symptoms may be related to the acute infection (fever and abdominal pain) or to the chronic infection (biliary colic, cholecystitis and cholangitis) [7]. However, there is a paucity of studies that evaluate the natural history of subjects infected with fascioliasis (chronic inflammation, liver fibrosis stages, and carcinogenesis) and in those who were treated (post-infectious liver damage). Therefore, the long-term of fascioliasis are unknown. Additionally, several studies reported an association between fascioliasis with other hepatic complications such as liver fibrosis, cirrhosis, and possibly also with cancer [8]. Due to these gaps in current knowledge regarding the natural history of fascioliasis, the aim of this study was to systematically review the literature to assess the role of F. hepatica in liver fibrosis, liver cirrhosis, and cancer. One of the authors (CM) designed and conducted the electronic search. We searched electronic databases to identify relevant studies (PubMed, LILACS, Scopus, Embase, Cochrane, and Scielo) from their inception through July 2015. The electronic search strategy was as follows: parasite (Fasciola hepatica) AND associated conditions (liver fibrosis, cirrhosis, tumor, cancer, neoplasia, malignancy, hepatocellular carcinoma, cholangiocarcinoma) [MeSH] AND associated terms (oncogene). The search term was adapted to the predominate language of the database. To identify additional candidate studies, we reviewed the reference lists of the eligible primary studies, narrative reviews, and systematic reviews. The search was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [9]. During the screening process, two reviewers working independently (CM, LAM) considered the potential eligibility of each of the abstracts and titles that resulted from executing the search strategy. We considered papers available in the following languages: English, Spanish, Italian, French, German, Turkish, Korean, Chinese, and Japanese. Those eligible publications were related to a possible direct relationship between fibrosis, cirrhosis, or cancer of the liver and presence of F. hepatica. During the elegibility process, all eligible studies based on their abstracts were then reviewed in full text versions. After reviewing the studies in full detail, we divided them into two groups: group 1, relevant and group 2, irrelevant. Relevant studies were those related to a possible direct relationship between the exposure and the disease. Irrelevant publications did not show a direct effect between the exposure and the disease. For example, a direct effect was when a parasite was found in liver fibrosis or cancer tissue, whereas an indirect effect (irrelevant) was if the parasite was mimicking a tumor without malignancy cells in the tissue. Any disagreement was resolved after consensus among all authors. For clinical studies, articles were selected based on the evidence (pathological assessment and/or images) of Fasciola on fibrosis, cirrhosis or cancer of the liver. For basic research studies, articles were selected considering any evidence of fascioliasis and genetic alteration events, either in vivo or in vitro. Those that met the inclusion criteria were included in the final analysis and discussion (included studies, Fig 1). Our analysis included those publications reporting fibrosis, cirrhosis as well as malignancy and identification of Fasciola either with direct or not relationship with fibrosis or tumor. Accordingly, any case report of liver fibrosis or malignancy was included in analysis only if a) shows a direct relationship with fascioliasis, b) occurs as a consequence of previous fascioliasis, or c) identification of the parasite in the tumor. Any reference meeting one of such criteria, whether national or international, was recorded, regardless of article type or quality. Data were then entered in a database covering: title, principal author, year of publication, type of study, method, results, and any additional comments. Details about the eligible studies are shown in S1 Table. Quality of studies was not used as a criterion to select or deselect the studies. Given that the primary outcome of interest was only to assess any relationship between Fasciola and the occurrence of liver fibrosis, cirrhosis or cancer; we did not explore the possibility of publication bias. Our search strategy allowed us to identify 1,557 papers, of which 1,461 were excluded by title, abstract evaluation, and duplication. Duplicate entries were identified by considering the author, the year of publication, the title of the article, and the volume, issue and page numbers of the source. In questionable cases, the abstract texts were compared. As a result, 96 studies were initially screened by reviewing the corresponding full-text papers. Then, 39 records were excluded due to lack of consistent evidence related to cancer, fibrosis or cirrhosis. Thus, 57 records were then assessed for eligibility according to the criteria outlined previously. Finally, 21 were eligible for our analysis. Type of source included 100% journal manuscripts. The included studies in the final analysis were divided into three categories based on the tissue effect gathered by F. hepatica exposure. Those three categories of publications were: (i) those publications related to the presence of F. hepatica and development of fibrosis; (ii) those publications showing data related to F. hepatica and consequent cirrhosis; (iii) those publications containing evidence of presence of F. hepatica and diagnosis of any cancer related to liver or biliary system. Based on such classification, we obtained a total of 1227,149 and 131 studies in the first, second and third categories, respectively (Fig 1). As mentioned above, 21 full-text papers were eligible for final analysis. Five were selected as related to F. hepatica and cancer, 12 were related to F. hepatica and fibrosis, and 4 were related to F. hepatica and cirrhosis. Information describing the studies analyzed are summarized in S1 Table, including author, publication year, study design, country, outcome, and evidence. We identified a total of 2 case reports and 2 case series (19. 1%), 8 in vitro studies (. 38. 1%), and 9 animal assay studies (42. 8%). The publication year ranged from 1974–2015. Geographically, 9 studies were conducted in Europe (42. 9%), 9 in the Americas (42. 9%), 2 in Asia (9. 5%), and 1 in Oceania (4. 8%). A total of 12 studies that reported liver fibrosis caused by F. hepatica infection (6 in animals, 4 in vitro studies, one human case series, and one human case report) were selected for this systematic review. A total of 3234 animals have been reported to have concomitant liver fibrosis and fascioliasis including bovine, sheep, calves, and pigs [10–21]. The first report of Fasciola and liver fibrosis was in 1977. Sheep infected with F. hepatica had fibrous tissue surrounding the hepatic lobules [12]. A higher number of parasites have a direct relationship with the degree of liver fibrosis in cattle (n = 10) [13]. In 3021 pigs, fascioliasis causes serious hepatic lesions mainly characterized by severe fibrosis [14]. In addition, periportal fibrosis and collagen deposition around the bile ducts has been also reported in cattle [15,16]. The more chronic injury around the bile ducts, the greater the periportal fibrosis. In an experimental animal model, the presence of collagen fibers around the bile ducts and cirrhosis with necrotic foci were found at 7 and 10 weeks post-infection with F. hepatica, respectively [17]; but the fibrosis may be reversible after effective anti-parasitic therapy. Whether liver fibrosis is entirely caused by Fascioliasis or by the host immune response is an open question not yet elucidated. Among in vitro studies, some identified the gene expression patterns of animals infected with F. hepatica [18,19]. Up-regulation of fibrosis-related genes in F. hepatica-infected rats including collagen I, alpha-smooth muscle-actin, platelet-derived growth factor beta receptor, tissue inhibitor of metalloproteinase II, and activated human stellated cells (HSCs), have been reported [18]. This suggests that this parasitic infection may be associated with the development of liver fibrosis by activating the HSCs, similar to other infections (i. e. hepatitis C viral infection). In infected sheep, microscopic analysis of the livers showed massive infiltration of inflammatory cells and deposition of collagen at 8-week post-infection [19]. In addition, the authors reported up-regulation of genes associated with fibrosis (including genes in the JAK-STAT pathway), tissue repair, remodeling and regeneration (including TNF-α, TGF-β, calponins, transgelins, osteopontin and adora2b) [19]. Inoculation of native GST of F. hepatica in goats caused portal fibrosis, inflammatory infiltration with plasma cells, formation of lymphoid follicles, accumulation of haemosiderin-laden macrophages and granulomatous foci [20]. Natural killer cells have been also found in infected rats around the portal space, centrilobular veins, periportal fibrosis areas and around collagen [21]. In human studies, one human case with severe fascioliasis was reported to have fibrosis of portal tracks with fibrosis extending into the parenchyma after a liver biopsy [10]. The other study included 87 patients with fascioliasis and aimed to characterize by imaging the long-term liver damage after effective anti-parasitary treatment. This showed that 9 patients continued having fibrotic liver lesions after 1 year of treatment, but neither cirrhosis nor cancer was documented on any of these patients during a mean follow-up period of 62 months [11]. Four records were selected that reported cirrhosis as a concomitant condition of fascioliasis, including 3 in vivo studies and 1 case report. Two additional studies reported sequential progression from liver fibrosis to cirrhosis in animals infected with F. hepatica [13,17]. Liver cirrhosis has been reported in wild animal models including cows, goats and alpacas [13,22,24]. Cirrhosis in fascioliasis has been also reported in experimental animal models [17,23]. The liver damage reported in animals infected -for at least 6 months (chronic infection) - were described as fibrotic nodules (stage IV of liver fibrosis or cirrhosis) in most of the liver with several degrees of inflammation [22,24]. In 25 goats infected by Fasciola at least for 3 months, the authors reported loss of the lobular pattern, proliferation of bile ducts and fibrosis in portal areas consistent with cirrhosis after examining both lobes of the liver [22]. In another study of 40 infected rats, a pool of their bile ducts was collected and it was found that collagen I and III were significantly increased when compared to controls [23]. This was found similar to what occurs in cirrhosis. In humans, one case has been reported with cirrhosis [25]. This was a 42-year-old American woman with fascioliasis who had an ERCP for biliary cirrhosis and cholangiogram suggesting sclerosing cholangitis [25]. The authors stated that the most direct cause for the primary biliary cirrhosis and other biliary complications on this patient was likely to be all caused by F. hepatica. A total of 5 studies were included in our systematic review. Four were in vitro animal studies, and one was an animal case series. We did not find reports of human cases of cancer explained by Fasciola. Chung (2012) investigated the role of TGF-β and IL-4 in the immunosuppression as a hypothetical mechanism of parasite evasion of host immune system [26]. The study demonstrated that TGF-β and IL-4 are up-regulated as a consequence of F. hepatica infection, TGF-β reaches its maximum levels of serum at week 2 post infection in each mouse [26]. TGF-β is a potent known proliferation factor that can also directly inhibit activation of immune system [27]. Both increased proliferation by growth factors and immune evasion are cancer hallmarks [28]. Some studies conducted by Motorna (2011) and Gentile (1998) used the lambda/lacI Big Blue transgenic mouse model to investigate if genetic damage, as a measure of lacI mutations, could result in liver tissue from infection by F. hepatica [29,30]. There was an increase of lacI mutations in mice with fascioliasis suggesting that the infection increases the risk for complex hepatic cell mutations rather than mutations stemming from more definable oxygen radical-associated events [31,32]. An additional publication reported an indirect relationship between F. hepatica and cancer by induction of CYP2A5 enzyme (from the parasitic infection) which participates in the metabolism of carcinogens like B1 (AFB1) and several nitrosamines [32]. The results of this study suggested that F. hepatica can alter the activity of key hepatic enzymes, which may contribute to accumulation or decrease clearance of carcinogenic compounds found in food products or environmentally [32]. We found one publication that reported hepatocellular carcinoma (HCC) in cattle with fascioliasis [33]. Our results show a lack of findings/evidence of fascioliasis and cancer in population-based studies; quantitative data to measure association with liver disease (Odds Ratio, Relative Risk); publication bias; quantify significance with a funnel plot or Egger´s regression asymmetry test; long-term follow-up of infected cases to asses for further liver damage, and studies in human and animals. As fascioliasis causes chronic infection in the liver, there is a need to elucidate the long-term clinical complications of this parasitic infection in humans. This systematic review summarizes the current evidence that may associate human fascioliasis with liver fibrosis, cirrhosis, and perhaps cancer. We showed that Fasciola plays an important role in the development of liver fibrosis, and cirrhosis in animal models as shown in Fig 2 [10–21,34]. The mechanism of this association may be due to the activation of HSCs by the cathepsins of the parasite [18]. The intensity of infection may play a role on the development of liver fibrosis during the infection. In addition, there is weak evidence in the role of Fasciola in liver fibrosis in humans. This is limited to case reports. The impact of this association has not been well established in populations highly prevalent with Fasciola. It will be important to identify at a population level whether patients with Fasciola are more prompted to develop liver fibrosis and cirrhosis compared to non-infected patients, after controlling for alcohol consumption and viral etiologies. Furthermore, the role of early detection, and early treatment of acute or chronic human fascioliasis has not been studied yet as a strategy to prevent development of fibrosis overtime. To further justify the plausibility that F. hepatica is associated with fibrosis, there are reports of fibrosis triggered by related organisms [34,35]. Livers of 35 fallow deer with fascioliasis, caused by the related parasite Fascioloides magna, were found to contain proliferation of the epithelium of bile ducts (biliary proliferation, an early stage for biliary cirrhosis) which were framed with a large quantity of connective tissue [35]. In that study, myofibroblasts especially HSCs were determined to play an important role in the synthesis of extracellular matrix components in the development of parasitic fibrosis and cirrhosis in the liver of these animals [35]. Similarly, fibrosis has been reported in 15 cows as a consequence of fascioliasis caused by F. gigantica with presence of proliferative and hyperplasic bile ducts [34]. In carcinogenesis, the evidence of the association of Fasciola infection and cancer is very limited and no conclusions can be unequivocally reached based on our findings. Fasciola infection in animal models has been demonstrated to overexpress a proliferative factor such as TGF-β, increase mutations (lacI) in mice, and induce CYP2A5 isoenzyme. The later may result in a reduction on the metabolism of carcinogenic agents. Our results remain informative since no animal (except for ref. 33) or human studies have shown a case of cancer and Fasciola. Furthermore, the evidence of Fasciola-related cancer in animals or humans has not been documented but the reasons for this are beyond of the aim from this study. We can speculate that as most endemic areas of fascioliasis are located in resource-poor settings, where the access to healthcare centers is limited, the chronic consequences from the infection by Fasciola are likely underreported and therefore, unknown. For instance, there has been a study to suggest an early presentation of liver cancer in young adults in Peru, but no etiology has been identified [36]. A recent study on HCC patients in Peru showed that all four K-RAS-mutated HCCs were unusual I21M mutants, uncommon K-RAS mutations different from codon 12 mutations have been associated with cholangiocarcinoma produced by viral infections or fluke infestations [37,38]. To the best of our knowledge, there has not been any association between liver cancer in Peru and fascioliasis to date. F. hepatica is able to induce DNA damage through action of mutational mediators such as reactive nitric species and reactive oxygen species [39]. Both cirrhosis and genomic instability combined to a protumorigenic environment caused by CYP2A5 mutated contribute to cell transformation. The results from this systematic review are interesting for several reasons. Most of the people chronically infected by Fasciola are asymptomatic but they are not necessarily free of inflammation [7]. There is a degree of inflammation into the liver in those chronically infected individuals by means of an increase in serum lipid peroxidation and a decrease in antioxidant enzymes, but the period between initial infection and developing of liver fibrosis in humans is still unknown [40]. Intensity of infection, length of infection, re-infections, other liver diseases, alcohol consumption, co-infection with chronic viral hepatitis, among others; are factors to be considered when assessing liver disease in infected individuals by Fasciola in endemic areas. For example, alcohol consumption can exacerbate cholangiofibrosis in hamsters infected by Opisthorchis, another liver fluke infecting bile ducts, but no studies in Fasciola have been performed [41]. One also might think that the inflammation from the infection would resolve after effective antiparasitic therapy but this has not yet been assessed. Future longitudinal studies in human from endemic areas may investigate further our findings. In addition, future studies may face the challenges of Fasciola resistant to triclabendazole, the only drug available for Fasciola nowadays [42]. The major limitation of our study is the absence of previously published population studies that assessed the role of F. hepatica in liver fibrosis, cirrhosis, and cancer. Therefore, our results come from basic science studies, animal models, case reports, and case series. However, no studies have previously systematically reviewed the literature in this important topic, and our study serves to suggest an association between F. hepatica with liver fibrosis and cirrhosis. Other limitations include publication bias, and lack of longitudinal follow-up of infected patients. Despite these limitations, we believe that our study makes an important contribution to recognize several potential severe chronic complications associated with human fascioliasis, and will be the base of future population studies that assess these associations. The results of our study and future studies will be of use for vulnerable populations affected by this fluke, in areas like the Peruvian Highlands, to prevent the complications caused by fascioliasis. Furthermore, it is relevant to investigate the existence of an inter-relationships between F. hepatica and gastrointestinal tract microbes that may affect the progression of fascioliasis. For instance, Helicobacter pylori infection has been closely associated with O. viverrini-associated cholangiocarcinoma suggesting that the liver fluke is a reservoir of the carcinogenic bacterium and thus making plausible that the co-infection may promote the pathogenesis of cholangiocarcinoma [43,44]. Additionally, studies of the biliary microbiota should reveal the role of host microbial content in the development of fascioliasis and whether alterations or changes in microbiota occur as a consequence of the liver damage. Most recently, compositional shifts in the tissue microbiome of O. viverrini-associated cholangiocarcinoma were identified suggesting that changes in the microenvironment occurred after parasite infection can trigger tumorigenesis [45]. Also, comparison of similarities and dissimilarities with the pathophysiological processes leading to liver cancer and cholangiocarcinoma, induced by infection with O. viverrini and C. sinensis, and including well-characterized liver fluke derived metabolites likely will advance this task [46,47]. We conclude that there is some evidence of an association between Fasciola infection with liver fibrosis and cirrhosis but no strong evidence between Fasciola and cancer. There is a need of long-term population studies to assess the association of F. hepatica with liver fibrosis, cirrhosis and cancer in endemic populations.

Fascioliasis is a neglected infectious disease caused by the trematode Fasciola. This parasite (liver fluke) is endemic in many parts of the world including countries from Asia, Africa, Europe and the Americas. High prevalence rates of fascioliasis has been repeatedly reported in the Andean region of South America. Liver complications in infected humans by this fluke have been sporadically reported in the literature. For instance, the relationship between F. hepatica infection and liver fibrosis has been suggested but its association with cancer is unclear. In this study, we found some evidence of an association between Fasciola infection with liver fibrosis and cirrhosis, but little between Fasciola and cancer. As Fasciola is highly endemic in some regions of the developing world, our study shed light on the complications of this parasitic infection which are not that different from flukes such as Schistosoma or Opisthorchis. We believe that further investigations are needed in order to elucidate the pathways in how F. hepatica infection causes liver damage.

Abstract Introduction Materials and Methods Results Discussion

invertebrates medicine and health sciences helminths tropical diseases fibrosis fascioliasis database searching parasitic diseases animals liver diseases trematodes developmental biology gastroenterology and hepatology neglected tropical diseases research and analysis methods liver fibrosis flatworms fasciola research assessment cirrhosis helminth infections database and informatics methods systematic reviews biology and life sciences organisms

2016

Association of Fasciola hepatica Infection with Liver Fibrosis, Cirrhosis, and Cancer: A Systematic Review

No vaccine is currently available for dengue virus (DENV), therefore control programmes usually focus on managing mosquito vector populations. Entomological surveys provide the most common means of characterising vector populations and predicting the risk of local dengue virus transmission. Despite Indonesia being a country strongly affected by DENV, only limited information is available on the local factors affecting DENV transmission and the suitability of available survey methods for assessing risk. We conducted entomological surveys in the Banyumas Regency (Central Java) where dengue cases occur on an annual basis. Four villages were sampled during the dry and rainy seasons: two villages where dengue was endemic, one where dengue cases occurred sporadically and one which was dengue-free. In addition to data for conventional larvae indices, we collected data on pupae indices, and collected adult mosquitoes for species identification in order to determine mosquito species composition and population density. Traditionally used larval indices (House indices, Container indices and Breteau indices) were found to be inadequate as indicators for DENV transmission risk. In contrast, species composition of adult mosquitoes revealed that competent vector species were dominant in dengue endemic and sporadic villages. Our data suggested that the utility of traditional larvae indices, which continue to be used in many dengue endemic countries, should be re-evaluated locally. The results highlight the need for validation of risk indicators and control strategies across DENV affected areas here and perhaps elsewhere in SE Asia. Dengue virus (DENV) is considered to be the most important arbovirus world wide, with a heavy disease burden in humans [1]. It is transmitted mainly by Aedes aegypti mosquitoes, but Ae. albopictus can also act as a vector [2–5]. Dengue is endemic in many countries around the world, especially in the tropics; moreover the number of endemic areas is increasing [6]. DENV belongs to the genus Flavivirus in the family Flaviviridae and consists of four antigenically distinct and medically relevant serotypes, with a possible fifth recently described (DENV1,2, 3,4 and 5) [1,6–9]. The clinical spectrum of DENV infection can vary from asymptomatic to more severe forms such as dengue hemorrhagic fever (DHF) and dengue shock syndrome (DSS) [8]. DENV is transmitted to humans following mosquito bite (horizontal transmission), and mosquitoes can become infected by ingestion of a DENV-containing blood meal [2]. However, DENV can also be maintained via vertical transmission i. e. passed into eggs and subsequently into the next generation of mosquitoes, thereby maintaining outbreaks in human populations. This has already been documented in both Ae. aegypti and Ae. albopictus in different countries including in SE Asia [10–13]. Since it was first identified in 1968 in the cities of Jakarta (capital of Indonesia) and Surabaya (East Java), dengue disease has been recognised as an important public health problem in Indonesia. Periodic outbreaks have occurred in Indonesia with an increasing number of cases and severity [14]. DENV incidence in Indonesia has been shown to peak during the rainy season (between the months of October and April) [15]. From 2004 onwards, Indonesia reported the highest number of DENV cases in the region. All four serotypes of DENV have been found to be circulating since and DENV3 infections associated with the most severe disease [16,17]. Despite dengue being a major concern remarkably little is known or done to control this virus in Indonesia, in spite of its size (in surface and population, as the world’s largest island nation but with high levels of poverty) and important economical position in SE Asia and the world [18]. As there are no vaccines or drugs available for DENV, control programmes for DENV transmission are often focused on managing mosquito populations. To determine the nature of mosquito populations, entomological surveys are usually conducted within routine control programmes [19]. For many years, the standard protocol has relied on traditional sampling (Stegomyia indices) which is based solely on the presence of larvae [20]. Indicators of DENV vector abundance (mainly Ae. aegypti) were based on larval surveys of container habitats and the calculation of various indices, namely House Index (HI: percentage of houses infested with larvae or pupae), Breteau Index (BI: number of larvae or pupae positive containers per 100 houses examined) and Container Index (CI: percentage of water-holding containers found to be infested with larvae or pupae) [20]. These indices can facilitate understanding of vector ecology in a given control area, but also serve as useful measures to determine the success of intervention strategies. However, these traditional sampling methods have shortcomings by measuring only the abundance of larvae and not determining species, which therefore may be poor predictors of the abundance of adult vector mosquitoes that are responsible for transmission [20]. Following consideration of these issues, Focks (2003) suggested that pupal/demographic survey methods were developed to replace the more traditional larval indices [20]. The pupae index is based on counting the number of pupae per container and identifying which container types are responsible for the largest number of adult mosquitoes. It assumes the ability to predict the potential of DENV epidemics more accurately than the traditional HI, which does not necessarily correlate with DENV transmission [21,22]. Despite the limitation of traditional sampling methods, many studies have continued to focus on indices from larval stages of the mosquito, e. g. [23–28]. However, the importance of developing improved and locally appropriate entomological surveys in DENV endemic areas is increasingly recognised [20]. In Indonesia, variations in dengue disease reporting make it important to further understand the entomological differences between areas with different DENV risk i. e. , in endemic areas (defined here as an area that has regularly reported DENV cases in the three years preceding this study, 2009–2011), sporadic areas (defined here as an area which has had an irregular number of DENV cases reported in the three years preceding this study, 2009–2011) and compare these to dengue free areas (defined here as an area with no reports of dengue disease in the three years preceding this study, 2009–2011). Comparisons of mosquito populations and local habits in these different areas are more likely to indicate the key entomological differences that can inform potential points of intervention, and the validity of the various indices and survey methods. In this study we applied traditional larvae indices, the pupae index as well as adult mosquito collections and species identification to compare and enhance the validity of entomological survey results in villages with different dengue endemicity in the Banyumas Regency of Java. By comparing these traditional indices to newer indicators with respect to their ability to predict dengue risk, we aimed to better understand the local dengue transmission processes. Overall our data help to fill important gaps in our knowledge of dengue transmission and associated ecology/human behaviour in this area of SE Asia and inform local prevention strategies. Our observations may be relevant beyond the study area by informing entomological surveys elsewhere. By determining the most representative factors to predict/analyse mosquito populations and transmission risk, entomology surveys can be done in an effective and more efficient manner. The study site used in this analysis is the Banyumas Regency, located in the southwest of Central Java Province, Indonesia (Fig 1). Coordinates for this location are as follows: 108" 39`17”–109" 27`15” East longitude, and 7" 15`05”–7" 37`10” South latitude. The total area is 132,760 km2, with a population of 1. 85 Million inhabitants at a male to female ratio of 50: 50. Banyumas Regency consists of 27 sub districts, and has 39 community health centres and a total of 331 villages. The environment in Java is characterized by a tropical monsoonal climate, with a dry season lasting approximately 6 months and a heavy monsoon the rest of year. Total annual precipitation averages at 1755 mm (69. 1 inches) and there are 2975 hours of sunshine on average per year. In order to determine the differences in mosquito population density between seasons, entomological surveys were carried out twice, once in the dry and once in the rainy season. During the dry season three independent villages were selected on the basis of differing endemicity status and their spread across the region (Fig 1): DENV endemic: Tanjung village, South Purwokerto Community Health Centre and Sokanegara, East Purwokerto Community Health Centre; DENV sporadic: Panusupan village, Cilongok I Community Health Centre; DENV free: Gunung Lurah village, Cilongok II Community Health Centre. The endemicity status criteria are based on “The Technical Manual Eradication of Dengue Mosquito-borne Diseases, Indonesian Ministry of Health” (1992) [29]. The determination of DENV status was made before the survey began, based on reported dengue cases from the Banyumas Regency Health Office. All suspected DENV cases are reported to the health office, however not all cases are confirmed. It is important to note that all reported cases are severe and require hospitalisation. Therefore it is likely there is under reporting of actual cases across the regency. The dry season survey was conducted from May-June 2012, while the rainy season survey was conducted from January-February 2013. One additional village, Sokanegara village (DENV endemic) was added to the survey in the rainy season. In each village, 100 houses were chosen by simple random sampling for the entomological surveys, resulting in a total of 300 houses being analysed in the dry season and 400 houses in the rainy season. The individual locations of the entomological surveys are shown in Fig 1. An overview of environmental conditions and characteristics of the four villages is shown in Table 1. Larvae collection was carried out in every container both inside and outside the participating houses. To measure the entomological parameters the House Index (HI: percentage of houses infested with larvae and/or pupae), Container Index (CI: percentage of water-holding containers infested with larvae or pupae), Breteau Index (BI: number of positive containers per 100 houses inspected), Pupae Index (PI: number of pupae per 100 houses inspected) and Free Larvae Index (FLI: the percentage of houses without larvae) were determined [20]. 100 houses is the sample size recommended by the Indonesian Ministry of Health, in “The Technical Manual Eradication of Dengue Mosquito-borne Diseases, Indonesian Ministry of Health” (1992) [29]. The interpretation of transmission risk levels of each village was made based on the larvae index, as described in the WHO document “A review of entomological sampling methods and indicators for dengue vectors” [20]. The survey also included a description of all containers, both artificial and natural in each participant’s house. Identification of the recovered larvae was based on the key identification criteria as described by Stojanovich and Scott, 1965 [30]. Insect collections were carried out using back-pack aspirators to capture adult mosquitoes in resting and flying positions. Areas inside the house where mosquitoes normally rest were focused on. For example, Ae. aegypti mosquitoes prefer to rest in dark, shielded, humid areas on hanging objects such as clothes and curtains and on walls. Adult mosquito capture was carried out between 8–11 AM for around 20 minutes per house (100 houses per village). Identification of adult mosquitoes was conducted by using key identification criteria as described earlier [30]. Sample sizes of 100 houses were used as recommended. Confidence intervals (C. I.) 95%, t test, chi square test were calculated by using IBM SPSS Statistic 21. Studies conducted here (data collection of mosquito breeding sites, mosquito egg collections) were carried out with ethical approval from the University of Glasgow (Project Number: 2012082) and the Ministry of National Education, Faculty of Medicine Gadjah Mada University, Medical and Health Research Ethics Committee (KE/FK/323/FC). No data involving human participants were collected in this study. Following the field surveys conducted during the dry and rainy seasons, the HI and BI indices in the rainy season were found to be higher than in the dry season: the average HI and BI in all villages in the rainy season were 24 and 31, respectively, higher than in the dry season (15 and 18, respectively). On the other hand, the CI was lower in the rainy season (Table 2) probably because more containers were found in the rainy seasons in all villages. Thus compared to the dry season, more mosquito larvae were found during the rainy season. Panusupan (DENV sporadic) showed the lowest free larvae indices (FLI), and was classed as a high risk level of DENV transmission compared to other villages using this index. To determine whether larvae density correlated with number of DENV cases occurring after the survey, updated information on the number of dengue cases from the Banyumas Regency Health Office was obtained (Table 3). Based on the report, Tanjung and Sokanegara (DENV endemic) which were classified as medium risk level according to the indices above, continued to have more dengue cases in 2012 and 2013 compared to the sporadic and free area. Meanwhile, Panusupan (DENV sporadic, but classed as high risk) reported no dengue cases in 2012 and 2013 and Gunung Lurah (reported as DENV free before 2012 and with a low or medium risk depending on index used) reported one dengue case in 2013. In addition to calculating the various indices as outlined above, species identification of the collected larvae was performed (Fig 2). Numbers of larvae in all villages were higher in the rainy season than in the dry season. Ae. aegypti and Ae. albopictus were the dominant species in Tanjung, Sokanegara (DENV endemic) and Panusupan (DENV sporadic). Culex sp. were identified in low numbers only in Panusupan (DENV sporadic) and Gunung Lurah (DENV free). Next pupae were assessed in each village and the Pupae Indices (PI) used in order to improve the entomology survey. Details of PI (house and container pupae indices) are indicated in Table 4. The DENV endemic and sporadic areas had higher CPI and HPI than the DENV free area (Gunung Lurah) (X2 = 6. 60, df = 1, p-value = 0. 01). This indicated that in the endemic/sporadic areas, mosquitoes tend to have a more conducive environment to survive from eggs to become pupae, and environments with greater survival of mosquitoes to the pupal stage correlated to a higher number of reported dengue cases in endemic areas (Table 3) compared to sporadic and free areas; the high CPI and HPI did not however, correlate with the zero reported cases in Panusupan in the months after the survey. According to Focks (2003) the threshold of dengue transmission is when the pupae/person index (Table 4) ranged between 0. 5–1. 5 with an optimum air temperature 28°C [20]. Containers found in each house were recorded in order to determine what the dominant mosquito breeding containers were in the various study areas. The main finding of the container survey was that more artificial containers were found in the four villages surveyed compared to natural containers (paired t-test, mean 195, SD 310, p = 0. 003) (summarized in Table 5). We found more natural containers in sporadic and free areas (Panusupan and Gunung Lurah) compared to endemic areas (Sokanegara and Tanjung), although there was no significant difference (p = 0. 5) (Table 5). Endemic areas (Tanjung and Sokanegara) are more urbanised (less vegetation, and more densely populated), as described in Table 1. Buckets, water storage containers and traditional bath-tubs were found to be the dominant breeding containers observed in all four villages. In fact, discarded tyres were the containers which had the highest proportion of infestation (53%), this finding is also consistent with other studies [31,32]. Moreover, other artificial containers such as aquariums, water dispensers and flower pots also showed high infestation rates. Measuring adult mosquito numbers is considered to be the most representative quantitative estimate to obtain information about mosquito abundance, as immature stages need to go through several developmental stages to become adult mosquitoes before they can transmit DENV [20]. After identification of the mosquito species in the dry and rainy seasons, the numbers of each species in each area are shown in Fig 3. The dominant adult mosquito species captured (both seasons combined) during the survey in Tanjung (where Culex sp. were dominant only in the rainy season) and Sokanegara (DENV endemic) were Ae. aegypti; in Panusupan (DENV sporadic), non dengue transmitting Culex sp. were identified as the dominant species (12 in the dry season and 144 in the rainy season); although comparable numbers (to endemic areas) of Ae. aegypti were identified. Moreover the combined numbers of Ae. aegypti and Ae. albopictus were higher in Panusupan than in the dengue endemic areas. However, because of its feeding preference, the role of Ae. albopictus has been called into question [3]. This adult collection result is in contrast with the larvae identification, where Ae. albopictus was found to be the dominant species in Panusupan and this is likely due to breeding behavior, as Ae. albopictus (and Ae. aegypti) species frequently breed in containers around housing while culicine mosquitoes use different types of habitats. In Gunung Lurah (DENV free), we captured very few adult mosquitoes, and Ae. albopictus was the dominant species collected while Ae. aegypti was completely absent. Many DENV-endemic countries such as Indonesia, Malaysia and Thailand use entomological surveys as a routine method recommended by WHO to record mosquito populations [19]. Information on mosquito density can then be used in mosquito control efforts and in prevention of DENV transmission [33,34]. Areas with high mosquito populations have usually been treated with larvicides such as organophosphates or temephos in an attempt to prevent outbreaks of DENV. In Indonesia, a Ministry of Health programme encourages community participation in carrying out routine entomology surveys in their homes [35]. Some villages in Indonesia also have trained village health volunteers (VHV) who regularly conduct entomological surveys. Traditional sampling methods i. e. larvae indices were routinely applied over many years to determine mosquito densities in defined areas and the subsequent risk of DENV transmission. However there can be limitations associated with traditional indices [20–22,36]. To assess the validity and usefulness of these methods, and improve the characterisation of vector populations in our study area, we combined the traditional larvae indices together with the pupae index, species identification and adult mosquito collections. Our results suggest that traditional larvae indices might not always be an appropriate way of quantifying mosquito populations and dengue transmission risk, as has been previously reported [36]. From the adult mosquito collections (and subsequent species identification), the high larvae indices in Gunung Lurah village (DENV free area; one recent case likely to have been imported from an area where transmission occured) did not support the transmission of DENV as very few adult mosquitoes were captured in this village. Nonetheless it is important to point out that both Ae. albopictus and Ae. aegypti were present in DENV-sporadic Panusupan and perhaps differences in vectorial capacity come into play locally. Clearly, our data indicated that larvae density was not always in accordance with the number of DENV cases reported in villages. The pupae survey in this study (Table 4), would suggest that the area of study has a low risk for dengue transmission according to factors previously defined by Focks (2003). The high presence of Ae. albopictus larvae in the dengue free area also suggests that presence of vectors alone may not predict transmission; Ae. albopictus in this area may have reduced capacity due to their feeding behaviour etc. or possibly reduced competence for DENV. That very few adult mosquitoes were found in 100 houses in Gunung Lurah might be due to a generally unfavourable environment for mosquitoes. These results suggested that high levels of adult Ae. aegypti in endemic (and sporadic) areas were a potential indicator of DENV transmission risk. These findings were in agreement with the real numbers of DENV cases which occured in Gunung Lurah; one reported case in 2013. Adult mosquito numbers (and species identification) may be a useful estimate to obtain information on dengue disease risk, at least in this part of Indonesia as immature stages need to go through several developmental stages in order to become adult mosquitoes able to transmit DENV [20,36]. However, these methods require specialist skills [37] and are not easily transferable to local surveillance programmes. Moreover, while our observations suggest that the usefulness of several indices should be questioned at local level, we stress that underreporting of dengue cases needs to be taken into consideration in the discussion of our results. Improved patient data collection and dengue diagnostics need to be developed, implemented and combined with future mosquito surveillance work in the Regency to support entomological surveillance studies whose accuracy relies on such data. Our findings may encourage such efforts and lead to a more in depth re-evaluation of the observations reported here. Based on the results of this study, mosquito populations in the regency are higher in the rainy season than in the dry season, for example more mosquito larvae and also adult mosquitoes in three villages were found during the rainy season compared to the dry season. This suggests that health officers and the community should focus their efforts on the beginning of the rainy season. Not surprisingly we also found more potential breeding containers in the rainy season with buckets and water storage containers as predominant water sources in all four villages surveyed. Our findings indicate that villagers can minimize the potential breeding sites for mosquitoes by reducing the presence of artificial containers such as traditional bath-tubs and buckets. It can be assumed that by reducing the number of these containers, DENV incidence could be minimized. The results from adult mosquito captures in the four villages indicated that Ae. aegypti still preferred urban areas (Tanjung and Sokanegara), although in Panusupan (DENV sporadic, rural), Ae. aegypti was also observed although the numbers of Culex sp. mosquitoes in this village were far higher. Ae. albopictus is more likely to be found in rural or suburban areas. These observations are also emphasized by our container survey, where we observed that natural containers were found more frequently in rural areas (Panusupan and Gunung Lurah) and Ae. albopictus is more prevalent than Ae. aegypti. These findings are in accordance with previous reviews on the differences in distribution and ecology between Ae. aegypti and Ae. albopictus which stated that Ae. albopictus prefers natural containers [38]. Species identification is important but rarely applied in the field, and often only for research purposes. Culex sp. mosquitoes were identified as the dominant adult mosquito type in Panusupan village; this is of interest since this species has not been shown to be a vector of DENV. Vazeille and colleagues stated that Ae. aegypti is the most effective vector for dengue viruses and is highly receptive to oral infection; they also demonstrated that Cx. quinquefasciatus can be infected by the parenteral route with DENV type 2 but the virus replicated to very low levels, therefore the authors concluded that Cx. quinquefasciatus should not be considered a biological vector of DENV [39]. A recent study carried out in Taiwan suggested that various vector indices alone were poor DENV outbreak indicators and each country should evaluate its own situation [40]. We agree with this statement, although we emphasize that better diagnostics needs to be implemented as part of any future studies on this subject in Java. The transmission risk by adult mosquitoes can be influenced by a number of factors that affect the extrinsic incubation period (EIP) and arbovirus/ vector interactions. Indeed, virus and vector genetics, but also gut microbiota and host responses are important factors in DENV-vector interactions [41–53]. Moreover, climatic factors such as temperature and humidity come into play. The influence of temperature on EIPs associated with DENV for example, has been analysed, and was shown to be important for EIP duration [54–56]. These, and other risk factors may vary locally, and could also change over time highlighting the importance of local assessments. At least in the case of Banyumas Regency, our findings also suggest that more prevention efforts should be carried out in the beginning of the rainy season to reduce dengue virus transmission, for example by clearing artificial containers. In summary the observations of this study can form the basis of a better understanding of dengue vector ecology in this part of Indonesia.

Geographically and economically, Indonesia is one of the most prominent countries in SE Asia, yet many of its endemic infectious diseases are poorly managed, controlled and understood. This includes dengue virus (DENV), which can result in serious human disease and is transmitted by mosquitoes. Dengue risk assessment is a key factor in managing the impact of infection on public health, and this often relies on assessing the presence and nature of mosquitoes through a number of indices associated with the occurrence of larvae and the location/availability of breeding containers. Here we assessed traditionally used indices in combination with other indicators including pupae indices, and the presence of adult mosquitoes in areas with different dengue status: endemic, sporadic or free. Our data suggested that traditional indices were poor indicators of reported local DENV transmission. This has important consequences for design and focus of risk management strategies and efforts to control DENV locally as well as elsewhere in the region.

Abstract Introduction Methods Results Discussion

invertebrates medicine and health sciences insect metamorphosis geographical locations animals indonesia java age groups developmental biology adults pupae infectious disease control insect vectors zoology infectious diseases epidemiology disease vectors insects arthropoda people and places mosquitoes asia entomology oceania metamorphosis biology and life sciences population groupings larvae organisms

2016

Dengue in Java, Indonesia: Relevance of Mosquito Indices as Risk Predictors

Fusarium fujikuroi causes bakanae (“foolish seedling”) disease of rice which is characterized by hyper-elongation of seedlings resulting from production of gibberellic acids (GAs) by the fungus. This plant pathogen is also known for production of harmful mycotoxins, such as fusarins, fusaric acid, apicidin F and beauvericin. Recently, we generated the first de novo genome sequence of F. fujikuroi strain IMI 58289 combined with extensive transcriptional, epigenetic, proteomic and chemical product analyses. GA production was shown to provide a selective advantage during infection of the preferred host plant rice. Here, we provide genome sequences of eight additional F. fujikuroi isolates from distant geographic regions. The isolates differ in the size of chromosomes, most likely due to variability of subtelomeric regions, the type of asexual spores (microconidia and/or macroconidia), and the number and expression of secondary metabolite gene clusters. Whilst most of the isolates caused the typical bakanae symptoms, one isolate, B14, caused stunting and early withering of infected seedlings. In contrast to the other isolates, B14 produced no GAs but high amounts of fumonisins during infection on rice. Furthermore, it differed from the other isolates by the presence of three additional polyketide synthase (PKS) genes (PKS40, PKS43, PKS51) and the absence of the F. fujikuroi-specific apicidin F (NRPS31) gene cluster. Analysis of additional field isolates confirmed the strong correlation between the pathotype (bakanae or stunting/withering), and the ability to produce either GAs or fumonisins. Deletion of the fumonisin and fusaric acid-specific PKS genes in B14 reduced the stunting/withering symptoms, whereas deletion of the PKS51 gene resulted in elevated symptom development. Phylogenetic analyses revealed two subclades of F. fujikuroi strains according to their pathotype and secondary metabolite profiles. The heterothallic ascomycete Fusarium fujikuroi Nirenberg is a member of the Fusarium fujikuroi species complex (FFC), a monophyletic lineage which includes at least eleven mating populations (MPs A-K) that are sexually infertile with one another, and numerous distinct anamorphic species [1]. F. fujikuroi (MP-C) is the causal agent of the rice disease bakanae (“foolish seedlings”), one of the most notorious seed-borne diseases with increasing economic importance in the major rice-growing countries in the world, including all rice-growing Asian and African countries, California, and more recently, Italy [1–3]. The fungus was one of the first fungal pathogens to be described, and bakanae is one of the oldest known diseases of rice being reported more than 100 years ago by Japanese scientists [4,5]. The most prominent symptoms of the disease are chlorotic, elongated and thin seedlings that are often several inches taller than healthy plants, and empty panicles leading to yield losses ranging from ca. 30–95% [6–9]. However, not all infected seedlings show the bakanae symptoms: sometimes they may be stunted or appear symptomless [10]. The incidence and severity of the bakanae or stunting disease symptoms varies with regions and isolate. The pathogen is dispersed predominantly with infected seeds, infected crop residues from the previous season in the soil, or by conidia on diseased stems which can be transmitted by rain and wind [6]. Disease control has become increasingly difficult due to rapidly developing fungicide resistance in the fungal population [4]. The enormous elongation of infected plants is caused by the ability of the pathogen to produce gibberellic acids (GAs), a family of plant hormones [11]. Fungal GAs are structurally identical to those synthesized by higher plants, but the respective biosynthetic pathways, genes and enzymes differ [12–14]. Previously we have shown that the ability of the fungus to produce GAs contributes to the efficient colonization in the rice roots [15]. In contrast to typical bakanae symptoms, it is unknown how the stunting phenotype of infected rice seedlings is triggered by F. fujikuroi. In addition, stunting of infected rice plants can also be caused by other Fusarium species, such as F. proliferatum, which can also be isolated from rice, though less abundantly than F. fujikuroi [16]. Recently, the first high-quality draft genome sequence of F. fujikuroi IMI 58289 has been published and the genetic capacity for biosynthesis of a whole arsenal of natural compounds has been demonstrated [15]. The genomes of additional F. fujikuroi isolates revealed some diversity regarding genome composition and virulence [17,18]. Here we present the genome sequences of eight additional F. fujikuroi strains, all but one isolated from infected rice from different geographic regions. The isolate FSU48 was obtained from maize. We provide a comparative analysis of the genome features, chromosome polymorphism, the ability to produce micro- and macroconidia, virulence, metabolome and transcriptome analyses under in vitro and in planta conditions. By the use of high-performance liquid chromatography coupled to a Fourier transform mass spectrometer (HPLC-FTMS) and genome-wide RNA-sequencing (RNA-seq), we demonstrate that in addition to species-specific common features there are differences between the isolates in all these aspects. Most importantly, we describe two pathotypes of F. fujikuroi on rice at the genomic, transcriptomic, and phylogenetic levels. Whereas most of the F. fujikuroi isolates caused typical bakanae symptoms with elongated chlorotic internodes, the isolate B14 caused stunting and withering of rice seedlings. We show that variations in the production of secondary metabolites (SMs), such as GAs and fumonisins, are crucial factors for the development of the bakanae or stunting pathotype, and that these two pathotypes are phylogenetically distinct groups among the field population of F. fujikuroi. F. fujikuroi is broadly distributed world-wide in all rice-growing countries. To gain insight to the level of variation regarding genome structure, morphology, asexual spore formation, virulence, expression profiles and secondary metabolism under laboratory conditions as well as on rice, we chose nine isolates from different areas of the world for comparative analysis of all these parameters (Table 1, Fig 1A). The high quality genome sequence of strain IMI 58289 that was assembled into twelve scaffolds corresponding to the twelve chromosomes [15] was used as master genome for structural annotation and for comparative analysis. The rice isolate V64-1 from Ruanda appeared to be a F. oxysporum strain when the genome was sequenced and analyzed. Therefore, this strain was used as outgroup in this study. A phylogenetic tree including all so far sequenced F. fujikuroi isolates and other FFC members, as well as more distantly related Fusarium species is shown in Fig 1B. The tree was generated based on the protein sequences of 5,181 single copy genes by the fast approximate likelihood ratio test to calculate branch support (aLRT) [19], which is a fast and accurate alternative to the time-consuming bootstrap analysis. Strain KSUX10626 [17] seems to be outside the F. fujikuroi clade. Although all newly sequenced strains except for V64-1 clearly belong to the species F. fujikuroi, visualization of chromosome content of the ten strains by pulse field gel electrophoresis (PFGE) combined with clamped homogeneous electric fields (CHEF) indicated that all strains contained multiple chromosomes of varying sizes (Fig 1C). However, the precise number of chromosomes in each strain could not be determined because several chromosomes had a similar size and could not be distinguished. Table 2 summarizes physical genome features of the newly and previously (IMI 58289) sequenced strains. The genome size for the F. fujikuroi strains is in the range of 43. 9 Mb (IMI 58289) to 46. 1 Mb (E282 and FSU48). The outgroup genome F. oxysporum V64-1 is smaller (49. 1 Mb) than the reference F. oxysporum strain 4287 (61. 4 Mb). Most of these differences are likely due to a different read coverage and different completeness of assemblies for repetitive regions. Despite the varying number of protein-encoding genes (14,817 for IMI 58289 to 16,088 for E282) which partially may be due to a manual gene structure validation, most key genome features of the newly sequenced strains are similar to those of F. fujikuroi IMI 58289 [15]. The completeness of the new draft genomes was explored by comparing each predicted proteome to two different, highly conserved eukaryote protein sets by BLAST [20,21]. Orthologs for all conserved proteins were found in all proteomes with the exception of a missing ortholog to ‘T-complex protein 1 subunit theta’ and an ortholog to ‘translation factor eIF6’ in F. fujikuroi B20. In F. oxysporum V64-1, an ortholog to ‘MTO1 –mitochondrial translation optimization’ is missing suggesting that also these genomes are more than 99% complete. In addition, the protein sets were subjected to BUSCO analysis and subsequently compared to published Fusarium proteome sets [22]. Of the 3,725 single-copy orthologs searched (library Sordariomyceta_odb9), 97. 9%–99. 2% were detected as complete and single-copy in the F. fujikuroi strains and F. oxysporum V64-1 which even exceeds the completeness of F. graminearum PH1 (94. 7%) and F. oxysporum (94. 2%) (S1 Table). Species of the FFC are heterothallic [23]. To determine the mating type of the nine F. fujikuroi isolates, we searched the genomes for the presence of genes either of the MAT1-1 or MAT1-2 mating types. Five of the strains contain the genes MAT1-1-1, MAT1-1-2 and MAT1-1-3, all belonging to the MAT1-1 mating type, while the other four contain the MAT1-2 mating type locus with the HMG-box type transcription factor (TF) -encoding gene MAT1-2-1 and an additional gene, MAT1-2-3, that has been recently identified as specific to the Hypocreales [23] (S2 Table). Differences between the isolates were found regarding the size of gene families prevalent in the genome of B14, which is the only one of the ten examined strains causing stunting of rice seedlings (see below). The number of transporter-encoding genes (954), Zn (II) 2Cys6 fungal type TFs (567), cytochrome P450- (173) and dehydrogenase-encoding genes (404) were higher when compared to the other F. fujikuroi strains. Also the number of polyketide synthase (PKS) genes (18 plus two truncated pseudogenes) was larger than in the genomes of the other strains (Table 3). The strain NCIM 1100 encodes less PKS (eight plus one truncated pseudogene) and nine terpene cyclase (TC) genes which results in the inability to produce gibepyrones, fujikurin and certain terpenes besides other unknown products (S3 Table and below). To gain a deeper insight into the variation at the chromosome ends between the ten analyzed strains, we used a PCR approach with primer pairs from the two ends of each chromosome based on the genome sequence of strain IMI 58289 [15]. This analysis revealed differences between the isolates either on one side or on both sides of the chromosomes (S1 Fig). The most variations were found for subtelomeric regions of chromosomes 3,5, 7,10 and 11. Two of the key enzyme-encoding genes of SM gene clusters, NRPS31 (apicidin F) and PKS16 (unknown product), are located at the far end of chromosome 1 and 11, respectively, and the PCR analysis clearly showed differences in the presence of these SM genes (S1 Fig). All isolates showed differences in colony morphology and pigmentation on solidified and in liquid media, respectively, indicating metabolic variation between them (Fig 2A–2E). There were differences in the ability to produce the red pigments, bikaverin and fusarubins, despite the presence of the respective gene clusters in all F. fujikuroi isolates (Fig 2C–2E). Previously, we have shown that both PKS-derived pigments are only produced under low nitrogen conditions. However, while bikaverin biosynthesis is induced at acidic pH, the perithecial pigments fusarubins are produced only under alkaline pH conditions [24,25]. Strain C1995 showed no coloration under bikaverin-production conditions (Fig 2C), and C1995, B14, FSU48 and NCIM 1100 showed no pigmentation under fusarubin production conditions (Fig 2E). To verify that the biosynthetic genes are not expressed in these isolates, we performed Northern blot analyses using the bikaverin and fusarubin biosynthetic genes as probes (S2 Fig). FSR2 (encoding an O-methyltransferase in the fusarubin gene cluster) was expressed in IMI 58289, but not in the remaining strains suggesting that the fusarubin genes are only slightly expressed in m567, MRC2276, B20, E282 and F. oxysporum V64-1, or alternatively that the red pigmentation in these strains under fusarubin production conditions (Fig 2F) might be due to the expression of the bikaverin biosynthetic genes. The species F. fujikuroi is described to produce slender macroconidia with three to five septa and oval or club-shaped microconidia, mostly without or with one septum [1]. However, the strains used in this work differ in their ability to produce asexual spores, i. e. micro- and macroconidia (Fig 2F). While most of the strains produce both types of conidia, others produce predominantly (B14 and NCIM 1100) or exclusively (IMI 58289) microconidia or predominantly macroconidia (E282), respectively, on V8 agar under light conditions. Strain m567 hardly forms any spores. In A. nidulans, asexual reproduction has been extensively studied for several decades [26–28]. Sequential activation of three major regulators, BrlA, AbaA and WetA, is necessary for the fungus to undergo asexual development. In addition, there are a number of fluffy regulatory genes (FLBA–FLBE) which regulate BRLA expression. Recently, it has been shown that an AbaA-WetA pathway is conserved in the distantly related species Fusarium graminearum [29–31]. There are highly conserved homologs for FlbB, FlbC, FlbD, FlbE and the central regulators WetA and AbaA also in the genomes of the F. fujikuroi isolates. However, no close relative to BrlA, which is responsible for the vesicle formation during conidiogenesis in A. nidulans, has been found in any Fusarium genome. In recent years, several new SMs have been identified in F. fujikuroi due to the deciphering of the fungal genome and the application of molecular techniques to activate silent gene clusters, e. g. those for apicidin F, fujikurins, beauvericin, trichosetin, and reversely N-prenylated tryptophan (r-N-DMAT) [32–36] as well as for the sesquiterpenes eremophilene and guaia-6,10 (14) -diene [37,38]. In addition, well known SMs such as fusaric acid, fusarins, fusarubins and gibepyrones have been linked to the respective biosynthetic gene clusters [39–42]. Bioinformatic analysis of the nine F. fujikuroi strains revealed several differences in the presence of PKS, NRPS, dimethylallyltryptophan synthase (DMATS) and TC gene clusters between them and also compared to closely related FFC members, such as F. mangiferae and F. proliferatum [43] (S3 Table). Altogether, there are 17 NRPS-, 23 PKS-, three DMATS-, and twelve TC-encoding genes present in the nine F. fujikuroi strains, i. e. 55 unique core-enzyme-encoding genes that could give rise to 54 distinct SMs (the fusaric acid cluster encodes two core enzymes, PKS6 and NRPS34). Only twelve of the 23 PKS-, 13 of the 18 NRPS-, eight of the twelve TC- and two of the three DMATS-encoding genes are present in the genomes of all F. fujikuroi isolates. The Korean strain B14 has three additional PKS genes (PKS40,43 and 51) not present in any other analyzed F. fujikuroi strain (S3 Table, Fig 3A and 3B). PKS51 is also not present in any other member of the FFC, while PKS40 is present in F. proliferatum ET1 and F. verticillioides, and PKS43 in F. mangiferae, respectively (S3 Table). In addition, B14 is the only F. fujikuroi isolate with a complete PKS5 cluster. This yet uncharacterized cluster is either absent or contains pseudogenes in the other isolates of this species, but seems to be functional in F. proliferatum and F. mangiferae (Fig 4A). The fujikurin gene cluster (PKS19) has been previously described as F. fujikuroi-specific [15], but was recently discovered in two newly sequenced strains of F. proliferatum [43]. However, the fujikurin cluster is absent in three of the nine analyzed F. fujikuroi isolates (S3 Table, Fig 4B). The apicidin F gene cluster (NRPS31) which is located at the far end of chromosome 1 in strain IMI 58289 [15] was shown to be unique for F. fujikuroi [32]. This cluster is present in all isolates but one, B14, most likely due to chromosome rearrangements in the subtelomere regions (S1 and S3A Figs). While the entire fumonisin gene cluster (PKS11) is present in most members of the FFC except for F. mangiferae (S3 Table), nine genes of the cluster, the homologs of FFUJ_09248 –FFUJ_09256, are missing in F. fujikuroi B20 (S3B Fig) and C1995. The Indian isolate NCIM 1100 has only 14 PKS-encoding genes (S3 Table). The absence of five PKS genes/clusters is due to large parts of chromosome 11 missing at both ends in this strain compared to IMI 58289 (Figs 3C and 4C, S1 Fig). The right border of chromosome 11 contains PKS8 (FFUJ_12090), PKS13 (FFUJ_12020), PKS17 (FFUJ_12066), and PKS18 (FFUJ_12074), while PKS16 (FFUJ_11199) and adjacent genes are absent at the left border of this chromosome (Figs 3C and 4C, S3 Table). Also the discontinuous distribution of some other SM gene clusters among the F. fujikuroi isolates is likely the result of gene loss and gene gain events, respectively, at the chromosome ends. However, gene clusters located in central parts of chromosomes, e. g. PKS19 (fujikurin) and PKS11 (fumonisin) clusters, are also only present in some and absent in other strains, while the genes at the right and left borders of the clusters show colinearity between all isolates. It is not clear whether these gene clusters are the result of horizontal gene transfer [44] or cluster duplication and loss (birth and death) as shown for the fumonisin gene cluster [45]. Thanks to the genome sequencing and subsequent identification of putative gene clusters [15], SM products have now been assigned to 22 of the 54 predicted PKS-, NRPS-, DMATS-, and TC-derived SM gene clusters in isolates of F. fujikuroi, and an additional two (ferrirhodin and depudecin) in members of the FFC. One reason for the inability to identify more products is probably due to the fact that many of the SM genes are not or only minimally expressed under laboratory conditions [46]. Only 15 of these predicted clusters with known products are present in the genomes of all nine F. fujikuroi isolates indicating the genetic diversity among isolates of one species. The genes which are commonly present in all F. fujikuroi genomes are four NRPS genes required for synthesis of the siderophores ferricrocin and fusarinine and the mycotoxins beauvericin and fusaric acid (together with PKS6); six PKS genes required for synthesis of fusarubins, bikaverin, fusaric acid (together with NRPS8), fusarins, trichosetin, and fumonisins; six TCs required for synthesis of GAs, phytoene, eremophilene, (–) -α-acorenol, (–) -guaia-6,10 (14) -diene, and (+) -koraiol, and one DMATS responsible for biosynthesis of r-N-DMAT (S3 Table). To study the variability in the production of the most prominent SMs, we cultivated the strains under three standardized culture conditions (6 mM and 60 mM glutamine, 6 mM NaNO3). The optimal conditions for the production of the different metabolites were previously shown to vary considerably regarding nitrogen availability and pH in strain IMI 58289 [15]. Therefore, we analysed these SMs in the three media by high-performance liquid chromatography coupled to a Fourier transform mass spectrometer (HPLC-FTMS) or by HPLC with a diode array detector (HPLC-DAD) in the case of GAs. We compared the production levels (Table 4) with transcriptome profiles of the SM biosynthetic genes generated by RNA-seq for two of these conditions (6 mM and 60 mM glutamine) (Table 5A and 5B). For most of the SMs, we observed similar regulatory mechanisms regarding nitrogen availability as previously described for strain IMI 58289. However, there were strain-specific differences. The most prominent and species-specific SMs are the GAs causing the F. fujikuroi-specific bakanae disease of rice. Although some recently sequenced species of the FFC such as F. mangiferae and F. proliferatum contain one or even two GA gene clusters, these species produce either no or only very small amounts of GAs [43]. In this work, we examined whether all analyzed F. fujikuroi strains produce GAs, and whether the GA levels can be correlated to the virulence on rice. Previously, the regulation of GA biosynthesis has been extensively studied for strain IMI 58289. It has been shown that GA gene expression is strictly regulated by nitrogen availability in an AreA- and AreB-dependent manner [47–49]. To examine whether low nitrogen conditions are optimal for GA production also in the other isolates, we performed HPLC-DAD analysis of all strains and analysed the expression by Northern blot analysis in addition to RNA-seq data. Accordingly, we observed the highest GA yields and GA gene expression levels under low nitrogen conditions for eight of the nine F. fujikuroi strains. The only exception among the F. fujikuroi isolates is strain B14 which showed no production of GAs and no visible expression of the GA genes despite the presence of the complete GA gene cluster in the genome (Table 5A and 5B; Fig 5A and 5B). Besides the very low expression of GA genes, strain B14 differs from the other strains by high expression of fumonisin biosynthetic genes under low nitrogen conditions (Table 5A, Fig 5A and 5C). Recently, we have shown that the fumonisin gene cluster in F. fujikuroi IMI 58289 is almost silent. Consequently, only very low amounts of fumonisins are produced in comparison to F. verticillioides. In F. fujikuroi IMI58289, the genes are only expressed and fumonisins are only produced, when the cluster-specific TF gene FUM21 is constitutively and strongly expressed [50]. Here, similar results were observed for eight of the nine F. fujikuroi isolates (except for B14) which all produce either no or only very small amounts of fumonisins and showed very low expression levels. In addition, B14 is the only isolate producing fujikurins under alkaline and low nitrogen conditions (Table 4). Previously, the fujikurin gene cluster (PKS19) was shown to be silent under all conditions tested in strain IMI 58289, and the products have been identified only after simultaneous over-expression of both the cluster-specific TF gene and the PKS19 gene itself [15,34]. In most of the strains, fusarins, fusaric acid, and apicidin F (except for B14) were only produced under high nitrogen conditions (60 mM glutamine) in accordance with the higher expression levels of the corresponding biosynthetic genes. The genes of the beauvericin cluster were recently shown to be silent under all conditions tested and activated only after deletion of the histone deacetylase gene HDA1 and knock-down of the histone methyltransferase gene KMT6 in strain IMI 58289 [33,38]. Most of the other strains analyzed here showed no or very low expression levels and no beauvericin production. The only exception was strain B20 that highly expressed the biosynthetic genes under both nitrogen conditions (low and high amounts of glutamine) and produced high levels of beauvericin at high nitrogen (Table 4; Table 5A and 5B) suggesting that the chromatin status around the beauvericin cluster differs in B20. To examine whether the isolates differ in their virulence on rice, and whether the different levels of GAs or the strain-specific production of other SMs (e. g. fumonisins in B14) correleate with the extent of symptom development, we performed assays with both germinating seeds and rice seedlings. To determine the ability of all ten strains to impair the seed germination, rice seeds were soaked for 18 h in spore suspensions of the respective strains. The percentage of germinated seeds was counted after 14 days post inoculation (dpi). Strains B14 and B20 induced a high percentage of seedling death indicating a high potential to kill the host seedlings (S4 Table). As B14 does not produce much GAs, the aggressiveness of this strain cannot be caused by these phytohormones. Next, we performed a rice seedling assay to compare the virulence of the different isolates to assess their ability to induce bakanae symptoms. The surface-sterilized seeds were first germinated, and the young healthy seedlings were then inoculated with spores. All strains except for B14 and F. oxysporum V64-1 (outgroup) induced the formation of slender, elongated and yellowish stems (Fig 6). Wheras V64-1-infected seedlings behaved like the water control, B14-infected seedlings were stunted and showed withering instead of the typical bakanae symptoms. The total seedling length and the length of the internodes were smaller than, or comparable to uninfected seedlings (water control). Previously, it has been reported that heavily infected seedlings can also be stunted and can show severe crown and root rot [51]. The type of symptoms and severity of disease depends on the fungal isolate and is thought to be affected by the proportions of GA and fusaric acid produced by the fungus, which potentially cause elongation of the plants or stunting, respectively [52,53]. However, this assumption has never been proved experimentally. To examine which of the SM gene clusters are similarly expressed in rice in all isolates, and which of them are specifically expressed or not expressed in strain B14, and therefore could be relevant for disease symptoms (especially stunting), we performed RNA-seq for all isolates also from rice seedlings at 7 dpi (Table 5C). The expression patterns were compared with those under in vitro conditions at high (60 mM glutamine) and low (6 mM glutamine) nitrogen (Table 5A and 5B). In general, the GA biosynthetic genes (shown for the key enzyme gene CPS/KS) are the most highly expressed genes in rice roots except for B14 in which they are hardly expressed (Table 5C). To check whether the expression profiles of SM genes correspond to the in planta production levels, we also performed SM analysis of infected rice plants by HPLS-FTMS at 7 dpi. In accordance with gene expression, the in planta GA production levels also differed between the isolates: B14-infected rice plants contained no detectable GA amounts in rice compared to the bakanae strains (Table 6). The most obvious difference of strain B14 compared to the other isolates was the high expression of the fumonisin genes resulting in significant levels of fumonisins (FB1 and FB2) in planta, similar to those observed in vitro (Table 5C; Table 6). Furthermore, while all isolates showed similar expression patterns for fusaric acid biosynthetic genes (PKS6; NRPS34) and similar fusaric acid production levels (Table 5C, Table 6), B14 was the only strain with low expression of the fusarin C genes. As expected, no fusarins were detectable in rice roots inoculated with this strain (Table 6). In addition, the yet uncharacterized PKS51 gene cluster, which is only present in B14, was exclusively expressed in planta, suggesting that the unknown product of this gene cluster might play a role during infection (Table 5C). Besides B20, B14 also gave high expression for beauvericin genes, but only low expression of gibepyrone (PKS13) and acorenol (STC6) biosynthetic genes. In conclusion, the very low expression of GA genes and the lack of detectable GA levels in B14-infected rice seedlings after 7 dpi are most likely responsible for the absence of bakanae symptoms in this isolate. Instead, the high levels of fumonisins which are exclusively produced only in this isolate, may overrule the growth-stimulating effect of the GAs and cause stunting/withering. To find out whether one of the gene clusters that are specifically expressed in B14 during infection on rice (Table 5C) might indeed cause the stunting effect of this isolate, the key enzyme-encoding genes for fumonisins (FUM1 = PKS11), and the yet uncharacterized B14-specific PKS51 gene were deleted in this strain. In addition, we also deleted the key gene for fusaric acid biosynthesis (FUB1 = PKS6) due to the speculation that fusaric acid production might cause the stunting pathotype [52]. The 5-day-old healthy rice seedlings were soaked in the conidial suspension of B14 or the mutant strains. At 5 dpi, the B14-infected seedlings were already stunted compared to the water control and showed withering symptoms which were even more obvious at 7 dpi and 9 dpi (Fig 7A). In contrast, the Δfum1 and Δfub1 strains of B14 appeared healthy at 7 dpi and caused delayed withering symptoms at 9 dpi only. The roots of all infected seedlings were heavily colonized with fungal mycelia at 9 dpi (Fig 7D). The double deletion mutants of B14 lacking both FUM1 and FUB1 behaved like the mock control and did not induce stunting at 5 dpi (Fig 7A) or occasionally upto 9 dpi. However, in some cases, they caused a similar delayed disease development as the single deletion strains at 9 dpi (Fig 7B and 7C). The add-back strains, which were generated by introducing the native copy of FUM1 and FUB1 into Δfum1 and Δfub1 mutants, respectively, caused typical stunting/withering symptoms similar to B14 (S4A Fig). The delay of stunting/withering symptom development by Δfum1 and Δfub1 strains indicates that the production of fumonisins and/or fusaric acid, in combination with the non-detectable levels of GAs, play an important role for the development of this specific pathotype in B14. For comparison, the genes FUM1 and FUB1 were also deleted in strain B20 (bakanae strain). Unlike the B14-derived mutants, all of the gene deletion strains of B20 we examined caused typical bakanae symptoms (elongated, slender and chlorotic shoots), as did B20. No withering of the seedlings or mycelial colonization on roots were visible suggesting that the absence of fumonisin production by B20 and B20-1 derived mutants protect the rice seedlings from development of B14-like symptoms (S5 Fig). Surprisingly, rice seedlings inoculated with the ΔPKS51 mutant strains showed an earlier and more severe symptom development compared to those with B14. The stunting/withering symptoms were clearly shown already at 5 dpi (Fig 7A). The more severe symptoms caused by the ΔPKS51 strains may suggest a possible role of the PKS51 product as an avirulence determinant. A similar role was described for the product of the Magnaporthe grisea PKS-NRPS gene ACE1 (Avirulence Conferring Enzyme1) which is also specifically expressed only on rice. Its yet unknown product is probably recognized by rice cultivars carrying a specific resistance gene [54,55]. However, further investigations will be needed to show whether the product of PKS51 acts in a similar way in F. fujikuroi strain B14. To further study the impact of these SMs on disease symptom development, fumonisin FB1, fusaric acid, or GA3 were exogenously supplied to rice seedlings infected with wild-type or mutant strains (Fig 7A). Addition of FB1 and fusaric acid to rice seedlings infected with Δfum1 and Δfub1, respectively, restored the WT phenotype and resulted in stronger withering symptoms compared to the deletion strains without the toxins. However, the even more reduced virulence of the double deletion strain (Δfum1/Δfub1) was not clearly complemented by exogenous supplies of both fumonisin and fusaric acid (Fig 7A). Interestingly, exogenous supply of GA3 did not cause bakanae symptoms on rice seedlings inoculated with the wild-type B14 strain, while addition of GA3 to the Δfum1/Δfub1 double deletion strain caused stem elongation of rice seedlings at 5 dpi, similar to the bakanae symptom caused by B20 (Fig 7A). Therefore, conversion of B14 into a bakanae pathotype by addition of GAs was only possible after deleting the key genes for the production of fumonisins and fusaric acid. Furthermore, exogenous supply of culture filtrate from strain B14 to seedlings infected with B14, B20 or B20 Δcps/ks caused stunting and withering symptoms while the culture filtrate of the B14 Δfum1/Δfub1 mutant caused milder symptoms (S4B and S4C Fig). Addition of culture fluid to B14-infected seedlings resulted in even more severe stunting than B14 alone (S4B Fig). Exogenous supply of B14 culture filtrate to seedlings inoculated with B20 or the GA-deficient B20 Δcps/ks mutant revealed severe stunting symptoms in both cases irrespective of the ability to produce GAs (S4C and S4D Fig). These data support our suggestion that fumonisins and fusaric acid play an important role for symptom development. To further investigate the role of fumonisins for causing stunting and withering, we performed the rice seedling pathotest with two F. verticillioides strains from corn which were shown to produce high amounts (more than 3,000 μg/g) of fumonisins and no GAs in rice seedlings. Both F. verticilloides isolates (Os35 and Os40) [56] caused withering at 7 dpi although the plants were not stunted compared to those of the mock control (S6A Fig). In addition, mycelia of both F. verticillioides strains colonized the roots of infected rice seedling as much as those of B14 (S6B Fig). This strong root colonization was not observed for roots infected with B20. Previously, it has been already reported that B14 triggered severe inhibition of root growth, and that its own growth rate in rice roots was more than 4 times higher compared to that of B20. Taking together the results of FUM1 and FUB1 deletion, exogenous addition of pure toxins or culture filtrate of B14 to rice seedlings inoculated with wild-type or mutant strains, and the pathotests with GA-deficient, highly fumonisin-producing F. verticillioides isolates provide strong indications that the biosynthesis of both fumonisins and fusaric acid and the lack of GA biosynthesis in B14 play cruicial roles for causing the stunting/withering phenotype on rice seedlings. However, additional factors are probably involved in inducing the stunting symptoms. Besides SMs, stunting/withering might be caused also by the different sets of TFs present in the genome of B14 and the bakanae strains, or by different expression levels of TF-encoding genes in rice. Therefore, we compared the expression levels of TF-encoding genes between B14 and the other eight F. fujikuroi isolates (S5A Table). There are 37 genes which are present in most of the nine genomes and which were specifically up-regulated during infection of rice (S5B Table). B14 had slightly higher expression levels only for three of them: FFB14_03090, FFB14_05980 and FFB14_01631. In addition, B14 has 28 strain-specific TFs which are not present in the genomes of the other strains (S5C Table). The most highly expressed gene in planta was FFB14_06367 encoding the pathway-specific TF of the putative PKS51 gene cluster. The high expression of PKS51 and the adjacent genes, including the TF-encoding gene, supports our assumption that this unique gene cluster is involved in determining the severity of disease symptom development. Because B14 was the only isolate causing the stunting pathotype among the analyzed ten strains, we attempted to determine how often this pathotype can be found in rice fields. Therefore, we inoculated rice seedlings with 15 field isolates, which were collected from rice grains and air above rice paddy fields in Korea between 2014 and 2016. Among the 15 isolates, we identified additional nine field isolates causing stunting and early withering symptoms similar to B14 while six isolates caused typical bakanae symptoms (Fig 8A). To determine whether these isolates can be phylogenetically distinguished from each other, we generated a phylogenetic tree using the nucleotide sequences of the RPB2 and EF1A genes [57] from the new field isolates, the ten isolates used in this study, and some closely related Fusarium species of the FFC. The F. fujikuroi clade, which was clearly separated from those of other closely related species such as F. proliferatum and F. verticillioides, contained two strongly supported subclades (with 76% bootstrap support: BS). Interestingly, the subclade with B14 consists of all of the field isolates causing stunting symptoms, while the other subclade contained B20 and the other F. fujikuroi strains used in this study and all bakanae-type field isolates (Fig 8B). This result indicates that the two pathotypes of F. fujikuroi exist as phylogenetically distinct groups within the population. To examine if there is a correlation between the pathotype and the presence of one of the SM clusters specific to B14 or B20 (PKS51 for B14, unknown product, and NRPS31 for B20, apicidin F) among the field isolates of each pathotype, we performed a PCR amplification using primer sets derived from PKS51 (unknown product) and NRPS31 (apicidin F), respectively. The PKS51-specific primer set amplified a fragment from all of the stunting-type isolates examined including B14, but not from the bakanae-type isolates. Similarly, the NRPS31-specific primer set amplified a fragment only from bakanae-type isolates examined (S7A Fig). Based on the correlation between the phylogenetically distinct pathotypes on the one hand, and the presence of either the PKS51 or the apicidin F (NRPS31) gene clusters, we examined additional field isolates from Korea using these primer sets. Among a total of 151 isolates examined, the B14-specific PCR amplification pattern was found in 69% of rice isolates, 100% of corn isolates, and 85% of airspora isolates. A phylogenetic tree generated from all of these isolates revealed a clustering of the stunting-type and bakanae-type isolates similar to that for the field isolates examined by pathogenicity tests (S7B Fig). However, it is currently unclear why the stunting-type population is predominant among the field isolates in Korea. To gain more information on the SM profile of these new stunting-type isolates, we examined the expression levels of the key genes of GA (CPS/KS) and fumonisin (FUM1) biosynthesis in the field isolates grown in liquid culture with 6 mM glutamine by use of quantitative real-time PCR (qPCR). The CPS/KS transcript levels from all the stunting-type isolates examined were similar to or even lower than that in B14, while those from all the bakanae-type isolates, including B20, were 8- to 24-fold higher than those in the stunting-type strains (S8A Fig). Furthermore, all stunting-type isolates showed clear expression of FUM1 in contrast to the bakanae isolates (shown for B20) (S8B Fig). These data indicate that both subgroups of isolates differ in a whole set of characteristic features (presence of either the PKS51 or NRPS31 genes, expression of either GA or fumonisin genes) which correlate with either the stunting or the bakanae pathotype. We also examined the production levels for both SMs in some of the new B14-like strains in vitro and in planta. As expected, the stunting-type isolates produced no or 10 to 15 times less GAs than the bakanae strain IMI 58289, while only the stunting-type strains produced fumonisins in submers cultures (S9A and S9B Fig). The analysis of GAs and fumonisins in rice roots and shoots revealed no GAs at all in seedlings infected with stunting-type isolates, while only the latter produced significant amounts of fumonisins in planta (S9C and S9D Fig). A similar correlation between low GA and high fumonisin levels on the one hand, and a pathotype called “dwarfism” on the other hand, was also described for the Italian F. fujikuroi isolate CSV1 [58]. In conclusion, we provide genome sequences of eight new F. fujikuroi isolates and one F. oxysporum isolate, all collected in different rice growing regions worldwide. We show that all these strains differ in genome and chromosome size, number of genes in major gene families such as TFs, transporters, SM biosynthetic genes and others. In addition, the isolates differ in colony morphology, pigmentation and the number and type of asexual spores (micro- and/or macroconidia). Major differences were identified in subtelomeric regions of the chromosomes where several SM gene clusters are located. Besides the differences in the presence of gene clusters, we observed variations in the ability to express SM genes and to produce the respective metabolites under in vitro and in planta (rice) conditions. Among the nine F. fujikuroi isolates analyzed, eight cause typical bakanae symptoms on rice seedlings due to their ability to produce GAs. Only one isolate, B14, does not cause elongation of infected plants, but instead causes stunting combined with early withering of rice seedlings. This isolate is the only one which does not produce GAs under in vitro and in planta conditions. Instead, B14 produces high amounts of fumonisins both under in vitro and in planta conditions. Furthermore, it is the only strain containing a putative gene cluster with PKS51 as key gene which is highly expressed in rice. To examine the determinant (s) of the stunting/withering pathotype, several key enzyme-encoding candidate genes were deleted in B14. The data demonstrate that the formation of fumonisins, and probably also fusaric acid, on the one hand, and the lack of GA production on the other hand, contribute to the stunting/withering pathotype, because the deletion of the respective key genes resulted in reduced virulence. Furthermore, the unknown PKS51 gene cluster seems to produce a SM which acts as an attenuator of disease, because the deletion of the PKS51 gene caused early withering of infected seedlings. Examination of more field isolates from Korean rice fields revealed a correlation between the pathotype and the ability to produce either fumonisins or GAs which is supported by the clear separation into two distinct phylogenetic clades. The fungal strains used in this work and their origin are shown in Table 1. Strain IMI 58289 was derived from Commonwealth Mycological Institute, Kew, United Kingdom. Strains m567 and FSU48 were provided by the Fungal Stock Center at the University Jena, Germany, C1995 by J. F. Leslie, Kansas State University, E282 by S. Tonti, University Bologna, Italy, MRC2276 by W. C. A. Gelderblom, South Africa, and NCIM 1100 was provided by the National Collection of Industrial Microorganisms, India. Strains B14 and B20 were provided by S. -H. Yun, Korea. Strain V64-1 was kindly provided by T. Kyndt from the University Ghent, Belgium. Escherichia coli strain Top10 F’ (Invitrogen, Groningen, The Netherlands) was used for plasmid propagation. The uracil-auxotrophic Saccharomyces cerevisiae FGSC 9721 (FY 834) was provided by the Fungal Genetics Stock Center (Kansas State University) and used for yeast recombination cloning. F. fujikuroi B20: Illumina TrueSeq genome sequencing by TheragenEtex, Suwon, Korea. The assembly was performed using Celera Assembler version 7. 0 [59], ‘overlap minimum length’ set to 150 bases. The assembly resulted in 318 scaffolds with a 21-fold coverage of the TrueSeq large reads. For all other strains sequencing was carried out by shot gun sequencing of an 8 kb library with paired end 100 bp read length using Illumina HiSeq 2000 by Eurofins MWG Operon, Germany. The assemblies were performed by ALLPATHS-LG [60] and the scaffolds were error corrected by mapping all Illumina shotgun paired-end data and further scaffolded using SSPACE [61] (Table 2). The data on the new genomes, including annotation, was submitted to the European Nucleotide Archive, study accession PRJEB14872 available at: http: //www. ebi. ac. uk/ena/data/view/PRJEB14872. Sample accession numbers are listed in Table 1. RNA-seq data are available at: https: //www. ncbi. nlm. nih. gov/gds/? term=GSE89480. Draft gene models for all genomes were generated by three de novo prediction programs: 1) Fgenesh [62] with different matrices (trained on Aspergillus nidulans, Neurospora crassa and a mixed matrix based on different species); 2) GeneMark-ES [63] and 3) Augustus [64] with Fusarium ESTs and RNA-seq based transcripts as training sets. Annotation was aided by exonerate [65] hits of protein sequences from F. fujikuroi IMI 58289 and F. oxysporum 4287 to uncover gene annotation gaps and to validate de novo predictions. Transcripts were assembled on the RNA-seq data sets using Trinity [66]. The different gene structures and evidences (exonerate mapping, RNA-seq reads and transcripts) were visualized in GBrowse [67] allowing manual validation of coding sequences with a focus on SM cluster genes and other genes of interest. The best fitting model per locus was selected manually and gene structures were adjusted by splitting or fusion of gene models and redefining exon-intron boundaries if necessary. tRNAs were predicted using tRNAscan-SE [68]. The predicted protein sets were searched for highly conserved single (low) copy genes to assess the completeness of the genomic sequences and gene predictions. Orthologous genes to all 246 single copy genes were searched for all proteomes by BLASTp comparisons (eVal: 10−3) against the single-copy families from all 21 species available from the FunyBASE [21]. Additionally, the proteomes were searched for the 248 core-genes commonly present in higher eukaryotes (CEGs) by BLASTp comparisons (eVal: 10−3) [20]. We also used BUSCO Version 3. 0. 1 in the Ubuntu virtual machine with the lineage specific profile library Sordariomyceta_odb9 (3. 725 BUSCO groups), downloaded from http: //busco. ezlab. org. The analysis was performed in gene set (protein) assessment mode running the python script run_BUSCO. py [22]. All genomes were analyzed using the PEDANT system [69]. To avoid misleading ortholog information based on similarity and bi-directional best hits, QuartetS [70] was applied to retrieve a reliable ortholog matrix which was used for all comparative representations. The phylogenetic tree of Fusarium species was calculated based on the protein sequences of 5,181 single copy genes that are shared among all analyzed species. Orthologs of the sequences were aligned separately using MAFFT [71]. After that, we concatenated the alignments and removed columns with gaps using Gblocks [72]. The evolutionary history was inferred using the Maximum Likelihood method PhyML [73] with default parameters and the amino acid substitution model LG. Branch support was tested using approximate likelihood ratio test (aLRT) based on the Shimodaira-Hasegawa-like (SH-like) procedure [74]. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. We calculated single copy genes in clustering proteins of all genomes and selecting clusters with exactly one representative from each genome. Protein clusters were calculated using usearch [75] (e-value cutoff: 0. 01) and mcl [76] (inflation value: 2). For phylogenetic analysis of field isolates of F. fujikuroi in Korea, TEF1α and and RPB2 were amplified from fungal genomic DNAs using the primer sets fRPB2-7cF/fRPB2-11aR [77] and EF1/EF2 [78], respectively (S6 Table). All nucleotide sequences from PCR products were edited with Lasergene (ver. 6. 0; DNASTAR, Madison, WI, USA) and aligned using ClustalW [79]. Maximum parsimony (MP), neighbor-joining (NJ), and unweighted pair group method with arithmetic mean (UPGMA) analyses were performed using MEGA (ver. 4. 02) with 1,000 bootstrap replications. Protoplasts from Fusarium strains were prepared as described previously. The protoplast suspension was mixed with 1. 2% InCert agarose (Lonza Group AG, Basel, Switzerland) and then loaded on a CHEF gel as described in [80]. Chromosomes of Schizosaccharomyces pombe and S. cerevisiae were used as a molecular size marker (Bio-Rad, Munich, Germany). To identify SM clusters in each genome, the InterPro scan results of the PEDANT analysis were used as described [81]. Essentially, predicted proteins with homology to a domain of a signature SM core enzyme (e. g. PKS, NRPS, TC or DMATS) were considered a marker for a gene cluster. A cluster was verified if any neighboring genes with homology to typical SM cluster enzymes, like P450 monooxygenases, oxidases, methyltransferases, MFS or ABC transporters or TFs, were identified. The extent of each putative gene cluster was then adjusted by comparison to previously published data and to homologous clusters in other Fusarium species. For SM production experiments and RNA preparation under in vitro conditions, strains were first cultivated for 3 days in 300 mL Erlenmeyer flasks with 100 mL Darken medium [82] on a rotary shaker at 180 rpm at 28°C. 500 μL of this culture were then used to inoculate 100 mL of ICI (Imperial Chemical Industries, UK) media [83] containing either 6 mM glutamine, 60 mM glutamine, or 6 mM NaNO3 for 3 days (RNA extraction) or 7 days (SM analysis), respectively. For RNA extraction, mycelia were flash-frozen with liquid nitrogen prior to lyophilization. For in planta expression studies by RNA-seq, lyophilized roots of infected rice plants were used for RNA preparation. For the generation of infected rice samples for transcriptome analysis by RNA-seq, rice seeds of the cultivar GSOR 100, Nipponbare, and Dongjin were used. The seeds of the former two were provided by Genetic Stocks-Oryza (GSOR) Collection, USDA ARS Dale Bumpers National Rice Research Center, Hwy, Arkansas, USA. Total RNA was extracted from mycelia grown for 3 days in liquid ICI media (containing either 6 or 60 mM glutamine) and from infected rice seedlings after 7 dpi using TRIzol Reagent (Life Technology, Karlsruhe, Germany) and purified using an RNeasy Plant MinElute Cleanup Kit (Qiagen, Hilden, Germany). The quality of DNase-treated RNA (28S: 18S > 1. 0; RIN≥ 6. 5; OD260/280 ≥1. 8; OD260/230 ≥ 1. 8) was determined using an Agilent Bioanalyzer. The high quality RNA was sent to BGI Tech Solutions Co. , Limited (Hong Kong) for library construction and sequencing by Illumina HiSeq2000 technology. RNA-seq reads were mapped on the reference genome using tophat2 (v2. 0. 8). The interval for allowed intron lengths was set to min 20 nt and max 1 kb [84–86]. We used cufflinks to determine the abundance of transcripts in FPKM (fragments per kilobase of exon per million fragments mapped) and calculated differentially expressed genes using cuffdiff [85,86]. The gene models were included as raw junctions. Genes with a minimum of fourfold increase or decrease in expression (|log2 of the FPKM values +1| ≥ 2) between the two experimental conditions were considered as regulated. The RNA-seq data has been deposited in NCBI' s Gene Expression Omnibus [87] and are accessible through GEO Series accession number GSE89480. The DNA constructs for deletion of FUM1 (FFB14_08440, FFB20_01984), FUB1 (FFB14_01651FFB20_13404), and PKS51 (FFB14_06372) from the genomes of F. fujikuroi strains B14 or B20 were created using a split-marker recombination procedure as previously described [88]. To delete FUM1, the 5' and 3' flanking regions of the FUM1 ORF were amplified with the primer pairs JFUM1f5/JFUM1rt5 and JFUM1ft3/JFUM1r3, respectively (in the first round of PCR), fused to the hygromycin B resistance gene (hygB) cassette, which was amplified from pBCATPH [88] with the primers HygB-for and HygB-rev (in the second round of PCR), and used as a template to generate split markers with the new nested primer sets, JFUM1fn/pUH-BC/H3 and JFUM1rn/pUH-BC/H2, respectively (in the third round of PCR) (S6 Table). Similarly, DNA constructs for targeted deletions of the other genes were created using the strategy described above. For the deletion of FUB1, the primer pairs JFUMB1f5/JFUM1Brt5 and JFUB1ft3/JFUB1r3 were used for amplification of 5' and 3' flanking regions of the FUB1 ORF, respectively, and JFUB1fn/pUH-BC/H3 and JFUB1rn/pUH-BC/H2 were used as the nested primer sets, respectively. For the deletion of PKS51, the primer pairs B14_6372For5/B14_6372rev5t and B14_6372for3t/B14_6372rev3 were used in the first round of PCR, and B14_6372forN/ pUH-BC/H3 and B14_6372revN/ pUH-BC/H2 for the third round of PCR. Additionally, for double deletion of FUM1 and FUB1, we generated a knock-out construct through which the 5′- and 3′-flanking regions of the FUM1 ORF were fused to a geneticin resistance gene cassette (gen) amplified from pII99 using the primer pair Gen-for and Gen-rev, as described above. The resulting construct was transformed into the deletion strain of FUB1. For the complementation of each deletion mutant of FUM1 or FUB1 derived from B14, intact copies of each gene were amplified from the genome of B14 using the primers JFUMf5/JFUMr3 and JFUB1f5/JFUBr3, respectively, and directly added into protoplasts of each deletion strain along with pSK660 including the geneticin resistance (gen) gene for co-transformation as previously described [89]. All primer sequences are listed in S6 Table. The effect of the single isolates on rice seedling germination was studied by infecting the seeds with conidia of each isolate. For spore formation, the fungal strains were cultured on PDA plates for 7 days at 28°C under light/dark (12 h/12 h) conditions. The plates were flooded with sterile water to obtain a conidial suspension (1 x 10−6). Seeds were soaked in the suspension for 18 h. Inoculated and non-inoculated (control) seeds were sown into 100 mL plastic pots. Eight days after sowing the number of germinated seeds was assessed. The number of dead, chlorotic and elongated seedlings was measured 15 days after sowing. For pathogenicity assays, healthy rice seeds were surface-sterilized by submersion in 70% of ethanol followed by 1% of sodium hypochlorite. Sterilized seeds were germinated in Murashige and Skoog (MS) agar [7] at 26°C for 5 days. The Fusarium isolates were first grown on oatmeal agar for one week. The pathogenicity assays were performed as previously described [7]. Agar plugs from the oatmeal plates were placed on top of 3 cm of sterilized vermiculite in a glass tube (18 cm x 1. 6 cm). The agar plugs were then covered with 3 cm of vermiculite. Five-day-old seedlings were transferred to the surface of the vermiculite layer to avoid the direct contact between seedlings and fungal inocula. Before covering the tubes with a cap, 4 mL of Yoshida solution was gently added to each test tube to help retain high humidity [90] and placed at 26°C for 3,5, 7 or 9 days. Their heights and internode lengths were measured and photographs of the seedlings and infected roots were taken. The symptom development caused by wild-type and mutant strains was examined in five independent pathogenicity tests. For pathotests with exogenous supply of culture filtrates, the wild-type B14 strain or its mutant lacking FUM1 and FUB1 (Δfum1/Δfub1) was inoculated into 50 mL of PDB (potato dextrose broth) and incubated for 5 days. The fungal liquid cultures were filtered through 2 × cheesecloth followed by filtration through 0. 25 μm membranes. The culture fluid was dried to 5 ml by lyophilization (10-fold concentration). For inoculation assay, 500 μl of the concentrated culture filtrate was exogenously supplied to a single rice seedling. F. fujikuroi B14 and B20 strains were transformed as previously described for F. graminearum [91]. Vector integration events were confirmed by diagnostic PCR (S10 Fig) using specific primers as indicated (S6 Table). PCR mixtures contained 25 ng of template DNA, 50 ng of each primer (S6 Table), 0. 2 mM deoxynucleoside triphosphates, and 1 U of Biotherm Taq polymerase (Genecraft, Lüdinghausen, Germany). The cDNA synthesis was performed using Superscript II (Invitrogen, Groningen, The Netherlands) and 1. 5 μg of total RNA as the template, according to the manufacturer' s instructions. The qPCR was performed using iTaq Universal SYBR Green Supermix (BioRad) and Superscript II cDNA as template, in a Biorad thermocycler iTaq. In all cases, the qPCR efficiency was between 90–110% and the annealing temperature was 60°C. Gene expression was measured in three biological replicates from each time point, and the relative expression levels were calculated using the ΔΔCt method [92]. The expression of a translation elongation factor α gene (EF1A), amplified by a primer pair (EF1-PS1 and EF1-PS2) (S6 Table), was used as an endogenous reference for data normalization. For analyses of the SMs, the strains were grown in submerged cultures as described above. After 7 days, mycelia were removed from the culture by filtration through Mirachloth (Calbiochem, Merck KGaA, Darmstadt, Germany). The culture filtrates were filter-sterilized using 0. 45 μm syringe filters (BGB, Schloßböckelheim, Germany). Fusaric acid and beauvericin were obtained from Sigma-Aldrich (Deisenhofen, Germany), GAs from Serva (Heidelberg, Germany) and methylparaben (MePa) was obtained from Fluka (Steinheim, Neu-Ulm, Germany) in analytical grade. The remaining standard substances were obtained as described in previous work [24,25,34,35,42,93–95]. All solvents were obtained in gradient or analytical grade from Sigma-Aldrich, VWR (Darmstadt, Germany) or Merck Schuchardt (Hohenbrunn, Germany). Water was purified by a Milli-Q Gradient A 10 system (Millipore, Schwalbach, Germany). Liquid culture samples were prepared as following: 10 μL of the culture filtrate and 10 μL of MePa (100 μg/mL) as internal standard were added to 80 μL of water. For in planta analysis, ten rice samples were combined and freeze-dried. The dried samples were treated with liquid nitrogen and pestled simultaneously, larger pieces were cut first with a scalpel. The samples were weighed and extracted with 1. 5 mL of the following mixture: ethyl acetate: methanol (MeOH): dichloromethane, 3: 2: 1. Precellys ceramic beads 1. 4/2. 8 (Peqlab, Erlangen, Germany) were added to the samples, and the mixture was vortexed for 1 min. Afterwards, the samples were shaken for 1 h on a rotary shaker with 150 rpm. After a short centrifugation step (3 min, 2900 g), 500 μL of the supernatant were transferred to a new vial and the solvent was evaporated to dryness under constant nitrogen flow. The residue was dissolved in 100 μL MeOH and put to an ultrasonic bath for 10 min. After vortexing the samples, they were centrifuged again with 5000 g, and 50 μL of the supernatant were collected. Afterwards, the samples were dried again under nitrogen flow and 1. 5 mL of MeOH/water, 3/1 (v/v), + 0. 1% formic acid (FA) were added. The extraction procedure described above was repeated. 10 μL of each extract were mixed with 10 μL MePa and 80 μL water for analysis. Some extracts were diluted again 1: 10; the corresponding values are labelled. The liquid culture samples were analyzed as following: A Shimadzu LC-20AD HPLC (Shimadzu, Kyoto, Japan) with a SIL-20ACXR autosampler coupled to a Sciex QTRAP 5500 (Sciex, Darmstadt, Germany) mass spectrometer was used. Separation was carried out on a Synergi Hydro-RP column from Phenomenex (Aschaffenburg, Germany) with 50 × 2. 0 mm and 2. 5 μm particle size, protected by a KrudKatcher classic filter (Phenomenex), and on a Nucleodur C18 Gravity-SB column from Macherey-Nagel (Düren, Germany) with 50 × 2. 0 mm and 3 μm particle size, protected by a KrudKatcher classic filter (Phenomenex), at 50°C. MeOH + 1% FA + 5 mM NH4Ac was used as eluent A, water + 1% FA + 5 mM NH4Ac was used as eluent B. A flow volume of 450 μL/min was applied, and the gradient started at 5% A. This condition was held for 1. 5 min. The gradient rose up to 98% A in 10. 5 min, and finally the column was rinsed for 3 min with 98%. After that, the column was equilibrated for 2. 5 min with 5% A. The integrated valco valve was used, discarding the first 2 min of the run, and the samples were cooled to 7°C. 5 μL of each sample was injected. Advanced scheduled multiple reaction monitoring (MRM) mode was used for acquisition. Both positive and negative ionization mode were applied. The curtain gas (CUR) was set to 35 psi, the collision gas was set to medium. The temperature of the heater gas (TEM) in the ion source was set to 450°C. Nebulizer gas (GS1) and auxiliary gas (GS2) were adjusted to 35 and 45 psi, respectively. In the positive MRM mode, the ion spray voltage (IS) was set to 4500 V, and the target scan time of this experiment was adjusted to 0. 3 s, resulting in a cycle time of 0. 5 s. The positive ionization mode was applied for the relative quantification of O-methyl-fusarubin, fusaric acid, gibepyrone A, apicidin F, beauvericin, fumonisins (FB1+FB2), fusarins, trichosetin and fujikurin A. The cell exit potential (CXP) was set to 11 V, the entrance potential (EP) was set to 10 V. For the negative MRM mode, the IS was set to -4500 V and the target scan time was adjusted to 0. 2 s, resulting in a cycle time of 0. 5 s. The negative ionization mode was applied for the relative quantification of MePa, and the gibberellins GA3, GA4 and GA7. The CXP was set to 11 V, the EP was set to 10 V. MRM transition for the quantification were as follows: FB1−722. 4–223. 1; FB2−706. 4–336. 3; GA3−345. 2–143. 0; GA4−331. 3–243. 1; GA7−331. 3–243. 1. The calibration curve for all the standards was prepared in a concentration range of 0. 0001–1 μg/ml. For bikaverin determination in liquid culture samples, HPLC-UV measurements were carried out on a Shimadzu LC-20AT pump system with a Shimadzu SIL autosampler and a photodiode array (PDA). A Gemini 5 u C6-Phenyl 110A, 250 × 4. 60 mm, 5 μm column (Phenomenex) was used, with water + 1% FA as eluent A and acetonitril + 1% FA as eluent B. The column oven was set to 40°C. The gradient started with 10% B with 1. 35 μL/min. After 3 min, the gradient rose up to 100% B during 17 min. The column was washed with 100% B for 6 min, and afterwards the column was equilibrated with 10% B again for 4 min. The wavelength for PDA analysis ranged from 220–600 nm. 100 μL of the sterile culture filtrate were injected. Peak areas were determined at 508 nm, and bikaverin obtained from Sigma Aldrich was used as standard substance. In planta sample analysis was performed with a different HPLC-system but the same mass spectrum. An Agilent 1260 HPLC system (Santa Clara, USA) was used, and the expected retention time for scheduled MRM analysis needed to be adjusted (S7–S10 Tables). Furthermore, bikaverin was analyzed in positive ionization mode.

Fusarium fujikuroi causes bakanae disease of rice. Infected seedlings appear to be taller and more slender when compared to healthy seedlings due to its ability to produce gibberellic acids (GAs). The disease is responsible for high yield losses, and its incidence varies with regions, rice cultivars grown and the aggressiveness of the fungal isolates. However, not all infected seedlings show bakanae symptoms: one of the isolates, B14, causes stunting and early withering of infected seedlings. The reason for the two pathotypes is not well understood. Researchers thought that the stunting phenotype was mostly caused by fungal-derived secondary metabolites such as fusaric acid, but there is no experimental evidence yet. B14 differs from the other strains by the presence of more PKS gene clusters, low expression of GA genes, lack of detectable levels of GAs and the production of high amounts of fumonisins in rice. Analysis of additional field isolates revealed a strong correlation between the pathotype (bakanae or stunting) and either GA or fumonisin production. Based on phylogenetic analyses, F. fujikuroi strains can be divided into two phylogenetically distinct subclades according to their pathotype and secondary metabolite profiles. This study provides new insights into the genomic variations and the population structure inside the species F. fujikuroi which will help to develop disease control strategies for this rice pathogen.

Abstract Introduction Results and discussion Material and methods

medicine and health sciences fusarium pathology and laboratory medicine pathogens microbiology hormones plant science rice plant hormones experimental organism systems seedlings molecular biology techniques plants fungal pathogens research and analysis methods mycology grasses artificial gene amplification and extension medical microbiology gene expression microbial pathogens gibberellins comparative genomics molecular biology biochemistry plant biochemistry eukaryota plant and algal models polymerase chain reaction genetics biology and life sciences genomics computational biology organisms

2017

Comparative genomics of geographically distant Fusarium fujikuroi isolates revealed two distinct pathotypes correlating with secondary metabolite profiles

During the first meiotic prophase, programmed DNA double-strand breaks (DSBs) are distributed non randomly at hotspots along chromosomes, to initiate recombination. In all organisms, more DSBs are formed than crossovers (CO), the repair product that creates a physical link between homologs and allows their correct segregation. It is not known whether all DSB hotspots are also CO hotspots or if the CO/DSB ratio varies with the chromosomal location. Here, we investigated the variations in the CO/DSB ratio by mapping genome-wide the binding sites of the Zip3 protein during budding yeast meiosis. We show that Zip3 associates with DSB sites that are engaged in repair by CO, and Zip3 enrichment at DSBs reflects the DSB tendency to be repaired by CO. Moreover, the relative amount of Zip3 per DSB varies with the chromosomal location, and specific chromosomal features are associated with high or low Zip3 per DSB. This work shows that DSB hotspots are not necessarily CO hotspots and suggests that different categories of DSB sites may fulfill different functions. During meiosis, the programmed formation of DNA double-strand breaks (DSBs) and their repair by homologous recombination ensures that crossovers (CO) occur between homologous chromosomes. COs promote the accurate segregation of homologs at the first meiotic division, thus avoiding aneuploidy, which is a common cause of birth defects and congenital diseases. In all species, two to 30 times more DSBs are formed than COs, indicating that only a subset of all DSBs formed in a cell are repaired through a pathway that will give rise to a CO. The remaining DSBs are repaired by other homologous recombination pathways, such as the synthesis dependent strand annealing (SDSA) mechanism, symmetrical Holliday junction resolution or Holliday junction dissolution [1], that result in non-crossovers (NCOs). In addition, a substantial fraction of meiotic DSBs is also repaired by homologous recombination using the sister chromatid as template, which is not productive for chiasmata and homolog segregation [2]. The repair pathway choice has thus to be tightly controlled to ensure the required number of COs per homolog pair. DSBs and COs tend to occur more frequently at preferred sites, or hotspots. It is not known whether DSB hotspots are also CO hotspots, or whether DSB repair is modulated by DSB localization on a chromosome. This question could be answered by comparing a high resolution genome-wide map of CO frequencies to the existing high resolution maps of DSBs, for instance in budding yeast (e. g. , [3], [4]). Nevertheless, several studies have suggested that the relative contribution of each DSB repair pathway may vary from site to site along the genome. For instance, using a small number of yeast meioses, Mancera et al noted that some sites gave rise to more COs and others to more NCOs per total recombination events [5]. Using a similar approach, Fung and colleagues showed that close to centromeres, COs and NCOs are strongly repressed although DSB activity was reported in these regions, suggesting that DSBs in centromere-proximal chromosomal regions are preferentially repaired by sister chromatid recombination [6], [7]. Analyses of human sperm recombination frequencies revealed that the CO/NCO ratio varied 30 times in the sites under study [8], [9], [10]. Finally, in the fission yeast Schizosaccharomyces pombe, strong discrepancies were found between the DSB map and the CO frequencies [11]. Thus, it is worth investigating if the map of meiotic DSBs truly reflects the map of COs along the genome, and what chromosomal features may influence the choice of DSB repair pathway. Several factors affect CO formation and their sites of action may reflect how a DSB is repaired. A group of proteins collectively termed “ZMM” is necessary for the formation of about 85% of all COs in budding yeast [12], [13]. During yeast meiosis, the ZMM proteins act by stabilizing the Single End Invasion (SEI) recombination intermediate, which once formed is transformed via capture of the second break end into a double Holliday junction (dHJ) that is mainly resolved as a CO [12], [14], [15]. The ZMM group comprises proteins that act directly on recombination intermediates in vitro, such as the Mer3 helicase, which promotes D loop extension and the Msh4–5 heterodimer, which stabilizes dHJs. This group also includes Zip1, the central element of the synaptonemal complex (SC), as well as Zip2, Zip3, Zip4 and Spo16 that might promote SC formation through Zip1 polymerization between homolog axes [13], [16]. Currently, it is hypothesized that the ZMM proteins, by promoting SC initiation and by directly acting on recombination intermediates, protect the CO-prone recombination intermediates (dHJ) from dissolution by anti-CO proteins, such as Sgs1 [17]. Zip3 has orthologs in C. elegans (ZHP-3) and in mammals (RNF212) and is considered to be a SUMO E3 ligase that sumoylates chromosome axis proteins, thus promoting SC polymerization. Indeed, the Zip3 sequence includes a SUMO Interacting Motif (SIM) and a C3H2C3 Ring-Finger Motif (RFM) that are important for Zip3 in vitro E3 ligase activity and necessary for SC polymerization and correct sporulation [18]. Indirect evidence suggests that ZMMs localize at CO-designated sites, but this has never been demonstrated. ZMMs form foci during meiotic prophase at the time of recombination [16], [19], [20] and the number of Zip3 foci is compatible with CO frequency in wild-type yeast strains [20]. Moreover, in hypomorphic spo11 mutant strains in which the number of DSBs but not of COs is reduced (a phenomenon known as CO homeostasis), the number of Zip3 foci follows the CO variation [21]. Finally, Zip2 foci are non-randomly distributed along chromosomes, like COs [22]. Among the ZMMs, Zip3 seems to be acting earlier because it is required for focus formation of all the other ZMMs [16]. We thus mapped Zip3 binding sites along individual genomic regions and genome-wide during budding yeast meiosis and then determined the features that influence its distribution. We show that Zip3 association with chromosomes is dynamic, occurring first with centromeres, in a DSB-independent manner, then with meiotic chromosome axes upon DSB formation and finally with DSB sites upon joint molecule formation, the preferred intermediate for CO production. These features establish Zip3 as a marker of CO-designated sites. Genome-wide mapping of Zip3 recruitment to DSB sites demonstrates the existence of different types of DSB hotspots based on CO production. Zip3 localization was previously investigated only by indirect immunofluorescence on chromosome spreads. To investigate Zip3 localization on meiotic chromosomes at about 1-kb resolution, we used chromatin immunoprecipitation (ChIP) and qPCR and yeast strains in which Zip3 was C-terminally tagged at its endogenous locus with three copies of the Flag epitope. Strains expressing the ZIP3-His6-FLAG3 allele showed normal meiotic progression and spore viability (98%, 205 tetrads dissected), showing that the tagged protein is functional. During a meiotic time-course, DSBs monitored at the BUD23 promoter hotspot on chromosome 3 form and reach a maximum at 3–4 hr, before getting repaired (Figure 1A). Zip3 showed a reproducible dynamic localization. It bound first to centromeres from 2 hr after meiosis induction and before DSB formation, then to axis-associated sites and finally to DSB sites, particularly at 4 hr (Figure 1E). At this time, DSB fragments, as detected by Southern blotting, started disappearing (Figure 1A), indicating that DSB ends were already engaged in homologous recombination repair. As Zip3 might be a SUMO E3 ligase, we investigated whether interaction with SUMO regulated Zip3 binding to the different chromosomal structures. To this aim, we mutated the Zip3 SIM (zip3I96K mutant) or the RFM (zip3H80A mutant) motif. Both mutated proteins were timely induced during meiosis, but they lacked the characteristic lower migrating bands that correspond to sumoylated Zip3 [18] (Figure 1B). In both mutants, early Zip3 binding to centromeres was abolished (Figure 1E), consistent with the previous suggestion that Zip3 recognizes sumoylated proteins at centromeres [18]. Moreover, recruitment to axis-associated and DSB sites was also mostly abolished (Figure 1E) and meiotic progression was impaired in both zip3 mutants (Figure 1D), similarly to what was observed in zip3 null mutants (data not shown). These findings indicate that Zip3 SUMO binding and E3 ligase activities are essential for Zip3 association with chromosomes and all its functions in meiosis. SUMO binding could be directly involved in Zip3 recruitment to all these chromosome locations or indirectly, if required only for the initial Zip3 enrichment at centromeres, and if this is an essential step for the subsequent recruitment of Zip3 to axes and DSB sites. We then mapped Zip3 binding sites genome-wide using microarrays at 3,4 and 5 hr during meiotic progression in two independent meiotic time-course experiments (Figure S1A and S2). Genome-wide profiling confirmed the results obtained by ChIP and qPCR (Figure 2A). We then compared the Zip3 maps with the maps of the axis-associated Rec8 cohesin [23] and of Red1, another meiotic axis component that does not show the strong centromere association characteristic of Rec8 [24]. The reference DSB map was the map established by genome-wide mapping of ssDNA in a repair-defective dmc1Δ mutant [3]. At 3 hr after meiotic induction, Zip3 was strongly associated with centromeres, as seen on individual chromosomes (Figure 2A and Figure S3) and in the genome-wide analysis (Figure 2B, Figure S1B and Table 1). All 16 centromeres contained a strong Zip3 peak at less than 1 kb away, and 16% of the 287 Zip3 peaks at this time were found at less than 10 kb from the centromeres. Moreover, 81% of Zip3 peaks at less than 10 kb from a centromere overlapped with an axis-associated Rec8 peak and 38% with a Red1 binding site. At 3 hr, Zip3 was weakly associated with chromosome arms and the Zip3 peaks at more than 10 kb from a centromere coincided with Rec8 (54% peaks) and Red1 (50%) enriched sites (Figure 2B and Figure S3). This is reflected by the overall strong correlation between the Zip3 signal at 3 h and the Rec8 and Red1 profiles (Table 1). At 4 hr, Zip3 association with Rec8 sites diminished (only 35% of its 966 binding sites occurred at Rec8 sites), while its association with DSB sites started to increase (Figure 2B, Figure S4, and Table 1). Concomitantly, the relative Zip3 binding to centromeres decreased (Figure 2B). Finally at 5 hr, Zip3 was almost exclusively associated with DSB sites. Indeed, none of the 557 Zip3 peaks was found at less than 1 kb from centromeres and only 15% of Zip3 peaks coincided with a Rec8 peak at this time (Figure 2B and Table 1). Thus, during meiosis, Zip3 associates first with centromeres. Centromeric Zip3 enrichment is then progressively reduced, whereas association with axis sites and particularly with DSB sites increases, in agreement with its previously described role in recombination. To investigate which events triggered these dynamic changes in Zip3 localization we used yeast mutants that affect precise steps of recombination (Figure 3A). Zip3 association with centromeres early in meiosis might occur independently of DSB formation. Indeed, by using the spo11Δ mutant in which DSBs are not formed, we could show that Zip3 associated transiently with centromeres, but not with axis or DSB sites (Figure 3B and 3C: ChIP and qPCR analysis of individual sites; Figure S3 and Table 1: genome-wide analysis). Thus, association of Zip3 with centromeres is independent of DSB formation, whereas DSB formation is required for Zip3 association with the chromosome arms. Moreover, in the rad50S mutant strain, where Spo11 DSBs are formed but not processed, Zip3 was recruited to centromeres and then chromosome axes, but not to DSB sites (Figure 3B and 3C). In the dmc1Δ mutant that is resection-proficient but deficient in strand invasion, Zip3 was transiently recruited to the axis-associated sites, with kinetics similar to those of wild-type cells, but associated rarely with DSB sites (at least eight times less than in wild-type cells), at the three sites examined (Figure 3B and 3C). Similarly, in the mnd1Δ mutant in which Dmc1 is loaded onto DSB ends but strand invasion does not occur [25], Zip3 was recruited to axes, but not to DSB sites (Figure 3B and 3C). We conclude that DSB formation is sufficient to trigger Zip3 localization at axis sites, whereas strand invasion is required for Zip3 association with DSB sites. In meiosis, rad52Δ mutants allow strand invasion by Dmc1 filaments, and wild-type levels of the Single End Invasion (SEI) intermediate, a crossover-specific intermediate, but are strongly impaired in the following step, second end capture, which leads to double Holliday junction formation and crossover resolution [26], [27]. In rad52Δ mutants, we detected centromere and axis association delayed but to nearly wild-type levels, but a strongly reduced binding of Zip3 to the three DSB sites (Figure 3B and 3C). This suggests that Zip3 requires the second end capture step, a crossover specific event, for associating with sites of DSB. Finally, we analyzed Zip3 association with chromosome structures in the ndt80Δ mutant in which dHJs are formed but not resolved [14]. Zip3 recruitment to DSB sites occurred, at levels even higher than in wild-type, suggesting that dHJ formation is the event that triggers or stabilizes Zip3 recruitment to DSB sites (Figure 3B and 3C). In addition, we reproducibly detected a very strong enrichment on the axis, perhaps a consequence of the aberrant turnover of dHJ intermediates in this mutant. Finally, we noticed that Zip3 remained bound with DSB sites longer than in wild-type (Figure 3B). This mutant analysis reveals that Zip3 associates with DSB sites only when they are engaged in dHJ intermediates, which are the CO precursors. Therefore Zip3 association with DSB sites can be considered as a marker for CO sites. We next investigated the role of Zip1, which is the central element of the SC and was previously described as not necessary for Zip3 focus formation [16], [20], in Zip3 localization by ChIP and qPCR analysis. In the absence of Zip1, Zip3 was recruited to centromeres, although less than in wild-type cells, and to axis-associated sites, but only rarely to DSB sites (about 10-fold reduction, Figure 3B and 3C). This may be linked to the suggested role of Zip1 in stabilizing the Smt3 chains that are good binding substrates for Zip3 ([18] and Discussion). Key events of meiosis are regulated by several kinases that are activated at different steps of meiosis. As Zip3 is phosphorylated in a DSB-dependent manner in meiosis ([18] and Figure 4A), we asked whether the dynamic Zip3 localization on chromosomes could be regulated by changes in its phosphorylation status. The CDK kinase Cdc28, together with the Cdc28-associated cyclins Clb5 and Clb6, is necessary for meiotic replication, DSB formation and SC formation [28] and can phosphorylate Zip3 in vitro [29]. In vivo, post-translational modifications of Zip3 are reduced in a clb5 and clb6 mutant [18], suggesting that Zip3 may be a CDK target. We mutated the six S/T-P CDK consensus motifs of Zip3 to A-P motifs (Figure S5) and found that mutant and wild-type Zip3 were similarly recruited and that meiotic divisions and spore viability were unaffected (Figure S5 and data not shown), demonstrating that Zip3 phosphorylation by CDK has no role in normal meiosis. We next investigated the role of Zip3 phosphorylation by the Tel1/Mec1 kinases, the budding yeast homologs of ATM/ATR. Tel1 and Mec1 are activated upon meiotic DSB formation and play important roles in several key meiotic processes, such as DSB end resection, inter-homolog recombination and regulation of meiotic prophase checkpoint [30]. To this aim, we mutated the four S/T-Q consensus motifs for Tel1/Mec1 to A-Q motifs (zip3-4AQ mutant). This led to a decrease of the low migrating forms of Zip3 due to phosphorylation (Figure 4B). Many of the Mec1-dependent phosphorylated proteins are substrates for the PP4 phosphatase, including histone H2A129 or the Zip1 protein in meiosis [31]. We found that the Zip3 lower migrating forms accumulated in a pph3Δ catalytic subunit PP4 phosphatase mutant, but not in a double zip3-4AQ pph3Δ mutant (Figure 4C). Together, these findings provide strong indication that Zip3 is phosphorylated by the Mec1/Tel1 kinases during meiosis. We next investigated the meiotic phenotypes of the zip3-4AQ mutant. Meiotic progression, spore viability (97%, 149 tetrads) and kinetics of DSB formation and repair were as in wild-type (Figure 5A and data not shown). At centromeres and axis sites, Zip3-4AQ was normally recruited. However, at the three tested DSB sites, loading of mutant Zip3 was 2- to 3-fold reduced in comparison to wild-type Zip3 (Figure 5B). Thus, the Mec1 consensus phosphorylation sites of Zip3 are important for its localization or stabilization on recombination intermediates. The reduced recruitment of Zip3-4AQ may result in lower CO frequencies. Indeed, in the EST3-FAA3 interval flanking a strong DSB site on chromosome 9, fewer COs were formed in the zip3-4AQ mutant than in the wild-type ZIP3 strain (Figure 5C and 5E). To test whether COs were reduced also at other loci, we performed tetrad analysis in a strain that contains genetic markers on chromosome 3,7 and 8 to measure the genetic distances in three intervals per chromosome. Genetic distances were significantly reduced in three of the nine intervals tested, demonstrating the effect of the zip3-4AQ mutation on CO frequency (Figure 5D and Table S1). The observation that the genetic distance was reduced at two intervals on chromosome 3 (the smallest chromosome tested) and at none on chromosome 7 (the largest chromosome) suggests that perhaps smaller chromosomes are more affected by the Zip3 mutation (Figure 5D). The residual association of Zip3-4AQ with DSB sites and the reduced CO frequency were still sufficient to promote full spore viability. We thus investigated whether the Zip3 S/T-Q motifs become essential for spore viability when DSBs are reduced. However, a mutant with reduced DSB levels did not show increased spore lethality when combined with the zip3-4AQ mutant (Figure S6). Finally, we hypothesized that the features of part of the COs in the zip3-4AQ mutant and of COs associated with wild-type Zip3 may be different. We thus measured CO frequency in the mus81Δ strain (wild-type Zip3), in which the alternative CO pathway is inactivated [32], and in the double zip3-4AQ mus81Δ mutants by physical analysis of the EST3-FAA3 DSB site with flanking markers. In our hands and at the hotspot examined, mutation of MUS81 did not affect CO formation in both strains, and CO was even slightly stronger in each case compared to its MUS81 counterpart (Figure 5E). We conclude that mutating Mec1/Tel1 consensus phosphorylation sites of Zip3 decreases its association with DSB sites and reduces CO frequency, and that the remaining CO are not dependent on the MUS81 pathway. In wild-type meiosis, Zip3 loading was not comparable at all DSB sites (see Figure S4). Specifically, although there was a high correlation between DSB and Zip3 sites at 4 and 5 hr after meiotic induction, Zip3 was enriched at DSB sites to various degrees (Figure S7). To test whether variations in Zip3 loading at DSBs correlated with changes in recombination frequencies, we chose DSB sites with differential Zip3 binding and flanked them with hemizygous recombination markers (Figure S8) to assess both DSB and CO frequencies. In the wild-type strain, we chose a DSB site with strong Zip3 enrichment (EST3-FAA3) and three sites with relatively lower Zip3 accumulation (ATG2-LAP3, COG7-LEU1 and ISF1-ADH3) (Figure 6A and Figure S7). The introduction of the flanking markers slightly lowered the DSB frequency in the interval (Figure S9) and we thus compared CO and DSB frequency in strains containing the flanking markers (Figure 6A and 6B and Table S2). The CO/DSB ratio varied among the sites and paralleled their relative Zip3 enrichment as measured on the ChIP-chip profiles: the three low-Zip3 DSB sites showed between 2. 5 and 5 times less COs per DSB than the EST3-FAA3 DSB site (Figure 6A and 6B). To investigate whether such differential loading could be observed also in a situation where the DSB profile and number were changed, we compared the genome-wide maps of DSBs and Zip3-Flag binding sites in the set1Δ strain, in which DSBs are reduced and redistributed to new sites [33]. ChIP followed by qPCR indicated that Zip3 localized at DSB sites at 6 h and 7 h after meiotic induction, as expected because DSB formation is delayed by about 2 hours in this strain [33] (Figure S10A). Conversely and like in the wild-type strain, few Zip3 binding sites coincided with Rec8 sites at the 6 and 7 h time-points (Figure S10B and S11). Moreover, like in wild-type cells, Zip3 loading onto DSB sites was variable. For instance, PES4, a strong set1Δ DSB site, was highly enriched in Zip3, whereas ARG3, another strong set1Δ DSB site, was not (Figure S10C). We flanked each of these two sites by hemizygous markers (Figure S8) and measured crossover frequencies. Similarly, like in the wild-type strain, the high-Zip3 PES4 site showed 2. 2 times more COs per DSB than the low-Zip3 ARG3 site (Figure S10C). These results are consistent with a positive effect of Zip3 loading on DSB repair by CO and shows that in the genome, there are DSB sites that are less bound by Zip3 and less frequently repaired by CO than the average. We then asked whether specific chromosome features were associated with these variations in Zip3 binding at DSB sites. We first investigated Zip3 loading at DSB sites close to centromeres as it was reported that inter-homolog CO frequency is usually low close to centromeres, although DSBs can form close to centromeres [7]. On several chromosomes, Zip3 did not bind to centromere-proximal DSBs (Figure S12) and, on average, the relative Zip3 signal at DSB sites close (less than 10 kb) to centromeres was significantly lower than in the rest of the genome (Figure 7A). To extend the analysis beyond centromere regions, we defined from our mapping data two categories of DSB sites. Among the 400 strongest DSB sites previously determined in the resection-proficient dmc1Δ strain (without DSBs at less than 10 kb from a centromere), we identified “low-Zip3” DSB sites (n = 166 sites) and “high-Zip3” DSB sites (n = 142 sites) (see Protocol S1 for details on the classification). In these two DSB populations, the mean DSB signal was not statistically different (Wilcoxon test, p = 0. 13). Similarly, several chromosome features, such as distance from a telomere or a centromere, and replication timing, were also not different (not shown). However, the strength of DSB signal measured in the resection-defective rad50S mutant was lower at low-Zip3 DSB sites than at high-Zip3 DSB sites [3] (Figure 6C and Figure 7B). Analysis of DSB formation by Southern blotting at the three low-Zip3 DSB sites ATG2-LAP3, ISF1-ADH3 and COG7-LEU1 (Figure 6D and 6E) and the low-Zip3 set1Δ DSB site ARG3 (Figure S10D) confirmed that at these sites fewer DSBs were detected in the rad50S than in the dmc1Δ background. By contrast, the high-Zip3 EST3-FAA3 and the high-Zip3 set1Δ PES4 DSB sites showed similar DSB frequency in both backgrounds (Figure 6D and 6E and Figure S10D). When we classified the DSBs in the rad50S mutant as high (157 sites) or low (113 sites), based on the peak signal intensity like we did for the Zip3 peaks, we found that the Zip3 signal was significantly lower in low-rad50S DSBs (Figure 7C). Overall, 66 DSB sites were present both in the low-Zip3 DSB and the low-rad50S DSB category, that is more than expected by chance (p<0. 01, Pearson' s Chi-square test). This further strengthens our observation that at least a subset of low-Zip3 DSB sites also shows reduced DSB formation in the rad50S mutant, suggesting that they have distinct properties. The second chromosomal feature that varied between high- and low-Zip3 DSB sites was the distance from an axis-associated site, defined as a Red1 peak (Figure 7B). Low-Zip3 DSB sites were significantly more distant from an axis site than high-Zip3 DSB sites (median distance from a Red1 peak: 5599 bp and 3660 bp, respectively). Conversely, no difference in the distance from an axis-associated site was observed between low and high rad50S DSB sites (Figure 7C). Furthermore, the low-Zip3 DSB sites that were NOT low rad50S DSBs were still much further away from an axis site than the high-Zip3 DSB sites (5709 bp and 3660 bp from a Red1 peak, Figure S13). We confirmed this observation in the set1Δ strain, in which the 200 strongest set1Δ DSB sites were classified as high- and low-Zip3 DSBs. High-Zip3 and low-Zip3 DSB sites did not show significant differences in their mean dmc1Δ DSB ChIP-chip signal (p = 0. 66), but the low-Zip3 DSBs were significantly further away from a set1Δ Rec8 peak or a Red1 peak than the high-Zip3 DSB sites (Figure S10E). Thus, we can distinguish two different categories of low-Zip3 DSB sites: sites with reduced DSB formation in the rad50S mutant and sites that are far from an axis-associated site, suggesting that proximity to an axis site favors DSB binding by Zip3 and resolution as a CO (see Discussion). Here we show that the ZMM protein Zip3 interacts dynamically with chromosomes, associating first with centromeres, then with chromosome axes upon DSB formation, and finally with DSB sites on the recombination intermediates engaged in CO formation. We thus propose that Zip3 is a molecular marker of CO sites. We then demonstrate that Zip3 association with chromosomes requires its SUMO E3 ligase motifs, thus implying that SUMO recognition and transfer are needed for Zip3 interaction with chromosomal proteins. Zip3 phosphorylation sites by Mec1/Tel1 kinase are also important for Zip3 full loading on DSBs and CO formation. Finally, we show the existence of DSB sites that are rarely bound by Zip3 and that produce fewer COs than the average of DSB hotspots. These low-Zip3 DSB sites are sensitive to the effect of the rad50S mutation and tend to be away from an axis-association site, where the recombination process takes place. A recent study showed that the proteins necessary for DSB formation reside on the chromosome axis, rather than at the sites of DSB formation in loop sequences [24]. This suggests that at the time of DSB formation, DSB hotspot sequences are already located on the chromosome axes. Indeed, using ChIP assays, we found that Zip3 first associates with axes and DSB sites, and later during the recombination process (when dHJs are formed at the pachytene stage) it becomes almost exclusively associated with DSB sites. We propose that at this stage, the recombination intermediates are located in the inter-homolog space and are detached from the axis, as previously seen cytologically in Sordaria [34]. Although recombination takes place close to the axis, axis-associated sites might be less immunoprecipitated by ChIP, because Zip3 is less intimately linked to these sites than to DSB sites. Our ChIP analysis of Zip3 localization in yeast mutants that affect defined steps of recombination indicates that DSB formation is sufficient to trigger Zip3 localization at axis sites, whereas Zip3 associates with DSB sites only when they are engaged in dHJ intermediates. Our results are in apparent contrast with previous cytological findings about Zip3 foci in various mutants. In the rad50S mutant, many Zip3 foci co-localized with Mre11, which associates with DSB sites in this strain [20], [35]. However, we found that Zip3 does not associate with DSB sites in this mutant. The previously described foci could correspond to Zip3 loading on chromosome axes where Mre11-enriched DSB sites may also be located in the rad50S mutant. Similarly, Zip3 foci were previously detected in the dmc1Δ mutant [36], whereas in our study Zip3 was normally associated with axis sites, but very little with DSB sites. This was not due to experimental artifacts due to a differential ability to immunoprecipitate Zip3 in these mutants, since we observed constant Zip3 recovery during the whole time-courses after immunoprecipitation (data not shown). These discrepancies underscore the complementarity between ChIP approaches and cytological studies and show that similar patterns of foci can underlie completely different protein localizations along chromosomes, as revealed by our study. The early Zip3 association with axes following DSB formation could be due to Zip3 binding to cleaved DSB sites that are located on the axis, or to a generalized recruitment of Zip3 on chromosome axes, maybe through interaction with a protein phosphorylated upon DSB formation. Our ChIP-chip data favor the second explanation because axis sites close to strong DSB sites were not more enriched in Zip3 and Zip3 binding to axes was rather homogenous along chromosomes (data not shown). The protein responsible for Zip3 loading onto axis sites could be an axis protein that is phosphorylated by the Tel1/Mec1 kinases, such as Hop1 [37]. We observed a reduced recruitment of Zip3 to all chromosomal regions in the zip1Δ mutant. It was proposed that at centromeres, Zip1 stabilizes Smt3 chains, made by other SUMO ligases acting in early meiosis, thus favoring Zip3 binding to centromeres. Our data confirm previous cytological observations [38] and suggest that Zip3 loading at centromeres may be a consequence of Zip1 localization at centromeres early in meiosis. Indeed, Zip1 association with centromeres is Zip3-independent and early centromere coupling mediated by Zip1 does not require Zip3 [39]. Our results in the zip3 SUMO ligase and the zip1Δ mutants are consistent with a previously proposed model [18]: after the initial Zip3 recruitment to DSBs, which requires its SUMO binding motif (our results), Zip1 binds to and stabilizes the Smt3 chains deposited by Zip3. This in turn induces a second wave of Zip3 recruitment to DSB sites via its SUMO binding motif [18]. Indeed, in the zip1Δ mutant, Zip3 association with DSB sites was strongly decreased. Interestingly, Zip3 foci persisted more on DSB sites in the ndt80Δ mutant than in the wild-type. The ndt80Δ mutant accumulates non-cleaved dHJs and thus our data are consistent with the proposed role of Zip3 and the ZMM in general to stabilize the crossover-designated intermediates from D-loop dismantling and later from dHJ dissolution by activities exerted by anti-crossover factors such as Sgs1 [40]. Strikingly, Zip3 association with the axis site reached very high levels in ndt80Δ cells. This may be due to a change of structure within the synaptonemal complex that persists in this mutant and that alters the association of sites undergoing dHJ with axis-associated sites, and renders these closer to strong DSB sites and thus more closely associated with Zip3. It would be interesting to determine if this increase of Zip3 association is seen for all axis-associated sites or only those that are close to strong DSB sites. We detected a negative influence of the centromere on the relative binding of Zip3 to DSB sites. However, Zip3 binding was not abolished, although these regions show few CO and NCO events and have been suggested to repair their DSBs mostly using the sister chromatid [7]. A previous study showed that during DSB repair by sister chromatid recombination, the formation of associated joint molecules still depends on the ZMM protein Msh4 [2]. Similarly, we found that when a DSB is forced to be repaired using the sister chromatid, it still binds to Zip3, albeit to a lesser extent than when it is repaired by the homolog (unpublished results). Thus, DSBs might bind to Zip3 also very close to centromeres if they form joint molecules with the sister chromatid, explaining why we see residual Zip3 association with DSB in these regions. In the rest of the genome, we detected qualitative differences among DSB sites. Specifically, for a chosen set of sites, we show that the CO frequency per DSB can vary from one DSB site to another and that this behavior can be predicted based on the relative Zip3 enrichment at the site. These DSB hotspots have peculiar properties: they form DSBs at a lower frequency in the rad50S mutant (our results and [3]) and they tend to overlap with coding regions (our results and [4]). Previous studies showed that in an artificially late replicating chromosomal region, meiotic DSBs also formed later. Interestingly, DSB formation at these sites is affected in rad50S mutants [41]. In the rad50S mutant DSB formation is impaired at many regions [3] and by extension these could be naturally late-occurring DSBs. Indeed, these “low-rad50S” DSBs tend to occur later, but the asynchrony of meiotic time-courses makes it difficult to reproducibly detect a delay ([3] and data not shown). Based on these data, we can hypothesize that the low-Zip3 DSBs that we have studied are naturally late-forming DSBs. This would imply that in a given chromosomal region, early-forming DSBs are the preferred substrate for CO designation. Indeed, CO designation is a very early event, much earlier than Zip3 association, which we defined as a CO marker in this study. Upon early DSB formation, the CO designation of one DSB might relieve the chromosomes from the experienced stress, thus locally disfavoring further CO designations and explaining CO interference [12], [42]. Thus, a DSB formed later in this region will have little chance to be chosen as a CO event. We also found that besides the rad50S effect on DSB frequency and the possibly associated differential timing of DSB formation, low-Zip3 DSBs are more distant from an axis-associated site. For their repair, and likely also for their formation, DSB sites interact with the chromosome axis, particularly where the Red1 and Hop1 proteins reside, and cytological studies showed that the association between Zip3 and Hop1/Red1 occurs prior to SC polymerization, likely at the future CO sites [36]. We propose that a DSB site away from the axis will be less efficiently brought or kept on the axis, making it less favorable for CO designation. Our data have important implications for the control of meiotic recombination and genetic distances at the level of DSB formation and repair outcome. It will be interesting to investigate whether the DSB sites with low CO frequency we identified are NCO hotspots being repaired via the homolog or if they are repaired via the sister chromatid and whether they are preferential binding sites for anti-crossover activities. These extra-DSB sites rarely repaired as crossover may be either used early for homolog pairing, which precedes crossover formation, or conversely, they may be later “safety” DSB made in case insufficient early DSB go into crossover (Figure 7D). Our work paves the way for further studies in other organisms, especially in mammals where the number of DSB largely exceeds that of COs. All yeast strains (Table S4) are derivative of the SK1 background. They were produced by direct transformation or crossing to obtain the desired genotype. Details of strain construction are in Protocol S1. All transformants were confirmed to have the flanking marker at the correct locus by PCR analysis to discriminate between correct and incorrect integrations. Synchronous meiosis in liquid culture was performed as described [43]. Progression through meiosis was monitored by scoring nuclear divisions after DAPI staining. Western blotting was performed as described [23] using the mouse monoclonal anti-FLAG antibody M2 (Sigma, 1∶1000), except for detecting phosphorylated Zip3 (Figure 4 and Figure S6) where samples were separated in 10% 150∶1 acrylamide-to-bisacrylamide gels. Dephosphorylation assays were carried out as described [18], using calf intestinal alkaline phosphatase in the presence or not of 20 mM of the phosphatase inhibitor sodium orthovanadate. For genetic distances on chromosomes III, VII and VIII, haploids were mated at 30°C on YPD supplemented with 1% Adenine for 5 hr before replica-plating on solid sporulation medium (1% potassium acetate) and incubated at 30°C for at least 48 hr. For recombination between hemizygous drug resistance markers, diploids were grown on YPD plates and then replica plated on sporulation medium at 30°C for at least 48 hr. Asci were dissected on YPD supplemented with 1% Adenine and replica-printed to the appropriate media to check for marker status. P (parental), NPD (non-parental) and T (tetratype) were scored to calculate the genetic distances as described in [44]. For calculation of the map distance, standard error calculations were performed using the Stahl Lab Online tools (http: //www. molbio. uoregon. edu/~fstahl/). For calculation of the ratio between CO and DSB per cell, we divided the % of cells that received a CO (genetic distance * 2) by the % of cells that received a DSB (%measured DSB frequency * number of chromatids per cell, i. e. 4). Genomic DNA was prepared, analyzed and DSB or CO frequency was determined as described [23]. The used restriction enzymes and probes are in Protocol S1. For each time-point, cells were processed and ChIP was performed as described [23], using 2 µg of the mouse monoclonal anti-FLAG antibody M2 (Sigma) and 30 µL Protein G magnetic beads (New England Biolabs). Quantitative PCR was performed using immunoprecipitated DNA or whole-cell extracts as described [23]. DSB1, DSB2 and DSB3 sites were chosen according to the genome-wide mapping of [3]. DSB1 is in the promoter of BUD23 on chromosome 3; DSB2 is in the promoter of ECM3 on chromosome 15 and DSB3 in the promoter of RIM15 on chromosome 6. Axis site was chosen from the Rec8 binding data of [45], on chromosome 3. Negative control site is neither a DSB site nor a Rec8 site, and is located in the promoter of CDC39, on chromosome 3. Primer positions are in Protocol S1. All time-courses and ChIP assays were repeated at least twice from independent experiments and gave similar results. Immunoprecipitated DNA and whole-cell DNA were amplified, labeled and hybridized to Agilent 44 k yeast whole genome oligonucleotide arrays as described [33]. Microarray images were read using an Axon 4000B scanner and analyzed using the GenePix Pro 6. 0 software (Axon Instruments). Files were converted to text files and analyzed using the R software. The signal intensities of profiles were normalized, by dividing all values by the mean of the lowest 10% ratio probes of the array (decile normalization, as described [24]). In this way, the 10% lowest values fall below 1, so that everything below and around this value can be interpreted as background. The resulting normalized data were next denoised and smoothed, as described before [23]. Raw data from [33], [3] and [24] were reanalyzed as described before [23]. Peaks were identified after denoising and smoothing with a 2 kb window (except for the data by [24], where a 300 bp window was used), and compared as described [23]. In the set1Δ Zip3-Flag 6 and 7 hr ChIP-chip assays, a very high signal was obtained, and we adjusted the threshold to 5 to obtain a number of Zip3 peaks comparable to that of the other experiments. High Zip3 DSB sites were DSB sites that coincide with a Zip3 peak the signal intensity of which differed by less than 50 ranks from that of the DSB site; Low Zip3 DSB sites were DSB sites either not bound by Zip3 or that coincide with a Zip3 peak the signal intensity of which was at least 100 ranks lower than that of the DSB site. For the chromosome coordinates, we used the Saccharomyces Genome Database features (http: //downloads. yeastgenome. org/curation/chromosomal_feature/) of the last update from July of 2010. The ChIPchip data generated in this study have been deposited at the Gene Expression Omnibus database, accession number GSE40563. Processed data for all chromosomes are provided in Table S3.

For sexual reproduction, meiosis is an essential step ensuring the formation of haploid gametes from diploid precursors of the germline. This reduction in the genome' s content is achieved through a specialized type of division, where a single round of DNA replication is followed by two successive rounds of chromosome segregation. The first round separates the homologs. For this to faithfully occur, homologous chromosomes pair with each other and experience recombination, catalyzed by the formation of programmed double-strand breaks (DSBs). Upon their repair, a subset of DSBs will generate crossovers, which result from an intermediate that creates a physical link between homologs and allows their correct segregation by the meiotic spindle. DSBs, as well as crossovers, do not occur randomly along chromosomes but at preferential places called hotspots. To ask if all DSB hotspots also give rise to high crossover frequency, we have systematically compared the map of DSBs with that of a protein, Zip3, which we show preferentially binds to DSB sites that are being repaired with a crossover. We discovered that several DSB hotspots rarely produce crossovers, meaning that the decision to repair a DSB with a crossover can be influenced by specific chromosomal features.

Abstract Introduction Results Discussion Materials and Methods

biology

2013

Differential Association of the Conserved SUMO Ligase Zip3 with Meiotic Double-Strand Break Sites Reveals Regional Variations in the Outcome of Meiotic Recombination

Schistosomiasis is endemic to many regions of the world and affects approximately 200 million people. Conventional adaptive T cell responses are considered to be the primary contributors to the pathogenesis of Schistosoma japonicum infection, leading to liver granuloma and fibrosis. However, the functional polarization of macrophages and the associated underlying molecular mechanisms during the pathogenesis of schistosomiasis remains unknown. In the present study, we found that excretory-secretory (ES) antigens derived from S. japonicum eggs can activate macrophages, which exhibit an M2b polarization. Furthermore, ES antigen-induced M2b polarization was found to be dependent on enhanced NF-κB signaling mediated by the MyD88/MAPK pathway in a TLR2-dependent manner. In addition, the cytokine profile of the liver macrophages from wild-type-infected mice are quite distinct from those found in TLR2 knockout-infected mice by quantitative PCR analysis. More importantly, the size of granuloma and the severity of the fibrosis in the livers of TLR2-/- mice were significantly reduced compared to that in WT mice. Our findings reveal a novel role for M2b polarization in the pathogenesis of schistosome infection. Schistosomiasis is one of the most important health problems in developing countries[1], and can be used as a chronic disease model for investigating the interplay between the immune response and parasite pathogenicity in the host[2]. Following a schistosome infection, the host immune response gradually switches from a predominant Th1 response to a Th2-dominated response following egg deposition [3]. The resulting Th2 cytokine secretion contributes to the development of hepatic fibrosis and portal hypertension[4,5]. A lipid fraction from Schistosoma mansoni eggs containing lysophosphatidylserine (lyso-PS) has been shown to induce dendritic cell (DC) activation that promotes Th2 and regulatory T-cell development via a Toll-like receptor (TLR) 2-dependent mechanism[6]. Moreover, soluble S. japonicum egg antigens (SjEA) can upregulate programmed death ligand 2 (PD-L2) expression on bone marrow-derived dendritic cells (BMDCs) in a TLR2-dependent manner to help inhibit T cell responses in S. japonicum infected mice[2]. More importantly, these data indicate that interactions between host TLRs and pathogen-associated molecular patterns (PAMPs) from schistosome eggs can initiate a Th2-biased immune response and contribute to the egg-induced immunopathology observed in schistosomiasis. Specialized pattern recognition receptors (PRRs) that recognize PAMPs, and the activation of such PRRs leads to an immediate innate immune response to infection and can profoundly influence the development of an adaptive immune response[7]. Among these PRRs, TLRs are type-1 transmembrane glycoproteins that can identify particular PAMPs and danger associated molecular patterns (DAMPs) [8]. TLRs are well-known to defend against pathogen invasion by triggering innate immune responses and subsequently priming adaptive immunity against infections, including Gram-positive and negative bacteria, as well as fungi, viruses, and parasites[9]. However, some TLRs can trigger a suppressive immune response through the binding of various ligands, which can help avoid excessive inflammation and develop chronic course of the disease, especially in helminth infections[10]. TLRs are expressed on various immune cells, including T cells, B cells, dendritic cells (DCs), and macrophages[11]. TLR engagement results in the activation of the mitogen-activated protein kinases (MAPKs), which, together with the NF-κB pathway, induce extracellular signaling to initiate specific cellular responses[12]. Macrophages, the most abundant mononuclear phagocytes in the human body, are heterogeneous, versatile cells that can undergo dynamic switches in phenotype or function in response to microenvironmental signals[13]. Functional macrophage polarization represents different extremes of a continuum ranging from M1, M2a, and M2b to M2c [14], which can cause different cell populations to display differential gene expression and distinct functions [15]. M1 polarization, driven by IFN-γ and LPS, typically acquires fortified cytotoxic and antitumoral properties, whereas M2 polarization generally obtains immunoregulatory activities, tissue repair, and remodeling[16]. In particular, M2a polarization is induced by IL-4 and IL-13, whereas M2b polarization, induced by immune complexes and TLR or IL-1R agonists, is characterized by an IL-10high and IL-12low phenotype, exerts immunoregulatory functions, and drives Th2 responses. In contrast, M2c polarization, induced by IL-10 and glucocorticoid hormones, results in immunosuppression and tissue-remodeling activities[14]. Although the critical role of macrophage activation in the pathogenesis of schistosomiasis has been validated[17], the precise phenotype and mechanism associated with functional macrophage polarization in schistosomiasis remains unclear. In this study, we identified a novel role for macrophages in liver pathogenesis using a S. japonicum-infected mouse model and present TLR2 signaling as a novel potential therapeutic target for schistosomiasis. All animal experiments were performed in strict accordance with the Regulations for the Administration of Affairs Concerning Experimental Animals (approved by the State Council of People’s Republic of China), and efforts were made to minimize suffering. All procedures performed on animals in this study were approved by the Laboratory Animal Welfare & Ethics Committee (LAWEC) of National Institute of Parasitic Diseases (Permit Number: IPD-2016-7). Female C57BL/6 mice (6- to 8-weeks-old) were purchased from the SLAC laboratory (Shanghai, China). TLR2-/- mice[18] were provided by Dr. Xiao-Ping Chen from the School of Medicine, Tongji University. All mice were maintained under specific pathogen-free conditions and fed with standard laboratory food and water. Gender and age-matched mice were infected percutaneously with 20 ± 1 cercariae of S. japonicum, which were shed from infected Oncomelania hupensis snails provided by the National Institute of Parasitic Diseases in Shanghai, China. S. japonicum eggs were isolated from the livers of female rabbits 6 weeks following infection with 800 − 1000 cercariae via abdominal skin penetration. ES antigens were prepared as described previously with modifications[19]. The collected eggs were washed twice in serum-free DMEM supplemented with 100 U/mL penicillin and 100 μg/mL streptomycin. The eggs were then resuspended in 24 mL DMEM, and 2 mL aliquots were placed in six-well culture plates (Corning, USA). The culture medium was harvested after 48 h and centrifuged for 10 min at 200 × g to remove eggs, and 10000 × g to remove any debris. The protein concentration of ES antigens was determined using a Bradford assay. The endotoxin level of ES antigens was <0. 03 EU/mL as determined by a Limulus amoebocyte lysate assay (Genscript, China) according to the manufacturer’s instructions. ES antigens were stored at -80°C until further use. To destroy the lipid structures, ES antigens were digested with phospholipase C (Sigma, USA) at 37°C for 12 h, followed by heat inactivation of the enzymes at 100°C for 10 min. To digest proteins, ES was treated with proteinase K (Sigma, USA) at 56°C overnight, followed by heat inactivation of the enzymes at 100°C for 10 min. Mock-treated ES was also performed by heat at 100°C for 10 min without the addition of enzymes. Protein disruption was regularly checked by SDS-PAGE and viewed by silver staining. BMDMs were prepared as previously described with modifications[20]. Briefly, bone marrow cells were isolated from the leg bones of wild-type and TLR2-/- mice and cultured in DMEM (Gibco, USA) supplemented with 10% FBS (Gibco, USA) and 50 ng/mL macrophage colony-stimulating factor (M-CSF) (Peprotech, USA) and maintained in a 5% CO2 incubator at 37°C. Six days after the initial BM cell culture, the medium was changed, and the purity of F4/80+ cells was > 99%, as determined by flow cytometry. In some experiments, BMDM cells (5 × 105 cells/mL) were pretreated with one of the following inhibitors: 10 μM BAY 11–7082 (NF-κB inhibitor, Beyotime biotechnoogy, China), 10 μM SP 600125 (JNK MAPK inhibitor, Beyotime biotechnoogy, China), 1 μM SB 203580 (p38 MAPK inhibitor, Beyotime biotechnology, China), 10 μM PD 98059 (ERK MAPK inhibitor, Beyotime biotechnology, China) and 10 μM ST 2825 (MyD88 homodimerization inhibitor, MedChem Express, USA). Furthermore, LPS from Escherichia coli serotype O111: B4 (Sigma, USA) and synthetic lipoprotein Pam3CSK4 (InvivoGen, USA) were used in some experiments. The total RNA was extracted from macrophages using Trizol reagent (Invitrogen, USA) and reversed transcribed using a cDNA reverse transcription kit (Takara, Japan). The reverse-transcribed cDNA was used as a template in qPCR reactions containing SYBR Green Real-time PCR Master Mix (Takara, Japan) and 0. 4 μM forward and reverse primers. Relative mRNA expression was calculated using the 2-ΔΔCt method and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primer sequences were prepared as previously described[16,21] and are listed in S1 Table. BMDMs were stimulated with different antigens for 24 h. The levels of IL-1β, IL-12p70, IL-6, MCP-1 (CCL2), TNF-α, and IL-10 in the supernatants were detected by ELISA (eBioscience, USA). Quantification of IFN-γ, IL-4 and IL-13 in the serum sample was also determined by an ELISA in accordance with the manufacturer’s instructions (eBioscience, USA) and expressed as pg per mL. Macrophages (5 × 105 cells/mL) were stimulated with various doses of ES (0. 1 μg/mL, 1 μg/mL, and 10 μg/mL), LPS (100 ng/mL), and Pam3CSK4 (4 μg/mL) for different time points. The treated cells were washed twice with PBS and then lysed for 30 min on ice in a RIPA solution containing a protease inhibitor cocktail and phosphatase inhibitors (Sigma, USA). The expression of proteins in the cell lysates were examined using anti-NF-κB phospho-p65, anti-phospho-JNK, anti-phospho-p38, and anti-phospho-ERK1/2 antibodies (Cell signaling technology, USA). Anti-GAPDH (Sungene Biotech, China) was used as an internal control. Statistical analysis was performed for band intensities and evaluated using image J (NIH, USA). The cellular suspension of liver leukocytes was prepared using the traditional method according to previous reported methods with modifications [22]. In brief, following a perfusion of 3 mL PBS via the portal vein, mouse liver fragments were pressed through a 70-μm cell strainer (BD, USA). The total liver cells were then resuspended in a 40% Percoll solution (GE Healthcare, USA), and centrifuged for 20 min at 800 × g. The leukocytes were resuspended in an erythrocyte-lysing buffer. The cells were washed and resuspended in a MACS separation buffer (Miltenyi Biotec, Germany), and anti-F4/80 microbeads (Miltenyi Biotec, Germany) were used to isolate macrophages from leukocytes. The purity of the isolated cells was confirmed at > 95%. In some experiments, liver macrophages with a purity of approximately 95% were used as the starting material for ES antigen stimulation. To assess the expression of activation and other biological markers on macrophages, flow cytometry was performed with FITC labeled anti-F4/80, APC labeled anti-Gr-1, Brilliant Violet 421 labeled anti-CD11b, PE labeled anti-MHC class II, PE-Cy7-labeled anti-CD40, PE-Cy7-labeled anti-CD80, APC-labeled anti-CD86, PE-labeled anti-CD16/32 and APC-labeled anti-mannose receptor (CD206) (Biolegend, USA). All flow cytometry data was acquired on an LSRFortessa X-20 (BD Biosciences, USA) and analyzed with FlowJo software (Tree star, USA). Fresh liver tissues were fixed in 4% formaldehyde overnight and routinely paraffin embedded. Paraffin sections (5 μm) were prepared from each liver tissue sample. H&E staining of liver tissue sections were performed according to the manufacturer’s instructions and assessed by a pathologist blinded to the treatment group. The liver tissue sections were stained with Masson’s trichrome staining to evaluate collagen content and distribution. The collagen fibers were stained blue, the cell nuclei were stained black, and the background was stained red. Each stained section was examined by optical microscopy with 100 × magnification and identical settings. Thirty pictures were taken of granulomas around single eggs from three sections in each tissue. Every picture was evaluated in a double-blind fashion by two independent investigators. The area featuring granulomas and fibrosis surrounding single eggs was evaluated using image J (NIH, USA). Data represented as the mean ± SEM were analyzed by a two-tailed Student’s t-test, or a one-way or two-way ANOVA using GraphPad Prism version 5. 0 (GraphPad Software, USA). Significant differences were accepted when the p-value was less than 0. 05. To confirm that ES stimulation could induce macrophage polarization, BMDMs were stimulated with different concentrations of ES in vitro. Enhanced CD86 expression was observed in the BMDMs that were treated with 0. 1 μg/mL or 1 μg/mL of ES compared with that of the control group (Fig 1A). However, CD206 (mannose receptor) and CD16/32 expression did not increase after ES stimulation (Fig 1A). Moreover, compared with 0. 1 μg/mL ES, stimulation with 1 μg/mL ES was found to upregulate MHC class II (I-A/I-E) and CD86 expression but downregulate CD80 expression. However, compared with 1 μg/mL of ES stimulation, 10 μg/mL of ES stimulation was found to downregulate MHC class II, CD16/32, and CD86 expression (Fig 1A). RT-qPCR analysis showed that BMDMs stimulated with ES antigens exhibited enhanced IL-6, IL-10, and Arg-1 mRNA levels in a dose-dependent manner (Fig 1B). Furthermore, 1 μg/mL of ES antigen stimulation increased the levels of IL-6, TNF-α, MCP-1, IL-10, and Arg-1 mRNA expression but decreased the levels of iNOS and Ym1 mRNA compared with 0. 1 μg/mL ES antigen stimulation. However, RT-qPCR revealed that the levels of TNF-α and IL-12 mRNA expression were downregulated in BMDMs stimulated with 10 μg/mL of ES antigens compared to BMDMs stimulated with 1 μg/mL of ES antigens (Fig 1B). Similar to these results, significantly higher levels of TNF-α, IL-12p70, MCP-1 (CCL2), and IL-1β were observed in the supernatants of BMDMs stimulated with 0. 1 μg/mL ES compared with that of the control macrophages (Fig 1C). Moreover, stimulation with 1 μg/mL ES antigens upregulated IL-6, IL-1β, and IL-10 expression but downregulated IL-12p70 expression compared with 0. 1 μg/mL ES antigen stimulation. However, compared with BMDMs stimulated with 1 μg/mL of ES antigens, the levels of IL-1β, TNF-α, MCP-1, and IL-6 were downregulated in the supernatants of BMDMs stimulated with 10 μg/mL ES (Fig 1C). Surprisingly, unlike LPS, ES stimulation did not increase iNOS protein expression but enhanced the production of Arg-1 protein (Fig 1D). According to the reported secretory products and biological surface markers associated with macrophage polarization[16], these data suggest that 1 μg/mL of ES antigen stimulation promotes TNF-α, IL-1β, IL-6, and IL-10, as well as promotes MHC class II and CD86 expression, whereas there are low levels of IL-12 production. These findings indicate that macrophages treated with 1 μg/mL ES display an M2b-polarized phenotype in vitro. Previous reports indicate that MAPK-NF-κB signaling contributes to macrophage activation[23,24]. To determine the expression pattern of MAPKs and NF-κB on activated macrophages, a Western blot was performed to analyze the levels of phospho-p38, phospho-p65, phospho-ERK, and phospho-JNK. As shown in Fig 2A, BMDMs treated with ES exhibited increased levels of phospho-p38, phospho-p65, phospho-ERK, and phospho-JNK expression in activated macrophages in a dose-dependent manner in vitro. As shown in Fig 2B, the RT-qPCR analysis revealed that BMDMs stimulated with ES antigens exhibited significantly decreased levels of IL-6, TNF-α, IL-10, MCP-1, and iNOS mRNA expression compared with the control BMDMs upon treatment with PD 98059 (ERK1/2 MAPK inhibitor), SP 600125 (JNK MAPK inhibitor), or BAY 11–7082 (NF-κB inhibitor). However, ST 2825 treatment markedly increased the levels of IL-6, RELMa, and Ym1 mRNA but decreased the levels of TNF-α, MCP-1, and Arg-1 mRNA compared with control group. Furthermore, SB 203580 (p38 MAPK inhibitor) treatment significantly decreased the levels of IL-10, MCP-1, and Arg-1 mRNA but increased the levels of RELMa mRNA compared with control BMDMs. Similar to these results, an ELISA was performed to assess the production of inflammatory markers in the supernatants of BMDMs exhibited remarkably decreased production of IL-6, MCP-1, TNF-α, and IL-10 upon treatment with PD 98059 (ERK1/2 MAPK inhibitor), SP 600125 (JNK MAPK inhibitor), or BAY 11–7082 (NF-κB inhibitor) (Fig 2C). However, BMDMs treated with ST 2825 exhibited significantly decreased levels of IL-10 and MCP-1. Furthermore, BMDMs treatment with SB 203580 (p38 MAPK inhibitor) displayed increased levels of IL-6, MCP-1, and TNF-α expression but remarkably decreased levels of IL-10 expression. These data demonstrate that the MyD88/MAPK/NF-κB signaling pathway facilitates ES-induced M2b polarization. Multiple previous studies have reported TLR2 to be an important PPR for soluble egg antigens (SEA) [2,25]. To further evaluate the role of the TLR2 receptor on macrophage M2b polarization induced by ES, BMDMs derived from wild-type and TLR2 knockout (KO) mice were stimulated with ES in vitro. As shown in Fig 3A, compared with the BMDMs from wild-type mice, BMDMs derived from TLR2 KO mice stimulated with ES did not exhibit increased protein levels of phospho-p38, phospho-p65, phospho-ERK, and phospho-JNK in a dose-dependent manner. However, BMDMs derived from TLR2 KO mice stimulated with a high dose of ES could increase the levels of phospho-ERK protein expression. Treatment BMDMs derived from TLR2 KO mice with ES failed to induce Arg-1 production (Fig 3B). Compared with that of BMDMs from wild-type mice, the ELISA analysis for inflammatory marker production in the supernatants of BMDMs from TLR2 KO mice showed that the levels of IL-6, MCP-1, TNF-α, and IL-10 markedly decreased (Fig 3C). More importantly, RT-qPCR analysis for inflammatory gene expression showed that liver macrophages purified from wild-type mice stimulated with 1 μg/mL of ES exhibited enhanced IL-10, TNF-a, IL-1β, IL-6, IL-12, MCP-1, Arg-1, and iNOS but reduced the RELMa and Ym-1 mRNA levels compared with liver macrophages purified from TLR2 KO mice (Fig 3D). This indicates that similar to BMDMs (S1 Fig), liver macrophages can be activated by ES antigens in a TLR2-dependent manner. Our results show that TLR2 is the pivotal receptor for M2b polarization following ES antigen stimulation. Previous studies have reported schistosome-specific lysophosphatidylcholine (lyso-PS) in SEA activated TLR2[6,26]. To clarify whether the lipids in ES contribute to the production of pro-inflammatory cytokines by macrophages, RT-qPCR was used to assess the levels of IL-6, TNF-α, IL-10, MCP-1, and iNOS mRNA, which were found to be decreased in BMDMs stimulated with both proteinase K-treated ES and phospholipase C-treated ES, compared with mock-treated ES (Fig 4A). However, phospholipase C-treated ES treatment increased the levels of Arg-1 mRNA but decreased Ym1 mRNA in BMDMs compared with mock-treated ES. Furthermore, proteinase K-treated ES enhanced the levels of RELMa and Ym1 mRNA but reduced Arg-1 mRNA compared with mock-treated ES. Similar to these data, an ELISA was used to further assess the levels of TNF-α, IL-6, IL-10, and MCP-1 in the supernatants of BMDMs. As shown in Fig 4B, compared to the mock-treated ES, the proteinase K-treated ES as well as phospholipase C-treated ES failed to induce BMDMs to produce high levels of TNF-α, IL-6, IL-10, and MCP-1. Thus, lipids or lipid conjugates contribute to M2b polarization. Although previous studies on schistosomiasis often focus on T and B lymphocytes, APCs (e. g. , macrophages) may play vital roles in the pathogenesis of the disease [27]. To evaluate the role of macrophages during the pathogenesis of liver granuloma formation and fibrosis in a murine model of schistosomiasis, liver leukocytes were isolated and analyzed for the presence of CD11b+F4/80+ cells. The absolute number and percentage of eosinophils (SSChighCD11b+F4/80+ cells) and infiltrated macrophages (SSClowCD11b+F4/80+ cells) were substantially increased in the liver issues of the infected mice (Fig 5A). Four weeks post-infection, a real-time PCR analysis of inflammatory gene expression revealed that purified macrophages from infected WT mice exhibited enhanced IL-10, Arg-1, MCP-1, RELMa, and IL-6 but reduced TNF-α, IL-12, and iNOS mRNA levels, which exhibited an M2 dominant- polarized phenotype (Fig 5B). However, compared with WT mice, the purified macrophages from infected TLR2-/- mice exhibited higher levels of TNF-α, IL-12, and iNOS mRNA, as well as lower levels of IL-10, Arg-1, and RELMa mRNA, which represents a dominant M1-polarized phenotype (Fig 5B). Moreover, a significant decrease in the percentage of infiltrated macrophages and neutrophils (CD11b+Gr-1+ cells) recruitment was observed in infected TLR2-/- mice, compared with that of wild-type mice at 6 weeks post-infection (Fig 5A and Fig 5C). More importantly, the area of granuloma formation and fibrosis surrounding single eggs in the livers of TLR2-/- mice were significantly lower compared with that of WT mice at 6 weeks post-infection (Fig 5D). Due to their inherent plasticity and heterogeneity, the ability of macrophages to functionally switch from killing pathogens to the promotion of tissue repair is likely critical for the host, especially when the host cannot eradicate a persistent infection but must limit tissue damage (e. g. , chronic helminth infection) [17]. The differential activation status of macrophages has the capability to promote, restrict, or resolve inflammation and fibrosis in schistosomiasis [17,27,28]. In the present study, we aimed to investigate the phenotypical and functional plasticity of various macrophage subtypes following infection with S. japonicum (e. g. , the presence of ES antigens derived from S. japonicum eggs). BMDMs stimulated with ES exhibit an activation status characterized by the production of multiple pro-inflammatory cytokines and abundant anti-inflammatory IL-10 production in vitro, following the activation of several MAPKs downstream of both TLR2 and MyD88. Several helminth antigens are known to be associated with the stimulation of various TLRs[29,30], whereas schistosoma antigens have been reported to interact with specific TLRs present on mononuclear phagocytes[25,31–33]. However, few of these studies describe the signaling pathways triggered in host cells or link them to the production of specific cytokines[25,33,34]. In the current study, we demonstrated that ES-stimulated BMDMs produce pro-inflammatory cytokines and the anti-inflammatory cytokine IL-10 in a dose-dependent manner via TLR2. Moreover, we found that lipids or lipid conjugates participate in macrophage polarization, since the levels of cytokines decreased significantly in the supernatants of BMDMs induced by phospholipase C-treated ES. More importantly, our data confirmed that BMDMs exposed to the ES of eggs were dependent on the phosphorylation of three MAPK cascades (p38, ERK1/2, and JNK1/2), an event often reported downstream of TLR activation[35]. Notably, these kinase cascades occur with identical activation profiles, leading to the phosphorylation of p65 reported to be downstream of p38, ERK1/2, and JNK1/2. However, the full details of these signaling pathways has not previously been reported in the context of ES antigens. Moreover, in the present study, we provide the complete molecular mechanism of specific cytokine production by macrophages in response to ES antigens. M2 macrophages can express Arg1, an enzyme adapted from the urea cycle which converts L-arginine to L-ornithine, enabling L-ornithine amino-transferase (OAT) to supply proline for collagen synthesis[36]. Although alternative macrophage activation is induced by IL-4/13 production, Arg1-expressing macrophages can downregulate inflammation, suppress Th2 cytokine production and reduce tissue scarring and pathology[37]. In addition, the anti-inflammatory cytokine, IL-10, plays a central regulatory role in the pathogenesis of schistosomiasis, and the maintenance of IL-10 expression during acute and chronic schistosome infection is critical for host survival[38]. In this study, TLR2 was found to be the key PRR associated with the production of pro-inflammatory cytokines, IL-10, and Arg-1 expression from BMDMs stimulated with ES in vitro. Surprisingly, we observed a significantly higher mRNA levels of IL-6, MCP-1 (CCL2), IL-10, and Arg-1 in the liver macrophages of infected wild-type mice compared to that of TLR2-/- mice, which was similar to M2b polarization in vitro. This finding suggests that ES antigens could induce the polarization of liver macrophages in vivo. However, the levels of IL-6, MCP-1 (CCL2), IL-10, and Arg-1 mRNA were found to be downregulated at 6 weeks post-infection. Thus, since similar serum levels of IL-4 and IL-13 were enhanced in both WT and TLR2 KO mice (S2 Fig), type II immunity may help to regulate levels of inflammatory cytokine mRNA in this study. Furthermore, the mRNA levels of RELMa were decreased in liver macrophages from TLR2 KO mice compared with that in WT mice at 4 weeks post-infection, which seemed different from that on stimulation with ES in vitro. Higher serum levels of IFN-γ that promoted M1 polarization and inhibited M2 polarization were observed in TLR2 KO mice compared with WT mice at 4 weeks post-infection (S2 Fig), thus type I immunity may help to downregulate the expression of RELMa. More importantly, IL-6 can recruit both neutrophils and monocytes during inflammation [39]. Moreover, MCP-1 accelerates liver fibrosis by promoting Ly-6C+ macrophage infiltration [40]. Our data indicate that TLR2 can increase the population of M2-type macrophages and neutrophil infiltration in the liver, perhaps because higher levels of MCP-1 (CCL2) and IL-6 are produced by ES-stimulated macrophages. Taken together, these data help explain the increased granuloma size and greater collagen deposition in the liver of wild-type mice. However, the mechanisms by which ES antigen-stimulated macrophages promote liver pathology require further investigation. In summary, we have presented a detailed study of the molecular events that occur in BMDMs following exposure to ES antigens derived from eggs, which led to the production of pro-inflammatory cytokines and IL-10. This mechanism involves the activation of p65 downstream of TLR2 via the phosphorylation of p38, ERK1/2, and JNK1/2. Additionally, we showed that this mechanism is also responsible for the immunoregulatory activity in schistosomiasis, which provides an early in vitro demonstration of the role of ES Ags in modulating immunopathology downstream of TLR2. Finally, we observed that the production of IL-6, IL-10, MCP-1, and Arg-1 by macrophages in response to ES extends beyond an in vitro phenomenon and is also evident in the liver macrophages of WT-infected mice, which rapidly produce high levels of IL-6, IL-10, MCP-1, and Arg-1 mRNA in vivo following egg deposition. This early and rapid release of IL-6, IL-10, MCP-1, and Arg-1 has the potential to greatly modulate the immune response to limit inflammation and tissue damage in the liver by conditioning the microenvironment. More importantly, M2 polarization dependent on enhanced TLR2 signaling would reduce the level of liver damage and promote fibrosis. While our data suggest that targeted TLR2 signaling inhibitors may have therapeutic effect during the acute phases of schistosomiasis, further study is still required to address the role of TLR2 signaling to better understand the potential benefits of its inhibitors in treatment of chronic disease and delineate novel insight into the immune interplay underlying schistosomiasis. For this reason, the results generated from this study might provide evidence supporting the necessity to include TLR2 signaling as a novel therapeutic target for schistosomiasis.

Schistosomiasis is a global health concern that affects primarily tropical and subtropical areas. During a schistosome infection, the eggs are trapped in the host liver and products derived from eggs induce a polarized Th2 response, resulting in granuloma formation and eventually fibrosis. Thus, it is important to elucidate the mechanism of granuloma formation and fibrosis development. Here, we show that activated macrophages play a novel role in the promotion of hepatic granuloma formation and liver fibrosis in a Schistosoma japonicum-infected mouse model. In addition, M2b polarization induced by egg products was dependent on enhanced NF-κB signaling mediated by the MyD88/MAPK pathway in a TLR2-dependent manner. Our findings reveal a novel role and mechanism of M2b polarization in the liver pathogenesis in S. japonicum-infected mice.

Abstract Introduction Methods Results Discussion

blood cells innate immune system medicine and health sciences immune cells immune physiology cytokines granulomas immunology tropical diseases parasitic diseases liver diseases animal models immune receptor signaling developmental biology model organisms membrane receptor signaling gastroenterology and hepatology experimental organism systems molecular development neglected tropical diseases research and analysis methods immune system proteins white blood cells animal cells animal studies proteins liver fibrosis mouse models immune system toll-like receptors biochemistry helminth infections schistosomiasis signal transduction cell biology physiology biology and life sciences cellular types immune receptors macrophages cell signaling

2018

Toll-like receptor-2 regulates macrophage polarization induced by excretory-secretory antigens from Schistosoma japonicum eggs and promotes liver pathology in murine schistosomiasis

Malaria parasites have been shown to adjust their life history traits to changing environmental conditions. Parasite relapses and recrudescences—marked increases in blood parasite numbers following a period when the parasite was either absent or present at very low levels in the blood, respectively—are expected to be part of such adaptive plastic strategies. Here, we first present a theoretical model that analyses the evolution of transmission strategies in fluctuating seasonal environments and we show that relapses may be adaptive if they are concomitant with the presence of mosquitoes in the vicinity of the host. We then experimentally test the hypothesis that Plasmodium parasites can respond to the presence of vectors. For this purpose, we repeatedly exposed birds infected by the avian malaria parasite Plasmodium relictum to the bites of uninfected females of its natural vector, the mosquito Culex pipiens, at three different stages of the infection: acute (∼34 days post infection), early chronic (∼122 dpi) and late chronic (∼291 dpi). We show that: (i) mosquito-exposed birds have significantly higher blood parasitaemia than control unexposed birds during the chronic stages of the infection and that (ii) this translates into significantly higher infection prevalence in the mosquito. Our results demonstrate the ability of Plasmodium relictum to maximize their transmission by adopting plastic life history strategies in response to the availability of insect vectors. All organisms experience some level of temporal variation in the quality of their environment. In response to these variations, many species have evolved specific strategies that allow them to limit or shut down growth and development until the conditions improve [1]. The best reported examples are dormancy in plants and diapause in insects, but similar strategies have also evolved in microbes. Bacteria can survive adverse conditions (e. g. desiccation, antibiotics) by entering a state of reduced metabolic activity called persistence [2], [3]. Several viruses (e. g. lambdoid phages, herpesviruses) have evolved the ability to integrate their host genome and enter a latent phase during which within-host replication is shut down, the infection is asymptomatic and transmission is very limited [4], [5]. Hence, the evolution of latent life cycle in pathogens may be viewed as an adaptation to temporal variations of the availability of susceptible hosts. For vector-borne pathogens the abundance of vectors is a key parameter determining the quality of their environment. Vector density may vary in space due to intrinsic heterogeneities of their habitat (e. g. temperature, hygrometry). In malaria, for instance, spatial variation in mosquito abundance has a direct impact on the geographic distribution of prevalence [6]–[8]. Vector abundance may also vary widely through time [9]. Although inter-tropical regions are characterized by a relatively constant density of vectors, regions from higher latitudes experience a broad range of climatic seasonality, and very far from the equator mosquitoes are present for only a fraction of the year [10]–[12]. From the parasite' s perspective, such temporal variation in vector density is analogous to the temporal variations in habitat quality experienced by other organisms. How have malaria parasites adapted to these temporal fluctuations in vector density? Malaria is caused by Plasmodium spp. , a prevalent vector-borne pathogen which is found infecting many vertebrate hosts, including humans, reptiles and birds. Plasmodium infections within the vertebrate host are characterized by drastic temporal changes in blood parasitaemia. After an initial acute phase, generally characterized by a very high number of parasites in the blood, the infection usually reaches a chronic phase where the parasitaemia stabilizes at low levels. During the chronic phase, however, blood parasites may go through short, intense, bouts of asexual replication during which parasitaemia increases temporarily. Little is known about the causes of such recrudescences but one potential trigger may be a weakening of the host' s immunity [13]. In some, but not all, Plasmodium species the infection may entirely disappear from the blood stream, hiding in other host cells in the form of (dormant) exoerythrocytic stages. After a period of latency that can last months or even years, parasites may reappear in the blood stream. These relapses are due to the differentiation of dormant parasite stages into new erythrocytic stages. The dormant stages of Plasmodium were first described in birds [14], [15] and, later, in humans [16], [17] and reptiles [18], [19]. Relapses and recrudescences have been puzzling researchers ever since the first clinical symptoms were described in P. vivax-infected humans in the late 19th century [20], [21]. Why do some malaria species (e. g. P. falciparum) completely lack the ability to produce dormant stages in the vertebrate host? What are the ultimate causes of the production of recrudescences and relapses? Is this diversity of life cycles due to the temporal variation in vector density? The ability to produce recrudescences and relapses may be a genetically fixed parasite strategy that has evolved as a way to match the dynamics of vector populations. Populations exposed to different fluctuations of vector density may thus evolve different strategies. In human malaria, the relapsing periodicity of different lineages of P. vivax supports this prediction [12], [22]. Indeed, lineages exhibiting frequent relapses have been sampled in Asia where the vector is present throughout the year. In contrast, long latency has been observed in lineages sampled in temperate zones where the mosquito vector is only present for a few months. In avian malaria, similarly, the differences in the within-host dynamics of Leucocytozoon spp. and Haemoproteus mansoni may have evolved to match the temporal fluctuations of their respective vector species (simuliid flies and Culicoides, respectively) [23]. Another explanation for these patterns may involve adaptive phenotypic plasticity. Phenotypic plasticity is the ability for a single genotype to exhibit different phenotypes in different environments [24], [25]. This contrasts with the above hypothesis (fixed strategy) where different relapsing strategies are associated with different genotypes. The ability to adopt a plastic exploitation strategy requires the ability to detect a change of the environment (i. e. cues) and the acquisition of such a sensing mechanism may be associated with direct fitness costs [24], [25]. In spite of these costs, phenotypic plasticity is often viewed as an adaptation to variable environments [24], [25]. Many pathogens have indeed evolved an unparalleled level of phenotypic plasticity in their life history traits to cope with the temporal variability of their habitat [26]–[28]. In Plasmodium, plasticity has been shown to be a response to various stressful conditions such as drug treatment and the presence of competitors [29], [30]. Some experimental evidence suggests that relapses may also be a plastic trait. P. vivax relapses are often observed in the spring and summer months irrespective of when the patients got the original infection [31], which suggests that the parasite may react to a change in the physiological state of the host or the environment. Relapses have also been observed in avian malaria, which has triggered several experimental studies to pinpoint the underlying environmental cues [32]. Some authors have proposed that spring relapses may result from increasing photoperiod and/or stress-induced hormonal changes [33]–[36]. Parasites may indirectly benefit from using hormonal and photoperiod cues because they often coincide with (or even anticipate) the appearance of vectors in temperate populations. Such indirect cues are, however, imperfect because vector abundance may be influenced by other, non-seasonal, factors. A more efficient strategy would be to react to direct cues such as mosquito bites which unambiguously indicate the presence of vectors [10], [31], [37]. Although there is some correlational evidence supporting this hypothesis, largely coming from longitudinal cohort studies [10], [37], [38], direct experimental evidence for this hypothesis is scarce and somewhat contradictory. In rodent malaria P. chabaudi, mice exposed to probing by Anopheles stephensi mosquitoes had higher and earlier parasite growth and gametocytaemia than control unexposed mice [39]. In contrast, however, Shutler et al. [40] found no evidence of facultative alteration in the timing or in the level of P. chabaudi or P. vinckei parasitaemia and gametocytogenesis as a consequence of mosquito probing. Rodent malaria is a laboratory model and, as such, may, not be the best system to test this hypothesis because An. stephensi is not the natural vector of rodent malaria [41]. In addition the parasites have been originally sampled from the tropical lowlands of the Congo Basin [42] an area where malaria transmission is high throughout the year [43] and thus the selective pressure for the evolution of plasticity in response to vector availability is expected to be weak. Finally, both rodent malaria experiments [39], [40] were carried out during the initial (acute) phase of the infection, i. e. when parasitaemia is already high (so no need to increase it further) and the infection recent (so the mosquitoes are probably still around). We contend that it is mainly in old (chronic state) infections that the parasite may accrue the greatest benefits from a plastic response to the bites of its vector. Finally, both of these studies used gametocyte density (the blood stages of Plasmodium that are transmissible to the vector) as a proxy for transmission but neither followed transmission all the way to the mosquito stage. Here, we first present a theoretical model that studies the evolution of parasite transmission in a variable environment. This model explores the effects of the seasonality of mosquito dynamics on the evolution of virulence and transmission strategies. In particular it clarifies the selective pressures acting on the evolution of temporally variable transmission strategies and identifies the conditions driving the evolution of costly plastic transmission strategies triggered by the exposure to mosquito bites. Then, we carry out an experiment to test the following two hypotheses: (1) Plasmodium parasites plastically react to the biting of uninfected vectors by enhancing their within-host replication, and (2) this effect yields higher rates of transmission to the mosquito vector. For this purpose, we studied the interaction between Plasmodium relictum (the aetiological agent of the most prevalent form of avian malaria which is commonly found infecting Passeriform birds in Europe) and its natural vector, the mosquito Culex pipiens. P. relictum is a very convenient malaria parasite to address this issue because it is known to have a long chronic phase marked by sudden events of recrudescences and relapses [44]. Strictly speaking, relapses originate from the division and differentiation of dormant stages (called phanerozoites) that infect the endothelial cells of different organs such as the spleen and liver, while recrudescences originate from an increased replication of the blood stages [44]. In practice, however, it is very difficult to distinguish between recrudescences and true relapses and in the following we will use the term relapse to encompass both cases. We investigate whether bites of uninfected Cx. pipiens mosquitoes trigger parasite relapses in the blood of domestic canaries (Serinus canaria) chronically infected by P. relictum (lineage SGS1), as well as the concomitant effects on transmission in terms of mosquito infectivity (see Box 1 and Fig. 1). To model the evolution of plastic transmission strategies we first need to model the epidemiological dynamics of malaria. For the sake of simplicity the vertebrate host population is assumed to be constant and equal to N = S (t) +I (t), where S (t) and I (t) are the densities of uninfected and infected hosts, respectively. In contrast, the density of the vector population may change through time. This may be particularly relevant in temperate environments where the mosquitoes do not reproduce in winter. In other words, the influx θ (t) of uninfected mosquitoes is assumed to change throughout the year (i. e. θ (t) fluctuates with a period T). As a consequence, the densities V (t) and VI (t) of uninfected and infected vectors also fluctuate through time. The following set of differential equations describes the temporal dynamics of the different types of hosts (the dot notation indicates differential over time): (1) Where d is the natural mortality rate of the vertebrate host and α is the virulence of malaria (the extra mortality induced by the infection); m and mI are the mortality rates of uninfected and infected vectors, respectively; β1 is the transmission rate from the vertebrate host to the vector; β2 is the transmission rate from the vector to the vertebrate host. Figure 2 presents a typical epidemiological dynamics in a seasonal environment. What are the consequences of the temporal variation in the availability of vectors on the evolution of malaria? More specifically, what is the effect of the shape of the function θ (t) on the evolution of the parasite? To study this question one can consider the fate of a mutant malaria strategy M that would alter its life history strategy in the vertebrate host. The replication in the vertebrate host is assumed to be governed by two traits of the parasite. The first trait, εF, governs the allocation to a fixed exploitation strategy that yields a within-host growth rate that does not vary with time. In contrast, the second trait, εP, governs allocation to a plastic exploitation strategy that may vary with time. More specifically we consider that when the parasite adopts this plastic trait within-host growth rate depends on the density of vectors in the population. In other words this plastic trait allows within-host growth rate to be conditional on the presence of vectors. Because within-host growth rate is assumed to affect virulence in the vertebrate host this yields: (2) As in classical models of virulence evolution [45], [46] more replication is costly because it may induce the death of the vertebrate host but it allows the parasite to transmit more efficiently. The parameters a and b govern the specific shape of the virulence-transmission trade-off (see equation (3) below). In addition we assume that the adoption of a plastic exploitation strategy requires the ability to acquire information regarding the availability of the vectors. The parameter c, therefore measures the fitness cost associated with a higher investment in the mechanisms allowing such plasticity. Only the transmission rate, β1, from the vertebrate host to the vector is assumed to be affected by the parasite exploitation strategy (i. e. β2 is assumed to be constant) which yields: (3) Note that in this model virulence and transmission vary in time only if the parasite allocates some resources in the development of a plastic trait (i. e. εP>0). Integrating the change in frequency of the mutant parasite genotype M over one period of the fluctuation allows deriving a condition for the invasion of the mutant (see Text S1): (4) where, , and. The first term in the above equation for sM is the classical benefit associated with higher investment in transmission. If the mutant invests more than the resident in transmission (i. e. Δβ1>0) the fitness increase depends on, which measures the availability of both uninfected hosts and vectors over the period of the fluctuation of the environment. The second term in sM is the classical cost of virulence if the mutant exploits the host more aggressively than the resident (i. e. Δα>0). The final term in sM measures the potential benefit associated with plastic transmission strategies. This term depends on the covariance between the availability of uninfected vectors, the availability of uninfected vertebrate hosts and the investment in transmission from vertebrate to mosquito hosts. The mutant may gain a fitness advantage if its conditional transmission rate can better track the fluctuations of the density of uninfected hosts. In other words this final term indicates that in a fluctuating environment it is adaptive to invest on transmission whenever uninfected hosts and mosquitoes reach high densities simultaneously. We can use this analysis to look at different evolutionary scenarios. The experimental design is presented in Box 1. In brief, we followed 20 experimentally infected birds over 300 days post infection and monitored within-host parasitaemia and transmission to vectors. Birds were assigned to two treatments: “exposed” or “control” (unexposed) to uninfected mosquito bites during 3 sessions (starting 34,122 and 291 days post infection, see Fig. 1A). During each session the exposed birds were bitten by a batch of 50 female mosquitoes every 3 days (see Fig. 1B). How do malaria parasites adapt to the density fluctuations of their insect vectors? To answer this question we started by studying the evolution of transmission strategies using a classical epidemiological model for a vector-borne pathogen. This theoretical approach helps clarify the multiple effects of temporal fluctuations of vector populations. We first considered the evolution of a fixed allocation to virulence and transmission. Our analysis shows that the effect of the temporal variation is driven by its effect on the average density of susceptible hosts and vectors over one period of the fluctuation. In particular we show that in more seasonal environments (e. g. higher latitudes), where the vectors can pullulate only for a few months, lower levels of virulence and transmission should be selected. This is because, in our model, seasonality reduces the average number of vectors. In the absence of the vector, investing in transmission becomes maladaptive because within-host reproduction is associated with higher virulence and host death. This result is very similar to the effect of periodic host absence on the evolution of phytopathogens when there is a trade-off between pathogen transmission and pathogen survival [50]. In addition, our predictions agree with recent models studying the effect of seasonality on virulence evolution [51], in that if the fluctuations of vector density do not affect the mean density of susceptible vectors over time, we expect no evolutionary consequences. Interestingly, our prediction on the effect of seasonality (Fig. 3) is consistent with the geographical distribution of relapsing strategies in P. vivax [22]. P. vivax genotypes sampled near the equator (where seasonality is minimal) invest in higher transmission strategies (higher rates of relapse) than P. vivax genotypes sampled in higher latitudes. In other words, in P. vivax malaria latitude is a very good predictor of the rate of relapses (Fig. 9). In a second step of the analysis we allowed plastic transmission strategies to evolve. In particular, we assumed that the malaria pathogens have the ability to sense the density of vectors through exposure to mosquito bites. We derived the condition promoting the evolution of such plastic behaviours when investment in this strategy is associated to a direct fitness cost on transmission. Koelle et al. [52] derived a similar result in a model of pathogen adaptation to seasonal fluctuations but without highlighting the force driving adaptive plasticity. Kumo and Sasaki [53] showed that the sensitivity to seasonality in a directly transmitted pathogen is driven by the correlation between the seasonal variation in transmission rate and the density of susceptible hosts. In our model the sensitivity to seasonality is governed by the fluctuation of mosquito density and plasticity. Similarly we show that what selects for plasticity is the covariance between transmission and the availability of hosts (both the vertebrate hosts and the vectors). In other words, plasticity evolves when mosquito bites provide accurate information on the availability of susceptible hosts. Cohen [54] obtained very similar results on the evolution of conditional dormancy strategies in randomly varying environments. The evolution of conditional dormancy depends on the correlation between the cue and the quality of the environment for individuals leaving the dormant state [54] (see also [55], [56]). In our model the correlation between the cue (mosquito bites) and the abundance of susceptible hosts depends on seasonality: as expected, plasticity evolves more readily when mosquitoes are only around for a short period of time. Have malaria parasites evolved the ability to respond plastically to mosquito bites on its vertebrate host? Previous work on acute rodent malaria infections has produced somewhat contrasting results [39], [40]. These earlier studies had in common that (i) they used an unnatural mosquito-Plasmodium combination, (ii) they were carried out using parasites collected in a high-transmission tropical environment several decades ago and (iii) were carried out when the infection is already at its highest level within the vertebrate host. Here we use a natural mosquito-Plasmodium combination to test the effect of mosquito bites on parasite transmission during the chronic phase of the infection. We used a P. relictum lineage (SGS1) which had been sampled from wild house sparrows in 2009 in a high latitude habitat (Dijon, France) where the environment is characterized by marked seasonal patterns, including variations in mosquito prevalence [57]. In addition, rather than inferring parasite transmissibility solely from the host' s parasitaemia, we also quantified the number of parasites that made it all the way to the gut (oocyst) stages of the biting mosquitoes. Our experiment confirmed our two main predictions. First, P. relictum SGS1 reacts to mosquito bites by increasing its overall parasitaemia in the blood. As expected, this effect was not present during the acute infection (first exposure session) because transmission is always high at this stage, but became apparent during the chronic stage of the infection (second and third exposure sessions). Second, this increased parasitaemia resulted in higher probability of infection to mosquitoes and thus in higher transmission rates. The results were consistent at the chronic stage of the infection (exposure sessions 2 and 3): there was a significant increase in mosquito infection rate after exposure to mosquito bites. Blood stage malaria infections comprise both asexual (replicating) and sexual (transmissible) stages. However, the molecular tools used to quantify overall parasitaemia in this study did not allow us to distinguish between these two parasite life stages. In other malaria systems the conversion rate between the asexual and the sexual (gametocyte) stages, and the resulting sex ratio of the gametocytes may be highly plastic [30], [58], so that overall parasitaemia may not necessarily be a good predictor of gametocyte density and/or transmission. Although nothing is yet known about the conversion rates or sex allocation strategies in P. relictum, in our experiment the increase in parasitaemia was accompanied by a significant increase in the number of infected mosquitoes, suggesting a concomitant increase in gametocyte density in birds exposed to mosquito bites. However, to directly test this hypothesis, we compared the gametocytaemia of exposed and unexposed birds by counting the visible gametocytes in the thick blood smears taken after the exposure (see Text S2). Contrary to expectations, however, we found no clear and consistent evidence that mosquito bites result in higher gametocytaemia. One potential explanation of this lack of consistency is an error in our estimate of gametocytaemia. The application of molecular techniques for the quantification of gametocytes has indeed called into question the use of microscopic methods to quantify Plasmodium gametocytes [59]. In particular, these studies have shown that submicroscopic gametocyte densities are common [60], [61] and can readily infect mosquitoes [62]. Unfortunately these molecular tools are currently only available for P. falciparum and P. chabaudi, and no equivalent tools exist to estimate gametocytaemia in P. relictum. A potential caveat of these results is that all the mosquitoes used in the same exposure session emerged roughly at the same date. As a consequence, females from the second and third mosquito batches were 3 and 6 days older (respectively) than mosquitoes used in the first batch. To control for a potential confounding effect of female age on transmission we therefore carried out another experiment using females of identical age (7 days old) at each exposure session. This experiment was carried out using laboratory (SLAB strain) mosquitoes and although the oocyst infection intensities were overall lower, we obtained qualitatively similar effects as in the main experiment (see Text S3). In addition, earlier studies have found either that age has little effect on mosquito vector competence [63] or that older mosquitoes have a lower prevalence and intensity of infection than their younger counterparts [64]. Hence, the increase in oocyst prevalence and intensity observed in consecutive exposure sessions in our main experiment cannot be explained by differences in the age of the mosquitoes used. The proximal mechanism governing this form of plasticity remains to be investigated. How do parasites in the blood or in tissues perceive mosquito bites? A plethora of substances and molecules present in the salivary fluid are injected when mosquitoes probe and feed [65]. The primary role of these molecules is to combat host homeostasis and to regulate inflammation at the biting site to facilitate blood uptake. Vector salivary lysates have been shown to stimulate within-host growth of Leishmania parasites [66] and may also trigger plastic life-history strategies in Plasmodium. In addition, host anaemia, erythropoeisis, and asexual density have all been shown to be associated with the onset of gametocytogenesis in rodent malaria [67]–[70]. Shutler et al. [40] suggested that blood feeding mosquitoes may cause host anaemia thereby triggering gametocytogenesis in P. chabaudi. Our data, however, do not support this hypothesis, because birds exposed to mosquitoes had similar or even higher haematocrit than control birds (see Text S3). In addition, previous findings obtained using P. gallinaceum have shown that in this avian malaria parasite, parasitaemia and gametocytaemia are not affected by host anaemia [71]. The study of the mechanisms governing plastic transmission strategies in avian malaria is hampered by the lack of available molecular tools to quantify and sex gametocytes (e. g. [72] for rodent malaria). Mechanistic studies can reveal fascinating pathogen strategies. For instance, a recent study on the Cauliflower mosaic virus (CaMV) has shown that when aphids feed on the infected plants the virus reacts instantly (and reversibly) to maximize its transmission. For this purpose it modifies the distribution of a specialised set of proteins which are essential for virus transmission [73]. In the absence of the vector, these proteins, which are toxic for the plant, are neatly packed away inside specialised structures called “transmission bodies”. This study not only represents an excellent example of the ability of some vector-borne pathogens to adopt plastic transmission strategies but it also demonstrates the sophisticated molecular and cellular mechanisms that may be involved. We identified the conditions that promote the evolution of plastic transmission strategies in a fluctuating environment. In line with our theoretical predictions, we show that P. relictum has the ability to boost its own transmission during the chronic phase of the vertebrate infection after being exposed to mosquito bites. Whether this ability extends to other Plasmodium species and in particular to human malaria remains to be investigated. In P. vivax the data presented in Figure 9 indicates a strong effect of latitude (i. e. seasonality) on relapses and transmission. The role of plastic transmission strategies on this pattern is unclear but it deserves further investigation. This may help define better control strategies, with more specific recommendations on both spatial and temporal implementations of targeted interventions against malaria hotspots [84]. The study of plastic transmission strategies may also be relevant to many other pathogens that are known to alternate between acute and dormant phases such as varicella zoster virus [85] Herpes Simplex virus [86], Mycobacterium tuberculosis [87] and HIV [88]. Such dormant parasites pose considerable therapeutic challenges and much would be gained from understanding the cues underlying the switch between dormant and acute stages in these pathogens [89]–[91]. In conclusion, a better understanding of the ecological determinants as well as the evolutionary forces governing parasite relapses is not only of academic interest: it is also urgently needed to improve the efficacy of public health strategies. Animal experiments were carried out in strict accordance with the National Charter on the Ethics of Animal Experimentation of the French Government, and all efforts were made to minimise suffering. Experiments were approved by the Ethical Committee for Animal Experimentation established by the authors' institution (CNRS) under the auspices of the French Ministry of Education and Research (permit number CEEA- LR-1051). Plasmodium relictum (lineage SGS1) is a generalist parasite and the most prevalent form of avian malaria in Europe, infecting over 30 birds species in the order Passeriformes [44], [83]. Our strain was originally isolated from wild house sparrows caught in the region of Dijon (France) and maintained in the laboratory via passages to naïve canaries either by intraperitoneal injection or through the bite of infected Culex pipiens mosquitoes. Experiments were conducted with wild Cx. pipiens pipiens mosquitoes. Cx. pipiens is the natural vector of P. relictum in the wild [44], [92], [93]. Thousands of L3 and L4 larvae were collected from a single sewage treatment lagoon in the village of Triadou (20 km north Montpellier, France) using a hand net and reared till adulthood under standard laboratory conditions [94]. We used females 7,10 and 13 days after emergence that had had no prior access to blood, had been maintained on glucose solution (10%) since their emergence, and had been starved (but provided with water) for 6 h before the experiment. Experiments were carried out using (1-year old) domestic canaries (Serinus canaria). Prior to the experiments, a small amount (ca. 15–25 µL) of blood was collected from the brachial vein of each of the birds and used for molecular sexing [95], as well as to verify that they were free from any previous haemosporidian infections [96]. Twenty birds were experimentally inoculated on the 3rd July 2010 (day 0, see Box 1 and Fig. 1) by means of an intraperitoneal injection of ca. 50–100 µL of an infected blood pool. The blood pool was constituted of a mixture of blood from 8 infected canaries that had been inoculated with the parasite 12 days previously following standard laboratory procedures [97]. Note that unlike what happens in some Plasmodium parasites such as P. vivax, the artificial infection with P. relictum via the inoculation of infected blood containing merozoites does not prevent the formation of exoerythrocytic stages [44], [98]. One bird failed to get infected and the remaining infected birds were assigned to two treatments: “exposed” (n = 10) or “unexposed” (control, n = 9) to mosquito bites. This assignment was made by balancing the gender of birds and the magnitude in the peak parasitaemia during the acute phase between the two treatments. This experimental design thus allowed mosquitoes to both probe and blood feed on the birds, and in this respect it contrasts with previous designs where only probing was allowed [39], [40]. Exposure to mosquito bites took place in August 2010 (first exposure session), and repeated in November 2010 and April 2011 (second and third exposure sessions). Each of these exposure sessions consisted of 3 “exposure days” separated by 3 day intervals: days 34,37 and 40 post-infection (dpi) for the first exposure session, days 122,125 and 128 dpi for the second exposure session, and days 291,294 and 297 dpi for the third exposure session (Box 1 and Fig. 1A). In the morning of each exposure day, a small (ca. 15–25 µL) amount of blood was taken from the brachial vein of all (“exposed” and “control”) birds to quantify parasitaemia (see below). In the evening, birds allocated to the “exposed” treatment were placed inside a cage (dimensions L40×W30×H30 cm) with a batch of 50 uninfected female mosquitoes for 2 hours (8–10pm). Around 30 females blood fed on the birds during this time (see Table S3) which is close to available estimations in the wild [99]. Tables S2 and S3 provide the full details of the number of replications (number of blood fed mosquitoes, number of mosquitoes dissected) for each exposure session. To minimize host defensive behaviours that may alter the mosquito biting process during the assay, we immobilized birds in a specially designed PVC tube that rendered their legs accessible to the mosquitoes while protecting the rest of the body from the bites [97]. “Control” birds were placed in identical conditions but without mosquitoes. Immediately after each exposure, blood-fed mosquitoes from each cage (n = 10) and time point (3 exposure sessions, 3 days per session) were collected, isolated in a new cage, and maintained under standard laboratory conditions for 7 days. Fifteen haphazardly chosen mosquitoes were dissected to check for the presence (prevalence) and number (intensity) of oocysts in the midgut [94]. In each exposure session, two further blood samples were taken from all experimental birds, 3–4 days and 7–8 days after the last exposure day (days 44 and 48 dpi, days 131 and 135 dpi and days 300 and 304 dpi, for the first to third exposure sessions, Fig. 1B). For each exposure session we therefore obtained 5 different blood samples (red arrows in Fig. 1A and 1B). These blood samples were used to quantify total parasitaemia using previously published qPCR procedures [97] and gametocytaemia by microscopic examination (see Text S2). In addition, blood samples were taken at regular intervals throughout the experiment to monitor parasitaemia before and between the exposure sessions (blue arrows in Fig. 1A). The statistical analyses were run using the R software (v. 2. 14. 0). Analyses were carried out separately for each exposure session. Variation in parasitaemia (log-transformed (RQ+1) ) was analyzed using linear mixed-effect models (lme function, nlme package) with bird as a random effect to account for the repeated sampling of individual hosts. A generalized linear mixed-effect models GLMM (glmer function, lme4 package, binomial distribution) was carried out to study variation in gametocytaemia (proportion of gametocytes). Bird and time were included as random and fixed factors, respectively. Variation in the infection prevalence (proportion of individuals harbouring at least 1 oocyst) and the oocystaemia (number of oocysts, only for infected mosquitoes) was analysed using GLMMs (glmer function, lme4 package, with binomial and Poisson distributions, respectively). Bird and time (i. e. time between the 5 different blood samples, see red arrows in Fig. 1b) were included as random and fixed factors, respectively. Here, time was considered as a factorial explanatory variable. When appropriate, a posteriori contrasts were carried out by aggregating factor levels that did not significantly differ from each other and by testing the fit of the simplified model [100]. The significance of explanatory variables was established by comparing the change in deviance with and without the term to a χ2 distribution. Degrees of freedom correspond to the difference in the number of terms in the model.

Seasonal fluctuations in the environment affect dramatically the abundance of insect species. These fluctuations have important consequences for the transmission of vector-borne diseases. Here we contend that malaria parasites may have evolved plastic transmission strategies as an adaptation to the fluctuations in mosquito densities. First, our theoretical analysis identifies the conditions for the evolution of such plastic transmission strategies. Second, we show that in avian malaria Plasmodium parasites have the ability to increase transmission after being bitten by uninfected Culex mosquitoes. This demonstrates the ability of Plasmodium parasites to adopt plastic transmission strategies and challenges our understanding of malaria epidemiology.

Abstract Introduction Results Discussion Materials and Methods

evolutionary ecology ecology and environmental sciences medicine and health sciences infectious disease epidemiology population dynamics microbiology parasitic diseases plant science plant pathology theoretical ecology evolutionary adaptation population biology infectious diseases epidemiology evolutionary theory disease vectors ecology disease dynamics vector biology biology and life sciences malaria evolutionary biology evolutionary processes

2014

Evolution of Plastic Transmission Strategies in Avian Malaria

Mutations in peroxin (PEX) genes lead to loss of peroxisomes, resulting in the formation of peroxisomal biogenesis disorders (PBDs) and early lethality. Studying PBDs and their animal models has greatly contributed to our current knowledge about peroxisomal functions. Very-long-chain fatty acid (VLCFA) accumulation has long been suggested as a major disease-mediating factor, although the exact pathological consequences are unclear. Here, we show that a Drosophila Pex19 mutant is lethal due to a deficit in medium-chain fatty acids (MCFAs). Increased lipolysis mediated by Lipase 3 (Lip3) leads to accumulation of free fatty acids and lipotoxicity. Administration of MCFAs prevents lipolysis and decreases the free fatty acid load. This drastically increases the survival rate of Pex19 mutants without reducing VLCFA accumulation. We identified a mediator of MCFA-induced lipolysis repression, the ceramide synthase Schlank, which reacts to MCFA supplementation by increasing its repressive action on lip3. This shifts our understanding of the key defects in peroxisome-deficient cells away from elevated VLCFA levels toward elevated lipolysis and shows that loss of this important organelle can be compensated by a dietary adjustment. Peroxisomes are vesicular organelles originally discovered and described by C. De Duve as catalase-containing organelles important for the degradation of hydrogen peroxide [1]. Recently, it has become more and more apparent that they harbor much more complex metabolic functions, which are still incompletely understood. In mammalian cells, they are involved in the β-oxidation of very-long-chain fatty acids (VLCFAs), the formation of ether phospholipids (e. g. , plasmalogens), the catabolism of branched-chain fatty acids, the production of bile acids, polyamine oxidation, and amino acid catabolism [2]. VLCFAs (chain length of C22 and more) do not enter the mitochondria via the carnitine shuttle carnitine palmitoyltransferase I (CPT I) and thus cannot be β-oxidized for energy gain. Instead, VLCFAs are exclusively oxidized in peroxisomes. They (and to a lesser extent, long-chain fatty acids [LCFAs], which are, however, predominantly oxidized in mitochondria) enter the peroxisomes after activation into acyl-CoA, where they are shortened by the peroxisomal β-oxidation machinery. The resulting short-chain fatty acids (SCFAs) and medium-chain fatty acids (MCFAs) are transported out of the peroxisome via the carnitine-shuttles carnitine O-acetyltransferase (CRAT) and carnitine O-octanoyltransferase (CROT) and enter the mitochondrion via carnitine shuttle or thiolase-dependent transport. In the mitochondria, they are further oxidized to acetyl-coA and feed the tricarboxylic acid (TCA) cycle and the electron transport chain [3]. It is unclear whether and to what extent the shortened peroxisomal products contribute to mitochondrial energy production, but overall, the energy gain from VLCFAs is minor in comparison to S/M/LCFAs because of their low abundance [4]. In yeast cells, peroxisomes are required for the β-oxidation of MCFAs [5], since there they are the only site of fatty acid oxidation. The complex metabolic functions of peroxisomes are reflected by the multiform pathologic symptoms of peroxisomal biogenesis disorders (PBDs) of the Zellweger syndrome spectrum, whose study has greatly contributed to our understanding of peroxisomal function. PBDs are rare genetic diseases caused by mutations in one of the approximately 16 peroxin (PEX) genes, which are involved in the assembly and maintenance of peroxisomes. One major disease-causing effect is thought to be the accumulation of VLCFA-containing lipids, which occurs not only in PBDs but also in defects specific to the peroxisomal β-oxidation machinery, like X-linked adrenoleukodystrophy (X-ALD) or single enzyme defects, and is thought to be a result of abolished degradation of VLCFAs in peroxisomes. While accumulation of VLCFAs in these peroxisomal diseases is undisputed, there are still some doubts as to the exact underlying mechanism. Mice lacking a functional VLC-acyl-CoA synthetase, which is needed to activate VLCFAs in order to enter the peroxisome, show normal VLCFA levels, although peroxisomal β-oxidation is strongly reduced [6]. Similarly, the exact pathological effect downstream of accumulating VLCFAs remains elusive, although detrimental effects on mitochondria are postulated by many authors. Mitochondrial defects are indeed often observed in peroxisomal diseases [7–10], and in vitro assays show mitotoxic effects of VLCFAs added to cells [11,12]. However, the majority of accumulating VLCFAs in peroxisomal diseases is shown to be contained in lipids and not as free fatty acids [13], also called nonesterified fatty acids (NEFAs), in which form they were added to the cells. Furthermore, a number of authors observe normal mitochondrial function and morphology in VLCFA-enriched tissues carrying an X-ALD causing ATP-binding cassette family D transporter (ABCD1) mutation [14–16], which suggests other pathological mechanisms involved in mitochondrial dysfunction in PBDs. The peroxisomal biogenesis and assembly machinery is well conserved in Drosophila [17–22]. In the present study, we use a D. melanogaster Pex19 mutant [23] as a genetically tractable in vivo model system to elucidate the impact of peroxisomal deficiency on cellular and organismal metabolism. Pex19 is a predominantly cytoplasmic peroxisomal core factor and essential for both the import of peroxisomal membrane proteins (PMPs) and the de novo formation of peroxisomes [24,25]. Together with Pex3 and Pex16, it is responsible for the translocation of membrane proteins and membrane vesicle assembly [26]. Pex19 loss of function specifically leads to Zellweger syndrome, the severest form of PBDs. We have previously shown that major hallmarks of Zellweger syndrome are recapitulated in Pex19 mutant flies, like absence of peroxisomes, VLCFA accumulation, mitochondrial abnormalities, and severely decreased viability. Furthermore, we have identified increased free fatty acids as a mitochondria-damaging agent in these flies. Free fatty acids accumulate as a result from a metabolic shift toward maximal lipid catabolism with severely increased lipolysis, which is caused by altered hepatocyte nuclear factor 4 (Hnf4) activity [23]. Here, we identify an imbalance in fatty acid composition in Pex19 flies as a pathology-causing effect. While VLCFAs accumulate as expected in Pex19 mutants, M/LCFAs are reduced to an extent far surpassing the relative effect caused by increased VLCFAs. This reduction is especially pronounced for fatty acids of 12 and 14 carbons (C12: 0, C14: 0, C14: 1), and we here present evidence that this MCFA gap is involved in pathology progression in the absence of peroxisomes. Feeding a diet rich in C12- and C14-containing coconut oil (or other natural oils of similar composition) rescues Pex19 flies into adulthood and ameliorates their metabolic imbalance, which is not achieved by feeding LCFA-containing oils. Similar rescues of Pex2 and Pex3 mutants unequivocally prove that peroxisomal absence, and not Pex19 loss itself, is responsible for the metabolic phenotype observed in Pex19 mutants. We show that MCFA shortage results in a shift in subcellular localization of the ceramide synthase Schlank. Schlank was recently reported to harbor a secondary function as a transcription factor and was shown to repress lipase 3 (lip3) [27]. Consistently, we observe severely increased expression of lip3 expression in Pex19 mutants. The feeding of coconut oil results in normalizing the activity of Schlank, concomitant with reduced lip3 transcription, ultimately resulting in reduced lipolysis, reduced free fatty acid levels, and amelioration of the mitochondrial phenotype. Furthermore, we present evidence that this general pathological mechanism is also present in a Pex19-deficient patient cell line, thereby opening up a new path in the search for future therapies for PBDs and other peroxisomal diseases. As we described previously, flies carrying a deletion of the Pex19 gene locus (Pex19ΔF7, from here on referred to as Pex19 mutants) lose their peroxisomes during larval development and die predominantly during the pupal stage [23]. When analyzing the fatty acid profile of Pex19 mutant pupae, we found an unexpected shortage in M- and LCFAs in addition to the well-described accumulation of VLCFAs (SCFAs, MCFAs, LCFAs, and VLCFAs here refer to chain lengths of 4–8,10–14,16–18, and ≥20 C-atoms, respectively). Direct comparison of the absolute concentration of all fatty acid methyl esters (FAMEs) revealed a significant reduction of M/LCFAs, especially C12: 0 and C14: 0, while FAMEs of C20: 0, C24: 0, C26: 0, and C28: 0 are enriched (Fig 1A), which is consistent with our previous findings in larvae [23]. M/LCFAs can fuel mitochondrial ATP production and do not depend on chain shortening by the peroxisome before they can enter the mitochondrion and the TCA cycle. Since we suspected an energy deficit in Pex19 mutants, we sought to fill the gap in MCFAs by dietary modifications. Freshly hatched larvae were kept by default on a high-caloric diet (apple juice agar plates with fresh yeast paste) rich in both carbohydrates and proteins. To avoid problems with toxicity or food avoidance, we used natural oils as sources for fatty acids with varying chain length. In order to minimize adverse effects of a high-fat diet [28], we added only 5% of the different natural oils to the yeast paste and analyzed Pex19 mutant survival. We found that oils containing high amounts of lauric acid (C12: 0) and myristic acid (C14: 0) —like coconut oil, babassu oil, or palm kernel oil [29–31]—have a positive effect on the survival rate of Pex19 mutants, yielding 55% of adult flies as compared to 20% on control food. By contrast, oils containing mainly LCFAs or VLCFAs have no or negative effects on the survival of Pex19 mutants (Fig 1B). For further studies, we used coconut oil, which contains up to 60% lauric acid and up to 18% myristic acid [29,30]. To show that the fatty acid composition rather than vitamins or secondary plant compounds is responsible for the observed rescue effect, we fed purified triacylglycerols (TAGs) from coconut oil to Pex19 mutants, which also resulted in a significant increase in adult flies (Fig 1C). Surprisingly, addition of synthetic TAGs containing only C12: 0 or C14: 0 does not copy the rescue effect of coconut oil, suggesting either impurities from the manufacturing process or other adverse effects of synthetic TAGs, since they slightly reduced the survival of control flies (Fig 1C). However, upon removal of the mitotoxic preservative nipagin from the food, synthetic TAGs also showed a rescue effect, while removal of nipagin alone does not enhance the survival of Pex19 mutants (Fig 1C). Supplementation with 5% coconut oil does not only increase the number of adults hatching from the pupal case from 20% to 55% but also the number of adults that survive for more than 24 hours after hatching, from 9% to 29% (Fig 1D). These adult flies survived up to 3 weeks without further supplementation of coconut oil. The inability to break down VLCFAs is considered the main disease-mediating factor in cells without functional peroxisomes, and lowering VLCFA levels is the aim of most therapies for patients with defective peroxisome metabolism, mostly by inhibiting elongases, e. g. , with erucic acid–containing Lorenzo’s oil [32,33]. We asked whether the dietary administration of 5% coconut oil had a rescue effect by lowering the VLCFA content and reanalyzed FAMEs in coconut oil–fed larvae by gas chromatography/mass spectrometry (GC/MS). We found that the MCFA-enriched diet leads to an increase in lipids containing C12: 0 and C14: 0 in both control and Pex19 mutant animals, both absolutely and relatively (Fig 2A, supplemental file S1 Fig). The total amount of fatty acids as a measure for lipids is reduced in Pex19 mutants, while, upon coconut oil administration, it is elevated in both wild types and Pex19 mutants (supplemental file S1 Fig). The relative amounts of lipid-contained fatty acids with a chain length of C16: 0 are reduced, despite the fact that coconut oil contains up to 9% of C16: 0, which suggests that the fatty acids from the rescue diet alter the FAME profile instead of simply representing the additional fatty acids taken up from the M/LCFA-enriched diet (supplemental file S1 Fig). Unexpectedly, lipids containing VLCFAs from C22: 0 to C26: 0 remain at elevated levels in Pex19 mutants and even increase in control animals on coconut oil–supplemented food (Fig 2A, supplemental file S1 Fig). Despite this increase, coconut oil supplementation has a beneficial effect on the survival of wild-typic flies as well as Pex19 mutants (Fig 1D). We calculated the average chain length from the absolute concentration of FAMEs and found that administration of the MCFA-enriched diet reduces the chain length from an average of 15. 93 to 14. 94 carbons in wild types and from an average of 16. 27 to 15. 35 carbons in Pex19 mutants (Fig 2B), showing that food-derived fatty acids have an influence on overall fatty acid content in flies, consistent with overall changes in the lipidome according to food source [34]. Adult Pex19 mutants appeared weak and did not properly inflate their wings. We tested their climbing ability in a negative geotaxis assay as a readout for neurodegeneration and found that Pex19 mutants climb slower than wild-typic flies and that most of them do not climb at all. Upon coconut oil supplementation, the climbing speed of Pex19 mutants increases, and the number of flies which are not able to climb decreases (supplemental file S1 Fig). Since we were limited in the number of adults because of the early lethality of the Pex19 mutants, we also conducted a crawling assay as a readout for larval neurodegeneration. We found that w- controls and Pex19 mutants with a genetic rescue construct show similar crawling performance on control and coconut oil–supplemented food, whereas Pex19 mutant larvae have locomotion deficits. The low crawling performance of Pex19 mutants significantly improves upon coconut oil supplementation (Fig 2C). To further characterize neurodegeneration in Pex19 mutants, we analyzed apoptosis in brains of 1-day-old adults by staining them with Annexin V-FITC (fluorescein isothiocyanate) (Fig 2D–2G). We found that Pex19 mutants have higher numbers of apoptotic cells in the brain. Supplementation with coconut oil results in a marked decrease of apoptotic cells in Pex19 mutants (Fig 2F and 2G). This shows that neurodegeneration due to peroxisome loss can be prevented or slowed down by dietary administration of MCFAs. Peroxisome loss provokes mitochondrial swelling due to elevated lipolysis and free fatty acid levels, as was shown by tetramethylrhodamine ethyl ester (TMRE) and mitotracker green stainings, as well as on the ultrastructural level (transmission electron microscopy, TEM) [23]. In order to assess the impact of the MCFA rescue diet on mitochondria, we stained mitochondria in the Malpighian tubules of third-instar larvae with MitoTracker Red CM-H2XRos. This dye emits fluorescence after oxidation in the cell and thus detects reactive oxygen species (ROS). We found that tissue of Pex19 mutants contains high amounts of ROS, as indicated by high numbers of enlarged CM-H2XRos-positive mitochondria. Upon coconut oil administration, size and number of CM-H2XRos-positive mitochondria decrease (Fig 3A–3D). To investigate the impact of the MCFA diet on the metabolism of Pex19 mutant animals, we analyzed several genes encoding for metabolic enzymes (Fig 3E). We found that expression of the acid lipase lip3, which is highly up-regulated in Pex19 mutants, is reduced from 250-fold to 66-fold upon feeding of the rescue diet. Previously, we could show that lip3 expression correlates with free fatty acid accumulation [23]. Hnf4 target genes such as hexokinase C (HexC), yippee-interacting protein 2 (yip2), and acyl-CoA synthetase long-chain (Acsl) are up-regulated in Pex19 mutants [23]. Upon coconut oil supplementation, expression of the glycolysis enzyme HexC is reduced from 33-fold to 11-fold, and the acyl-CoA synthetase Acsl is reduced from 15-fold to 2-fold (Fig 3E). By contrast, the acetyl-CoA acyltransferase yip2 remains at high levels on rescue food, suggesting that mitochondrial β-oxidation stays at a maximum in Pex19 mutants on MCFA rescue food and that the transcriptional response upon coconut oil rescue is markedly different than upon genetic reduction of Hnf4 with reduced yip2 expression [23]. Isocitrate dehydrogenase (IsoDH), a target gene of the regulator of mitochondrial abundance and peroxisome proliferator–activated receptor gamma coactivator 1-α (PGC1-α) ortholog spargel [35], is slightly up-regulated in Pex19 mutants and remains at this level upon feeding of the rescue diet, while another spargel target gene, mt Complex I, is down-regulated and also remains at this level upon treatment with coconut oil (Fig 3E). Pex19 mutants have increased β-oxidation rates, probably in response to the high free fatty acid load, and presumably without gaining additional energy from it (Fig 3F and 3H). We measured the mitochondrial β-oxidation in wild-type and Pex19 mutants fed with coconut oil as the etomoxir-sensitive oxygen consumption of permeabilized larval tissue. Amounts of porin and citrate synthase activity were determined to exclude effects due to higher mitochondrial mass (supplemental S2A and S2B Fig). We found that the rescue diet enhances the β-oxidation rate in wild types and found that it stays at high levels in Pex19 mutants under both feeding conditions. This suggests that MCFAs increase mitochondrial β-oxidation in general while possibly improving the energy gain in Pex19 mutants, since they already have maximal β-oxidation rates under normal food conditions (Fig 3G and 3H). Since mitochondrial β-oxidation is increased in Pex19 mutants, the question arises if the gap in MCFAs and the rescue effect of MCFA-containing coconut oil is caused by their preferential degradation in mitochondrial β-oxidation for energy gain. To show that the MCFAs from coconut oil target the mitochondria, we cotreated the larvae with the CPT I–inhibitor etomoxir [36]. Etomoxir indeed abolishes the rescue effect of coconut oil with respect to survival (Fig 3I). This suggests that MCFAs from the rescue diet are efficiently metabolized by the mitochondria. The striking rescue of the loss of a whole organelle by a dietary adjustment provoked the question whether it was specific for Pex19 mutants or general for the loss of peroxisomes. We tested the effect of the dietary rescue on the survival of other Pex mutants: transheterozygous Pex2HP35039/Pex2f01899 [37], referred to as Pex2HP/f; Pex32[19] crossed over a deficiency (Df (32) 6262), referred to as Pex3/Df; Pex5MI06050 (Bloomington stock collection), referred to as Pex5−/−; and Pex10MI04076 (Bloomington stock collection), referred to as Pex10−/−. Of note, it has been shown that Pex2 and Pex16 mutants have reduced levels of C12: 0 [22], consistent with our observations in Pex19 mutants. We also reanalyzed the survival and dietary rescue of the Pex19ΔF7 mutant allele crossed over a deficiency (Df (2L) esc p3-0). We found that addition of 5% coconut oil to the food improves the survival of all of these Pex mutants (Fig 4A). To exclude unspecific effects due to enhanced food uptake, we performed a feeding assay but found no significant differences between genotypes or feeding condition (supplemental file S2 Fig). We concluded that the dietary rescue compensates for the loss of peroxisomes rather than for the loss of Pex19. To strengthen the hypothesis that the dietary rescue with coconut oil supplementation is a peroxisome rather than a Pex19-specific effect, we analyzed mitochondrial swelling in Malpighian tubules of Pex2HP/f, Pex3/Df, and Pex19 mutant larvae. Similar to Pex19 mutants, Pex2 and Pex3 mutants display large, balloon-shaped mitochondria, whereas mitochondria are small in wild-typic tissue. Upon administration of 5% coconut oil with the diet, we observed a rescue of this phenotype in all mutants: mitochondria were small and numerous, similar to wild-typic mitochondria (Fig 4B–4I). Similarly, the concentration of lipotoxic free fatty acids, which can cause mitochondrial swelling [23], is elevated in Pex2 and Pex3 mutants as in Pex19 mutants. Feeding of the rescue diet reduces the free fatty acid concentration in all three Pex mutants. A Pex19 mutant with a genetic rescue construct does not show elevated free fatty acid levels (Fig 4J). Lip3 expression is drastically increased in Pex19 mutants [23], which is reduced upon feeding of coconut oil (Fig 3E). Recently, it was discovered that the ceramide synthase Schlank acts as a transcriptional regulator of lip3 [27]. Schlank shuttles from the nuclear envelope to the endoplasmic reticulum (ER) membrane, thereby releasing its repression of lip3 transcription. We assessed the subcellular localization of Schlank in Pex2, Pex3, and Pex19 mutant fat body tissue by immunohistochemical stainings with anti-Schlank and anti-Lamin to stain the nuclear membrane. We found that the nuclear localization is indeed reduced (Fig 5C, 5E, 5G and 5I). This is consistent with the observed elevated transcript levels of lip3 and free fatty acid levels. We wanted to know if Schlank plays a role in the rescue effect induced by MCFA supplementation. Of note, upon addition of 5% coconut oil to the diet, Schlank indeed regained its localization to the nuclear membrane (Fig 5D, 5F, 5H and 5I) in the Pex2, Pex3, and Pex19 mutants we analyzed and would thus be enabled to repress lip3, explaining the observed decrease in lip3 expression, reduced free fatty acid levels, ameliorated mitochondrial morphology, and rescue to adulthood. To further prove this hypothesis, we overexpressed a shortened version of Schlank (Schlankaa1–138) [38,27], which is constitutively nuclear (supplemental file S3 Fig), in the Pex19 mutant background. We found that forcing Schlank into the nucleus in this manner rescues the lethality of Pex19 mutants (Fig 5J). Furthermore, we generated a double mutant for Pex19 and a Schlank knock-in (KI) mutant with a mutation in one of the nuclear localization sequences (NLS2), SchlankKINLS2 [28,27]. SchlankKINLS2 mutants show derepression of lip3, since the mutant Schlank protein is excluded from the nuclear membrane [27]. Without the ability to enter the nucleus, SchlankKINLS2 protein should not be able to react to MCFA supplementation, and a rescue with coconut oil should no longer be possible. Indeed, we found that SchlankKINLS2; Pex19 double mutants are larval lethal and are not rescued by coconut oil (Fig 5K), which confirms a role of Schlank in conferring the rescue effect of MCFA supplementation: Schlank releases its repression on lip3 due to MCFA shortage in Pex19 mutants, thereby provoking lipotoxicity and increased mitotoxic free fatty acid levels. Filling the MCFA gap via the diet restores Schlank nuclear localization and thus its repression of lip3, which ameliorates the damaging lipolytic program. If Schlank lacks the nuclear localization sequence, the dietary rescue no longer works, since the mutated Schlank cannot confer lip3 repression by shuttling to the nuclear membrane. We posed the question whether rescue of peroxisome loss was specific for D. melanogaster or a more universal effect. To address this question, we made use of two other models: Caenorhabditis elegans and, to assess the clinical relevance of our findings, a human skin fibroblast cell line from a Zellweger syndrome patient with a mutation in PEX19 (Δ19T cells, [39], supplemental file S3 Fig). We fed a C. elegans strain expressing mitochondrial GFP in muscle tissue (myo-3p: : mtGFP) with bacteria expressing RNA interference (RNAi) against prx-19, the C. elegans Pex19 homolog. Knock-down of prx-19 led to massively swollen mitochondria. When coconut oil was added to the bacterial lawn, control worms showed some fragmented and enlarged mitochondria, while mitochondrial swelling in the prx-19 knock-down animals was reduced (supplemental file S3 Fig). We analyzed the mitochondrial morphology in control fibroblasts from a healthy person and in Δ19T cells by staining them with TMRE and found that mitochondria were organized in long, stretched, filamentous networks in control cells, while they were more fragmented and swollen in Δ19T cells. To mimic the M/LCFA rescue diet, we conjugated coconut oil to bovine serum albumin (BSA, final concentration approximately 1 mM) and added 10% to the media. Similar to our results in C. elegans, this leads to mitochondrial fragmentation in control cells but has an ameliorative effect on the mitochondrial swelling in Δ19T cells (Fig 6A–6D, supplemental file S3 Fig). Free fatty acid levels are elevated in Δ19T cells (Fig 6E). Upon addition of 10% BSA-conjugated coconut oil, free fatty acid levels increased in both control and Δ19T cells. By contrast, when Δ19T fibroblasts were cultured with 10% BSA-conjugated coconut oil for several passages, and samples were taken in the absence of coconut oil in the medium, they showed reduced levels of free fatty acids (Fig 6E). Our results suggest a conserved mechanism of the impact of Pex19 mutation on mitochondria and the pathology of the disease, converging on free fatty acid toxicity. Free fatty acid levels and mitochondrial swelling are ameliorated across vertebrate and invertebrate models (flies and human cells) upon supplementation with MCFA from coconut oil. We show that supplementation with MCFAs, rather than removal of VLCFAs, rescues a peroxisome-deficient PBD model by acting on the ceramide synthase Schlank to prevent lipotoxicity. The Pex19 mutant was generated by imprecise excision following Drosophila-standard techniques. The line Pex19ΔF7 was chosen from a jump-out screen and tested as a transcript null. To detect homozygous animals, they were kept with a CyO, twi-GFP balancer. As control flies, we used the strain w1118 (Bloomington stock #3605). Wild-type and heterozygous, balanced Pex19ΔF7 flies were reared on standard fly food. Pex2HP35039, Pex2f01899, Pex32, Df (32) 6262, Pex5MI06050, Pex10MI04076, and Df (2L) esc p3-0 were obtained from the Bloomington stock center; SchlankKINLS2 /FM7, Kr-GFP, and UAS-Schlankaa1-138 flies were kindly provided by Reinhard Bauer. The latter insertion was recombined with the Pex19ΔF7 chromosome using standard Drosophila techniques. SchlankKINLS2/Y; Pex19 ΔF7/ΔF7 males were selected by absence of fluorescent markers. As control, we crossed FM7, Kr-GFP/Y; pex19 ΔF7/CyO, twi-GFP males with +; Pex19ΔF7/CyO, twi-GFP virgins and used the resulting homozygous +/Y; Pex19 ΔF7/ΔF7 males. For all assays, eggs were collected on apple juice agar plates (2% agar, 2. 5% sucrose, 25% apple juice, 1. 5% nipagin) with fresh yeast paste (42 g of yeast mixed with 10 ml of tap water), and first-instar larvae were transferred to fresh plates with yeast as control or yeast and supplements: 5% coconut oil, babassu oil, palm kernel oil, cocoa butter, rapeseed oil, olive oil, safflower oil, or broccoli seed oil. Oils of different manufacturers gave similar results. Except for babassu and palm kernel oil, which were refined and deodorized, native natural oils were used. Synthetic TAGs were from Sigma Aldrich. Nipagin was omitted from the apple juice agar plates for nipagin-free feeding assays. TAGs were purified from native coconut oil with a solid phase extraction column, and 5% TAGs were added to the yeast paste for feeding experiments. For the etomoxir assay, 100 μl of a 250-μM etomoxir (Sigma-Aldrich) solution was added to 500 mg of the yeast or coconut oil–containing yeast. For survival assays, 25 first-instar larvae were collected for each condition, and at least 5 independent experiments were conducted (exact numbers of n given in figure legends). The number of surviving pupae, adults (including pharates), and viable adults (survivors) that were able to move and lived at least 24 hours was counted. Five larvae at a time were put with a soft brush onto a prewarmed PBS-agar plate (10-mm petri dish) placed on top of millimeter graph paper and left for 1 minute to acclimate. Then, they were filmed using a Panasonic camcorder (HC-V380) for 40 seconds. Their movement was tracked manually by marking the position of the head every 2 seconds. The millimeter graph paper was used to scale the resulting image so that track lengths could be measured with ImageJ (Fiji). Larvae that did not move were excluded from analysis. Human fibroblast control and Δ19T cells were kept in Dulbecco’s modified eagle medium (DMEM, Gibco) with 10% FBS, 10,000 units of penicillin, and 10 mg streptomycin per ml. For coconut oil treatment, coconut oil was coupled to BSA at 37 °C, following an adapted protocol of Seahorse Bioscience for BSA-conjugated palmitate. In brief, 65. 94 mg of coconut oil was added to 44 ml of 150 mM NaCl and stirred at 37 °C, and 11. 2 ml was added to 14 ml BSA solution in 150 mM NaCl and stirred for 1 hour at 37 °C. For NEFA measurement, 1 × 105 cells were seeded in 6-well plates and harvested after 48 hours. Cells were pelleted and washed with PBS. Cell pellets were treated like larval tissue (see Free fatty acid section). For experiments with coconut oil–BSA, cells were treated with 10% coconut oil–BSA 24 hours after seeding. Additionally, Δ19T cells were cultured with 10% coconut oil–BSA for several passages and seeded for NEFA measurement in normal medium. For live cell stainings, cells were seeded into 8-well slides for microscopy (Ibidi). After 48 hours, cells were washed carefully 3 times with PBS and stained with 50 nM TMRE for 15 minutes. Afterward, cells were washed again 3 times with PBS and immediately analyzed in the microscopy slide with PBS using a Zeiss LSM 710 with a 63× water objective (Plan-Apochromat, Zeiss). We used an α-Schlank antibody [44], α-Lamin Dm0 (Developmental studies hybridoma bank, DSHB), and DAPI for (immune) stainings of Drosophila fat bodies. Secondary antibodies coupled to Alexa dyes were from molecular probes, and DAPI from Sigma-Aldrich. For (immune) histochemistry, we dissected tissue of interest from third-instar larvae. Tissue was fixed for 30 minutes in 3. 7% formaldehyde and washed with PBT before and after incubation with primary antibody and Alexa dye–coupled secondary antibody. Tissue was mounted in Fluoromont G and analyzed using a Zeiss LSM 710 confocal microscope. Fat body cells were analyzed with a 25× water objective (Plan-Neofluar, Zeiss) and a pinhole of 1 Airy unit, and Malpighian tubules were analyzed with a 63× water objective (Plan-Apochromat, Zeiss) and a pinhole of 1 Airy unit. Human fibroblasts stained with TMRE were analyzed with a 63× water objective (Plan-Apochromat, Zeiss) and a pinhole of 1 Airy unit at a resolution of 1024 × 1024 pixels and at 1× zoom. For apoptosis assays, brains of adult flies were dissected and stained using an annexin V-FITC apoptosis detection kit (Sigma Aldrich) according to the manufacturer’s instructions. Imaging was done with identical imaging parameters for all conditions analyzed, using a 25× water lens (Plan-Neofluar, Zeiss) and a pinhole of 1 Airy unit on a Zeiss LSM710. Maximum intensity projections of 5 consecutive optical sections were generated for each genotype analyzed. For staining of mitochondria, 96-hour-old larvae or 5-day-old adults were dissected in ice-cold PBS, and their Malpighian tubules were stained for 20 minutes at RT with 50 nM TMRE (Sigma-Aldrich) in PBS according to the manufacturer’s protocol. The Malpighian tubules were then directly mounted in Fluoromount G and analyzed with a Zeiss LSM 710. Picture analysis and quantification were done using ImageJ. For quantification of Schlank subcellular localization, the corrected total cell fluorescence (CTCF) was measured with ImageJ using the Integrated Density and Area parameters. CTCF was calculated as CTCF = IntDen (Area × Mean gray value of background). The nuclear signal was determined by analyzing the area within the Lamin staining. For quantification of mitochondria, the particle size was measured with ImageJ. Each staining was done at least 5 times. For quantification of FAMEs, 15 third-instar larvae were homogenized in 1 N MeHCl in a Precellys 24 homogenizer (peqlab). A minimum of n = 7 was analyzed for each condition. C15: 0 and C27: 0 standards were added, and samples were incubated for 45 minutes at 80 °C. Methyl esters were collected by addition of hexane and a 0. 9% NaCl solution. The hexane phase was collected in a new glass vial and concentrated by vaporization. Samples were analyzed by GC/MS using an Agilent HP 6890 with an HP-5MS column. For FAME analysis in pupae, 100 pupae at n = 3 were processed as described above. NEFAs were measured by an adaptation of the copper-soap method [45]. In brief, 3 third-instar larvae were weighed and homogenized in 20 μl of 1 M phosphate buffer per mg tissue. Then, 25 μl of the supernatant were transferred to 500 μl of Chloroform/Heptane 4: 3, and lipids were extracted by shaking the vial for 5 minutes. Unspecific background provoked by phospholipids was circumvented by addition of 23 mg of activated silicic acid. Next, 300 μl of the chloroform phase was transferred to 250 μl of Cu-TEA (copper-triethanolamine). After shaking and centrifuging, 150 μl of the organic phase was transferred to fresh cups. Liquid was evaporated in a 60 °C heat block, and lipids were dissolved in 100 μl of 100% ethanol. Copper was detected by complexation with a mixture of dicarbazone–dicarbazide, and the color intensity was measured in a 96-well plate at 550 nm in a TECAN plate reader. Each experiment was conducted at a minimum of 5 biological replicates. Six larvae per genotype were washed with PBS, and their weight was recorded for normalization purposes. The larvae were inverted in ice-cold PBS and permeabilized in ice-cold BIOPS buffer (2. 77 mM CaK2EGTA, 7. 23 mM K2EGTA, 5. 77 mM Na2ATP, 6. 56 mM MgCl2. 6H2O, 20 mM taurine, 15 mM Na2. phosphocreatine, 20 mM imidazole, 0. 5 mM DTT, 50 mM MES) containing 100 μg/mL saponin (fresh) at 4 °C with gentle rocking for 10 minutes. Then, the larvae were equilibriated in respiration medium (MiR05,0. 5 mM EGTA, 3 mM MgCl2*6H2O, 60 mM K-Lactobionate [lactobionic acid is dissolved in H2O, and pH is adjusted to pH 7. 4 with KOH], 20 mM Taurine, 10 mM KH2PO4,20 mM HEPES, 110 mM sucrose, 1 g/L fatty acid–free BSA) supplemented with 0. 5 mM carnitine. The larvae were added into the oxygraph chambers, and oxygen concentration was brought to around 500 μM by using catalase and H2O2. After basal respiration was recorded, 5 μM palmitoyl-CoA was added to the chamber. Fatty acid β-oxidation was induced by adding complex I substrates, electron transfer flavoprotein (ETF) substrates, and ADP (10 mM proline, 10 mM pyruvate, 5 mM malate, 5 mM glutamate, 2 mM ADP, and 15 mM glycerol-3-phosphate). After that, etomoxir was added at the indicated concentrations to block fatty acid transfer into mitochondria via CPT I, thereby blocking β-oxidation and leaving complex I–dependent respiration. Finally, residual oxygen consumption (ROX) was measured by inhibiting complex III with antimycin A. All values were corrected for ROX. β-oxidation was calculated by subtracting etomoxir-resistant respiration from respiration in the presence of all substrates. Each measurement was repeated at least 3 times (biological replicates). Whole RNA of third-instar larvae was isolated using TriFast reagent (peqlab). Tissue was homogenized using a Precellys 24 homogenizer (peqlab). Transcription to cDNA was performed using the Quantitect Reverse Transcription Kit (Quiagen). qPCR was performed with a CFX Connect cycler (biorad). A minus-RT was analyzed in a PCR for each cDNA. qPCR was performed with a CFX Connect cycler (biorad) using GoTaq SYBR Mix (Promega). Values were normalized against two housekeeping genes (actin5c and rp49) and against wild-type control (ΔΔCq). See Table 1 for primer sequences. Each experiment was repeated at least 5 times. We used the software GraphPad InStat for statistical analysis of our data. We applied the unpaired, two-tailed Student t test, assuming heteroscedasticity for single comparisons, and ANOVA with Tukey posttests and Bartlett’s test for homoscedasticity. Error bars represent standard deviation. Asterisks represent * p < 0. 05, ** p < 0. 01, *** p < 0. 001.

Peroxisomes are organelles that contain several enzymes and fatty acids required for many metabolic tasks in the cell, and upon peroxisome loss, their educts accumulate. One example is the accumulation of very-long-chain fatty acids (VLCFAs) with a chain length of more than 20 carbons. These fatty acids cannot be oxidized in mitochondria but are exclusively degraded in peroxisomes. Lowering increased VLCFA levels is sometimes attempted as a treatment option for human disorders with peroxisomal dysfunction, although its effectiveness remains unclear. Here, we have analyzed this process in Drosophila melanogaster and found that peroxisomal loss results not only in VLCFA accumulation but also in a reduction of medium-chain fatty acids (MCFAs). We could show that this is due to a state of high lipolysis and increased mitochondrial activity. By supplementation with MCFAs from coconut oil, we were able to rescue mitochondrial damage and lethality observed in peroxisome-deficient flies. We found that this process is mediated by the ceramide synthase Schlank, which acts as a transcription factor and shuttles between nuclear membrane and endoplasmic reticulum (ER) in response to MCFA availability. We conclude that peroxisome loss triggers the accumulation of free fatty acids and mitochondrial damage in flies and that these effects can be rescued by a diet rich in MCFAs.

Abstract Introduction Results Discussion Methods

lipolysis medicine and health sciences plant products diet developmental biology nutrition mitochondria bioenergetics peroxisomes cellular structures and organelles lipids crop science life cycles chemistry agriculture biochemistry vegetable oils agronomy hydrolysis cell biology biology and life sciences chemical reactions fatty acids energy-producing organelles physical sciences larvae

2018

Dietary rescue of lipotoxicity-induced mitochondrial damage in Peroxin19 mutants

Understanding how hepatitis C virus (HCV) induces and circumvents the host' s natural killer (NK) cell-mediated immunity is of critical importance in efforts to design effective therapeutics. We report here the decreased expression of the NKG2D activating receptor as a novel strategy adopted by HCV to evade NK-cell mediated responses. We show that chronic HCV infection is associated with expression of ligands for NKG2D, the MHC class I-related Chain (MIC) molecules, on hepatocytes. However, NKG2D expression is downmodulated on circulating NK cells, and consequently NK cell-mediated cytotoxic capacity and interferon-γ production are impaired. Using an endotoxin-free recombinant NS5A protein, we show that NS5A stimulation of monocytes through Toll-like Receptor 4 (TLR4) promotes p38- and PI3 kinase-dependent IL-10 production, while inhibiting IL-12 production. In turn, IL-10 triggers secretion of TGFβ which downmodulates NKG2D expression on NK cells, leading to their impaired effector functions. Moreover, culture supernatants of HCV JFH1 replicating Huh-7. 5. 1 cells reproduce the effect of recombinant NS5A on NKG2D downmodulation. Exogenous IL-15 can antagonize the TGFβ effect and restore normal NKG2D expression on NK cells. We conclude that NKG2D-dependent NK cell functions are modulated during chronic HCV infection, and demonstrate that this alteration can be prevented by exogenous IL-15, which could represent a meaningful adjuvant for therapeutic intervention. Natural Killer (NK) cells are effectors of the rapidly acting antiviral innate immune system. They kill virally infected cells and are an important source of antiviral cytokines such as IFNγ. In addition, they establish an early and efficient dialogue with professional antigen presenting cells (APCs) that in turn, orchestrate the adaptive immune response towards Th1-type antiviral immunity [1]. NK cell activation is tightly regulated by the integration of signals emanating from a diverse array of inhibitory and activating receptors [2]. Inhibitory receptors, including Killer cell Immunoglobulin-like receptors (KIRs) and CD94/NKG2A, gauge expression of MHC class I molecules which can be compromised by viral immune subversion, and thus serves as an indicator of the integrity of cells. Activating receptors, including the natural cytotoxicity receptors (NCRs) and NKG2D, usually detect the presence of infectious non-self and/or stress-induced self ligands at the surface of infected cells. Hepatitis C virus (HCV), which replicates in hepatocytes, mediates a chronic liver infection in the majority of infected individuals. NK cells abound in the normal liver, where they make up to 30% of resident hepatic lymphocytes [3]. This huge amount of NK cells in the liver suggests that they are important sentinel cells, surveying the liver for signs of damage or cellular stress. However, it also implies that HCV must divert NK cell-mediated responses in order to establish persistent infection. The importance of NK cells in the resolution of HCV infection is illustrated by the influence of genetic polymorphisms of KIR and their HLA ligands on the outcome of HCV infection [4]. Various alterations of NK cell phenotype have been described during chronic HCV infection, but results are often contradictory regarding the experimental conditions used (ex vivo or in vitro cytokine-stimulated), the modifications involved and their consequences on effector functions [5], [6], [7], [8], [9], [10], [11]. The NKG2D activating receptor is constitutively expressed on human NK and CD8 T cells [12]. Its ligands, the MHC class I chain-related A and B proteins (MICA and MICB) and UL-16 binding proteins (ULBP1–4), are almost undetectable in normal tissues, but are induced on the cell surface by various stresses such as DNA damage, tumor transformation and intracellular infection. The importance of the NKG2D defense system is highlighted by the observation that tumors and viruses have developed several mechanisms for evading NKG2D-mediated recognition [13], [14], [15], [16], [17]. The overall contribution of the NKG2D pathway in the control of HCV infection is unclear [7], [10]. We show here that NKG2D is downmodulated on circulating NK cells, and consequently NK cells are functionally impaired. This defect is mediated by the HCV-NS5A protein, which disturbs the equilibrium between pro- and anti-inflammatory monocyte-derived cytokines. MIC proteins are induced at the cell surface upon exposure to various pathogens [11], [18], [19], [20], serving as a warning signal that alerts NK cells to mediate effector functions through NKG2D signaling. We thus examined if MIC was expressed in the liver during chronic HCV infection. While staining of control livers showed a faint expression of MIC in the cytoplasm of some hepatocytes and Kupffer cells only, HCV-infected livers displayed a strong and diffuse expression of MIC in the cytoplasm and at the surface of HCV-infected hepatocytes, and also in some uninfected hepatocytes and large mononuclear cells in portal spaces resembling macrophages (Figure 1). NKG2D is constitutively expressed on NK cells, and should therefore mediate recognition and destruction of MIC-expressing cells. Due to the restricted availability of fresh HCV-infected liver samples to isolate infiltrating NK cells, we examined the expression of NKG2D on circulating NK cells. No significant difference in the percentage of circulating NK cells, or in the proportion of CD56bright/CD56dim cells was detectable between patients and controls (data not shown). The percentage of NKG2D-expressing NK cells was similar in HCV patients and healthy controls (>95% of NK cells in both groups). However, a decreased expression of NKG2D on both CD56bright and CD56dim NK cells was detected in HCV viremic patients as compared with healthy controls (mean MFI: 61±15 versus 93±25, P<10−4), HCV patients with sustained viral response (SVR) after treatment (81±14, P<10−3) or patients with non-infectious chronic inflammatory liver disease (87. 4±24. 5, P<10−3) (Figure 2A). Although showing some variability among viremic patients, NKG2D levels were not correlated with age, sex, HCV viral load, genotype, ALT levels, liver fibrosis or activity score. To evaluate the functional consequences of this NKG2D reduction, we quantified NK cell IFNγ production and CD107a degranulation by flow cytometry. Freshly purified circulating NK cells from HCV patients showed impaired IFNγ production in the presence of MHC class I-negative K562 as compared to NK cells from healthy controls. In addition, NK cells from HCV patients showed a two-fold decreased degranulation in the presence of K562 targets (mean CD107 expression 29. 3%±2. 7 in controls compared to 15. 9%±2. 9 in patients, P = 0. 003) (Figure 2B). This defective NK cell function was at least in part dependent on NKG2D, as shown using C1R-MICA as target cells. CD107a expression on NK cells positively correlated with NKG2D levels (Spearman rho (r) = 0. 62, P = 0. 008, Supplementary Figure S1). Moreover, NKG2D blocking by anti-NKG2D antibody largely inhibited inhibited NK cell degranulation in both HCV patients and healthy controls (Figure 2C). That NK cell degranulation was not fully abrogated indicates however, that it likely involves other activating receptor (s) in addition to NKG2D. Altogether, these results suggest that the signaling pathway initiated by NKG2D on target exposure may not operate properly in HCV patients due to NKG2D reduction on NK cells. Systemic NKG2D downregulation on immune effector cells has been related to the release of soluble factors such as MIC molecules (sMIC) in the serum of cancer patients [13]. We measured sMIC in the serum of HCV patients and healthy controls, and found similar low levels of sMIC in both groups (data not shown). TGFβ is another mediator of systemic NKG2D downregulation [21], [22], [23]. Total TGFβ levels were higher in the serum of HCV-infected patients compared to controls and SVR patients (Figure 3A). An inverse linear relationship was observed between TGFβ and NKG2D levels on NK cells (Figure 3B). To investigate whether serum of HCV patients could mimic the effect of exogenous TGFβ on NKG2D expression, we incubated control NK cells with recombinant TGFβ or with serum from representative HCV patients with known TGFβ concentration, and analyzed NKG2D expression. NKG2D levels were reduced in a TGFβ concentration-dependent manner, and were largely restored when incubation was performed in the presence of neutralizing anti-TGFβ antibody (Figure 3C). Engagement of the HCV receptor CD81 by the major HCV envelope protein E2 was shown to block NK cell functions triggered by NKG2D engagement [24]. We thus hypothesized that HCV-E2 might modulate NKG2D expression. PBMCs from normal donors were exposed to recombinant HCV-E2, as well as to other structural and non-structural HCV proteins for 6 to 48 hr, and NKG2D levels were measured on NK cells (Figure 4A). While HCV-E2, -core, -NS3, or -NS4 proteins had no or minor effect, HCV-NS5 induced a dose-dependent reduction of NKG2D on NK cells, which was maximal at 48 hr. Using different recombinant NS5A preparations (E. Coli-derived full length NS5, yeast-derived NS5 2054-2995 or E. Coli-derived NS5A amino acid 2061–2392), we reproducibly identified the NS5A protein as being responsible for this effect. At a concentration of 0. 5 µg/ml, NS5A reduced NKG2D expression by 40% (P = 0. 001), which was of the same order of magnitude as 10 ng/ml of recombinant TGFβ used as positive control. Of note, NS5A stimulation also induced downmodulation of the NKp30 activating receptor (Supplementary Figure S2), in line with the previously described effect of TGFβ on NKp30 expression [23]. The β2-microglobulin, used as control for irrelevant protein produced in E. Coli, had no effect on NKG2D expression. All recombinant proteins used were tested for the absence of significant lipopolysaccharide (LPS) contamination (0. 054 endotoxin unit/µg recombinant protein in the case of NS5A, i. e. 5. 4 pg/µg protein). In addition, pretreating NS5A by 10 µg/ml of polymyxin B was without effect, ruling out the possibility of contaminating LPS being the factor responsible for NKG2D reduction. Inactivation of the NS5A protein by freeze/thaw before incubation with PBMCs abolished the NS5A-mediated NKG2D reduction, suggesting that it required intact protein conformation. To verify that NS5A-mediated downregulation of NKG2D on NK cells was accompanied by a decrease in their functional capacity, PBMCs were exposed to NS5A, NS4 or medium alone for 48 h, after which NK cell degranulation capacity in the presence of K562 target cells was evaluated by flow cytometry. The presence of NS5A in PBMC culture induced a significant decrease of CD107a expression on NK cells, while NS4 had no effect, confirming that NS5A is responsible for a decreased functionality of NK cells (Figure 4B). We then measured the TGFβ concentration in culture supernatants from PBMCs exposed to NS5A or medium alone for 12 to 48 h. Levels of TGFβ progressively increased in the presence of NS5A (Figure 4C). To confirm that the NS5A effect was indeed related to TGFβ production, we pretreated PBMCs with anti-TGFβ antibody prior to stimulation with NS5A. Blocking TGFβ abrogated the NS5A-induced reduction of NKG2D on NK cells in a dose-dependent way (Figure 4D). When similar experiments were performed on freshly purified NK cells, NS5A stimulation failed to downmodulate NKG2D, suggesting that TGFβ was likely produced by a distinct cell population among PBMCs. To identify the source of TGFβ, we cocultured purified NK cells with different components of autologous PBMCs, including adherent or non-adherent cells, monocyte-depleted PBMCs, or purified monocytes. Cells were stimulated for 6–48 hr with NS5A, after which NKG2D was measured on NK cells. At the same times, culture supernatants were recovered and were assayed for TGFβ production. As shown in Figure 4E, only monocyte-containing populations induced a significant decrease of NKG2D expression on NK cells (p<0. 002). Notably, downregulation of NKG2D was completely lost in the Transwell system, indicating that monocyte-NK cell contacts were required for this effect. NKG2D reduction was accompanied by an increased production of TGFβ in cell supernatants, which was maximal after 40 hr of NS5A stimulation (Figure 4C). In addition, neutralization of TGFβ in the coculture system of monocytes and NK cells fully restored NKG2D expression on NK cells. Altogether, these results indicate that NS5A protein induces TGFβ production by monocytes, which in turn affects NKG2D expression and inhibits NK cell functions. Activation of monocytes by microbial products usually induces the production of IL-12, and to a lesser extent of IL-10. We thus wondered if the ability of NS5A to regulate the production of TGFβ was more global, and measured the production of IL-10 and IL-12 in supernatants from control monocytes stimulated for 24 to 48 h with NS5A, NS4, medium alone, or LPS as a positive control. High levels of IL-10 (>1400 pg/ml) were detected 24 hr after NS5A stimulation (Figure 5A, left panel). These levels were of the same order of magnitude as those induced by 1 µg/ml of LPS. We then wondered if NS5A-induced TGFβ production was related to autocrine IL-10 release. Blocking IL-10 or its receptor abrogated the NS5A-induced TGFβ secretion in a dose-dependent manner, but did not modify the basal TGFβ production by monocytes (Figure 5A, right panel). Furthermore, we found elevated IL-10 levels in the sera of HCV patients, that were positively correlated with TGFβ levels (r = 0. 39, P = 0. 016). By contrast, very low amounts of IL-12 were detected in supernatants of NS5A-stimulated monocytes, while LPS induced high levels of IL-12 as expected (Figure 5B). We thus hypothesized that NS5A might inhibit the LPS-induced production of IL-12 by monocytes. Indeed, pretreatment of monocytes by NS5A strongly inhibited IL-12 production upon LPS stimulation. Taken together, these findings demonstrate that NS5A potently increases the production of anti-inflammatory cytokines IL-10 and TGFβ, while concurrently suppressing the production of proinflammatory IL-12. The lack of NS5A-induced IL-12 secretion also confirms that LPS contamination is not responsible for the NS5A-mediated effect in monocytes. NK cell activation requires signals provided by APCs that sense pathogen products through conserved pattern-recognition receptors such as Toll-like receptors (TLRs). In particular, TLR2 and TLR4 are involved in extracellular sensing of several viral proteins by monocytes and dendritic cells [25], [26], [27]. We thus hypothesized that NS5A might interact with TLR2 or TLR4 signaling in monocytes. Pretreating monocytes with blocking anti-TLR4 antibody, or with antibody to the TLR4 associated molecule CD14, fully abolished NS5A-mediated IL-10 production, while blocking anti-TLR2 antibody had no effect (Figure 5C). This suggested that NS5A might interact with TLR4 on monocytes. To support this hypothesis, freshly purified monocytes were incubated at 4°C with NS5A or NS4, and binding was revealed by staining with anti-NS5A antibody and flow cytometry analysis. A significant binding of NS5A on monocytes was observed, that was inhibited by almost 50% in the presence of blocking anti-TLR4 antibody (Figure 5D). TLR4 signaling results in the downstream activation of NF-kB, MAPK (p38 and JNK) and PI3K pathways[28]. TLR4 activation may contribute to IL-10 production via p38 and PI3K [29], while PI3K is an endogenous suppressor of IL-12 production triggered by TLR4 [30]. We pretreated monocytes with pharmacological inhibitors of signaling molecules prior to stimulation with NS5A, and measured IL-10 production. Inhibition of p38 or PI3K suppressed NS5A-induced production of IL-10 in a dose-dependent way, while other inhibitors had no or minor effects (Figure 5E). Altogether, our results indicate that, upon NS5A interaction with TLR4, monocytes preferentially secrete IL-10 through activation of the p38 and PI3 kinase pathways, but are prevented from secreting IL-12. NS5A is not found in viral particles secreted by infected cells, which raises the question of its availability in the extracellular medium. However, it is becoming increasingly clear that HCV infection of hepatocytes has direct cytopathic effects, suggesting that NS5A might be released from apoptotic/necrotic infected cells [31], [32], [33]. To verify this hypothesis, we determined whether supernatants of HCV-infected cells induced downmodulation of NKG2D. To this aim, we used Huh-7. 5. 1 hepatoma cells transfected or not with the infectious genotype 2a JHF1 replicon [34], [35]. Control PBMCs were cultured for 48 h with Huh-7. 5. 1 culture supernatants or with recombinant NS5A as positive control, after which NKG2D levels on NK cells were measured (Figure 6A). Supernatants from Huh-7. 5. 1 uninfected cells, or those recovered as soon as day 3 post transfection, did not modify NKG2D levels. By contrast, supernatants recovered on days 13,15 and 18 post transfection reproducibly induced downmodulation of NKG2D at levels similar to those obtained using 0. 5 µg/ml rNS5A. At these time points, the majority of cells in the culture expressed HCV proteins, infectivity titers in culture supernatant were maximal, and cytopathic effects were observed, as reported [31]. Notably, there was no effect of supernatants recovered at day 23, i. e. at a time cells were cytologically normal and levels of extracellular infectious virus had declined. To rule out the possibility that this effect was mediated by TGFβ, we quantified TGFβ in Huh-7. 5. 1 culture supernatants. In all conditions studied, TGFβ levels were below 40 pg/ml and could thus not be responsible for the observed NKG2D-downregulating activity. Taken together, these data indirectly suggested that NS5A is released by HCV-replicating cells, most likely among cell debris generated by infection. Since we previously observed a specific binding of recombinant NS5A on monocytes, we incubated control PBMCs with supernatants from uninfected or infected Huh-7. 5. 1 cells, or with recombinant NS5A or NS4 proteins, after which NS5A binding to monocytes was evaluated by staining with anti-NS5A antibody. Supernatants from uninfected cells, or supernatants recovered on days 2 and 23 post transfection did not show any binding signal, in accordance with their lack of NKG2D-downregulating activity. However, supernatants recovered on day 12 post transfection gave a binding signal that was of the same order of magnitude than that observed with recombinant NS5A (Figure 6B). Notably, this effect was abrogated when day 12 supernatants were filtered in order to eliminate cell debris, suggesting that NS5A was not released from JHF1-replicating Huh-7. 5. 1 cells in a soluble form, but was rather associated with apoptotic-cell components. Our attempts to corroborate the presence of NS5A in day 12 supernatants using ELISA or Western blot techniques, or to deplete NS5A from these supernatants using anti-NS5A antibody, were unsuccessful (data not shown). This could be due to the fact that NS5A was not easily recognized by the 9E10 mAb when associated with apoptotic cell debris in the supernatants. Alternatively, it could be that a fraction of bioactive NS5A becomes liberated after interaction of the cell debris with monocytes. Given that apoptotic HCV-replicating cells seem to release NS5A, we made an effort to resolve the issue of NKG2D expression on NK cells in the infected liver. We analyzed liver-infiltrating NK cells and paired circulating NK cells in 11 additional HCV viremic patients who underwent liver biopsy for diagnostic purpose. NKG2D levels on peripheral NK cells fully matched with those in the first series of patients (MFI 59 ±7. 7 and 61±15, respectively). Of note, the proportion of NK cells among liver-infiltrating mononuclear cells was very low (2. 7%±0. 7%), as already reported in HCV-infected livers [9], [36], [37]. To our surprise - but in line with previous observations in the rat and human ([10], [38], [39]- liver NK cells expressed higher NKG2D levels than their circulating counterpart (mean MFI 115. 5 ±17. 4 versus 59±7. 7, P = 0. 002, Wilcoxon matched-pairs test). NKG2D analysis in a representative liver sample is shown in Figure 6C (left panel) and Supplementary Figure S3. By comparison, NKG2D levels on liver NK cells from 8 control patients with non-infectious chronic inflammatory liver disease were not significantly different from those observed on circulating NK cells (mean MFI 106. 2 and 87, respectively). Staining of HCV-infected liver sections showed that NKG2D+ cells were indeed very scarce even among large portal infiltrates. Only few cells were found positive in sinusoidal tracts (Figure 6C, right panel). Similar to what was observed for circulating NK cells, the proportion of liver NK cells and their NKG2D levels were not correlated with any HCV disease marker (not shown). Whether these intrahepatic NKG2Dhigh NK cells were more functionally competent than their circulating counterpart could not be evaluated due to their too small number. In attempt to analyze NKG2D-mediated functions in a clinically relevant target cell system, we sought to use JHF1-infected Huh-7. 5. 1 cells. However, it turned out that these cells were not pertinent for NKG2D ligand expression studies. Not only HCV infection did not induce MIC expression on Huh7. 5. 1 cells, but none of the stimuli known to be potent inducers of NKG2D ligands (heat-shock, oxidative stress, γ-radiation, retinoic acid, inhibitors of histone deacetylase) was able to induce MIC surface expression on Huh7. 5. 1 cells (data not shown). In contrast with TGFβ, IL-15 up-regulates surface NKG2D expression [40]. Because of the reciprocal antagonism of IL-15 and TGFβ on intracellular signaling pathways [41], [42], [43], we wondered if decreased NKG2D expression on circulating NK cells from HCV patients might be amplified by a resistance of NK cells to endogenous IL-15 due to TGFβ or by a defective production of IL-15 in response to infection. PBMCs from healthy controls or HCV patients were pretreated with IL-15 for 24 hr after which NKG2D staining was performed (Figure 7A). IL-15 restored NKG2D expression on patients' NK cells at levels similar to those usually observed in controls, indicating that NK cells from HCV patients were normally responsive to IL-15. Moreover, IL-15 fully antagonized the NS5A-induced downmodulation of NKG2D on NK cells. Even in the presence of TGFβ-containing serum, IL-15 could prevent NKG2D expression. Furthermore, both NK cells from HCV patients and TGFβ-stimulated control NK cells exhibited a significant enhancement of cytotoxicity upon IL-15 stimulation (Figure 7B). Given that pathogen components are among the stimuli that elicit production of IL-15, one could expect elevated levels of IL-15 in the serum of chronic HCV patients. IL-15 levels were not higher in patients than in controls (mean 7. 61±1. 93 pg/ml, and 7. 98±1. 40 pg/ml, respectively; ns). It must be noted, however, that IL-15 is mostly present in membrane-bound IL-15/IL-15Rα complexes [44], so that free IL-15 is unlikely to represent a reliable marker of systemic IL-15 production. Altogether, these data suggest that overexpression of TGFβ contributes to the reduction of NKG2D and defective functions of circulating NK cells in HCV patients, a defect which can be antagonized by exogenous IL-15. HCV uses a repertoire of dampening signals to subvert immune responses, a significant number of which target the innate system [45], [46]. We report here an altered expression of the NKG2D receptor as an additional HCV strategy to avoid NK-cell mediated recognition. HCV-NS5A protein, through monocyte-derived TGFβ production, downregulates expression of NKG2D on NK cells, thus reducing their cytotoxic potential and IFNγ production. Some previous studies have reported defective NK cell function in HCV infection [47], [48], [49], although others have not seen this [5], [50], [51]. Different methodologies, including the use of total PBMC or purified NK cells, fresh or cryopreserved cells, unstimulated or cytokine-stimulated cells, chromium release or flow cytometry assays, and small sample sizes, might explain why some of the findings in these studies differed from our own. We performed all experiments on freshly purified unstimulated NK cells, and measured NK cell degranulation rather than overall K562 cell lysis, because it has the advantage to shift the focus from the fate of target cells to the true response of NK cells, as previously demonstrated [52]. Different virally encoded products have been shown to impair NKG2D-mediated detection of infected cells, usually by targeting the ligands of NKG2D rather than the receptor itself [14], [16], [53], [54]. HCV only encodes a small number of structural and non-structural proteins. Consequently, each HCV gene product must have pleiotropic functions rather than highly specialized ones. Targeting NKG2D rather than its numerous ligands is at lower cost for HCV. However, it is likely that HCV must also target other receptors to escape NK cell recognition. Indeed, reduced NKp30 levels have been observed on NK cells from HCV patients [6]. Interestingly, TGFβ not only downmodulates NKG2D, but also reduces expression of NKp30 [23] and we observed a TGFβ-mediated reduction of NKp30 on control NK cells upon NS5A stimulation. This pleiotropic effect of TGFβ thus represents an economical way for HCV to shift the overall balance of NK signals towards an inhibitory phenotype. Our results are in contrast with a recent report by Oliviero et al. showing an increased proportion of NKG2D+ NK cells in HCV patients compared with healthy controls [10]. A potential explanation to this discrepancy is the surprisingly low frequency of NKG2D-positive NK cells in controls (60%) from Oliviero' s study, although NKG2D is usually reported to be constitutively expressed on all NK cells [12]. In our study, the differences in patients and controls only affected NKG2D expression levels, but not the frequency of NKG2D-positive cells. We show that TGFβ production results from NS5A interaction with the TLR4 complex on monocytes, which leads to a dysregulated equilibrium of inflammatory cytokines, i. e. increased IL-10 and defective IL-12 production. IL-10 is a potent suppressor of TLR-induced inflammatory responses, and an important target of immune subversion for some pathogens. IL-10 signaling activates STAT3, which positively regulates TGFβ promoter activity [55]. Previous studied identified that HCV core, NS3 or NS4, but not E2 protein induced monocyte-derived IL-10 production [56], [57]. In the case of core and NS3, this effect was mediated through TLR2 signaling [58]. Unfortunately, NS5A was not tested in these studies. We think that NS5A signals through TLR4 in monocytes, because preincubation with a blocking anti-TLR4 mAb inhibited NS5A-mediated IL-10 production. Also, binding experiments showed that NS5A interacted directly with TLR4 on monocytes. The likelihood of contaminating LPS contributing to the NS5A-mediated effect was ruled out by the lack of concomitant IL-12 production. The downregulation of NKG2D required direct monocyte-NK cell contacts, as it was completely lost in the Transwell system. This suggests that, in addition to produce soluble TGFβ, NS5A-stimulated monocytes might express membrane-bound TGFβ, which would further participate in NKG2D modulation through direct contact with NK cells. In support of this hypothesis, myeloid-derived suppressor cells (MDSCs), a subpopulation of immature myeloid cells with suppressor functions, were shown to downregulate NKG2D expression and inhibit liver NK cell cytotoxicity in cancer-bearing mice, through expression of membrane-bound TGFβ and direct contact with NK cells [59]. Furthermore, MDSCs were recently shown to inhibit NK cell functions through direct cell-cell contact in the context of hepatocellular carcinoma in humans [60], [61]. Whether a subpopulation of NS5A-binding monocytes characterizes MDSCs able to suppress NK cell activity through TGFβ production is under investigation in our laboratory. Further studies will be needed to determine the role of MDSCs in HCV-infected patients. The effect of NS5A on monocytes is reminiscent of other proteins from persistent viruses. Interaction of HTLV-1 p30 protein with TLR4 signaling stimulates the release of IL-10 and hampers the release of pro-inflammatory cytokines from macrophages [25]. HIV Tat-induced IL-10 production by monocytes is regulated by p38 MAPK [62]. The LMP1 protein of EBV also induces IL-10 via p38 and PI3 kinase activation [63]. The vaccinia virus A52R protein activates p38 and JNK, and promotes TLR4-induced IL-10 production, while inhibiting NFkB-dependent genes IL-8 and RANTES [64]. Human major group rhinoviruses [26] downmodulate the accessory function of monocytes by inducing IL-10 production and inhibiting IL-12 production. Together, those reports and our findings open the idea that engagement of TLR4 may generate negative signals that are necessary for immune subversion and viral persistence. NS5A is localized in the perinuclear regions of the infected cell, but is not present in the circulating virions. However, there is now increasing evidence that HCV mediates hepatocyte apoptosis [31], [32], [33], [65], [66], which may allow HCV proteins to be released in the extracellular milieu. We observed that supernatants of Huh7. 5. 1 human hepatoma cells transfected with the JFH1 infectious replicon reproduced the effect of recombinant NS5A on NKG2D downmodulation, suggesting that NS5A might be released in the culture medium from apoptotic cells. However, direct proof for the presence of NS5A protein in HCV-infected cell supernatants is still lacking, as it was not accessible by anti-NS5A mAb in Western blot or depleting experiments. The possibility that a fraction of bioactive NS5A only becomes liberated after interaction of the cell debris with monocytes is supported by our observation of a comparable NS5A binding on monocytes of recombinant protein and of supernatants from JFH1-replicating Huh7. 5. 1 cells. To reconcile this idea with our finding that the rare NK cells present in HCV-infected liver expressed high NKG2D levels, one may envision a scenario in which the local cytokine microenvironment of the liver sinusoids, in particular IL-15 (which is produced by Kupffer cells) can inhibit the effect of TGFβ and enhance NKG2D expression. Indeed, IL-15 antagonizes the TGFβ immunosuppressive effects through blockade of the Smad3 signaling pathway [41], [42], [43]. Expression of IL-15 within HCV-infected livers was reported to show a sinusoidal distribution [67]. We found that the rare intrahepatic NKG2D-positive cells were located in sinusoidal tracts, but not in parenchymatous areas or necroinflammatory lesions where NS5A release by apoptotic infected cells is likely to occur. Moreover, we showed that IL-15 could fully prevent the TGFβ-mediated modulation of NKG2D and NK cell functions in vitro. A possibility is that once having migrated to areas of inflammation in the liver, most NK cells would be induced to apoptosis. This phenomenon might be favored by the abnormal expression of the Programmed-Death 1 (PD-1) molecule on NK cells, which has been observed in chronically-infected HCV patients [68]. Only NK cells expressing high levels of NKG2D would preferentially home to the liver, or could survive in the liver due to their resistance to apoptosis under inflammatory conditions. This hypothesis is the matter of current investigation in our laboratory. That NKG2D levels, either on peripheral or liver-infiltrating cells, were not correlated with virological or histological markers of the liver disease has also been observed by others [39] and might reflect such complex interactions. Unfortunately, our attempt to clarify whether increased NKG2D expression on liver NK cells is associated with enhanced cytotoxic activity was hindered by the highly restricted access to fresh liver biopsy tissue from chronically infected patients. The availability of noninvasive biomarkers for first-line assessment of liver fibrosis has led to a dramatic decrease in the use of liver biopsy for patients with chronic hepatitis C. Regrettably, functional analysis of liver-infiltrating immune cells in the few patients still undergoing liver biopsy is probably not representative of natural HCV infection. Altogether, our observations raise the idea that reducing IL-10 and/or TGFβ bioavailability could be a suitable means to restore NK cell functions in chronic hepatitis C. However such approach could dangerously modify the overall equilibrium between effector and regulatory mechanisms. Rather, we propose the use of IL-15 - or biologically active soluble IL-15/IL-15Rα complexes [69] - as an adjuvant therapeutic agent to restore NKG2D-mediated NK cell functions. Notably, the pathways triggered by IL-15 receptor signaling are required for the NKG2D-mediated signal transduction and cytotoxicity [70]. Jinushi et al. showed that dendritic cells (DCs) from HCV-infected patients have impaired IL-15 production upon stimulation by IFNα [71]. Given the role of NK cells in promoting optimal initiation of adaptive CD8 T cell responses, and the role of IL-15 in the proliferation and survival of NK and CD8 T cells, IL-15 might help not only in establishing strong innate responses, but also in inducing more robust antiviral CTL responses. The HCV viremic patient group consisted of 34 chronically infected patients (anti-HCV antibodies and HCV RNA positive) who were naive of treatment, or who discontinued treatment at least 6 months before study. The HCV aviremic group was composed of 9 subjects with sustained viral response (SVR) following IFNα and/or ribavirin therapy, with viremia remaining undetectable for at least 6 months at the time of study. The main clinical characteristics of the patients are shown in Table 1. Patients with primary biliary cirrhosis (n = 4) or autoimmune hepatitis (n = 5) were used as non-infectious chronic inflammatory liver disease controls. An additional series of 11 HCV viremic patients was studied for paired analysis of circulating and liver-infiltrating mononuclear cells. The control group consisted of 23 age and sex matched blood donor volunteers seronegative for HCV. The study was performed in accordance with the Declaration of Helsinki and French legislation, and received approval of the Grenoble University Hospital ethical committee (03/APTF/1). All study participants provided written informed consent. Blood samples were processed within 2 h of collection and PBMCs were separated by Lymphoprep gradient centrifugation (Biowest). NK cells or monocytes were freshly purified from PBMCs by negative selection using magnetic microbead separation kits (Miltenyi Biotec) with purity higher than 90%. Liver-infiltrating mononuclear cells were isolated from fresh biopsy as reported [72] and processed immediately for staining and flow cytometry. Cells were incubated for 20 min at 4°C with combinations of the following antibodies: CD3-FITC, CD56-PE, CD8- or CD4-PerCP (BD Biosciences); NKG2D-APC or isotype-matched control antibodies of irrelevant specificity (BD PharMingen). Cells were fixed in 1% formaldehyde and analyzed on FACSCalibur (BD Biosciences), collecting a total of 100,000 events in a live gate, and data were analyzed using FlowJo software. NK cell cytototoxic potential was studied using CD107a as a marker of degranulation. Freshly isolated NK cells were incubated in the presence or absence of K562 cells, C1R cells or C1R-MICA transfectants (a generous gift from A. Toubert, Hopital St-Louis, Paris, France) as target cells. CD107a-Pe-Cy5 antibody (BD) was added directly to the tubes at 20 µg/mL. After 1 hour at 37°C in 5% CO2, brefeldin A (10 µg/ml, Sigma) and monensin (6 µg/ml, Sigma) were added for additional 5 hr, and cells were stained with CD3-FITC and CD56-PE antibodies, fixed and analyzed by flow cytometry. Where indicated, NKG2D blocking antibody (20 µg/ml, Coulter Immunotech) was added. For intracellular IFNγ analysis, NK cells were incubated for 6 hr with K562 cells, fixed following staining with anti-CD3 and anti-CD56, permeabilized with 0. 2% saponin and stained with IFNγ-PE antibody (BD) for an additional 30 min. Recombinant TGFβ and IL-15 were purchased from R&D Systems. Cytokine levels were quantified using ELISA (TGFβ, IL-10 and IL-15 quantikine kits from R&D Systems; IL-12 ELISA kit from Diaclone). Soluble MICA was measured in the sera with a sandwich ELISA as described [73]. Recombinant soluble MICA was consistently detected at concentration of 0. 2 ng/ml. The following genotype 1a-derived recombinant HCV proteins were used: E. Coli-derived full length core, NS3, NS4 and NS5 (Axxora LKT). In confirmatory experiments, we used E. Coli-derived rNS5A amino acid 2061–2392 (Axxora LKT) and yeast-derived rNS5 2054–2995 (ibtsystems). Recombinant HCV-E2 protein (Immunodiagnostics) was purified from baculovirus-infected insect cells. β2microglubulin was used as control for E. Coli-purified protein. All proteins were used at a final concentration of 0. 1 to 1 µg/ml. Endotoxin levels determined by the limulus amebocyte lysate assay (BioWhittaker Cambrex) were between 0. 05 and 0. 2 endotoxin unit/µg protein (0. 054 EU/µg for the full length NS5A protein from Axxora LKT used in most experiments). To ensure that trace amount of endotoxin did not contribute to the observed responses, rNS5A was subjected to polymyxin B (10 µg/ml) (Sigma–Aldrich, St. Louis, MO, USA) for 15 min at room temperature. For blocking experiments, cells were incubated with 10 µg/ml of neutralizing mAb to TLR4, TLR2 or CD14 (eBioscience), soluble IL-10 receptor, anti-IL-10 neutralizing antibody (R&D Systems) before the addition of HCV protein. Isotype-matched antibodies were used as controls (Coulter Immunotech). Inhibitors of the signaling molecules JNK (SP600125), p38 (SB203580), PI3 kinase (LY294002), and MEK1 (U0126) were from Calbiochem. Huh-7. 5. 1 cells were kindly provided by Pr. Francis V. Chisari (The Scripps Research Institute, La Jolla, CA), and grown in Dulbecco' s modified Eagle' s medium-based medium as described [35]. Productive HCV infection was achieved as described [34], [35]. Briefly, Huh-7. 5. 1 cells were transfected with genomic HCV RNA transcribed in vitro from the plasmid pJFH1 [34] (a kind gift from Takaji Wakita, National Institute of Infectious Diseases, Tokyo, Japan) used as template, and cells were then passaged when necessary to maintain subconfluent cultures throughout the experiment. Cultures were probed for the frequency of HCV protein-expressing cells by in situ immunofluorescence, and infectivity titers in culture supernatant were assessed by focus-formation assay [35]. For binding experiments, monocytes were incubated for 30 min at 4°C with supernatants from non-infected or JFH1-replicating Huh-7. 5. 1 cells, or with 0. 5 µg/ml of recombinant NS5A (positive control) or NS4 (negative control). Where indicated, culture supernatants were passed through a 0. 45-µm filter following low-speed centrifugation to remove cellular debris. After washing and blocking with human IgG, cells were incubated for 40 min at 4°C with the mouse 9E10 mAb specific for genotype 2a NS5A (a generous gift from C. M. Rice, Rockefeller University, NY, USA) [74], followed with PE-labeled goat anti-mouse Ig, and analyzed by flow cytometry. Expression of MIC was evaluated on liver biopsy samples submitted to the Department of Pathology for diagnostic purpose. Paraffin-embedded liver biopsy sections (12 patients) were stained with anti-MIC (clone SR99 [75]) or anti-NKG2D (R&D Systems) mAb, followed with biotinylated goat anti-mouse Ig. For double immunofluorescence staining, cryosections (3 patients) were stained overnight at 4°C with anti-HCV-NS5A mAb (clone 7-D4, BioDesign), followed with FITC-labeled goat anti-mouse IgG, then incubated with biotinylated anti-MIC mAb, followed with streptavidin-Cy3. Slides were mounted with DAPI-containing medium (Vector Laboratories) and analyzed by immunofluorescence (Eclipse E888, Nikon) or confocal laser scanning (TCS SPS AOBS model, Leica) microscopy. All statistical tests were performed using Stata software (version 8. 0). Qualitative values between groups were compared using the chi-square test or Fisher' s exact test, and quantitative values were compared using the non-parametric Mann-Whitney U test. The Wilcoxon test was used to compare matched pairs. Correlation between two variables was determined using Spearman' s coefficient (rho). Two-sided P values less than 0. 05 were considered significant. NKG2D, P26718 (NKG2D_HUMAN); Toll-like receptor 4, O00206 (TLR4_HUMAN); NS5 protein, Q81596 (Q81596_9HEPC).

Natural killer (NK) cells are part of the innate immune response against virus infection. Their activation is the net result of signals emanating from a panel of inhibitory and activating receptors, among which the NKG2D activating receptor plays a major role. NKG2D ligands, the MHC class I related Chain (MIC) molecules, are induced on HCV-infected hepatocytes. In this paper, we show that NKG2D expression is decreased on NK cells from chronically infected HCV patients. As a consequence, NK cell cytolytic and IFNγ-producing functions are impaired. We show that this phenomenon is mediated by TGFβ produced by monocytes upon stimulation by the non-structural HCV-NS5A protein. NS5A could bind to TLR4 on monocytes, thus inducing the production of IL-10 and TGFβ, while inhibiting the production of IL-12. We further showed that TLR4-dependent IL-10 production by monocytes upon NS5A stimulation was mediated through the p38 and PI3 kinase pathways. In addition, we demonstrated that IL-15 could inhibit the TGFβ-mediated effects on NKG2D expression and NK cell functions. Collectively, these results identify a new dampening signal used by HCV to subvert innate immune response, and may provide new insights into the design of new strategies to restore NK cell functions in chronic hepatitis C.

Abstract Introduction Results Discussion Materials and Methods

immunology/immune response immunology/innate immunity immunology/immunity to infections

2010

Hepatitis C Virus (HCV) Evades NKG2D-Dependent NK Cell Responses through NS5A-Mediated Imbalance of Inflammatory Cytokines

When visual contrast changes, retinal ganglion cells adapt by adjusting their sensitivity as well as their temporal filtering characteristics. The latter has classically been described by contrast-induced gain changes that depend on temporal frequency. Here, we explored a new perspective on contrast-induced changes in temporal filtering by using spike-triggered covariance analysis to extract multiple parallel temporal filters for individual ganglion cells. Based on multielectrode-array recordings from ganglion cells in the isolated salamander retina, we found that contrast adaptation of temporal filtering can largely be captured by contrast-invariant sets of filters with contrast-dependent weights. Moreover, differences among the ganglion cells in the filter sets and their contrast-dependent contributions allowed us to phenomenologically distinguish three types of filter changes. The first type is characterized by newly emerging features at higher contrast, which can be reproduced by computational models that contain response-triggered gain-control mechanisms. The second type follows from stronger adaptation in the Off pathway as compared to the On pathway in On-Off-type ganglion cells. Finally, we found that, in a subset of neurons, contrast-induced filter changes are governed by particularly strong spike-timing dynamics, in particular by pronounced stimulus-dependent latency shifts that can be observed in these cells. Together, our results show that the contrast dependence of temporal filtering in retinal ganglion cells has a multifaceted phenomenology and that a multi-filter analysis can provide a useful basis for capturing the underlying signal-processing dynamics. Sensory systems have to encode stimuli over wide input ranges, and neurons therefore adapt their processing characteristics to the encountered stimulus statistics. In the vertebrate retina, ganglion cells adjust their sensitivity and temporal filtering characteristics when visual contrast changes [1–9]. While several studies have identified different mechanisms that contribute to sensitivity changes of ganglion cells [9–13], the origins of the contrast-dependent changes in temporal filtering are much less understood. Early studies typically investigated changes in temporal filtering in the frequency domain by observing how the encoding of sinusoidal signals at different frequencies is affected by contrast [7,8, 14]. More recent work [1–3,6, 9] has shifted the focus towards measuring the filter characteristics in the time domain by using white-noise stimuli and computing the filter as the spike-triggered average (STA). In agreement with the frequency-domain studies, the STA analyses have shown that higher contrast leads to faster kinetics and a shift from low-pass to band-pass filtering characteristics. Here, we take a new perspective on the contrast dependence of temporal filtering in the retina by taking multiple parallel filters into account. As a direct extension of the STA, spike-triggered covariance (STC) analysis allows the extraction of a set of multiple relevant filters from experimental data [15–19]. Indeed, STC and related analyses of retinal ganglion cells have typically revealed several relevant filters, corresponding to different stimulus features that influence the cell’s spiking response [20–25]. We therefore here ask how the set of multiple stimulus features is affected when visual contrast changes. For example, one expectation may be that the entire set of features shifts so that it covers different regions of stimulus space for different contrast levels. Alternatively, a fixed set of features may suffice to capture the relevant stimulus space across contrast levels, but the relative importance of the individual features may be altered, which then results in the contrast dependence of temporal filtering as measured by the STA. Based on multielectrode-array recordings from isolated salamander retinas, we investigated these possibilities by applying spike-triggered analysis, both STA and STC, to assess the temporal filtering characteristics and the set of relevant features under high and low visual contrast. In order to investigate how the visual features represented by a single ganglion cell are affected when contrast changes, we conducted multielectrode-array recordings from isolated salamander retinas. Visual stimuli consisted of a spatially uniform white-noise flicker of light intensity. The contrast level of this flicker stimulus was alternated between a low level (12%), lasting for 100 sec, and a high level (32%), lasting for 20 sec. The relatively longer presentation of low-contrast stimulation allowed us to collect comparable numbers of spikes under both conditions, despite the higher firing rates for high contrast (Fig 1A), thus providing the basis for reliable filter estimation at both high and low contrast. We restricted our analysis to cells with Off-type filter shapes, which represent the vast majority of ganglion cells in the salamander retina [26–28]. Note, though, that under full-field steps of light intensity, many of such Off-type ganglion cells in the salamander retina show On-Off response characteristics to various degrees [26,29,30]. We obtained temporal filters for each contrast level by computing the STAs (Fig 1B). The filters showed the typical changes known from other studies of contrast adaptation [1–3,6, 9]; for higher contrast, the STA displayed a shorter time-to-peak and became more biphasic. Though we focus in this work on analyzing the filter changes, we also checked the contrast-induced sensitivity changes by computing the nonlinearities that describe the relation between the filter output and the resulting spike rate [31]. As expected, the nonlinearities were shifted to the right for higher contrast (Fig 1C), corresponding to reduced sensitivity. In order to determine whether multiple stimulus features affected a ganglion cell’s response, we performed STC analysis for each contrast level. We collected spike-eliciting stimuli, defined as those stimulus segments that preceded a spike. We then computed the covariance matrix of these spike-eliciting stimuli and subtracted the prior stimulus covariance matrix. From this covariance matrix difference, relevant features are identified by an eigenvalue analysis [15–18,32]. In brief, an eigenvalue that is significantly different from zero means that, along a particular direction in stimulus space, stimulus segments that elicited spikes and those that did not are distributed differently. The corresponding stimulus direction, which can be identified by the corresponding eigenvector, therefore denotes a stimulus feature to which the cell is sensitive. As observed previously for salamander retinal ganglion cells [20], the obtained spectra of eigenvalues (Fig 1D) typically displayed two or more eigenvalues that deviated substantially from the baseline at zero, as also confirmed by statistical testing (see Methods). Indeed, under the high-contrast condition, none of the analyzed 345 cells revealed fewer than two significant eigenvalues, reconfirming that ganglion cell responses are affected by multiple visual features. Similar to previous observations [20], we observed different types of eigenvalue spectra, in particular with respect to the occurrence of a significant positive eigenvalue (Fig 1D). To analyze the contrast dependence of the feature sets separately for different types of eigenvalue spectra, we sought a practically useful subdivision of our recorded cells. To do so, we obtained a measure of the importance of each identified feature by computing how much information it carried about the ganglion cell’s response (see Methods). We then compared the information values corresponding to the features from the high-contrast data with the most positive eigenvalue, v1, and with the two most negative of the 20 eigenvalues, v19 and v20 (Fig 2A). The scatter of the information values supported the existence of different types of eigenvalue spectra, although the number of types and their exact boundaries are not entirely clear from this data. Yet, for the purpose of the subsequent analyses, we found the following classification to be useful. Type I cells were defined as those for which the two most informative features were v19 and v20 (see Fig 1, top row, for an example). These cells were by far the most frequently observed ones in our recordings, and they typically showed no substantial information in v1. By contrast, Type II cells were those for which v1 was particularly informative (Fig 1, center row), here defined as having an information rate of at least 0. 8 bits/spike. Finally, Type III cells were defined as the cells in between, with a v1 feature that was less informative than those of Type II cells, but more informative than v19 (Fig 1, bottom row). Altogether, this procedure separated the analyzed cells into 293 cells of Type I, 24 cells of Type II, and 28 cells of Type III. Note that we do not claim that these types match actual ganglion cell types, and we cannot exclude that the data form a continuum. Type III cells, for example, border on the large group of Type I cells, yet an analysis of the differences in information values between v1 and v19 (Fig 2B) speaks against a single broad distribution and justifies treating Type III cells as a separate group. The analysis of the information rates also showed that information deteriorated quickly after the first two most informative features (Fig 2C) and that the combination of the two most informative filters captured around 80% of the total information contained in the occurrence of individual spikes (Fig 2D), similar to previous findings [20]. Given the good performance of the two-feature model and the small information values of further features, we focused our subsequent analysis for each cell on the two most informative visual features. This choice allowed us to analyze all cells across both contrast levels in a unified fashion, despite the fact that the actual number of significant features depended sensitively on the number of analyzed spikes and on the cutoff used to determine significance and varied between cells and contrast levels. In the following, we denote the two most informative features by k1 and k2, respectively. For cells of Type I, k1 and k2 corresponded to v20 and v19, respectively, whereas k1 and k2 were defined as v1 and v20 for Type II cells and as v20 and v1 for Type III cells. How the identified features relate to a cell’s spiking activity can be illustrated by projecting all spike-eliciting stimuli onto the two identified features k1 and k2 (Fig 1E). This shows that the positive eigenvalue for Type II cells corresponds to the occurrence of distinct clusters of spike-eliciting stimuli (Fig 1E, center row), whereas Type III cells display a continuously elongated region of spike-eliciting stimuli along the dimension of the positive eigenvalue (Fig 1E, bottom row). The distinct clusters of spike-eliciting stimuli, as found here for Type II cells, had previously been shown to correspond to spikes elicited by the On and the Off pathway, respectively [20,21,23,24,33]. Note that responses to full-field steps of light intensity did not distinguish these cells from the other two types because many cells in our recordings from all three types had some level of On-Off response characteristics under light-intensity steps (Fig 2E). Yet, the appearance of two clusters in the spike-eliciting stimuli suggested that cells of Type II have particularly strong and distinct On responses even under the applied flicker stimuli. This occurrence of two clusters of spike-eliciting stimuli was a defining characteristic of Type II cells; in fact, our criterion for separating between Type II and Type III cells based on the information of v1 assigned all cells with such clusters to Type II. For each cell, the basic structure of the eigenvalue spectrum was generally similar for the high-contrast and low-contrast data (Fig 1D). In particular, the occurrence of a significant positive eigenvalue did not depend on the contrast level. Yet, closer inspection revealed subtle differences; in general, slightly more eigenvalues deviated significantly from the baseline at zero when contrast was higher, in particular for cells of Type I and II. These differences did not result from different spike numbers under the two contrast conditions, as they persisted when spike numbers were matched by discarding surplus spikes for either of the two contrast conditions. This yielded an average of 4. 5±0. 1 significant eigenvalues (mean±SEM) at high contrast and 4. 1±0. 1 at low contrast. To evaluate this further, we calculated for each cell the difference in the number of significant eigenvalues, ∆EV, between high and low contrast, taking into account equal numbers of spikes for both conditions. We found that there were more significant eigenvalues at high contrast for Type I cells (∆EV = 0. 4±0. 1, mean±SEM, p<10−3, Wilcoxon signed-rank test) as well as for Type II cells (∆EV = 0. 7±0. 2, p<10−2), but not for Type III cells (∆EV = 0. 0±0. 2, p>0. 5). Based on the STC analysis of the three distinguished cell types, we then asked whether the identified stimulus features provide a useful basis for explaining the contrast-induced changes in the STA. To approach this question, we explored whether the STA could be described as a linear combination of the features k1 and k2. In particular, we asked whether a single such two-feature basis is sufficient to describe the STAs from both contrast conditions. This would mean that for both contrast conditions the cell remained sensitive to the same stimulus subspace, spanned by the STC-derived features, and that only the relative contributions of the features changed under contrast adaptation. To avoid using the same spikes for computing the STAs as well as the features, we split the data for each contrast level into a training set, from which we obtained k1 and k2, and a test set, from which we computed the STA (see Methods). Fig 3 shows how the features k1 and k2, obtained either under high contrast (Fig 3A, 3B and 3C) or under low contrast (Fig 3D, 3E and 3F), fitted the STAs from both contrast conditions for the three sample cells of Fig 1. Note that the STA fits with STC features derived from the same contrast level are expected to be good, as long as the relevant feature space is well described by just k1 and k2, because the STA is generally a linear combination of all relevant features [15,34]. The actual analysis of interest, on the other hand, is to fit the STAs with STC features derived from a different contrast level. If, for example, contrast adaptation alters the relevant feature space itself, these fits across contrast conditions should fail. To evaluate the goodness of fit for the STAs of all recorded cells, we computed the coefficient of determination (R2) for each fit (Fig 4A and 4D). This coefficient measures the fraction of variance in the STA that is captured by the fit, so that values near unity correspond to a perfect fit. We also inspected the distribution of the weights for k1 and k2 obtained from the fits (Fig 4B and 4E). Note that the weights implicitly also denote the fit quality: because the features and the STA are normalized and because the weights correspond to the projections of the STA onto the features, the radius of the data points in these plots is bounded by the unit circle, which is reached only if the STA is a linear combination of k1 and k2, i. e. , if the fit is perfect. For Type I cells, we found–as expected–that the high-contrast basis provided a good fit for the high-contrast STA, as shown by the sample cell (Fig 3A) as well as by the coefficients of determination (Fig 4A). Yet, the same basis, obtained from the high-contrast STC analysis, also provided an excellent fit for the low-contrast STA, showing that a single basis served to capture temporal filtering at both contrast levels. The good fits for both STAs were achieved with different contributions from the two features; the k2 feature became relatively more important for the high-contrast STA, as reflected by the change in weights for k1 and k2 (Fig 4B). Surprisingly, the fit of the low-contrast STA with the high-contrast basis was even systematically better than the fit of the high-contrast STA with the same basis (Fig 4A). This implies that, while the high-contrast STC features provide a useful basis for either contrast level, additional features must become more relevant at high contrast. The low-contrast basis, on the other hand, provided a good fit only for the low-contrast STA, but not for the high-contrast STA (Figs 3D and 4D), indicating that the low-contrast basis lacks some of the structure that becomes relevant at high contrast. For Type II cells, in contrast to Type I cells, the fits of the STAs were altogether poor; neither the high-contrast (Figs 3B and 4A) nor the low-contrast basis (Figs 3E and 4D) provided a good fit to either of the two STAs. This indicates that additional dynamics beyond those captured by the feature basis k1 and k2 are required to describe temporal filtering in these cells. For Type III cells, the high-contrast basis generally provided good fits, similar to the case of Type I cells (Figs 3C and 4A). Unlike for Type I cells, however, the low-contrast basis for Type III cells also yielded good fits for both STAs (Figs 3F and 4D). This indicates that, for Type III cells, the relevant dynamics that mediate contrast adaptation are already present under low-contrast stimulation, in line with our earlier observation that the number of significant features did not increase with contrast for Type III cells, in contrast to the other two types. To investigate the effect of using only two features in these fits, we also fitted the STAs with all features detected as significant in the STC analysis (Fig 4C and 4F). Naturally, fits improved as compared to using the two-feature basis. The improvement was particularly pronounced for Type II cells, confirming that additional dynamics beyond the first two features are important for capturing temporal filtering in these cells. Otherwise, the findings remained similar to the two-feature fits. In particular, for Type I cells, low-contrast STAs were still generally better fitted than high-contrast STAs, confirming that temporal filtering at high contrast is affected by a larger number of emerging filters, not all of which reach significance in the STC analysis. The analyses above show that, for most cells, a single basis can fit the STAs from both contrast levels. In principle, this could result from STAs that just do not vary much with contrast so that a good fit at one contrast immediately implies a good fit at the other contrast. Thus, to check whether the features actually capture the variations of the STAs with contrast, we computed how much of the contrast-induced STA variance was explained by the fits with k1 and k2 (Fig 4G) by measuring the coefficient of determination for the difference of the STAs and the difference of their fits (see Methods). This analysis corroborated the distinction between Type I and Type III cells; for the former, the high-contrast basis captured the contrast-induced STA changes much more accurately, whereas the low-contrast basis was clearly superior for Type III cells. In order to explore the origins of the observed differences between cell types further, we compared our results to computational models of contrast adaptation. Mechanistically, the primary source for contrast adaptation is thought to be activity-dependent gain control in the form of synaptic depression at bipolar cell terminals [9,12,13] and adaptation in the spike generation mechanism of the ganglion cell [10,11]. We therefore performed the spike-triggered analyses on two established models that contain such activity-dependent gain control in order to check which of the three distinguished cell types are consistent with these models or require other dynamics. Both models have previously been shown to accurately capture many phenomena of contrast adaptation, in particular with respect to the observed changes in temporal filtering. The first model [35], termed here spike-feedback model, is based on a linear temporal filter, followed by a threshold evaluation to obtain spikes, together with negative feedback that is subtracted from the filtered signal after each spike (Fig 5Ai). The model can be used to predict individual spikes of retinal ganglion cells [35] and has been shown to capture the phenomenology of contrast adaptation surprisingly well [36]. The second model [37], termed LNK model, is based on a linear-nonlinear (LN) cascade, composed of a linear temporal filter and a nonlinear transformation, which is followed by a first-order kinetic process (Fig 5Aii, Methods). The kinetic process matches well the dynamics of synaptic depression at bipolar cell terminals, and the model has been shown to provide an accurate account of contrast adaptation at the level of the ganglion cell membrane potential [37]. When stimulated with high- and low-contrast white noise, both models showed changes in temporal filtering (Fig 5B) similar to experimental observation; STAs became narrower and more biphasic at high contrast. In both cases, STC analysis revealed significant negative eigenvalues, but no significant positive eigenvalue, reminiscent of Type I cells. We then analyzed how the two STAs could be fitted by the STC-derived features. For the spike-feedback model, differences between the STAs and STC-derived features were small, and good fits were obtained in all cases. For the LNK model, the STA fits showed similar behavior as observed for Type I cells. Only the high-contrast basis provided good fits for both STAs (Fig 5Cii and 5Dii), with a better fit for the low-contrast STA. Like for Type I cells, the latter follows from the growing importance of more than just the first two features at higher contrast (Fig 5Bii). Thus, these gain-control models are consistent with our findings for Type I cells, but not for the other two types. In particular, the models did not produce a positive eigenvalue, even when we varied model parameters over considerable ranges. This appears to be a generic feature, for which the following reasoning provides some intuition: The core element of the models is a single temporal filter followed by a threshold or a fairly steep nonlinearity, necessary to account for the experimentally observed response sparseness. Thus, only sufficiently strong activation of the filter triggers spikes, leading to a reduced variance of the spike-triggered ensemble along this stimulus dimension, corresponding to a negative eigenvalue. The negative-feedback dynamics in the models provide suppression of spikes, leading to reduced variance in the spike-triggered ensemble and thus further negative eigenvalues for the corresponding suppressive filters [15,38,39]. Therefore, we will in the following consider other dynamics for capturing the contrast dependence of temporal filtering in Type II and Type III cells. STAs of Type II cells were generally poorly fitted when using the two features derived from the STC analysis. To investigate this cell type further, we come back to the finding that Type II cells exhibited two clusters of spike-eliciting stimuli (Fig 1E), which had previously been attributed to represent inputs from both On and Off pathways [20,21,23,24,33]. We therefore separated the two clusters (see Methods) and analyzed their spikes separately to compute STAs and corresponding nonlinearities for each cluster at both high contrast (Fig 6A) and low contrast (Fig 6B). For each contrast, the two STAs yielded filter shapes that are characteristic for the On and Off pathway, respectively, including the known relative delay of the On-type filter [24]. Furthermore, while the original nonlinearity showed a slightly non-monotonic shape, indicative of two contributing pathways [23], the separate analysis of the two clusters yielded monotonically increasing nonlinearities. These findings supported the notion that the filters of the two clusters represent inputs through the On and Off pathway, and we therefore termed the filters kON and kOFF. The cells were dominated by the Off pathway; the corresponding cluster always contained many more spikes than that of the On pathway. The relative contribution of the On pathway, however, depended on the contrast level; at high contrast, a larger percentage of spikes was part of the On cluster (20±3% at high contrast, 8±2% at low contrast, p<10−3, Wilcoxon signed-rank test). Thus, contrast controls the relative effectiveness of the On and Off pathway in these cells, thereby influencing temporal filtering and how the filtered signals relate to the cell’s activity. To check whether these changes in contributions from the two pathways are sufficient to explain the contrast dependence of temporal filtering in these cells, we asked whether a single set of filters kON and kOFF can fit the STAs from both contrast conditions. Not surprisingly, the On and Off filters from the high-contrast data provided excellent fits of the STAs from the same contrast level (Fig 6C); by construction, the STA is a weighted average of the STAs from the two clusters. However, the fits of the low-contrast STAs with the high-contrast On and Off filters were typically poor and led to negative weights for the On filter (Fig 6C), inconsistent with the supposed representation of the physiological On pathway. Analogously, the low-contrast On and Off filters provided good fits only for the low-contrast STAs. This analysis indicated that contrast-dependent changes in stimulus filtering in these cells does not only follow from modified relative contributions of the On and Off pathways. We therefore hypothesized that, in addition, the two pathways adapt individually. To test this hypothesis, we performed spike-triggered analysis only for those spikes of Type II cells that were part of the Off-pathway clusters. The STC analysis of these spikes showed that the Off pathway of Type II cells resembled a Type I cell, with all significant eigenvalues being negative (inset of Fig 6Di). Furthermore, the Off-pathway STAs from both contrast conditions were now well fitted by the two most relevant features obtained from the high-contrast STC analysis (Fig 6D). Also, the high-contrast STC features provided a better fit of the low-contrast than of the high-contrast STA, as had been the case for Type I cells. We also attempted to perform a similar analysis with the spikes from the On-pathway cluster, but the much lower number of spikes contributed by this pathway rendered the STC analysis too noisy for a reasonable analysis of STA fits. Nonetheless, our findings suggest that the contrast-dependent changes in temporal filtering of Type II cells follow from independently adapting On and Off pathways, with the alterations in their relative contributions resulting from stronger adaptation in the Off pathway. This finding is in line with the previously reported stronger contrast adaptation in Off-type as compared to On-type ganglion cells in the salamander retina [3], suggesting that in general the Off pathway adapts more strongly in this system. For Type III cells, the positive eigenvalue was not associated with the occurrence of distinct clusters of spike-triggered stimuli. Furthermore, Type III cells differed from the other two cell types in that they allowed particularly good fits of the high-contrast STA by the low-contrast STC features and did not increase the number of relevant features at higher contrast. This suggests that contrast adaption of temporal filtering is less associated with additional emerging features. We therefore looked for other dynamics that might be of particular importance for these cells. Theoretical studies had shown that the occurrence of positive eigenvalues in the STC analysis together with an elongated distribution of spike-eliciting stimuli, as we had found for Type III cells, occur for the Hodgkin-Huxley model and are associated with effects on the precise timing of spikes [40]. A previous study [20] had hypothesized that positive STC eigenvalues observed in some retinal ganglion cells may similarly arise through such spike-timing dynamics. Furthermore, spike-time jitter and spike-latency shifts are known to affect temporal filters and eigenvalue spectra of spike-triggered analyses [41–43]. For studying contrast adaptation, such spike-timing dynamics may be particularly relevant, as contrast has a strong effect on the amount of jitter [44,45] as well as on spike latency [46–49]. We therefore hypothesized that Type III cells are cells with particularly strong spike-timing dynamics. In particular, activity-dependent shifts in spike latency might play an important role for the contrast dependence of temporal filtering because such shifts can lead to delays in the filters, consistent with our experimental observations (Fig 1B), whereas jitter primarily broadens the filters. We therefore analyzed whether there exists a systematic effect of the stimulus on spike latency. For our white-noise stimuli, even at fixed contrast, the activation level of a ganglion cell fluctuates over time, and some spikes are elicited when the activation of the cell is strong while others result from episodes of smaller activation level. Thus, we aimed at assessing the relationship between activation level and timing of associated spikes for each ganglion cell. To obtain an estimate of the activation level, we used the STA of the cell to filter the stimulus sequence and then detected the peaks in this filter output (Fig 7A), as these peaks likely play a major role in generating spikes. We then gathered all spikes in the vicinity of each peak and constructed histograms of relative spike timing for different ranges of the activation level (Fig 7B). This analysis indeed revealed a strong effect of activation level on spike timing. First, spike time histograms were broader at low contrast than at high contrast, indicating that spike time jitter was larger at low contrast. Second, for both high and low contrast, stronger activation levels led to relatively earlier spikes, indicating that spike latency directly depended on the activation level. These effects of activation level on spike jitter and on spike latency were observed for cells of all three types. We then tested whether these spike timing dynamics quantitatively differed between the cell types. To do so, we quantified the effect of the activation level on the spike latency by extracting the relative spike timing for which each of the histograms reached its maximum. We then plotted this spike-time shift against the mean activation level for the corresponding histogram (Fig 7C). These plots generally revealed a continuous, approximately linear decrease of the spike-time shift with increasing activation level. For each cell, we then computed the slope of the spike-time shift versus the activation level and compared the slope values between cell types (Fig 7D). For both high and low contrast, this analysis showed that Type III cells indeed displayed a much stronger dependence of spike timing on activation level than either Type I cells or Type II cells (p<10−6 in all cases, t-test). In fact, the stimulus-induced shifts in spike timing were about twice as big for Type III cells as compared to the other two types. The stronger dependence of spike timing on activation level in Type III cells can also be illustrated by averaging the dependence of spike-time shifts on activation levels over all cells of a given type (Fig 7E), showing that Type III cells span a bigger range of spike-time shifts than the other two types. The analysis of spike timing indicated that Type III cells display a particularly strong dependence of spike latency on activation level. To explore whether such spike timing dynamics could contribute to the contrast dependence of temporal filtering in these cells, we analyzed a simple model that included stimulus-dependent spike-time shifts, but no explicit contrast adaptation or negative feedback components. To do so, we generated spikes according to an LN model, followed by an inhomogeneous Poisson process (Fig 8A). Subsequently, we shifted the spike times, mimicking the spike-timing dependence observed in the data. Concretely, we used the output of the LN model as a measure of the activation level associated with each of the generated spikes. Depending on the activation level, each spike was then shifted to a later time, with smaller activation level corresponding to a larger shift (see Methods). Analysis of spike trains obtained from this model revealed a slower STA under low contrast (Fig 8B) and a significant positive eigenvalue in the STC analysis (Fig 8C), with a continuous, elongated distribution of spike-eliciting stimuli in the space spanned by the identified features k1 and k2 (Fig 8D and 8E). All these characteristics are strikingly similar to the experimental data of Type III cells. Note that without the explicit spike-timing shifts, the applied LN model would only show a single negative significant eigenvalue. The positive eigenvalue of the model was thus indeed a product of the spike-time shift. Finally, the STC-derived features k1 and k2 provided good fits of the model STAs (Fig 8F and 8G). In particular, the high-contrast STA was fitted well by the low-contrast basis (Fig 8Gi), which was a characteristic feature of TypeType III cells in the experimental data (cf. Fig 4D). This shows that the particularly strong spike-time shifts observed in Type III cells represent a viable model for explaining the specifics of temporal filtering and its contrast adaptation in these cells. We based our analysis of the multi-filter structure of temporal filtering and its contrast dependence on spike-triggered covariance (STC) analysis, rather than on likelihood-based [50,51] or information-theoretic [22,34,52,53] alternatives because the STC presents a straightforward extension of the STA, which has commonly been used as a standard technique for analyzing contrast adaptation. In addition, STC analysis requires relatively few assumptions about the underlying stimulus-response relationship, whereas fitting the data with a more rigid modeling framework might have made it difficult to detect the particularly strong influence of spike-timing dynamics in a small subset of neurons. Furthermore, we chose to perform the STC analysis separately for high and low contrast in order to test whether either contrast alone could deliver a subspace suited for generalization across contrast levels. The separate STC analyses turned out useful by revealing that Type III cells were specific in that the low-contrast basis fitted the high-contrast STA and that the number of relevant features did not increase with higher contrast. Using a simple white-noise stimulus and separately analyzing periods with different stimulus variance is a standard approach for assessing contrast adaptation [1–3,6, 9]. An interesting future direction may be to generalize the contrast-induced changes to more complex stimuli with spatial structure [54] or with continuous variations of contrast over time. A low-dimensional description of the contrast-induced changes in temporal filtering as investigated in the present work may be a useful starting point for this endeavor. A disadvantage of the STC approach is that there is no straightforward way of connecting the derived features to actual dynamics or mechanisms in the investigated system. In particular, the STC-derived features are constrained to be orthogonal to each other and bear significance primarily as a basis of the relevant stimulus subspace, which may alternatively be represented by any rotational transformation of this basis [15,34]. This is the reason why we did not compare the STC-derived features directly, but rather tested whether the STAs from both contrast conditions are contained in the same two-dimensional subspace. In principle, one might aim at identifying a more easily interpretable basis of the relevant subspace through subsequent analyses that incorporate appropriate prior assumptions, such as locality [55,56] or independence [57] of features. Yet, while locality seems an obvious assumption for spatial features, it is less clear which prior assumptions should guide the search for an appropriate feature basis of temporal filtering. Alternatively, one could apply a concrete model framework based on multiple filters and fit it to data via STC [38] or other methods [58,59]. Instead, we here chose to compare the STC analysis of the experimental data to the same analysis for different models to provide a qualitative comparison rather than aiming at improved interpretability of individual features. A central mechanism of contrast adaptation in the retina is synaptic depression at bipolar cell terminals [5,9, 12,13]. Indeed, our findings for Type I cells, which constituted the vast majority of cells in our recordings, were nicely reproduced by a kinetic model [37] that is well matched to the biophysics of vesicle depletion. In addition, spike generation contributes to contrast adaptation in salamander retinal ganglion cells through inactivation of sodium channels [10]. Such spike-dependent gain control is captured by the spike-feedback model. Phenomenological models similar to the ones considered here have also been used to include stimulus-driven feedforward suppression [7,60–65]. Schwartz et al. [38], for example, showed for a sample ganglion cell that a model with suppressive stimulus features, derived via an STC analysis, can capture the cell’s contrast-induced changes in the STA. In the retina, a typical source of stimulus-driven suppression is feedforward inhibition. Yet, inhibitory mechanisms alone cannot account for contrast adaptation, as the adaptive effects have been shown to persist under pharmacological blockade of inhibition in the retina [9,54,66]. While we cannot exclude that alternative gain-control models might provide a good match for the results of Type II and Type III cells, specific observations led us to consider other dynamics for these cells. For Type II cells, the relevant ingredient was suggested by the clustered structure of the spike-triggered stimuli, indicating that convergence of On and Off pathway signals is relevant. Indeed, these cells typically showed responses to both increments and decrements of light intensity, yet this property was shared by a considerable fraction of cells from the other two types (Fig 2E). This is not surprising, as On-Off response characteristics by ganglion cells is a common feature of the salamander retina [23,26]. Type II cells are thus not an exclusive class of On-Off cells, but rather appear to correspond to cells with relatively pronounced On-type excitation under white-noise stimulation. In addition to the effects of On pathway signals, the Off pathway of Type II cells showed the contrast-adaptation characteristics of Type I cells. The need to separate spikes from the On and Off pathway for further analyzing Type II cells suggests that contrast adaptation occurs independently in each of these pathways. The simplest explanation is that a major component of contrast adaptation occurs before On and Off signals converge inside the ganglion cell, consistent with the implied role of synaptic depression between bipolar cells and ganglion cells. Type III cells were distinct from the other two types in that they did not show emergence of additional features at higher contrast: the number of significant eigenvalues in the STC analysis did not differ between contrast levels, and the low-contrast STC analysis already provided features that captured temporal filtering even at high contrast. Instead, we found that these cells have a particularly strong dependence of spike timing on the activation level. Different mechanisms could contribute to the observed systematic latency shifts. For example, nonlinear gain control can effectively shift spikes forward in time for higher contrast by truncating the response at its tail end [67]. A similar scenario has been shown to occur in the spike-feedback model [36]: stronger activation leads to earlier threshold crossing, whereas later response parts are suppressed by the accumulation of negative feedback. Yet, this model alone produces a positive eigenvalue in the STC analysis only at fairly large levels of noise with spike jitter on the same time scale as the filter [20]. Another contribution could come from the temporal dynamics of spike generation itself. For example, models that describe spike generation through a saddle-node bifurcation are characterized by a strong dependence of spike delay on input current [68]. In biophysical terms, the weaker drive of a smaller stimulus allows slow inactivation processes, such as potassium currents or sodium-channel inactivation, to keep up with the cell’s depolarization and thereby create arbitrary long delays for sufficiently weak activation. The most basic classification of ganglion cells into subtypes distinguishes between On, Off, and On-Off cells, depending on the cells’ responses to steps in light intensity [69]. This scheme still represents the most fundamental cell-type classification also in the salamander retina [26], but alternative classification schemes have been suggested based on the analysis of the shapes of STAs [27,28,70] as well as based on the characteristics of multiple STC-derived features [20]. Yet, a clear segmentation of salamander retinal ganglion cells into distinct cell types or even a reliable estimate of the number of cell types is still lacking. In particular, tiling of receptive fields, which represents a hallmark of an identified cell type in the retina, has been observed in only few individual cases [28,71]. We here therefore followed a pragmatic course by classifying the analyzed ganglion cells into three broad classes based on the cells’ STC eigenvalue spectra, whose different structures are directly relevant for choosing the appropriate basis of the STA fits in this work. Clearly, some or all of these classes should be expected to contain multiple actual types of ganglion cells. The three classes also showed other differences than their STC spectra, though none of these clearly separated the groups. Type II cells had, on average, the strongest On-type responses under full-field steps in light intensity, and Type III cells tended to have slower filters than cells from the other two groups. Fairhall et al. [20] previously investigated salamander retinal ganglion cells with STC analysis and described five cells types, though stating that this did not represent a rigorous clustering, but rather a suitable representation of a continuum of response types. Still, similarities between the five response types described by Fairhall et al. and our three types suggest a rough match: The filter-and-fire and complex filter-and-fire cells of Fairhall et al. , which both had no positive eigenvalue and only differed in the number of negative eigenvalues, would thus correspond to our Type I cells. Furthermore, their bimodal cells showed two clusters of spike-eliciting stimuli with correspondence to On and Off stimuli and thus match our Type II cells. Finally, the Hodgkin-Huxley-like cells and the ring cells then correspond to our Type III cells, with a positive eigenvalue and a continuous distribution of spike-eliciting stimuli. Interestingly, Fairhall et al. observed particularly large spike-time jitter for their ring cells [20], matching another property of our Type III cells. Recently, different types of short-term plasticity had been reported in the salamander retina, with some cells showing sensitization, described as elevated sensitivity for several seconds after strong stimulation [71,72]. Here, we did not distinguish between sensitizing and classically adapting cells because our analysis focused on temporal filtering during the steady state of a given contrast level, rather than on transient sensitivity changes. It also seems that the comparatively long periods between contrast switches and the smaller difference between the contrast levels in our study might trigger sensitization less effectively. Similar analyses of how adaptation affects the multi-filter structure of sensory neurons have previously been performed in the somatosensory [73,74] and the auditory system [75]. Neurons in the somatosensory barrel cortex were found to be selective to multiple features under white-noise whisker stimulation. Changes in stimulus variance were then shown to lead to a gain change for each feature, but the features themselves remained invariant, and the gain change was approximately equal for all features [73]. Changes in the correlation structure of multi-whisker stimulation, on the other hand, altered the relative contributions of the features [74]. Likewise, for auditory neurons in songbirds, changes in variance of acoustic stimuli left the relevant features invariant, but changed the relative importance of the features [75], similar to our observations in the retina. Changing the mean stimulus intensity, on the other hand, fundamentally altered the features of the auditory neurons [75], indicating different functional roles of adaptation to mean and variance in this system. Together with the present work, these studies indicate that investigating adaptation phenomena from the perspective of multiple parallel filters can be a fruitful endeavor. The occurrence of multiple relevant stimulus features appears to be a ubiquitous finding, including, besides the retina, the downstream visual system [39,76–82], as well as other sensory systems, such as auditory [75,83,84], somatosensory [73,74,85], olfactory [86,87], and mechanosensory [88]. Furthermore, adaptive changes in sensory processing are found throughout different sensory systems. In particular, adaptation to stimulus variance in different auditory systems [75,89–93] and in the somatosensory system [73,85,94] as well as adaptation to luminance [95,96] and to other stimulus correlations [97,98] in the early visual system bear striking similarity to retinal contrast adaptation. Thus, the question whether adaptation alters the different stimulus features themselves or merely their relative contributions to signal processing pertains to a wide range of systems. All experimental procedures were performed in accordance with national and institutional guidelines and approved by the institutional animal care committee of the University Medical Center Göttingen (protocol number T11/35). Retinas were obtained from dark-adapted adult axolotl salamanders (Ambystoma mexicanum; pigmented wild type) of either sex. After enucleation of the eyes, retinas were isolated from the eyecup and cut in half. One retina half was placed ganglion-cell-side-down on a planar multielectrode array (Multichannel Systems, 252 channels, 10-μm electrode diameter, 60-μm spacing) for extracellular recording, while the other retina pieces were stored in oxygenated Ringer’s solution (110 mM NaCl, 2. 5 mM KCl, 1. 6 mM MgCl2,1. 0 mM CaCl2,22 mM NaHCO3,10 mM D-glucose, equilibrated with 95% O2 and 5% CO2) for later recording. The preparation was performed with infrared illumination under a microscope equipped with night-vision goggles. During recordings, retinas were continuously perfused with the Ringer’s solution at room temperature (20°C-22°C). The measured voltage signals were amplified, band-pass filtered between 300 Hz and 5 kHz, and digitized at a sampling frequency of 10 kHz. Potential spikes were detected by threshold crossing. Separation from noise and sorting into units representing individual ganglion cells was achieved by custom-made software, based on a Gaussian mixture model and an expectation-maximization algorithm [99]. Only well-sorted units with a clear refractory period were used for further analysis. Data are available from the Dryad Digital Repository (doi: 10. 5061/dryad. 7r7n7). To visually stimulate the retina, the screen of a gamma-corrected miniature OLED monitor (eMagin, OLED-XL series, 800x600 pixels) was focused through a telecentric lens onto the photoreceptor layer. The projection of the screen covered the recorded piece of retina. The size of individual image pixels on the retina was 7. 5 μm x 7. 5 μm. The stimulus screen was updated with a frame rate of 60 Hz and controlled through custom-made software, based on Visual C++ and OpenGL. Visual stimuli for the contrast-adaptation analysis were composed of a spatially uniform flicker of light intensity around a mean intensity of M = 39. 5 mW/m² in the photopic range. New intensity values were drawn randomly from a Gaussian distribution with standard deviation σ at a rate of 30 Hz. The contrast level C = σ/M alternated between 100-sec episodes of low contrast (C = 12%) and 20-sec episodes of high contrast (C = 32%). For each experiment, we typically recorded a total of 120–240 such episode pairs of high and low contrast. To measure responses to steps in light intensity, we used 5–10 min of alternating 2-sec periods of bright and dark illumination at 100% contrast around the same mean illumination as for the flicker stimulation. To calculate average firing rates, spikes were counted over the duration of each illumination level, excluding the first 100 ms after each switch in illumination. For each recorded ganglion cell and both contrast conditions, the spike-triggered average (STA) was computed [31,100] and compared to the relevant features obtained from an eigenvalue analysis of the spike-triggered covariance (STC) matrix [15–19]. STC analysis provides a straightforward extension of the STA, which has been commonly used to characterize contrast adaptation. Furthermore, this analysis allowed us to extract multiple relevant stimulus features for assessing temporal filtering without having to assume a particular model of the origins of these features or their interactions. For both spike-triggered analyses, the stimulus sequence s (t) was defined as the sequence of contrast values, corresponding to the deviations from mean light intensity, normalized to unit standard deviation, and sampled at the stimulus update rate of 30 Hz. Spike trains were binned at the same sample rate. We used spikes from the entire span of each contrast episode (100 sec for low contrast, 20 sec for high contrast). This decision is justified by the fact that the investigated changes in temporal filtering are known to occur immediately after a switch in contrast [1], whereas sensitivity of the ganglion cells changes more slowly, as reflected by changes in the average firing rate over several seconds after a contrast switch. In order to verify that these non-stationary effects of sensitivity do not affect our results, we also performed the entire analysis with only the second half of each contrast episode, where sensitivity is nearly stationary. This added noise to the estimated filters because of the reduced data but otherwise had no effect on the results (S1 Fig). To calculate the STA and STC, we collected for each spike time tn the stimulus sequence sn (τ) = s (tn − τ) that preceded the spike, where the lag τ covers 20 time bins. The STA is then obtained as the average over all spikes, s¯ (τ) =〈sn (τ) 〉n. In practice, if there is more than one spike in a time bin, the corresponding stimulus sequence preceding that bin is multiplied by the number of spikes in the bin. STC analysis is based on the covariance matrix C of spike-triggered stimulus sequences, Cτ1, τ2=〈[sn (τ1) −s¯ (τ1) ][sn (τ2) −s¯ (τ2) ]〉n. We subtracted the prior covariance matrix of all stimulus sequences, Cprior, which here is just the identity matrix because of the white-noise statistics of the stimulus. From the resulting matrix ΔC = C − Cprior, relevant features are found by diagonalizing ΔC to obtain its eigenvalues and corresponding eigenvectors. In essence, this analysis is similar to a principal component analysis of the set of spike-triggered stimulus sequences, except for the subtraction of the prior covariance matrix, which shifts the baseline of eigenvalues to zero and allows for the occurrence of negative eigenvalues. Significant deviations of eigenvalues from the baseline indicate that the corresponding eigenvectors represent relevant visual features. To compare shapes of the obtained temporal filters from STA and STC analysis, all filters were normalized so that the sum of squared filter components equaled unity. For plotting filters, we upsampled all filters from the original 20 sample points to 96 sample points by cubic spline interpolation. To assess eigenvalue significance, we applied a spike-shuffling technique [15,17] by randomly time-shifting individual spikes within their corresponding stimulus episode of either high or low contrast and performing the STC analysis with the shifted spike train to get an eigenvalue spectrum that is determined by sampling noise. After repeating the time-shift analysis 1,000 times, we estimated the 95% confidence interval for the eigenvalue spectrum and compared it to the true eigenvalue spectrum. If the maximal or minimal eigenvalue lay outside the confidence interval, it was regarded as significant. This analysis was then repeated in a nested fashion [15] by projecting out the corresponding significant eigenvector from all stimulus sequences and repeating the analysis in the reduced stimulus space to test for further significant eigenvalues. This procedure was iterated until both the maximal and minimal eigenvalue of the remaining spectrum lay inside the confidence interval. We computed nonlinearities corresponding to individual filters by a histogram method [31]. To do so, stimuli were first convolved with the respective temporal filters to obtain a generator signal. This generator signal was then binned into 40 bins in a way so that each bin contained the same number of data points. We displayed the nonlinearity by plotting the average generator signal against the average spike rate for each bin. We evaluated the relevance of different features by computing the information transmitted by a feature k about individual spikes [101] according to Ifeature (k) = ∫ ds P (s | spike) log2 (P (s | spike) / P (s) ), where s represents the projections of the stimulus sequences onto the feature k, P (s) is the probability distribution of the prior stimulus ensemble along the direction k, and P (s | spike) is the probability distribution of spike-triggered stimuli along this direction. The integral was evaluated by discretizing the projected stimulus values s with a bin size equal to 0. 1 of the stimulus standard deviation. We also computed the information captured by two features k1 and k2 with the same formula by replacing the distributions of s with the two-dimensional distributions over the two corresponding stimulus projections s1 and s2 [101]. For a subset of the recorded cells, we also measured the total information transmitted by single spikes [101]. To do so, we used a stimulus where the same white-noise sequence of 100 sec at low contrast and 20 sec at high contrast was repeated for 50 trials or more to obtain the time-varying firing rate r (t). We then computed the total single-spike information as Ione spike=1r¯T∫0Tdtr (t) log2r (t) r¯ where r¯ is the mean firing rate and T is the stimulus duration. The integral was evaluated by discretizing time according to the stimulus update rate. All information values were corrected for bias that arises from finite sampling by analyzing subsections of the data (80–100%) and using linear extrapolation to estimate the information value at infinite sample size [101,102]. The primary use of the information-theoretic analysis in this work was to order the features obtained from the STC analysis according to their information content and to provide an absolute measure of how relevant the features were. In particular, we used this analysis to classify the ganglion cells. We calculated the information for each eigenvector vi of the STC analysis, with vi denoting the eigenvector of the i-th largest eigenvalue (i = 1, …, 20). To reduce the effect of noisy cells on our population analysis, we discarded cells that had no eigenvector with an information rate of at least 0. 8 bits/spike or for which the two most informative eigenvectors did not lie at the ends of the eigenvalue spectrum, for example, if the most informative feature was v20, but the second-most one was neither v19 nor v1. We thereby discarded 98 out of 443 recorded cells. For further analysis, we selected those two eigenvectors with maximal information transmission as the visual features k1 and k2. These then always corresponded to a set of two of the features v1, v19, and v20, with vi denoting the eigenvector of the i-th largest eigenvalue (i = 1, …, 20). To distinguish between cell types, we thus focused our analysis on how the cells were distributed in the space spanned by information values corresponding to v1, v19, and v20, as discussed in the Results section. To fit an STA by visual features derived from the STC analysis, we subdivided the data for each contrast condition into a training set (80% of the contrast episodes, randomly chosen) to obtain the STC-derived features and a test set (the remaining 20% of the episodes) to obtain the STA. The STA s¯ (τ) was then fitted by a linear combination of the two features k1 and k2: s^ (τ) =α1⋅k1 (τ) +α2⋅k2 (τ). The weights α1 and α2 were optimized according to a least-squares criterion. The goodness of fit was quantified by the coefficient of determination R2, based on the STA s¯ (τ) and its fit s^ (τ): R2=1−∑τ (s¯ (τ) −s^ (τ) ) 2∑τ (s¯ (τ) −〈s¯ (τ) 〉) 2, where 〈⋅〉 denotes the average over time. For Type II cells, we also obtained an alternative basis for fitting the STAs by identifying filters kON and kOFF, which represent signal processing by the On and Off pathway of the retina, respectively. These filters were computed by subdividing the spike-triggered stimuli according to whether they corresponded to the On or Off pathway. Specifically, all spike-triggered stimuli were projected onto k1 and k2 and separated in this two-dimensional space into two clusters by k-means clustering. The average spike-triggered stimulus segment of each cluster then yielded the On and Off filters kON and kOFF, respectively [20,23,24,48]. To assess how much of the contrast-induced variations of the STAs were captured by each feature basis, we quantified how well the difference between the high-contrast and the low-contrast STA was captured by the differences of their fits. To do so, we computed the coefficient of determination for the STA difference. Concretely, based on the STAs s¯high (τ) and s¯low (τ) obtained under high and low contrast and of the corresponding fits s^high (τ) and s^low (τ) obtained with a single basis, we computed the coefficient of determination as R2=1−∑τ[ (s¯high (τ) −s¯low (τ) ) − (s^high (τ) −s^low (τ) ) ]2∑τ[ (s¯high (τ) −s¯low (τ) ) −〈s¯high (τ) −s¯low (τ) 〉]2. The delay of a spike after the spike-eliciting stimulus (i. e. , the spike latency) as well as its timing variability (spike jitter) can depend on the level of activation of a cell and thereby affect temporal filtering as assessed by spike-triggered analyses [41–43]. In order to compare such spike-timing dynamics for different ganglion cell types under white-noise stimulation, we estimated the dependence of spike timing on the activation level in the following way: for each cell and each contrast level, we filtered the white-noise stimulus with the corresponding STA and used the resulting filter output as an estimate of the activation level. We then identified all local maxima of this filter output and binned them according to their peak value into 40 bins so that each bin contained the same number of local maxima. For each bin, we collected the spike trains in a window of 367 ms around the local maxima. Averaging these spike trains provided us with a histogram of spike probabilities at different times relative to the local maximum of the filter output. Histograms for small peak values of the filter output suffered from noise, owing to the paucity of contributing spikes. We therefore restricted further analysis to the histograms corresponding to the highest activation levels, using ten and four histograms for the high-contrast and low-contrast stimulus, respectively. We then normalized the histograms to their peak values in order to analyze their shapes independently of the number of contributing spikes. For each spike-probability histogram, the position of its peak was obtained from fitting a 2nd-order polynomial through its largest data point and the two neighboring data points. The peak position yielded a characteristic spike-time shift for the corresponding activation level. We found that the relation between activation level and characteristic spike-time shift over the analyzed range of activation levels could be well fitted by a straight line. We therefore used the slope of this line, ∆Spike-time shift/∆Activation level (Fig 7D), as a measure of the sensitivity of spike-timing on activation level. We explored two computational models of ganglion cell activity that had previously been shown to capture essential aspects of contrast adaptation in these cells, in particular with respect to the temporal filtering characteristics of the cells. As stimuli for both models, we used Gaussian white-noise sequences, similar to the experiments, sampled in discrete time steps of ∆t = 33. 3 ms, with zero mean and standard deviation 0. 32 for high contrast and 0. 12 for low contrast, respectively. The first model [35,36] is based on a temporal filter of the stimulus, followed by a threshold operation to determine spikes. In addition, a feedback signal is subtracted from the feedforward filter signal before application of the threshold. Each spike increments the feedback signal by a fixed amount. In between spikes, the feedback decays exponentially with a given time constant. For the temporal stimulus filter, we used the experimentally measured STA, normalized to unit power, of a sample ganglion cell at low contrast. (Note, though, that the model did not aim at reproducing the particular responses of the sample cell. Rather, we aimed at analyzing the generic behavior of the model, based on parameters that matched the time scales of our data.) The choice of the other model parameter values was guided by the values used in [36], but using slower feedback decay to account for the relatively slow filter time scales of the salamander retina. Specifically, we used a threshold of 0. 2, a spike-triggered increase of the feedback signal by 3. 0, and a time constant for the decay of the feedback signal of 250 ms. The second model [37], called linear-nonlinear-kinetic (LNK) model, uses a sequence of linear stimulus filtering, nonlinear transformation, and a biophysically plausible first-order kinetic process to transform the stimulus into the membrane potential of a ganglion cell. Concretely, for the first two stages of the model, we used the same temporal filter as for the spike-feedback model, combined with the sigmoidal nonlinearity of the form 1 / (1 + exp (−a (x − θ) ) ) with a = 2. 5, θ = 2. The kinetic stage consisted of a resting state, an active state, and two inactive states, with transition rates between them. The value of the active state corresponds to the model’s output. As an input to the kinetic stage, the signal after the nonlinearity, termed u (t), modulates both the transition from the resting state to the active state as well as the transition from the first to the second inactive state. The transition rates between the states were slightly modified as compared to [37] to make recovery of adaptation slower because of the comparatively slower filters in our data. The applied values are indicated in the corresponding figure in the Results section. For both models, we also tested variations of the parameter values over some range and found that the obtained results and conclusions are robust to these variations. To selectively explore the effect of spike-timing dynamics on STC analysis and contrast-induced changes in temporal filtering, we also investigated a simple extension of an LN model, with added systematic spike-timing shifts, but no activity-dependent gain control. First, we implemented an LN model by again using the same stimulus structure and temporal filter as in the other two models. We then applied a threshold-linear transformation a[x − θ]+ with a = 20 Hz and θ = 0. 08, and determined spikes from the resulting firing rate according to an inhomogeneous Poisson process. Finally, each spike was delayed by an amount that depended on the value of the firing rate when the spike had been triggered. Specifically, we linearly divided the range of firing rates from zero to 15 Hz into four levels. Firing rates larger than 15 Hz were treated the same as those in the largest of these four levels. For each level L (L = 1, …, 4, with larger L corresponding to larger firing rates), we shifted the corresponding spikes by 4 − L units of the time step ∆t. Thus, spikes that had been generated (by chance) when the firing rate was actually small experienced a larger delay than spikes that had been generated by a higher firing rate.

Our sensory systems have to process stimuli under a wide range of environmental conditions. To cope with this challenge, the involved neurons adapt by adjusting their signal processing to the recently encountered intensity range. In the visual system, one finds, for example, that higher visual contrast leads to changes in how visual signals are temporally filtered, making signal processing faster and more band-pass-like at higher contrast. By analyzing signals from neurons in the retina of salamanders, we here found that these adaptation effects can be described by a fixed set of filters, independent of contrast, whose relative contributions change with contrast. Also, we found that different phenomena contribute to this adaptation. In particular, some cells change their relative sensitivity to light increments and light decrements, whereas other cells are influenced by a strong contrast-dependence of the exact timing of their responses. Our results show that contrast adaptation in the retina is not an entirely homogeneous phenomenon, and that models with multiple filters can help in characterizing sensory adaptation.

Abstract Introduction Results Discussion Methods

2015

Spike-Triggered Covariance Analysis Reveals Phenomenological Diversity of Contrast Adaptation in the Retina

Leishmania (L.) killicki (syn. L. tropica), which causes cutaneous leishmaniasis in Maghreb, was recently described in this region and identified as a subpopulation of L. tropica. The present genetic analysis was conducted to explore the spatio-temporal distribution of L. killicki (syn. L. tropica) and its transmission dynamics. To better understand the evolution of this parasite, its population structure was then compared with that of L. tropica populations from Morocco. In total 198 samples including 85 L. killicki (syn. L. tropica) (from Tunisia, Algeria and Libya) and 113 L. tropica specimens (all from Morocco) were tested. Theses samples were composed of 168 Leishmania strains isolated from human skin lesions, 27 DNA samples from human skin lesion biopsies, two DNA samples from Ctenodactylus gundi bone marrow and one DNA sample from a Phlebotomus sergenti female. The sample was analyzed by using MultiLocus Enzyme Electrophoresis (MLEE) and MultiLocus Microsatellite Typing (MLMT) approaches. Analysis of the MLMT data support the hypothesis that L. killicki (syn. L. tropica) belongs to the L. tropica complex, despite its strong genetic differentiation, and that it emerged from this taxon by a founder effect. Moreover, it revealed a strong structuring in L. killicki (syn. L. tropica) between Tunisia and Algeria and within the different Tunisian regions, suggesting low dispersion of L. killicki (syn. L. tropica) in space and time. Comparison of the L. tropica (exclusively from Morocco) and L. killicki (syn. L. tropica) population structures revealed distinct genetic organizations, reflecting different epidemiological cycles. Leishmaniases are vector-borne diseases caused by several Leishmania species that cycle between their phlebotomine sandfly vectors and mammalian reservoir hosts [1]. Leishmania parasites, like many other microorganisms, have a high adaptation capacity that allows them to invade and survive in various ecosystems. The spread of a parasitic genotype or group of genotypes in new ecosystems can lead to population differentiation. Consequently, new Leishmania taxa have regularly been described during the last decades [2–4]. Leishmania killicki could be considered as a typical example of this evolutionary process. Rioux et al. [5] identified this parasite in the Tataouine province (South Eastern Tunisia) for the first time in 1980. Then, sporadic cases were reported in Kairouan and Sidi Bouzid (Center of Tunisia), Gafsa (South Western Tunisia) and Séliana (Northern Tunisia) [6–8]. Besides Tunisia, this taxon was described in Libya [9] and Algeria [10–12]. The probable zoonotic transmission of this parasite, with the Ctenodactylus gundi rodent as reservoir and Phlebotomus (P.) sergenti as vector, was suggested but needs to be confirmed [13–17]. Data on L. killicki are scarce and the few available studies mainly focused on the detection and identification of this taxon using isoenzymatic or genetic approaches (PCR-RFLP, PCR-sequencing and PCR-SSCP) [18–21]. The isoenzymatic characterization using the MultiLocus Enzyme Electrophoresis (MLEE) technique identified four zymodemes for L. killicki. Zymodeme MON-8 (the most frequently identified) was found in isolates from Tunisia and Libya [5,9]; zymodemes MON-301 and MON-306 were identified in Algeria [10,11,18], and MON-317 was characterized in Tunisia for the first time [22]. In a recent taxonomic study, we confirmed that L. killicki is included within the L. tropica complex and we suggested calling it L. killicki (syn. L. tropica) [22]. Nevertheless, L. killicki (syn. L. tropica) epidemiology, transmission dynamics and why it is essentially described in Tunisia are still not well understood. The specific objective of this study was to provide new insights on the molecular epidemiology and transmission of L. killicki (syn. L. tropica). To this aim, we carried out a genetic study based on the analysis of nine microsatellite loci by MultiLocus Microsatellite Typing (MLMT) in a sample of 198 isolates from different Maghreb regions to explore the population structure of L. killicki (syn. L. tropica) and to compare the data with those of L. tropica populations from Morocco. A total of 198 samples were included in this study. They were composed by 154 Leishmania strains selected from the Leishmania collection of Montpellier, France (BRC-Leish, BioBank N° BB-0033-00052) and 44 samples collected by the research group of the Laboratoire de Parasitologie—Mycologie Médicale et Moléculaire (Monastir, Tunisia) during epidemiological investigations. These samples belonged to L. killicki (n = 85) and L. tropica (n = 113) and were identified over a period of 34 years (from 1980 to 2013). L. killicki samples were collected in Algeria (n = 7), Tunisia (n = 77) and Libya (n = 1). All the L. tropica strains were from Morocco since we have recently suggested that L. killicki and L. tropica from Morocco could have originated from a same L. tropica ancestor. Among the 198 samples, 168 were isolates from infected patients (Morocco [n = 113]; Tunisia [n = 47]; Algeria [n = 7]; Libya [n = 1]), 27 were DNA samples from human skin lesion biopsies (Tunisia), two were DNA samples from Ctenodactylus gundi bone marrow (Tunisia) and one was a DNA sample from a Phlebotomus sergenti female (Tunisia) (see supplementary data S1 Table). The L. killicki samples from Tunisia (n = 77) and the L. tropica samples Morocco (n = 113) were classified according to the area and period of isolation (S1 Fig). Although some of the isolates included in this study were previously characterized [5,7, 9,10,18,20,23,24], they were all (n = 168) analyzed again at the Centre National de Référence des Leishmanioses (CNRL), Montpellier (France) using the MLEE technique and 15 enzymatic systems, according to Rioux et al. [25]. Genomic DNA was extracted from the isolates using the QIAamp DNA Mini Kit, according to the manufacturer’s instructions, and eluted in 150 μl of AE buffer. The DNA samples from the 27 human skin biopsies, the two C. gundi and the P. sergenti were identified by polymerase chain reaction (PCR) amplification followed by digestion with BstU1 and Taq1, according to Haouas et al. [19]. The produced fragments were separated by electrophoresis on 3% agarose gels and compared with those of the WHO reference strains of L. major MON-25 (MHOM/MA/81/LEM265), L. infantum MON-1 (MHOM/FR/78/LEM75) and L. killicki MON-8 (MHOM/TN/80/LEM163). First, few randomly selected L. killicki (n = 10) and L. tropica (n = 25) strains were genotyped by amplifying the 21 microsatellite loci already used by Schwenkenbecher et al. [26] in order to select the best markers. All 21 loci could be amplified in the L. tropica samples. Conversely, only nine loci (six described by Schwenkenbechet et al. [27] and three by Jamjoom et al. [28]) were amplified in the tested L. killicki strains. These nine loci were used for genotyping the 198 samples under study (see supplementary S2 Table). All samples were amplified using the PCR conditions described by Schwenkenbecher et al. [27]: 2 min at 94°C and then 40 cycles of 94°C for 30 s, annealing temperature of each locus-specific primer set (2) for 30 s, 72°C for 1 min and a final extension step of 72°C for 10 min. The amplification products were visualized on 1. 5% agarose gels. Multiplex genotyping was done using 1 μl of PCR-amplified DNA added to the Genescan 500LIZ internal size standard and 13. 5μl of formamide in an automated sequencer. Genotyping data were analyzed with the Genemapper software v. 4. 0 to determine the fragment sizes. Fstat v. 2. 9. 3. 2 [29], updated from Goudet [30], was used for statistical analysis of the sample genetic polymorphism based on Nei’s unbiased estimator of genetic diversity (Hs) [31], the number of alleles per locus (N) and the mean allelic richness. The same software was also used for calculating the Wright’s F statistics [32] according to the Weir and Cockerham’s method [33]. The Fst coefficient reflects the inbreeding that results from the subdivision of the population into sub-populations of limited size, and measures the genetic differentiation between sub-populations. It varies between 0 and 1; values > 0. 25 reflect a high genetic differentiation [34]. Fst is considered significant when the p-value is ≤ 0. 05. The Fis coefficient estimates the inbreeding of individuals due to the local non-random union of gametes in each subpopulation. Fis values range between -1 and 1. A negative value indicates an excess of heterozygotes, a positive value corresponds to heterozygote deficiency. Genotypes obtained from the concatenated sequences of the nine microsatellite loci were used to calculate the global genotypic diversity Dg (Dg = number of genotypes per population/total number of genotypes). The Neighbor-Joining (NJ) phenetic tree was constructed using the MEGA 5. 10 software [35] from a Cavalli-Sforza and Edwards [36] genetic distance matrix obtained using the POPULATIONS software (http: //www. legs. cnrs-gif. fr/bioinfo/populations). Leishmania strains were obtained from the Leishmania collection (BRC-Leish, Montpellier, France, BioBank N° BB-0033-00052) which is part of the French network of Biological Resources Centres for Microorganisms (FBRCMi). This parasite collection is isolated over a period of many years and is completely independent of patients from which strains were isolated. All samples taken from humans were anonymized. Isoenzymatic characterization of the L. killicki (n = 55) and L. tropica (n = 113) strains was performed to confirm (n = 166) or to identify (n = 2) their zymodemes. Ten zymodemes were obtained (three for L. killicki and seven for L. tropica). L. killicki was represented by three zymodemes. MON-8 was identified in 44 Tunisian isolates and in the Libyan sample, while MON-301 was found in the seven Algerian isolates (see S1 Table). The newly described zymodeme MON-317 was identified in three strains (MHOM/TN/2009/MET122, MHOM/TN/2010/MET300 and MHOM/TN/2010/MET301) isolated from the focus of Gafsa (South West of Tunisia) (S1 Table). L. tropica was mainly represented by the zymodeme MON-102 (n = 76), followed by MON-113 (n = 22) and MON-107 (n = 6). The four remaining zymodemes were found only in few strains: MON-109 (n = 3), MON-112 (n = 3), MON-264 (n = 1) and MON-311 (n = 2) (S1 Table). The 198 samples were amplified using primers for the nine investigated loci. Clear electropherograms and two alleles per locus and per sample were obtained (see supplementary data S3 Table). Twenty nine alleles were obtained, ranging from two for the GA1, GM2 and LIST7027 loci to five for the GA11 and LIST7036 loci (mean: 3. 22 alleles per locus). The global genetic diversity was moderate (Hs = 0. 261) and the global genotypic diversity was high (Dg = 0. 53). The Fis values were positive at all loci and ranged from 0. 120 for the LIST7040 locus to 0. 920 for the GA6 locus (mean value = 0. 664) (S2 Table). Analysis of the genotyping data concerning all the L. killicki samples (n = 85) revealed 22 alleles ranging from a single allele for the GA1 and LIST7027 loci to five for the GA11 locus. The mean number of alleles per locus was 2. 55 and the value of the mean allelic richness was 1. 23. The global genetic diversity was low (Hs = 0. 185) and the global genotypic diversity was moderate (Dg = 0. 38) (Table 1). Comparison of the data for the L. killicki samples from Tunisia (n = 77) and from Algeria (n = 7) indicated that their genetic diversity was low (Hs = 0. 215 for the Tunisian strains and Hs = 0. 15 for the Algerian strains) and that the genetic differentiation between these populations was low, but significant (Fst = 0. 11, p = 0. 03) (Table 2). The Hs and Fst values were not calculated for L. killicki from Libya because only one specimen was available. Moreover, estimation of the genetic differentiation between the different Tunisian populations (strains from Gafsa, Tataouine and Kairouan-Séliana) and the Algerian samples showed that the genetic differentiation was important between the populations from Tataouine and Algeria (Fst = 0. 34, p = 0. 005) and lower but still significant between the samples from Gafsa and Algeria (Fst = 0. 09, p = 0. 01). No genetic differentiation was found between the Kairouan-Séliana and Algerian populations (Fst = 0. 1, p = 0. 18), possibly due to the small number of specimens from Kairouan-Séliana (n = 3) (Table 2). Analysis of the genetic diversity within the different L. killicki populations from Tunisia showed that the Gafsa strains (Hs = 0. 22) were more polymorphic than the Tataouine strains (Hs = 0. 15). The Hs value for the Kairouan-Séliana population was certainly biased because of the low number of strains and was not considered in this analysis. Finally, the genetic differentiation between the Gafsa and Tataouine populations was also high (Fst = 0. 3, p = 0. 002) (Table 2). Analysis of the genetic diversity of the L. killicki samples classified based on the time of isolation indicated that specimens isolated during the 1980–1989 period were less diversified (Hs = 0. 13) than those isolated between 2000 and 2009 (Hs = 0. 24) or 2010 and 2013 (Hs = 0. 2). The Hs value was not calculated for the 1990–1999 period because only one strain was collected during that time window. Genetic differentiation was important between the population isolated during the 1980–1989 period and the other populations. Conversely, no genetic differentiation was found between the populations collected between 2000 and 2009 and between 2010 and 2013 (Table 2). The Fst value was not estimated for the 1990–1999 window because only one strain was isolated in that period. Analysis of the genetic diversity of the L. killicki samples classified based on the region and time of isolation revealed relatively higher Hs values for the specimens collected in Gafsa at different times (Hs Gafsa [2000–2009] = 0. 26, Hs Gafsa [2010–2013] = 0. 28) than for those collected in Tataouine (Hs Tataouine [1980–1989] = 0. 13, Hs Tataouine [2000–2009] = 0. 16). The Kairouan-Séliana strains isolated at different periods and the Tataouine strains collected during the 1990–1999 period were not included in this analysis due to their limited number. Analysis of the genetic differentiation between these populations showed high Fst values that reflected temporal and geographical differences. However, a moderate genetic differentiation was found in the samples collected in Tataouine between 1980 and 1989 and between 2000 and 2009 and no genetic differentiation was observed between the strains isolated in Gafsa and Tataouine during the 2000–2009 period (Table 2). Finally, thirty-six genotypes were found. Genotype 24 was the most frequent (17. 95%) in the samples from Gafsa and Tataouine (Fig 1). Analysis of the genotype distribution in each focus and according to the time of isolation showed that most genotypes were specific to a locality or to a period of isolation (see Figs 1 and 2). Twenty six alleles were identified ranging from two for the GA1 and LIST7027 loci and five for LIST7036. The mean number of alleles per locus was 2. 88 and the mean allelic richness was 1. 98. The global genetic diversity (Hs = 0. 38) and genotypic diversity (Dg = 0. 63) were high (Table 1). Genetic diversity was also high when strains were classified according to the area of isolation in Morocco (Hs Azilal = 0. 34, Hs Essaouira = 0. 44, Hs Ouarzaezate = 0. 44, Hs Taza = 0. 38). For the strains from the locality of Salé, the Hs was not estimated because of their limited number (n = 3). Genetic differentiation was mainly not observed between strains from different localities; however, few Fst values were significantly different, although they were very low (from Fst = 0. 025, p = 0. 035 to Fst = 0. 05, p = 0. 05) (Table 2). Genetic diversity was high also when the L. tropica strains were classified according to the period of isolation ([1980–1989] Hs = 0. 35; [1990–1999] Hs = 0. 35; [2000–2009] Hs = 0. 43), whereas genetic differentiation was moderate but significant (Table 2). Comparison of the genotyping data showed strong genetic links between the L. killicki and L. tropica populations with 19 shared alleles among the 29 detected. Moreover, the NJ tree showed that L. killicki forms a monophyletic cluster inside the L. tropica complex (see Fig 3). Overall, the L. killicki population was characterized by lower genetic and genotypic diversity, fewer alleles per locus and lower allelic richness than the L. tropica population (Table 1). Analysis of the population structure showed an important genetic differentiation between the L. tropica population and the entire L. killicki sample (Fst = 0. 53, p = 0. 01) and also between the L. tropica population and the L. killicki populations from Tunisia [Fst = 0. 53, p = 0. 01] and from Algeria [Fst = 0. 5, p = 0. 01]) (Table 3). This result was confirmed also when the L. tropica population was compared with the L. killicki populations from the different locations in Tunisia (Gafsa, Tataouine, Kairouan Séliana) (Fst > 0. 4, p < 0. 05) (Table 3). Despite a great knowledge on Leishmania parasites, many taxa, such as L. killicki (syn. L. tropica), are still not completely characterized. The main objective of this study was to understand the epidemiology and transmission dynamics of L. killicki (syn. L. tropica) by analyzing its population structure and by comparing the genetic patterns of L. killicki (syn. L. tropica) and L. tropica populations in Maghreb. The comparison of L. killicki (syn. L. tropica) and L. tropica revealed a strong genetic differentiation associated with a lower genetic polymorphism within L. killicki (syn. L. tropica). Furthermore, the NJ tree showed that L. killicki (syn. L. tropica) creates a homogeneous and monophyletic cluster within the L. tropica complex. These data support the recently obtained MultiLocus Sequence Typing (MLST) results [22] suggesting that L. killicki (syn. L. tropica) emerged from L. tropica by a founder effect. The strong genetic differentiation indicates an independent evolution and an absence of gene flow between the two taxa after the founder event. The geographic distance and the ecological barriers between Morocco (area of isolation of all L. tropica specimens) and Tunisia, Libya and Algeria (regions of origin of all L. killicki (syn. L. tropica) samples) as well as the different transmission cycles can explain this diversification. Maghreb countries are essentially separated by mountains and the Sahara desert that could prevent the circulation and migration of Leishmania vectors and reservoirs. Furthermore, L. killicki (syn. L. tropica) transmission cycle is most probably zoonotic [14,15], whereas that of L. tropica appears to be both zoonotic or anthroponotic [24]. The comparison of L. killicki (syn. L. tropica) samples from Tunisia and Algeria revealed also a differentiation within this taxon, but lower than the one detected with L. tropica. This result supports the idea that L. killicki (syn. L. tropica) spread recently and may be still spreading between the different countries after the founder event. It is not known yet where the L. tropica subpopulation emerged to generate L. killicki (syn. L. tropica), but the number of reported cases suggests Tunisia. Despite the low sample size from Algeria, we detected a strong and significant genetic differentiation between the population from Tataouine and the samples from Algeria and a low genetic differentiation between the Gafsa and Algerian populations. These results seem to indicate a more recent diversification between the Gafsa and Algerian populations, supporting the hypothesis of a recent L. killicki (syn L. tropica) dispersion from Gafsa to Algeria. Conversely, the only isolate from Libya is genetically closer to the Tataouine than to the Gafsa population. This pattern is in agreement with the geographical distances/characteristics of these regions. Indeed, the mountains in the Gafsa area, where the probable reservoir (s) of L. killicki (syn L. tropica) live (s), belong to the Atlas Mountain chains, while mountains in the Tataouine region are connected to the Libyan mountains. Concerning the L. killicki (syn. L. tropica) populations from Gafsa and Tataouine, despite their low genetic diversity indices, they show a strong and significant genetic differentiation with a lower genetic diversity among the Tataouine samples. These data suggest that the Tataouine population is more recent and that these two populations are genetically isolated. The presence of geographical barriers separating the South West and South East of Tunisia (the Sahara desert and the Chott Djerid salt lake) could explain this structuring. Analysis of the spatio-temporal evolution of L. killicki (syn. L. tropica) in Tunisia shows a low circulation of genotypes between the different populations not only in space, but also in time within a region. Based on this observation and because most isolates were from infected humans, we can hypothesize that L. killicki (syn. L. tropica) mainly circulates in the reservoir host C. gundi and humans are accidentally infected. This is in agreement with the zoonotic character of L. killicki (syn. L. tropica) compared to L. tropica, which is known to be an anthropozoonotic or zoonotic pathogen, according to the infection focus [24]. Comparison of L. tropica from Morocco and L. killicki (syn. L. tropica) from Tunisia revealed that the population structures of these two taxa are different. Indeed, L. killicki (syn. L. tropica) populations from Tunisia showed an important genetic differentiation and differences in terms of genetic diversity, whereas the L. tropica populations from Morocco were genetically more homogeneous and only slightly differentiated. These data suggest that L. killicki (syn. L. tropica) poorly disperses (except for rare migration events from a region to another) compared to L. tropica from Morocco. This finding might reflect different ecological patterns, such as epidemiological cycles, infection of the reservoirs or vector behavior. To conclude, this detailed study on L. killicki (syn. L. tropica) population genetics allowed exploring the evolutionary history of this parasite and highlighting its different genetic patterns compared to L. tropica. Despite the probable recent divergence between these taxa, they seem to evolve differently in terms of epidemiological cycle and thus transmission dynamics. Particularly, this study supports the hypothesis of a zoonotic transmission cycle for L. killicki (syn. L. tropica). Our data also suggest that Gafsa could be the historical focus of L. killicki (syn. L. tropica), although the sample size from the other regions was too small to firmly validate this hypothesis. It is now essential to study the P. sergenti vector populations in Tunisia and their susceptibility to L. killicki (syn. L. tropica) and the parasite biology in C. gundi to better understand the transmission cycle of this parasite. Although for the moment, L. killicki (syn. L. tropica) should be still considered a L. tropica subpopulation, our analyses indicate that in the future, this taxon position may have to be reconsidered.

Leishmania killicki (syn. L. tropica) was discovered in 1986. Few studies have been conducted on this parasite exclusively described in Maghreb. Consequently, many elements on its epidemiology, transmission, population structure and dynamics remain unknown. To better understand the evolution of this parasite, its population structure has been compared with that of L. tropica populations from Morocco using Multilocus Enzyme Electrophoresis (MLEE) and MultiLocus Microsatellite Typing (MLMT) typing. MLMT data support the hypothesis that L. killicki (syn. L. tropica) belongs to the L. tropica complex despite the strong genetic differentiation between them. Despite the probable recent divergence between L. killicki (syn. L. tropica) and L. tropica, they seem to evolve differently. Indeed, L. killicki (syn. L. tropica) appears slightly polymorphic and highly structured in space and time, while L. tropica was genetically heterogeneous, slightly structured geographically and temporally. The different population structures revealed distinct genetic organizations, reflecting different epidemiological cycles. Several parameters could explain these opposite epidemiological and genetic patterns such as ecosystems, vectors and reservoirs.

Abstract Introduction Materials and Methods Results Discussion

2015

Comparison of Leishmania killicki (syn. L. tropica) and Leishmania tropica Population Structure in Maghreb by Microsatellite Typing

Changes in climate and environmental conditions could be the driving factors for the transmission of hantavirus. Thus, a thorough collection and analysis of data related to the epidemic status of hemorrhagic fever with renal syndrome (HFRS) and the association between HFRS incidence and meteorological factors, such as air temperature, is necessary for the disease control and prevention. Journal articles and theses in both English and Chinese from Jan 2014 to Feb 2019 were identified from PubMed, Web of Science, Chinese National Knowledge Infrastructure, Wanfang Data and VIP Info. All identified studies were subject to the six criteria established to ensure the consistency with research objectives, (i) they provided the data of the incidence of HFRS in mainland China; (ii) they provided the type of air temperature indexes; (iii) they indicated the underlying geographical scale information, temporal data aggregation unit, and the data sources; (iv) they provided the statistical analysis method that had been used; (v) from peer-reviewed journals or dissertation; (vi) the time range for the inclusion of data exceeded two consecutive calendar years. A total of 27 publications were included in the systematic review, among them, the correlation between HFRS activity and air temperature was explored in 12 provinces and autonomous regions and also at national level. The study period ranged from 3 years to 54 years with a median of 10 years, 70. 4% of the studies were based on the monthly HFRS incidence data, 21 studies considered the lagged effect of air temperature factors on the HFRS activity and the longest lag period considered in the included studies was 34 weeks. The correlation between HFRS activity and air temperature varied widely, and the effect of temperature on the HFRS epidemic was seasonal. The present systematic review described the heterogeneity of geographical scale, data aggregation unit and study period chosen in the ecological studies that seeking the correlation between air temperature indexes and the incidence of HFRS in mainland China during the period from January 2014 to February 2019. The appropriate adoption of geographical scale, data aggregation unit, the length of lag period and the length of incidence collection period should be considered when exploring the relationship between HFRS incidence and meteorological factors such as air temperature. Further investigation is warranted to detect the thresholds of meteorological factors for the HFRS early warning purposes, to measure the duration of lagged effects and determine the timing of maximum effects for reducing the effects of meteorological factors on HFRS via continuous interventions and to identify the vulnerable populations for target protection. Hemorrhagic fever with renal syndrome (HFRS) is a rodent-associated zoonosis and also a legally mandated notifiable disease in China [1,2]. China has the largest number of HFRS cases in the world [3,4]. After the epidemic of HFRS in mainland China in the 1980s, the HFRS incidence fluctuated periodically with a cycle of about 5 to 10 years, but overall with a descending trend. In the 1990s, the annual number of HFRS cases reported was between 40,000 to 60,000 cases [5]. Since the beginning of the 21st century, the HFRS incidence has continued to decline in mainland China, and after reaching the lowest level in nearly 20 years in 2009, it has gradually increased [6]. In the past three years, the number of HFRS incident cases was maintained at around 11,000 cases per year [7]. Climate and environmental changes could affect the reservoir ecology and dynamics of rodent carriers and hence trigger the spread of hantavirus transmission [5,8, 9]. With the acceleration of China’s urbanization process, especially in the process of rapid transition of China’s agriculture-related landscapes to urban landscapes, the dual role of climate change and environmental change has led to a leap in the epidemic area range of HFRS. Exploring or clarifying the relationship between HFRS epidemic and those environmental factors may help to grasp the spread and epidemic pattern of HFRS and then the pattern could serve as the partial basis of accurate HFRS incidence prediction and corresponding health care resources allocation. Due to the above-mentioned background and consideration, a comprehensive and in-depth collection, collation and analysis of data related to the epidemic status of HFRS is needed and this has become the consensus of researchers in the field of HFRS surveillance and prevention. Since the last century, an extensive literature has accumulated on the association between global climate change and infectious diseases. The involved meteorological factors included seasonal and climate change, greenhouse effect, tropical climate, ultraviolet light and etc. , while in the field of HFRS prevention and control, the most studied meteorological factor was air temperature [10,11]. The choice of appropriate statistical methods according to the type of HFRS epidemic area, the availability, fineness and data distribution characteristics of historical data such as epidemic situation or meteorological data in various regions is being another challenge. The researchers adopted various statistical methods to explore the epidemiological links between HFRS incidence and air temperature, at different spatial and temporal scale and among those studies, the length of lag period also varied. Hence, to characterize the variety of data aggregation scale and statistical methods adopted for the epidemiological links between the HFRS epidemic status and air temperature, we conducted the present systematic literature review for a broader appreciation. In this review, we assessed the current landscape of these studies in terms of user preferences, information needs of HFRS incidence data and air temperature data, and the considerations in the adoption of scale. Due to the challenges of rapidly changing situation of integrating and analyzing data, the review placed the emphasis on publication within the recent five years. A systematic search in PubMed (www. ncbi. nlm. nih. gov/pubmed), Web of Science, Chinese National Knowledge Infrastructure (www. cnki. net), Wanfang Data (www. wanfangdata. com. cn) and VIP Info (www. vipinfo. com. cn) was performed to collect the publications related to the correlations between HFRS incidence and air temperature in mainland China. Data from Jan 1st, 2014 to February 28th, 2019 was retrieved by means of the keyword ‘hemorrhagic fever with renal syndrome’ in combination with the keywords ‘HFRS’, ‘epidemic hemorrhagic fever’, temperature’, ‘climate’, ‘meteorological factor’ and ‘China’ in both English and Chinese. References cited in the retrieved articles were also evaluated to maximize article recovery. The last search was conducted on March 20th, 2019. Eligible studies had to meet all the following six criteria: (i) they provided the data of the incidence of HFRS in mainland China; (ii) they provided the type of temperature indexes; (iii) they indicated the underlying geographical scale information, temporal data aggregation unit, and the data sources; (iv) they provided the statistical analysis method that had been used; (v) from peer-reviewed journals or dissertation; (vi) the time range for the inclusion of data exceeded two consecutive calendar years. Therefore, epidemiological studies providing HFRS incidence data only, or studies outside the boundaries of mainland China were excluded, review articles were also excluded. And when the studies were duplicately reported, the publication with larger sample size or more detailed data were included. As for each included study, the following information was extracted according to a self-designed data extraction form: first author and publication year, location and geographical scale (national, provincial, municipal and county/local-level), type of HFRS incidence data (incidence or number of HFRS incident cases), study period chosen, the air temperature indexes involved (average, maximum, minimum), temporal data aggregation unit (annual, seasonal, monthly, weekly and daily), the statistical method adopted, lag time considered, the correlation between HFRS incidence and air temperature that each included study concluded. To ensure the reliability, two investigators (XHB and CP) independently screened each publication and the literature screening process was checked by a third reviewer (TJ). Two investigators (CP and TJ) independently summarized the data before discussing the results together, and all the discrepancies resolved by the principal investigator (PG). According to the PRISMA Statement (S1 Table) [12], STROBE checklist [13], the bibliometric review by Dufault et al. [14] and the systematic review by Betran et al. [15], a self-designed quality assessment item list (S2 Table) was adopted to evaluate the quality of the included ecological studies. The risk of bias in the included ecological studies was evaluated with a total of ten risk-biased items regarding external validity (items 1 to 4 assessed the domain of selection) and internal validity (items 5 to 9 assessed the domain of measurement bias and bias of extrapolation or interpretation, and item 10 assessed the bias related to the funding). Two investigators (CP and ZMH) negotiated and completed the quality assessment. For the disagreements, the decision was made by the principal investigator (PG). After the data was extracted and double-checked, articles containing the study period, geographical location, type of HFRS incidence data, air temperature indexes involved, temporal data aggregation unit, description of statistical method adopted, correlation concluded between HFRS incidence and air temperature were considered to be qualified, and all the articles included in this systematic review met these quality requirements. At the end of the overall risk assessment of study bias, according to the previously proposed criteria [16], studies with a “No” score ≤3 were classified as low risk, studies with a “No” score 4–6 were classified as moderate risk and studies with a “No” score ≥7 were classified as high risk. Studies with overall low and moderate risk of study bias were included in this present systematic review. A total of 111 articles related to the topic that published between Jan 1st, 2014 and Feb 28th, 2019 were identified, including 62 publications in Chinese and 49 publications in English. Three duplicated articles were subsequently removed, after intensive reading the titles, abstracts and full-texts of these article, 81 publications were also excluded. Thus, 27 publications (18 in Chinese and nine in English) were finally included in the systematic review, and among them 21 studies were with low risk of study bias and six studies were with moderate risk of study bias. The literature selection process is shown in Fig 1 and the PRISMA checklist is provided in S1 Table, the risk bias and assessment results are provided in S2 Table. Among the 27 publications included, there were 22 journal articles and five dissertations; Those 22 journal articles scattered over 17 kinds of journals, with the journal ‘PLoS Neglected Tropical Diseases’ was identified as the most active journal about the topic during the study period. These 17 journals could be grouped into two categories, public health (thirteen kinds of journals), and natural sciences, including environmental sciences (four kinds of journals). Among the included 27 studies, one study analyzed the HFRS data at the national level which included the data from 31 provinces, autonomous regions and municipalities in mainland China, and the remaining 26 studies involved 12 provinces and autonomous regions. Studies from Shandong Province accounted for 37. 0% of all the included studies. Five studies collected HFRS data at provincial level, 17 studies collected HFRS data at municipal level, and four studies collected HFRS data at county (local) level. The study period of the data nested in the included studies ranged from 3 years to 54 years with a median of 10 years, as shown in Table 1. Regarding the temporal unit of data aggregation, in the included studies, fifteen studies were based on monthly HFRS incidence or the number of monthly incident HFRS cases, five studies were based on the number of daily reported HFRS cases, two studies based on annual HFRS incidence, four studies based on both monthly and annual HFRS incidence, and one study was based on the number of weekly reported HFRS cases. As for the corresponding air temperature indicators, seven studies adopted all the three indicators of average air temperature, average maximum air temperature and average minimum air temperature, and the rest of the studies were based on either average air temperature only or average maximum air temperature only. Twenty-one studies considered the lagged effect of air temperature factors on the HFRS activity and the longest lag period considered in the included studies was 34 weeks. With regards to the statistical methods adopted by the researchers to explore the correlation between air temperature and HFRS incidence, Spearman correlation analysis, Pearson correlation analysis, generalized additive model, seasonal differential autoregressive moving average model, negative binomial multivariate regression analysis, distributed lag nonlinear model conditions, conditional logistic regression analysis and wavelet analysis have been indicated in the included studies, as shown in Table 1. The associations observed at one scale were not present at another one. Fifteen studies indicated the negative correlation between air temperature and HFRS activity, while seven studies found the positive correlation between air temperature and HFRS activity. There were also two studies that defined a certain temperature as the dividing point between HFRS activity and air temperature, the correlation curve was inverted ‘U-shaped’ [23,29]. Also, three studies did not find statistically significant associations between HFRS activity and air temperature. Because the current China’s National Notifiable Infectious Disease Reporting system (NIDR) is still unable to distinguish the type of hantaviruses infected by HFRS patients, the direction and magnitude of the effects of temperature on HFRS activity in different seasons were also inconsistent. The study based on the monthly temperature index at county scale in Shandong Province found that the increase of average air temperature in spring was the risk factor of SEOV-type HFRS outbreak, but similar results could not be found in other seasons and in the HTNV-type HFRS [27]. The study from Changsha City indicated that HFRS incidence was highly correlated with the previous air temperature [33]. The study from Changchun City, Jilin Province [18] and Changsha City [33] found that HFRS was prone to outbreaks and epidemics during relatively dry, high temperature and low wind conditions from late July to late September. The study in Guangdong Province [41] indicated that rainy or cloudy conditions in the previous season might contribute to the growth of the number of HFRS cases. The study in Guangzhou City found that the incidence of HFRS was negatively correlated with the daily average air temperature, and when the average daily temperature was 15. 2 °C, the relative risk reached highest with the lag period of ten days [19]. In addition to collecting data on the incidence of HFRS in the whole population, a certain study also collected data on HFRS incidence (age, gender, occupation, etc.) from different populations based on the national legal infectious disease surveillance system. The study in Huludao City, Liaoning province found that HFRS activity in the population aged 35–59 years was significantly affected by air temperature, but this phenomenon could not be found in the population of other age groups [25]; the authors of this Huludao study attributed the phenomenon to the increased virus exposure in agricultural labor work in the middle-aged population. This study in Huludao also indicated that the correlations between HFRS activity and air temperature were statistically significant in both males and females, while with different length of period, two months in the male and no lag in the female population [25]. It is clear from the major part of the included studies that air temperatures are indirectly associated with HFRS activity, however, the temperature-HFRS association findings were inconsistent and location-dependent. Our systematic review indicated that the ecological effects of air temperature on HFRS incidence could be affected by the spatial or temporal scale of the data and also the study period involved, which might help to partly understand the contradicting observations in the included studies. The researchers need to consider or identify which temperature indicator and data aggregation unit are more appropriate to explain the correlations between HFRS incidence and air temperature at different geographical scales [43]. Seeing that the effect of air temperature on the HFRS activity varied among different populations [25], it is suggested that the demographic characteristics of the local target population should be considered when the correlation between meteorological factors, such as air temperature, and HFRS incidence is estimated or predicted to assist precise HFRS prevention and control. The ecological correlation analysis is not only data-driven, but also technology-concentrated, the integration of the collection of high-quality HFRS incidence data and the multi-discipline development can open vast vistas for the correlation analysis techniques’ application in the field of infectious diseases epidemiology [44]. In the real world, only when the statistical methods that can be understood and adopted by the HFRS incidence data collectors and the users of infectious disease surveillance system, the relationship between HFRS incidence and meteorological factors could be better clarified. And then, these correlations could be possibly incorporated into the early warning, prevention and control of HFRS. According to the observed correlations between HFRS activity and air temperature, with special considerations of geographical scale, data aggregation unit, length of lag period and study period, appropriate statistical methods nested in the National Notifiable Infectious Disease Reporting system will be of great importance for the users. The present systematic review might be improved if the following information could be considered in the future HFRS-related ecological studies. All the included studies in the present systematic review are descriptive, further exploratory analysis, explanatory analysis and statistical inference are needed. The distribution of Hantaan and Seoul type of HFRS was not available in the China’s surveillance system of the legally mandated notifiable infectious diseases. The data of HFRS vaccine coverage could not be obtained in the included studies, and HFRS vaccine coverage did affect the magnitude of HFRS incidence. Only one single included study investigated the correlation between air temperatures and HFRS incidence stratified by age group, thus the characteristic of the HFRS vulnerable population that was most affected by air temperature could not be obtained. Besides the different data scale and the differences in the hantaviruses type of HFRS, the factors relating to the dynamics of the rodent hosts and human activities, such as urbanization indicators, should be considered when understanding the results from these ecological studies, given the fact that China is a topographically heterogeneous country. It should also be emphasized that air temperature as an isolated indicator that cannot explain the HFRS incidence fully, confounding factors should always be considered. Caution should be used when studying the associations between HFRS activity and the isolated or the combination of meteorological variables, because of the possible multicollinearity. Therefore, meteorological factors and the impact of climate changes on the pathogenesis of HFRS still need to be further deepened, especially in the process of rapid transition of China’s agriculture-related landscapes to urban landscapes [45]. In summary, the present systematic review first described the heterogeneity of geographical scale, data aggregation unit and study period chosen in the ecological studies that seeking the correlation between air temperature indexes and the incidence of HFRS in mainland China during the period from January 2014 to February 2019. The appropriate adoption of geographical scale, data aggregation unit, the length of lag period and the length of incidence collection period should be considered when exploring the relationship between HFRS incidence and meteorological factors such as air temperature. Further investigation is warranted to detect the thresholds of meteorological factors for the HFRS early warning purposes, to measure the duration of lagged effects and determine the timing of maximum effects for reducing the effects of meteorological factors on HFRS via continuous interventions and to identify the vulnerable populations for target protection.

China has the largest number of hemorrhagic fever with renal syndrome (HFRS) cases in the world. With the acceleration of China’s urbanization process, especially in the process of rapid transition of China’s agriculture-related landscapes to urban landscapes, the dual role of climate change and environmental change has led to a leap in the epidemic area range of HFRS. Exploring or clarifying the relationship between HFRS epidemic and those environmental factors may help to grasp the spread and epidemic pattern of HFRS and then the pattern could serve as the partial basis of accurate HFRS incidence prediction and the corresponding allocation of public health resources. The present systematic review first described the heterogeneity of geographical scale, data aggregation unit and study period chosen in the ecological studies that seeking the correlation between air temperature indexes and incidence of HFRS in mainland China during the period from January 2014 to February 2019. Raising the awareness of the appropriate adoption of geographical scale, data aggregation unit, the length of lag period and the length of incidence collection period is of great importance when exploring the relationship between HFRS incidence and meteorological factors such as air temperature.

Abstract Introduction Materials and methods Results Discussion

medicine and health sciences spring statistics china atmospheric science geographical locations seasons hemorrhagic fever with renal syndrome mathematics research and analysis methods climate change infectious diseases geography mathematical and statistical techniques research assessment people and places climatology asia urban areas earth sciences geographic areas systematic reviews viral diseases physical sciences statistical methods statistical data

2019

Distribution of geographical scale, data aggregation unit and period in the correlation analysis between temperature and incidence of HFRS in mainland China: A systematic review of 27 ecological studies

Nod-like receptors (NLRs) comprise a large family of intracellular pattern- recognition receptors. Members of the NLR family assemble into large multiprotein complexes, termed the inflammasomes. The NLR family, pyrin domain-containing 3 (NLRP3) is triggered by a diverse set of molecules and signals, and forms the NLRP3 inflammasome. Recent studies have indicated that both DNA and RNA viruses stimulate the NLRP3 inflammasome, leading to the secretion of interleukin 1 beta (IL-1β) and IL-18 following the activation of caspase-1. We previously demonstrated that the proton-selective ion channel M2 protein of influenza virus activates the NLRP3 inflammasome. However, the precise mechanism by which NLRP3 recognizes viral infections remains to be defined. Here, we demonstrate that encephalomyocarditis virus (EMCV), a positive strand RNA virus of the family Picornaviridae, activates the NLRP3 inflammasome in mouse dendritic cells and macrophages. Although transfection with RNA from EMCV virions or EMCV-infected cells induced robust expression of type I interferons in macrophages, it failed to stimulate secretion of IL-1β. Instead, the EMCV viroporin 2B was sufficient to cause inflammasome activation in lipopolysaccharide-primed macrophages. While cells untransfected or transfected with the gene encoding the EMCV non-structural protein 2A or 2C expressed NLRP3 uniformly throughout the cytoplasm, NLRP3 was redistributed to the perinuclear space in cells transfected with the gene encoding the EMCV 2B or influenza virus M2 protein. 2B proteins of other picornaviruses, poliovirus and enterovirus 71, also caused the NLRP3 redistribution. Elevation of the intracellular Ca2+ level, but not mitochondrial reactive oxygen species and lysosomal cathepsin B, was important in EMCV-induced NLRP3 inflammasome activation. Chelation of extracellular Ca2+ did not reduce virus-induced IL-1β secretion. These results indicate that EMCV activates the NLRP3 inflammasome by stimulating Ca2+ flux from intracellular storages to the cytosol, and highlight the importance of viroporins, transmembrane pore-forming viral proteins, in virus-induced NLRP3 inflammasome activation. The innate immune system, the first line of defense against pathogens, utilizes pattern-recognition receptors (PRRs) to detect pathogen-associated molecular patterns (PAMPs). RNA viruses are detected by host PRRs including Toll-like receptors (TLRs), retinoic acid-inducible gene-I (RIG-I) -like helicases (RLHs), and Nod-like receptor (NLR) family, pyrin domain-containing 3 (NLRP3) [1], [2], [3], [4]. NLRP3 plays an important role in the secretion of proinflammatory cytokines interleukin 1 beta (IL-1β) and IL-18 after viral infections. Upon activation, NLRP3 forms the protein complex termed “NLRP3 inflammasome” by recruiting the apoptosis-associated speck-like protein containing a caspase recruitment domain (ASC) and pro-caspase-1, which is activated by autocatalytic cleavage within the complex [5]. The active caspase-1 catalyzes proteolytic processing of pro-IL-1β and pro-IL-18 into active cytokines that are then released across the plasma membrane by poorly understood mechanisms [6]. Secretion of these two cytokines requires upregulations of pro-IL-1β, pro-IL-18, and NLRP3, which are induced by signals from TLRs, IL-1 receptor or tumor necrosis factor receptor (signal 1), in addition to the activation of caspase-1 through inflammasome activation (signal 2) [7], [8]. In influenza virus infection, the signal 1 is provided by TLR7 that recognizes influenza virus RNA, whereas the signal 2 comes from the function of the virus-encoded proton-selective ion channel M2 protein, but not from viral RNA [9]. Encephalomyocarditis virus (EMCV), a member of the genus Cardiovirus within the family Picornaviridae, is a nonenveloped, positive single-stranded RNA virus. This virus has a ∼7. 8 kb viral genome covalently linked to a viral protein VPg at its 5′ end that serves as a primer for viral RNA synthesis. The viral genome encodes a polyprotein precursor, which is divided into the P1, P2 and P3 regions and processed mainly by the virus-encoded 3C protease. Processing of the P1 region produces the structural capsid proteins 1A (VP4), 1B (VP2), 1C (VP3), and 1D (VP1), whereas the P2 and P3 regions are processed into the nonstructural proteins 2A, 2B, 2C, 3A, 3B (VPg), 3C and 3D as well as cleavage intermediates (2BC, 3AB, and 3CD) [10]. The EMCV 2A protein has been implicated in the shutoff of host protein synthesis and viral pathogenesis [11], [12], but little is known about the roles of the EMCV 2B and 2C proteins. Picornavirus 2B proteins have been reported to act as viroporins, transmembrane pore-forming viral proteins that alter the membrane permeability to ions by forming membrane channels, and participate in a range of viral functions [13]. The avian encephalomyelitis virus (AEV) 2C protein is known to induce apoptosis [10], [14]. EMCV is detected by at least three classes of PRRs in the host innate immune system. First, double-stranded RNA (dsRNA) produced in EMCV-infected cells is recognized by TLR3 [15]. Second, EMCV RNA is recognized by melanoma differentiation associated gene 5 (MDA5), a member of RLHs, unlike many other RNA viruses which are recognized by RIG-I [16], [17], [18]. Activation of RLHs leads to the induction of type I interferons in infected cells. Third, the NLRP3 inflammasome detects EMCV by unknown mechanisms [19], [20]. In this study, we examined how EMCV activates the NLRP3 inflammasome. Our results indicate that EMCV increases the local Ca2+ concentration in the cytoplasm by stimulating Ca2+ flux from intracellular storages through the action of its viroporin 2B, thereby activating the NLRP3 inflammasome. Together with the data with the influenza virus M2 protein, our study reveals the importance of viroporins in virus-induced NLRP3 inflammasome activation. In order to determine whether EMCV infection activates the inflammasomes, we measured IL-1β secretion from unprimed or lipopolysaccharide (LPS) -primed mouse bone marrow-derived macrophages (BMMs) infected with EMCV or influenza virus PR8 (Figure 1A). LPS induces pro-IL-1β in the cytosol (signal 1) [7]. In agreement with a previous report [20], EMCV infection induced release of IL-1β from LPS-primed BMMs, but not from unprimed BMMs. The amount of IL-1β secretion from BMMs was increased in a multiplicity of infection (MOI) -dependent manner (Figure 1B). Western blot analysis demonstrated that the p17 subunit, the mature processed form of IL-1β, was secreted in the supernatant (Figure 1C). Furthermore, the secretion of IL-1β after EMCV infection was inhibited by yVAD-CHO, a specific peptide inhibitor of caspase-1, without affecting the secretion of IL-6 in both bone marrow-derived dendritic cells (BMDCs) and BMMs (Figure 1D and Figure S1). To dissect the importance of various components of the inflammasome complex in IL-1β secretion in response to EMCV infection, we generated RAW264. 7 cells stably expressing short hairpin RNA (shRNA) against murine NLRP3, ASC or caspase-1. Quantitative RT-PCR and Western blot analysis confirmed knockdown of NLRP3, ASC, and caspase-1 at the mRNA and protein levels (Figure S2A–B). Furthermore, these cells produced comparable levels of IL-6 after LPS stimulation to EGFP-knockdown cells (Figure S2C), indicating that innate immunological responses are not generally affected by the knockdown. Like that after influenza virus infection [9], IL-1β secretion after EMCV infection was found to be dependent on NLRP3, ASC and caspase-1 (Figure 1E), indicating that EMCV infection activates the NLRP3 inflammasome. This is consistent with previous studies demonstrating that EMCV-induced IL-1β production was abrogated in NLRP3−/−, ASC−/−, and caspase-1−/− BMDCs [19], [20]. We next examined whether viral replication is required for NLRP3 inflammasome activation by EMCV. Unlike untreated virions, UV-irradiated EMCV failed to induce IL-1β secretion from LPS-primed BMDCs (Figure 2A), indicating that the viral particles or genomic RNAs per se are insufficient to activate the NLRP3 inflammasome. To test whether viral RNA translation is needed to elicit EMCV-induced NLRP3 inflammasome activation, LPS- primed BMDCs were treated with a translation inhibitor cycloheximide (CHX) prior to stimulation with ATP or infection with EMCV. The CHX-pretreated cells produced IL-1β normally in response to extracellular ATP (Figure 2B, left), indicating that NLRP3 inflammasome activation by ATP stimulation does not require de novo translation. This is consistent with a previous report showing that CHX dose not inhibit nigericin-induced NLRP3 activation in primed macrophages [21]. In contrast, pretreatment of cells with CHX significantly inhibited EMCV-induced IL-1β secretion (Figure 2B, right). These data indicate that virus-encoded proteins or viral RNA transcripts, not viral genomic RNAs, activate the NLRP3 inflammasome. Next, we tested whether viral RNAs were able to activate the NLRP3 inflammasome after infection. To this end, we examined the abilities of viral genomic RNAs and transcripts from EMCV-infected cells to trigger NLRP3 inflammasome activation, by measuring IL-1β secretion from BMMs transfected with these RNAs. Although transfection with RNAs from the virions or virus-infected cells induced robust expression of interferon-β (Figure 2C), it failed to stimulate secretion of IL-1β from BMMs (Figure 2D). It is possible that the amount (50 µg) of transfected viral RNA was not enough to trigger NLRP3 inflammasome activation. However, we ruled out this possibility by measuring the amounts of viral genomic RNA obtained from virions. We collected about 12 µg of viral genomic RNA from 4×108 plaque forming unit (PFU) of virions (Figure 2E). In contrast, we used 1. 2×106 PFU of EMCV for the infection at an MOI of 1. 5, in which IL-1β secretion was clearly detected (Fig. 1B). Furthermore, since EMCV is a positive single-stranded RNA virus, transfection of cells with viral genomic RNA would produce virions in the supernatant which could stimulate the NLRP3 inflammasome. Indeed, we detected about 600 PFU of EMCV in the supernatant of the cells transfected with 50 µg of viral genomic RNA at 24 h post transfection (Figure 2F). Infection of 8×105 cells with 600 PFU of EMCV corresponds to the MOI below 0. 001, which is insufficient to induce IL-1β secretion (Figure 1B). Thus, these results indicate that viral RNA genome and transcripts are insufficient to trigger robust activation of the NLRP3 inflammasome and that the signal 2 is probably derived from virus-encoded proteins. We previously demonstrated that a mutant M2 protein of influenza virus, which has lost its proton selectivity and enabled the transport of other cations (Na+ and K+), induced enhanced IL-1β secretion compared with the wild-type M2 protein [9]. In addition, picornavirus 2B proteins have been reported to act as viroporins [13]. Thus, we hypothesized that EMCV viroporin 2B protein may trigger inflammasome activation by altering intracellular ionic concentrations. To test this idea, we stimulated BMMs with LPS (signal 1) and transduced them with the lentivirus expressing the EMCV 2A, 2B, or 2C protein. IL-1β was released from LPS-primed BMMs transduced with the 2B-expressing lentivirus, but not from those transduced with other lentiviruses (Figure 3A). These data indicated that the expression of EMCV viroporin 2B is sufficient to activate the NLRP3 inflammasome. Next, we determined the subcellular localization of EMCV 2B protein responsible for EMCV-induced inflammasome activation by using a confocal microscopy. To this end, we first generated plasmids expressing influenza virus hemagglutinin (HA) - or Flag-tagged proteins and confirmed their expression in HEK293T cells by immunoblot analysis (Figure 3B). When each protein was expressed in HeLa cells, the EMCV 2B and 2C proteins were localized to the Golgi apparatus and other cytoplasmic structures, and the EMCV 2A protein to the nucleus (Figure 3C and Figure S3). We also examined the intracellular localization of NLRP3. In agreement with a previous report [8], stimulation of cells with LPS induced NLRP3 expression in the cytosol (Figure 3D). Upon infection with EMCV, NLRP3 was redistributed to the perinuclear region or cytoplasmic granular structures, which is also observed in cells stimulated with other NLRP3 ligands such as monosodium urate (MSU), alum or nigericin (Figure 3D) [22] and considered as a hallmark of NLRP3 activation. Although resting cells or cells expressing the EMCV 2A or 2C protein uniformly expressed NLRP3 throughout the cytoplasm, it was redistributed to the perinuclear region in EMCV 2B- or influenza virus M2-expressing cells (Figure 3E). In addition, the nonstructural 2B proteins from poliovirus and enterovirus 71, members of the genus Enterovirus within the family Picornaviridae, also induced the redistribution of NLRP3 to the perinuclear region (Figure S4). Notably, the redistributed NLRP3 was largely co-localized with the EMCV 2B, influenza virus M2, poliovirus 2B, and enterovirus 71 2B proteins (Figure 3E and Figure S4). We also demonstrated that NLRP3 was redistributed to the perinuclear region and co-localized with EMCV 2B protein in BMMs transduced with the lentivirus expressing the Flag-tagged EMCV 2B but not 2A or 2C protein (Figure S5). Together, these data provide evidence that the EMCV viroporin 2B alone is sufficient to trigger NLRP3 inflammasome activation and IL-1β secretion from LPS-primed BMMs. The redistribution of NLRP3 after EMCV infection was not inhibited by yVAD-CHO (Figure 3D). This result is consistent with the activity of yVAD-CHO operating at the step after NLRP3 activation. It has been demonstrated that picornavirus 2B proteins, including EMCV 2B, are mainly localized to the endoplasmic reticulum (ER) and Golgi apparatus, and reduce Ca2+ levels in those organelles ([Ca2+]ER and [Ca2+]Golgi) [23], thereby presumably elevating the local concentration of cytoplasmic Ca2+ ([Ca2+]cyt). To examine whether the elevation of [Ca2+]cyt by the EMCV 2B protein plays a role in NLRP3 inflammasome activation, we first measured the kinetic changes in the [Ca2+]cyt after infection with EMCV by using a calcium-dependent fluorescent probe. Infection with EMCV resulted in a significant increase in the fluorescence intensity of the cells compared with that of mock-infected cells (Figure 4A). The elevation of fluorescence intensity became apparent around 8 hours post infection (Figure 4A), which corresponded to the initiation of IL-1β secretion in the culture supernatant (Figure 4B). To better understand the role of the elevation of [Ca2+]cyt in NLRP3 inflammasome activation, we undertook three different approaches. First, we tested the effects of [Ca2+]cyt-increasing drugs on IL-1β secretion (Figure 4C and D). Thapsigargin (TG) is a non-competitive inhibitor of the sarco-ER Ca2+ ATPase (SERCA) that transports cytoplasmic Ca2+ into the ER [24]. Treatment with TG induced IL-1β secretion from LPS-primed BMMs (Figure 4C). Similarly, the Ca2+ ionophore ionomycin induced significant release of IL-1β (Figure 4C). When cells were treated with TG or ionomycin, NLRP3 was redistributed to the perinuclear region or cytoplasmic granular structures (Figure 4D), as in EMCV-infected cells (Figure 3D) or EMCV 2B- or influenza virus M2 protein-expressing cells (Figure 3E). Thus, these data indicate that an increase in [Ca2+]cyt is important in NLRP3 inflammasome-mediated IL-1β release. Second, we investigated main sources for Ca2+ flux required for EMCV-induced inflammasome activation. Addition of EGTA in the extracellular medium, which blocked ionomycin-induced Ca2+ influx (Figure S6) and Listeria monocytogenes-induced secretion of IL-1α [25] (Figure 4E, right), had no effect on EMCV- or EMCV 2B-induced inflammasome activation (Figure 4E and Figure S7A). Next, we tested whether the cell-permeable Ca2+-chelator BAPTA-AM inhibits EMCV-induced inflammasome activation. The release of lactate dehydrogenase (LDH) from BAPTA-AM treated cells was measured as an index of cytotoxicity. Treatment of BMMs with 5–10 µM of BAPTA-AM (LDH release <3%) significantly blocked IL-1β secretion by EMCV or EMCV 2B protein (Figure 4F and Figure S7B) and the redistribution of the NLRP3 (Figure S8). Furthermore, EMCV 2B protein, but not 2A or 2C protein, specifically reduced [Ca2+]ER/Golgi (Figure 4G), as demonstrated using an EGFP chimera that modulates its fluorescence in response to changes in the Ca2+ concentration [26]. We confirmed the fluorescence changes of the EGFP chimera after treatment of transfectants with TG (Figure S9) and the comparable expression levels of the EGFP chimera in different transfectants (Figure 4H). These results indicate that Ca2+ flux from intracellular storages, but not from extracellular medium, is important in EMCV-induced inflammasome activation, as expected from the function of the 2B protein. Third, we examined whether Ca2+ flux-induced activation of inflammasome requires the action of calpain, a Ca2+-dependent cysteine protease. To this end, we examined the effects of carbobenzoxy-valyl-phenylalanial (MDL-28170, calpain inhibitor III) on EMCV-induced inflammasome activation. While 12–100 µM of MDL-28170 significantly blocked L. monocytogenes-induced processing of IL-1α [25] (Figure S10B), even 100 µM of the drug had no effect on EMCV- or EMCV 2B-induced inflammasome activation (Figure S7A and Figure S10A). These data indicate that EMCV infection increases [Ca2+]cyt via the flux from intracellular storages and thereby activates NLRP3 inflammasome in an calpain-independent manner. We finally tested the role of previously identified factors that can activate the NLRP3 inflammasome [22], [27], [28], [29]. Since mitochondrial ROS was found to be important in NLRP3 inflammasome activation by MSU, alum, and ATP [22], [27], we first measured the kinetic changes in ROS-producing mitochondria after infection with EMCV. The level of ROS-producing mitochondria in LPS-primed BMMs infected with EMCV peaked at 3 h after infection and gradually decreased to the basal level by 24 h (Figure 5A). The kinetics of the change was not correlated with the levels of mature IL-1β in culture supernatants (Figure 5A). Furthermore, treatment with antioxidant Mito-TEMPO, a scavenger specific for mitochondrial ROS [30], [31], had no effect on the secretion of IL-1β in BMMs infected with EMCV (Figure 5B) or transduced with EMCV 2B (Figure S7A), but significantly inhibited IL-1β secretion in response to ATP or MSU (Figure 5B), as reported previously [27]. Cathepsin B, a specific lysosomal cysteine protease, has been implicated in NLRP3-mediated IL-1β production in response to nonviral signals [28], [29] and influenza virus infection [32]. We therefore examined the effect of the cathepsin B-specific inhibitor, Ca-074-Me, on EMCV-induced inflammasome activation. In agreement with a previous report [29], alum-induced release of IL-1β was significantly inhibited by the Ca-074-Me (Figure 5C). In contrast, IL-1β production in response to EMCV or EMCV 2B was not affected in Ca-074-Me-treated BMMs (Figure 5C and Figure S7A). Thus, these data indicate that EMCV activates the NLRP3 inflammasome independent of mitochondrial ROS and cathepsin B. In this study, we demonstrated that EMCV, a positive strand RNA virus, activates the NLRP3 inflammasome by increasing [Ca2+]cyt through the action of virus-encoded viroporin 2B (Figure 6). Inflammasome-mediated cytokine release via NLRP3 requires two signals: signal 1 drives pro-IL-1β synthesis while signal 2 induces activation of caspase-1 and cleavage of pro-IL-1β. In the present study, we found that EMCV-induced IL-1β secretion requires LPS priming as signal 1, in agreement with a previous report [20]. This may be partly explained by the fact that EMCV has a strategy to shut off host mRNA translation [33]. LPS signaling may be able to overcome this viral inhibitory activity. In contrast, Poeck et al. have demonstrated that EMCV induces pro-IL-1β synthesis via MDA5 [19]. The reasons for these discrepant results are unclear, but could be different EMCV strains used. Currently, two distinct inflammasome complexes have been shown to be involved in RNA virus-induced caspase-1 activation (signal 2): the NLRP3 inflammasome and the RIG-I inflammasome. Poeck et al. reported that vesicular stomatitis virus (VSV), a negative strand RNA virus of the family Rhabdoviridae, activates RIG-I, which in turn recruits the adaptor ASC and pro-caspase-1 to form the RIG-I inflammasome in an NLRP3-independent manner [19]. In contrast, Rajan et al. showed that like influenza virus [9], VSV activates the NLRP3 inflammasome independently of RIG-I and MDA5 [20]. Measles virus, a negative strand RNA virus of the family Paramyxoviridae, which induces type I interferons through its recognition by RIG-I and MDA5 [34], [35], activates the NLRP3 inflammasome but not the RIG-I inflammasome [36]. Our present study also show that IL-1β secretion from EMCV-infected cells is mediated by the NLRP3 inflammasome, consistent with previous reports [19], [20]. Thus, the NLRP3 inflammasome appears to play a central role in caspse-1-dependent IL-1β secretion after RNA virus infections. Although TLRs and RLHs are known to recognize viral RNAs, it remains unclear whether viral RNAs are also required to activate the NLRP3 inflammasome. In the case of influenza virus, transcriptional induction of genes encoding pro-IL-1β, pro-IL-18, and NLRP3 is activated through TLR7 that recognizes viral RNA, whereas the NLRP3 inflammasome activation is caused by the virally-encoded proton-selective M2 ion channel [9]. Consistent with this observation, Murube et al. showed that treatment of cells with poly (I∶C) failed to elicit inflammasome activation [37]. Similarly, our present study demonstrated that transfection with viral RNAs from EMCV virions or EMCV-infected cells did not stimulate secretion of IL-1β from BMMs, although it induced robust expression of type I interferons. The failure of the viral RNA from EMCV to activate inflammasome activation could be explained by the fact that the adaptor protein ASC specifically interacts with RIG-I but not MDA5 [19]. Furthermore, inactivation of EMCV by ultraviolet irradiation completely abrogated IL-1β secretion, suggesting that viral entry and replication is needed for NLRP3 inflammasome activation. The same findings have been reported for influenza virus [38] and measles virus [36]. Together, these observations suggest that viral genomic RNA, viral transcripts and poly (I∶C) do not act as direct ligands for NLRP3. Then, how does NLRP3 detect RNA virus infections? The NLRP3 inflammasome can be activated by a wide range of stimuli, such as endogenous danger signals from damaged cells, bacterial components, and environmental irritants, besides viruses [7]. Three models for activation of the NLRP3 inflammasome have been proposed thus far [5], [39], [40]. One model proposes that the binding of extracellular ATP to the purinergic receptor P2X7 on the cell surface plays an important role. Activation of this ATP-gated ion channel triggers K+ efflux and the recruitment of pannexin 1 to form a large non-selective pore, which may enable the entry of NLRP3 agonists into the cytosol. In the second model, upon phagocytosis of large crystals and environmental irritants (such as asbestos), lysosomal rupture and cytoplasmic release of lysosomal contents such as cathepsin B may occur in cells, which activates the NLRP3 inflammasome. In the third model, production of ROS from damaged mitochondria may stimulate NLRP3 inflammasome activation [22], [27]. Our data clearly showed that EMCV-induced inflammasome activation is mitochondrial ROS- and cathepsin B-independent. The peak of accumulation of ROS-producing mitochondria in EMCV-infected cells was not correlated with the appearance of mature IL-1β in the supernatant. In addition, EMCV-induced IL-1β was not inhibited by inhibitors of mitochondrial ROS and cathepsin B, which effectively blocked ATP- and Alum-induced IL-1β secretion, respectively. Instead, the elevation of [Ca2+]cyt after EMCV infection seemed to be important in NLRP3 inflammasome activation. Picornavirus viroporin 2B proteins have been reported to reduce Ca2+ levels in the ER and Golgi apparatus [10], presumably causing Ca2+ flux into the cytosol and increasing the local [Ca2+]cyt. This suggests that the ionic imbalance in the cytoplasm through the action of EMCV 2B may activate the NLRP3 inflammasome. Indeed, we found that EMCV 2B protein specifically reduced [Ca2+]ER/Golgi, stimulated IL-1β secretion, and induced the redistribution of NLRP3 to the perinuclear region. Notably, the nonstructural 2B proteins from poliovirus and enterovirus 71 also induced the redistribution of NLRP3. Since enterovirus 2B proteins, but not EMCV 2B protein, were found to inhibit protein trafficking, resulting in accumulation of proteins in the Golgi complex [23], the mechanism by which EMCV 2B induces NLRP3 activation could be different from that of poliovirus 2B- and enterovirus 71 2B-induced activation. The notion of the viral nonstructural protein-induced inflammasome activation is indeed supported by our earlier observation that the proton-selective influenza virus M2 protein and its mutant capable of transporting Na+ and K+ [41], both caused NLRP3 inflammasome activation [9]. Since influenza virus increases [Ca2+]cyt [42], [43], [44], it is possible that Ca2+ flux into the cytosol also provokes NLRP3 inflammasome activation in influenza virus-infected cells. While further studies are required to better understand how RNA viruses exactly activate the NLRP3 inflammasome, our present study indicates that viroporins and disturbances in the intracellular ionic milieu following viral infections are important in RNA virus-induced NLRP3 inflammasome activation. Many viruses are known to encode viroporins [13]. For instance, human immune deficiency virus type-1 has an accessory protein Vpu whose transmembrane domain acts as the potassium channel [45], to counteract the host antiviral protein tetherin [46]. The hepatitis C virus p7 protein, a 63-amino acid polypeptide important in assembly and release of infectious virions [47], was found to form hexamers with ion channel activity [48], [49]. Thus, we suspect that in addition to more drastic disruption of membranes by pore-forming toxins [50], [51] and membrane rupture [28], [29], [52], [53], the ionic imbalance via the action of virus-encoded ion channels may be a main target for the NLRP3 inflammasome as a sensor for cellular stress. Knowledge of the exact mechanisms by which NLRP3 detects viruses and thereby affects pathogenesis will provide us with a better understanding of viral diseases to design effective interventions and treatments. Animal experiments were carried out in strict accordance with the recommendations in Guidelines for Proper Conduct of Animal Experiments of Science Council of Japan. The protocol was approved by the Committee on the Ethics of Animal Experiments of the Kyushu University, Japan (Permit Number: A23-087-0). All surgery was performed under sevoflurane anesthesia, and all efforts were made to minimize suffering. C57BL/6 mice used were 8 to 10 weeks of age. BMMs were prepared as described previously [38]. Briefly, bone marrows from the tibia and femur were obtained by flushing with Dulbecco' s modified Eagle' s medium (DMEM; Wako Pure Chemical Industry). Bone marrow cells were cultured with DMEM supplemented with 10% heat-inactivated fetal bovine serum (FBS), L-glutamine and 30% L929 cell supernatant containing the macrophage colony-stimulating factor at 37°C for 5 days. For BMDCs, bone marrow from the tibia and femur was obtained as described above, and bone marrow cells were cultured in RPMI 1640 medium containing 10% heat-inactivated FBS, L-glutamine and 5% J558L hybridoma cell culture supernatant containing the granulocyte macrophage colony-stimulating factor (GM-CSF) in 24-well plate for 5 days. The culture medium containing GM-CSF was replaced every other day. RAW264. 7 cells, HEK293T cells and HeLa cells were maintained in DMEM supplemented with 10% FBS. PLAT-gp cells (a gift from M. Shimojima and T. Kitamura) containing the retroviral gag and pol genes [54] were maintained in DMEM supplemented with 10% FBS and blasticidin (10 µg/ml; Invitrogen). 293FT cells (Invitrogen) for lentivirus production were maintained in DMEM supplemented with 10% noninactivated FBS, 6 mM L-glutamine, 0. 1 mM non-essential amino acid (NEAA) and 500 µg/ml Geneticin. HT1080 cells were maintained in E-MEM (Eagle' s minimal essential medium) supplemented with 10% FBS, 2 mM L-glutamine and 0. 1 mM NEAA. EMCV used for all experiments was grown in L929 cells for 15 h at 37°C. Influenza virus A/PR8 (H1N1) was grown in allantoic cavities of 10 day-old fertile chicken eggs for 2 days at 37°C. Viruses were stored at −80°C, and viral titer was quantified by a standard plaque assay using L929 cells for EMCV and Madin-Darby canine kidney cells for influenza virus. UV inactivation was performed by exposing viruses to 1. 0 J of UV light/cm2 with a Stratalinker UV closslinker (Stratagene). L. monocytogenes EGD (a gift from Y. Yoshikai, Kyushu University) was grown in tryptic soy broth (BD 211825) at 37°C overnight, washed repeatedly, resuspended in PBS containing 10% glycerol, and stored at −80°C in small aliquots. The concentration of bacteria was determined by plating 10-fold serial dilutions of bacterial suspension on tryptic soy agar plates and counting the colonies after cultivation at 37°C overnight. The cDNAs encoding EMCV 2A, 2B, and 2C proteins and mouse NLRP3 were obtained by reverse transcription and PCR of total RNA from EMCV-infected L929 cells and LPS-primed RAW264. 7 cells, respectively. The cDNAs encoding nonstructural 2B proteins from poliovirus type 1 (Mahoney strain) and enterovirus 71 (SK-EV006/Malaysia/97) were obtained by PCR from their full-length genomic cDNAs (a gift from S. Koike, Tokyo Metropolitan Institute) [55], [56]. These cDNAs were cloned into the eukaryotic expression vectors pCA7 [57] (a derivative of pCAGGS [58]), pCA7-Flag, or pCA7-HA to produce untagged or Flag- or HA-tagged proteins. The cDNA encoding the ER targeting sequence (calreticulin signal sequence) fused to the 5′ end of G-GECO1 were obtained by PCR using specific primers and the pTorPE-G-GECO1 (Addgene), which expresses an EGFP chimera that modulates its fluorescence in response to changes in the Ca2+ concentration [26]. The cDNA was cloned into the Nhe I and Bgl II sites of pDsRed2-ER (Clontech) to generate an expression plasmid encoding ER/Golgi-targeted G-GECO1 (pDsRed2-ER/Golgi-G-GECO1). Target sequences were designed using BLOCK-iT RNAi Designer (Invitrogen) or had been described previously [36]: 5′-GGA TCT TTG CTG CGA TCA ACA-3′ for NLRP3,5′-GCT CAC AAT GAC TGT GCT TAG-3′ for ASC, 5′-GGA CAA TAA ATG GAT TGT TGG-3′ for caspase-1, and 5′-GGC ACA AGC TGG AGT ACA ACT-3′ for EGFP. To generate shRNA-expressing retroviruses, pRS-shNLRP3, pRS-shASC, pRS-shCaspase-1, and pRS-shEGFP were generated by inserting the DNA fragments containing a mouse RNA polymerase III promoter (from the plasmid pBSsi-mU6 (Takara) ) and each target sequence into pRS-U6/puro vector (OriGene). PLAT-gp cells in collagen-coated 10-cm dishes were transfected with 20 µg of each shRNA-expressing pRS vector and 2 µg of pCVSV-G, which encodes the VSV G protein [59], using PEI-Max (Polysciences, Inc). Culture medium was replaced with fresh medium 6 h later, and supernatants containing retroviruses were harvested and filtered through a 0. 45-µm filter (Millipore) at 48 h post-transfection. To generate RAW264. 7 cells constitutively expressing shRNA targeting NLRP3, ASC, caspase-1, and EGFP mRNA, respectively, RAW264. 7 cells were infected with each shRNA-expressing retrovirus in the presence of polybrene (8 µg/ml) at 37°C. Then, cells were washed with phosphate-buffered saline (PBS) and cultured for further 24 h in complete DMEM in the presence of polybrene. Cells were cultured for 2 to 3 weeks in complete DMEM containing puromycin (0. 5 µg/ml) to kill non-transduced cells. Levels of expression of targeted genes were analyzed by real-time PCR and Western blot analysis (Figure S2). To generate lentiviruses expressing untagged or Flag-tagged EMCV 2A, 2B, or 2C protein, the full-length cDNA encoding each viral protein was cloned into pLenti6. 3/V5-TOPO vector (Invitrogen). 293FT cells in collagen-coated 10-cm dish were transfected with 3 µg of each viral protein-expressing pLenti6. 3/V5-TOPO vector together with ViraPower Packaging Mix (Invitrogen) using Lipofectamine 2000 (Invitrogen). Culture medium was replaced with fresh medium 24 h later, and supernatants containing lentiviruses were harvested and filtered through a 0. 45-µm filter (Millipore) at 72–96 h post-transfection. Lentivirus encoding an irrelevant protein (GFP) served as a control. The viral titer was quantified using HT1080 cells according to the manufacturer' s instructions (Invitrogen). Briefly, aliquots of serial 10-fold dilutions of the stock virus containing polybrene (8 µg/ml) were inoculated into HT1080 cells in six-well plates. Culture medium was replaced with fresh medium 24 h later. Then, cells were cultured in complete medium containing blasticidin (10 µg/ml) to kill non-transduced cells. After 10 to 14 days postinfection, cells were washed with PBS and the number of plaques was counted. BMMs, BMDCs, or RAW264. 7 cells were infected with EMCV at an MOI of 1. 5 or L. monocytogenes at an MOI of 100 for 1 h at 37°C, washed with PBS, and cultured with complete DMEM for 18 to 24 h. Unless otherwise stated, all experiments were performed in LPS-primed BMM, BMDCs, or RAW264. 7 cells. To inhibit caspase-1 activation, cells were pretreated with Ac-YVAD-CHO (Bachem) for 30 min. The cells were then infected with EMCV in the presence of Ac-YVAD-CHO for 1 h at 37°C, washed with PBS, and cultured with complete DMEM containing Ac-YVAD-CHO. Cell-free supernatants were collected at 18–24 h postinfection or transfection with viral RNAs. The supernatants were analyzed for the presence of IL-1α, IL-1β, or IL-6 using an enzyme-linked immunosorbent assay (ELISA) utilizing paired antibodies (eBiosciences) [9]. LDH release assays (#G1780, Promega) were performed according to the manufacturer' s instructions. LDH release data were used to account for cells death. The data are expressed as percentage of maximum LDH release. Mitochondria-associated ROS levels were measured by staining cells with MitoSOX (MolecularProbes/Invitrogen) according to manufacturer' s instructions. Cells were then washed with PBS, trypsinized, and resuspended in PBS containing 2% FBS. Flow cytometric analysis was performed on a FACSCalibur instrument (BD Bioscience). Viral RNAs were isolated from virions or L929 cells infected with EMCV at 10 h post infection by using TRIzol reagent (Invitrogen). The RNA extracts were treated with RQ1 DNase (Promega). The concentration of the resultant RNA was determined by NanoDrop technology (Thermo Scientific). BMMs were transfected with total RNA from infected or mock-infected L929 cells, genomic RNA from EMCV, or poly (dA∶dT) (10 µg/ml) using Lipofectamine 2000 (Invitrogen). After 24 h posttransfection, cell-free supernatants were collected and analyzed for the presence of IL-1β using ELISA utilizing paired antibodies (eBiosciences) [9]. At the same time, total RNA was extracted from the cells by using TRIzol reagent (Invitrogen), treated with RQ1 DNase (Promega), and reverse transcribed into cDNAs by using SuperScript III reverse transcriptase (Invitrogen) with an oligo (dT) primer, according to the manufacturer' s instructions. The SYBR Premix Ex Taq II (Takara) and a LightCycler (Roche Diagnostics, Indianapolis, IN) were used for quantitative PCR with the following primers: mouse IFN-β forward, 5′-GCACTGGGTGGAATGAGACTATTG-3′, and reverse, 5′-TTCTGAGGCATCAACTGACAGGTC-3′; mouse GAPDH forward, 5′-ACCACAGTCCATGCCATCA-3′, and reverse, 5′-TCCACCACCCTGTTGCTGTA-3′. Subconfluent monolayers of HEK293T or HeLa cells in six-well cluster plates were transfected with pCA7-Flag-2A, pCA7-Flag-2B, pCA7-Flag-2C, pCA7-Flag-NLRP3, pCA7-HA-2A, pCA7-HA-2B, pCA7-HA-2C, or pDsRed2-ER/Golgi-G-GECO1. At 48 to 72 h posttransfection, the cells were washed with PBS and lysed in 1 ml of coimmunoprecipitation buffer [50 mM Tris (pH 7. 5), 150 mM NaCl, 1% Triton X-100,1 mM EDTA, 10% glycerol] containing protease inhibitors (Sigma). The lysates were centrifuged at 20,630×g for 5 min at 4°C. Each supernatant was mixed with sodium dodecyl sulfate (SDS) loading buffer [50 mM Tris (pH 6. 8), 100 mM DTT, 2% SDS, 0. 1% bromophenol blue, 10% glycerol] and boiled for 5 min. These samples were fractionated by SDS-polyacrylamide gel electrophoresis using 12% polyacrylamide gel and electroblotted onto polyvinylidene difluoride (PVDF) membranes (Hybond-P; Amersham Biosciences). The membranes were incubated with anti-Flag (F7425; Sigma), anti-HA (sc-7392; Santa Cruz), anti-GFP (JL-8; Clontech), or anti-actin (sc-8432; Santa Cruz) antibody, followed by incubation with horseradish peroxidase-conjugated anti-rabbit IgG (Invitrogen) or anti-mouse IgG (Invitrogen) for detection of the Flag-tagged or HA-tagged proteins, respectively. The PVDF membranes were treated with Chemi-Lumi One Super (Nacalai Tesque) to elicit chemiluminescent signals, and the signals were detected and visualized using a VersaDoc 3000 imager (Bio-Rad). HeLa cells were seeded on coverslips in six-well cluster plates and transfected with 0. 5 µg each of pCA7-Flag-2A, pCA7-Flag-2B, pCA7-Flag-2C, pDsRed-monomer-Golgi (Clontech), pDsRed2-ER (Clontech), pCA7-HA-2A, pCA7- HA-2B, pCA7- HA-2C, pCA7- HA-M2, or pCA7-Flag-NLRP3. At 24 h posttransfection, the cells were fixed and permeabilized with PBS containing 2. 5% formaldehyde and 0. 5% Triton X-100. The cells were then washed with PBS and incubated with anti-Flag (F1804; Sigma) and anti-HA (561; Medical & Biological Laboratories Co.), followed by incubation with Alexa Fluor 488-conjugated donkey anti-mouse IgG (H+L) and Alexa Fluor 594-conjugated donkey anti-rabbit IgG (H+L) (Molecular Probes, Eugene, OR). To analyze subcellular localization of NLRP3, BMMs were seeded on coverslip in six-well cluster plates and treated with LPS, ionomycin, TG, EMCV, or lentivirus expressing Flag-tagged nonstructural protein from EMCV. At 24 h poststimulation, BMMs were fixed and permeabilized with PBS containing 2. 5% formaldehyde and 0. 5% Triton X-100. The cells were then washed with PBS and incubated with anti-NLRP3 (Cryo-2, AdipoGen), anti-calnexin (ab22595, Abcam), or anti-GM130 (EP892Y, Abcam), followed by incubation with Alexa Fluor 488-conjugated donkey anti-mouse IgG (H+L) antibodies. Nuclei were stained with DAPI (4. 6-diamidino-2-phenylinodole). The stained cells were observed using a confocal microscope (Radiance 2100; Bio-Rad or A1Rsi; Nikon). HeLa cells were seeded on 35-mm glass bottom dishes (Matsunami, Osaka, Japan) and infected with EMCV or influenza virus PR8 for 1 h, and washed with PBS and incubated with 1 ml Hanks' balanced saline solution (HBSS) containing 200 mM HEPES and 1 µg/ml Fluo8-AM for 1 h at 37°C. The cells were washed with PBS and incubated in HBSS containing 200 mM HEPES for 30 min at room temperature. Fluorometric cell images were recorded with an ICCD camera/image analysis system, and the intensities were determined. Fluo8-AM fluorescence was measured using excitation at 490 nm and emission at 514 nm. Statistical significance was tested by one-way ANOVA followed by Tukey' s post test using GraphPad PRISM software. P values of less than 0. 05 were considered statistically significant.

The innate immune system, the first line of defense against invading pathogens, plays a key role not only in limiting microbe replications at early stages of infection, but also in initiating and orchestrating antigen-specific adaptive immune responses. The innate immune responses against viruses usually rely on recognition of viral nucleic acids by host pattern-recognition receptors such as Toll-like receptors and cytosolic helicases. In addition, recent studies have indicated that certain viruses activate the NLRP3 inflammasome, a multiprotein complex containing the intracellular pattern-recognition receptor NLRP3, which in turn induces secretion of proinflammatory cytokines. We have previously revealed the role of the NLRP3 inflammasome in innate recognition of influenza virus, in which the influenza virus proton-selective ion channel M2 protein, but not viral RNA, is required. Here, we demonstrate that another RNA virus, encephalomyocarditis virus (EMCV), also activates the NLRP3 inflammasome in a viral RNAindependent manner. Instead, the EMCV viroporin 2B, which is involved in Ca2+ flux from intracellular storages into the cytosol, activates the NLRP3 inflammasome. Our results highlight the importance of viroporins, virusencoded transmembrane pore-forming proteins, in recognition of virus infections by NLRP3.

Abstract Introduction Results Discussion Materials and Methods

immunity innate immunity immunology biology

2012

Encephalomyocarditis Virus Viroporin 2B Activates NLRP3 Inflammasome

Computational methods for predicting evolutionarily conserved rather than thermodynamic RNA structures have recently attracted increased interest. These methods are indispensable not only for elucidating the regulatory roles of known RNA transcripts, but also for predicting RNA genes. It has been notoriously difficult to devise them to make the best use of the available data and to predict high-quality RNA structures that may also contain pseudoknots. We introduce a novel theoretical framework for co-estimating an RNA secondary structure including pseudoknots, a multiple sequence alignment, and an evolutionary tree, given several RNA input sequences. We also present an implementation of the framework in a new computer program, called SimulFold, which employs a Bayesian Markov chain Monte Carlo method to sample from the joint posterior distribution of RNA structures, alignments, and trees. We use the new framework to predict RNA structures, and comprehensively evaluate the quality of our predictions by comparing our results to those of several other programs. We also present preliminary data that show SimulFold' s potential as an alignment and phylogeny prediction method. SimulFold overcomes many conceptual limitations that current RNA structure prediction methods face, introduces several new theoretical techniques, and generates high-quality predictions of conserved RNA structures that may include pseudoknots. It is thus likely to have a strong impact, both on the field of RNA structure prediction and on a wide range of data analyses. Many RNA genes function by assuming a distinct three-dimensional structure in which the molecule folds back onto itself. Contacts are formed by hydrogen bonds between non-consecutive nucleotides that are complementary to each other. These hydrogen bonds are weak compared with covalent bonds. The three possible consensus pairs of complementary nucleotides are {A, U}, {G, C}, and {G, U}. It turns out that many properties of the three-dimensional RNA molecule can already be studied even if we know only the positions in the RNA sequence that form base-pairs. This is the level of abstraction that is predominantly chosen for studying RNA structure. For our purposes, an RNA structure is unambiguously defined by the set of base-pairing positions in the RNA sequence. This set of base-pairing sequence positions defines the RNA secondary structure. We count pseudoknotted structures, i. e. , structures that contain non-nested base-pairs (e. g. , two pairs i–j and i′–j′ whose sequence positions are in order i < i′ < j < j′) as secondary structures. The RNA structure allows us to draw conclusions about the molecule' s potential function and often even the mechanism by which it acts. It is therefore of fundamental importance to be able to predict an RNA' s structure from its sequence alone. Most RNA structure prediction programs investigate only secondary structures that do not contain pseudoknots. In addition, most of the structure prediction programs aim to predict the pseudoknot-free secondary structure that minimizes the free energy of the entire RNA molecule. The first empirical and theoretical investigations of the free energies of RNA secondary structures were conducted by Tinoco and his colleges in the early 1970s [1,2]. Because the number of possible secondary structures grows exponentially with the length of the RNA sequence, algorithmic tricks have to be employed to render the calculation of the minimum free energy (MFE) secondary structure tractable. The first fast algorithm—based on a primitive scoring scheme—for finding the pseudoknot-free MFE secondary structure was proposed by Nussinov and Jacobson [3]. A few years later, Zuker and Sankoff [4] showed how similar ideas can be used to define an algorithm that calculates the pseudoknot-free MFE secondary structure using the Tinoco energy model. This algorithm still forms the basis of several of today' s best MFE secondary structure prediction programs, e. g. , Mfold [5–7] and the programs RNAfold and RNAalifold of the Vienna package [7–11]. The MFE approach has, however, a number of limitations. One conceptual limitation is the underlying assumption that a given RNA sequence will assume its MFE structure in the cell, i. e. , its thermodynamic RNA structure. This assumption is not well supported in the general case. Theoretical, comparative studies of RNA molecules [12] show that the thermodynamic structure of even moderately long RNA molecules need not correspond to the “functional RNA structure, ” i. e. , the RNA structure that confers the observed functionality to the molecule and that is conserved during evolution. This may, for example, be due to co-transcriptional folding [13–15], i. e. , the folding of the RNA molecule as it is being transcribed. During co-transcriptional folding, a succession of kinetic RNA structures forms along a folding pathway. None of these kinetic RNA structures needs to correspond to the MFE structure. The observed discrepancies between the observed functional structure and the predicted MFE structure may also be due to other molecules binding the RNA molecule, which can obviously influence the structure formation process. The functional secondary structure may also differ from the MFE structure by having unstructured regions that do not comprise any base-pairs. As we are interested in investigating the functional roles of RNA molecules in the cell, we therefore focus on predicting the evolutionarily conserved RNA structure rather than the thermodynamic one. There exist several programs that aim to simulate the dynamic folding process of an RNA molecule in the cell to predict RNA structures that may contain pseudoknots [16–19]. However, these programs study only a single RNA sequence at a time, and their predictive power decreases with increasing sequence length because the error is multiplicative. Another conceptual limitation of the MFE approach is that the Zuker–Sankoff algorithm cannot handle pseudoknotted secondary structures, i. e. , structures with non-nesting base-pairs. However, pseudoknotted RNA structures are known to fulfill diverse and important functional roles in the cell [20]. We should thus aim to include them in RNA structure predictions. The prediction of pseudoknots has received more attention in the last few years, but remains very difficult. Pseudoknot prediction is, in the most general case, NP-hard even for binary strings [21,22]. For special classes of pseudoknots, structure predictions can be made more efficient from O (L4) to O (L6) for an RNA sequence of length L [23–30]. However, these algorithms are still too slow for practical purposes. The best information for predicting the functional RNA structure can be derived from functionally equivalent RNA sequences of evolutionarily related organisms. This is due to the fact that evolutionarily related RNA sequences that serve the same purpose in the cell are likely to employ the same mechanism for exerting this function. In particular, if the function of these RNA sequences depends on their structure, this RNA structure (but not necessarily the RNA sequences themselves) should be highly conserved. If we therefore align the RNA sequences such that structurally equivalent parts are grouped together, we can detect pairs of columns in the sequence alignment where the primary sequence conservation may be low, but the functional conservation in terms of base-pairs is high. These base-pairing columns in the alignment where compensatory mutations occur in a correlated way are called co-varying or co-evolving columns. They provide the main sequence signal that many comparative structure prediction programs detect to predict the base-pairs of evolutionarily conserved RNA secondary structures. RNAalifold [10] of the Vienna package combines this type of information with a traditional MFE structure prediction, whereas Pfold [31,32] incorporates it in a score-based approach that also takes the known evolutionary relationship of the sequences in the input alignment explicitly into account. RNA-Decoder [33,34] uses an approach similar to Pfold' s, but allows for extra evolutionary constraints due to known protein-coding regions in the input alignment and is the only one of the three programs capable of explicitly modeling unstructured regions. However, none of these three programs can predict pseudoknotted secondary structures. There already exist comparative programs that attempt pseudoknot prediction. These programs use the maximum weighted matching algorithm (MWM algorithm) [35,36] to extract an RNA secondary structure that may contain pseudoknots from a given set of base-pairs with different weights. Tabaska et al. [37] use a non-comparative approach to obtain these weighted base-pairs, whereas Witwer [38] analyzes a fixed input alignment with the comparative RNA structure prediction program RNAalifold [10] (which cannot predict pseudoknotted secondary structures) to obtain weighted base-pairs. The MWM algorithm requires O (L3) time to analyze an RNA sequence of length L, but requires a post-processing step to extract a bi-secondary structure [39]. Both programs have the same problems as the underlying algorithms and often have a low prediction accuracy [37]. Ruan et al. [40] developed a program that can utilize thermodynamic or comparative information or both. The overall accuracy is 80% for identifying base-pairs. However, this performance is achieved only with input alignments of very high quality that cannot be established without already knowing the conserved RNA secondary structure. The fundamental conceptual problem that all of these comparative programs face is that they require an input alignment of high quality to be able to predict the conserved RNA secondary structure. However, such an alignment can often only be established if we already know the conserved RNA secondary structure. These comparative structure prediction programs thus face a major chicken-and-egg problem. If the sequences are very well conserved and easy to align based on primary sequence similarity, the resulting alignment may contain no or few co-varying columns. If, at the other extreme, the sequences are only distantly related, a trustworthy sequence alignment that would exhibit many co-varying columns is impossible to establish based on primary sequence similarity alone. Comparative RNA structure prediction methods that take a fixed alignment as input can therefore analyze only a very limited range of available data successfully. This chicken-and-egg problem has been addressed by several comparative structure prediction programs that do not require a fixed input alignment, Dynalign [41,42], Foldalign [43,44], CARNAC [45,46], ComRNA [47], Stemloc [48], and CONSAN [49]. However, these programs find only very conserved local structures and do not model the evolutionary relationship between the sequences. ComRNA, CARNAC, and Stemloc can analyze several input sequences (Stemloc achieves this by calculating progressive pairwise alignments), whereas Dynalign, Foldalign, and CONSAN are limited to only two input sequences. ComRNA is the only one of these programs that can predict pseudoknotted secondary structures. The predictions of ComRNA rely on the calculation of maximal cliques, a problem that is known to be NP-complete. In the general case, it thus requires exponential time to run analyses, but it may be fast enough to analyze short sequences. To summarize, all of the existing RNA structure prediction programs face at least one of the following challenges: (1) the MFE structure rather than the evolutionarily conserved structure that is likely to correspond to the functional structure is predicted, (2) unstructured regions of the RNA are not explicitly modeled, (3) input alignments are fixed and cannot be altered and improved, (4) pseudoknotted structures are either completely ignored or computationally too expensive to predict, (5) only two evolutionarily related RNA sequences are used as input, or (6) the evolutionary relationship between the RNA sequences is not explicitly modeled. There are several good reasons to convince ourselves that many of these problems can be best solved simultaneously. For example, a good structure prediction should improve the prediction of a good alignment, and vice versa. Likewise, the prediction of a good alignment should improve the prediction of the correct evolutionary relationship of the RNA sequences, and vice versa. The idea of co-estimating RNA secondary structures, multiple sequence alignments, and evolutionary trees was first suggested in a theory paper by David Sankoff in 1985 [50]. As the proposed strategy is computationally very demanding, this approach did not receive much attention until the mid-1990s, when Eddy and Durbin [51] and Sakakibara et al. [52] introduced covariance models (CMs). CMs employ stochastic context-free grammars (SCFGs) [52] to align a given RNA sequence to a fixed multiple sequence alignment via a consensus RNA secondary structure that may not contain pseudoknots. CMs do not explicitly model the evolution of the sequences in the alignment, but only consider different nucleic acids. Such a model can also be used for pseudoknot-free secondary structure prediction [32,34]. Modeling the insertion–deletion process with constraints on a pseudoknot-free consensus RNA secondary structure is a much harder problem, and, so far, only a few studies [53,54] have tried to address the problem. By considering both alignment and secondary structure, other studies extended the pioneering work of Sankoff without explicitly considering an evolutionary model [55,56]. As pseudoknots are context-dependent structures that cannot be modeled with SCFGs, CMs cannot be used to model pseudoknotted RNA structures. We here propose a novel theoretical framework for solving the problem of co-estimating RNA secondary structures including pseudoknots, multiple sequence alignments, and evolutionary trees. We introduce a joint distribution of RNA structures, alignments, and trees in a Bayesian framework. As it is not feasible to analytically calculate any interesting statistics in this model in reasonable computational time, we propose a Markov chain Monte Carlo (MCMC) method with which we can sample from the posterior distribution. According to elementary probability theory, the following equation holds: where D stands for data, i. e. , the individual, un-aligned RNA sequences, Z: = P (D) denotes the so-called partition function, S is a consensus RNA secondary structure that may contain pseudoknots, A is a multiple sequence alignment, and T is an evolutionary tree relating the sequences. Equation 1 is also called Bayes' theorem. The aim of the MCMC algorithm is to sample from the posterior probability distribution, i. e. , from P (S, A, T|D), using the terms on the right hand side of Equation 1. For the sampling, we need to know only the ratio of the probabilities, P (S1, A1, T1|D) /P (S2, A2, T2|D). We thus have to be able to calculate P (S, A, T|D) and P (S, A, T) only up to a constant factor and can, for example, omit the calculation of Z. We now explain how we calculate the different terms on the right hand side of Equation 1. We also introduce models and explain how to employ them to calculate the terms on the right hand side of the equation. The definitions that we propose for the prior probabilities merit a detailed discussion as there is currently no widely accepted consensus on how to define these prior distributions. Concerning the calculation of the likelihood, we make a conscious decision to use the widely known Felsenstein likelihood. The analytical calculation of the posterior distribution is computationally too expensive. Instead, we employ a Bayesian MCMC method [62,63] to sample from the posterior distribution. This corresponds to a random walk on the possible states (S, A, T), whose stationary distribution is the posterior distribution, P (S, A, T|D). The random walk is constructed in two steps. In the first step, a new state, Xnew: = (S, A, T) new in our case, is drawn from a proposal distribution, P, and in the second step, the discrepancy between the proposal and the target distribution, π, is corrected by accepting the proposal with probability where X = (S, A, T) is the actual state of the chain. The chain remains in state X with probability 1 − Paccept [62,64]. The mixing of the Markov chain depends on how closely the proposal distribution resembles the target distribution. Gibbs sampling is a special case of MCMC sampling, where each state X can be described as a multidimensional vector X = (x1, x2, . . . xn) and where it is possible to draw a new random coordinate xi from the conditional distribution P (xi|X[−i]) for any state X and any coordinate i. X[−i] denotes the vector of coordinates without coordinate i. As the newly drawn coordinate is always accepted, the Gibbs sampler is an MCMC method with an acceptance probability of one. As it is generally not possible to sample from an arbitrary conditional distribution, the Gibbs sampling strategy can only rarely be used. However, it is possible to mimic the conditional distribution with an auxiliary distribution. This strategy is employed in partial importance sampling; see MacKay [65] for an overview of different sampling strategies. Importance sampling has been successfully used to model different distributions that occur in the context of bioinformatics [66,67], and we employ it for proposing alignments and RNA secondary structures. We define a Markov chain that converges to the desired distribution and then use this chain to sample from the posterior. As we have seen in the section before, the posterior distribution is a joint distribution on RNA structures, multiple sequence alignments, and trees. The challenge is to define moves on this joint distribution that are reversible and ergodic, and that satisfy detailed balance [63]. It is possible to define tree moves that are independent from the actual alignment and RNA structure. However, it is generally impossible to alter the alignment without disturbing the RNA structure. We therefore use the following three types of moves: changing the length of an edge in the tree, changing the tree topology, and using a complex move that alters both the RNA structure and the alignment. The primary result of an MCMC run is a large set of simulated (S, T, A) triples that are distributed according to the posterior distribution. This data needs post-processing to characterize and visualize the posterior distribution. As we are interested in deriving the RNA structure that is best supported by the posterior distribution, we therefore marginalize with respect to the RNA structure. We project the RNA structure, S, of each sampled (S, T, A) triple onto the RNA sequences in D and thereby obtain a set of RNA structures for each individual RNA sequence. The challenge is to combine these structures into a single RNA structure that captures the prominent features of the set of structures. There already exist a number of programs that determine a consensus structure for a given set of RNA structures, e. g. , RNAdistance of the Vienna package [7,73] and RNA-Forester [74,75]. RNA-Forester computes a global structural alignment for several unaligned sequences with known secondary structures using a dynamic programming procedure that depends on scores with combined information on structure and sequence similarity. However, RNA-Forester and RNAdistance both require the input RNA structures to be secondary structures and cannot handle pseudoknots. SimulFold is to our knowledge the first program that predicts an RNA structure including pseudoknots while simultaneously estimating an alignment as well as an evolutionary tree for several, evolutionarily related input RNA sequences. It was therefore not possible to present a comparison to a truly equivalent program. Instead, we compare the RNA structures predicted by SimulFold to those predicted by RNAalifold [10], Hxmatch [78], Pfold [31,32], and CARNAC [45,46]. RNAalifold takes a fixed alignment as input and predicts a consensus RNA secondary structure without pseudoknots. It extends the MFE algorithm employed by the non-comparative MFE methods Mfold and RNAfold by interpreting the fixed input alignment as a hyper-sequence and by simultaneously minimizing the overall free energy while taking the primary sequence conservation and co-varying columns in the fixed alignment into account. The optimization is implemented in a dynamic programming procedure that combines free energy parameters with conservation scores. Hxmatch is an extension of RNAalifold. Like RNAalifold, it takes a fixed alignment as the only input and predicts a consensus RNA secondary structure. However, unlike RNAalifold, it is capable of predicting secondary structures with pseudoknots. Hxmatch employs a two-step procedure. In the first step, the fixed input alignment is analyzed with RNAalifold [10] of the Vienna package, which calculates the base-pairing probability for each possible pair of columns in the alignment by considering all possible secondary structures without pseudoknots. In the second step, these weighted base-pairs are used as input to the MWM algorithm. The MWM algorithm derives the highest-scoring subset of mutually compatible base-pairs, requiring O (L3) time and O (L2) memory to analyze an input alignment of length L. As these base-pairs need not correspond to a bi-secondary structure, a heuristic, greedy algorithm is then employed to extract a bi-secondary structure. The MWM algorithm and the greedy algorithm are repeatedly used until the resulting bi-secondary structure remains unchanged or 30 iterations have been completed. Hxmatch does not take the evolutionary relationship of the input RNA sequences explicitly into account. Pfold takes as input not only a fixed alignment, but also an evolutionary tree relating the sequences, and predicts a consensus secondary structure which does not contain pseudoknots, as Pfold cannot handle pseudoknots. Pfold employs an SCFG, i. e. , a probabilistic rather than an MFE model, to derive the consensus secondary structure. Similar to RNAalifold, it takes the primary sequence conservation and co-varying columns in the fixed input alignment into account. Unlike RNAalifold, Pfold also takes the known evolutionary relationship of the input sequences, i. e. , the input tree, explicitly into account. Both, Pfold, and RNAalifold require O (L3) time and O (L2) memory to analyze an input alignment of length L. CARNAC is also a comparative RNA structure prediction method. It takes several unaligned RNA sequences as input and predicts an RNA structure for each individual RNA sequence which does not contain pseudoknots, as CARNAC cannot handle pseudoknots. Similarly to Hxmatch and RNAalifold, it does not take the evolutionary relationship of the input sequences explicitly into account. CARNAC employs a multi-step procedure for generating predictions. In the first step, potential helices are predicted for each RNA sequence separately. In the second step, an optimal consensus secondary structure is extracted from these helices for every possible pair of RNA sequences. In the third and last step, the different secondary structures that were predicted in a pairwise fashion for each individual RNA sequence are combined into one secondary structure using graph theoretical techniques. This is the final RNA structure reported by CARNAC for that RNA sequence. In the most general case, the algorithms underlying CARNAC would require O (L6) time and O (L4) memory to analyze input sequences of length L [45]. These requirements can be reduced by a number of computational tricks. For the data investigated by Perriquet et al. [45], the empirically observed requirements were approximately O (L2) time and memory. We compiled a large and diverse dataset from previously published data [78–80] to thoroughly investigate SimulFold' s ability to correctly predict RNA structures. Our dataset consists of 16 sets of evolutionarily related sequences that cover a wide range of average pairwise sequence identities (pids) and sequence lengths. Half of the sets contain a pseudoknotted reference structure; the other half contain a reference structure without pseudoknots. The number of sequences in each set ranges from five to 15 sequences. We automatically generated Clustal-W alignments for all 16 sets to serve as input alignment for those programs that take a fixed input alignment. The resulting alignments are 74–1,601 nucleotides long, and their average pid ranges from 40% to 91%. Table 1 summarizes the main characteristics of each set (see the columns “Structure” and “Alignment” and the caption of the table). Table 1 shows the performance values for the RNA structures predicted by SimulFold, RNAalifold, Hxmatch, Pfold, and CARNAC for all 16 sets. We used the Clustal-W alignments as fixed input alignments for Hxmatch and Pfold. The same alignments were also used as initial alignments for SimulFold. To evaluate the structure prediction performance, we compared the known RNA structure of the reference organism in each set with the corresponding predicted RNA structure. We measured the quality of the structure predictions in terms of the number of correctly predicted base-pairs (true positive base-pairs, or TP, see Table 1), the number of incorrectly predicted base-pairs (false positives, or FP), and the number of known base-pairs that have not been correctly predicted (false negatives, or FN). We also calculated Mathews' s correlation coefficient (MCC) (see Table 1), which is defined as CARNAC generally shows a high specificity, i. e. , a low number of incorrectly predicted base-pairs, often in combination with a low sensitivity, i. e. , a low number of true positive base-pairs. This low sensitivity can even be found for sets whose average pid is fairly high, e. g. , set U5 (high) with an average pid of 88%. CARNAC' s performance is naturally limited by the fact that it cannot predict pseudoknotted structures. Besides SimulFold, Hxmatch is the only other investigated program that is capable of predicting pseudoknotted secondary structures. Hxmatch has the tendency to over-predict base-pairs, as indicated by the high number of false positive base-pairs. This happens for low average pids (e. g. , set tRNA [low]) and for high pids (e. g. , set SSU [high]). It is interesting to note that RNAalifold often does better than Hxmatch at predicting the base-pairs of pseudoknotted reference structures, e. g. , the results for RNaseP (medium), RNase P8, SSU (medium), and SSU (high). However, there are also examples, see the corona set, where the reverse holds. Like RNAalifold and Hxmatch, Pfold is a program that takes a fixed alignment as input. Its performance tends to be low for the low average pid range, e. g. , the U5 (low), tRNA (low), and rRNA (low) sets, which all have average pids below 50%. For the high pid range, its performance can be limited because of the fact that Pfold cannot model pseudoknots, e. g. , in the corona, entero, and hepatitis delta virus (HDV) sets. Its performance for the RNaseP (medium) set constitutes a notable exception to this trend. SimulFold is the only program that simultaneously co-estimates alignments, structures, and trees. It clearly outperforms all other programs in terms of overall performance for eight out of 16 sets: U5 (low), group II intron (low and high), tRNA (low), rRNA (low and high), entero, and HDV. It also shows a competitive performance for the sets U5 (high) and tRNA (high). These sets cover a wide range of average pids, from 40% to 91%. The results for the two SSU sets show that SimulFold has problems analyzing these two sets, whose reference alignments span more than 1,500 nucleotides. However, the results for the RNase P8 set show that SimulFold can successfully predict structures with high sensitivity even for comparatively long sequences (the reference alignment of the RNase P8 set has a length of 472 nucleotides). The results for the RNase P8 and the HDV sets show the benefits of parallel tempering. When investigating the predictions for the HDV set, we concluded from the loglikelihood plot (see Figure 5) that the MCMC chain got stuck in local minima. We therefore implemented a more sophisticated version of SimulFold that employs the MCMC technique of parallel tempering [81] to address the problem. As the grey line in Figure 5 shows, parallel tempering solves the mixing problem for the HDV set and significantly improved the sensitivity, while at the same time reducing the number of incorrectly predicted base-pairs. Our initial motivation for devising a novel method that simultaneously co-estimates RNA structures, alignments, and evolutionary trees was to improve the prediction of RNA structures, in particular those with pseudoknots. A very interesting additional benefit of our approach is that SimulFold can also be used as an alignment and phylogeny prediction program. We here present preliminary results for sequences from the HDV set that show SimulFold' s potential as an alignment and phylogeny prediction program. By the same argument that we made above for RNA structure prediction, we should also be able to derive better alignments and trees if we co-estimate all three, interdependent quantities together rather than in isolation. The HDV dataset contains 15 sequences of HDV ribozymes from several strains (their NCBI accession numbers are shown on the figures showing the alignments and the consensus networks). The ribozyme contains one pseudoknot and a variable helix. We calculated posterior probabilities of alignment columns from the multiple alignments that the MCMC method sampled, i. e. , the probability of seeing a particular alignment column in a sampled multiple alignment. We calculated the maximum posterior decoding (MPD) alignment using a dynamic programming procedure [72]. It has been shown [67] that MPD yields better estimates than maximum a posteriori alignment estimation from an MCMC sample. The MPD alignment is shown in Figure 6, together with the estimated posterior probabilities for each column as well as the reference secondary structure that includes a pseudoknot. The figure clearly highlights three regions in the alignment where lower posterior probabilities are due to an ambiguity in the estimation. The first region overlaps a hairpin loop. Even though the MPD alignment contains no gaps in this region, the sequences vary a lot and there exist several plausible explanations that relate the sequences in this region in terms of evolutionary indel events. The remaining two regions overlap the two base-paired sides of a variable helix. The low posterior probabilities indicate that several plausible alignments exist for these regions. These observations are in line with our difficulty to correctly predict these parts of this helix. We conjecture that this helix may be shorter or contain bulges in some of the sequences of the HDV set. If we compare our MPD alignment to the alignment generated by Clustal-X [61] shown in Figure 7, we observe two main differences. First, Clustal-X does not take secondary structure into account when predicting the alignment. This yields several nonsense base-pairs with respect to the known reference structure (highlighted in green in Figure 7). Second, Clustal-X does not evaluate the reliability of the different regions in the predicted alignment. We thus do not know which parts of the predicted alignment are particularly well or poorly supported by the data. We calculated consensus networks based on the evolutionary trees sampled from the posterior distribution by the MCMC using the method of Holland and Moulton [82] implemented in the SplitsTree4 program [83]. We set the threshold for splits to 0. 1, i. e. , we retained only splits that were present in at least 10% of the sampled trees, and generated the two networks shown in Figure 8. The two networks have the same topology, but differ in the lengths of their edges, which represent different kinds of information. In the left network, the length of each edge is proportional to the probability of the split that is represented by the edge in the posterior distribution (the unit in the top left corner shows 1,000 occurrences in 2,000 sampled trees). In the right network, the length of each edge is equal to the average length of the edge in the sampled trees that contain that edge. There are five groups of strains: the lone strain AJ309873; a group containing U81988, M28267, X77627, and M92448; another group containing U81989, AF104263, AF104264, and X85253; and finally two relatively close groups containing AB088679, AF018077, and AF309420, and L22063, AB03748, and AJ309880. There is not enough phylogenetic signal to infer the relationship between the union of the two last groups and the other three groups. As Figure 8 indicates, there are several plausible explanations for how strains in the first two groups could have evolved. These preliminary results show that SimulFold not only allows us to derive consensus multiple alignments and evolutionary trees, but even enables us to highlight particularly well or poorly estimated parts of these alignments and trees. It is easy to think of situations where one does not want to simultaneously co-estimate RNA structures, alignments, and trees, e. g. , because a high-confidence RNA structure (or alignment or tree) has already been established. It is straightforward to employ SimulFold in these situations, as the program can be easily told to keep the input RNA structure or alignment or tree (or any combination thereof) fixed. An MCMC can suffer from a low efficiency for three main reasons: (1) the acceptance ratio is low, (2) the Markov chain gets stuck in local optima, or (3) the computational time to perform each step is large. We introduced partial Metropolis importance sampling to quickly propose moves that replace only part of the data and to keep the rejection probability and autocorrelation low. For the HDV set, the initial loglikelihood plot shows poor mixing (see Figure 5). We therefore implemented the more sophisticated MCMC technique of parallel tempering in SimulFold. This option can be switched on whenever the mixing properties need to be improved. Figure 5 shows, for the HDV set, how parallel tempering can considerably improve the mixing properties. The parallel tempering run took 1. 5 d and used 50 MB of memory on an Intel Xeon dual 3 GHz machine for the HDV set, involving seven parallel chains, 10,000 steps for burn-in, and 2,000 samples, and making 100 steps between two samplings. Shorter sequences, multiple alignments with fewer sequences, and MCMC runs without parallel chains took proportionally less time. In terms of computational complexity, SimulFold takes O (N · L) time to propose a change in phylogeny for N sequences and an alignment of length L. Changing the alignment takes O (N · W2) time, where W is the length of the window that is re-sampled. Changing the structure takes O (K · N · L) time, where K is the number of changed helices. In its current implementation, SimulFold uses the same number of moves to update trees, alignments, and structures. The mixing of the chain decreases with the length and number of sequences. We did not investigate this behaviour in detail, but know that for multiple sequence alignments, the mixing time is proportional to N · L. We propose here a novel theoretical framework for co-estimating an RNA structure including pseudoknots, S, a multiple-sequence alignment, A, and an evolutionary tree, T, given several evolutionarily related RNA sequences, D, as input. We also present an implementation of this framework in a new computer program, called SimulFold, and evaluate the quality of the predicted RNA structures relative to those predicted by existing programs. Our novel theoretical framework allows us to sample (S, T, A) triples from the posterior distribution, P (S, A, T|D), in a computationally very efficient way using a Bayesian MCMC. For every RNA sequence in D, we then extract the RNA structure that is best supported by the posterior distribution using the MWM algorithm and a post-processing step implemented in an auxiliary program called bp2bistruc. Our work is significant in a number of ways. SimulFold overcomes several limitations of existing RNA structure prediction methods, in particular the conceptual limitations of SCFG-based methods. SimulFold does not rely on a fixed input alignment or tree, it can predict pseudoknotted RNA structures, it can take any number of related RNA sequences as input, it aims to predict the evolutionarily conserved RNA structure rather than the thermodynamic or MFE structure, it explicitly models the evolutionary relationship between the RNA sequences, it is a fully probabilistic method that is capable of quantifying the reliability of its predictions, and, most important for the majority of users, it works in a computationally efficient way and can be used on any standard desktop computer. Furthermore, SimulFold derives the RNA structure that is best supported by the posterior distribution, rather than the RNA structure that maximizes the likelihood, which is what SCFG-based structure prediction methods do. We use a number of novel theoretical and computational tricks to achieve the above. We devised a new expression for the prior P (S, A, T), in particular a function that models the contribution of trees, a function that incorporates information on structures and alignments, and a function that quantifies the contribution of gaps in the alignment. For sampling from the posterior distribution, we propose a new way of sampling trees and a fairly sophisticated new way of jointly sampling structures and alignments in a computationally very efficient way. After O (N · L2) pre-processing time, we do an MCMC step modifying the base-pairs that requires O (N · L) time, where N is the number of sequences and L is the average length of the sequences. We also introduce a new type of MCMC sampler that we call a partial Metropolis importance sampler. We implemented the sophisticated MCMC technique of parallel tempering into SimulFold, which can be switched on whenever the loglikelihood plot indicates poor mixing properties. Finally, we introduce a new program, bp2bistruc, that derives an RNA structure that may include pseudoknots (a bi-secondary structure, to be precise) from an input table of base-pairing probabilities. The performance of SimulFold in predicting RNA secondary structures with and without pseudoknots compares very well to the performance of RNAalifold, Hxmatch, Pfold, and CARNAC across a wide range of average pids and sequence lengths. We also present encouraging preliminary results that show SimulFold' s potential as an alignment and phylogeny prediction program. It is not only possible to derive a consensus alignment and tree, but also to highlight those parts of the alignment and tree that can be particularly well or poorly estimated. This information is very valuable for interpreting the results in great detail. It is easy to think of situations where one does not want to simultaneously co-estimate RNA structures, alignments, and trees, e. g. , because a high-confidence RNA structure (or alignment or tree) has already been established. We therefore implemented special flags in SimulFold that allow the user to keep the input RNA structure or alignment or tree (or any combination thereof) fixed. We hope that this feature will make SimulFold a useful program for a wide range of interesting tasks and data analyses. In the future, we intend to investigate different models and priors for use in SimulFold, e. g. , a co-transcriptional folding prior. We also hope to further improve the properties of the sampling, e. g. , partial importance sampling of tree or an even better structure sampler, to improve the performance for very long sequences. SimulFold opens up a large number of possibilities for exciting data analysis. Most importantly, we can now start analyzing data whose low primary sequence conservation has so far prevented their analysis with methods that require a high-quality input alignment. We hope that our work inspires other researchers to also develop methods that predict or investigate the functional structure of RNA sequences so that we learn more about how RNA sequences play their diverse functional roles in the cell. SimulFold as well as information on the input and output files of this analysis can be found at http: //www. cs. ubc. ca/∼irmtraud/simulfold/.

Not only is the prediction of evolutionarily conserved RNA structures important for elucidating the potential functions of RNA sequences and the mechanisms by which these functions are exerted, but it also lies at the core of RNA gene prediction. To get an accurate prediction of the conserved RNA structure, we need a high-quality sequence alignment and an evolutionary tree relating several evolutionarily related sequences. These are two strong requirements that are typically difficult to fulfill unless the encoded RNA structure is already known. We present what is to our knowledge the first method that solves this chicken-and-egg problem by co-estimating all three quantities simultaneously. We show that our novel method, called SimulFold, can be successfully applied over a wide range of sequence similarities to detect conserved RNA structures, including those with pseudoknots. We also show its potential as an alignment and phylogeny prediction method. Our method overcomes several significant limitations of existing methods and has the potential to be used for a very diverse range of tasks.

Abstract Introduction Methods Results Discussion

evolutionary biology genetics and genomics vertebrates computational biology

2007

SimulFold: Simultaneously Inferring RNA Structures Including Pseudoknots, Alignments, and Trees Using a Bayesian MCMC Framework

Biophysical models of cardiac tension development provide a succinct representation of our understanding of force generation in the heart. The link between protein kinetics and interactions that gives rise to high cooperativity is not yet fully explained from experiments or previous biophysical models. We propose a biophysical ODE-based representation of cross-bridge (XB), tropomyosin and troponin within a contractile regulatory unit (RU) to investigate the mechanisms behind cooperative activation, as well as the role of cooperativity in dynamic tension generation across different species. The model includes cooperative interactions between regulatory units (RU-RU), between crossbridges (XB-XB), as well more complex interactions between crossbridges and regulatory units (XB-RU interactions). For the steady-state force-calcium relationship, our framework predicts that: (1) XB-RU effects are key in shifting the half-activation value of the force-calcium relationship towards lower [Ca2+], but have only small effects on cooperativity. (2) XB-XB effects approximately double the duty ratio of myosin, but do not significantly affect cooperativity. (3) RU-RU effects derived from the long-range action of tropomyosin are a major factor in cooperative activation, with each additional unblocked RU increasing the rate of additional RU’s unblocking. (4) Myosin affinity for short (1–4 RU) unblocked stretches of actin of is very low, and the resulting suppression of force at low [Ca2+] is a major contributor in the biphasic force-calcium relationship. We also reproduce isometric tension development across mouse, rat and human at physiological temperature and pacing rate, and conclude that species differences require only changes in myosin affinity and troponin I/troponin C affinity. Furthermore, we show that the calcium dependence of the rate of tension redevelopment ktr is explained by transient blocking of RU’s by a temporary decrease in XB-RU effects. Tension generation in cardiac muscle is a highly cooperative process, with significant increases in tension caused by relatively small increases in the calcium concentration. The Hill coefficient (nH) describing the degree of cooperativity of the force-calcium relationship is typically around nH = 3 in experiments on skinned muscle cells [1], and as high as nH = 10 in intact cells [2]. Our understanding of the molecular mechanisms giving rise to this cooperative activation and the precise regulation of tension generation required for effective cardiac pump function remains incomplete. However, there is a general agreement on the potential types of interactions involved in cooperative activation between regulatory units (RU) and crossbridges (XB) [3–5]. Each half-sarcomere in a myocyte contains 26 RU’s, and each RU consists of 7 actin monomers, one long tropomyosin molecule spanning the actin monomers, and a complex of troponin (troponin I, troponin C and troponin T) which regulates local activation. Within an RU, calcium (Ca2+) bind to troponin C (TnC), causing a conformational change in tropomyosin, unblocking actin for myosin crossbridge (XB) binding [6]. Underlying cooperative activation, three types of interactions are proposed between regulatory units and crossbridges. Cooperative effects between RU’s are known as ‘RU-RU cooperativity’, where unblocking of tropomyosin in one RU leads to an increased probability of unblocking in a nearby RU, due to overlap of tropomyosin molecules between neighbouring RU’s. Evidence in support of these effects includes experimental data which shows a significant decrease in cooperativity when the overlap between neighbouring tropomyosin units is removed or reduced [7–9], dependence on nearest neighbour interactions in cardiac muscle [10], and modifications to long-range cooperativity by phosphorylation of tropomyosin [11]. In addition there are cooperative interactions in which the binding of crossbridges increases the rate at which further crossbridges bind, known as ‘XB-XB cooperativity’ [12,13]. Although XB-XB interactions can increase the steady-state force per activated RU, more complex interactions with neighbouring RU’s are involved in their effect on cooperativity. These more complex interactions by which crossbridges affect RU activation are known as ‘XB-RU cooperativity’ [13,14]. Evidence for the importance of these effects on muscle activation can be seen from various experiments in which calcium sensitivity is affected by changes to crossbridge affinity using crossbridge inhibitors and enhancers [1,15–17]. A potential factor in XB-RU cooperativity are the proposed effects of tension generation on the affinity of TnC for Ca2+ [18,19]. The mechanisms and significance of this interaction remain controversial, with some researchers claiming this effect appears mainly from non-physiological rigor crossbridges [1], while others point to it as a key component of normal muscle function [18,20,21]. In addition to uncertainties in the biophysical basis for cooperativity, the exact link between calcium binding to TnC and the movement of tropomyosin has remained obscure, troponin I (TnI) is known to play a key role in transmitting this signal [22–24], and in recent years this link has been clarified with research on crystal structures of troponin [25,26]. These studies showed that calcium binding to TnC opens up a hydrophobic patch on TnC which has a high affinity for the switch region of TnI [27]. The movement of the switch region also moves the nearby inhibitory (‘C-terminal’) region of TnI which is responsible for pinning tropomyosin in the blocking position on actin in resting muscle [28]. The competitive binding of these TnI regions to both TnC and actin results in the unblocking of actin at higher Ca2+ concentration, allowing myosin crossbridges to bind and generate force. Further support for the critical role of TnI is given by its numerous phosphorylation sites and role in regulating muscle function through β-adrenergic stimulation and the response to length-dependent activation [28–30]. Fig 1 gives an overview of a regulatory unit (RU) and its states in this competitive binding framework. In the three-state framework proposed by McKillop and Geeves [31] RU’s can be either in the ‘blocked’ state with TnI pinning tropomyosin to actin, in the neutral ‘closed’ state where myosin crossbridges are able to bind, or in the ‘open’ state with crossbridges having moved tropomyosin in the opposite direction compared to TnI binding. The continuous flexible chain models represent the spatial deformation of tropomyosin along the whole thin filament. At points along the chain with a TnI binding site or crossbridge, the chain is in a fully ‘blocked’ or ‘open’ position respectively. However, in the space between bound sites, the chain can occupy a continuum of intermediate states. Due to the single TnI binding site per RU, we can still unambiguously refer to an RU as blocked based on TnI binding. Describing an RU itself as ‘open’ becomes more problematic in this modelling framework, as there are 2–3 crossbridges per RU and any combination of these can be bound to actin at any one time. In the rest of this paper we refer to the state of RU’s only as ‘blocked’ and ‘unblocked’ based on TnI-actin binding, regardless of the tropomyosin deformation induced or number of crossbridges bound near the RU. There are several challenges in applying these advances in physiology to create a computational model of cardiac contraction that is both tractable for a wide range of simulation and analysis, and captures the critical physiological features of the underlying proteins. Firstly, computational models which include tension-dependent feedback mechanisms often suffer from non-physiological hysteresis, in which tension generation is higher for decreasing calcium compared to increasing calcium [32]. Secondly, in the absence of a clear mechanistic explanation for cooperativity, computational models based on ordinary differential equations (ODE) tend to use phenomenological representation of cooperativity to achieve adequate tension development [5,33–36]. Some recent developments have begun to address these shortcomings, including detailed models of the thin filament based on spatial interaction of tropomyosin [37,38]. Firstly, the work by Campbell et al. includes a model of tropomyosin interaction between neighbouring RU [38], and is based on ODEs. However, it is limited to approximately 9 RU’s, and requires the assumption that calcium bound to TnC does not unbind in the tropomyosin ‘closed’ state. Extending the model beyond these assumptions quickly leads to an increase in the required number of states beyond what is computationally tractable. Nevertheless, the model is arguably the most biophysically detailed contraction model to have been applied in the context of a whole-organ cardiac mechanics [39]. Secondly, a more detailed underlying model of cooperativity is given by models of tropomyosin as a continuous flexible chain, based on the work by Smith et al. [40–43]. These approaches assume that tropomyosin, which consists of many molecules overlapping end-to-end to form a long filament, can be modelled as a homogeneous flexible chain. The deformation of the tropomyosin chain in these models is determined by a combination of weak electrostatic interactions with actin and elastic deformation of the chain. Although still a simplification that ignores potential inhomogeneities arising from end-to-end overlap, these models provide a more detailed description of tropomyosin kinetics which are able to describe longer range interactions in the thin filament, compared to models which assume only nearest-neighbour interactions. Solving these more detailed models remains computationally challenging, and results are typically given by developing approximations for the equations of the deformation of tropomyosin, or applying stochastic approaches to predict a steady-state force-calcium relationship. Our goal in this article is to create an ODE-based model of cardiac contraction with a biophysically detailed representation of cooperativity based on the competitive binding model of troponin I and the continuous flexible chain model for tropomyosin. The formulation of an ODE-based model facilitates modelling of a wide range of simulations of dynamic function of muscle, and will allow us to link this model to whole organ mechanics in the future. This paper is organized as follows: We start with a general theory on modelling tropomyosin as a continuous flexible chain and the use of Boltzmann’s law. The section “Steady-state models” describe our model for the steady-state blocking and unblocking of RU’s in the absence of myosin crossbridges. We extend this model to include myosin crossbridges, and develop techniques to make this approach computationally tractable. This extended steady-state model is then used to explain the sources of cooperativity, and the effects on myosin binding in producing XB-RU effects and changes in Ca2+-TnC affinity. The section “Dynamic models” develops the dynamic models of cardiac contraction, which we use to investigate the role of cooperative activation in isometric tension development across different species, as well as the influence of cooperative effects on the rate of tension redevelopment and its dependence on Ca2+. The steady-state models developed in the previous sections replicate a range of experimental measurements related to steady-state cooperative activation. Cooperative effects also have an important impact on beat-to-beat dynamic tension generation, and may have different roles in different species due to differences in heart rate and calcium dynamics. To be able to investigate the role of cooperative activation in dynamic tension generation, in this section we extend our proposed framework to simulate dynamic changes in tension in response to transient changes in Ca2+. For a dynamic model of n RU’s and m crossbridges, we use: A regular grid of (n + 1) ⋅ (m + 1) state variables TmXBi, j which represent the fraction of half sarcomeres with i RU’s unblocked and j crossbridges bound. The state variable TnCB, the fraction of RU’s that are blocked with Ca2+ bound to TnC The state variable TnCU, the fraction of RU’s that are unblocked with Ca2+ (but not TnI) bound to TnC, The state variable TnITnC, the fraction of RU’s that are blocked with Ca2+ and TnI bound to TnC Several dependent variables are useful in the formulation in the differential equations for these states: U = ∑ i = 0 n ∑ j = 0 m i n TmXB i, j (fraction of RU’s in the unblocked state) (10) B = 1 - U (fraction of RU’s in the blocked state) (11) TnI = U - TnITnC (fraction of RU’s with TnI not bound to actin or TnC) (12) The kinetics of the tropomyosin and crossbridge states can now be defined using a standard Markov model approach, with transition rates defined using the ratio of Boltzmann terms: dTmXB i, j dt = k i-1, j b → u TmXB i-1, j - k i, j u → b TmXB i, j ︷ if i ≠ 0 + k i+1, j u → b TmXB i+1, j - k i, j b → u TmXB i, j ︷ if i ≠ n + k i, j-1 xb+ TmXB i, j-1 - k i, j xb- TmXB i, j ︷ if j ≠ 0 + k i, j+1 xb- TmXB i, j+1 - k i, j xb+ TmXB i, j ︷ if j ≠ m (13) k i, j xb+ / k i, j+1 xb- = SE i, j + 1 SE i, j · 1 K DM (14) k i, j b → u / k i+1, j u → b = SE i + 1, j SE i, j · K DA · U TnI (15) Where k i, j u → b, k i, j b → u are the transition rates from unblocked to blocked, and blocked to unblocked, respectively, for the state with i unblocked RU’s. The ratio k i, j b → u / k i+1, j u → b follows naturally from the ratio of Boltzmann terms between the states, requiring only the addition of the probability that TnI is free to bind to actin P (TnI free∣RU unblocked) = TnI/U. The framework based on Boltzmann’s law can be used to determine the ratios of transition rates, but does not result in an absolute on- and off-rate. To determine how the energy difference influences the on- and off-rates, we use a similar approach as proposed by Campbell et al. [38]. Given two states S1, S2 with energies E1, E2 and Boltzmann terms B 1 = e − E 1 k B T, B 2 = e − E 2 k B T, the transition rates between them are given by: k S 1 → S 2 / k S 2 → S 1 = B 2 / B 1 (16) k S 1 → S 2 = (B 2 / B 1) r (17) k S 2 → S 1 = (B 2 / B 1) - (1 - r) (18) The parameter r represents how strongly the on-rate and off-rate depend on the difference in energy between the states E1/E2, ranging from only the off-rate (r = 0) to only the on-rate (r = 1). We apply this model to the effect of tropomyosin deformation on the rates k i, j b → u, k i, j u → b, with the assumption that both rates are equally affected (r = 0. 5). The affinity of TnI for actin KDA is handled separately from the influence of tropomyosin, and is split into two rate constants kA-, kA+. In addition, as our state TmXBi, j is a combination of several sub-states with i unblocked RU’s, what remains is to take into account is the difference in the number of transitions in the average state, given by (n − i) potential RU’s for a blocked-to-unblocked transition and i potential RU’s for a unblocked-to-blocked transition, for any state. Combined, these considerations result in the following equations for the rate constants: ki, jb→u=kA-· (n−i) · (SEi+1, j/ (n−i) SEi, j/ (i+1) ) r (19) ki, ju→b=kA+·i· (SEi, j/ (n−i+1) SEi−1, j/i) − (1−r) ·TnIU (20) For crossbridge binding and unbinding rates we introduce a parameter q to represent the effect of tropomyosin deformation on the crossbridge binding and unbinding rate. Two different choices will be considered. Firstly, tropomyosin deformation affecting both on- and off-rate equally (q = 0. 5) similar to RU unblocking. Secondly, the choice of a constant unbinding rate (q = 1), where only the on-rate is affected by tropomyosin deformation. The impact of this choice will be considered in the next section. Taking into account the number of potential crossbridges, equations for the transition rates are given by: ki, jxb+=kM+· (m−j) · (SEi, j+1/ (m−j) SEi, j/ (j+1) ) q (21) ki, jxb-=kM-·j· (SEi, j/ (m−j+1) SEj−1, i/j) − (1−q) (22) We model TnC and TnI kinetics using simplified global state variables for the blocked and unblocked regulatory units. The equations for these kinetics are given by: dTnITnC dt = k I+ TnC U - k I- TnITnC (23) dTnCB dt = k C+[Ca2+] (B - TnC B) - k C- TnC B - J bu + J ub (24) dTnCU dt = k C+[Ca2+] (U - TnC U - TnITnC) - k C- TnC U - dTnITnC dt + J bu - J ub (25) These equations represent standard Michaelis-Menten kinetics, apart from the terms Jbu and Jub, which represent the ‘flux’ of Ca2+ bound to TnC between the global TnC buffers for blocked and unblocked RU’s. Each time an RU blocks or unblocks, we need to consider the probability of a calcium ion moving between TnCB and TnCU. These fluxes are given by: J bu = 1 n TnC B B ∑ j = 0 m ∑ i = 0 n - 1 k i, j b → u TmXB i, j (26) J ub = 1 n TnC U TnI ∑ j = 0 m ∑ i = 1 n k i, j u → b TmXB i, j (27) The sums represent the total transition rates between blocked and unblocked RU’s from Eq 13, multiplied by 1/n to represent one RU out of n changing for each transition. This is multiplied by the probability of a calcium ion being present on a closing or blocking RU, which is TnC B B for blocked units and TnC U TnI for unblocked units as this latter probability needs to be considered over unblocked RU’s which do not have TnI bound to TnC·Ca2+ (and k i, j u → b already contains a factor TnI/U). For a numerical implementation, quantities such as TnI/U (in Eq 20) should be calculated as TnI/ max (U, ɛ) for a small constant ɛ to avoid undefined 0/0 quantities. Secondly, many of the states TmXBi, j are not populated in practice and can be removed from the formulation for improved efficiency and stability. Specifically, for every i we include states TmXBi, j up to the value of j for which SEi, j < 10−6∑i SEi, j, which reduces the number of states from 1794 to ∼ 750 without significantly affecting the solution. The model’s initial condition should be determined by pacing for a specific calcium transient and parametrization, starting from the completely de-activated state (all state variables set to 0 except TmXB0,0 = 1). In a physiological setting, cooperativity is a key component of normal activation and relaxation of the heart. Thus, an important test of the effectiveness of a model in reproducing physiological cooperativity is its ability to reproduce realistic tension based on experimentally measured calcium transients. In this section, we investigate if tension development across different species is consistent with identical cooperativity, despite significant differences in heart rate. Although the equations in the dynamic model appear to have a large number of free parameters, all transition rates follow from the 9 parameters listed in Table 3. We first parametrize our model to reproduce twitch tension at 37°C in mouse, as we have found this the most challenging test case in practice, and start with the q = 0. 5 case. Active tension was calculated by setting the maximal tension developed to 120 kPa as in previous work [36] (see S1 Text for details). We vary kA-, kI-, kM-, kC+ between 0. 01/ms and 100/ms, and found that tension development and relaxation are all sensitive to the choice of these parameters. As kA-, kI- are generally not thought to be rate-limiting [56], we set both of them to 10/ms. Parameters kM-, kC+ are then determined by requirements for time to peak tension and relaxation times according to the ranges of experimental measurements determined in previous work [36] and summarized in Table 4, resulting in kC+ = kM- = 0. 5/ms. Based on this initial parametrization, we investigate dynamic function of the model across species, applying it to both tension generation in isometric twitches and tension redevelopment. Firstly, we parameterize the model for three different species as well as both choices of crossbridge binding rates (q = 1 or q = 0. 5). Troponin C is a highly conserved protein and its kinetics are not sensitive to temperature or (mammalian) species [56] while crossbridge cycling rates are highly dependent on species, temperature and myosin heavy chain isoform. However, despite the lack of variation in TnC properties, calcium transients vary between the different species while a similar peak isometric tension of approximately 40 kPa is required to be consistent with whole organ contraction across different species [36,70]. Inspired by these observations we use our model to test the hypothesis that contraction across different mammalian species (mouse, rat and human) can be reproduced using only differences in TnI-TnC affinity as given by KDI and the rate of crossbridge kinetics as given by kM-. We determined these two parameters by using a two-dimensional parameter sweep which shows the influence of different constraints on both parameters. Constraints used include species-specific constraints for time to peak tension and relaxation times, and constraints on minimum force < 1 kPa and maximum force between 35–45 kPa in all species. Table 4 shows the parametrization for the three different species [57–64], and Fig 7 shows the corresponding force transients. We were able to capture the different twitch kinetics between species with variations in only KDI and kM-. The resulting parameters show that changes in crossbridge kinetics are consistent with differences in heart rate (mouse > rat > human), while changes to the parameter KDI for TnI affinity for TnC⋅Ca2+ correspond to differences in the calcium transients (c. f. Fig 7B). With respect to the choice of crossbridge unbinding rate on the thin filament state (q = 1 or q = 0. 5), the parameter KDI could be kept the same for the different choices although it was not fixed a priori. However, kM- needs to be significantly higher when constant unbinding rates are used. Overall, model results suggest contractile function across these different species are consistent with a common mechanism and kinetics for thin filament cooperative activation. The rate of tension redevelopment (ktr) has been shown to vary significantly depending on [Ca2+], with experiments showing a range of 4–8× between the lowest and highest rates observed [55,71–74]. These differences in tension redevelopment rates have been linked to the effect of thin filament activation kinetics [55], and is often not well reproduced by computational models. To test our model’s ability to replicate and explain these more complex dynamic effects which involve multiple contractile proteins, and isolate the role of cooperativity in tension redevelopment, we have performed simulations of the calcium dependence of the redevelopment rate of tension. As our model does not account for dynamic length changes, we apply the following procedure to simulate the ktr protocol. For each calcium level the model is run to steady-state, and 50% of crossbridges are instantly detached to simulate the state of the filament after a rapid shorten-relengthen protocol (c. f. trace in [71]). Crossbridges are detached by setting TmXBi, j/2 to TmXBi, j for even j, and TmXBi, ⌈j/2⌉ and TmXBi, ⌊j/2⌋ to TmXB i, j 2 for odd j, where ⌈⋅⌉, ⌊⋅⌋ denote rounding up and down, respectively. Subsequently the model is run with normal binding rates to simulate tension redevelopment, and mono-exponential curve is fitted to the resulting force: F F max = 1 - a e - k tr t (28) To investigate the importance of XB-RU cooperative activation, this procedure is repeated with disabled TnI-actin dynamics (kA- = kA+ = 0) during tension redevelopment to prevent RU (un) blocking. Results in Fig 8 show that a constant unbinding rate independent of tropomyosin deformation (q = 1) shows a larger difference between the minimal ktr and the value at high Ca2+, with around a 6. 5× difference in mouse and rat. Interestingly, unlike experimental data, our results show a clear minimum rather than the typical monotonically increasing ktr. We attribute this effect to the smaller range of [Ca2+] investigated in experiments and the tendency for noise to dominate measurements at low force levels. For a variable unbinding rate, this ratio is only around 3–4 in mouse and rat. In addition, ktr rises steeply at low [Ca2+], as the unbinding rate becomes very fast. In both cases, transient blocking of RU’s is the main mechanism leading to the minimum in ktr. At very low and high [Ca2+], the state of the thin filament is not much affected by crossbridges, being mostly blocked or unblocked regardless. This can be seen by the similar ktr for results with and without RU (un) blocking disabled. However, at intermediate [Ca2+], unblocking of crossbridges will cause RU’s to start moving to the ‘blocked’ position, requiring additional time to be re-activated by XB-RU interaction resulting in the lower ktr. In this paper we developed a novel model of cardiac contraction and used it to investigate the effects of different cooperative mechanisms on both the steady-state and dynamic behaviour of cardiac muscle. We have been able to show the relative importance of the different potential mechanisms for generating the steeply cooperative force-calcium relationship in muscle. Firstly, RU-RU cooperativity is the clear dominant mechanism for cooperativity near and above Ca50, caused by progressively easier unblocking for any individual RU as more of their close neighbours are unblocked. This can be seen from the cooperativity of the RU unblocking in the absence of crossbridges (Fig 4) and RU activation (Fig 5), with moderate cooperativity which is not significantly biphasic. Although RU-RU cooperativity deriving from end-to-end interactions between tropomyosin is a major part of cooperative activation, it does not explain all of the cooperative activation seen. Secondly, cooperativity between crossbridges, i. e. of the ‘XB-XB’ type, approximately doubles the number of crossbridges bound at maximal activation (Fig 4), and in doing so increases the effects of XB-RU interactions, but does not in itself significantly increase the steepness of the force-calcium relationship. Thirdly, we have shown that activation of regulatory units by crossbridges, i. e. ‘XB-RU’ cooperativity, is important for determining calcium sensitivity. Our model reproduces the effects of changes in the affinity of myosin XB on calcium sensitivity shown experimentally [1,16,54] (Fig 6). Despite differences in experimental conditions due to temperature and permeabilization of muscle, there is good quantitative agreement in the shift in calcium sensitivity (ΔpCa50, Table 2). Our results show a mild decrease in cooperativity for simulations with decreased crossbridge affinity (nH = 5. 1 → 4. 2) which is almost entirely in the lower half (below Ca50) of the force-calcium relationship (n2 = 7. 5 → 6. 1), with nearly identical values for the upper half (n1 = 2. 7 → 2. 4). The modest change in Hill coefficient may explain some of the contradictory results in the literature so far, as Hill fits are inherently sensitive to the choice of fitting method and calcium concentrations used, in addition to experimental noise. The ratio of n2/n1 varies between 2. 5–2. 7 in these simulations, compared to approximately 2. 0 in experiments [52,53]. This ratio is very sensitive to the choice of fitting methods and [Ca2+] window used for fitting, which could explain these differences. Alternatively, these differences could be the result of the simplified sarcomere geometry in the current model. In addition, our model reproduces the activation of muscle by high-affinity ‘rigor’ crossbridges, such as in conditions of low ATP [45] or special experimental preparations with NEM-S1 myosin [75], even in the complete absence of Ca2+. Although results from these experiments are less important for contraction models to reproduce as they represent conditions far from physiological, they are a direct result of XB-RU cooperativity in our model and thus increases confidence in the choice of biophysical framework. In addition this is a novel feature for computational models, as in the vast majority of models in this area RU’s are mathematically unable to activate in the absence of Ca2+ [34–36,38]. Our model also explains effects of force on Ca2+-TnC binding. The proposed effects of crossbridges on Ca2+-TnC affinity appear in our model as an emergent property of the competitive binding of TnI and pinning of tropomyosin by both TnI and crossbridges. Specifically, crossbridge binding prevents tropomyosin to move to the blocked position, and this effectively prevents TnI from binding to tropomyosin-actin. This in turn increases the relative time TnI is bound to TnC⋅Ca2+, thus increasing the effective affinity of TnC for Ca2+, as Ca2+ is highly unlikely to dissociate from TnC when TnI is bound. As a result, our framework represents the observed tension-dependent feedback on TnC affinity without hysteresis in the steady-state force-pCa relationship. Finally, we have identified an important effect by which approximately five neighbouring unblocked tropomyosin units are required to significantly bind myosin (Fig 5). These effects are also partly responsible for the very steep force-calcium relationship in the region below Ca50 and the strongly biphasic shape of the force-calcium curve. Specifically, it allows for near-zero force in a state with a significant fraction of unblocked RU’s with Ca2+ bound to TnC, as there are not enough consecutive unblocked RU’s to generate significant force. Achieving low force at physiological diastolic [Ca2+] of approximately 0. 1–0. 2 μM is particularly important for effective diastolic filling in the heart. In addition this result shows that the common modelling assumptions of crossbridge binding properties being linearly proportional to the number of unblocked RU’s is potentially inaccurate. Thus, cooperative activation of muscle derives in part from the end-to-end interactions of tropomyosin, with additional cooperativity mostly below Ca50 driven by XB-RU effects and nonlinear crossbridge binding properties. Building on the steady-state models, we developed different models for dynamic muscle function in mouse, rat and human, to investigate the ability of our model of cooperative activation in reproducing physiological activation and relaxation of muscle. We have parameterized the model to these three different species using only changes in the affinity of TnI for TnC (KDI) and crossbridge cycling rates (kM-). Our results in Table 4 and Fig 7 show it is possible to reproduce the different kinetics and accommodate large changes in Ca2+ transients without changing the properties of TnC which are generally thought to be highly conserved. Our results are consistent with a highly conserved underlying mechanism for cooperative activation, while the relative order of crossbridge cycling rates is highly species dependent as expected from the differences in heart rates, with mouse faster than rat, and rat faster than human. The model also naturally reproduces the Ca2+-dependence of the rate of force redevelopment ktr, and is able to explain the strong Ca2+ dependence of force redevelopment as a result of transient blocking of RU’s (Fig 8). Specifically, detachment of crossbridges causes a steady-state activation that is lower (similar to the blebbistatin experiments in the steady-state models), and results in blocking of RU’s due to a decrease in XB-RU effects. At high and low Ca2+, the filament state is mostly blocked or unblocked, and is not much affected by the number of crossbridges. However, near Ca50, this state is particularly sensitive to crossbridges and XB-RU effects. As crossbridges are detached by the fast length change, some RU’s move to a blocked position, and require additional time to become unblocked again through XB-RU cooperative interactions. This causes a higher ktr compared to that seen at high Ca2+ concentrations. The choice of constant or tropomyosin-deformation dependent crossbridge unbinding rates significantly affects our results for ktr. Firstly, the Ca2+-dependence is stronger for the case with constant unbinding rate, and better represents experimental observations of 4–8 fold changes, which suggests that the choice of a constant crossbridge unbinding rate may be the more physiological one. However, reproducing physiological relaxation rates requires particularly high unbinding rates (≈ 2/ms in mouse). A potential explanation for this is that velocity dependent effects are important even in isometric tension relaxation due to internal sarcomere shortening [76]. Secondly, the behaviour at low Ca2+ is different, with the variable unbinding rate showing increasing ktr. This can be explained by considering the high energy barrier for a single crossbridge binding to a fully blocked thin filament. Experimental results rarely show this region, most likely due to the experimental noise dominating the near-zero force. Regardless, our results and interpretation suggest a Hill fit as applied previously (e. g. [55]) may not be a suitable choice for these results, as the Ca2+-ktr curves show a clear theoretical minimum. The current model is designed to investigate the protein-protein interactions responsible for cooperativity. As a computational model can not capture every detail, we have made a range of assumptions to make the model computationally tractable, such as the regular spacing of fixed myosin crossbridges, homogeneity of tropomyosin properties, the choice of two global TnC buffers, and several techniques for reducing the number of states. For a general model of contraction, the most important limitations in the current formulation of the model is the absence of velocity- and length-dependent effects. The current model uses a very simplified sarcomere geometry with regularly spaced crossbridges which are all capable of binding to actin, and does not include the effects of filament overlap which prevents crossbridges from binding to specific regions on the thin filament, even at resting length. Extending the model to represent length-dependence of tension would require an accurate model of filament overlap and other details of sarcomere geometry. This in turn would require changes to the current Monte Carlo sampling strategies and representative state approach, as currently equivalent states would have different crossbridge binding properties depending on the location of unblocked units on the thin filament. However, with respect to the effects of length-dependent activation, the explicit representation of troponin I in our model makes it especially suitable for testing current hypotheses relating to length-dependent modulation of troponin I [77–79]. The addition of velocity dependence would allow us to investigate the interaction between velocity-dependent unbinding and XB-RU effects in ventricular relaxation. Although spatially detailed velocity dependence would likely be too computationally demanding for a practical model, the current dynamic model framework is suitable for extensions using averaged strain-distortion approaches [80,81]. To limit model complexity, the current model uses simple two-state crossbridge kinetics in which the transitions which affect tropomyosin deformation are considered to be most relevant. Another possible extension of the model is to incorporate more detailed crossbridge dynamics, using a 3-state (weak, strong, unbound) [35] or more complex model which incorporates ATP/ADP/Pi kinetics [82]. Such extensions are not expected to affect the steady-state results, as our current model can be interpreted as a steady-state approximation of a more complex modelling framework. These extensions could significantly affect dynamic model results and would also allow for more detailed velocity-dependent binding rates between different crossbridge states. Extensions with ATP kinetics would also allow for more quantitative comparison with experimental data showing ATP-dependence of force, as shown in the rigor crossbridge tests in Fig 6B. However, spatially detailed effects would have to be simplified for an extended dynamic model to keep such an extended model tractable, using techniques such as those applied for blocked and unblocked TnC states (TnCU, TnCB) in the current model. Finally, the model currently consists of 750 ODEs, which is a relatively high number compared to many phenomenological contraction models currently available. Models which explicitly represent interactions between RU’s range in complexity from a few ODEs to several thousands. For example, the work by Razumova et al. uses the assumptions of a uniform spatial distribution of RU states, which allows them to reproduce cooperative effects with a lower number of ODEs [83]. However, cooperative activation of RU’s breaks this assumption as unblocked units tend to cluster together. Furthermore, states with identical number of unblocked RU’s can have very different crossbridge binding properties (see Fig 2). More recent work from this group no longer assumes a uniform spatial distribution, but requires several thousand ODEs to represent 9 RU’s [38], as does the model by Dobrunz et al. which takes a similar approach [84]. Although a high number of ODEs has been used before in whole organ models [39], and is thus not an immediate computational problem, it limits their portability and ease of use. Thus, model reduction strategies will form an important part of future work. A potential advantage of our modelling approach is its basis in an underlying biophysical model of the tropomyosin chain. This approach allows for longer-range interactions than these previous ODE models including only nearest-neighbour interaction, and results in fewer free parameters compared to approaches where RU-RU, XB-RU and XB-XB effects are defined by independent parameters. A potential disadvantage of our model is the requirement for a pre-calculation using Monte-Carlo sampling, which is computationally costly. To increase the models usability, an implementation including the results of the Monte Carlo sampling procedures are made available online at cemrg. co. uk. Future work in both model reduction and deformation-dependence will further increase the usability of this biophysically detailed model in a whole-organ setting, and improve the predictive power of multi-scale cardiac models in biological and clinical applications.

Force generation in cardiac muscle cells is driven by changes in calcium concentration. Relatively small changes in the calcium concentration over the course of a heart beat lead to the large changes in force required to fully contract and relax the heart. This is known as ‘cooperative activation’, and involves a complex interaction of several proteins involved in contraction. Current computer models which reproduce force generation often do not represent these processes explicitly, and stochastic approaches that do tend to require large amounts of computational power to solve, which limit the range of investigations in which they can be used. We have created an new computational model that captures the underlying physiological processes in more detail, and is more efficient than stochastic approaches, while still being able to run a large range of simulations. The model is able to explain the biological processes leading to the cooperative activation of muscle. In addition, the model reproduces how this cooperative activation translates to normal muscle function to generate force from changes in calcium across three different species.

Abstract Introduction Models Results Discussion

2015

A Spatially Detailed Model of Isometric Contraction Based on Competitive Binding of Troponin I Explains Cooperative Interactions between Tropomyosin and Crossbridges

The genus Henipavirus in the family Paramyxoviridae contains two viruses, Hendra virus (HeV) and Nipah virus (NiV) for which pteropid bats act as the main natural reservoir. Each virus also causes serious and commonly lethal infection of people as well as various species of domestic animals, however little is known about the associated mechanisms of pathogenesis. Here, we report the isolation and characterization of a new paramyxovirus from pteropid bats, Cedar virus (CedPV), which shares significant features with the known henipaviruses. The genome size (18,162 nt) and organization of CedPV is very similar to that of HeV and NiV; its nucleocapsid protein displays antigenic cross-reactivity with henipaviruses; and it uses the same receptor molecule (ephrin- B2) for entry during infection. Preliminary challenge studies with CedPV in ferrets and guinea pigs, both susceptible to infection and disease with known henipaviruses, confirmed virus replication and production of neutralizing antibodies although clinical disease was not observed. In this context, it is interesting to note that the major genetic difference between CedPV and HeV or NiV lies within the coding strategy of the P gene, which is known to play an important role in evading the host innate immune system. Unlike HeV, NiV, and almost all known paramyxoviruses, the CedPV P gene lacks both RNA editing and also the coding capacity for the highly conserved V protein. Preliminary study indicated that CedPV infection of human cells induces a more robust IFN-β response than HeV. Henipaviruses were first discovered in the 1990s following investigation of serious disease outbreaks in horses, pigs and humans in Australia and Malaysia [1], [2] and comprise the only known Biosafety Level 4 (BSL4) agents in the family Paramyxoviridae [3]. Depending upon the geographic locations of outbreaks, and the virus and animal species involved, case mortality is between 40% to 100% in both humans and animals [4], [5], making them one of the most deadly group of viruses known to infect humans. The genus Henipavirus in the subfamily Paramyxovirinae currently contains two members, Hendra virus (HeV) and Nipah virus (NiV) [6]. Fruit bats in the genus Pteropus, commonly known as flying foxes, have been identified as the main natural reservoir of both viruses although serological evidence suggests that henipaviruses also circulate in non-pteropid bats [7], [8], [9], [10]. The discovery of henipaviruses had a significant impact on our understanding of genetic diversity, virus evolution and host range of paramyxoviruses. Paramyxoviruses, such as measles virus and canine distemper virus, were traditionally considered to have a narrow host range and to be genetically stable with a close to uniform genome size shared by all members of Paramyxovirinae [3]. Henipaviruses shifted this paradigm on both counts having a much wider host range and a significantly larger genome [6]. Identification of bats as the natural reservoir of henipaviruses also played an important role in significantly increasing international scientific attention on bats as an important reservoir of zoonotic viruses, including Ebola, Marburg, SARS and Melaka viruses [11], [12], [13], [14]. Since the discovery of the first henipavirus in 1994, much progress has been made in henipavirus research, from identification of functional cellular receptors to the development of novel diagnostics, vaccine and therapeutics [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25]. By contrast, there is little understanding of the pathogenesis of these highly lethal viruses. This is due in part to the requirement of a high security BSL4 facility for any live infection studies and in part to the limited range of research tools and reagents for the current small animal models. Research into the mechanisms of henipavirus pathogenesis is also hampered by the lack of related, but non-pathogenic or less pathogenic viruses, thus preventing targeted comparative pathogenetic studies. Early serological investigations in Australia and more recent studies in other regions (e. g. , China) indicated the presence of cross-reactive, but not cross-neutralizing, antibodies to henipaviruses in bats of different species [8]. These findings were further supported by the detection of henipavirus-like genomic sequences in African bats [26]. Discovery and isolation of these related viruses will be highly important to our further understanding of henipavirus evolution, mechanism of cross-species transmission, and pathogenesis in different animal species. Here we report the isolation and characterization of a new bat henipavirus which, based on preliminary infection studies, is non-pathogenic in two of the small animal infection models currently used in henipavirus research. We believe that this new virus will provide a powerful tool to facilitate our future study into different aspects of henipaviruses, especially in the less advanced area of pathogenesis. As part of our on-going field studies on HeV genetic diversity and infection dynamics in the Queensland flying fox populations, urine samples were collected on a regular basis for PCR and virus isolation. Since the establishment of the Pteropus alecto primary cell lines in our group [27], we have intensified our effort to isolate live virus from these urine samples by routinely inoculating separate primary cell lines derived from kidney, spleen, brain, and placenta, as well as Vero cells. Syncytial CPE was observed in kidney cell (PaKi) monolayers 5 days post inoculation (dpi) with two different urine samples (Fig. S1) collected in September 2009 from a flying fox colony in Cedar Grove, South East Queensland (see Fig. S2 for map location). No CPE was observed in any of the four other cell lines. Supernatant harvested 6 dpi was used to inoculate fresh PaKi cell monolayers. After two passages in PaKi cells, the virus was able to infect and cause CPE in Vero cells. However, the CPE morphology of CedPV infection in Vero cells was different from that of HeV infection. Further analysis using HeV-specific PCR primers indicated that the new bat virus was not an isolate of HeV. Considering the formation of syncytial CPE by this new virus and the previous success in isolating paramyxoviruses from bat urine [28], [29], [30], paramyxovirus family-specific and genus-specific primers were used to determine whether this new virus was a member of the family Paramyxoviridae. Positive PCR fragments of the expected sizes were obtained from the Paramyxovirinae and Respirovirus/Morbillivirus/Henipavirus primer sets developed by Tong et al [31]. Sequencing of the PCR products indicated that it was a new paramyxovirus most closely related to HeV and NiV. Based on these preliminary data, the virus was named Cedar virus (CedPV) after the location of the bat colony sampled. Full length genome sequence was determined by a combination of three different approaches, random deep sequencing using 454 technology, sequencing of PCR products obtained using degenerate primers designed based on known henipaviruses, and RACE to determine the precise genome terminal sequences. As shown in Fig. 1, the genome of CedPV is 18,162 nt in length most similar to that of HeV in the family. The full genome sequence has been deposited to GenBank (Accession No. JQ001776). The genome size is a multiple of 6, hence abiding by the Rule-of-Six observed for all known members of the subfamily Paramyxovirinae [3]. It has a 3-nt intergenic sequence of CTT absolutely conserved at all seven positions and highly conserved gene start and stop signals similar to those present in HeV and NiV (Fig. S3). Also similar to the HeV genome is the presence of relatively large non-coding regions in the CedPV genome (Fig. 1 and Table 1). The overall protein-coding capacity of the CedPV genome is 87. 41% which is significantly lower than the average of 92. 00% for other family members but higher than HeV at 82. 12%. As the genome size of CedPV and HeV is very similar, the increased coding capacity of CedPV is attributed to an increase in protein sizes for five of the six major proteins, with the L protein being 257-aa larger (Table 1). At 2,501 aa, the CedPV L protein is the largest, not only in the family Paramyxoviridae but also for all known viruses in the order Mononegavirale. Phylogenetic analysis based on the full length genome sequence and the deduced amino acid sequences of each structural protein confirmed the initial observation that CedPV is most closely related to henipaviruses in the family. A phylogenetic tree based on the deduced sequences of the nucleocapsid protein (N) is presented in Fig. 2. Phylogenetic tree based on whole genome sequences gave similar results (Fig. S4). CedPV is more closely related to HeV and NiV than henipavirus-like sequences detected in African bats [26], [32] as shown in a phylogenetic tree based on the only sequences available of a 550-nt L gene fragment (Fig. S5). First discovered for the parainfluenza virus 5 (PIV5, previously known as simian virus 5), almost all members of Paramyxovirinae have a P gene which produces multiple proteins through an RNA editing mechanism by addition of non-templated G residues leading to production of N-terminal co-linear proteins from different reading frames downstream from the editing site [3], [33]. These multiple gene products are known to play a key role in antagonizing the innate response of susceptible hosts [3]. A search of CedPV for open reading frames (ORF) in the P gene revealed a 737-aa P protein and a 177-aa C protein, but failed to find the highly conserved, cysteine-rich V ORF present in most other paramyxoviruses. The RNA editing site with the sequence of AAAAGGG, which is absolutely conserved in all known HeV and NiV isolates discovered to date, is also missing from the CedPV P gene sequence. To further verify that there are no multiple mRNAs produced from the CedPV P gene, direct sequencing of P gene transcripts was conducted from CedPV-infected Vero cells using multiple sets of primers generating overlapping fragments covering the entire coding region of the P gene. Each produced uniform trace files indicating a lack of RNA editing activities, which is very different from the mixed peaks generated by HeV and NiV immediately after the editing site (Fig. S7). To our knowledge, CedPV is the first member of Paramyxovirinae that lacks both RNA editing and any V-related coding sequence in its P gene. Further investigation is required to exclude the possibility that the P-gene editing in CedPV is cell- or tissue-specific and not present or present at an extremely low level in the current virus-cell system. The striking similarity in genome size and organization and the presence of highly conserved protein domains among the N, M and L proteins between CedPV and henipaviruses prompted us to investigate the antigenic relatedness of these viruses. Staining of CedPV- infected Vero cells using rabbit anti-henipavirus antibodies indicated the presence of cross-reactivity. This cross-reactivity was further confirmed in reverse by staining of HeV-infected Vero cells using a rabbit serum raised against a recombinant CedPV N protein (Fig. 3). However, analysis by virus neutralization test using either polyclonal or monoclonal antibodies found that henipavirus-neutralizing antibodies were unable to neutralize CedPV. Similarly, CedPV-neutralizing antibodies obtained in our infection studies (see below) also failed to neutralize either HeV or NiV. It can therefore be concluded that CedPV and henipaviruses share cross-reactive antigenic regions, but not cross-neutralizing epitopes. To further investigate the relationship between CedPV and recognized henipaviruses, we investigated the use of the henipavirus receptors, the ephrin-B2 and -B3 host cell proteins, as potential receptors for CedPV infection. Our previous studies have demonstrated that the ephrin-B2 and -B3 expression negative HeLa-USU cell line could support henipavirus infection and formation of syncytial CPE only when either the ephrin-B2 or -B3 gene was transiently expressed in the cells [22], [34]. For CedPV, similar observations were made with respect to the ephrin-B2 receptor. As shown in Fig. 4, CedPV failed to infect HeLa-USU, but was able to infect and cause syncytial CPE when the human ephrin-B2 gene was expressed. In contrast, when ephrin-B3 molecule was introduced, there was no evidence of infection. Ferrets, guinea pigs, and mice exhibit differing responses to the previously described henipaviruses HeV and NiV, with ferrets and guinea pigs, but not mice developing severe disease characterized by systemic vasculitis [20], [35], [36], [37], [38]. In contrast, ferrets and guinea pigs exposed to CedPV by, respectively, oronasal and intraperitoneal routes remained clinically well although neutralizing antibody was detected in serum between 10 to 21 days pi (Table 2). Balb-C mice exposed to CedPV by the oronasal route remained clinically well and did not develop neutralizing antibody in serum by day 21 pi. In ferrets electively euthanized at earlier time-points, there was reactive hyperplasia of tonsillar lymphoid tissue, retropharyngeal and bronchial lymph nodes, accompanied by edema and erythrophagocytosis. CedPV antigen was detected in bronchial lymph node of one animal euthanized on day 6 pi, consistent with viral replication in that tissue; cross-reactive immunostaining against anti-NiV N protein antibodies was also noted (Fig. 5). No other significant histological lesions were identified. Viral RNA was detected in selected lymphoid tissues of 3 (of 4) ferrets sampled day 6 to 8 pi, including pharynx, spleen, and retropharyngeal and bronchial lymph nodes, as well as the submandibular lymph node of the ferret euthanized on day 20 pi. This pattern of lymphoid involvement suggests that there may be transient replication in the upper and lower respiratory tracts although CedPV genome was not recovered from nasal washes, oral swabs, pharynx or lung tissue of affected animals. Virus isolation was unsuccessful for all PCR positive tissues. As a first step towards the understanding of the pathogenicity difference between CedPV and HeV, we examined the IFN responses in human HeLa cells upon virus infection. As shown in Fig. 6, while the induction of IFN-α was similar in cells infected with HeV or CedPV, there was a significant difference of IFN-β production upon infection by HeV or CedPV, with CedPV-infected cell producing a much higher level of IFN-β. To investigate the CedPV exposure status of pteropid bats in Queensland and potential co-infection (either concurrent or consecutive) of CedPV with HeV, we tested 100 flying fox sera collected previously for other studies for antibody against the two viruses. Due to the cross-reactivity observed above, virus neutralization tests were conducted to obtain more accurate infection data for each virus. Overall, 23% of the sera were CedPV-positive and 37% HeV-positive (Table S1). Co-infection was reflected in 8% of the sera tested. The emergence of bat-borne zoonotic viruses (including HeV, NiV, Ebola, Marburg, and SARS) has had a significant impact on public health and the global economy during the past few decades. With the rapidly expanding knowledge of virus diversity in bat populations around the world, it is predicted that more bat-borne zoonotic viruses are likely to emerge in the future. The discovery of a novel ebolavirus-like filovirus in Spanish microbats demonstrates that the potential for such spill over events is not limited to Africa or Asia [39]. It is therefore important to enhance our preparedness to counter future outbreaks by conducting active pre-emergence research into surveillance, triggers for cross-species transmission, and the science of identification of potential pathogens. Henipaviruses represent one of the most important bat-borne pathogens to be discovered in recent history. Although CedPV displays some differences from existing members of the genus Henipavirus, we propose that CedPV be classified as a new henipavirus based on the following shared features with known henipaviruses: 1) it is antigenically related to current henipaviruses; 2) its genome size and organization is almost identical to those of HeV and NiV; 3) it has a similar prevalence in flying foxes; and 4) it uses ephrin-B2 as the cell entry receptor. The lack of cross-neutralization between CedPV and HeV or NiV was not unexpected from the comparative sequence analysis of all the deduced proteins, especially the G protein (see Table 1). It is clear that the genetic relatedness of CedPV with HeV or NiV is much lower than between HeV and NiV. However, the percentage sequence identities of the major viral proteins between CedPV and HeV/NiV are on average at least 10% higher than that between HeV/NiV and any other known paramyxoviruses. Also, there was no antigenic cross-reactivity observed between CedPV and representative viruses of the other paramyxovirus genera in the subfamily Paramyxovirinae (Fig. S6). Like other paramyxoviruses, the P gene of henipaviruses produces multiple proteins which play a key role in viral evasion of host innate immune responses [4], [40], [41]. One of these is the Cys-rich V protein: all members of the subfamily Paramyxovirinae produce the V protein with the exception of the human parainfluenza virus 1 (hPIV1). Although a putative RNA editing sequence (AAGAGGG) is present at the expected editing site of the P gene, the hPIV1 RNA polymerase does not produce an edited mRNA of the P gene [42]. There are remnants of the V ORF easily detectable in the hPIV1 P gene although the predicted 68-aa ORF region is interrupted by multiple in-frame stop codons. Of the 7 Cys residues conserved between bovine parainfluenza virus 3 and Sendai virus, four are still present in the non-functional V ORF of hPIV1[42]. In contrast, an extensive ORF and sequence homology search of the CedPV P gene only identified one aa coding region with minimal sequence identity to the V ORFs of HeV and NiV (see Fig. S8). In this region, out of the 9 Cys residues conserved between HeV and NiV V proteins, only 2 are present in the CedPV P gene. Furthermore, the sequence (AGATGAG) upstream from this putative ORF V coding region does not match the consensus RNA editing site. It can therefore be concluded that CedPV is the only member of Paramyxovirinae which lacks both the functional V mRNA/protein and the coding capacity for the RNA editing site and ORF V. The evolutionary significance of this finding needs further investigation. Our in vitro study indicated that ephrin B2, but not ephrin B3, was able to restore CedPV infection in the ephrin B2-deficient HeLa cells. While this is highly suggestive that ephrin B2 is the functional entry receptor for CedPV, it should be emphasized that this was not a direct proof that ephrin B2 is the receptor. Further investigation is required to confirm this. In our preliminary studies, it was shown that CedPV was able to replicate in guinea pigs and ferrets, but failed to cause significant clinical diseases, unlike that of the closely related HeV and NiV. These first infection experiments were conducted with a high dose if virus to establish whether the CedPV could replicate in these animals and determine the degree of any clinical disease. A second experiment was then carried out in ferrets to determine the site of replication and tissue tropism in sequentially sacrificed animals. A lower dose was used to gain better comparison with similar infection experiments using HeV and NiV [18], [35]. Although these initial experimental infection studies indicate that CedPV is less or non-pathogenic in these species, it is possible that CedPV may be pathogenic in other hosts, such as horses. We hypothesize that the lack of a V protein may impact on the pathogenicity. In this regard, it was encouraging to observe that infection of human cells by CedPV induced a much more robust IFN-β response than HeV. Further study is required to dissect the exact molecular mechanism of this observed difference. Due to the close relationship between CedPV and HeV, it was important to investigate the possibility of co-infection by these two viruses in the Australian bat population. Based on the detection of neutralizing antibodies at 23% for CedPV, 37% for HeV and 8% for both, it can be concluded that the co-infection rate is very close to the theoretical rate of 8. 5% (the product of the two independent infection rates). Based on this limited preliminary analysis, it appears that infection of bats by one henipavirus neither prevents nor enhances the likelihood of infection by the other. In summary, the discovery of another henipavirus in Australian flying foxes highlights the importance of bats as a significant reservoir of potential zoonotic agents and the need to expand our understanding of virus-bat relationships in general. Our future research will be directed at determining whether spill-over of CedPV into other hosts has occurred in the past in Australia, whether CedPV is pathogenic in certain mammalian hosts, and whether CedPV exists in bat populations in geographically diverse regions. Cell lines used this study were Vero (ATCC), HeLa-USU [22], and the P. alecto primary cell lines derived from kidney (PaKi), brain (PaBr), (spleen) PaSp and placenta (PaPl) recently established in our group [27]. Cells were grown in Dulbecco' s Modified Eagle' s Medium Nutrient Mixture F-12 Ham supplemented with double strength antibiotic-antimycotic (Invitrogen), 10 µg/ml ciprofloxacin (MP Biomedicals) and 10% fetal calf serum at 37°C in the presence of 5% CO2. Urine (approximately 0. 5–1 ml) was collected off plastic sheets placed underneath a colony of flying foxes (predominantly Pteropus alecto with some P. Poliocephalus in the mixed population) in Cedar Grove, South East Queensland, Australia and pooled into 2-ml tubes containing 0. 5 ml of viral transport medium (SPGA: a mix of sucrose, phosphate, glutamate and albumin plus penicillin, streptomycin and fungizone). The tubes were temporarily stored on ice after collection and transported to a laboratory in Queensland, frozen at −80°C, and then shipped on dry ice to the CSIRO Australian Animal Health Laboratory (AAHL) in Geelong, Victoria for virus isolation. The samples were thawed at 4°C and centrifuged at 16,000×g for 1 min to pellet debris. Urine in the supernatant (approximately 0. 5–1 ml) was diluted 1∶10 in cell culture media. The diluted urine was then centrifuged at 1,200×g for 5 min and split evenly over Vero, PaKi, PaBr, PaSp and PaPl cell monolayers in 75-cm2 tissue culture flasks. The flasks were rocked for 2 h at 37°C, 14 ml of fresh cell culture media was added and then incubated for 7 d at 37°C. The flasks were observed daily for toxicity, contamination, or viral cytopathic effect (CPE). Cells showing syncytial CPE were screened using published broadly reactive primers [31] for all known paramyxoviruses and a subset of paramyxoviruses. PCR products were gel extracted and cloned into pGEM T-Easy (Promega) to facilitate sequencing using M13 primers. Sequences were obtained and aligned with known paramyxovirus sequences allowing for initial classification. Whole genome sequence was determined using a combination of 454 sequencing [43] and conventional Sanger sequencing. Virions from tissue culture supernatant were collected by centrifugation at 30,000×g for 60 min and resuspended in 140 µl of PBS and mixed with 560 µl of freshly made AVL for RNA extraction using QIAamp Viral RNA mini kit (Qiagen). Synthesis of cDNA and random amplification was conducted using a modification of a published procedure [44]. Briefly, cDNA synthesis was performed using a random octomer-linked to a 17-mer defined primer sequence: (5′-GTTTCCCAGTAGGTCTCNNN NNNNN-3′) and SuperScript III Reverse Transcriptase (Invitrogen). 8 µl of ds-cDNA was amplified in 200 µl PCR reactions with hot-start Taq polymerase enzyme (Promega) and 5′-A*G*C*A*C TGTAGGTTTCCCAGTAGGTCTC-3′ (where * denotes thiol modifications) as amplification primers for 40 cycles of 95°C/1 min, 48°C/1 min, 72°C/1 min after an initial denaturation step of 5 min at 95°C and followed by purification with the QIAquick PCR purification kit (Qiagen). Sample preparation for Roche 454 sequencing (454 Life Sciences Branford, CT, USA) was according to their Titanium series manuals, Rapid Library Preparation and emPCR Lib-L SV. To obtain an accurate CedPV genome sequence, 454 generated data (after removing low quality, ambiguous and adapter sequences) was analysed by both de novo assembly and read mapping of raw reads onto the CedPV draft genome sequence derived from Sanger sequencing. For 454 read mapping, SNPs and DIPs generated with the CLC software were manually assessed for accuracy by visualising the mapped raw reads (random PCR errors are obvious compared to real SNPs and DIPs especially when read coverage is deep). Consensus sequences for both 454 de novo and read mapping assembly methods were then compared to the Sanger sequence with the latter used to resolve conflicts within the low coverage regions as well as to resolve 454 homopolymer errors. Sequences of genome termini were determined by 3′- and 5′-RACE using a protocol previously published by our group [45]. Briefly, approximately 100 ng of RNA was ligated with adaptor DT88 (see reference for sequence information) using T4 RNA ligase (Promega) followed by cDNA synthesis using the SuperScript III RT kit (Invitrogen) and an adaptor-specific primer, DT89. PCR amplification was then carried out using DT89 and one or more genome-specific primers. PCR products were sequenced directly using either DT89 or genome specific primers by an in-house service group on the ABI Sequencer 3100. The CLC Genomics Workbench v4. 5. 1 (CLC Inc, Aarhus, Denmark) was used to trim 454 adapter and cDNA/PCR primer sequences, to remove low quality, ambiguous and small reads <15 bp and to perform de novo and read mapping assemblies all with default parameters. Clone Manager Professional ver 9. 11 (Scientific and Educational Software, Cary, NC, USA) was used to join overlapping contigs generated by de novo assembly. Phylogenetic trees were constructed by using the neighbor-joining algorithm with bootstrap values determined by 1,000 replicates in the MEGA4 software package [46]. Quantitative PCR assays (qPCR) were established based on CedPV-specific sequences obtained from the high throughput sequencing. A TaqMan assay on the P gene was developed and used for all subsequent studies. The sequences of the primer/probe are as follows: forward primer, 5′-TGCAT TGAGC GAACC CATAT AC; reverse primer, 5′-GCACG CTTCT TGACA GAGTT GT; probe, 5′-TCCCG AGAAA CCCTC TGTGT TTGA-MGB. The coding region for the CedPV N protein was amplified by PCR with a pair of primers flanked by AscI (5′ end) and NotI (3′ end) sites for cloning into our previously described GST-fusion expression vector [47]. The expression and purification by gel elution was conducted as previously described [48]. For antibody production, purified protein was injected subcutaneously into 4 different sites of 2 adult (at a dose of 100 µg per animal) New Zealand white female rabbits at days 0 and 27. The CSIRO' s triple adjuvant [49] was used for the immunization. Animals were checked for specific antibodies after days 5 and 42 and euthanized at day 69 for the final blood collection. For immunofluorescence antibody test, Vero cell monolayers were prepared in 8-well chamber slides by seeding at a concentration of 30,000 cells/well in 300 µl of cell media and incubating over night at 37°C. The cell monolayers were infected with an MOI of 0. 01 of CedPV, HeV or NiV and fixed with 100% ice-cold methanol at 24 h post-infection. The chamber slides were blocked with 100 µl/well of 1%BSA in PBS for 30 min at 37°C before adding 50 µl/well of rabbit sera against CedPV N or NiV N diluted 1∶1000. After incubation at 37°C for 30 min, the slides were washed three times in PBS-T and incubated with 50 µl/well of anti-rabbit 488 Alexafluore conjugate (Invitrogen) diluted 1∶1000 at 37°C for 30 min. The slides were then washed three times in PBS-T and mounted in 50% glycerol/PBS for observation under a fluorescence microscope. For virus neutralization test, serial two-fold dilutions of sera were prepared in duplicate in a 96-well tissue culture plate in 50 µl cell media (Minimal Essential Medium containing Earle' s salts and supplemented with 2 mM glutamine, antibiotic-antimycotic and 10% fetal calf serum). An equal volume containing 200 TCID50 of target virus was added and the virus-sera mix incubated for 30 min at 37°C in a humidified 5% CO2 incubator. 100 µl of Vero cell suspension containing 2×105 cells/ml was added and the plate incubated at 37°C in a humidified 5% CO2 incubator. After 4 days, the plate was examined for viral CPE. The highest serum dilution generating complete inhibition of CPE is defined as the final neutralizing titer. Human ephrin B2 and B3 genes were cloned into pQCXIH (Clontech) and the resulting plasmids packaged into retrovirus particles in the GP2–293 packaging cell line (Clontech) and pseudotyped with vesicular stomatitis virus G glycoprotein (VSV-G) following the manufacturer' s instructions. HeLa-USU cell line [22] was infected with the VSV-G pseudotyped retrovirus particles in the presence of 1 µg/ml polybrene (Sigma). 8 h post infection, the medium was changed and the cells were allowed to recover for 24 h, allowing time for the retroviral insert to be incorporated into the cell genome and for expression of the hygromycin resistance gene. 24 h post infection, cells transformed by the retrovirus were selected for by the addition of 200 µg/ml hygromycin in the media. Stocks of cells that were resistant to hygromycin were prepared and frozen. HeLa-USU and ephrin-expressing HeLa-USU cells were seeded in 6-well tissue culture plates at a density of 250,000 cells/well overnight. The viruses (HeV and CedPV) were diluted to give an MOI of 0. 01 and inoculated into the wells. The cell monolayers were examined daily for syncytial CPE. Animal studies were carried out in the BSL4 animal facility at AAHL. Ferrets, guinea pigs and mice were used on the basis of their known and varying responses to exposure to other henipaviruses. Firstly, 2×106 TCID50/ml CedPV passaged twice in bat PaKi cells was administered to 2 male ferrets (1 ml oronasally); 4 female guinea pigs (1 ml intraperitoneally); and 5 female Balb-C mice (50 µl oronasally). Guinea pigs and mice were implanted with temperature sensing microchips (LifeChip Bio-thermo, Destron Fearing) and weighed daily. Ferret rectal temperature and weight was recorded at sampling times. Animals were observed daily for clinical signs of illness and were euthanized at 21 d post-inoculation. Sera were collected on days 10,15 and 21 to test for neutralizing antibody against CedPV. Secondly, on the basis of asymptomatic seroconversion to CedPV noted in ferrets in the first study, 7 further female ferrets were exposed by the oronasal route to a lower dose of 3×103 TCID50. Two animals were euthanized on each of days 6,8 and 10 post-inoculation and one on day 20. Nasal washes, oral swabs, and rectal swabs were collected on days 2,4, 6,8 and 10 and urine was sampled on the day of euthanazia; each specimen was assessed for CedPV genome. A wide range of tissue samples were collected at post mortem examination and assessed by routine histology, immunohistochemistry (using rabbit antibodies raised against recombinant CedPV and NiV N proteins, respectively), qPCR (see above) and virus isolation using reagents and procedures previously established in our group [16]. HeLa cells were infected with Hendra and Cedar viruses at an MOI 0. 5 for 24 hours, at which time total cellular RNA was extracted and IFN-α and IFN-β mRNA levels were quantified by real-time PCR using Power SYBR Green RNA-to-CT 1-Step Kit (Applied Biosystems). Primers were as previously described [50]. Sera from 100 flying foxes collected during 2003–2005 from Queensland, Australia were screened for neutralizing antibodies to CedPV. Virus neutralization test was conducted as described above (antibody tests). All serum samples were tested at a dilution of 1∶20.

Hendra and Nipah viruses are 2 highly pathogenic paramyxoviruses that have emerged from bats within the last two decades. Both are capable of causing fatal disease in both humans and many mammal species. Serological and molecular evidence for henipa-like viruses have been reported from numerous locations including Asia and Africa, however, until now no successful isolation of these viruses have been reported. This paper reports the isolation of a novel paramyxovirus, named Cedar virus, from fruit bats in Australia. Full genome sequencing of this virus suggests a close relationship with the henipaviruses. Antibodies to Cedar virus were shown to cross react with, but not cross neutralize Hendra or Nipah virus. Despite this close relationship, when Cedar virus was tested in experimental challenge models in ferrets and guinea pigs, we identified virus replication and generation of neutralizing antibodies, but no clinical disease was observed. As such, this virus provides a useful reference for future reverse genetics experiments to determine the molecular basis of the pathogenicity of the henipaviruses.

Abstract Introduction Results Discussion Materials and Methods

veterinary diseases emerging infectious diseases virology emerging viral diseases veterinary virology biology microbiology veterinary science

2012

Cedar Virus: A Novel Henipavirus Isolated from Australian Bats

Zygotic gene expression programs control cell differentiation in vertebrate development. In Xenopus, these programs are initiated by local induction of regulatory genes through maternal signaling activities in the wake of zygotic genome activation (ZGA) at the midblastula transition (MBT). These programs lay down the vertebrate body plan through gastrulation and neurulation, and are accompanied by massive changes in chromatin structure, which increasingly constrain cellular plasticity. Here we report on developmental functions for Brahma related gene 1 (Brg1), a key component of embyronic SWI/SNF chromatin remodeling complexes. Carefully controlled, global Brg1 protein depletion in X. tropicalis and X. laevis causes embryonic lethality or developmental arrest from gastrulation on. Transcriptome analysis at late blastula, before development becomes arrested, indicates predominantly a role for Brg1 in transcriptional activation of a limited set of genes involved in pattern specification processes and nervous system development. Mosaic analysis by targeted microinjection defines Brg1 as an essential amplifier of gene expression in dorsal (BCNE/Nieuwkoop Center) and ventral (BMP/Vent) signaling centers. Moreover, Brg1 is required and sufficient for initiating axial patterning in cooperation with maternal Wnt signaling. In search for a common denominator of Brg1 impact on development, we have quantitatively filtered global mRNA fluctuations at MBT. The results indicate that Brg1 is predominantly required for genes with the highest burst of transcriptional activity. Since this group contains many key developmental regulators, we propose Brg1 to be responsible for raising their expression above threshold levels in preparation for embryonic patterning. Vertebrate BAF protein complexes remodel chromatin with the mutually exclusive help of Brahma (brm) or Brahma-related gene 1 (brg1) ATPase subunits. SWI/SNF complexes are known to participate broadly in nucleosome-based aspects of DNA metabolism in normal and malignant cells [1–4], but their specific ATPase subunits designate them for different functions. Brm-/- mice are viable although heavier than normal, suggesting Brm to be a negative regulator of cell proliferation [5]. In contrast, brg1-/- mice die during early embryogenesis and brg1 heterozygotes are predisposed to exencephaly and tumor formation [6]. These results suggest unique functions for BAF complexes carrying the different ATPases. Brg1 containing BAF complexes become further subspecialized in a tissue specific manner by association with co-factors of the BAF60 protein family during cell differentiation [7,8]. Specific functions have also been described in murine embryonic stem cells, where a specialized esBAF complex containing Brg1, Baf155 and Baf60a regulates aspects of ES self renewal, pluripotency and cell priming for differentiation [9–11]. How these findings for esBAF relate to normal mouse embryogenesis is not fully clear. Embryos lacking maternal Brg1 protein arrest at two cell stage and are compromised in zygotic genome activation [12], while embryos, lacking only zygotic Brg1 protein, die before implantation [6]. Since ATP dependent chromatin remodelers, including Brg1, are conserved among vertebrates [10,13] deeper insight into Brg1’s developmental functions could be derived from non-mammalian vertebrate model organisms. Although Brg1 is expressed throughout development, several reports from Xenopus and Zebrafish have shown only relatively late requirement of Brg1 in development, i. e. during differentiation of heart, neural plate and brain [13–15]. We had obtained precedence for specific involvement of chromatin remodelers in developmental processes as early as germ layer formation in Xenopus, where Mi2-beta/NuRD remodeling activity is needed to position the boundary between mesoderm and neuroectoderm [13]. These findings let us expect also earlier functions for BRG1/BAF complexes. In search for such early functions, we have investigated the transcriptional and embryonic consequences of Brg1 depletion in the closely related species X. tropicalis and X. laevis. To generate a Brg1 loss of function situation in Xenopus, we designed three Morpholino (MO) oligonucleotides against mRNAs of both X. laevis brg1 homoeologs (Fig 1A). We determined their relative translation blocking activities in X. laevis embryos with a recombinant brg1/luciferase transcript, containing ~700bp of the brg1 cDNA sequences with the morpholino targeting regions fused in frame to the luciferase ORF. The blocking efficiencies of the Morpholinos increased about three-fold from 5’ to 3’ direction on the target mRNA. BMO1 had the strongest effect and reduced luciferase activity approximately seven-fold (S1 Fig). Based on these results, we selected BMO1 and BMO2 for further analysis. We investigated the consequences of systemic Brg1 protein knockdown in X. tropicalis embryos, where the target region for BMO1 and BMO2 is conserved (Fig 1A). To achieve a homogenous protein knockdown, Morpholinos were injected four times into the animal pole region at the two- to four-cell stage (“radial” injection type). In titration experiments, we determined a dose of 30 ng BMO1/embryo to reduce Brg1 protein levels to one-third, while even 60ng of BMO2 reduced them only two-fold (Fig 1C). Whereas more than 90% of the control morphants survived until hatching, the majority of the BMO1 injected embryos died during gastrulation (Fig 1B). Consistent with less efficient Brg1 depletion, BMO2 morphant embryos died later than BMO1 morphants and survived better (Fig 1B). The survival rate of CoMO and BMO1 morphants was comparable until late blastula, although the Brg1 protein levels were already diminished in the latter case to about 30% of CoMO injected embryos (Fig 1C). The residual BRG1 protein is very likely of maternal origin [14] and therefore insensitive to MO knockdown. Notably, immunostaining for activated Caspase-3 showed no signs of apoptosis at early gastrula stage indicating that morphant blastulae are healthy and initiate gastrulation without visible defects or delay (S2 Fig). Bulk zygotic transcription commences in Xenopus at the midblastula transition (MBT), about three hours before gastrulation starts. Because under our conditions BMO1 morphants die mostly during gastrulation or neurulation, but not before, it was possible to assess the consequences of Brg1 protein knockdown on embryonic transcription at late blastula. Using the same conditions described above, we compared CoMO and BMO1 injected X. tropicalis embryos by microarray analysis. Although these conditions were ultimately lethal, at the investigated late blastula stage more than 90% of the mRNAs were expressed at normal levels. A total of 872 transcripts responded to the Brg1 protein knockdown, with 211 of them being upregulated, and 661 being downregulated, relative to control morphants (Fig 1D, S1 and S2 Tables). Gene Ontology analysis revealed an enrichment for the terms “chromatin assembly”, “cellular complex assembly”and “macromolecular complex assembly” in the upregulated cohort (S3 Fig, panel A). In contrast, the downregulated responders were strongly enriched in several GO terms related to various developmental and pattern specification processes (S3 Fig, panel B). Here the most enriched term was “nervous system development”, consistent with a known requirement for Brg1 during vertebrate neural differentiation [14–17]. Nineteen of the 61 genes from this GO category were reduced in our genome-wide data set already at the blastula stage, i. e. before neural plate formation. From these 19 genes, fifteen were found reduced by independent qRT/PCR analysis (Fig 1E). In addition we reproduced the microarray results for a variety of important developmental regulatory genes by qRT/PCR (S3 Fig, panels C and D). Furthermore, we investigated by whole-mount RNA in situ hybridization (WMISH) the expression patterns of genes involved in neural induction at late blastula, confirming the microarray results for downregulated foxD4l1 and noggin expression, and unaffected zic2 expression in BMO1 morphants (S3 Fig, panels E-K’). These independent analyses confirmed the microarray data in a robust manner. In summary, our results indicate an essential function for Brg1 protein before the onset of gastrulation, detailing primarily an enhancement of gene transcription. We repeated the morphological analysis in X. laevis by injecting radially the BMO1 morpholino at 60 ng, 40 ng and 20ng per embryo, together with a fluorescent lineage tracer (S4 Fig). At the two higher doses, BMO1 injections caused again embryonic death in the majority of the embryos before late neural tube stage (NF22; S4 Fig panel C). All remaining embryos were arrested in gastrulation (S4 Fig, panel A), occasionally surviving in this state until the heartbeat stage (NF34; S4 Fig, panel B). At a dose of 20 ng, more than half of the embryos survived until NF22; while about 80% of the survivors did not finish gastrulation, 20% became arrested at the open neural plate stage and remained in this condition until NF34 (S4 Fig, panels B and C). In summary, embryonic survival is correlated with the BMO1 dose. Moreover, the formation of dorso-anterior structures is still completely blocked under conditions (20ng BMO1/radial), in which the majority of the embryos survives beyond neurulation. Therefore, we wondered, whether local ablation of Brg1 protein by a reduced Morpholino dose might improve overall embryonic differentiation and thus provide information on developmental pathways requiring Brg1 activity. We tested this hypothesis in X. laevis embryos by targeted microinjections. When CoMO or BMO1 oligonucleotides (10ng/embryo) were injected at the 4-cell stage into the marginal zone of either the two dorsal (DMZ), or the two ventral (VMZ) blastomeres, most of these embryos survived until tadpole stage. β-Galactosidase staining for coinjected nlacZ mRNA confirmed the correct targeting of the injections and demonstrated viability of the injected cell progeny. DMZ injected control morphants grew up into phenotypically wildtype tadpoles (Fig 2A). In contrast, 80% of DMZ injected BMO1 morphants displayed stunted antero-posterior body axes with severely truncated heads carrying the nlacZ stain (Fig 2B and 2F). The remaining 20% of BMO1 morphants showed no morphological abnormalities. The major phenotype was a specific consequence of Brg1 protein depletion, since dorso-anterior structures were largely rescued by coinjection of wildtype human brg1 mRNA that contains four mismatches in the BMO1 target region. The heads of these rescued embryos contained well-developed eyes with lenses and recurrent retinal pigment (Fig 2C). Interestingly, this rescue of the BMO1 morphant phenotype was neither achieved with human brm nor Xenopus iswi mRNAs (S3 Table). When we inspected VMZ injected embryos, their overall morphology was much less affected. Although we have not investigated any internal organs, these embryos were at least capable to develop a well-structured body axis including heads (Fig 2D, 2E and 2G). A smaller fraction (30%) of the BMO1 morphants was weakly anteriorized, displaying enlarged heads and eyes, but concomitantly deficient in posterior tissues such as the fin (Fig 2E’ and 2G). Several conclusions can be drawn from these results. First, the presence of β-Gal positive cells in tadpoles demonstrates that BMO1 injections are not cell-lethal per se. Second the inability of Brm and Iswi to rescue the BMO1 phenotype argues for distinct remodeling events that specifically require Brg1 protein. Finally, the observed morphological phenotypes suggest that Brg1 is involved in axis formation. Particularly on the dorsal side of the embryo, Brg1 protein seems required to unfold the dorsalizing gene expression program (DGEP) during germ layer patterning. This assumption was investigated by several experiments. First, we coinjected Xenopus brg1 and nlacZ mRNAs into one ventral blastomere at the 4-cell stage of wildtype embryos. At the tadpole stage, these embryos had formed with high penetrance a secondary axis rudiment, which contained somites with differentiated muscle tissue (S5 Fig). Moreover, the chordin gene is one of the early developmental regulators, downregulated in radial X. tropicalis BMO1 morphants (S3 Fig, panel C, S2 Table). This gene was ectopically induced by ventral injections of human brg1 mRNA (S6 Fig). Notably, hBrg1 also efficiently restore dorso-anterior development in DMZ injected BMO1 morphants (Fig 2C and 2F). These two observations indicate that Brg1 protein overexpression can initiate de novo formation of axial structures, apparently through activation of DGEP genes like chordin. Normally, formation of dorsal structures is initiated by maternal Wnt/β-Catenin signaling on the prospective dorsal side of the embryo [18–20]. Coinjection of β-catenin mRNA at a dose, which alone was insufficient to cause morphological consequences, reestablished quite efficiently head structures, including eyes, as well as longer body axes and tails in DMZ-injected BMO1 morphants (Fig 3A–3D). When injected ventrally, β-Catenin frequently induced a secondary embryonic anlage with complete heads, which was reduced to single-axis status by coinjection of BMO1 (Fig 3E–3H). These last experiments demonstrate a cooperation between canonical Wnt signaling and Brg1 in early embryonic patterning, which had not been observed before. At blastula stage, DGEP is initiated by two newly induced, local signaling centers–the dorso-animal Blastula-Chordin-Noggin-Expressing (BCNE) Center and the overlapping, but more vegetally located Nieuwkoop Center (NC). Both regions contribute to head-formation and are induced by maternal WNT-signaling [21]. The BCNE signature genes, including nodal3. 1/nr3, chordin and noggin, encode secreted BMP inhibitors. All these genes, in particular chordin, were downregulated by radial Brg1 protein knockdown in X. laevis (Fig 4, panels A-C). Expression of the BCNE genes was largely restored by coinjection of human Brg1 mRNA (Fig 4A”–4C”), consistent with the previously described morphological rescue of dorso-anterior tissues (see Fig 2). The downregulation of chordin and noggin mRNAs matches our results in X. tropicalis (S3 Fig, panels C, and G to H’). We note that the expression of nr3 was upregulated in X. tropicalis blastulae (S3C Fig), while it is downregulated in the X. laevis BCNE. Possible mechanisms for this species-specific difference are discussed later. The NC marker gene sia1 is a direct Wnt-target, a key regulator of axial patterning, and is needed for proper gene expression in BCNE and Spemann’s organizer [22–24]. Notably, expression of sia mRNA in radial BMO1 morphants was not downregulated compared to control morphants (Fig 4D and 4D’), consistent with the results for sia1 and sia2 from the microarray analysis in X. tropicalis (S3C Fig). The insensitivity of sia1/sia2 genes excludes a role of Brg1 protein as general coactivator of Wnt/β-Catenin signaling in Xenopus, which had been suggested by earlier studies [25]. Notably, Brg1 is involved in both direct (nr3; [26]) and indirect [chd, nog; ref. [27]] Wnt-mediated gene activation events in the BCNE, which is consistent with the morphological phenotypes observed in DMZ-injected BMO1 morphants. The process of germ layer patterning, which defines the future body plan, occurs during gastrulation and requires Spemann’s organizer, which overlaps with and succeeds the BCNE and NC territories. Cells of the organizer are the first ones to involute during gastrulation. They secrete a panoply of proteins, which inhibit BMP as well as Wnt and Nodal signaling pathways [28–30]. These organizer properties generate gradients of signaling activities that dynamically establish gene expression domains of appropriate size within the morphogenetic field of the forming germ layers [31,32]. Since we had discovered that BCNE gene expression depends on normal BRG1 protein levels, we sought to extend the analysis to gene expression domains of organizer and non-organizer mesoderm. We evaluated the expression of critical regulatory factors by WMISH in early to mid gastrula stage (NF10. 5 to NF11), i. e. before development becomes typically arrested in X. laevis BMO1 morphants (S4 Fig). Among the organizer genes was the BMP inhibitor nr3 [33], which at blastula stage was downregulated in the BCNE. The nr3 mRNA levels were reduced in most gastrulae within its normal domain (Fig 5A–5C). The otx2 gene is first transcribed in the organizer and specifies at later stages anterior tissues in all three germ layers [34]. Upon Brg1 knockdown, the intensity of otx2 staining and the size of its domain were reduced. This was true both for preinvoluted otx2 expression at the blastopore lip as well as for the involuted part, where otx2 mRNA is confined to a narrow stripe in BMO1 morphants (Fig 5D–5F). Finally, foxA4 mRNA was frequently downregulated in its proper domain at the lip (Fig 5G–5I). Some non-organizer genes in the neighboring dorso-lateral mesoderm were also misregulated. This included the homeobox genes vent1 and vent2, which mediate the ventro-posteriorizing activity of BMP ligands [35]. In BMO1 morphants, transcripts from both genes invaded the organizer territory (Fig 5K–5P). This dorsal expansion of vent gene expression indicates a severe functional impairment of Spemann’s organizer [36]. In addition, the muscle regulatory genes myoD and myf5 were reduced (S7 Fig, panels A-F), while gsc, t/bra and xpo were unaffected (S7 Fig, panels G-S). Also chordin transcription was unimpaired in the organizer, despite the fact that it is downregulated in the BCNE region at blastula stage (compare S7K–S7M Fig with Fig 4A). While chordin is activated by maternal Wnt signaling in the BCNE, it is controlled in the organizer by additional regulators including nodal signaling and the mesodermal transcription factors gsc and not [37,38]. These findings suggest a context-dependent role for Brg1 in target gene regulation. In summary, a systemic depletion of BRG1 protein leads to a misbalance in dorso-ventral patterning and an unusual coexistence of non-organizer (vent1/vent2) and organizer transcripts within the dorsal blastopore lip. Our morphological and molecular analyses defined the earliest defects in BRG1-depleted embryos to the late blastula stage, when the BCNE center is established. Kuroda and colleagues have demonstrated by tissue transplantation that the BCNE contributes to brain formation [21]. Therefore, we decided to address by orthotopic transplantation, whether the absence of eyes and forebrains in BMO1 morphants arises autonomously from the BCNE region, or results from a defective crosstalk between germ layers. For this purpose we transplanted wildtype or morphant BCNE grafts into wildtype host embryos (Fig 6). Grafts were marked by fluorescent dextran to distinguish them from host tissues (experimental workflow see S8A Fig). At the tadpole stage, two-thirds of the embryos transplanted with a WT-BCNE had generated tadpoles with heads containing well-developed eyes with lenses (n = 14/21; Fig 6A–6A”). In contrast, almost 90% of the embryos transplanted with a morphant BCNE lacked eyes completely or had only remnants of retinal pigmentation without lenses (n = 22/25; Fig 6B–6B”). The morphological differences between the two conditions were significant (Fig 6C). To visualize the major brain domains we stained the transplanted tadpoles for otx2 mRNA (S8 Fig, panels B-E). Half of the morphant transplants showed a strong reduction in otx2 expression, in which forebrain, midbrain and hindbrain areas were collapsed to an amorphous tissue mass. This result was particularly obvious in specimen, in which the lineage tracer of the transplanted BCNE populated only part of the brain. In these cases, otx2 staining was structured comparatively normal in the host-derived parts of the brain, while it was severely reduced in the transplanted area (S8 Fig, compare panels C, D with C’ and D’). The clearest results were observed for otx2 expression in the retina, which was present in 90% of the WT transplants, but only in 25% of the morphant transplants. In summary the results from this experimental series indicate an autonomous defect within the Brg1-depleted BCNE region for neuronal differentiation and brain patterning, which cannot be compensated by secreted factors from the wildtype host environment, including mesodermal Chordin (S7K–S7M Fig). We have demonstrated that several BCNE genes, in particular chordin, are specifically downregulated in BMO1 morphants and are responsible for defective head formation. However, the global transcriptome analysis had revealed a much larger number of genes responding to BRG1 knockdown, suggesting that other regions of the embryo also contribute to the BMO1 phenotype. In a new series of experiments, we compared side by side the consequences of dorso-animal (”DA”/BCNE center) with dorso-vegetal (“DV”/Nieuwkoop Center) blastomere injections at the eight-cell stage. The majority of control morphants developed completely normal in both types of injections (Fig 7, panels A, D, C and F). As expected, DA-injections of BMO1 resulted in embryos with shorter, tailless axes, and strongly reduced heads (Fig 7, panels B and C). Targeting of the BMO1 to DV-blastomeres maintained the length of the main body axis much better than DA-injections, but still reduced head and eye formation in a large fraction of the embryos (Fig 7, panels E and F). RNA in situ hybridisation and ß-Galactosidase staining at the late blastula stage confirmed that both DA and DV blastomeres contribute to the BCNE expression zone, as it has been described before [39,40]. Consequently, chordin and noggin mRNAs were downregulated with both injections within the overlap (S9 Fig, panels A-J). The genes hhex and cer1 are expressed in the dorso-vegetal region and are known to promote anterior development and head formation, respectively [41,42]. DA-injections of BMO1, which do not overlap with the hhex and cer1 expression domains, had no effect on these mRNAs (Fig 7, panels G-L). In contrast, DV-injections strongly downregulated both genes in a statistically significant manner (Fig 7, panels M-Q). Most importantly, the results from targeted 8-cell injections demonstrate an additional role for Brg1 in the Nieuwkoop Center, the prospective anterior endoderm region of the embryo, where it is required for hhex and cer1 transcription. The morphological and molecular analysis of its protein knockdown phenotype demonstrated Brg1 to be essential for embryonic vitality and germ layer patterning. Targeted injections of BMO1 oligo to different regions of the embryo support this conclusion in a consistent manner and revealed functional connections of Brg1 to several signaling pathways (Wnt/Bmp) and embryonic regions (BCNE/anterior mesendoderm). In search for a common denominator of Brg1 function, which could explain both the diverse impact on embryonic gene expression at late blastula stage (>800 altered transcripts) and the developmental functions in dorsal and ventral signaling centers, we investigated the gene response in Brg1 morphants in relation to the zygotic genome activation (ZGA) at MBT. Precedence for this assumption comes from work in mice, where absence of maternal BRG1 protein has been reported to cause developmental arrest at 2-cell stage and to impair transcription during ZGA [12]. While originally identified as global onset of zygotic transcription, MBT is now recognized as a continuous reorganization of the embryonic mRNA pool from early to late blastula stages, consisting of a major turnover of maternal mRNAs, coupled to broad, but not genome-wide initiation of transcription [43–45]. As shown in Fig 8A, three prototypic transcript classes can be operationally defined—i) maternal mRNAs, whose abundance declines; ii) transcripts with relatively constant abundance, and iii) mRNAs with increasing abundance through de novo transcription. We decided to characterize the three mRNA classes through global transcriptome analysis at immediate pre-MBT (NF8) and late Blastula stages (NF9) in X. tropicalis embryos (see S10 Fig, panels A and B for details). We then compared these data with the transcriptional changes observed in the BMO1 morphants at late blastula. About 104 genes are expressed at the two time points (Fig 8B). After setting a threshold for transcript levels with an adjusted p-value ≤0. 05, the abundance of 1357 transcripts changes between pre-MBT and late blastula. Decreasing levels characterize 761 transcripts as maternal mRNAs, whereas 596 transcripts classify as zygotic mRNA due to their increasing abundance (Fig 8B and S4 Table). By plotting the difference in mRNA levels between Brg1 morphants versus control morphants, we found that transcripts from the 596 genes activated at the MBT responded much stronger to the Brg1 knockdown than other transcripts (Fig 8C). A Gene Ontology search associated these zygotic transcripts with “Regulation of Transcription”, “RNA Metabolic Process”, and “Pattern Specification Process” (S10 Fig, panel C). When genes of the GO-term “Pattern specification process” were ranked in a heat map according to their magnitude of transcriptional activation, many of them were significantly downregulated by Brg1 protein knockdown (red asterisks, Fig 8D). A similar correlation between Brg1 dependence and transcriptional activation at MBT was found for the GO-term “Nervous System Development” (S10 Fig, panel D). In both cases, Brg1-dependent genes were enriched in the upper half of the heat maps, where genes with the highest fold-activation are located. The correlation was even more striking, when only the top-activated zygotic RNAs (≥5. 65-fold increase after MBT; n = 324 genes) were considered—over 40% of these genes were Brg1 dependent (Fig 8E). Based on this analysis Brg1 protein is needed to amplify a transcriptional burst at MBT, which is necessary to initiate embryonic patterning. To investigate potential developmental functions for Brg1 between MBT and neurula stages, we have carefully optimized the conditions for Brg1 protein depletion using three overlapping Morpholino oligonucleotides. These MOs differed significantly in their ability to block Brg1 protein synthesis, with the most upstream located targeting site of BMO3 having the least effect. The binding site of the Morpholino oligo used in earlier studies [46] starts 15 nucleotides upstream of BMO3 and does not overlap with BMO1/-2 target regions, presumably resulting in suboptimal targeting of brg1 mRNAs. We consider this the most likely explanation for the much earlier and more severe effects we report in this study. The demonstrated reduction of endogenous Brg1 protein levels in combination with the robust rescue of both morphological and molecular aspects of BMO1 morphants by human brg1 mRNA identifies the reported phenotype as a specific consequence of the Brg1 protein knockdown. Furthermore, both Brm and Iswi fail to compensate the Brg1 deficiency, suggesting specific remodeling events for Brg1-containing SWI/SNF complexes as the underlying molecular cause. These findings are in agreement with other reports [46–48]. Our results extend in a significant manner the current knowledge about the role of Brg1 during vertebrate embryogenesis. Transcriptome analysis from X. tropicalis late blastula stage (NF9) indicates that nearly 9% of transcripts (Fig 1D) were sensitive to Brg1 depletion. By comparison of the pre- and post-MBT transcriptomes we classified almost 600 transcripts to be de novo expressed at MBT, a number well in agreement with current estimates of X. tropicalis zygotic genome activation [47,48]. The majority of these newly activated genes is Brg1-sensitive, detailing a significant impact for Brg1 on the first wave of zygotic gene expression of the embryo. In general we found that the gene response to Brg1 depletion is conserved between the two frog species. One clear exception is nr3/nodal3. 1, which in the absence of Brg1 is downregulated in X. laevis, but upregulated in X. tropicalis (Figs 4,5 and S2). Such a differential response could involve a species-specific modulation of the functional outcome of Brg1 activity on nr3 gene transcription, and/or reflect differences in the transcriptional regulation of these orthologs. Indeed, there is evidence for chromosomal rearrangements and gene amplifications at the nr3/nodal3. 1 gene locus [49] which may have changed the regulation of nr3 transcription in the X. laevis L genome. The short time span between MBT and late blastula (~3 hrs) implies that many, if not all misregulated genes are direct targets of Brg1-SWI/SNF. We expect this assumption to be valid for up- and downregulated gene cohorts alike, based on abundant evidence implicating Brg1 complexes with both gene activation and repression. Generally, the outcome of its action on a target gene is dictated by Brg1’s protein partners within SWI/SNF multiprotein complexes and by the gene-specific context. The many protein interactors of Brg1 and the versatile effector functions of SWI/SNF complexes pose a remarkably difficult challenge to investigate Brg1 acitivity on the mechanistic level in the embryo. Notably, Brg1-SWI/SNF complexes are found associated with both active and repressed chromatin elements [50] through interactions with DNA, histone modifications, and a large number of specific transcription factors (for detailed information on Brg1-SWI/SNF see reviews in [2,4, 17]. In our study, we have not investigated genes that become upregulated/derepressed in BMO1 morphants. However, recent studies indicate that in human and mouse ESCs, BRG1-SWI/SNF complexes repress enhancer elements of lineage specification factors by modulating H3K27 acetylation or methylation levels [51,52]. Based on this evidence, some of the genes, which are repressed by Brg1 in Xenopus before gastrulation, may also be important for embryonic development. Highly informative for us were the GO-terms enriched in the downregulated gene cohort. They guided our analysis to establish Brg1 protein–and by inference corresponding SWI/SNF complexes–as regulator of early embryonic patterning. In evaluating the role of Brg1, one needs to distinguish carefully early and late phenotypes. The aberrant patterns of gene expression in BCNE, Nieuwkoop center and Spemann’s organizer are immediate consequences of the lack of Brg1 protein by temporal linkage with ZGA. Some phenotypes assessed late, like the axial defects characterizing the DMZ-injected BMO1 morphants, can still be attributed to direct effects, since the state of the underlying regulatory gene network is fixed at gastrulation. This has been demonstrated by classical experiments, in which the normal body plan of a tadpole is severely and irreversibly altered through treatments between the late one-cell to 32-cell stage, which modulate maternal Wnt-signaling activity [18,53] The situation is different for late phenotypes that arise from defective cell differentiation over time, such as the malformed brains in embryos transplanted with a morphant BCNE (Fig 6) or injected dorso-animally with BMO1 (Fig 7). Since Brg1 is continuously expressed in the neuroectoderm and neural crest, these phenotypes may indicate a later requirement for Brg1 in brain and retina development [50]. What are the main developmental functions of Brg1 and how are they implemented in the process of embryogenesis? While previous studies implicated Brg1 in neuroectoderm differentiation [15–17,54], our results demonstrate that this ATPase is required for neural plate formation. Programming the neural ground state involves a conserved gene regulatory network [24,54]. Essential members include FoxD4I1, Zic1, Zic3 and Iroquois2, which are all significantly downregulated in Brg1 depleted embryos at late blastula (Figs 1 and S2, S2 Table). The installation of this network requires inhibition of Bmp signals in the prospective neural plate through secreted antagonists, three of which (chordin, noggin, nr3,) are expressed first in the BCNE region and subsequently in the organizer under the control of maternal Wnt/β-Catenin signaling [31,55]. For these genes, our data define Brg1 as a coactivator of Wnt signaling, which helps install DGEP as first step to the formation of dorso-anterior tissues. Whether Brg1 could be even sufficient to induce DGEP alone, as suggested by secondary axis formation and ectopic chordin induction after ventral overexpression of Xenopus and human Brg1 proteins, is not clear at the moment. The ventral side of the embryo contains residual amounts of maternal Wnt11 protein, with which Brg1 might cooperate [20,56]. In dorso-vegetal cells, the combination of maternal Wnt signaling plus high Nodal signaling establishes the Nieuwkoop Center [57]. This mode of regulation separates it from the BCNE, even though it overlaps with the BCNE region by cell lineage [39,40]. The multifunctional BMP inhibitor cerberus, expressed in the Nieuwkoop Center is also BRG1-sensitive, as shown by dorso-vegetal BMO1 injections. Taken together, these results have identified three different embryonic territories to require Brg1 activity, and prevent a simple assignment of Brg1 function to either ventralizing or dorsalizing gene expression programs. Interestingly not all Wnt targets depend on coactivation by Brg1-SWI/SNF, for instance the transcription factor gene sia1 (Fig 4D). This helps to explain the peculiar observation that in BMO1 morphants chordin mRNA is selectively reduced in the BCNE, but is unaffected in the organizer (compare Figs 4 and S7). According to current models, maternal Wnt signaling induces transiently sia1 expression (Brg1-independent; Fig 4D), which in turn activates chordin transcription (Brg1-sensitive; Figs 1,4, S6 and 7) at the blastula stage (see Carnac et al. , Development 1996); subsequently, chordin transcription is maintained through Nodal signaling (Brg1-independent; see Figs 5 and S7) in the gastrula organizer [27]. This model would predict that sia1 maintains chordin in a Nodal-responsive state without transcriptional activation. A poised state is also a common theme for Wnt target genes in Xenopus, which are frequently bound by β-Catenin without eliciting a transcriptional response [58]. Transcriptional activation of Wnt targets has been proposed to occur through context-specific mechanisms, downstream of β-Catenin binding to chromatin. One such context could be recruitment of a Brg1-SWI/SNF chromatin remodeler. It should be highly informative to identify mechanisms, by which Brg1-SWI/SNF recognizes its targets at the blastula stage. Gene expression domains in the gastrula organizer are generally less affected in BMO1 morphants than in the BCNE. Nevertheless, our data indicates a third function for Brg1 in activating transcription of the core BMP synexpression group (bmp4, vent1 and vent2; Fig 5, S2 Table), which specifies ventro-posterior tissues [59]. Interestingly, while their mRNA levels are globally reduced in radial BMO1 morphants at late blastula (S2 Table), the overlapping vent1/vent2 expression domains are expanding into the organizer field. Here, they become coexpressed with DGEP genes like gsc and otx2 (which specify anterior position) and the bmp antagonists chordin and nr3 (Figs 5 and S7). This highly unusual pattern of genes in the organizer had also been generated by simultaneous knockdown of three BMP antagonists (chordin, noggin, follistatin) [36]. It indicates a significant weakening of the organizer function, which is reflected in our data by reduced expression of nr3 and otx2 within the organizer, and of the myogenic bHLH transcription factors myoD and myf5 in non-organizer mesoderm (Figs 5 and S7). That the reduced bmp4/vent1/vent2 expression indeed contributes to the aberrant body plan, is apparent in radial morphants injected with 20ng BMO1 (S4 Fig). Although their blastomeres received the same amount of BMO1 oligo as embryos, which were injected only in DMZ or VMZ (i. e. 5ng/blastomere), the radial morphants were arrested at gastrulation and thus morphologically much stronger affected than DMZ or VMZ morphants (compare S4B Fig with Fig 2). It is plausible to assume that these perturbations in the organizer and non-organizer activities, together with reduced cerberus transcription in the Nieuwkoop Center are responsible for the massive loss of dorso-anterior structures seen in DMZ-injected BMO1 morphants (Figs 2 and 3) and for the posteriorized character of Brg1 depleted embryos, arrested permanently at the gastrula stage (S4 Fig). The invasion of vent1/vent2 gene expression into the organizer field suggests a basal failure in the cell determination process, which normally prevents activation of non-compatible gene expression programs in cells [60]. Indeed, non-compatible gene expression could mount a substantial problem given that almost 900 genes are misregulated in Brg1 morphants. In summary, we have identified at least two independent functions for Brg1-SWI/SNF, which are essential for Xenopus embryonic patterning, namely being i) selective coactivator of maternal Wnt signaling on the prospective dorsal side of the embryo and ii) coactivator of the core bmp synexpression group on the prospective ventral side. However, more than 800 genes are misregulated in Brg1 depleted embryos at the blastula/gastrula transition. Is there a common denominator to explain Brg1’s impact on development? By quantitative filtering of global mRNA fluctuations at MBT, we have shown that Brg1 is predominantly required for genes with the highest burst of transcriptional activity. Mechanistically, Brg1-SWI/SNF could be involved to catalyze transitions from transcriptional silent to active chromatin states at promoters or facilitate long range interaction between distal enhancers and promoters [61], which in general become engaged at the blastula/gastrula transition [62]. Since many of the bursting genes are key developmental regulators, this may put BRG1 in a key position to raise their expression above threshold levels in preparation for the embryonic patterning process. Notably, both mathematical models and experimental evidence have detailed an enormous self-regulatory capacity within embryonic fields [63–65]. Why BRG1-depleted embryos cannot compensate a quantitatively insufficient ZGA and fail to restore axis formation through self-regulation, constitutes a key question for the future. We have shown that BRG1 protein is essential for early Xenopus development. BRG1 is involved in de novo activation of transcription at the Midblastula transition and is needed to achieve the transcriptional amplitude of genes with the highest-fold activation. Among these bursting genes are many regulators of embryonic axis formation. Targeted depletion of BRG1 protein levels results in the specific downregulation of key genes of the BCNE Center (chordin, noggin), the Nieuwkoop Center (hhex, cer) and the bmp-controlled ventral signaling territory (vent1, vent2). We propose that Brg1 fulfills a systemic function for late blastula stage transcription in preparation of embryonic pattern formation. Animal work has been conducted in accordance with Deutsches Tierschutzgesetz; experimental use of Xenopus embryos has been licensed by the Government of Oberbayern (AZ: 55. 2. 1. 54–2532. 6-7-12). The ORF of X. laevis brg1 cDNA was generated by PCR from overlapping ESTs (Genbank acc. Nrs. AW766934, BG234591, BQ7288178) and subcloned into pCS2+. The full-length cDNA was verified by sequencing and deposited into GenBank (AY762376). For testing morpholino targeting efficiencies, the cDNA region from -77 to +617 of X. laevis brg1 (“BISH”) was fused in frame to the luciferase ORF in a gateway-compatible pCS2+ vector. All primer sequences are provided in the supplemental data section, S5 Table, part a. Open reading frames of human brg1, brm and X. laevis iswi (kind gift from Anthony Imbalzano and Paul Wade) were sub-cloned into pCS2+ for in vitro transcription. Capped mRNA for microinjection was synthesized as described [66]. Three antisense morpholino oligonucleotides against the translational start site of Brg1 mRNAs were purchased from GeneTools: All three are fully complimentary to transcripts from both X. laevis homeologs (NM_001086740. 1 and BG554361); BMO1 and BMO2 also match perfectly the mRNA sequence of the S. tropicalis homolog (BG554361). BMO1: 5’- CCATTGGAGGGTCTGGGGTGGACAT-3’; BMO2: 5’-CAGGGAGAAGATCCAGTCACTGCTA-‘3; BMO3: 5’-GACATCACTGCAGGGAGAAGATCCA-‘3. The unrelated standard control Morpholino served as control for specificity. Morpholino targeting efficiencies was determined in vivo with a Brg1-luciferase fusion mRNA. Xenopus laevis embryos were radially injected at the 2 cell stage with individual morpholino oligonucleotides (60ng/embryo). At the 8 cell stage, the four animal blastomeres were superinjected with synthetic BISH-luciferase mRNA (25pg/embryo). The embryos were cultivated until gastrulation (NF11) and luciferase activity was measured with Dual-Luciferase® Reporter Assay System (Promega). Samples consisted of cleared protein lysates from 5 pooled embryos per condition. X. laevis and X. tropicalis eggs were collected, in vitro fertilized, microinjected and cultivated following standard procedures. Embryos were staged according to Nieuwkoop and Faber (1967). Radial injections were performed at the 2–4 cell stage, targeted injections were performed either at the 4 cell, 8 cell or 16 cell stage. For lineage tracing they were either injected with Alexa Fluor-488 Dextran, Alexa Fluor-594 Dextran (Invitrogen) or with 25-100 pg/blastomere of either nuclear lacZ or eGFP mRNA. Tissue transplantations were carried out in 0. 8x MBS + Gentamycin in agarose-coated dishes. The transplant was kept in place with a cover slip for one hour, after which the transplanted embryo was transferred to a new dish in 0. 1x MBS + gentamycin. Whole-mount RNA in situ hybridizations were performed as described [67]. Embryos were photographed with a Leica M205FA stereomicroscope. For immunocytochemistry anti-active Caspase3 antibody (1: 20000, Promega) and anti-rabbit alkaline phosphatase-conjugated (1: 1000, Chemicon) secondary antibody was used. For the production of xBrg1 specific monoclonal antibodies, N-terminal domain (amino acid 202–282) was cloned into pGEX4T3 expression vector (Amersham), expressed in E. Coli and purified to immunize rats. For quantitative measurement of Brg1 knockdown 15 embryos or eggs of each condition were collected and lysed in 75 μl NOP buffer [68] and centrifuged for 20 min at 14000 rpm to remove yolk plates. The supernatant was mixed with Roti®-Load 1 (Roth) and loaded onto an 8% SDS-PAGE. The separated proteins were blotted on nitrocellulose membrane and blocked for minimum 1 h at RT in 5% milk in PBSw. The membrane was incubated over night at 4°C with anti-Brg1 mab 3F1 (1: 3) and as loading control anti α-tubulin (1: 8000, Sigma). As secondary antibody the LiCor α-rat 800 and α-mouse 700 (1: 10000, respectively) was used. The membrane was developed using the LiCOR system and the intensities were measured and quantified against the loading control. Total RNA of 10 embryos was extracted using Trizol (Ambion) and phenol/chloroform. The RNA was precipitated with 70% Isopropanol and cleaned using the RNeasy Cleanup Kit (Qiagen) including DNAseI-on-column digestion. For qPCR analysis 1 μg of total RNA was transcribed with the DyNAmo CDNA Sythesis Kit (Bioenzym). For qPCR 5–20 ng cDNA was mixed with the Fast SYBR Green Master mix (Applied Biosystems) and amplified with a Lightcycler (Roche). Primer sequences are given in S5 Table, part b. For comparative MicroArray analysis X. tropicalis embryos were injected radially at the 2–4 cell stage with 30ng BMO1 or 60ng CoMO and cultivated until late Blastula stage. Per condition, 10 embryos were collected and RNA was extracted as described above. For the pre/postMBT Microarray, wildtype X. tropicalis embryos were collected. Ten embryos were pooled per sample. After fertilization we collected one sample in the 4cell stage as negative GS17 control. For the preMBT time-point we collected embryos from around the ~1000 cell stage and then every 20–25 minutes for approximately 60-100min. We choose 20–25 min breaks depending on how long the embryos need in average for the first cell divisions. 120–150 minutes after collecting the last preMBT sample we started again to collect every 20–25 min samples until we observed dorsal lip formation. For all samples we extracted RNA the way it was described before. In order to find the preMBT sample closest to the MBT we performed qPCR analysis with MBT-marker gs17. For the preMBT time-point we took the last sample without gs17 expression. As postMBT we choose the sample ~40min before dorsal lip formation, in accordance to the developmental age, at which the comparative Microarray analysis was performed. The quality of the extracted RNA was controlled for both experimental setups with the Bioanalyzer and handed to the “Facility of Functional Genomics” at the Gene Center, Munich for microarray performance on an Affymetrix Xenopus tropicalis genome Array. Microarray preprocessing was conducted separately for the two experimental sets (Brg1 knockdown and MBT) using R/Bioconductor (www. bioconductor. org). If not indicated otherwise, we used standard parameters in all functions calls. Expression values were calculated using ‘gcrma’. Probe sets were kept for differential expression analysis if there were more ‘present’ calls (calculated using ‘mas5calls’) in one of the treatment groups than non-‘present’ calls, if their expression level variance was higher than zero across all arrays and if the probe set had an Entrez identifier annotation according to the Entrez database with a date stamp of 2011-Mar16. One gene to many probe set relationships were resolved by retaining only the probe set with the highest interquartile range across all arrays. Differential expression statistics were obtained using a linear model (library ‘limma’). A significant response was defined if the adjusted p-value was smaller than 0. 05. For all embryonic quantitative analysis (morphological phenotype, WMISH, qRT/PCR) SEM are displayed and the statistical analysis was performed using two-tailed, Paired Student’s t-test. For transcriptome analysis see microarray section.

Brahma-related-gene-1 (Brg1) is a catalytic subunit of mammalian SWI/SNF chromatin remodeling complexes. Loss of maternal Brg1 protein arrests development in mice at the 2-cell stage, while null homozygotes die at the blastocyst stage. These early requirements have precluded any analysis of Brg1’s embryonic functions. Here we present data from X. laevis and X. tropicalis, which for the first time describe a role for Brg1 during germ layer patterning and axis formation. Brg1-depleted embryos fail to develop past gastrulation. Genome-wide transcriptome analysis at late blastula stage, before the developmental arrest, shows that Brg1 is required predominantly for transcriptional activation of a limited set of genes involved in pattern specification processes and nervous system development shortly after midblastula transition. Mosaic analysis by targeted microinjection defines Brg1 as an essential amplifier of gene expression in dorsal (BCNE and Nieuwkoop center) and ventral (BMP/Vent) signaling centers, being required and sufficient to initiate axial patterning by cooperating with canonical Wnt signaling. Since Brg1-dependent genes share a high burst of transcriptional activation before gastrulation, we propose a systemic role for Brg1 as transcriptional amplifier, which balances the embryonic patterning process.

Abstract Introduction Results Discussion Conclusion Methods

morpholino rna interference gene regulation messenger rna vertebrates nucleotides animals xenopus blastulas animal models dna transcription developmental biology model organisms amphibians experimental organism systems epigenetics embryos research and analysis methods embryology genetic interference gene expression antisense oligonucleotides biochemistry rna nucleic acids oligonucleotides genetics biology and life sciences frogs organisms

2017

Brg1 chromatin remodeling ATPase balances germ layer patterning by amplifying the transcriptional burst at midblastula transition

The mechanisms of evolution of plant viruses are being unraveled, yet the timescale of their evolution remains an enigma. To address this critical issue, the divergence time of plant viruses at the intra- and inter-specific levels was assessed. The time of the most recent common ancestor (TMRCA) of Rice yellow mottle virus (RYMV; genus Sobemovirus) was calculated by a Bayesian coalescent analysis of the coat protein sequences of 253 isolates collected between 1966 and 2006 from all over Africa. It is inferred that RYMV diversified approximately 200 years ago in Africa, i. e. , centuries after rice was domesticated or introduced, and decades before epidemics were reported. The divergence time of sobemoviruses and viruses of related genera was subsequently assessed using the age of RYMV under a relaxed molecular clock for calibration. The divergence time between sobemoviruses and related viruses was estimated to be approximately 9,000 years, that between sobemoviruses and poleroviruses approximately 5,000 years, and that among sobemoviruses approximately 3,000 years. The TMRCA of closely related pairs of sobemoviruses, poleroviruses, and luteoviruses was approximately 500 years, which is a measure of the time associated with plant virus speciation. It is concluded that the diversification of RYMV and related viruses has spanned the history of agriculture, from the Neolithic age to the present. The mechanisms of evolution of plant viruses are being progressively unraveled [1]–[3], yet the timescale of their evolution remains an enigma. Even the order of magnitude is unknown [4]. Several viruses showed few genetic changes between isolates separated in space and time, sometimes for centuries [5]–[8]. In contrast, recent evidence from statistical analyses of sequences of dated isolates of Tomato yellow leaf curl virus (genus Geminivirus) [9], Rice yellow mottle virus (genus Sobemovirus) (RYMV) [10] and Zucchini yellow mosaic virus (genus Potyvirus) [11] indicated rapid evolution, similar to that of most animal viruses. The paradox is addressed here by calculating the divergence time of plant viruses at the intra- and inter-specific levels using RYMV and related viruses. Molecular-dating techniques provide insights into the history of lineages that have a poor or non-existent fossil record, such as viruses [12], [13]. These techniques were originally based on the assumption of a strict molecular clock reflecting steady accumulation of genetic changes over time. Recently, new methods enable the incorporation of variable rates into molecular dating [13]. Here, we applied a Bayesian Markov Chain Monte-Carlo method for performing relaxed phylogenies that is able to co-estimate phylogeny and divergence times under uncorrelated relaxed-clock models [14]. RYMV causes an emergent disease that was first observed in 1966 in Kenya. Since then, it has been reported in nearly all rice-growing countries of sub-Saharan Africa. RYMV is transmitted by coleopterous insects and is also disseminated abiotically. It has a narrow host range limited to wild and cultivated rices and a few related grasses [15]. There is no evidence of recombination between RYMV isolates [16], [17]. The rate of evolution of RYMV was recently evaluated using the coat protein (CP) sequences of 253 isolates collected between 1966 and 2006 from all over Africa [10]. The same group of sequences is analyzed here to assess the time of their most recent common ancestor (TMRCA), which is a measure of the divergence time of RYMV. The TMRCA was calculated by a Bayesian coalescent analysis of the sequences using several molecular clock and population genetic models [14]. Sobemoviruses infect both monocotyledonous and dicotyledonous plants, but the host range of each virus species is narrow and confined to a few plant species of the Poaceae or Fabaceae. Sobemoviruses are transmitted by beetle vectors, seeds and direct contact [18]. They share a common genomic organization, as found after re-sequencing some of the virus species [19], [20]. Ten sobemovirus species have been fully sequenced, nine of them are currently registered by ICTV [18] and a tentative one, Imperata yellow mottle virus (IYMV), was recently isolated from Imperata cylindrica in Africa [56]. Their genomes contain four open reading frames (ORFs). ORF1, located at the 5′ end of the genome, encodes a protein involved in virus movement and gene silencing suppression. ORF2 comprises two overlapping ORFs. ORF2a encodes a serine protease and a viral-genome-linked protein. ORF2b is translated through a -1 ribosomal frameshift mechanism through a fusion protein. It encodes the RNA-dependent RNA polymerase (RdRp). The coat protein gene (ORF4) is expressed by a sub-genomic RNA at the 3′ end of the genome. No evidence of recombination between sobemoviruses has been found either in phylogenetic [21], [56] or experimental studies [22]. The genus Sobemovirus is not assigned to a family. However, the RdRp of the sobemoviruses is phylogenetically related to that of the poleroviruses and enamoviruses (family Luteoviridae) [23], and to Poinsettia latent virus (PnLV), a putative polerovirus-sobemovirus hybrid [24]. Sobemoviruses, luteoviruses (family Luteoviridae) and dianthoviruses (family Tombusviridae) are more distantly related. The CPs of sobemoviruses are related to those of the necroviruses (family Tombusviridae), and the CPs of the poleroviruses to those of the luteoviruses [18]. Recombination events between ancestors of these genera are the likely causes of the present situation [25], [26]. Altogether, this led to the proposal of a “supergroup” to include these related genera [25]. The RdRp of the sobemoviruses also shows similarities with that of Mushroom bacilliform virus (MBV) (genus Barnavirus, family Barnaviridae) which infects mushrooms [27]. The divergence time of sobemoviruses was assessed from the full-length sequences using the age of RYMV under relaxed molecular clock models for calibration. The divergence time of the sobemoviruses with members of related genera was inferred from RdRp sequences with the same methodology. The time associated with plant virus speciation was assessed by calculating the TMRCA of closely related pairs of sobemoviruses, poleroviruses and luteoviruses. Collectively, these studies provide estimates of the diversification time of a plant virus species, the time associated with plant virus speciation, and the TMRCA of plant viruses of the same genus and of different genera. The intra- and inter-specific plant virus diversification was found to span the history of agriculture from the Neolithic age to the present. The estimates of the TMRCA of RYMV inferred from the 253 dated CP sequences were dependent on both molecular clock and demographic models. Models enforcing relaxed molecular clocks performed better than the strict clock model, whatever the population genetic model selected (Table 1). The average substitution rates ranged from 5. 1×10−4 to 12. 3×10−4 nucleotides (nt) /site/year among the models (data not shown). The highest marginal likelihood was obtained with the model implementing the relaxed uncorrelated exponential molecular clock and the exponential growth model. The Bayes Factor (BF) gave strong support to this model when compared to other clock and population models. Under this model, the average TMRCA of RYMV was 195 years and the substitution rate was 11. 7×10−4 nt/site/year. The median was 182 years. The highest density probability (HPD) interval ranged from 107 years to 308 years, with an approximate lognormal distribution of the estimates. Subsequently, a lognormal distribution with a lognormal mean of 5. 2 and a standard deviation of 0. 3 was applied as the prior distribution of TMRCA of RYMV for the upward calibration of nodes of sobemoviruses and related viruses. The two most divergent RYMV isolates, Ma10 and Tz202, were collected 5,000 km apart in Mali and Tanzania, respectively, and differed by 10. 5% in the full genome. They were subsequently referred to as isolates RYMV-1 and RYMV-2. The distribution of the estimates of the TMRCA of RYMV calculated from the dated CP sequences was taken as the prior of their divergence time (node 1 in all figures). The full sequences of these two RYMV isolates and of nine other sobemoviruses were considered (Table 2). A total of 4,798 characters was analyzed, 3,432 of them (72%) being parsimony-informative. The lognormal clock model performed better than the strict model (marginal likelihoods in loge units were −50237 and −50260, respectively), whereas the exponential model failed to converge. The deviation from the hypothesis of a strict clock was limited (coefficient of variation = 0. 23). The TMRCA and the substitution rates of the sobemoviruses under the lognormal and the strict clock models were close: 3,137 vs. 3,326 years, and 4. 0×10−4 vs. 3. 7×10−4 nt/site/year, respectively. The Yule speciation process and the constant population size coalescent model as tree priors yielded similar estimates. Among sobemoviruses, RYMV is most closely related to IYMV (node 2). The TMRCA of RYMV and IYMV was 1262 years (523–2248) (Figure 1). Cocksfoot mottle virus (CfMV, genus Sobemovirus) also infects monocotyledonous plants but without overlap in host or geographical range. CfMV is the species the most closely related to RYMV and IYMV (node 3). The TMRCA of CfMV, IYMV and RYMV was 2,317 years (921–3,929). The root height of all sobemoviruses (node 4) was 3,137 years (1,133–5,295). The divergence time of sobemoviruses and related viruses was assessed from the RdRp sequences (Table 2). Again, the distribution of the estimates of the TMRCA of RYMV calculated from the dated CP sequences was taken as the prior of the divergence time of RYMV-1 and RYMV-2 (node 1). A total of 2,199 characters were analyzed, 1,607 being parsimony-informative (73%). The model enforcing the lognormal clock model performed better than the strict model (marginal likelihoods in loge units were −29663 and −29677, respectively), whereas the exponential model failed to converge. Again, the deviation from the hypothesis of strict clock was limited (coefficient of variation = 0. 28) and the estimates were close. For instance, the basal root height under the lognormal and the strict clocks were 8,772 vs. 10,440 years, and the substitution rates were 3. 2×10−4 and 2. 8×10−4 nt/site/year, respectively. However, the HPD interval was wider with the lognormal model (2,929–15,671 years) than with the strict model (4,971–18,060 years), i. e. , a 1∶5. 4 ratio for the lognormal model vs. a 1∶3. 6 ratio for the strict clock model. The age of sobemoviruses (node 4) calculated on the full genome and on the RdRp sequences were similar (3,137 and 3,056 years, respectively) despite the difference in number of parsimony-informative characters considered (3,432 vs. 1,607 characters). The age of sobemoviruses calculated on the CP sequences was similar (2,884 years) although the dN/dS ratios of the RdRp and of the CP genes were 0. 18 and 0. 39, respectively, reflecting the differences in functional constraints operating on the two genes. The TMRCA of the sobemoviruses and MBV (node 5) was 4,418 years (1,480–8,092) (Figure 2). The divergence time of sobemoviruses, poleroviruses, and MBV (node 6) was 5,118 years (1,840–9,050). The root height of sobemoviruses, MBV, poleroviruses, and luteoviruses (node 7) was 8,772 years (2,929–15,671). The TMRCA of these viruses and Red clover necrotic mosaic virus (RCNMV) (genus Dianthovirus) was 9,059 (3,370–16,260) (node 8), a value not substantially different from node 7 (Figure 3). Several isolates of Subterranean clover mottle virus (SCMoV, genus Sobemovirus), which caused a disease restricted to southwest Australia, were fully sequenced [28]. The highest divergence between two isolates collected in 1991 and 1996, respectively, was 1. 2%. Accordingly, the divergence time of SCMoV was estimated to be 20 years with a HPD interval of 6–44 years, indicating a date of diversification between 1952 and 1990 (Figure 3). Southern bean mosaic virus (SBMV, genus Sobemovirus) and Sesbania mosaic virus (SeMV, genus Sobemovirus) differed by 31. 6% in their complete genome and thus are the two most closely related sobemoviruses (Figure 1). Their divergence time (node “a”) was 526 years (169–938). Cereal yellow dwarf virus CYDV-RPV and CYDV-RPS, two closely related poleroviruses, differed by 22% in their RdRp sequences (Figure 2). Their TMRCA was 531 years (180–1,018) (node “b”). Barley yellow dwarf virus BYDV-MAV and BYDV-PAS, two closely related luteoviruses, differed by 21. 1% in their RdRp. Their divergence time (node “c”) was 451 years (141–813). Altogether, the TMRCA of these closely related pairs of sobemoviruses, poleroviruses and luteoviruses ranged from approximately 450 to 550 years. The 253 RYMV isolates collected in 16 countries represent the diversity of the species [10], [17]. Accordingly, the TMRCA of these 253 isolates provides a reliable estimate of the divergence time of RYMV. By contrast, the 10 sobemovirus species probably underestimate the number of sobemoviruses in cultivated and wild plants [29]. However, theoretical studies indicated that numerous samples are not necessary to date old coalescent events. It was calculated that the coalescence time of a sample of 10 taxa was 90% of the expected coalescent time of the entire population [30]. Consequently, although calculated on a limited number of species, the TMRCA of sobemoviruses and members of related genera provide reliable estimates of their divergence times. Relaxed molecular clock models incorporate the rate variation among lineages in estimates of divergence time. Accordingly, any punctuated evolution, as might occur in species jump, should be accounted for in the relaxed clock models. Results from relaxed clocks should be evaluated in relation to those of strict clocks [31]. In our study, the lognormal relaxed clock model performed better at the inter-specific level than the strict clock model. However, the deviation from a strict clock model was limited. This explained why the TMRCA estimates under strict and relaxed clock models were close. There was, however, a 1∶3 ratio between the lower and upper bounds of the HPD intervals of the TMRCA of RYMV (308 and 107 years, respectively). The variance of this estimate, further enlarged after relaxation of the molecular clock assumption, accounted for the large HPD intervals of divergence times at the inter-specific level. However, the HPD of RYMV divergence time is still substantially narrower than those of the other plant viruses studied with dated sequences [9], [11]. This is likely to be due to the larger number of isolates used and the wider range of dates encompassed with RYMV. This could also reflect the fact that the RYMV isolates were collected, sequenced and analyzed by the same group of scientists, subsequently reducing the uncertainties associated with the use of data sets from various and heterogeneous sources. Assessing the divergence time of RYMV from dated sequences does not suffer from the limitations of alternative approaches. Measuring RYMV evolution rate from experimental studies or from old virus specimens was previously found to be inappropriate [10]. Applying epidemiological evidence is not adequate either. Symptoms of RYMV were first described in 1966, i. e. , 40 years ago, a value inconsistent with the 107 to 308 years of the HPD interval for RYMV diversification. This means that RYMV diversified decades before the disease symptoms were reported. It also suggests that RYMV caused epidemics long before it was recognized as a disease. The first report of symptoms should better be considered as a lower bound of virus diversification, i. e. , the minimum time since the virus diversified. Exceptions are viruses in localized and well-surveyed regions such as SCMoV in southwest Australia. From dated sequences, SCMoV diversification was estimated to occur between 1952 and 1990. This interval includes 1979, the year when the first symptoms were reported [32]. Biogeographical evidence to estimate divergence time can be misleading too. Madagascar was separated from mainland Africa approximately 100 millions years ago. The timescale of evolution of RYMV excludes the possibility that the divergence between isolates from Madagascar and from East Africa reflects vicariance events [33]. Altogether, the set of CP sequences of 253 dated isolates of RYMV currently provides the most reliable approach to date plant virus diversification. The divergence time of RYMV was approximately 200±100 years, whereas symptoms were reported for the first time in 1966 in East Africa [34] and in 1975 in West Africa [35]. The African rice Oryza glaberrima was domesticated in West Africa approximately 3,000 years ago, whereas the Asiatic rice O. sativa was introduced in the 10th and 16th centuries in East and West Africa, respectively [36], [37]. Consequently, RYMV diversified centuries after rice was domesticated or introduced in Africa, and decades before epidemics were reported. The 19th century was a period of extension of the rice culture in Africa [37]. This may have favored the spread of RYMV from is primary host to rice, followed by its dissemination throughout Africa. The divergence time between sobemoviruses and related viruses was estimated to be approximately 9,000 years, that between sobemoviruses and poleroviruses approximately 5,000 years, and that among sobemoviruses approximately 3,000 years (Figure 3). The estimates of the age of sobemovirus diversification did not depend on the sequence length or on the gene considered. Even considering their HPD, these time-scales encompassed the Neolithic “agricultural revolution. ” This period was the transition from nomadic hunting and gathering communities to agriculture and settlement. It occurred independently in several prehistoric human societies between 10,000 and 4,000 years before present (BP) [38], [39]. Ancient peoples completed the domestication of all major plant species upon which human survival depends ca. 4,500 years BP [40], [41]. One likely consequence of agricultural expansion is the dramatic increase of opportunities for encounters between wild and cultivated plant species, between cultivated plants at various stages of domestication, and between plants and potential insect vectors. These new encounters must have facilitated the emergence of plant viruses. This is still apparent nowadays when crop species are moved from their center of origin into new regions. They are exposed to infection by indigenous viruses to which they have not previously been adapted [4], [42], [43]. Further crowding of plants associated with agricultural development, especially monoculture, facilitated the build-up of vector populations and the disease spread, as is still apparent at the present time [43]. Similarly, the Neolithic age was critical for the emergence of infectious human diseases, a period referred to as the first epidemiologic transition [44]. This was attributed to the increased contacts between humans and wild fauna, and among humans themselves. Our results suggest that the Neolithic age was also a period of epidemiological transition for plant pathogens such as viruses, intrinsically for the same reason: increased contacts between hosts, pathogens and vectors. The hypothesis that the emergence of plant viruses is linked to the development of agriculture is consistent with the view that RNA viruses have a recent origin [12], and also that humans have become the world' s greatest evolutionary force [45]. The divergence time of the RdRp of sobemoviruses and poleroviruses bounded the dates of the recombination events between the genera. They must have occurred after the diversification of the common ancestor of the RdRp of sobemoviruses and poleroviruses approximately 5,000 years ago, and before the diversification of each of the two genera approximately 3,000 years ago. These recombination events, which necessarily involved the co-existence of different genomes in the same plant, must have been favored by the increased opportunities of co-infections associated with agricultural expansion that started during the Neolithic age. Events occurring at this period also possibly led to virus diversification outside the plant kingdom, as suggested by the divergence time of the sobemovirus and MBV estimated to be approximately 4,500 years. Much effort has been recently devoted to the numerical taxonomy of plant viruses to set thresholds in percentage of nucleotide divergence for demarcation criteria at the intra- and inter-specific levels [46]. In this study, nucleotide divergence illuminates the timescales associated with these demarcation criteria (Figure 3). The limited deviation from the strict clock model allowed the comparison of these timescales. The inter-generic divergence time between sobemoviruses, poleroviruses and luteoviruses exceeded approximately 3,000 years. The inter-specific divergence of sobemoviruses ranged from approximately 500 to 3,000 years. Consistent divergence times of approximately 500 years were obtained between closely related pairs of sobemoviruses, luteoviruses and poleroviruses, which were first considered as strains and later ranked as different species. This provides an estimate of the time associated with speciation of plant viruses. The intra-specific divergence time of RYMV was approximately 200 years, which is 2 to 3 times less than the speciation time of plant viruses. Overall, this range of values revealed that plant diversification at the intra- and inter-specific levels occurred within the Holocene, and has spanned the entire history of agriculture, from the Neolithic age to the present. The CP genes (720 nucleotides) of 253 isolates from 16 countries in Africa collected over a 40-year period, and the complete genome of two isolates of RYMV were previously sequenced [10], [17]. The complete sequences of the sobemoviruses, the sequences of the RdRp of the poleroviruses, luteoviruses, PnLV, and MBV were downloaded from GenBank (Table 1). The sequences were aligned using CLUSTAL W with default parameters [47]. The parameters of interest were estimated within a Bayesian coalescent framework by a Markov Chain Monte Carlo (MCMC) method using the Bayesian Evolutionary Analysis by Sampling Trees (BEAST) program (http: //beast. bio. ed. ac. uk/) [48]. The Bayesian MCMC method estimates a parameter as the mean of its posterior distribution while simultaneously incorporating uncertainty in the underlying genealogy or phylogeny and other parameters. The length and number of MCMC chains were chosen so that the effective sample size for the root height parameter and other parameters was >200, indicating that the parameter space was sufficiently explored. The convergence of the parameters to a stationary distribution was assessed with TRACER [49], and the statistical uncertainties were summarized in the 95% HPD intervals. Comparison of models was performed by calculating the Bayes Factor (BF), which is the ratio of the marginal likelihood of each model [50]. A value of loge (BF) >2. 3 was taken as evidence of a strong support for the model with the highest marginal likelihood. The coefficient of variation of the evolution rates calculated under the uncorrelated lognormal relaxed clock model was used to assess the degree of deviation from the strict molecular clock model. In earlier studies, the evolution rate was the target parameter [10], whereas here the TMRCA or the root height was the parameter of interest. It was taken as a measure of the divergence time of RYMV. The root height was estimated by enforcing strict and relaxed (uncorrelated lognormal and uncorrelated exponential) molecular clocks as implemented in BEAST [48]. Four demographic models were applied as coalescent priors: constant population size, exponential growth, expansion growth, and a piece-wise Bayesian skyline plot [49]. Default values were used for the other priors. The uncertainty in the TMRCA of RYMV is summarized by the highest posterior density interval that contains 95% of the marginal posterior distribution. The full sequences of 10 sobemoviruses were considered for the intra-generic analysis (Table 2). The RdRp sequences of related viruses were added for the inter-generic analysis. The total number of characters and the number of parsimony informative characters were calculated with PAUP [51]. The dN/dS ratios were calculated under the MG94 model [52] as implemented in Hyphy (http: //www. hyphy. org/) [53]. The poleroviruses listed by ICTV [23], Pea enation mosaic virus (genus Enamovirus) and PnLV were screened for recombination signals. Putative recombinant genomes were searched using the RDP3 package (http: //darwin. uvigo. es/rdp/rdp. html). It implements six recombinant detection programs: RDP, GENECONV, MaxChi, Chimera, Bootscan and Siscan [54]. The default detection thresholds were applied. Five poleroviruses showing no signals of recombination were subsequently selected: Beet chlorosis virus (BchV), Beet mild yellowing virus (BMYV), Potato leaf roll virus (PLRV), CYDV-RPS and CYDV-RPV (Table 2). Similarly, the RdRP sequences of two luteoviruses were chosen: BYDV-PAS and BYDV-MAV. The best-fitting nucleotide substitution model was evaluated by hierarchical likelihood ratio testing [55], as implemented in HyPhy [53]. The best-fitting model was the HKY model with gamma rate heterogeneity. The dates of isolation of the virus species were considered as contemporaneous as they differed by a few years only, whereas our study dealt with inter-specific divergence times ranging from hundreds to thousands of years. The maximum clade credibility tree was reconstructed by Bayesian inference under the relaxed molecular clock models as implemented in BEAST. A Yule speciation process was selected as a tree prior. The distribution of the estimates of the TMRCA of RYMV was subsequently used as the prior of the RYMV node for upward calibration of the nodes of the trees. The HPD intervals of the TMRCA of sobemoviruses and related viruses subsequently summarized both the uncertainties of the phylogenetic signal and of the prior (the RYMV age). A uniform distribution with bounds of 5×10−5 and 5×10−3 nt/site/year was applied as the prior of the uncorrelated lognormal relaxed clock mean. A similar prior was applied for the Yule speciation process birth rate. A uniform distribution with bounds of 0. 2 and 5 was applied as the prior of the gamma shape parameter. A Jeffrey prior with initial value of 1 was applied for the HYK transition-transversion parameter.

The timescale of the evolution of plant viruses is an enigma, and even its order of magnitude is unknown. This critical issue is addressed here by calculating the age of plant viruses. An accurate estimate of the age of Rice yellow mottle virus (RYMV) was obtained by statistical analysis of a set of dated sequences. The age of RYMV provides a reliable calibration of related viruses, applying recently developed relaxed molecular clock models. It was found that RYMV diversified approximately 200 years ago, and that inter-specific diversification ranged from 500 years to 9,000 years. Altogether, plant virus diversification has spanned the history of agriculture from the Neolithic age to the present. This suggests that the Neolithic was a period of epidemiological transition for plant virus diseases, as already proposed for infectious human diseases. Intrinsically, it is for the same reason: increased contacts between hosts, pathogens, and vectors. This is consistent with the view that RNA viruses have a recent origin, and that humans have become the world' s greatest evolutionary force.

Abstract Introduction Results Discussion Materials and Methods

virology/emerging viral diseases virology/virus evolution and symbiosis evolutionary biology/evolutionary ecology

2008

Diversification of Rice Yellow Mottle Virus and Related Viruses Spans the History of Agriculture from the Neolithic to the Present

The olfactory (OR) and vomeronasal receptor (VR) repertoires are collectively encoded by 1700 genes and pseudogenes in the mouse genome. Most OR and VR genes were identified by comparative genomic techniques and therefore, in many of those cases, only their protein coding sequences are defined. Some also lack experimental support, due in part to the similarity between them and their monogenic, cell-specific expression in olfactory tissues. Here we use deep RNA sequencing, expression microarray and quantitative RT-PCR in both the vomeronasal organ and whole olfactory mucosa to quantify their full transcriptomes in multiple male and female mice. We find evidence of expression for all VR, and almost all OR genes that are annotated as functional in the reference genome, and use the data to generate over 1100 new, multi-exonic, significantly extended receptor gene annotations. We find that OR and VR genes are neither equally nor randomly expressed, but have reproducible distributions of abundance in both tissues. The olfactory transcriptomes are only minimally different between males and females, suggesting altered gene expression at the periphery is unlikely to underpin the striking sexual dimorphism in olfactory-mediated behavior. Finally, we present evidence that hundreds of novel, putatively protein-coding genes are expressed in these highly specialized olfactory tissues, and carry out a proof-of-principle validation. Taken together, these data provide a comprehensive, quantitative catalog of the genes that mediate olfactory perception and pheromone-evoked behavior at the periphery. Olfaction is used for locating and discriminating between food sources, but also plays a fundamental role in social communication between individuals. Mice heavily rely on their sense of smell to distinguish between animals from their own and different species, and to determine their identity [1]. Additionally, upon detection of specific semiochemical cues, these animals show certain behavioral responses, many of which are stereotypical and have been well characterized [2]–[6]. The mammalian olfactory system is formed by the olfactory mucosa (OM) and the vomeronasal organ (VNO) and is dedicated to sensing odorants and pheromones present in the environment. These cues are detected via olfactory (OR), trace-amine associated (TAAR), vomeronasal (VR) and formyl-peptide (FPR) receptors expressed by the sensory neurons in the epithelia of these organs. The OM detects mainly airborne molecules while the VNO identifies both volatile and non-volatile compounds [7]. The importance of a finely tuned olfactory system is reflected in the amount of genes specialized for the detection of odorants and pheromones. In the most recent assembly of the reference mouse genome (GRCm38) over 1,200 genes are annotated as coding for ORs and around 530 for VRs with a smaller number of TAAR and FPR genes. Together they comprise almost 5% of the complete gene catalog. A large proportion of OR and VR gene repertoires have been identified through computational methods, based on homology searches to a few experimentally described reference receptors [8]. Accordingly most only have their protein coding sequences annotated in genomes. Indeed, for some of the genes annotated as VRs or ORs there is a complete absence of supporting evidence for them being expressed in the VNO or OM. ORs are also expressed in non-olfactory tissues, including the kidney, heart, lung, and testes [9], [10], where they have been shown to work as chemoreceptors in human sperm chemotaxis [11], [12]. This has raised the question of whether all OR genes encode true olfactory receptors [13]. Most olfactory sensory neurons express only one OR or VR gene, in a monoallelic fashion [14]. Consequently only a small proportion of cells in each epithelium express any given receptor, which makes their detection challenging. Furthermore, high levels of sequence similarity within OR, and particularly VR subfamilies, means it is very difficult to ensure specificity when using hybridization based detection methods [9], [15]. Among the behavioral responses elicited through olfactory signals, many are clearly distinct between adult male and female mice, including sexual conduct [3], [4], [6], aggressive responses to intruders [2], and parental care [16], but the mechanisms that ensure such differentiated responses have not yet been fully elucidated in mammals [17]. In silk moths, this is achieved by only males expressing the receptor BmOR-1 in their antenna, which detects the female-specific sex pheromone bombykol [18]. In contrast, the Drosophila sex pheromone cVA is detected by both sexes and elicits dimorphic behavior by routing the signal via different third order neuronal circuits deep in the brains of males and females [19]. Sexual dimorphism in pheromone receptor expression has been reported in rats [20], but the best defined mammalian example is the detection of the male-specific pheromone ESP1 by Vmn2r116, which is capable of eliciting lordosis behavior specifically in female mice. Mice of both sexes appear capable of detecting this pheromone, suggesting the differential response is due to modifications in the downstream neural circuitry [4]. To determine the full receptor repertoire expressed in the mouse VNO and OM, and assess whether sexual dimorphism in olfactory-mediated behavior can be explained by differential gene expression in these organs, we used RNAseq to profile their transcriptomes in male and female mice. We show that a very high proportion of the annotated receptors are indeed expressed in the olfactory system, we experimentally characterize their full length transcripts for the first time and compare their abundances to previous estimates. There are a few differences in expression between the two sexes but only very minor distinctions in the levels of the receptor repertoires. However, genome-wide expression analysis revealed a large number of novel genes in olfactory tissues and some inter-individual variation for subsets of genes. We conducted deep RNA-sequencing in whole VNO and OM of three adult male and three adult female biological replicates. Due to their small size, each VNO replicate was pooled from three genetically identical animals; each OM replicate was from a single animal. The VNO samples, composed of the sensory neuroepithelium, progenitor and non-neuronal supporting cells, underlying glandular and cavernous tissue and a blood vessel with blood [21], [22], yielded a mean of 37. 1 million (±3. 6 million) short paired-end fragments per sample (Table 1). The whole OM samples, including the main olfactory epithelium and underlying lamina propria, non-neuronal supporting cells, glandular tissue and blood vessels with blood [23], yielded 46. 4 million (±4. 3 million) fragments on average (Table 1). From these, approximately 84% were mapped unambiguously to the genome. To estimate expression levels, we counted the number of uniquely mapped fragments assigned to each annotated gene. We then normalized to account for the length of the gene and the depth of sequencing to obtain FPKM values (fragments per kilobase of exon sequence per million fragments) [24]. The expression estimates for all genes in each replicate are listed in Datasets S1, S2. We first assessed the variation in gene expression among the three biological replicate samples for each sex and tissue; the correlation values were highly significant between them all (Spearman' s rank correlation coefficients of at least 0. 95, p-value<2. 2×10−16; Figure S1). Only small sets of genes are unusually variable among replicates (Figure 1A–B) and the distribution of gene expression is very similar between males and females (density plots in Figure 1C). We therefore averaged the FPKM values for each gene in each sex and tissue. In both tissues a few genes are extremely highly expressed. For example, in the VNO the 14 most abundant genes account for almost 50% of the fragments obtained from the whole tissue. The highest, Lcn13, has an average expression of around 97,300 FPKM, but more than 85% of the genes have values below 10 FPKM. A similar distribution is observed in the OM, though less extreme. The most abundant gene, Bpifb9b, is expressed at about 22,300 FPKM and the top 261 genes account for half the total expression; again, the overall majority of genes (83. 9%) are expressed below 10 FPKM. A total of 10,552 (28. 35%) and 9,881 (26. 54%) genes in the VNO and OM respectively have no fragments mapped in any replicate suggesting they are not expressed in that tissue. The expression of the remaining genes shows a bimodal distribution of low- and high-expressed genes (density plots in Figure 1C), characteristic of RNAseq datasets [25]. These can be decomposed into two normal-like overlapping distributions, and each gene can be assigned to either distribution with a degree of confidence. Low-expressed genes typically do not have active chromatin marks, are enriched in non-functional mRNAs and, unlike the high-expressed genes, lack correlative protein expression data [25]. We therefore focused our analysis of differential gene expression on those genes that have at least a 25% probability of being within the highly-expressed distribution: 17,698 genes in the VNO and 17,983 in the OM (see Materials and methods for details). Among the 19,579 genes that are expressed in either tissue, 63. 14% (12,363) are differentially expressed with a false discovery rate (FDR) of less than 5%. To explore these further, we selected the genes showing a fold change of four or higher and searched for enriched functional terms. As expected, those expressed higher in the VNO are enriched for VR genes, which are involved in the response to pheromones, odorant-binding and lipocalin-related proteins. Additionally, the calcium signaling pathway, ionic and voltage-gated channel activity, regulation of blood pressure and the immune response are significantly enriched. For the OM, enriched genes are dominated by those encoding ORs and, those involved in the olfactory transduction pathway and sensory perception. In addition, there is enrichment of ionic and ligand-gated channels. In contrast, ‘housekeeping’ genes are expressed at similar levels in both tissues (scatter plot in Figure 1C). Another widely used technology to profile gene expression levels is microarrays. For comparison with our RNAseq data we used commercial Illumina expression microarrays to profile six more biological replicate VNO and OM samples. For both tissues, the overall expression values are correlated (Spearman correlation = 0. 71 for the VNO, 0. 72 for the OM, p-value<2. 2×10−16). However, the microarray intensity values reach saturation for the highly expressed genes while the RNAseq values keep increasing over two orders of magnitude (Figure 2A–B). Quantitative real time PCR (qRT-PCR) is accepted as the gold standard for expression profiling, so we next compared both our RNAseq and the microarray expression estimates to a panel of qRT-PCR TaqMan gene expression assays (Figure 2C–F). We included genes with and without a known function in olfactory and vomeronasal signaling that cover the whole range of expression values observed (Table S1). The correlation is considerably higher with the RNAseq data (Pearson correlation r2 = 0. 81 for the VNO and 0. 9 for the OM) than with the microarray values (Pearson correlation r2 = 0. 58 for the VNO and 0. 52 for the OM), indicating that RNAseq is better suited for transcriptome profiling in the olfactory system. Furthermore, the strong correlation between the qRT-PCR and the RNAseq data gives us confidence that these expression estimates are reproducible and specific, and provide a comprehensive characterization of olfactory transcriptomes. To investigate whether sex-specific responses to olfactory cues can be accounted for by transcriptional differences in the VNO and OM, we searched for sexually dimorphic gene expression patterns. We found that the overall transcriptomes are very similar between males and females. In the VNO 282 genes (1. 59% of all expressed) are differentially expressed by sex at a 5% FDR. In the OM, only 81 genes (0. 45%) reach statistical significance (Figure 3). Furthermore, just 51 and 34 respectively show log2-fold changes greater than 2, while the remaining show very slight deviations towards one sex. Only 11 genes are sexually dimorphic in both olfactory tissues. Among these are genes expected to be differentially expressed by sex, such as the X-inactive specific transcript, Xist, and four genes on the Y chromosome (Kdm5d, Ddx3y, Eif2s3y, Uty). The differential expression analysis between males and females for all genes in each tissue is provided in Dataset S3. We noted that 110 (39. 0%) and 45 (45. 5%) of the genes identified as sexually dimorphic in the VNO and OM respectively show unusually high variance (at least a three-fold difference) between any two replicates of the same sex, a likely contributing factor to the dimorphism. Moreover, some subsets of these genes had very similar patterns of variation. For example, a group of eight lipocalins (six of which are significantly dimorphic) all showed at least a 130 fold increase in abundance in one male VNO sample over the two other replicates (Figure 3A, S2A). All other lipocalins do not (Figure S2B), suggesting this variation is unlikely to be a consequence of sample contamination. Due to the small size of the organ, the VNO samples we sequenced were pooled from three mice. Therefore we next extracted RNA from the VNO of 15 individual group-housed males and assessed the expression of four of the most variable lipocalin genes by TaqMan qRT-PCR. Two thirds of the animals showed equivalent expression values but the remaining five had increased expression levels up to ten fold higher than the mean expression across all animals (Figure S2C). Consistent with the RNAseq data, the expression dynamics across each individual animal was the same for the four lipocalins. The monogenic expression of receptors in olfactory sensory neurons means each individual receptor is expressed in only a small subset of cells. Therefore the expression of any given receptor within the whole epithelium is low and this has hindered their study in a comprehensive manner. The GRCm38 mouse assembly contains 1,249 annotated OR genes and 530 VR genes. To ensure that these represent the complete repertoires, we took the cDNA sequences for the mouse VR genes as previously reported [26], [27], and aligned them to the genome with BLAST. We recovered 35 Ensembl genes that were not annotated as a VR gene, but that perfectly matched a VR cDNA sequence. These were included in subsequent analyses (Table S2). A similar procedure performed with all OR genes annotated in Ensembl provided four additional genes that have high identity alignments but that had not been annotated. We first analyzed the overall expression distribution for each class of receptors in their cognate tissue. In both cases the receptors in the repertoire do not have equal abundances, as may be expected if receptor choice was a random process. Instead we observe a large dynamic range of expression: a few receptors are expressed at high levels and the vast majority of the receptors are expressed at relatively low levels. For the VR genes, the most highly expressed receptor, Vmn2r89, has a value of 131. 86 FPKM. 42 receptors are expressed above 20 FPKM and the median expression for V2R genes is 0. 66 FPKM with 0. 45 FPKM for V1R genes (Figure 4, S3A). 416 VR genes (77. 6%) have at least one fragment mapped uniquely. From the other 120,82 are pseudogenes, and 61 have reads that map to several genes (also called multireads). 59 VR genes have no mapped fragments, either unique or multi-mapped, but these are all annotated as pseudogenes. In the case of the OR genes the most abundant, Olfr1507, is expressed at 87. 54 FPKM, and 11 genes are above 20 FPKM. The median expression is 0. 95 FPKM (Figure 5, S3B). Despite their relatively low abundance, 1,180 (94. 48%) of all the OR genes have at least one fragment mapped to their exonic region. Of the remaining 69 genes, 50 are annotated as pseudogenes and 16 have multireads that could indicate expression. We found only 9 putatively functional OR genes that have no evidence of expression in the OM whatsoever (Olfr115, Olfr141, Olfr504, Olfr564, Olfr574, Olfr834, Olfr1053, Olfr1061, Olfr1367). Importantly, the expression estimates for both OR and VR genes are consistent between biological replicates suggesting the uneven distribution we observe is stereotypical. We identify no overt pattern in the expression of either VR or OR genes based on cluster or genomic location (Figure 4,5). We next asked if any VR genes are expressed in the OM, or if OR genes are found in the VNO, as these may be indicative of specialized olfactory circuits [28]. We found one VR gene, Vmn2r29, is expressed in the OM at a level that is higher than the median OR gene expression (1. 04 FPKM). This is consistent across all six replicates, suggesting there may be a previously unrecognized mechanism of pheromone detection in the OM (Figure S4). In the case of the VNO, 17 OR genes are expressed higher than the VR gene median (Figure S5), with Olfr124 as the highest (14. 9 FPKM) followed by Olfr692 and Olfr1509 (7. 4 and 3. 1 FPKM respectively). Both Olfr124 and Olfr692 consistently display higher FPKM values in the VNO than the OM. In 2012, Plessy et al. reported the expression of 955 OR genes using nanoCAGE, a methodology that captures the 5′ end of transcripts and generates short sequence reads around that region [29]. Additionally, Khan et al. (2013) profiled 531 OR genes using NanoString nCounter [23]. Both used C57BL/6 sub-strains of mouse, allowing direct gene-level comparison with the data reported here. The NanoString counts are consistent with RNAseq expression estimates (Spearman correlation = 0. 81; Figure 6A). The agreement of these two technologies, which are based on very different detection principles, provides support for the accuracy in the quantification of expression of these genes by RNAseq. However, our receptor gene data is only moderately similar to nanoCAGE (Spearman correlation = 0. 38; Figure 6B) and the abundance estimates of the OR and VR genes represented on the Illumina microarrays (Spearman correlation = 0. 54 for OR and 0. 29 for VR genes; Figure S6). An advantage of RNAseq over the other expression profiling techniques described here is that it is not restricted to a catalog of known transcripts. We therefore used Cufflinks to generate de novo assemblies from the sequencing reads in order to identify full length transcripts for OR and VR genes [30]. We sequenced at sufficient depth to produce new, extended receptor gene models for 913 (73. 1%) OR and 246 (45. 9%) VR genes (the models and their sequences are provided in Datasets S4, S5, S6, S7). We identified additional exons for many of the receptor genes: 866 and 68 OR genes have exons 5′ and 3′ to the coding sequence, respectively; and 163 and 79 VR genes have exons 5′ and 3′ to the coding sequence (Figure 7A–B). OR and V1R genes typically have coding regions that span a single exon, but we identified 54 OR and 15 V1R genes where at least one of the reconstructed transcripts has an intron within the protein coding sequence (as annotated in Ensembl). The predicted open reading frames (ORFs) for most of these transcripts are truncated, due to a premature stop codon. But for 17 OR and 3 V1R genes the ORF is of typical length, and could encode a putatively functional receptor. All but one (Olfr332) of these gene models are reported in Ensembl and classified as protein coding (Dataset S8). We investigated cases of alternative splicing by retaining all the multi-exonic receptor gene models and counted the number of alternative isoforms produced. 70% of VR genes have between 1 and 4 isoforms while 85% of OR genes have 1 to 3 isoforms (Figure S7A). A few receptor genes have more than 8 different isoforms (38 VRs and 10 ORs), however in most of these cases this is due to the presence of several transcription start sites (TSS) or exons that differ in length by just a few nucleotides, so several of the final transcripts differ only very slightly (Figure S7B). We next calculated the length for each receptor gene based on the existing Ensembl and our new reconstructed models. The median length for both the Ensembl OR and V1R gene models is about 950 nucleotides, while the corresponding reconstructed gene models are now around 2,500 nt long. The median length of Ensembl V2R genes is 2,559 nt, while for the V2R reconstructed gene models it is 2,912 nt (Figure 7C). The lack of experimentally validated UTRs has been a major hindrance for the design of hybridization probes to discriminate between highly similar OR and particularly VR transcripts [9], [15], [31], [32]. We therefore assessed whether our new gene models will help resolve this by determining the proportion of each gene sequence that is unique in the genome. We find a large increase in the proportion of unique sequence in our new extended V1R (P<0. 0001, Mann Whitney test) and V2R gene models (P<0. 0001, Mann Whitney test); a more modest increase is apparent in OR genes (P = 0. 044, Mann Whitney test; Figure 7D). We next compared the 5′ ends of the OR gene models reconstructed here using Cufflinks, to the proposed transcription start sites (TSS) reported by Plessy et al. (2012) using nanoCAGE [29]. A third of the ORs differ in less than 20 nucleotides, and almost 85% are within a 500 nucleotide window (Figure 6C). However 34 OR genes have a discrepancy of more than 5 kb, where the 5′ end proposed by nanoCAGE is upstream of the one found by Cufflinks. We closely examined the sequencing data for the 25 genes with the biggest 5′ differences (Figure 6D). For 24 genes, we were unable to find any sequencing fragments consistent with the TSS proposed by nanoCAGE. In 12 of these cases, the nanoCAGE TSS overlaps with the 3′ UTR of an adjacent OR gene and 2 actually represent the TSS of a different gene (Figure 6D). Only one TSS is correctly inferred by nanoCAGE, where Cufflinks failed to reconstruct the full-length model. Examples of these different scenarios can be found in Dataset S9. Clowney et al. (2011) also defined the 5′ end of OR genes using tiling microarrays [33]. We similarly compared the 5′ ends in our reconstructed models to these data and found that a third of the receptor genes differ in less than 100 nucleotides and 80% of the data is contained within a 1. 5 kb window (Figure S8). We extended the analysis of the de novo assembly performed by Cufflinks to the full olfactory transcriptomes. This revealed 5,562 and 6,228 loci that have evidence of transcription in the VNO and OM respectively, that do not overlap any annotated genes in the Ensembl database. 40% of these loci are found in both tissues (Table 2). Many of these are located in close proximity to the start or end of annotated genes and are likely to represent unannotated UTRs. Therefore to search for new genes we first excluded all those predictions that lie within 5 kb of cataloged genes. Of the remaining features, about 75% represent single-exon transcripts, leaving 756 and 847 putatively novel multi-exonic genes expressed in the VNO and OM respectively. The genomic coordinates of these are provided in Datasets S10. We cross-referenced these loci with the Ensembl databases to search for alignments to known protein features or overlaps with computationally predicted transcripts. About 30% of these putative genes have known protein domains and 80% lie within transcripts predicted in silico. Finally, we sought to validate a selection of these putative genes experimentally. We focused on a de novo six exon transcript that is extremely highly expressed in the VNO (the 6th most abundant in the transcriptome) and a second, less abundantly expressed novel transcript located adjacent to it in the genome that has OM expression. We cloned full–length transcripts from both these genes and identified ORFs on opposite strands that encode two closely related proteins (Figure 8A). We identified these as novel members of the lipocalin gene family, and named them Lcn16 and Lcn17. A phylogeny of all mouse lipocalins reveals these genes form a distinct sub-clade (Figure S9A), and in situ hybridization analyses confirm Lcn16 is expressed abundantly in glandular tissues of the VNO (Figure 8B, S9B), while Lcn17 is expressed in a small number of cells in the main olfactory epithelium (Figure 8C, S9C). We compared the transcriptomes from VNO and OM of both male and female mice to assess whether differences in gene expression in these tissues could underpin sexually dimorphic behaviors [17]. One mechanism could be to differentially regulate the molecular components involved in cue detection, such as ORs or VRs [20] or in known elements of their signal transduction pathways, under the control of sex-specific hormones. We identified 9 VR genes and 2 OR genes that were significantly more abundant in one gender, but most of these displayed only marginal differences. Only one receptor, Olfr1347, had a fold-change in FPKM greater than 2 and none of the other dimorphic transcripts we identified are known to be involved in olfactory or vomeronasal neuron signal transduction. Therefore, we consider it unlikely that the striking dimorphic behavioral responses to some mouse semiochemicals [2], [4], [6], [34] can be solely accounted for by transcriptional differences at the level of detection. It remains to be elucidated whether differences in translation and/or protein modification of receptors or signal transduction machinery underlie sexually dimorphic detection of pheromones. Alternatively, both sexes may detect all mammalian olfactory signals equally, but interpret them differently due to sexually dimorphic central circuits [35]. A high proportion of the most abundantly expressed genes in our datasets, including S100a5, Obp1a, Obp1b, Lcn3, Lcn4, Mup4, Mup5, Dmbt1, Bpifa1 and 5430402E10Rik, have been previously detected in olfactory tissues using other methods [36]–[40], but a major benefit of RNAseq is that it permits novel gene discovery. Olfactory tissue transcriptomes are likely to be a rich source of novel genes for three reasons. Firstly, they are not widely used in transcription based gene discovery and annotation pipelines. Secondly, olfactory organs tend to be enriched in specialized genes with highly restricted expression patterns. Third, genes involved in pheromone detection are often species-specific and functional orthologues are typically lacking in the human genome, which confound their detection by comparative genomic methods [22]. We found a surprisingly high number of novel transcripts that map some distance away from known genes, and encode consistent multi-exonic gene models. Over 200 of these have protein features, suggesting they are indeed novel genes. As a proof of principle we cloned two, Lcn16 and Lcn17, which encode new members of the lipocalin protein family. Consistent with other lipocalins expressed in the VNO Lcn16 is extremely abundantly expressed in acinar cells of the vomeronasal gland [37], [39], while Lcn17 is expressed in cells of unknown function that are scattered throughout the main olfactory epithelium. Orthologous ORFs for Lcn16 and Lcn17 are found in the same orientation in the rat genome, but synteny is disrupted around this location in the primate lineage and there are no orthologues present in primates or the human genome. A major goal of our study was to investigate the expression of all the OR and VR genes in parallel. To do so requires a technology that is both sensitive enough to detect highly diluted signals and that is capable of distinguishing between very similar paralogues. In an early attempt to characterize the expression of the receptor repertoire, Young et al. screened a cDNA library constructed from the olfactory epithelium, using degenerate olfactory receptor probes, and identified the expression of 419 distinct ORs [41]. This approach, however, suffers from biases in the library construction and screening which hinders the identification of certain classes of receptors. High-density oligonucelotide arrays were designed to target the computationally predicted 3′ UTRs of OR and VR genes and probe the expression of all receptors annotated in an early genome assembly [9], [31]. Expression was confirmed for probes against 817 OR and 266 VR genes. Unfortunately these studies used a different strain of mouse and/or the gene-level expression data is not publically available, thus we were unable to compare those abundance estimates with the data reported here. To address this we used a commercially available expression microarray to estimate abundances. Compared with RNAseq we found that microarrays suffer from high levels of noise, possibly due to non-specific hybridization, and reach saturation with highly expressed genes [42]. We were able to detect expression above threshold for only 39. 8% of the 1107 OR genes present in the microarray and for 57. 4% of the 197 VR genes represented in the array, consequently there are only moderate correlations between microarray and RNAseq receptor abundance estimates. Surprisingly, we also found that previously published nanoCAGE estimates of OR gene expression correlated poorly with our RNAseq data [29]. This is partly because some 5′ nanoCAGE tags were apportioned to the wrong OR transcript (Figure 6D), but other factors may also contribute to the disparity. The mice used by Plessy et al. were younger than used here and their tissue was collected by laser capture microdissection which could result in incomplete sampling of the whole epithelium. Moreover, nanoCAGE tags that mapped to multiple locations were distributed by algorithm, while we took a more conservative approach and did not include multi-mapped reads in abundance estimates. More recently, NanoString nCounter technology has been used to detect OR expression in the OM [15], [23]. Probes could be designed to only approximately half of the predicted OR gene repertoire with confidence, resulting in expression quantification values for 531 OR genes in whole olfactory mucosa. NanoString nCounter is a hybridization probe based method, thus relative measures of abundance between different OR genes are not necessarily accurate. Nevertheless, we found OR gene expression estimates using this very different technology were consistent with our RNAseq data, lending support to both methods. We obtained evidence of expression for all putatively functional VR genes and all but 9 potentially functional OR genes by RNAseq. We cannot rule out the possibility that these may be expressed at levels below the threshold of detection in our experiments. Indeed one (Olfr504) is present in the NanoString dataset where it was reported to be expressed, albeit at a low level [23]. However, some ORs are known to be ectopically expressed in mice [10], [43], [44], thus it is possible they may have evolved extra-olfactory functions. Alternatively, they could be expressed at a different age [23], or they may be cryptic pseudogenes that have disrupted promoter elements and thus are no longer recognized by the machinery regulating olfactory receptor choice. Supporting this is our observation that approximately one third of both OR and VR genes with interrupted ORFs are not expressed in olfactory tissues, a bias that had been noted previously [41]. Experimental disruption of the ORF of an OR allele does not ablate its expression, instead another OR allele is co-expressed in the same cell [45], [46]. However, this phenomenon clearly occurs less frequently with naturally occurring pseudogenes, which probably reflects a parallel degeneration of their regulatory sequences. By comparing the promoter sequences of expressed with non-expressed OR and VR genes and pseudogenes, it may now be possible to identify key genomic motifs that control receptor choice. The generation of RNAseq data for a majority of ORs and VRs enabled us to obtain new, significantly extended gene models. The vast majority of receptor genes contain several exons and it is common to observe differential inclusion of these, diversifying the transcript set produced from each gene. In most cases the putative coding sequences of OR and V1R genes span a single exon and the additional exons contains only UTRs. In these instances the functional consequence of alternative splicing is unclear, as the same receptor protein would be generated from each transcript. In other cases alternative transcripts have introns interrupting the coding sequence, resulting in a truncated ORF that is likely to encode a non-functional receptor. However 1. 5% and 1. 3% of OR and VR genes, respectively, generate transcripts that can theoretically encode multiple, putatively functional receptor proteins. Further work will be necessary to verify the existence of transcripts, and determine their functional consequences. Most receptors are annotated from comparative genomic studies [8], which means non-coding UTRs are frequently missing. Identifying UTRs is especially useful for VR and OR genes, since they provide additional sequence that is typically more divergent than the ORF. When we compare the complete gene models we have reconstructed with those currently annotated, both the amount of the receptor transcript sequence, and the proportion that is unique between receptors increases substantially. We anticipate this resource will permit the design of more specific probes for NanoString nCounter, TaqMan qRT-PCR, microarray or in situ hybridization and thus increase the utility of these techniques in the olfactory system. The use and care of animals used in this study was approved by the Wellcome Trust Sanger Institute Animal Welfare and Ethics Review Board in accordance with UK Home Office regulations, the UK Animals (Scientific Procedures) Act of 1986. All mice used were C57BL/6J, 8 to 10 weeks old and group housed. The VNO was dissected from nine male and nine female animals and the tissue from three animals was pooled to obtain 5 ug of RNA for each biological replicate. Each OM sample was obtained from a single animal. RNA was extracted using the RNeasy mini kit (Qiagen) with on-column DNAse digestion, using a disposable RNAse free plastic grinder to homogenize the sample. All RNA was subsequently quantified with a spectrophotometer and visualized for quality by RNA integrity analysis. mRNA was prepared for sequencing using the TruSeq RNA sample preparation kit (Illumina) with a selected fragment size of 200–500 bp. The VNO samples were sequenced on the Illumina Genome Analyzer II platform and the OM samples on an Illumina HiSeq 2000; both generated 75 bp paired-end reads. Using STAR 2. 3 [47], sequencing reads were aligned to the GRCm38 mouse reference genome, annotation version 68 of the Ensembl mouse genome database (http: //jul2012. archive. ensembl. org/info/data/ftp/index. html). The number of fragments aligned to each gene was counted using the HTSeq package with the script htseq-count, mode intersection-nonempty. Any reads that map to multiple locations in the genome (also called multireads) are not counted towards the expression estimates since they cannot be assigned to any gene unambiguously, but these provide evidence of transcription in at least one of the loci to which they map. To compare the expression values across genes and conditions, raw count data was transformed into fragments per kilobase of exon per million fragments (FPKM) with the formula: We assessed GC-content biases in our RNAseq data as previously described [48], but observed no correlation between GC-content and fold-change in a differential expression test. Therefore, the FPKM values were not adjusted. Plotting, linear regression and computation of the Pearson' s and Spearman correlation was carried out in the R environment (http: //www. R-project. org) and sequencing reads were visualized using IGV [49]. RNA was extracted from the VNO and OM of six C57BL/6J males of 10 weeks of age as previously described. Profiling was performed on the Illumina MouseWG-6 v2. 0 Expression BeadChip following the manufacturer' s instructions. Variance stabilizing transformation was applied to the data obtained from BeadStudio, which was then quantile normalized using the Bioconductor R package, lumi [50]. We are aware that some of the receptor genes are not properly annotated in the Ensembl transcriptome, but are reported as novel genes. To recover the entirety of the VR gene repertoire, we took the cDNA sequences as reported [26], [27] and locally aligned them to the mouse genome with BLAST. Then we identified those alignments that overlap genes not annotated as VRs with 100% identity, and changed their name while preserving the Ensembl identifier. In all cases the coordinates obtained from the alignments were concordant with the annotation. A list detailing the gene names that were changed is reported in Table S2. Furthermore, 19 additional predicted genes have high identity alignments to other VR sequences. Similarly, we aligned with BLAST all the OR cDNA sequences present in Ensembl and recovered four predicted genes that share high similarity to other ORs. To compare the expression estimates form RNAseq and microarrays to those from qRT-PCR, RNA from OM and VNO was extracted, as previously described, from four individual male and four individual female C57BL/6J mice. Predesigned TaqMan gene expression assays were selected to target genes across the full range of expression values obtained by RNAseq (Table S1). They were used on a 7900HT Fast Real-Time PCR System (Life Technologies) according to the manufacturer' s instructions. To test for correlation between technologies, mean cycle threshold (Ct) values were obtained from three technical replicates and each normalized to Actb and Eef1a1 expression (chosen because of its similar abundance in both OM and VNO) using the ΔΔCt method. Relative quantity (RQ) values were calculated using the formula RQ = 2−ΔΔCt. For validating the inter-individual variation in lipocalin genes, the mean Ct values were obtained from two technical replicates and each normalized to Actb using the ΔCt method. RQ values were calculated using the formula RQ = 2−ΔCt. The overall distribution of expression values obtained from RNAseq data is bimodal. It has been proposed that such distribution is the combination of two normal-like distributions of low- and high-expressed genes [25]. Gaussian mixture models can be used to estimate such underlying normal distributions. We used the expectation-maximization algorithm provided in the Mixtools Bioconductor package [51], using all genes with at least one fragment count in one replicate, for each tissue. For both transcriptomes, the algorithm converged to optimal values and two distributions were fitted. The algorithm reports, for each gene, its probability of being part of either distribution. Based on this, we arbitrarily considered genes to be expressed if they had a 25% or greater probability of falling in the distribution containing the highly-expressed genes. To test for differential expression between sexes and tissues, we used DESeq [52] on the genes defined as expressed. Variation between replicates was calculated with the function estimateDispersions, using per-condition as the method. Genes were considered to be differentially expressed if they had an adjusted p-value of 0. 05 or less (equivalent to a false discovery rate of 5%). The Database for Annotation, Visualization and Integrated Discovery (DAVID) was used to search for enrichment of functional terms and biological processes [53]. In all analyses a background was provided, comprising only the expressed genes used for the relevant analysis. We considered significant those with an adjusted (false discovery rate) p-value smaller than 0. 1. To search for unannotated genes we performed Reference Annotation Based Transcript (RABT) Assembly [30], using Cufflinks v2. 1. 1 guided by the Ensembl annotation (version 68). Assembled transcripts from the different replicates were combined with Cuffmerge. In order to extract the candidates with greatest probability of encoding protein coding genes, we cross-referenced all predicted loci to the Ensembl databases using the API [54]. Ad hoc perl scripts were used to further refine the gene models produced for VR and OR genes, deleting those predictions that fuse adjacent receptor genes or that are antisense to the annotated gene. Full length transcripts of Lcn16 and Lcn17 were obtained using 5′ and 3′ RACE kits (Invitrogen) on RNA from VNO and OM tissue following the manufacturer' s instructions. All genes with a lipocalin domain (IPR000566) were extracted from Ensembl version 68 and a phylogeny was reconstructed in MEGA using the neighbour-joining method with the Kimura-2 parameter model of substitution [55]. DNA corresponding to the probe-specific regions were synthesized and integrated in pIDT (Integrated DNA Technologies). Templates for in situ hybridization probes were amplified by PCR from those plasmids by using the forward primers and reverse primers with or without a T3 promoter site (TATTAACCCTCACTAAAGGGAA) attached to their 5′ end. The primers used were: Lcn16_fw, TGACCATAAGCCTGACCGTG; Lcn16_rv, AGTGCCACATCCACAGAGTG; Lcn17_fw, TTACCCCACTGCCTCCCTTT; Lcn17_rv, TTGTTGGCGTTGGTGCCATA. Digoxigenin (DIG) -labeled sense and anti-sense probes for Lcn16, and fluorescein (FLU) -labeled sense and anti-sense probes for Lcn17, were synthesized according to the manufacturer' s instructions (Roche). Adult mice (C57BL/6J, 8 weeks old) were anesthetized, and then perfused with 4% paraformaldehyde (PFA). Snouts were dissected out and, post-fixed at 4°C for 2 hours in 4% PFA, then decalcified for another 72 hours by immersion in a 50∶50 mixture of 4% PFA and 0. 5 M EDTA. This was followed by immersion in 30% sucrose for 16 hours. The snouts were then embedded in TissueTek O. C. T. (Sakura), and frozen at −20°C. 14 µm cryosections were thaw mounted onto Superfrost Plus slide glasses (Thermo), dried at 55°C for 2 hours, and kept at −20°C until use. Hybridizations were performed overnight at 58°C using standard protocols [56]. Probes were visualized with the direct TSA Kit (FITC or Cy3, Perkin Elmer), or HNPP-Fast Red (Roche), according to the manufacturer' s instructions. Slides were mounted with VectaShield (Vector). Sections were observed and photographed with a Leica DM 400B fluorescent microscope, attached to an Olympus DP72 camera. RNAseq data are available in the European Nucleotide Archive (ENA) under accessions PRJEB2572 and PRJEB1365. Microarray data are available in the ArrayExpress database under accession number E-MTAB-2163. The sequences for Lcn16 and Lcn17 are available in GenBank under accessions KJ004569 and KJ004570.

The sense of smell in mice involves the detection of odors and pheromones by many hundreds of olfactory and vomeronasal receptors. The genes that encode these receptors account for around 5% of the whole gene catalog, but they are poorly understood because they are very similar to each other, and are thought to be turned on randomly in only a small number of cells. Here we use multiple gene expression technologies to curate and measure the activity of all the genes involved in the detection of odors and find evidence of many new ones. We show that most genes encoding olfactory and vomeronasal receptors have complex, multi-exonic structures that generate different isoforms. We find that some receptors are consistently more abundant in the nose than others, which suggests they are not turned on randomly. This may explain why mice are particularly sensitive to some odors, but less attuned to others. We find that overall males and females differ very little in gene expression, despite having altered behavioral responses to the same odors. Thus diversity in receptor expression can explain differences in odor sensitivity, but does not appear to dictate whether sex pheromones are differentially detected by males or females.

Abstract Introduction Results Discussion Materials and methods

genome expression analysis neuroscience animal models model organisms genome analysis bioassays and physiological analysis molecular genetics research and analysis methods gene expression mouse models microarrays gene identification and analysis olfactory system transcriptome analysis genetics biology and life sciences sensory systems genomics computational biology

2014

The Olfactory Transcriptomes of Mice

Host restriction factors constitute a formidable barrier for viral replication to which many viruses have evolved counter-measures. Human SAMD9, a tumor suppressor and a restriction factor for poxviruses in cell lines, is antagonized by two classes of poxvirus proteins, represented by vaccinia virus (VACV) K1 and C7. A paralog of SAMD9, SAMD9L, is also encoded by some mammals, while only one of two paralogs is retained by others. Here, we show that SAMD9L functions similarly to SAMD9 as a restriction factor and that the two paralogs form a critical host barrier that poxviruses must overcome to establish infection. In mice, which naturally lack SAMD9, overcoming SAMD9L restriction with viral inhibitors is essential for poxvirus replication and pathogenesis. While a VACV deleted of both K1 and C7 (vK1L-C7L-) was restricted by mouse cells and highly attenuated in mice, its replication and virulence were completely restored in SAMD9L-/- mice. In humans, both SAMD9 and SAMD9L are poxvirus restriction factors, although the latter requires interferon induction in many cell types. While knockout of SAMD9 with Crispr-Cas9 was sufficient for abolishing the restriction for vK1L-C7L- in many human cells, knockout of both paralogs was required for abolishing the restriction in interferon-treated cells. Both paralogs are antagonized by VACV K1, C7 and C7 homologs from diverse mammalian poxviruses, but mouse SAMD9L is resistant to the C7 homolog encoded by a group of poxviruses with a narrow host range in ruminants, indicating that host species-specific difference in SAMD9/SAMD9L genes serves as a barrier for cross-species poxvirus transmission. Emerging and reemerging infectious diseases have continued to pose a major threat to public health. In particular, zoonotic viral infections have caused such lethal human diseases as SARS, avian influenza, human monkeypox, and Ebola [1]. For many viruses, including coronaviruses and influenza viruses, host species-specific difference in viral entry receptors presents a major hurdle for cross-species transmission [2]. Poxviruses, however, can enter nearly any animal cell [3]. Why many poxviruses show strict host species specificity and what it would take for them to jump to new hosts are less clear [4]. Poxviruses include many lethal animal and human pathogens [5], the most infamous of which is the smallpox-causing variola virus. Smallpox was successfully eradicated mainly through a global immunization program with vaccinia virus (VACV), and routine VACV vaccination had since discontinued. The human population is now vulnerable to zoonotic orthopoxvirus infection, as some extant poxviruses related to variola virus are capable of infecting a wide variety of wild and domestic animals. There are also many poxviruses with a more restricted host range [6]. For example, capripoxviruses, consisting of sheeppox virus, goatpox virus, and lumpy skin disease virus, have a very narrow host-range in ruminants, causing economically significant diseases in sheep, goats, and cattle, respectively. Host-restricted poxviruses have been exploited as safe vectors for vaccines, gene therapy or oncolytic viral therapies, although the basis for their host restriction is largely unknown [4]. Poxvirus host range at the cellular level is governed by a group of poxvirus genes referred to as the host range genes [4,6]. The first discovered and perhaps the most important host range genes are K1L and C7L of VACV [7,8]. VACV replication in most mammalian cell lines requires either K1L or C7L [7], and the deletion of both genes from VACV aborts the replication prior to viral late gene expression [9]. K1L is only present in VACV and a few related orthopoxviruses, but a C7L homolog that functions nearly identically to VACV C7L is present in most mammalian poxviruses [10]. SAMD9 (Sterile Alpha Motif Domain-containing 9) was found to be the restriction factor in human cell lines that blocked the replication of poxvirus mutants that lack K1L and C7L-like genes [11]. K1 and C7 are structurally distinct [12,13], but K1 and many C7 homologs independently bind and inhibit human SAMD9 [11,12,14]. SAMD9 was initially identified as a tumor suppressor whose loss-of-function mutations in humans cause normophosphatemic familial tumoral calcinosis (NFTC) [15,16]. Humans and many other mammals also encode a chromosomally adjacent paralog of SAMD9 named SAMD9-like (SAMD9L). Human SAMD9 and SAMD9L are ubiquitously expressed in many tissues [17], and their expression can be further induced by interferons [18,19]. Evolutionary analysis suggested that the two paralogs derived from the duplication of an ancestral gene early in mammalian evolution and that some mammalian species suffered a loss of either SAMD9 or SAMD9L [20]. Notably, mice lack SAMD9, while many ruminants lack SAMD9L. Mouse SAMD9L is also a tumor suppressor, and haploinsufficiency of mouse SAMD9L resulted in myeloid malignancies [21]. The molecular functions of SAMD9 family of proteins remain largely elusive, but recent sequence analysis predicted a complex domain architecture suggestive of a regulative function of a putative signal transduction network [22]. Although SAMD9 has been established as a critical poxvirus restriction factor in human cell lines, whether this restriction is important at the organismal level and in other mammalian species is unknown. Moreover, whether SAMD9L plays a role in host defense is unknown. We studied the function of SAMD9L from a host that lacks SAMD9 (mice) as well as a host that maintains both paralogs (humans). We found that SAMD9L functions nearly identically to SAMD9 as a host restriction factor and that overcoming SAMD9/SAMD9L (SAMD9/L) restriction is essential for poxvirus replication and pathogenesis. We also discovered some host species-specific difference in SAMD9/L and some poxvirus species-specific difference in antagonizing SAMD9/L, suggesting that these differences contribute to the barrier for cross-species poxvirus infection. The importance of human SAMD9 as a poxvirus restriction factor became evident only when viral inhibitors of SAMD9 were removed from poxviruses, so we used a VACV mutant deleted of both K1L and C7L (vK1L-C7L-) as the model poxvirus to study poxvirus restriction factors in different host species. Mouse is one of the mammalian species that have lost SAMD9 gene, and a previous study suggested that mouse SAMD9L was not a functional paralog of human SAMD9 in terms of the tumor suppressor function [23]. However, mouse cells, such as NIH 3T3 cells and mouse embryonic fibroblasts (MEFs), restricted the replication of vK1L-C7L- [24]. To test whether mouse SAMD9L (mSAMD9L) might be the restriction factor for vK1L-C7L-, we used the CRISPR-Cas9 technology to knock out mSAMD9L from 3T3 cells. To control for potential off-target effect, we performed two independent knockouts with different guide sequences. To confirm the gene knockout, cell clones were isolated, and the mSAMD9L genotype of representative clones from the two knockouts (named ΔmSAMD9L#1 and ΔmSAMD9L#2) was determined by sequencing. All mSAMD9L alleles of the cell clones were found to contain indels, resulting predominantly in frameshift (S1A Fig). While vK1L-C7L- was unable to replicate in the parental 3T3 cells, it replicated well in ΔmSAMD9L#1 (Fig 1A) and ΔmSAMD9L#2 (S1B Fig) cells, resulting in nearly 100-fold increase of viral titer after 24 hours of infection. Moreover, while vK1L-C7L- was sensitive to interferons (IFNs) in permissive human cells [24,25], its growth in ΔmSAMD9L cells was not inhibited by pretreating the cells with IFN-β (Fig 1A). As an alternative to the CRISPR-Cas9 knockout, we prepared MEFs from SAMD9L+/+ and SAMD9L-/- mice. While vK1L-C7L- failed to grow in SAMD9L+/+ MEFs, its replication was completely restored in SAMD9L-/- MEFs (Fig 1B), confirming that mSAMD9L is the restriction factor for vK1L-C7L- in mouse cells. vK1L-C7L- is highly attenuated in mice [10]. To determine whether mSAMD9L functions as a restriction factor at the organismal level, we tested whether vK1L-C7L- would regain virulence in SAMD9L-/- mice. Intranasal infection of SAMD9L+/+ and SAMD9L+/- mice with 106 plaque-forming unit (PFU) of vK1L-C7L- did not cause any disease symptom or sustained body weight loss, similar to the mock infection (Fig 2A). The infected mice developed an antibody response to VACV (S2A Fig), indicating that they were properly infected. In contrast to their SAMD9L+/+ and SAMD9L+/- littermates, all SAMD9L-/- mice lost close to 25% of body weight by day 5 post infection and had to be euthanized. A similar lethal effect on SAMD9L-/- mice was observed when the infectious dosage was reduced to 105 or 104 PFU (Fig 2B), indicating that vK1L-C7L- was highly virulent in SAMD9L-/- mice. In mice, SAMD9L haploinsufficiency caused myeloid malignancies [21]. To test whether SAMD9L haploinsufficiency would increase susceptibility to vK1L-C7L- infection, we infected SAMD9L+/+ and SAMD9L+/- mice with the highest dosage of vK1L-C7L- we could reasonably obtain (107 PFU in 20 μl). Again, neither group of mice developed any disease symptoms or had sustained body weight loss (Fig 2C), and no virus was detected in lungs when the mice were euthanized at five days post infection (S2B Fig), indicating that one copy of mSAMD9L gene is sufficient for restricting vK1L-C7L-. Altogether, the results from SAMD9L knockout cells and mice demonstrate that mSAMD9L constitutes an essential host barrier for vK1L-C7L- replication and pathogenesis. We previously used vK1L-C7L- as the parental virus in constructing a panel of recombinant viruses that expressed different C7 homologs from representative mammalian poxviruses [10,24]. This panel of viruses were used for comparing the function of different C7 homologs in the same viral background, and they served as a surrogate model for different mammalian poxviruses in terms of their ability to overcome SAMD9 restriction. Studies of these viruses showed that the C7 homologs from many mammalian poxviruses, including myxoma virus (MYXV; infect rabbits) MYXV-M62, Yaba-like diseases virus (YLDV; infect monkeys) YLDV-67, swinepox virus (SWPV) SWPV-064, and sheeppox virus (SPPV) SPPV-063, could bind human SAMD9 and overcome its restriction [12]. All these C7 homologs, except for SPPV-063, could also overcome the restriction of vK1L-C7L- by mouse cell lines [24]. To determine whether the defect of SPPV-063 in mouse cell lines reflects a host species-specific defect, we studied the mouse virulence of vSPPV-063, the vK1L-C7L--derived recombinant virus that expressed SPPV-063. Swinepox virus SWPV-064 is the closest homolog of SPPV-063, so vSWPV-064, the vK1L-C7L--derived recombinant virus that expressed SWPV-064, was used as the control. Similar to vK1L-C7L-, vSPPV-063 did not result in significant body weight loss in SAMD9L+/- or SAMD9L+/+ mice but caused lethal intranasal infection in SAMD9L-/- mice (Fig 2D). In contrast, the same dose of vSWPV-064 was lethal to SAMD9L+/+ and SAMD9L+/- mice as well as to SAMD9L-/- mice (Fig 2E). In a separate experiment, vVACV-C7, the vK1L-C7L--derived recombinant virus that expressed VACV-C7, was also lethal to SAMD9L+/+ and SAMD9L+/- mice (Fig 2F). As SPPV-063 can overcome SAMD9 restriction in human cells, these data show that SPPV-063 has a host species-specific defect in mice. A biochemical basis for SAMD9 antagonism by the poxvirus proteins is their binding of SAMD9 [12], so we next studied the binding of mSAMD9L by the viral proteins. Due to the lack of a specific antibody to mSAMD9L, a plasmid encoding a flag-epitope tagged mSAMD9L was used in transfection of 293 cells, which were subsequently infected with recombinant viruses expressing different V5-tagged viral proteins. Pulldown of VACV-K1, VACV-C7, MYXV-M62, SWPV-064 or YLDV-67 also precipitated mSAMD9L (Fig 3A). However, SPPV-063 failed to co-precipitate mSAMD9L. As expected, MYXV-M63 and MYXV-M64, which are two additional C7 homologs from MYXV previously known not to bind human SAMD9 [12], also failed to co-precipitate mSAMD9L. Residue 134 and 135 of SPPV-063 were previously shown to be largely responsible for the defect of SPPV-063 in mouse cells, and substituting them with the corresponding ones found in SWPV-064 and VACV-C7 was sufficient to restore the host-range function in mouse cells without compromising the function in human cells [24]. The same substitution was found to result in the binding of SPPV-063 with mSAMD9L (Fig 3B). As controls, individual substitutions of two neighboring residues (residue 129 or 132) did not result in mSAMD9L binding, correlating with their lack of effect on the host-range function in mouse cells [24]. Interestingly, residues 134 and 135 are only adjacent to the three conserved loops of the C7-like proteins (Fig 3B) that are critical for the binding with SAMD9 [12]. The finding that mouse SAMD9L functions similarly to human SAMD9 as a poxvirus restriction factor raised the question about the function of human SAMD9L (hSAMD9L). Since knockdown of human SAMD9 (hSAMD9) was sufficient for abolishing the host restriction for vK1L-C7L- in HeLa and A549 cells [14,26], hSAMD9L was not previously suspected to play a role in host restriction. To find out whether hSAMD9L expression might be impaired in cell lines, we performed Western blot on HeLa cells and normal human fibroblasts derived from foreskin, and found that both hSAMD9 and hSAMD9L were constitutively expressed and their expression was further induced by IFN-β (Fig 4A). To study hSAMD9L function independent of hSAMD9, we knocked out hSAMD9 from HeLa cells with CRISPR-Cas9. In a HeLa cell clone with hSAMD9 knockout (ΔhSAMD9), no hSAMD9 was detected by Western blot even after the cells were treated with IFN-β (Fig 4A). ΔhSAMD9 cells were fully permissive for the replication of vK1L-C7L-, as reported previously [14]. However, pretreating the ΔhSAMD9 cells with IFN-β restored the host restriction for vK1L-C7L-, reducing the viral yield by more than 100-fold compared to that in untreated cells (Fig 4B). To test whether hSAMD9L was responsible for the restriction, we knocked out hSAMD9L from ΔhSAMD9 and the parental HeLa cells. Both the hSAMD9L single-knockout cells (ΔhSAMD9L) or hSAMD9 and hSAMD9L double knockout cells (ΔhSAMD9&L) had no detectable level of hSAMD9L protein in untreated cells and only a trace amount of the protein after IFN-β stimulation (Fig 4A). In contrast to ΔhSAMD9 cells, ΔhSAMD9&L cells remained permissive for vK1L-C7L- after IFN-β treatment (Fig 4B), and the growth of vK1L-C7L- was similar to that of the wild type (WT) VACV WR strain (S3 Fig), indicating that IFN-β-induced hSAMD9L restricted the replication of vK1L-C7L- in the absence of hSAMD9. ΔhSAMD9L cells were similar to the parental HeLa cells in restriction of vK1L-C7L- (Fig 4B). To ensure that the HeLa cell results are representative of the phenotype in human cells, we performed similar knockout studies in a variety of human cells that we identified to be restrictive of the replication of vK1L-C7L-. These include normal human foreskin fibroblasts (HFFs) and cancer cells derived from skin (A431), breast (HS578T and MDA-MB-231), cervix (HT-3), prostate (PC-3) and ovary (SKOV3). We transduced these cells with a lentivirus expressing a gRNA targeting either hSAMD9 or hSAMD9L and pooled the stably transduced cells. The SAMD9 and SAMD9L protein level in the pooled cells was reduced but not eliminated as in the clonally selected HeLa knockout cells (S4A & S5A Figs), presumably because the targeted gene was repaired with in-frame indels in a fraction of the cells. Thus, the cells were conservatively speaking gene knockdown instead of knockout cells. Nevertheless, the results were similar to that in knockout HeLa cells. The HFFs were less stringent than the HeLa cells in restricting vK1L-C7L-, allowing the viral titer increase by ~3-fold after 24 hours of infection (S4B Fig). The knockdown of hSAMD9 or hSAMD9L from HFFs increased the viral yield by ~110-fold or ~6-fold, respectively, indicating that hSAMD9 is the dominant restriction factor in HFFs. Correspondingly, robust viral late protein synthesis was only observed in hSAMD9-knockdown HFFs (S4A Fig). IFN treatment of hSAMD9-knockdown HFF cells restored the restriction for vK1L-C7L- (S4B Fig). The results from the knockout studies in all the cancer cell lines were also similar to that in HeLa cells in that the knockdown of hSAMD9, but not hSAMD9L, abolished the host restriction for vK1L-C7L- (S5B Fig). Moreover, IFN treatment of hSAMD9-knockdown cells restored the restriction for vK1L-C7L- (S5C Fig). Thus, in all human cells that we have tested, the basal level of hSAMD9 as well as IFN-induced hSAMD9L are both capable of restricting vK1L-C7L- replication. To find out whether mammalian poxviruses could overcome the restriction of hSAMD9L, we induced hSAMD9L expression in ΔhSAMD9 HeLa cells with 200 U/ml of IFN-β and then infected the cells with our panel of vK1L-C7L--derived VACVs. Viruses that expressed VACV-K1, VACV-C7, MYXV-M62, SPPV-063, SWPV-064 and YLDV-67 (but not MYXV-M63 and MYXV-M64) grew in IFN-treated ΔhSAMD9 cells (Fig 5A), indicating that all known SAMD9 antagonists could also antagonize hSAMD9L. We then studied these viral proteins for binding of hSAMD9L. While MYXV-M63 (S6 Fig) and MYXV-M64 did not co-precipitate hSAMD9L, VACV-K1, VACV-C7, MYXV-M62, YLDV-67, SPPV-063 and SWPV-064 co-precipitated hSAMD9L (Fig 5B). Among them, SPPV-063 precipitated a lower amount of hSAMD9L, indicating a reduced affinity. This defect was again due to residue 134 and 135 of SPPV-063, as substitution of these two residues increased the precipitation of hSAMD9L without affecting the precipitation of hSAMD9 (S6 Fig). To determine whether the weaker hSAMD9L binding affinity by SPPV-063 resulted in reduced inhibitory potency, we induced an increasingly higher level of hSAMD9L from ΔhSAMD9 HeLa cells with IFN-β and then infected the cells with either vSPPV-063 or vSWPV-064. The two viruses grew equally well in untreated ΔhSAMD9 cells, reaching similar titers after 24 hours of infection (Fig 5C). The viral yields were gradually reduced by the increasing concentrations of IFN-β, and the magnitudes of the reduction were significantly larger for vSPPV-063 than for vSWPV-064 when the interferon concentration was greater than 250 U/ml, indicating that SPPV-063 is less effective than SWPV-064 at antagonizing hSAMD9L. SAMD9 was recently identified as a restriction factor for poxviruses in human cell lines [11,14]. However, whether SAMD9 is important for host defense against poxvirus pathogenesis and whether similar antiviral defense exists in other mammals, especially those that lack a SAMD9 ortholog, were not established. In this study, we showed that the SAMD9 paralog, SAMD9L, from mammalian species as diverse as mice and humans, functions similarly to human SAMD9 as a poxvirus restriction factor. Since at least one of the two paralogs is present in all mammals with a completely sequenced genome [20], our finding indicates that SAMD9 and/or SAMD9L-mediated antiviral defense is conserved in mammals. This conservation allowed us to use the mouse model to reveal the critical role of SAMD9/L in host defense against poxvirus pathogenesis. Studies of SAMD9/L from two different mammalian species and the SAMD9/L inhibitors from diverse mammalian poxviruses revealed some host species-specific difference in SAMD9/L, which could serve as a barrier for cross-species poxvirus infection. This knowledge could be useful in assessing the potential of a given poxvirus in switching or expanding its host range. Mouse is one of the mammalian species that have lost SAMD9 but maintained SAMD9L. Mouse SAMD9L is 53% and 71% identical to human SAMD9 and SAMD9L at amino acid level, respectively. Through gene knockout with both the conventional technique as well as the CRISPR-Cas9 technique, we found that mouse SAMD9L was essential for restricting the replication of a model poxvirus, a vaccinia virus mutant deleted of both K1L and C7L (vK1L-C7L-). 3T3 cells with CRISPR knockout of SAMD9L as well as SAMD9L-/- MEFs were permissive for the replication of vK1L-C7L-. Furthermore, vK1L-C7L- caused lethal intranasal infection only in SAMD9L-/- mice but was completely avirulent in SAMD9L+/+ and SAMD9L+/- mice, demonstrating the potency of SAMD9L as a restriction factor at the organismal level. Human is one of the mammalian species that have both SAMD9 and SAMD9L, located head-to-tail in adjacent positions of the same chromosome. Previous phylogenetic studies suggested that SAMD9 and SAMD9L originated from a common ancestor through an ancient gene duplication event [20]. Gene duplications have a major role in evolution of new biological functions [27]. With about 60% amino acid sequence identity, human SAMD9 and SAMD9L could have diverged to take on different functions. This idea would be consistent with the previous findings that loss-of-function mutations in only SAMD9 cause NFTC in humans [15] and that knockdown of SAMD9 was sufficient for abolishing the restriction for vK1L-C7L- in several human cell lines [14,26]. We were thus initially surprised to find that human SAMD9L functioned similarly to SAMD9 as a restriction factor for poxviruses. The advent of CRISPR-Cas9 genome editing technique made it possible to readily knock out one or both of the human paralogs from various human cells. The knockouts in tumor cell lines from various tissues and the normal human foreskin fibroblasts showed that both SAMD9 and SAMD9L, when sufficiently expressed, could inhibit vK1L-C7L- replication. In many human cells, the basal level of SAMD9 is sufficient for restricting vK1L-C7L-, but SAMD9L has to be induced by IFN to have the same effect. This probably only reflects a difference in gene regulation of the two paralogs in human cells but not any real difference in their potency or mechanism of action. Mouse SAMD9L is also an interferon-stimulated gene [28], but the basal level of SAMD9L in mouse cells is sufficient for restricting vK1L-C7L-, again suggesting that the role of IFN in SAMD9L function is merely to induce SAMD9L to a sufficient level in some cell types. Similarly, the only contribution made by K1 and C7 in vaccinia virus antagonism of IFN or IRF1 [24,25] is likely the inhibition of IFN- or IRF1-induced SAMD9/L, as vK1L-C7L- was not sensitive to IFN in cells that had SAMD9/L knockout. Both human SAMD9 and SAMD9L have strong anti-proliferative function, and gain-of-function (GoF) mutations of SAMD9 or SAMD9L cause multisystem developmental disorder. GoF SAMD9 mutations cause MIRAGE (myelodysplasia, infection, restriction of growth, adrenal hypoplasia, genital phenotypes, and enteropathy) disorder [29], while GoF SAMD9L mutations cause a similar disorder characterized with cytopenia, immunodeficiency, and neurological symptoms [19,30]. In both cases, the GoF mutations predispose to myelodysplastic syndrome by facilitating the selection for a loss of the chromosome that contain the mutations [19,30]. So why are both SAMD9 and SAMD9L kept in humans and many other mammals? We can only postulate that the duplication of SAMD9 gene and the divergence of sequence and expression pattern represent a strategy for retaining two drastically different “alleles” of SAMD9/L in the same host without the deleterious effect of both genes being constitutively active. Having two different alleles of SAMD9/L would have given the hosts more flexibility to rapidly evolve SAMD9/L to evade the viral inhibitors, perhaps during a time of mammalian evolution when the challenge from poxviruses was particularly rampant. Phylogenetic analysis has showed that both SAMD9 and SAMD9L genes have been subjected to sustained positive selection [20]. Corresponding to the conservation of SAMD9/L in mammals, the distribution of SAMD9/SAMD9L inhibitors in mammalian poxviruses are also very broad. Nearly all mammalian poxviruses encode a C7 homolog that provides a host-range function similar to VACV C7 [10,24,31], and a few poxviruses also encode K1 (Fig 6). All the functional C7 homologs and K1 were previously shown to bind and inhibit human SAMD9 [11,12,14]. The breadth of their antagonistic activities is now expanded to include human and mouse SAMD9L. Among the C7 homologs from a wide variety of different mammalian poxvirus, the one from sheeppox virus, SPPV-063, stands out as the only one that displays a specificity for SAMD9. SPPV-063 could bind and inhibit human SAMD9, but it has reduced potency against human SAMD9L and failed to inhibit mouse SAMD9L. Remarkably, its binding to mouse SAMD9L could be restored simply by substituting two residues, suggesting that subtle difference between SAMD9 and SAMD9L might be responsible for their different ability in resisting SPPV-63. Interesting, these two residues are only adjacent to the “three-fingered molecular claw” that is critical for SAMD9 binding [12], indicating that they may modulate the conformation of “the claw” to influence its binding specificity. Sheeppox virus, goatpox virus, and lumpy skin disease virus (LSDV) are members of the capripoxvirus genus with a narrow host-range in ruminants. All capripoxviruses encode a nearly identical C7 homolog (>97% identical) as SPPV-063, with the same residues at position 134 and 135, suggesting that they all preferentially antagonize SAMD9. This correlates with the loss of SAMD9L in ruminants. Thus, one evolutionary scenario is that the ancestor of SPPV-063 could tolerate genetic drift that only disrupted SAMD9L binding, after the ancestor of extant capripoxviruses established a niche host in ruminants and was no longer restricted by SAMD9L. While the resulted failure of capripoxviruses in antagonizing mouse SAMD9L could account for the host restriction of capripoxviruses in mice, the resulted reduction in potency against interferon-induced human SAMD9L could only partly explain the host restriction of capripoxviruses in humans. We suspect that similar genetic drift might have occurred to other capripoxvirus host-range genes, resulting in their specialization to the ruminant hosts and altogether contributing to the host restriction of capripoxviruses in non-ruminant hosts. This idea is supported by a previous report that the sheeppox virus homolog of VACV host range gene E3 failed to provide the host range function in HeLa cells [32]. That mouse SAMD9L is completely resistant to SPPV-063 while human SAMD9L is only partially resistant could have been the result of a stronger positive selection in rodents, perhaps by some poxviruses that have similar inhibitory profile as the capripoxviruses. Overall, our studies suggest that host species-specific difference in SAMD9/L gene repertoire contributes to the barrier for cross-species poxvirus transmission. The animal studies reported in this paper were approved by the Institutional Animal Care and Use Committee at the University of Texas Health Science Center at San Antonio (Protocol no. 14040x), and adhere to the guidelines and recommendations from the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health and the Office of Laboratory Animal Welfare. The Institution has an Animal Welfare Assurance on file with the NIH Office of Laboratory Animal Welfare (assurance number A3345-01). The care and use of animals is in accordance with the NRC Publication, as revised in 2011, “Guide for the Care and Use of Laboratory Animals, ” and other applicable Federal regulations. VERO (ATCC CCL-81), HeLa 229 (ATCC CCL-2. 1), NIH/3T3 (ATCC CRL-1658), A431 (ATCC CRL-1555), HS578T (ATCC HTB-126), HT-3 (ATCC HTB-32), MDA-MB-231 (ATCC HTB-26), SKOV3 (ATCC HTB-77), and PC-3 (ATCC CRL-1435) were originally from ATCC. HEK 293FT was from Thermo Fisher Scientific (cat. no. R70007). Mouse embryonic fibroblasts (MEFs) were generated from embryos of SAMD9L−/− and SAMD9L+/+ mice according to standard protocols. Human foreskin fibroblasts (HFFs), described in [33], were kindly provided by Dr. Zhilong Yang. WT VACV WR strain, K1L and C7L deletion VACV (vK1L-C7L-) and a panel of vK1L-C7L--derived recombinant viruses expressing VACV-K1 (vVACV-K1L) or a C7 homolog from different poxviruses (vVACV-C7L, vYLDV-67R, vMYXV-M62R, vMYXV-M63R, vMYXV-M64R, vSPPV-063 and vSWPV-064) were described before [10,25,34]. vSPPV-063 with mutations at residue 129,132,134&135 of SPPV-063 were also described before [24]. All viruses were propagated and titrated on VERO cells. The plasmids used for the gene knockout were constructed from lentiCRISPRv2 (a gift from Feng Zhang, Addgene plasmid #52961) according to the published protocol [35]. In brief, lentiCRISPRv2 was digested with BsmBI and ligated with a pair of oligonucleotides with the specific guide sequence. For each target gene, two guide sequences were chosen from the human or mouse genome-wide sgRNA library [36]. They were as follows: 5’-TATCCGGGAACCACGGTTCG-3’ (mSAMD9L #1), 5’-ACACCAACAATTTCCCCGTG-3’ (mSAMD9L #2), 5’-TTGACTGAACAAGACGTGAA-3’ (hSAMD9 #1), 5’-GAGAATTTGTTCTTCGATAC-3’ (hSAMD9 #2), 5’- CCTGACCAGTTAGACGACGC-3’ (hSAMD9L #1), and 5’-GGCTAGCTCTAGGGATATCC-3’ (hSAMD9L #2). The lentiCRISPRv2-derived plasmid was either transfected directly to the target cells or used in making lentiviruses for transduction of the target cells. In the latter case, the lentiviral plasmid and the packaging plasmids pMD2. G and psPAX2 (gifts from Didier Trono, Addgene plasmid # 12259 & 12260) were transfected into HEK 293FT cells with lipofectamin 3000 (Thermo Fisher Scientific). 60 hr post transfection, culture supernatants were collected, clarified by centrifugation, passed through a 0. 45 μm filter, and used for transduction. For lentiviral transduction, the lentiviruses and the target cells in medium containing 10 μg/ml polybrene were centrifuged at 1,500 rpm for 2 hr. 24 hr after either transfection or transduction, the cells were subjected to puromycin selection for 2 (transfection method) or 7 (transduction method) days. The puromycin concentrations are 3 μg/ml for HeLa, 10 μg/ml for 3T3, and 2 μg/ml for all other cells. For each gene knockout in 3T3 and HeLa cells, two separate knockouts with different guides were performed. Clones of the knockout cells were isolated and validated by Western blotting for hSAMD9 or hSAMD9L or genotyping for mSAMD9L. For genotyping, the genomic DNA of the clones was extracted using the QuickExtract DNA extraction kit (Epicentre). ~500 bp of DNA flanking the target site was PCR-amplified with the primer pair (5’-GGCCACTCAATCTCATTGACCCAAT-3’ and 5’-TGCCCAGGATATTCTTAGAGCTAGC-3’) and cloned to pGEM vector (Promega). The sequence of 10–20 pGEM clones was determined by Sanger sequencing. For knockout of hSAMD9 or hSAMD9L in additional human cell lines and HFFs, only guide #1 described above was used for the knockout. The stably transduced target cells were pooled without clonal selection, and the knockout was validated by Western blot. Mice with different SAMD9L genotypes used in the infection experiment were generated from breeding pairs of SAMD9L+/- mice [21]. At around 4 to 5 weeks old, the mice were infected intranasally with viruses in 20 μl PBS as described previously [10]. The viruses used were purified through a sucrose cushion sedimentation according to the standard protocol [37]. Individual mice were weighed every day and euthanized when 25% of the body weight was lost. Statistical comparisons of body weight changes among groups were analyzed by two way ANOVA using GraphPad Prism 5. 0. Values of p < 0. 05 were considered statistically significant. The anti-VACV serum antibody titers of some infected mice were determined by ELISA with purified VACV virions as described previously [38]. All mouse protocols were approved by UTHSCSA IACUC. mSAMD9L ORF was PCR-amplified with the primer pair (5’-AGTGGACAAGTAACTCAACCAAAATTG-3’ and 5’-GATCACTTTTATGCCATATGCC-3’) from cDNA synthesized from 3T3 cellular mRNAs. A 3XFlag tag sequence and a HA tag sequence were then appended to the 5’ and 3’ end of the mSAMD9L ORF by recombinant PCR, and the final PCR product was inserted between KpnI and SacII sites of the pcDNA3. 1/V5-His-topo vector (Thermo Fisher Scientific). The mSAMD9L ORF was completely sequenced and found to have one amino acid difference (Val instead of Ile at 459) compared to the mSAMD9L reference sequence in GenBank (NP_034286. 2). The difference was not corrected, as SAMD9L from most rodents also has Val at this position. Cells in 12-well plates were incubated with 1 PFU per cell of different viruses for 2 h at room temperature. Following adsorption, the cells were washed twice with phosphate-buffered saline (PBS). One set of the cells was harvested immediately as the 0 hr post infection sample, while the other set were moved to 37°C incubator to initiate viral entry and harvested at 24 hr post infection. The viral titers in the cell lysates were determined by plaque assays on VERO cells. For testing the effect of IFN, human or murine cells were treated with 200 U/ml of human or murine IFN-β (PBL Biomedical Laboratories) for 24 hr, before the cells were infected with VACV as described above. For assessing viral late protein expression, Western blot with a mAb against the VACV late protein WR148 (clone HE7) [39] was performed. For immunoprecipitation of hSAMD9 or hSAMD9L, HeLa cells or IFN-treated ΔhSAMD9 HeLa cells were infected with different VACV viruses. For immunoprecipitation of mSAMD9L, 293FT cells were transfected with the mSAMD9L expression plasmid and then infected with different VACVs at 48 hr post transfection. After 8 hr of infection, the cells were lysed on ice with a lysis buffer (0. 1% (w/v) NP-40,50 mM Tris, pH 7. 4,150 mM NaCl) supplemented with protease inhibitor cocktail tablets (Roche Molecular Biochemicals). The cleared cell lysates were mixed with 50 μl of 50% (vol/vol) V5-agarose beads (Sigma-Aldrich) for 30 min at 4°C. After washing with lysis buffer, the beads were resuspended in SDS sample buffer, the eluted proteins were resolved by SDS-PAGE and detected with Western blot as described previously [34]. The detection antibodies were mouse monoclonal antibodies (mAb) against V5 (Sigma-Aldrich; clone V5-10), Flag tag (Sigma-Aldrich) or HSP70 (Santa Cruz), and rabbit polyclonal antibodies against hSAMD9 (Sigma-Aldrich, HPA-21319) and hSAMD9L (Proteintech, 25173-1-AP).

Zoonotic viral infections represent a major threat to public health. For many viruses, host species-specific difference in viral entry receptors presents a major hurdle for cross-species transmission. Poxviruses, however, can enter nearly any animal cell. Why many poxviruses show strict host species specificity and what it would take for them to jump to new hosts are less clear. Here, we present data suggesting that SAMD9 and its paralog, SAMD9L, constitute a critical host barrier against poxvirus infection and pathogenesis. We also discovered some host species-specific difference in SAMD9/SAMD9L and some poxvirus-specific difference in antagonizing SAMD9/SAMD9L, suggesting that these differences serve as a barrier for cross-species poxvirus infection. The knowledge is fundamental for understanding the determinants of poxvirus host-range.

Abstract Introduction Results Discussion Materials and methods

medicine and health sciences poxviruses pathology and laboratory medicine hela cells pathogens biological cultures microbiology viruses immunoprecipitation dna viruses cell cultures research and analysis methods proteins medical microbiology microbial pathogens vaccinia virus cell lines viral replication precipitation techniques biochemistry nih 3t3 cells host restricted organisms virology viral pathogens interferons biology and life sciences cultured tumor cells organisms

2018

A paralogous pair of mammalian host restriction factors form a critical host barrier against poxvirus infection

Wolbachia are maternally transmitted intracellular bacterial symbionts that infect approximately 40% of all insect species. Though several strains of Wolbachia naturally infect Drosophila melanogaster and provide resistance against viral pathogens, or provision metabolites during periods of nutritional stress, one virulent strain, wMelPop, reduces fly lifespan by half, possibly as a consequence of over-replication. While the mechanisms that allow wMelPop to over-replicate are still of debate, a unique tandem repeat locus in the wMelPop genome that contains eight genes, referred to as the “Octomom” locus has been identified and is thought to play an important regulatory role. Estimates of Octomom locus copy number correlated increasing copy number to both Wolbachia bacterial density and increased pathology. Here we demonstrate that infected fly pathology is not dependent on an increased Octomom copy number, but does strongly correlate with increasing temperature. When measured across developmental time, we also show Octomom copy number to be highly variable across developmental time within a single generation. Using a second pathogenic strain of Wolbachia, we further demonstrate reduced insect lifespan can occur independently of a high Octomom locus copy number. Taken together, this data demonstrates that the mechanism/s of wMelPop virulence is more complex than has been previously described. Symbiosis has played a pivotal role in arthropod diversification, speciation, and the ability of these animals to occupy a variety of niches in the natural world. Symbionts fall into two broad categories: infectious symbionts that often impose severe fitness costs to their host in order to complete their lifecycle, and vertically transmitted endosymbionts that form lifelong infections with their host, which range from beneficial to commensal in nature [1]. A key determinate of endosymbiont: host interactions is the infection density that the symbiont establishes in their host [2]. If symbiont density is too high, the endosymbiont risks inducing pathology in the host and reducing host fitness. On the other hand, if density is too low, the endosymbiont may not be transmitted to the next generation [3]. Density of endoysmbionts may be regulated by a combination of microbe or host mechanisms, as well as external factors including nutritional status of the host or temperature [2,4–6]. One symbiont that has been shown to be influenced by all of these factors is Wolbachia pipientis, a gram-negative alpha-proteobacteria that infects numerous invertebrate species, such as filarial nematodes and at least 40% of insect species [7–10]. Most Wolbachia manipulate host reproductive systems to enhance their maternal transmission through host populations, with a smaller number of strains shown to provide protection against microbial infections [11–13] or impose fitness costs to their host [14–16]. The strength of these phenotypes correlates with Wolbachia density [12,17–19], which itself has been shown to be largely strain and host dependent. Wild populations of Drosophila melanogaster are often infected by one of two Wolbachia strains, wMelCS or more commonly, wMel [20]. A third strain, wMelPop, was recovered from a mutant Drosophila laboratory stock [15]. The abundance and effect each strain has on Drosophila lifespan, or protection against viral infection, correlate with Wolbachia density. The wMel strain, which establishes the lowest density in the host, has no impact on lifespan but provides the lowest level of protection against viral infection; conversely wMelPop establishes the highest infection density and significantly reduces adult-lifespan but provides the highest level of virus protection [12]. The pathogenicity of wMelPop and its ability to over-replicate appear to be independent of host factors, with pathology and associated high infection densities observed in novel transinfected hosts [19], suggesting that genetic factors are responsible. Comparative genomic analyses between wMelPop and wMelCS have identified an 8-gene region, referred to as the “Octomom” locus, which is triplicated in wMelPop [19,21,22]. A recent study by Chrostek and Teixeira also correlated increased copy number of the Octomom locus with both increased Wolbachia infection densities and pathology. How the Octomom locus influences pathology was undetermined [22]. A second determinate of wMelPop pathology is the extrinsic temperature the host is exposed to, with pathology positively correlating to an increase in temperature [23]. Intriguingly, when flies are reared at 19°C no pathology is observed, presumably because wMelPop does not over-replicate or the rate at which it over-replicates was too slow to reduce host fitness [23]. While wMelPop pathology has been correlated to increasing temperature [23], or bacterial density and Octomom copy number [22], no studies to date have investigated the effect temperature has on Wolbachia density and Octomom copy number. If Wolbachia density determines the strength of pathology, we would expect to observe decreasing Wolbachia infection densities as the extrinsic temperature decreased. Similarly, if the Octomom copy number determines Wolbachia density, and consequently pathology, we would expect to observe a decrease in Octomom copy number as the extrinsic temperature decreases. Here we evaluated the lifespan of adult Canton-S Drosophila infected with wMelPop that were reared at four different temperatures, and as expected, observed increased pathology as temperature increased. When flies were reared at 18°C, however, we observed an extension to adult lifespan, similar to that previously observed at 16°C [24]. Estimates of wMelPop infection densities showed a bi-modal pattern across the different rearing temperatures, with a distinct shift in bacterial growth in the host. Absolute density and the rate of growth of Wolbachia were decoupled from the strength of pathology. Copy number of the Octomom region was dynamic across time but no correlation between temperature and Octomom copy number was observed. Similarly, no correlation between Octomom copy number and bacterial density, or strength of pathology was observed. Finally, we describe a new pathogenic strain of Wolbachia, wMel3562, which maintains the Octomom region at low frequency, established a high bacterial infection and reduced adult-lifespan in flies. Taken together, these observations challenge recent evidence of how expansion of the Octomom locus leads to the breakdown of mutualism between Wolbachia and host. A standard survival assay [15] was used to confirm that an increase in environmental temperature correlated with a reduction in adult lifespan of Drosophila infected with wMelPop. We compared the lifespan of adult D. melanogaster infected by wMelPop across a range of rearing temperatures (29°C, 24°C, 21°C, 18°C), to Wolbachia-free controls (Fig 1; Table 1). As expected, we observed the greatest pathology associated with the highest rearing temperature; median survival increased for both infected flies and uninfected controls as rearing temperatures decreased. Additionally, hazard ratios, the ratio of fly death between infected and uninfected controls, decrease with a reduction in temperature (24°C and 21°C), demonstrating that temperature, infection, and their interaction affect Drosophila survival (Table 1). When adult wMelPop-infected D. melanogaster were reared at 18°C, wMelPop infection was associated with an extended adult lifespan. Similar results have been previously observed for adult flies reared at 16°C [24]. Thus we conclude that there is a temperature dependent effect of wMelPop on Drosophila survival. The capacity of a bacterium to cause disease reflects its relative pathogenicity and the degree of virulence is directly influenced by the ability of the organism to cause disease despite host resistance mechanisms; it is affected by different variables such as the number of infecting bacteria [25]. Genes that influence bacteria virulence are also regulated by temperature, which acts as an' on-off' mechanism for bacterial growth [26]. Using a standard qPCR assay [27], we determined if temperature affected wMelPop replication over time in adult Drosophila, reared at 29°C, 24°C, 21°C, and 18°C (Fig 2). Flies reared at 29°C and 24°C had equivalent bacterial densities until the last day of fly survival at 29°C (Day 11; F (1,8) = 0. 59, p = 0. 61). Over-replication of wMelPop continued as flies reared at 24°C aged, reaching an average maximum density of approximately 181. 6 Wolbachia genomes to 1 Drosophila genome. This demonstrated a high bacterial density before death, despite outliving flies at a similar density, reared at 29°C. Flies reared at 24°C and 21°C had significantly different bacterial density until the last day of fly survival at 24°C (F (1,8) = 0. 02, p < 0. 0001). Flies reared at 21°C and 18°C had equivalent bacterial densities until the last day of fly survival at 21°C (F (1,8) = 0. 64, p = 0. 22). Wolbachia density displayed a bimodal trend between flies raised at high (29°C and 24°C), and low (21°C and 18°C) temperatures. To determine what temperature would shift wMelPop replication from low to high densities, or if an intermediate growth profile existed, we estimated wMelPop density in flies reared at 23°C and 22°C. wMelPop density in adult Drosophila reared at 23°C was not significantly different compared with density in flies reared at 24°C until the last day of fly survival at 24°C (Day 30; F (1,8) = 1. 11, p = 0. 42) a similar result was found when density was measured in flies reared at 22°C was compared to flies reared at 21°C through to the last day of fly survival at 24°C (Day 40; F (1,8) = 0. 60, p = 0. 94), demonstrating a bimodal relationship between density and temperature. These results suggest wMelPop density is temperature-dependent and that density is not correlated with host pathology under these conditions. Chrostek and Teixeira have previously postulated a positive correlation between increasing copy number of the Octomom locus / wMelPop bacterial density with pathology [22]. Given that temperature does influence bacterial density, but that bacterial density does not always correlate with pathology, we set out to determine if Octomom copy number correlated with differences observed in Wolbachia density at different rearing temperatures, or pathology. Octomom copy number was determined by estimating the ratio of WD0508, a single copy gene within the Octomom locus, to WD0550, a single copy gene in the wMelPop genome. First we compared copy number of WD0508 in 11-day-old flies, the earliest time point in which the bimodal density trend was observed from six rearing temperatures (29°C, 24°C, 23°C, 22°C, 21°C, and 18°C; Fig 3A). On average, between 1 to 1. 5 copies of the Octomom locus were observed in 11-day-old flies reared at 29°C, 24°C, 23°C, 22°C, and 21°C (Fig 3A; S1 Table). These results suggest Wolbachia density or pathology is not dependent on Octomom copy in infected Drosophila. When wMelPop infected flies were reared at 18°C there was a significantly lower prevalence of the insert, with at least half of the bacterial genomes lacking the Octomom locus (Fig 3A). As all flies were maintained at 24°C from embryo to one-day-adult flies prior to being reared at different temperatures, these results suggest either temperature directly affected genome instability at this locus within 11 days or that low temperatures had selected for wMelPop variants that lacked or had low copy number of the Octomom locus. To further explore how stable the Octomom locus was over time, Octomom copy number was estimated in a mixed population of flies reared at 24°C or 21°C, at different time points throughout their lifespan (Fig 3B). If Octomom copy number affected the strength of pathology, flies reared at 21°C should have lower Octomom frequency for a greater period of time when compared to flies reared at 24°C. No significant difference was observed for average Octomom copy number in wMelPop infected flies when reared at 24°C and 21°C (Fig 3B; S2 Table). Copy number was initially low in one-day old wMelPop infected flies, but increased in frequency by Day 5 and was maintained until Day 17 across both temperatures. Interestingly, as wMelPop infected flies aged (Day 22 –Day 40), Octomom copy number was not significantly different from that observed in 11-Day-old flies reared at 18°C (18°C (-2. 43 ± 0. 46); 24°C (-3. 03 ± 0. 42), [F (1,9) = 1. 486 p = 0. 92]; 21°C (-2. 20 ± 0. 35), [F (1,9) = 2. 06, p = 0. 59). These results demonstrate that temperature and increasing adult age, not Octomom copy number, influence wMelPop bacterial density. Furthermore, we observed Octomom copy number to be highly variable over developmental time and were not correlated to Wolbachia density or host pathology. To date only two Wolbachia strains that infect D. melanogaster are known to establish higher infection densities and reduce adult lifespan: wMelCS [12] and wMelPop [15] and both were recovered from CantonS Drosophila fly-stocks. To determine if other life-shortening Wolbachia strains existed we screened short-lived D. melanogaster fly-stocks for the presence of Wolbachia using a standard PCR assay [27]. All Wolbachia positive fly-lines were cured of their infection and survival was compared. From fly-stock 3562, a mutagenized CantonS fly-stock which contains a known mutation in the Hyperkinetic gene [28], we recovered a pathogenic strain of Wolbachia that reduced adult lifespan at both 29°C (Table 1; Fig 4; Hazard Ratio = 11. 29; 8. 23–15. 48) and 24°C (Table 1; Fig 4; Hazard Ratio = 1. 68; 1. 34–2. 11). Genotyping of the strain, hereto referred as wMel3562, confirmed it is a member of the wMelCS/wMelPop clade sharing all known genetic markers with both strains [12,13,20]. We then estimated the Wolbachia density (Fig 5) and the copy number of the Octomom locus (Fig 6). Similar to wMelPop, wMel3562 over replicated in infected flies (Fig 5) at both 24°C and 29°C, with densities equivalent to those of wMelPop observed at 21°C (F (1,8) = 2. 61; p = 0. 36), and has a higher density than wMelCS when reared at 29°C. As with wMelPop pathology, the strength of wMel3562 pathology was influenced by temperature rather than absolute bacterial density or the rate of bacterial growth (Fig 5). Measurements of the Octomom locus in 11-day old flies showed that at 24°C, the frequency of the locus within the wMel3562 population was low (Fig 6; F (1,8) = 0. 14, p < 0. 001) compared to wMelPop at 24°C, but equivalent to the copy number observed in 22 day-old wMelPop infected flies (F (1,8) = 0. 10, p = 0. 86). Mean Octomom copy number in wMel3562 relative to a single copy Wolbachia gene in 11-day old flies reared at 24°C. Errors bars represent SE. This study provides evidence that wMelPop induces host pathology in a temperature dependent manner, which is independent of bacterial density or rate of bacterial growth. Over replication of wMelPop was observed in Drosophila reared at all evaluated temperatures–the rate at which this was observed was bimodal. Flies reared at high temperatures (29°C—23°C) shared similar bacterial density across time and established the highest bacterial density in the shortest period of time. Flies reared at low temperatures (22°C—18°C) had similar bacterial densities to each other and had a markedly different growth rate and final bacterial density. Interestingly, despite having the same infection density (e. g. 21/18°C or 29/24°C), infected flies lived significantly longer as rearing temperature decreased and even outlived their uninfected counterparts at extremely low temperatures, a phenotype that has been previously described at 19°C and 16°C [23,24]. This suggests that the interaction of wMelPop and host rearing conditions is a major determining factor in pathology, and that strength of pathology was not determined by either absolute bacterial density or the rate of growth within the host. Chrostek and Teixeira recently described a correlation between high and low Octomom copy number, wMelPop density and pathology [22]. Critical to these observations were a set of selection experiments whereby they were able to select for high or low-copy Octomom wMelPop Drosophila flylines and in turn observed altered pathology in accordance with their predictions. A similar selection experiment had been previously conducted, however unlike Chrostek and Teixeira’s experiments, this study concluded that the changes to pathology were due to host effects and not selection acting upon wMelPop [29]. The difference between these two studies can be attributed to the design of the selection experiment. While both selected for increased or decreased wMelPop pathology for a minimum of 14 generations, based on high/low Octomom copy number [22] or on survival [30], only Carrington and colleague’s experiment comprised a series of backcrosses to the unselected parental stock in order to determine if selection had acted upon the nuclear host genome or the wMelPop genome. As the changes to pathology persisted with the host nuclear background they concluded that the observed changes to pathology were due to selection upon the host genome and not the wMelPop genome [30]. In the absence of this additional experiment, it is difficult to determine if the selection applied by Chrostek and Teixeira, and its associated changes to pathology, has acted upon wMelPop, the host genome or both. While all estimates of Octomom copy number using qPCR are an average of all wMelPop bacteria that infect an individual fly (thus a single fly could harbour both low and high-copy Octomom wMelPop bacteria) our data showed that Octomom copy number is highly variable over the fly lifespan. Low copy Octomom wMelPop variants were observed in larval, pupal, and late adult insects, and higher Octomom copy wMelPop variants present in younger adults. Octomom’s variability over time poses a number of questions and can be explained by a number of scenarios. The first is that low- or high-copy Octomom wMelPop variants might be tolerated at different developmental stages, with low-copy number wMelPop selected for in larval and pupal stages, while adult flies might tolerate Wolbachia with higher copy numbers of Octomom. Variation of Octomom copy number from high- to low-copy number could also be the result of individual flies that harbour high-copy Octomom wMelPop dying faster than those infected by low-copy Octomom wMelPop. Thus only low-copy Octomom wMelPop strains could be recovered in flies older than 17-days. If this were true, then the death rate for flies with low-copy Octomom wMelPop should be identical to that of uninfected flies, yet we observed both mortality and wMelPop density continue to increase after 17-days of age while at the same time the ratio of the Octomom locus to the rest of the wMelPop genome decreased three fold, from a ratio of 1: 1 to 1: 3 (Fig 3B). A third possibility is that all flies were infected by both low and high-copy Octomom wMelPop bacteria, over time the high-copy strains simultaneously induced cellular damage to the host, leading to pathology, and died off faster than low-copy number strains. A final possibility is that due to frequent recombination at the repetitive sequences that flank the Octomom locus, its copy number within the wMelPop genome changes over-time as the fly develops and ages. The role of Octomom copy number variation in pathology is unclear, regardless of how observed variation may arise. We observed no difference in Octomom copy number in wMelPop genomes when flies were maintained at different temperatures and harboured different bacterial densities or experienced different levels of pathology. We also showed that in addition to wMelPop-CLA [27], the wMel3562 strain establishes an infection density higher than wMelCS and induces pathology in the fly host, however, the Octomom region is absent in wMelPop-CLA [21] and uncommon within mixed D. melanogaster populations of wMel3562 (Fig 6). Furthermore, a third strain related to wMel, wAu establishes significantly higher infection densities than wMel in D. melanogaster [30] and also lacks the Octomom region [31,32]. Consequently we conclude that wMelPop density, its rate of growth, and the strength of pathology induced is unrelated to the copy number of the Octomom locus. The mechanisms by which pathogenic strains of Wolbachia such as wMelPop, wMelPop-CLA, and wMel3562 over-replicate and induce pathology as flies age, are still unclear, however temperature appears to be a significant force. It is well established that temperature affects Drosophila biology and lifespan. When reared at high temperatures, adult Drosophila suffer from a general degeneration of cytoplasmic organelles in nerve cells while at the same time there is an intense loss of ribosomes in the Malpighian tubules [33]. As both tissues are heavily infected by wMelPop [15,34–36], the presence of bacteria may exacerbate these physiological responses, leading to the observed host pathology. Temperature has also been shown to affect Drosophila immune function and their response to bacterial pathogens [37–39]. When reared at 17°C adult Drosophila display increased gene expression of the heat shock protein Hsp83, as well as several immune genes, which both correlate with decreased bacterial growth and pathology when compared to flies reared at 25°C or 29°C [38]. Given the decrease in host pathology in wMelPop-infected Drosophila maintained at low temperatures, it is tempting to speculate that similar host immune responses act to attenuate wMelPop pathology. Despite initial hopes of describing an environment-genotype-to-phenotype link among extrinsic temperature, Octomom copy number and pathology, we have demonstrated that Octomom copy number is highly variable over time, is unresponsive to extrinsic rearing temperature and is not correlated to either bacterial density or pathology. The density of wMelPop does not appear to determine pathology as equivalent bacterial densities were observed at different rearing temperatures but the strength of pathology differed. Instead it appears that a combination of the rate of bacterial growth and temperature determines wMelPop pathology. The pathogenic Wolbachia strain wMelPop [40,41], was introgressed into the Drosophila melanogaster Canton-S [42] stocks as described previously [43] D. melanogaster fly-strain 3562, known to have reduced lifespan and a mapped mutation to hyperkinetic (Hk1) [44,45] was obtained from the Bloomington (Indiana, USA) stock centre. The Wolbachia infection from the 3562 flyline, henceforth referred to as wMel3562, was introgressed into two different Drosophila genetic backgrounds: BNE, a wild-type strain collected from Brisbane, Australia [46] and w1118 a heavily inbred white eyed mutant strain. Briefly, virgin 3562 females were mated with Wolbachia-free BNE males (BNE-T); female progeny that maintained the FM6 balancer were collected and crossed to BNE-T males. Wildtype female progeny were collected and backcrossed to the BNE-T background for an additional five generations to create the BNE-wMel3562 line. The wMel3562 strain was introgressed into the w1118 background from the BNE-wMel3562 flyline by crossing virgin females with w1118 males for five generations. Tetracycline treatments were performed as described previously [47] to generate genetically identical fly lines that lacked the Wolbachia infection; hereto referred as wMelPop-T or wMel3562-T. Gut flora was reconstituted using standardised methods [12] and all experiments were conducted at a minimum of seven generations post tetracycline treatment. All flylines were reared from embryo to 1-day-old adults at 24°C on a 12: 12 hour light-dark cycle. The lifespan of approximately 200 adult male Drosophila derived from wMelPop or wMelPop-T fly-lines reared as described previously were determined at 29°C, 24°C, 21°C and 18°C. The lifespan of approximately 200 adult male Drosophila derived from wMel3562 or wMel3562-T fly-lines reared as described previously were determined at 24°C and 29°C. All adults were collected by CO2 anaesthesia immediately following eclosion and separated into groups of 20 before being transferred to the desired temperature, maintained on standard food medium and kept on a 12: 12 hour light-dark cycle. Survival was determined every 3-days, until all flies had died. Data was analysed using LogRank analysis (Mantel-Haenszel method; proportional hazards model; SPSS). Both Wolbachia infection density in adult flies and the copy number of the Octomom repeat locus in the Wolbachia genome were estimated using relative qPCR assays [27]. Five adult flies reared at 29°C, 24°C, 23°C, 22°C, 21°C, or 18°C were collected at regular time points. Genomic DNA was isolated using a QIAGEN DNeasy Blood & Tissue Kit according to manufacturer instructions (QIAGEN, Doncaster, VIC). Total DNA was estimated using an ND-1000 Nanodrop Spectrophotometer (Analytical Technologies, Collegeville, PA). To estimate the relative abundance of Wolbachia bacteria in each sample, we compared the abundance of the single-copy Wolbachia ankyrin repeat gene WD0550 to that of the single-copy D. melanogaster gene Act88F [27]. To estimate the copy number of the Octomom repeat locus, we compared the abundance of the single copy gene WD0550 to WD0508 (WD0508F: 5’ TGAGGAAGAAAGTGGAAAGGCA 3’ WD0508R: 5’ ACATGAGCAGAAACTCCTTCCT 3’), a single copy gene located within the Octomom repeat locus. Each qPCR contained 12. 5 μl of 2x SYBR pre-mix (QIAGEN), 1 μl of Forward primer, 1 μl of Reverse primer, 100 ng of DNA and H2O to a final volume of 25 μl. [27]. The relative abundance of Wolbachia bacteria to Drosophila or Octomom repeat region to the Wolbachia genome was determined using the delta-delta CT method [48]. Statistical significance was established with Two-Way ANOVA for bacterial density and One-Way ANOVA for Octomom copy number (SPSS).

Wolbachia are obligate intracellular, symbiotic bacteria that infect approximately 40% of insect species, as well as filarial nematodes, arachnids and terrestrial isopods. While the vast majority of Wolbachia strains impose few fitness costs to their host, one strain wMelPop is unique as it lacks the ability to regulate its growth, and as consequence can reduce host lifespan by half. The strength of pathology induced by wMelPop has been linked to either increased bacterial density or copy number of an eight gene tandem repeat region referred to as the “Octomom” locus. To date no study has determined the effect changes to temperature have on Octomom copy number or bacterial density. Here we demonstrate that while the Octomom locus is unstable within a single generation of its host, changes to Octomom copy number did not occur in response to temperature. Furthermore, Octomom copy number or bacterial density does not correlate to the strength of pathology. These results indicate that the underpinning genetics of pathology are unclear, and the mechanisms by pathology is induced are more complex than previously realised.

Abstract Introduction Results Discussion Methods

invertebrates medicine and health sciences pathology and laboratory medicine microbiology invertebrate genomics animals wolbachia animal models bacterial diseases developmental biology drosophila melanogaster model organisms microbial growth and development bacteria drosophila research and analysis methods infectious diseases microbial physiology pathogenesis genetic loci animal genomics insects bacterial growth arthropoda host-pathogen interactions genetics biology and life sciences genomics organisms

2016

Intensity of Mutualism Breakdown Is Determined by Temperature Not Amplification of Wolbachia Genes

Proline–tyrosine nuclear localization signals (PY-NLSs) are recognized and transported into the nucleus by human Karyopherin (Kap) β2/Transportin and yeast Kap104p. Multipartite PY-NLSs are highly diverse in sequence and structure, share a common C-terminal R/H/KX2–5PY motif, and can be subdivided into hydrophobic and basic subclasses based on loose N-terminal sequence motifs. PY-NLS variability is consistent with weak consensus motifs, but such diversity potentially renders comprehensive genome-scale searches intractable. Here, we use yeast Kap104p as a model system to understand the energetic organization of this NLS. First, we show that Kap104p substrates contain PY-NLSs, demonstrating their generality across eukaryotes. Previously reported Kapβ2–NLS structures explain Kap104p specificity for the basic PY-NLS. More importantly, thermodynamic analyses revealed physical properties that govern PY-NLS binding affinity: (1) PY-NLSs contain three energetically significant linear epitopes, (2) each epitope accommodates substantial sequence diversity, within defined limits, (3) the epitopes are energetically quasi-independent, and (4) a given linear epitope can contribute differently to total binding energy in different PY-NLSs, amplifying signal diversity through combinatorial mixing of energetically weak and strong motifs. The modular organization of the PY-NLS coupled with its combinatorial energetics lays a path to decode this diverse and evolvable signal for future comprehensive genome-scale identification of nuclear import substrates. Karyopherinβ proteins (Kapβs; Importins/Exportins) mediate the majority of nucleocytoplasmic protein transport. There are 19 known Kapβs in human and 14 in yeast [1,2]. Kapβs bind substrates through nuclear localization or export signals (NLSs or NESs) and transport them through the nuclear pore complex, and Ran GTPase regulates Kapβ–substrate interactions [3–6]. Ten Kapβs are known to function in nuclear import, each recognizing at least one distinct NLS. The best-known NLS is the short, basic, classical NLS, which is recognized by Kapα/Kapβ1 [4], and this pathway is conserved functionally from human to yeast [7,8]. Classical NLSs can be divided into monopartite and bipartite NLSs. Monopartite NLSs contain a single cluster of basic residues, whereas bipartite sequences contain two clusters of basic residues separated by a 10–12 amino acid linker. Thermodynamic dissection by scanning alanine mutagenesis of monopartite NLSs from the SV40 large T antigen (PKKKRKV) and the c-myc proto-oncogene (PAAKRVKLD) [9–11] confirmed a previously determined consensus sequence of K (K/R) X (K/R) [8,12]. Binding energies of these small signals are dominated by a single lysine residue, in the third position of the SV40 large T antigen and in the fourth position of c-myc, which makes numerous interactions with Kapα [9]. Thus, in the monopartite classical NLS, it is well-known that a relatively small motif is recognized, and binding energy is concentrated in stereotypical fashion across small sequences. Although numerous structures are available for bipartite NLSs [13–15], thorough thermodynamic analysis of this subclass is not available, and its consensus is less well-defined (one example is KRX10–12KRRK) than that for the monopartite NLS. Furthermore, a nonfunctional SV40 NLS mutant was rescued by a bipartite-like addition of a two-residue N-terminal basic cluster [9], suggesting that bipartite classical NLSs can accommodate larger sequence diversity than their monopartite counterparts. Recently, structural and biochemical analyses of human Kapβ2 (Transportin) bound to the hnRNP A1 NLS revealed physical rules that describe Kapβ2′s recognition of a diverse set of 20–30-residue-long NLSs that we termed PY-NLSs [16]. These rules are structural disorder of a 30-residue or larger peptide segment, overall basic character, and weakly conserved sequence motifs composed of a loose N-terminal hydrophobic or basic motif and a C-terminal RX2–5PY motif. The composition of the N-terminal motifs divides PY-NLSs into hydrophobic and basic subclasses (hPY- and bPY-NLSs). The former contains four consecutive predominantly hydrophobic residues, while the equivalent region in bPY-NLSs is enriched in basic residues. Approximately 100 different human proteins have been identified as potential Kapβ2 substrates [16–25]. Table 1 summarizes previously reported validated and potential PY-NLSs. Although many of these potential substrates were predicted by bioinformatics [16] and still need experimental testing, more than 20 have been validated for Kapβ2 binding (Table 1) [16–25]. Comparison of in vivo and in vitro validated PY-NLSs shows large sequence diversity, which is reflected in weak consensus sequences [16]. Structures of five different Kapβ2-bound PY-NLSs also show substantial variability, with structurally diverse linkers separating the convergent consensus regions [16,26,27]. The PY-NLS is significantly larger than the short monopartite classical NLS. The well-defined consensus and concentrated binding energy of the latter may reflect compactness of the signal. In contrast, the binding energy of the PY-NLS is spread over a much larger sequence. Physical properties of the multipartite PY-NLS may be more similar to those of the less-studied, larger, and sequentially more diverse bipartite classical NLS. Diverse PY-NLSs are described necessarily by weak consensus motifs. Therefore, instead of the traditional way of describing a linear recognition motif with a strongly restrictive consensus sequence, PY-NLSs were described by a collection of individually weak physical rules that together were able to provide substantial limits in sequence space for reasonable predictions of new Kapβ2 substrates [16]. However, the currently predicted substrates are most likely only a fraction of all PY-NLS-containing proteins because narrow sequence patterns were used in the initial search to achieve optimal accuracy. In fact, the sequence patterns used [16] were too narrow to predict PY-NLSs in known substrates HuR, TAP, hnRNP F, and JKTBP-1. The coverage of conventional sequence-based bioinformatics searches is expected to be severely limited due to PY-NLS diversity. Although sequence patterns obviously need to be expanded, we do not yet understand the limits of sequence diversity within motifs or how the different motifs may be combined. Knowledge of how binding energy is parsed in PY-NLSs will shape future efforts to decode these highly degenerate signals. Furthermore, physical understanding of how diverse PY-NLS sequences can achieve common biological function also will provide unique insights into many biological recognition processes that involve linear recognition motifs with weak and obscure consensus sequences, such as vesicular cargo sorting and protein targeting to the mitochondria and the peroxisome [28–33]. The yeast homolog of Kapβ2 is Kap104p (32% sequence identity) [34]. Only two Kap104p substrates, the mRNA processing proteins Nab2p and Hrp1p, are known. Several groups have mapped and validated NLSs of these substrates using both in vivo and in vitro methods to arginine–glycine (RG) -rich regions that were termed rg-NLSs [35–37]. Little sequence homology was detected between NLSs recognized by Kapβ2 and Kap104p. Furthermore, substrate recognition by the two karyopherins appears nonanalogous, as Kap104p does not recognize human substrate hnRNP A1 [35,37]. Given the recent physical understanding of Kapβ2–NLS interactions, we seek to examine the evolutionary conservation and energetic organization of signals in this pathway through studies of Kap104p–NLS interactions. First, we present biochemical and biophysical analyses showing that RG-rich substrates of yeast Kap104p share similar physical characteristics to those of human PY-NLSs. Kap104p recognizes the basic but not hydrophobic PY-NLS subclass, and structural analyses of Kapβ2–NLS complexes suggested the origin of this specificity, enabling prediction of PY-NLS subclass specificity for all eukaryotic Kapβ2s. Thermodynamic analyses of Kap104p–NLS interactions revealed biophysical properties that govern binding affinity of PY-NLSs. These signals contain at least three energetically significant binding epitopes that are also linear motifs. Each linear epitope accommodates significant sequence diversity, and we have characterized some of the limits of this diversity. The linear epitopes are also energetically quasi-independent, a property that is probably due to intrinsic disorder of the free signals. Finally, in different PY-NLSs, a given epitope can vary significantly in its contribution to total binding energy. When combined with multivalency, this energetic variability can amplify signal diversity through combinatorial mixing of energetically weak and strong motifs. In vivo validated RG-rich NLSs of Hrp1p and Nab2p (or rg-NLSs) are located at residues 494–534 and 201–250, respectively (Figure 1A) [35–38]. Examination of their sequences revealed physical characteristics similar to those of human PY-NLSs. Hrp1p and Nab2p NLSs are located within structurally disordered segments of 120–190 residues (DisEMBL structural disorder probabilities of 0. 72 and 0. 63 for Hrp1p and Nab2p, respectively [39]) in the full-length proteins (Figure 1A). 506RSGGNHRRNGRGGR519 of Hrp1p and 216KNRRGGRGGNRGGR229 of Nab2p contain many basic residues, like basic N-terminal motifs in human Kapβ2 substrates hnRNP M, PQBP-1, and YB-1 (Table 1) [16]. Closer to the C termini, the Hrp1p 525RNNGYHPY532 and the Nab2p 235RFNPL239 segments either match or are homologous to the C-terminal RX2–5PY consensus. Immobilized full-length Hrp1p, Nab2p, and their NLSs bound Kap104p in stoichiometric proportions in pull-down binding assays (Figure 1B). Although it was previously reported that Ran could not dissociate substrate from Kap104p [36], we observed efficient dissociation of both full-length substrates and NLSs by RanGTP, possibly due to higher activity and GTP loading of the recombinant Ran. Our results suggest that Kap104p–NLS interactions and regulation by Ran are similar to other characterized Kapβ-mediated nuclear import processes in human [3–6]. Thermodynamic parameters for Kap104p binding to Hrp1p and Nab2p NLSs were obtained by isothermal titration calorimetry (ITC) (Figure S1). Both NLSs bound Kap104p with high affinity (KD of 32 nM for Hrp1p and 37 nM for Nab2p (Tables 2 and 3) ), and extensive mutagenesis of NLSs is discussed below. Thus, on the basis of their sequence characteristics, high affinity for karyopherin, and dissociation by RanGTP, yeast NLSs recognized by Kap104p resemble PY-NLSs. To investigate the PY-NLS subclass specificity of Kap104p, we examined its interaction with several human hPY- and bPY-NLSs as well as several predicted (see below) yeast hPY- and bPY-NLSs. Splicing factor hnRNP A1 and mRNA transport factor TAP/NXF1 contain hPY-NLSs, and splicing factor hnRNP M and FUS contain bPY-NLSs (Figure 2A). All four human PY-NLSs interacted with Kapβ2 [16], but only bPY-NLSs from hnRNP M and FUS bound yeast Kap104p in GST pull-down assays (Figure 2B). Both yeast Hrp1p and Nab2p NLSs bound equally well to Kap104p and Kapβ2 (Figure S2). Hrp1p and Nab2p are the only two known Kap104p substrates [34–36]. We needed to identify additional yeast sequences to test the preference of Kap104p for bPY-NLS. Because Nab2p has a C-terminal PL instead of PY motif, suggesting that PL motifs also may be present in other functional PY-NLSs, we used the program ScanProsite [40] and sequence patterns Φ1-G/A/S-Φ3-Φ4-X7–12-R/K/H-X2–5-P-Y/L (where Φ1 is a hydrophobic residue and Φ3 and Φ4 are hydrophobic residues or R or K) [16] to search for potential hPY-NLSs within Saccharomyces cerevisiae proteins in the UniProtKB/Swiss-Prot protein database [41]. A consensus sequence for the N-terminal motif of bPY-NLSs is not available due to lack of an apparent specific pattern. As a result, we modified a previously used sequence pattern that is consistent with the basic motifs of hnRNP M and PQBP-1 [16] to accommodate additional validated human bPY-NLSs and NLSs in Nab2p and Hrp1p (Table 1). The resulting sequence pattern K/R-X0–6-K/R-X0–6-K/R-X0–6-K/R-X2–5-R/K/H-X1–5-PY/L is used to search for potential yeast bPY-NLSs. The resulting lists were filtered for structural disorder [39] and overall basic character. Six hPY/L-containing fragments were tested, but none bound Kap104p (Figure 2A and 2C). However, 11 of 20 bPY/L-containing fragments tested bound Kap104p and were dissociated by RanGTP (Figure 2A and 2D and Figure S3a and S3b). Two bPY/L-containing full-length substrates, Tfg2p and Rml2p, were tested, and both bound Kap104p and were dissociated by RanGTP (Figure 2D). Of the 11 bPY/L-containing proteins in yeast that bound Kap104p (Table 1), 7 (or 64%) have been shown to be predominantly nuclear or show both nuclear and cytoplasmic localization. Thus, recognition of the basic subclass of PY-NLS is conserved between human and yeast. However, human Kapβ2 has evolved to recognize an additional hydrophobic PY-NLS subclass, enabling it to transport a broader range of substrates. Alternatively, Kap104p may have evolved to be more specific and lost its ability recognize hPY-NLSs. Kapβ2 and Kap104p sequences were aligned and examined in the context of crystal structures of Kapβ2 bound to NLSs of hnRNPs A1 (hPY-NLS) and M (bPY-NLS) [16,26]. Kapβ2 has 20 HEAT repeats, each consisting of two antiparallel helices A and B. Both PY-NLSs bind the Kapβ2 interface lined with B helices of HEAT repeats 8–18 (abbreviated H8B–H18B), converging structurally at three spatially distinct binding sites: (1) overlapping portions of the N-terminal hydrophobic and the larger basic motifs, (2) the arginine residue, and (3) the PY residues, both of the C-terminal RX2–5PY motifs [26]. Correspondingly, both structures share many common Kapβ2 interface residues, especially those that contact the conserved C-terminal RX2–5PY motif (Figure 3A). Approximately half of the Kapβ2–NLS interface residues are conserved in Kap104p. Interfaces with the RX2–5PY motifs (H8B–H12B) are mostly invariant, while differences occur at the structurally overlapping interfaces with the basic/hydrophobic N-terminal motifs (H15B–H17B) and at linker regions (H12B–H14B) (Figure 3A). Here, Kapβ2 residues I722, S723, N726, E734, T766, and I773 that contact the hnRNP A1 hydrophobic motif are replaced with T, P, I, L, S, and V, respectively, in yeast (Figure 3A and 3B) such that many hydrophobic contacts with the FGPM N-terminal motif of hnRNP A1 are expected to be lost in yeast (detailed description in Text S1). In contrast, among Kapβ2 residues that contact basic side chains of bPY-NLSs, only E653 of Kapβ2 is different in yeast (Figure 3A and 3C), and several amino acids have been replaced by more electronegative amino acids in Kap104p (Figure 3C), further supporting bPY-NLS recognition in yeast. Comparison of individual HEAT repeats of Kapβ2 and Kap104p showed high identity (∼50%) at H8–H10, but the similarity dropped to ∼20% at H17 (Figure 3D). The B helices that line the interface are generally more conserved than the outer A helices. However, even in the former, sequence identities in H16B–H17B dipped significantly below 40% (Figure 3D). These observations suggest that both helical orientations and interface functional groups are better conserved at recognition sites for the C-terminal PY motif (H8–H10) than at the N-terminal basic/hydrophobic motifs (H16–H17). Consequently, the loss of Kap104p recognition for the N-terminal hydrophobic motif is most likely due to critical interface residue changes in H16B–H17B and to changes in helical orientations in this region. We have aligned sequences of Kapβ2 homologs, tracked interface residues and potential overall helical similarities at the N-terminal hydrophobic motif interfaces in different organisms, and used this information to predict species in which Kapβ2 would recognize hPY-NLSs. Results of these studies are discussed in Text S2 and shown in Figure S4A and S4B. We have performed scanning alanine mutagenesis covering residues 506–532 of the Hrp1p NLS (Figure 1A, Table 2, and Table S1). In the N-terminal region of the Hrp1p NLS, none of the four mutants 506RSGG509/AAAA, 512RRNG515/AAAA, 516RGG518/AAA, and 519RGGYN523/AAAAA (Table 2) affected Kap104p binding, suggesting that this N-terminal basic-enriched region may contribute little to total binding energy. However, these mutations may be misleading as glycine to alanine mutations may decrease the entropy of the unbound NLS, thus decreasing the entropic penalty of binding and offsetting affinity loss from arginine mutations. Therefore, we also generated a quadruple mutant where all of the arginines (R512, R513, R516, and R519) were mutated to alanines. This quadruple mutant decreased Kap104p binding by a marginal 5-fold (Figure 4A and Table 2), suggesting that positive charges in the N-terminal basic region are somewhat important for Kapβ–NLS interaction. Quadruple mutant R512, R513, R516, R519/KKKK did not affect Kap104p binding (Table 2), further suggesting that stereospecific interactions with arginine guanido groups are not important for Kap104p binding. Kap104p binding was not affected significantly when both arginine residues, 524RR525, in the C-terminal RX2–5PY motif of the Hrp1p NLS were mutated to alanines (KD, mutant/KD, wild type = 1. 7; Figure 4A and Table 2). In contrast, the C-terminal 531PY532/AA mutation abolished detectable Kap104p binding (Table 2). The enthalpies of binding for all of the PY-NLSs that we have measured by ITC are similar, and the weakest measurable KD in this series was 10 μM [26]. Therefore, we assume that the affinity of the Hrp1p 531PY532/AA mutant is likely weaker than 10 μM and its KD, mutant/KD, wild type > 200 (Figure 4A). Thus, the Hrp1p NLS contains one strong binding hotspot at its PY motif, similar to the single significant hotspot at the C-terminal PY motif of the human substrate hnRNP M (KD, mutantPY/AA/KD, wild type = 500 for the hnRNP M NLS) [26]. Interestingly, we also located a modest binding hotspot at residue Y529 (KD, mutant/KD, wild type = 4 for Y529A; Figure 4A and Table 2) in the linker between the arginine and the PY of the RX2–5PY C-terminal motif. However, the Y529L mutation did not affect Kap104p binding (Table 2), suggesting that a hydrophobic, but not necessarily aromatic, moiety at this position might be important. We have performed scanning alanine mutagenesis covering residues 210–239 of the Nab2p NLS (Figure 1A, Table 3, and Table S1). Binding energy along the Nab2p NLS appears quite distributed compared to that of the Hrp1p NLS, with no single binding hotspot that stands out above others (Figure 4B and Table 3). In its basic N-terminal region, 216KNRR219,222RGG224, and 226RGGRN230 each were mutated to alanines, but only 216KNRR219/AAAA showed a small 3-fold decrease in Kap104p affinity (Table 3). None of the single mutants K216A, R218A, R219A, R222A, R226A, or R229A decreased Kap104p binding (Table S1), and simultaneous mutation of all of the arginines to lysines also did not decrease Kap104p binding. In contrast, mutation of all five arginines to alanines decreased affinity by 60-fold (KD = 2. 25 μM; Figure 4B and Table 3), suggesting that the collective basic character of this region contributes significantly to the total binding energy of the NLS. Comparison of single arginine to alanine mutants (KD, mutant/KD, wild type ≈ 1. 0) to the pentamutant R218, R219, R222, R226, R229/AAAAA (KD, mutant/KD, wild type = 60. 8) indicated a binding cooperativity of at least 60-fold within the N-terminal basic motif of Nab2p. When R235 of the Nab2p C-terminal RX2–5PL motif was mutated to an alanine, Kap104p affinity decreased by 5-fold (Figure 4B and Table 3). Crystal structures of Kapβ2 bound to NLSs of hnRNPs A1 and M showed the equivalent arginine residues making electrostatic interactions with numerous aspartate and glutamate residues, suggesting the importance of a positively charged residue at this position [16,26]. We also mutated R235 to lysine and histidine, but neither mutant affected Kap104p binding significantly (KD, mutant/KD, wild type are 1. 0 and 1. 7, respectively; Table 3). The C-terminal 238PL239/AA mutation in the Nab2p NLS decreased Kap104p binding by 10-fold (Figure 4B and Table 3). The energetic significance of this mutation suggests its equivalence to the PY motif in human Kapβ2 substrates and in Hrp1p. Furthermore, the Nab2p 238PL239/PY mutant bound Kap104p with a slightly higher affinity at a KD value of 13 nM. Mutagenesis of residue L239 to all other amino acids is described below. The measurable 238PL239/AA mutation in the Nab2p NLS (KD = 376 nM) provided an opportunity to explore cooperativity across binding sites or epitopes. Mutations in the Nab2p triple mutant R222A, 238PL239/AA (KD 411 nM; KD, mutant/KD, wild type = 11. 1; Table 3) show almost perfect additivity when compared to a single R222A mutant (did not affect Kap104p binding; Table S1) and double mutant 238PL239/AA (KD, mutant/KD, wild type = 10. 2; Table 3). A second Nab2p triple mutant R235A, 238PL239/AA (KD = 544 nM; KD, mutant/KD, wild type = 14. 7; Table 3) also was compared to a single R235A mutant (KD, mutant/KD, wild type = 5. 5; Table S1) and double 238PL239/AA mutant (KD, mutant/KD, wild type = 10. 2; Table 3). Strict additivity between the R and the PL sites would give a calculated KD, mutant/KD, wild type value of 56. 1 for the triple mutant. Thus, the experimental KD, mutant/KD, wild type value of 14. 7 for the triple mutant indicated 3. 8-fold cooperativity between the two epitopes. Similarly, Hrp1p triple mutant R512A, 524RR525/AA and double mutant R512A, Y529A showed cooperativity of approximately 1. 4- and 2-fold between epitopes, respectively. The couplings between binding epitopes observed here for both Nab2p and Hrp1p are still more than an order of magnitude lower than that observed within the N-terminal basic region of Nab2p (>60-fold cooperativity). We also located a new binding hotspot at F236 in Nab2p (KD, mutant/KD, wild type = 8 for F236A; Figure 4B and Table 3), which is located in the linker between the R and the PL of the RX2–5PL C-terminal motif. This site is analogous to Y529 of Hrp1p discussed in the previous section, and both residues are located two residues N-terminal of the PY/L motifs. As in the Hrp1p NLS Y529L mutant, the F236L mutation in Nab2p did not affect Kap104p binding (Table 3). Aromatic or hydrophobic residues occur at this position in many human PY-NLSs, including hnRNPs M, D, and F, JKTBP, TAP, HMBA-inducible protein, PABP2, PQBP-1, RB15B, and WBS-16 [16,22,23,27]. Aromatic side chains at this position overlap in the crystal structures of Kapβ2 bound to the NLSs of hnRNPs M and D and TAP [26,27]. The F61 of the hnRNP M NLS, Y352 of the hnRNP D NLS, and Y72 of the TAP NLS make hydrophobic interactions with Kapβ2 W460A and with the backbones of the PY motifs. A hydrophobic residue here may contribute to binding energy through both favorable enthalpy and a decrease of entropic penalty upon binding by preorganizing the PY motif. Thus, if present, a hydrophobic residue here may be considered as an extension of the PY motif. Hrp1p contains a single very significant binding hotspot at its PY motif. In contrast, binding energy in Nab2p is more evenly distributed across its N-terminal basic region and the R, F, and PL residues of its C-terminal consensus motif. Thus, distributions of binding energy in the two yeast NLSs are very different. From the N to C terminus, energetic distribution across the three epitopes (N-terminal basic region, R, and PY/L of the C-terminal motif) of Hrp1p and Nab2p can be described roughly as medium–weak–strong and strong–medium–medium, respectively (ΔΔG < 0. 9 kcal/mol is categorized as weak, 0. 9 ≤ ΔΔG ≤ 1. 7 kcal/mol as medium, and ΔΔG > 1. 7 kcal/mol as strong; Figure 4A and 4B). Similarly, in previously characterized PY-NLSs of hnRNPs A1 and D, TAP, and JKTBP [16,27], energetic distributions at the three epitopes also are quite varied, with rough patterns of strong–weak–weak, strong–medium–medium, weak–weak–weak, and weak–medium–strong, respectively (Figure 4C–G). In summary, all three PY-NLS epitopes are energetically highly variable, the N-terminal basic/hydrophobic and the C-terminal PY motifs appear to cover the entire energetic continuum from strong to weak, and the arginine of the RX2–5PY motif is medium to weakly energetically significant. To examine the effect of PY-NLS mutations on nucleocytoplasmic localization of Hrp1p and Nab2p in vivo, we expressed GFP-tagged full-length Hrp1p and Nab2p wild-type and mutant proteins in yeast. Wild-type Hrp1–GFP and Nab2p–GFP are localized in the nucleus as has been reported previously (Figure 6A–D) [34–36]. Mutations in the C-terminal PY motif (531PY532/AA) of Hrp1p, which abolished detectable Kap104p binding, resulted in mislocalization of the GFP fusion protein to the cytoplasm (Figure 6A and 6C). The N-terminal basic motif of Hrp1p is also important for nuclear localization of Hrp1p: the R512, R513, R516, R519/AAAA mutant, which decreased Kap104p binding by a marginal 5-fold, also is mislocalized (Figure 6A and 6C). Xu and Henry have shown previously that substitutions of R516 and R519 with glutamines mislocalized Hrp1p, but proteins with lysine substitutions are properly localized [38,42]. This further suggests that basic charges rather than stereospecific interactions are necessary for Kap104p interactions. In the case of Nab2p, mutations in either the N-terminal motif (pentamutant R218, R219, R222, R226, R229/AAAAA; decreases Kap104p binding by 60-fold) or the C-terminal PY motif (238PL239/AA; decreases Kap104p binding by 10-fold) resulted in increased cytoplasmic localization of the GFP fusion protein (Figure 6B and 6D). Arginine methylation of Nab2p by Hmt1p is required for its export from the nucleus, possibly explaining some nuclear accumulation of the N-terminal mutant despite its low affinity for Kap104p [38,43]. Combined mutations of both the N- and the C-terminal motifs resulted in diffuse localization of the fusion protein, consistent with further affinity reduction for Kap104p (Table 3 and Figure 6B and 6D). We have shown here that mutations in the PY-NLSs of Hrp1p and Nab2p that decrease binding affinity to Kap104 also affect nuclear localization in yeast cells. Structures of PY-NLSs from hnRNPs A1, M, and D, TAP, and JKTBP converge spatially at three distinct binding sites or epitopes separated by structurally variable linkers: (1) the N-terminal hydrophobic/basic motif, (2) the arginine residue of the C-terminal RX2–5PY sequence motif, and (3) the PY of the C-terminal RX2–5PY motif [16,26,27]. We have shown here that all three structural epitopes can be energetically significant. The N-terminal basic-enriched motifs of Hrp1p and Nab2p NLSs constitute epitope 1, where collective basic character and likely charge density drive Kap104p binding. Mutations of all of the arginines in this region to alanines decreased binding energy by 0. 9–2. 3 kcal/mol for both NLSs. Similarly, the N-terminal hydrophobic motif of the hnRNP A1 NLS and the equivalent region of the hnRNP D NLS that contains both hydrophobic and basic residues are also energetically significant, with mutations decreasing binding energy by ∼2 kcal/mol [26]. Epitopes 2 and 3 are contained within the C-terminal RX2–5PY/L sequence motifs. Two linkers of variable lengths, compositions, and structures connect epitope 1 to epitope 2 and epitope 2 to epitope 3 [16,26]. Epitope 2 is located at Hrp1p 524RR525 and Nab2p R235 at the first consensus position of the C-terminal RX2–5PY/L sequence motifs. Of the three PY-NLS epitopes, epitope 2 tends to contribute the least to binding energy, with mutations decreasing binding energy maximally by ∼1 kcal/mol in Nab2p, hnRNP D, and JKTBP (Figure 4B, 4E, and 4G). Epitope 3 is located at Hrp1p 531PY532 and Nab2p 238PL239. Mutations at these terminal positions are generally energetically significant, decreasing binding energy by 1. 3–4 kcal/mol in Hrp1p, Nab2p, hnRNPs M and D, and JKTBP. However, exceptions are seen in hnRNP A1 and TAP, where PY mutations decreased binding modestly by only ∼0. 7 kcal/mol. Because free PY-NLSs are structurally disordered and adopt extended Kapβ2-bound conformations, epitopes 1–3 are presented as peptides that can be represented by sequence patterns or linear motifs [44–46]. In epitope 1, the N-terminal basic motif may be represented by a collection of sequence patterns covering 5–19 residues, and the N-terminal hydrophobic motif by sequence patterns of approximately 4 residues. Epitopes 2 and 3 are both relatively smaller and simpler and together can be described by a single sequence pattern. Comparison of validated and potential PY-NLSs in Table 1 [16,26] show that sequences within each of the three linear epitopes can be quite variable. The N-terminal basic/hydrophobic motif is the largest and most variable epitope. Mutagenesis of yeast PY-NLSs has provided more information on the diversity and also suggested some limits to the diversity of individual epitopes. In particular, positive charges within the N-terminal basic motifs are important, but arginine and lysine residues are interchangeable, and the exact positions of basic groups may not be important (Tables 2 and 3 and Table S1). Additional biochemical and structural studies will be needed to understand requirements of charge density, segment size, and negatively selected amino acids in this epitope. The consensus for this basic region remains elusive. The 55% accuracy for bioinformatics-derived potential yeast bPY-NLSs binding to Kap104p may reflect high sequence variability and undiscovered physical characteristics of this region. Epitope 2 is usually composed of a single residue. Examination of validated PY-NLSs (Table 1) shows that arginine is most prevalent in this position, although histidines are found in this position in hnRNP D, JKTBP, and HuR and lysines in potential yeast NLSs of Naf1p, Sbp1p, Arp8p, and Ste20p (Figure S3A). Mutagenesis has shown that arginine, lysine, and histidines are interchangeable in this position. Thus, the appropriate sequence pattern here is R/K/H. Human Kapβ2 substrate HuR (Table 1) has a PG dipeptide, and yeast Nab2p and eight bioinformatics-derived potential yeast NLSs contain PL dipeptides at the C-terminal positions of their NLSs (epitope 3). In some cases, epitope 3 matters energetically more than in others. It is unclear why the dipeptide motif is energetically significant in some peptides and relatively silent in others. We speculate that a hydrophobic amino acid two residues N-terminal of the PY motif may be necessary (though probably not sufficient) and should be included in the sequence pattern for an energetically strong epitope 3. A hydrophobic residue at this position may preorganize the short peptide segment for binding, lowering both strain and entropic penalties. We also note that if epitope 3 is energetically very significant, then the terminal site tends to be phenylalanine, histidine, and methionine. If the dipeptide motif is fairly silent energetically, then many other amino acids are allowed in the terminal position. Mutations within a linear epitope such as within the N-terminal basic region of Nab2p show large cooperativity of >60-fold (Table 2 and Table S1). Mutations within the N-terminal basic region of the hnRNP M NLS also show cooperativity, in a similar regime, of ∼40-fold [26]. In contrast, seven examples of simultaneous mutations between different linear epitopes in Hrp1p, Nab2p (Tables 2 and 3), and hnRNPs A1 and M [16,26] show only modest cooperativities of 1. 0–3. 8-fold. Cooperativity between linear epitopes in PY-NLSs is also very small compared to that typically observed between spatially distinct sites in conformational epitopes. For example, in the interaction of human growth hormone with human growth hormone receptor, mutations at distant sites in the interface showed large cooperativity of ∼60-fold [47]. Thus, by comparison, the linear epitopes in PY-NLSs are energetically quasi-independent. In an analogous system, a bipartite interaction in a linear sorting signal in a SNARE and COPII coat also exhibited energetic quasi-independence, showing only a 1. 5–2-fold cooperative effect between the two distant sites [48]. In both PY-NLSs and vesicular sorting signals, minimal coupling between linear epitopes, and thus energetic modularity of those epitopes, may be attributed to flexible or structurally variable linkers that connect the epitopes. Finally, the fourth biophysical property that governs PY-NLS affinity stems from the observation that binding energy is distributed very differently amongst the three linear epitopes in all seven thermodynamically characterized PY-NLSs [16,26,27]. In different PY-NLSs, a given linear epitope can vary significantly in its contribution to total binding energy. For example, the N-terminal basic motif in Hrp1p contributes much less to Kap104p binding than the equivalent epitope in Nab2p (compare Figure 4A and 4B). Similarly, PY in hnRNP A1 contributes only weakly to Kapβ2 binding, while PY motifs in hnRNP M and Hrp1p are the sole binding hotspots in the NLSs (Figure 4A, 4C, and 4D). We previously had taken advantage of the energetic variability of PY-NLS epitopes by harnessing the avidity effect of the NLS hotspot at epitope 1 of hnRNP A1 fused to the NLS hoptspot at epitope 3 of hnRNP M, which resulted in a chimeric peptide inhibitor that bound Kapβ2 200-fold tighter than both substrates and RanGTP [26]. Despite the wide energetic variability of individual linear epitopes, the total binding energies are very similar for various PY-NLS-containing substrates. Therefore, evolution has not combined epitopes randomly but rather tuned them to a range for appreciable Kapβ2 binding and efficient Ran dissociation. The extremely tight-binding chimeric peptide inhibitor of Kapβ2 [26] is evidence of such evolutionary pressure. Although very high affinity can be achieved easily, nuclear import function is lost as RanGTP can no longer dissociate substrates. Binding energy in the PY-NLS is distributed over a large sequence, with three different elements contributing differently in various substrates. It is this feature that makes the PY-NLS fundamentally different from the well-known monopartite classical NLS. A relatively small motif is recognized in a monopartite NLS, and binding energy is concentrated in a stereotypical fashion across small sequences. In PY-NLSs, the three distinct linear sequence elements are presented on peptides that exhibit intrinsic structural disorder and bind Kapβ2 with extended structurally diverse conformations. This modular and flexible display of multiple sequence motifs is relatively free of spatial constraints that usually relate multiple binding sites within a folded ligand. Furthermore, when binding energy is variably distributed among multiple epitopes in PY-NLSs, single mutations or mutations within single NLS epitopes are likely to have decreased chances of abolishing karyopherin binding. Thus, the modular, flexible, and energetically combinatorial architecture of PY-NLSs may allow significant evolvability to form new interactions while maintaining Kapβ2 recognition. Similar “multifaceted” interactions, where different ligands make energetically significant interactions with different subsets of interface residues, were recently studied in a theoretical context [49] and also suggested to be more tolerant to mutations and are therefore quite evolvable. Multiple functions have been identified in fact in several PY-NLSs. In Nab2p, the RGG region that overlaps NLS epitope 1 is a putative RNA binding region [50]. The PY-NLSs in Nab2p, Hrp1p, EWS, and FUS interact with and are methylated by arginine methyltransferases [43,51–54]. Phosphorylation sites also have evolved within PY-NLSs to regulate nucleocytoplasmic localization. Serine phosphorylation in the hnRNP A2 NLS and tyrosine phosphorylation in the SAM68 NLS [55] both alter subcellular localization of the proteins. A PY-NLS also may evolve additional NLSs within its sequence. This could generate redundancy in nuclear import pathways and also provide a path to switch substrates from one karyopherin to another and ultimately from one cellular process to another. We have identified a potential classical NLS [56] in the N-terminal basic motifs of eight human bPY-NLSs in Table 1. It is not clear what overlapping NLSs mean in the cellular context, but this question will need to be explored in the future. Identifying correct sequences that will account for most of the very diverse PY-NLS is an extremely challenging task. The core problem is that binding energy is distributed across three epitopes or motifs in many different ways. Thus, simply relaxing sequence constraints in a global search will also increase “noise” and result in many wrong answers. We predict that if a PY motif (epitope 3) is energetically very significant, then the sequence tolerance for this motif is small, and sequence content of the other two epitopes will likely not matter. Thus, this subset of the PY-NLSs should be identified easily upon identification of PY motifs that can provide large binding energies. Given the relatively small size of this motif, the task of finding strong PY motifs should be experimentally accessible. A similar situation should apply for an energetically strong N-terminal basic/hydrophobic motif (epitope 1). However, as the need for affinity from the PY motif decreases and as more binding energy is provided by the two other motifs, sequence tolerance relaxes. The problem of multiple motifs with varying sequence tolerances seems very complex, but the relatively small size of each motif and energetic independence of the motifs allow the problem to be divided into manageable pieces. Our current inability to identify sequences of individual epitopes that are energetically strong may contribute to the 55% accuracy for bioinformatics-derived potential yeast bPY/L-NLS binding to Kap104p. For example, individual epitopes in bioinformatics-derived sequences that did not bind Kap104p may be energetically weak and thus did not provide sufficient binding energy when combined. First, the range of energies for PY-NLSs that are import-competent in vivo (and to what degree) will need to be determined. The range of suitable binding energies likely will vary depending on cellular concentrations of substrates but should not be unbounded [57]. For example, a designed peptide with a KD of 100 pM binds Kapβ2 too tightly for in vivo nuclear import [26], thus providing a high-affinity boundary for Kapβ2 import. Second, binding energies of putative PY-NLSs will need to be predicted. Unfortunately, the accuracy of calculating binding affinity for protein–small molecule interaction is still questionable, and predictions of binding energies for protein–protein interactions are even further behind [58]. Our studies here suggest that we can get around this problem by handling each epitope independently and then combining them to assess for functional NLSs. We may use computational alanine-scanning mutagenesis [59] to predict binding energy differences for each of the three PY-NLS linear epitopes and then empirically determine combinations that are functional. Such predictions could be tested against a future experimental thermodynamic database obtained from the initial predicted PY-NLSs [16], and the method was refined iteratively. Binding energy calculation remains problematic. We expect that prevalent sequence- and physical-characteristics-based bioinformatics methods are limited to successful prediction of potential NLSs with at least one energetically strong linear epitope but will miss those composed of multiple weak or intermediate epitopes. A computational method that combines bioinformatics, structural modeling, and prediction of binding energies may be a solution. Many more Kapβ2–NLS structures will be necessary to expand a structural database to facilitate modeling interactions of new sequences by homology modeling and/or physical energy function-based predictions of protein–protein interactions [60–62]. PY-NLSs are very diverse in sequence and structure and thus cannot be described sufficiently by their weak consensus motifs. Instead, PY-NLSs are described by a collection of weak physical rules that also include requirements for intrinsic structural disorder and overall positive charge [16]. Here, we examined the energetic organization of PY-NLSs through mutagenic and thermodynamic analyses of these signals in yeast. These studies have revealed physical properties that govern the binding affinity of this variable signal. The PY-NLS is a modular signal composed of three spatially distinct but structurally conserved linear epitopes that can be represented by a series of sequence patterns. Although each linear epitope can accommodate substantial sequence diversity, we have begun to define limits for each. More importantly, in addition to structural modularity, the three linear epitopes also exhibit energetic modularity. Modular organization of the PY-NLS suggests that the daunting search for these very diverse sequences can be performed in parts. Finally, each linear epitope can contribute very differently to total binding energy in different PY-NLSs, explaining how signal diversity can be achieved through combinatorial mixing of energetically weak and strong motifs while maintaining affinity appropriate for nuclear import function. This collection of physical rules and properties describes how functional determinants of the PY-NLSs are organized and lays a path to decode this diverse and evolvable signal for future genome-wide identification of Kapβ2 import substrates. More generally, many biological recognition processes involve linear recognition motifs with weak and obscure sequence motifs. Physical understanding of how diverse PY-NLS sequences can achieve common biological function may serve as a model for decoding many other weakly conserved and complex signals throughout biology. The Kap104p gene (gift from J. Aitchison) was subcloned into the pGEX-Tev vector [34]. Yeast substrate genes were obtained by PCR from a S. cerevisiae genomic DNA library (Novagen) and subcloned into the BamHI and NotI sites of the pGEX-Tev and/or pMAL-Tev vectors [63,64]. Site-directed mutagenesis of Nab2p 201–251 and Hrp1p 494–534 were performed using the QuikChange method (Stratagene) and confirmed by nucleotide sequencing. Full-length Nab2p and Hrp1p wild-type and mutant genes were subcloned into the SpeI and SmaI sites of a modified pRS415 (CEN6, ARS, LEU2, and APR) shuttle vector containing a C-terminal GFP gene [65]. BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) cells harboring pRS415 plasmids were grown at 30 °C in SC-Leu media to mid-logarithmic phase [66]. Cells were transferred to a 1. 5% low-melting-point agarose pad made with SC-leu in a coverslip bottom Wilco dish. Cells were observed on an Olympus IX-81 inverted microscope (60× objective), and images were acquired with a Hamamatsu ORCA-ER camera. All images were analyzed in Image-Pro Plus software (Media Cybernetics). To obtain the N/C ratio, mean fluorescence intensity in a 36-pixel box was measured in the nucleus and cytoplasm for at least 50 cells of each mutant. The GST–Kap104p protein was expressed in Escherichia coli Rosetta (DE3) pLysS cells (Novagen). Cells were lysed using an EmulsiFlex-C5 homogenizer (Avestin). The supernatant was applied to glutathione sepharose (GE Healthcare) and washed extensively with Tris buffer (50 mM Tris, pH 7. 5,100 mM NaCl, 1 mM EDTA, 2 mM DTT, and 20% glycerol). GST–Kap104p was eluted with Tris Buffer plus 20 mM glutathione, pH 8. 1. The GST tag was cleaved using 0. 5 ml of TEV protease in a total volume of 10 ml and separated from Kap104p using an anion exchange column (GE Healthcare). Kap104p was purified further by gel filtration chromatography in TB buffer (20 mM HEPES, pH 7. 3,110 mM KAc, 2 mM MgAc, 1 mM EGTA, 2 mM DTT, and 20% glycerol). Yeast substrates and NLSs were expressed in E. coli BL21 (DE3). The maltose-binding protein (MBP) NLSs were lysed as above and purified by affinity chromatography using amylose resin (New England Biolabs). After extensive washing with Tris buffer, protein was eluted with Tris buffer plus 10 mM maltose. The protein was purified further by cation exchange chromatography. The GST substrates were lysed by sonication and immobilized on glutathione sepharose. The protein was washed with TB buffer and left on the beads for binding assays. Human substrates were expressed and purified as previously reported [16]. Potential Kap104p substrates were identified as described in Lee et al. [16]. Sequence patterns Φ1-G/A/S-Φ3-Φ4-X7–12-R/K/H-X2–5-P-Y/L (Φ1 is a hydrophobic residue and Φ3 and Φ4 are hydrophobic residues or R or K) and K/R-X0–6-K/R-X0–6-K/R-X0–6-K/R-X2–5-R/K/H-X1–5-PY were used in ScanProsite [40] to screen S. cerevisiae proteins in the UniProtKB/Swiss-Prot database [41]. Approximately 30 μg of Kap104p was added to ∼10 μg of GST protein immobilized on 20 μl of glutathione sepharose followed by extensive washes with TB buffer and a second incubation with either buffer or RanGTP (5-fold molar excess). Immobilized proteins were visualized with SDS-PAGE and Coomassie staining. Affinities of wild-type and mutant MBP–Nab2p NLS and MBP–Hrp1p NLS binding to Kap104p were determined by ITC using a MicroCal Omega VP-ITC calorimeter (MicroCal). Proteins were dialyzed against buffer containing 20 mM Tris, pH 7. 5,100 mM NaCl, 2 mM β-mercaptoethanol, and 10% glycerol. The 90–350 μM MBP-NLS proteins were titrated into a sample cell containing 9–35 μM Kap104p. All ITC experiments were done at 20 °C with 35 rounds of 8-μl injections. Data were plotted and analyzed with a single-site binding model using MicroCal Origin software (version 7. 0). Accession numbers for genes mentioned in this paper from the National Center for Biotechnology Information (http: //www. ncbi. nlm. nih. gov) are: Bbp1p (855820), Clg1p (852657), hnRNP A1 (3178), Hrp1p (853997), Kap104p (852305), Kapβ2 (3842), Nab2p (852755), Nam8p (856486), Pos5p (855913), Rml2p (856660), Sin3p (854158), Sko1p (855554), Snp1p (854749), TAP (10482), and Tfg2p (852888).

To travel between the cytoplasm and nucleus, proteins rely on a family of transport proteins known as the karyopherinβ family. Karyopherinβ2, the human version of a family member, recognizes cargo proteins containing a class of nuclear localization signal known as the PY-NLS. The yeast homolog of Karyopherinβ2, Kap104p, also recognizes PY-NLSs, indicating that this pathway has been conserved between evolutionarily distant species. We mutated residues in the PY-NLSs of two Kap104p cargo proteins and analyzed how tightly these mutants bound Kap104p. These experiments revealed three PY-NLS regions, or epitopes, that are important for binding Kap104p. Each epitope is composed of amino acids that vary between cargoes. The epitopes are energetically independent and bind Kap104p with varying strengths in different PY-NLSs, such that mutating the epitope of one PY-NLS may mistakenly direct cargo to the cytoplasm, while a similar mutation in a different PY-NLS has little effect on cargo localization. This flexible, energetically modular, and combinatorial architecture of PY-NLSs may confer higher tolerance to mutations, but it also allows greater sequence diversity, making prediction of new PY-NLSs difficult. The characteristics of PY-NLSs reported here will assist in the identification of new Kap104p cargoes. And the approach used may be applicable to other biological recognition pathways.

Abstract Introduction Results Discussion Materials and Methods Supporting Information

biophysics biochemistry

2008

Modular Organization and Combinatorial Energetics of Proline–Tyrosine Nuclear Localization Signals

One of the most intriguing questions in evolution is how organisms exhibit suitable phenotypic variation to rapidly adapt in novel selective environments. Such variability is crucial for evolvability, but poorly understood. In particular, how can natural selection favour developmental organisations that facilitate adaptive evolution in previously unseen environments? Such a capacity suggests foresight that is incompatible with the short-sighted concept of natural selection. A potential resolution is provided by the idea that evolution may discover and exploit information not only about the particular phenotypes selected in the past, but their underlying structural regularities: new phenotypes, with the same underlying regularities, but novel particulars, may then be useful in new environments. If true, we still need to understand the conditions in which natural selection will discover such deep regularities rather than exploiting ‘quick fixes’ (i. e. , fixes that provide adaptive phenotypes in the short term, but limit future evolvability). Here we argue that the ability of evolution to discover such regularities is formally analogous to learning principles, familiar in humans and machines, that enable generalisation from past experience. Conversely, natural selection that fails to enhance evolvability is directly analogous to the learning problem of over-fitting and the subsequent failure to generalise. We support the conclusion that evolving systems and learning systems are different instantiations of the same algorithmic principles by showing that existing results from the learning domain can be transferred to the evolution domain. Specifically, we show that conditions that alleviate over-fitting in learning systems successfully predict which biological conditions (e. g. , environmental variation, regularity, noise or a pressure for developmental simplicity) enhance evolvability. This equivalence provides access to a well-developed theoretical framework from learning theory that enables a characterisation of the general conditions for the evolution of evolvability. Explaining how organisms adapt in novel selective environments is central to evolutionary biology [1–5]. Living organisms are both robust and capable of change. The former property allows for stability and reliable functionality against genetic and environmental perturbations, while the latter provides flexibility allowing for the evolutionary acquisition of new potentially adaptive traits [5–9]. This capacity of an organism to produce suitable phenotypic variation to adapt to new environments is often identified as a prerequisite for evolvability, i. e. , the capacity for adaptive evolution [7,10,11]. It is thus important to understand the underlying variational mechanisms that enable the production of adaptive phenotypic variation [6,7, 12–18]. Phenotypic variations are heavily determined by intrinsic tendencies imposed by the genetic and the developmental architecture [18–21]. For instance, developmental biases may permit high variability for a particular phenotypic trait and limited variability for another, or cause certain phenotypic traits to co-vary [6,15,22–26]. Developmental processes are themselves also shaped by previous selection. As a result, we may expect that past evolution could adapt the distribution of phenotypes explored by future natural selection to amplify promising variations and avoid less useful ones by evolving developmental architectures that are predisposed to exhibit effective adaptation [10,13]. Selection though cannot favour traits for benefits that have not yet been realised. Moreover, in situations when selection can control phenotypic variation, it nearly always reduces such variation because it favours canalisation over flexibility [23,27–29]. Developmental canalisation may seem to be intrinsically opposed to an increase in phenotypic variability. Some, however, view these notions as two sides of the same coin, i. e. , a predisposition to evolve some phenotypes more readily goes hand in hand with a decrease in the propensity to produce other phenotypes [8,30,31]. Kirschner and Gerhart integrated findings that support these ideas under the unified framework of facilitated variation [8,32]. Similar ideas and concepts include the variational properties of the organisms [13], the self-facilitation of evolution [20] and evolution as tinkering [33] and related notions [6,7, 10,12]. In facilitated variation, the key observation is that the intrinsic developmental structure of the organisms biases both the amount and the direction of the phenotypic variation. Recent work in the area of facilitated variation has shown that multiple selective environments were necessary to evolve evolvable structures [25,27,34–36]. When selective environments contain underlying structural regularities, it is possible that evolution learns to limit the phenotypic space to regions that are evolutionarily more advantageous, promoting the discovery of useful phenotypes in a single or a few mutations [35,36]. But, as we will show, these conditions do not necessarily enhance evolvability in novel environments. Thus the general conditions which favour the emergence of adaptive developmental constraints that enhance evolvability are not well-understood. To address this we study the conditions where evolution by natural selection can find developmental organisations that produce what we refer to here as generalised phenotypic distributions—i. e. , not only are these distributions capable of producing multiple distinct phenotypes that have been selected in the past, but they can also produce novel phenotypes from the same family. Parter et al. have already shown that this is possible in specific cases studying models of RNA structures and logic gates [34]. Here we wish to understand more general conditions under which, and to what extent, natural selection can enhance the capacity of developmental structures to produce suitable variation for selection in the future. We follow previous work on the evolution of development [25] through computer simulations based in gene-regulatory network (GRN) models. Many authors have noted that GRNs share common functionality to artificial neural networks [25,37–40]. Watson et al. demonstrated a further result, more important to our purposes here; that the way regulatory interactions evolve under natural selection is mathematically equivalent to the way neural networks learn [25]. During evolution a GRN is capable of learning a memory of multiple phenotypes that were fit in multiple past selective environments by internalising their statistical correlation structure into its ontogenetic interactions, in the same way that learning neural networks store and recall training patterns. Phenotypes that were fit in the past can then be recreated by the network spontaneously (under genetic drift without selection) in the future or as a response to new selective environments that are partially similar to past environments [25]. An important aspect of the evolved systems mentioned above is modularity. Modularity has been a key feature of work on evolvability [6,29,41,42] aiming to facilitate variability that respects the natural decomposable structure of the selective environment, i. e. , keep the things together that need to be kept together and separate the things that are independent [6,12,20,41]. Accordingly, the system can perform a simple form of generalisation by separating knowledge from the context in which it was originally observed and re-deploying it in new situations. Here we show that this functional equivalence between learning and evolution predicts the evolutionary conditions that enable the evolution of generalised phenotypic distributions. We test this analogy between learning and evolution by testing its predictions. Specifically, we resolve the tension between canalisation of phenotypes that have been successful in past environments and anticipation of phenotypes that are fit in future environments by recognising that this is equivalent to prediction in learning systems. Such predictive ability follows simply from the ability to represent structural regularities in previously seen observations (i. e. , the training set) that are also true in the yet-unseen ones (i. e. , the test set). In learning systems, such generalization is commonplace and not considered mysterious. But it is also understood that successful generalisation in learning systems is not for granted and requires certain well-understood conditions. We argue here that understanding the evolution of development is formally analogous to model learning and can provide useful insights and testable hypotheses about the conditions that enhance the evolution of evolvability under natural selection [42,43]. Thus, in recognising that learning systems do not really ‘see into the future’ but can nonetheless make useful predictions by generalising past experience, we demystify the notion that short-sighted natural selection can produce novel phenotypes that are fit for previously-unseen selective environments and, more importantly, we can predict the general conditions where this is possible. This functional equivalence between learning and evolution produces many interesting, testable predictions (Table 1). In particular, the following experiments show that techniques that enhance generalisation in machine learning correspond to evolutionary conditions that facilitate generalised phenotypic distributions and hence increased evolvability. Specifically, we describe how well-known machine learning techniques, such as learning with noise and penalising model complexity, that improve the generalisation ability of learning models have biological analogues and can help us understand how noisy selective environments and the direct selection pressure on the reproduction cost of the gene regulatory interactions can enhance evolvability in gene regulation networks. This is a much more sophisticated and powerful form of generalisation than previous notions that simply extrapolate previous experience. The system does not merely extend its learned behaviour outside its past ‘known’ domain. Instead, we are interested in situations where the system can create new knowledge by discovering and systematising emerging patterns from past experience, and more notably, how the system separates that knowledge from the context in which it was originally observed, so that it can be re-deployed in new situations. Some evolutionary mechanisms and conditions have been proposed as important factors for improved evolvability. Some concern the modification of genetic variability (e. g. , [36,44,45] and [46]), while others concern the nature of selective environments and the organisation of development including multiple selective environments [36], sparsity [47], the direct selective pressure on the cost of connections (which can induce modularity [27,44] and hierarchy [48]), low developmental biases and constraints [49] and stochasticity in GRNs [50]. In this paper, we focus on mechanisms and conditions that can be unified and better understood in machine learning terms, and more notably, how we can utilise well-established theory in learning to characterise general conditions under which evolvability is enhanced. We thus provide the first theory to characterise the general conditions that enhance the evolution of developmental organisations that generalise information gained from past selection, as required to enhance evolvability in novel environments. The main experimental setup involves a non-linear recurrent GRN which develops an embryonic phenotypic pattern, G, into an adult phenotype, Pa, upon which selection can act [25]. An adult phenotype represents the gene expression profile that results from the dynamics of the GRN. Those dynamics are determined by the gene regulatory interactions of the network, B [38,39,47,53,54] (see Developmental Model in S1 Appendix). We evaluate the fitness of a given genetic structure based on how close the developed phenotype is to the target phenotypic pattern, S. S characterises the direction of selection for each phenotypic trait, i. e. , element of gene expression profile, in the current environment. The dynamics of selective environments are modelled by switching from one target phenotype to another every K generations. K is chosen to be considerably smaller than the overall number of generations simulated. Below, we measure evolutionary time in epochs, where each epoch denotes NT × K generations and NT corresponds to the number of target phenotypes. (Note that epoch here is a term we are borrowing from machine learning and does not represent geological timescale.) In the following experiments all phenotypic targets are chosen from the same class (as in [25,34]). This class consists of 8 different modular patterns that correspond to different combinations of sub-patterns. Each sub-pattern serves as a different function as pictorialised in Fig 1. This modular structure ensures that the environments (and thus the phenotypes that are fittest in those environments) share common regularities, i. e. , they are all built from different combinations from the same set of modules. We can then examine whether the system can actually ‘learn’ these systematicities from a limited set of examples and thereby generalise from these to produce novel phenotypes within the same class. Our experiments are carried out as follows. The population is evolved by exposure to a limited number of selective environments (training). We then analyse conditions under which new phenotypes from the same family are produced (test). As an exemplary problem we choose a training set comprised of three phenotypic patterns from the class (see Fig 2a). One way to evaluate the generalisation ability of developmental organisations is to evolve a population to new selective environments and evaluate the evolved predisposition of the development system to produce suitable phenotypes for those environments (as per [34]). We do this at the end of experimental section. We also use a more stringent test and examine the spontaneous production of such phenotypes induced by development from random genetic variation. Specifically, we examine what phenotypes the evolved developmental constraints and biases B are predisposed to create starting from random initial gene expression levels, G. For this purpose, we perform a post-hoc analysis. First, we estimate the phenotypic distributions induced by the evolved developmental architecture under drift. Since mutation on the direct effects on the embryonic phenotypes (G) in this model is much greater than mutation on regulatory interactions (B) (see Methods), we estimate drift with a uniformly random distribution over G (keeping B constant). Then we assess how successful the evolved system is at producing high-fitness phenotypes, by seeing if the phenotypes produced by the evolved correlations, B, tend to be members of the general class (see Methods). In this section, we focus on the conditions that promote the evolution of adaptive developmental biases that facilitate generalised variational structures. To address this, we examine the distributions of potential phenotypic variants induced by the evolved developmental structure in a series of different evolutionary scenarios: 1) different time-scales of environmental switching, 2) environmental noise and 3) direct selection pressure for simple developmental processes applied via a the cost of ontogenetic interactions favouring i) weak and ii) sparse connectivity. We next asked why costly interactions and noisy environments facilitate generalised developmental organisations. To understand this, we monitor the match between the phenotypic distribution induced by the evolved developmental process and the ones that describe the past selective environments (training set) and all potential selective environments (test set) respectively over evolutionary time in each evolutionary setting (see Methods). Following conventions in learning theory, we term the first measure ‘training error’ and the second ‘test error’. This demonstrates predictions c, e and f from Table 1. The dependence of the respective errors on evolutionary time are shown in Fig 3. For the control scenario (panel A) we observe the following trend. Natural selection initially improved the fit of the phenotypic distributions to both distributions of past and future selective environments. Then, while the fit to past selective environments continued improving over evolutionary time, the fit to potential, but yet-unseen, environments started to deteriorate (see also Fig B in Supporting Figures in S1 Appendix). The evolving organisms tended to accurately memorise the idiosyncrasies of their past environments, at the cost of losing their ability to retain appropriate flexibility for the future, i. e. , over-fitting. The dashed-line in Fig 3A indicates when the problem of over-fitting begins, i. e. , when the test error first increases. We see that canalisation can be opposed to the evolution of generalised phenotypic distributions in the same way over-fitting is opposed to generalisation. Then, we expect that preventing the canalisation of past targets can enhance the generalisation performance of the evolved developmental structure. Indeed, Fig 3B, 3C and 3D confirm this hypothesis (predictions a-c from Table 1). In the presence of environmental noise, the generalisation performance of the developmental structure was improved by discovering a set of regulatory interactions that corresponds to the minimum of the generalisation error curve of 0. 34 (Fig 3B). However, natural selection in noisy environments was only able to postpone canalisation of past targets and was unable to avoid it in the long term. Consequently, stochasticity improved evolvability by decreasing the speed at which over-fitting occurs, allowing for the developmental system to spend more time at a state which was characterised by high generalisation ability (see also Fig A in The Structure of Developmental Organisation in S1 Appendix). On the other hand, under the parsimony pressure for weak connectivity, the evolving developmental system maintained the same generalisation performance over evolutionary time. The canalisation of the selected phenotypes was thus prevented by preventing further limitation of the system’s phenotypic variability. Note that the outcome of these two methods (Fig 3B and 3C) resembles in many ways the outcome as if we stopped at the moment when the generalisation error was minimum, i. e. , early stopping; an ad-hoc solution to preventing over-fitting [51]. Accordingly, learning is stopped before the problem of over-fitting begins (see also Fig A in The Structure of Developmental Organisation in S1 Appendix). Under parsimony pressure for sparse connectivity, we observe that the generalisation error of the evolving developmental system reached zero (Fig 3D). Accordingly, natural selection successfully exploited the time-invariant regularities of the environment properly representing the entire class (Fig 2h). Additionally, Fig D in Supporting Figures in S1 Appendix shows that the entropy of the phenotypic distribution reduces as expected over evolutionary time as the developmental process increasingly canalises the training set phenotypes. In the case of perfect generalisation to the class (sparse connectivity), this convergence reduces from 16 bits (the original phenotype space) to four bits, corresponding to four degrees of freedom where each of the four modules vary independently. In the other cases, overfitting is indicated by reducing to less than four bits. As seen so far, the generalisation ability of development can be enhanced under the direct selective pressure for both sparse and weak connectivity and the presence of noise in the selective environment, when the strength of parsimony pressure and the level of noise were properly tuned. Different values of λ and κ denote different evolutionary contexts, where λ determines the relative burden placed on the fitness of the developmental system due to reproduction and maintenance of its elements, or other physical constraints and limitations, and κ determines the amount of extrinsic noise found in the selective environments (see Evaluation of fitness). In the following, we analyse the impact of the strength of parsimony pressure and the level of environmental noise on the evolution of generalised developmental organisations. Simulations were run for various values of parameters λ and κ. Then, the training and generalisation error were evaluated and recorded (Fig 4). This demonstrates prediction (g) from Table 1. We find that in the extremes, low and high levels of parsimony pressures, or noise, gave rise to situations of over-fitting and under-fitting respectively (Fig 4). Very small values of λ, or κ, were insufficient at finding good regulatory interactions to facilitate high evolvability to yet-unseen environments, resulting in the canalisation of past targets, i. e. , over-fitting. On the other hand, very large values of λ over-constrained the search process hindering the acquisition of any useful information regarding environment’s causal structure, i. e. , under-fitting. Specifically, with a small amount of L1-regularisation, the generalisation error is dropped to zero. This outcome holds for a wide spectrum of the regularisation parameter ln (λ) ∈ [0. 15,0. 35]. However, when λ is very high (here λ = 0. 4), the selective pressure on the cost of connection was too large; this resulted in the training and the generalisation errors corresponds to the original ‘no model’ situation (Fig 4C). Similarly, with a small amount of L2-regularisation, the generalisation error quickly drops. In the range [10,38] the process became less sensitive to changes in λ, resulting in one optimum at λ = 38 (Fig 4B). Similar results were also obtained for jittering (Fig 4A). But the generalisation performance of the developmental process changes ‘smoothly’ with κ, resulting in one optimum at κ = 35 × 10−4 (Fig 4A). Inductive biases need to be appropriate for a given problem, but in many cases a moderate bias favouring simple models is sufficient for non-trivial generalisation. Lastly we examine whether generalised phenotypic distributions can actually facilitate evolvability. For this purpose, we consider the rate of adaptation to each of all potential selective environments as the number of generations needed for the evolving entities to reach the respective target phenotype. To evaluate the propensity of the organisms to reach a target phenotype as a systemic property of its developmental architecture, the regulatory interactions were kept fixed, while the direct effects on the embryonic phenotype were free to evolve for 2500 generations, which was empirically found to be sufficient for the organisms to find a phenotypic target in each selective environment (when that was allowed by the developmental structure). In each run, the initial gene expression levels were uniformly chosen at random. The results here were averaged over 1000 independent runs, for each selective environment and for each of the four different evolutionary scenarios (as described in the previous sections). Then, counts of the average number of generations to reach the target phenotype of the corresponding selective environment were taken. This was evaluated by measuring the first time the developmental system achieved maximum fitness possible. If the target was not reached, the maximum number of generations 2500 was assigned. We find that organisms with developmental organisations evolved in noisy environments or the parsimony pressure on the cost of connections adapted faster than the ones in the control scenario (Fig 5). The outliers in the evolutionary settings of moderate environmental switching, noisy environments and favouring weak connectivity, indicate the inability of the developmental system to express the target phenotypic pattern for that selective environment due to the strong developmental constraints that evolved in those conditions. This corresponds to the missing phenotype from the class we saw above in the evolved phenotypic distributions induced by development (Fig 2e, 2f and 2g). In all these three cases development allowed for the production of the same set of phenotypic patterns. Yet, developmental structures evolved in the presence of environmental noise or under the pressure for weak connectivity exhibited higher adaptability due to their higher propensity to produce other phenotypes of the structural family. In particular, we see that for the developmental process evolved under the pressure for sparsity, the rate of adaptation of the organisms was significantly improved. The variability structure evolved under sparsity to perfectly represent the functional dependencies between phenotypic traits. Thus, it provided a selective advantage guiding phenotypic variation in more promising directions. The above experiments demonstrated the transfer of predictions from learning models into evolution, by specifically showing that: a) the evolution of generalised phenotypic distributions is dependent on the time-scale of environmental switching, in the same way that generalisation in online learning algorithms is learning-rate dependent, b) the presence of environmental noise can be beneficial for the evolution of generalised phenotypic distributions in the same way training with corrupted data can improve the generalisation performance of learning systems with the same limitations, c) direct selection pressure for weak connectivity can enhance the evolution of generalised phenotypic distributions in the same way L2-regularisation can improve the generalisation performance in learning systems, d) noisy environments result in similar behaviour as favouring weak connectivity, in the same way that Jittering can have similar effects to L2-regularisation in learning systems, e) direct selection pressure for sparse connectivity can enhance the evolution of generalised phenotypic distributions in the same way that L1-regularisation can improve the generalisation performance in learning systems, f) favouring weak connectivity (i. e. , L2-regularisation) results in similar behaviour to early stopping, g) the evolution of generalised phenotypic distributions is dependent on the strength of selection pressure on the cost of connections and the level of environmental noise, in the same way generalisation is dependent on the level of inductive biases and h) in simple modularly varying environments with independent modules, sparse connectivity enhances the generalisation of phenotypic distributions better than weak connectivity, in the same way that in problems with independent features, L1-regularisation results in better generalisation than L2-regularisation. Learning is generally contextual; it gradually builds upon what concepts are already known. Here these concepts correspond to the repeated modular sub-patterns persisting over all observations in the training set which become encoded in the modular components of the evolved network. The inter-module connections determine which combinations of (sub-) attractors in each module are compatible and which are not. Therefore, the evolved network representation can be seen as dictating a higher-order conceptual (combinatorial) space based on previous experience. This enables the evolved developmental system to explore permitted combinations of features constrained by past selection. Novel phenotypes can thus arise through new combinations of previously selected phenotypic features explicitly embedded in the developmental architecture of the system [25]. Indeed, under the selective pressure for sparse connectivity, we observe that the phenotypic patterns generated by the evolved developmental process consisted of combinations of features from past selected phenotypic patterns. Thus, we see that the ‘developmental memories’ are stored and recalled in combinatorial fashion allowing generalisation. We see that noisy environments and the parsimony pressure on the cost of connections led to more evolvable genotypes by internalising more general models of the environment into their developmental organisation. The evolved developmental systems did not solely capture and represent the specific idiosyncrasies of past selective environments, but internalised the regularities that remained time-invariant in all environments of the given class. This enabled natural selection to ‘anticipate’ novel situations by accumulating information about and exploiting the tendencies in that class of environments defined by the regularities. Peculiarities of past targets were generally represented by weak correlations between phenotypic characters as these structural regularities were not typically present in all of the previously-seen selective environments. Parsimony pressures and noise then provided the necessary selective pressure to neglect or de-emphasise such spurious correlations and maintain only the strong ones which tended to correspond to the underlying problem structure (in this case, the intra-module correlations only, allowing all combinations of fit modules). More notably, we see that the parsimony pressure for sparsity favoured more evolvable developmental organisations that allowed for the production of a novel and otherwise inaccessible phenotype. Enhancing evolvability by means of inductive biases is not for granted in evolutionary systems any more than such methods have guarantees in learning systems. The quality of the method depends on information about past targets and the strength of the parsimony pressure. Inductive biases can however constrain phenotypic evolution into more promising directions and exploit systematicities in the environment when opportunities arise. In this study we demonstrated that canalisation can be opposed to evolvability in biological systems the same way under- or over-fitting can be opposed to generalisation in learning systems. We showed that conditions that are known to alleviate over-fitting in learning are directly analogous to the conditions that enhance the evolution of evolvability under natural selection. Specifically, we described how well-known techniques, such as learning with noise and penalising model complexity, that improve the generalisation ability of learning models can help us understand how noisy selective environments and the direct selection pressure on the reproduction cost of the gene regulatory interactions can enhance context-specific evolvability in gene regulation networks. This opens-up a well-established theoretical framework, enabling it to be exploited in evolutionary theory. This equivalence demystifies the basic idea of the evolution of evolvability by equating it with generalisation in learning systems. This framework predicts the conditions that will enhance generalised phenotypic distributions and evolvability in natural systems. We model the evolution of a population of GRNs under strong selection and weak mutation where each new mutation is either fixed or lost before the next arises. This emphasises that the effects we demonstrate do not require lineage-level selection [61–63]—i. e. , they do not require multiple genetic lineages to coexist long enough for their mutational distributions to be visible to selection. Accordingly a simple hill-climbing model of evolution is sufficient [25,36]. The population is represented by a single genotype [G, B] (the direct effects and the regulatory interactions respectively) corresponding to the average genotype of the population. Similarly, mutations in G and B indicate slight variations in population means. Consider that G′ and B′ denote the respective mutants. Then the adult mutant phenotype, P a ′, is the result of the developmental process, which is characterised by the interaction B′, given the direct effects G′. Subsequently, the fitness of Pa and P a ′ are calculated for the current selective environment, S. If f S (P a ′) > f S (P a), the mutation is beneficial and therefore adopted, i. e. , Gt+1 = G′ and Bt+1 = B′. On the other hand, when a mutation is deleterious, G and B remain unchanged. The variation on the direct effects, G, occurs by applying a simple point mutation operator. At each evolutionary time step, t, an amount of μ1 mutation, drawn from [−0. 1,0. 1] is added to a single gene i. Note that we enforce all gi ∈ [−1,1] and hence the direct effects are hard bounded, i. e. , gi = min{max{gi + μ1, −1}, 1}. For a developmental architecture to have a meaningful effect on the phenotypic variation, the developmental constraints should evolve considerably slower than the phenotypic variation they control. We model this by setting the rate of change of B to lower values as that for G. More specifically, at each evolutionary time step, t, mutation occurs on the matrix with probability 1/15. The magnitude μ2 is drawn from [−0. 1/ (15N2), 0. 1/ (15N2) ] for each element bij independently, where N corresponds to the number of phenotypic traits. Following the framework used in [64], we define the fitness of the developmental system as a benefit minus cost function. The benefit of a given genetic structure, b, is evaluated based on how close the developed adult phenotype is to the target phenotype of a given selective environment. The target phenotype characterises a favourable direction for each phenotypic trait and is described by a binary vector, S = 〈s1, …, sN〉, where si ∈ {−1,1}, ∀i. For a certain selective environment, S, the selective benefit of an adult phenotype, Pa, is given by (modified from [25]): b = w (P a, S) = 1 2 1 + P a · S N, (1) where the term Pa ⋅ S indicates the inner product between the two respective vectors. The adult phenotype is normalised in [−1,1] by Pa ← Pa/ (τ1/τ2), i. e. , b ∈ [0,1]. The cost term, c, is related to the values of the regulatory coefficients, bij ∈ B [65]. The cost represents how fitness is reduced as a result of the system’s effort to maintain and reproduce its elements, e. g. , in E. coli it corresponds to the cost of regulatory protein production. The cost of connection has biological significance [27,64–67], such as being related to the number of different transcription factors or the strength of the regulatory influence. We consider two cost functions proportional to i) the sum of the absolute magnitudes of the interactions, c = ∥ B ∥ 1 = ∑ i = 1 N 2 | b i j | / N 2, and ii) the sum of the squares of the magnitudes of the interactions, c = ∥ B ∥ 2 2 = ∑ i = 1 N 2 b i j 2 / N 2, which put a direct selection pressure on the weights of connections, favouring sparse (L1-regularisation) and weak connectivity (L2-regularisation) respectively [68]. Then, the overall fitness of Pa for a certain selective environment S is given by: f S (P a) = b - λ c, (2) where parameter λ indicates the relative importance between b and c. Note that the selective advantage of structure B is solely determined by its immediate fitness benefits on the current selective environment. The χ2 measure is used to quantify the lack of fit of the evolved phenotypic distribution P t ^ (s i) against the distribution of the previously experienced target phenotypes Pt (si) and/or the one of all potential target phenotypes of the same family P (si). Consider two discrete distribution profiles, the observed frequencies O (si) and the expected frequencies E (si), si ∈ S, ∀i = 1, …, k. Then, the chi square error between distribution O and E is given by: χ 2 (O, E) = ∑ i (O (s i) - E (s i) ) 2 E (s i) (3) S corresponds to the training set and the test set when the training and the generalisation error are respectively estimated. Each si ∈ S indicates a phenotypic pattern and P (si) denotes the probability of this phenotype pattern to arise. The samples, over which the distribution profiles are estimated, are uniformly drawn at random (see Estimating the empirical distributions). This guarantees that the sample is not biased and the observations under consideration are independent. Although the phenotypic profiles here are continuous variables, they are classified into binned categories (discrete phenotypic patterns). These categories are mutually exclusive and the sum of all individual counts in the empirical distribution is equal to the total number of observations. This indicates that no observation is considered twice, and also that the categories include all observations in the sample. Lastly, the sample size is large enough to ensure large expected frequencies, given the small number of expected categories. For the estimation of the empirical (sample) probability distribution of the phenotypic variants over the genotypic space, we follow the Classify and Count (CC) approach [69]. Accordingly, 5000 embryonic phenotypes, P (0) = G, are uniformly generated at random in the hypercube [−1,1]N. Next, each of these phenotypes is developed into an adult phenotype and the produced phenotypes are categorised by their closeness to target patterns to take counts. Note that the development of each embryonic pattern in the sample is unaffected by development of other embryonic patterns in the sample. Also, the empirical distributions are estimated over all possible combinations of phenotypic traits, and thus each developed phenotype in the sample falls into exactly one of those categories. Finally, low discrepancy quasi-random sequences (Sobol sequences; [70]) with Matousek’s linear random scramble [71] were used to reduce the stochastic effects of the sampling process, by generating more homogeneous fillings over the genotypic space.

A striking feature of evolving organisms is their ability to acquire novel characteristics that help them adapt in new environments. The origin and the conditions of such ability remain elusive and is a long-standing question in evolutionary biology. Recent theory suggests that organisms can evolve designs that help them generate novel features that are more likely to be beneficial. Specifically, this is possible when the environments that organisms are exposed to share common regularities. However, the organisms develop robust designs that tend to produce what had been selected in the past and might be inflexible for future environments. The resolution comes from a recent theory introduced by Watson and Szathmáry that suggests a deep analogy between learning and evolution. Accordingly, here we utilise learning theory to explain the conditions that lead to more evolvable designs. We successfully demonstrate this by equating evolvability to the way humans and machines generalise to previously-unseen situations. Specifically, we show that the same conditions that enhance generalisation in learning systems have biological analogues and help us understand why environmental noise and the reproductive and maintenance costs of gene-regulatory connections can lead to more evolvable designs.

Abstract Introduction Results and discussion Methods

learning organismal evolution acoustics social sciences neuroscience learning and memory developmental biology cognitive psychology sound pressure evolutionary adaptation evolutionary genetics physics psychology phenotypes natural selection genetics biology and life sciences physical sciences evolutionary biology cognitive science evolutionary processes evolutionary developmental biology

2017

How evolution learns to generalise: Using the principles of learning theory to understand the evolution of developmental organisation

Pathogenic bacterial infections of the lung are life threatening and underpin chronic lung diseases. Current treatments are often ineffective potentially due to increasing antibiotic resistance and impairment of innate immunity by disease processes and steroid therapy. Manipulation miRNA directly regulating anti-microbial machinery of the innate immune system may boost host defence responses. Here we demonstrate that miR-328 is a key element of the host response to pulmonary infection with non-typeable haemophilus influenzae and pharmacological inhibition in mouse and human macrophages augments phagocytosis, the production of reactive oxygen species, and microbicidal activity. Moreover, inhibition of miR-328 in respiratory models of infection, steroid-induced immunosuppression, and smoke-induced emphysema enhances bacterial clearance. Thus, miRNA pathways can be targeted in the lung to enhance host defence against a clinically relevant microbial infection and offer a potential new anti-microbial approach for the treatment of respiratory diseases. Pathogenic bacterial infections of the respiratory tract are major causes of morbidity, are difficult to treat and can be life-threatening [1]. They also play a critical role in the pathogenesis of many inflammatory conditions of the lung (e. g. chronic obstructive pulmonary disease (COPD) and cystic fibrosis) and are major causes of acute exacerbations of pre-existing disease [2–4]. Current approaches to the treatment of these diseases and associated exacerbations are often ineffective. A potential explanation is that bacteria are becoming increasingly resistant to antibiotics and effective microbicidal activity requires a robust host defence response, which is impaired in infection-prone patients by underlying disease processes and immunosuppressive therapies (e. g. corticosteroids) [5–9]. A possible new treatment approach is to develop ways of boosting the innate host response, which bacteria cannot easily circumvent. Recently, important roles for microRNA (miRNA) in regulating innate host defence responses and acquired immunity have been identified [10,11]. In particular, miRNA expression is intimately linked to activation of pathogen recognition pathways (e. g. Toll-Like Receptors (TLR) ) that sense invading pathogen and promote immune cell recruitment which in turn leads to elimination of infectious agents. MiRNAs are small non-coding RNAs of approximately 22 nucleotides in length and individual miRNA have the capacity to bind to a multitude of mRNA molecules in a sequence specific manner to inhibit their translation [12]. Thus, a single or set of miRNAs has the potential to control an entire cellular pathway and its related networks. Key examples of miRNAs that are known to be activated by pathogen associated molecular patterns (PAMPs) include miR-9 [13], miR-146 [14] and miR-155 [14], which are important in regulating inflammatory pathways in macrophages and neutrophils by controlling TLR signaling. In addition, recent studies have demonstrated that miRNAs such as miR-155 [15,16], miR-21 [17], and miR-29 [18] are involved in regulating bacterial infections through the innate immune system. In search of new anti-microbial therapeutic approaches to treat respiratory infections we used non-typeable Haemophilus Influenzae (NTHi) as a model to investigate the roles of miRNA in regulating the innate host immune response to infection. NTHi is a commonly isolated bacterium from patients with chronic lung disease and is often linked to exacerbations of COPD [19,20]. During NTHi infection, macrophages and neutrophils are the key innate immune cells recruited to the lung to combat the bacterium. Activation of these cells induces phagocytosis and cytokine secretion that recruits immune cells and facilitates bacterial clearance [21,22]. Here we demonstrate that down-regulation of miR-328-3p (termed miR-328 hereafter) is a key element of the innate host defence response to NTHi infection, which facilitates bacterial clearance. Moreover, pharmacological inhibition of miR-328 profoundly enhances the clearance of the infection by increasing bacterial uptake by phagocytes, the production of reactive oxygen species (ROS), and microbicidal activity. Notably, inhibition of miR-328 in the lung was effective in amplifying the clearance of infection even in models of corticosteroid-induced immunosuppression and cigarette smoke-induced emphysema. Our studies provide the first proof-of-principle data that miRNA pathways can be manipulated in the lung to enhance host defence against microbial infection, and suggest a potential new anti-microbial approach to the treatment of infection induced respiratory diseases. We first analyzed the kinetics of bacterial clearance and cellular infiltration into the lungs of NTHi infected BALB/c mice. Mice were inoculated intratracheally (i. t.) with NTHi (5x105 CFU) and infection peaked between 6–12 hours (h) and was cleared within 24 h (Fig 1A). The numbers of total cells (Fig 1B) and neutrophils (Fig 1C) in BAL fluid was significantly increased 6 h post NTHi infection (p. i.). From 24–48 h p. i. total cells (Fig 1B), neutrophils (Fig 1C), and macrophages (Fig 1D) were increased significantly, which coincided with clearance of the infection. To assess the expression and regulation of miRNA during bacterial infection, we performed array analysis on total RNA isolated from the airways of NTHi-infected versus sham-exposed mice 24 h p. i. , when cells were actively clearing bacteria. The expression of 15 miRNAs were up-regulated, while 49 were down-regulated >2. 5-fold (S1 Fig). We validated the microarray data for several miRNAs using TaqMan PCR and both the pattern of expression and quantitative changes were confirmed. Among these differentially expressed miRNA we selected miR-328, specifically the 3p strand, as a candidate miRNA for further study as its baseline expression was among the highest observed and most significantly down-regulated following infection. Additionally, the roles of miR-328 in immunity to pathogens and in inflammation had not been characterised. We validated the miRNA array data for miR-328 using TaqMan PCR and observed a ~2-fold reduction in expression 24 h after infection (Fig 1E). Interestingly, miR-328 was decreased within 3 h in infected airways and remain decreased over a 48 h period (S2A Fig). Similarly, levels of miR-328 are diminished and remain lower in macrophages isolated from the NTHi infected lung over a similar time period (S2B Fig). To examine the function of miR-328 in microbial host defence responses in the lung, we isolated the two major innate immune cells that respond to infection and were elevated in the lungs; macrophages and neutrophils. Macrophages were purified from lungs of naïve mice and pre-treated with a miRNA inhibitor (antagomir (ant) ) with perfect complementarity to miR-328 (ant-328). This blocked miR-328 function before the cells were infected with NTHi in vitro. An antagomir with a scrambled sequence (Scr) was used as a negative control. Antagomir concentrations were titrated such that at the dose used approximately 100% of macrophages (or neutrophils) in the cultures contained cytoplasmic antagomir. Exposure of macrophages to NTHi and Scr resulted in a decrease in the levels of miR-328 by ~25% as assessed by TaqMan qPCR (Fig 2A) (this level was similar to NTHi exposure alone). Administration of ant-328 inhibited expression of miR-328 (Fig 2A). Macrophages treated with ant-328 had a significantly reduced bacterial load in the culture supernatant and increased bacterial uptake compared to the Scr treated controls (Fig 2B and 2C). Similar results were obtained with neutrophils purified from the bone marrow of naïve mice and exposed to the above treatments. Inhibition of miR-328 in neutrophils (S3A Fig) resulted in a 3-fold decrease in bacterial load in the culture supernatant as early as 1 h p. i. and a dramatic increase in bacterial uptake compared to Scr control treatment (S3B–S3C Fig, respectively). To determine whether the effects of miR-328 inhibition were due to altered phagocytosis of bacteria, or increased permissibility of phagocytes to active infection, we conducted a phagocytosis assay. Macrophages or neutrophils were exposed to heat-killed NTHi (to remove its ability to infect cells) labelled with CFSE and treated with ant-328 (or Scr control) in vitro for 1 h, and uptake assessed both by flow cytometry and fluorescence microscopy. Inhibition of miR-328 function in macrophages (Fig 2D and 2E) and neutrophils (S3D Fig) substantially enhanced phagocytosis of heat-killed NTHi (significantly increased levels of intracellular CFSE). The use of trypan blue in these experiments to quench CFSE fluorescence from extracellular bacteria attached to the cell surface confirmed the increased intracellular uptake after ant-328 treatment. To directly demonstrate that ant-328 increased the phagocytosis of bacteria we pre-treated macrophages with both antagomir and cytochalasin D (a potent phagocytosis inhibitor) before exposing the cells to heat-killed CFSE-labelled bacteria. Inhibition of phagocytsosis by cytochalasin D significantly reduced the effect of ant-328 on uptake of bacteria (S4 Fig). The marked effect on bacterial clearance by ant-328 was not related to apoptosis since there were no significant difference in the rate of cell death following antagomir treatment in either macrophages or neutrophils (S5A and S5B Fig). Treatment of macrophages with the mimetic to miR-328 had no effect on bacterial clearance likely because the basal expression level of miR-328 in lung macrophages is already high and so any additive effect becomes masked (S6A and S6B Fig). We next determined how the effects of ant-328 on enhancing phagocytosis may be regulated by examining specific bacterial binding and uptake pathways. We observed that treatment of macrophages with ant-328 significantly increased the expression of multiple bacterial uptake pathways including the LPS binding molecule CD14, the non-opsonic scavenger receptor CD36, and the bacterial adhesion integrin CD11b (S7A and S7C Fig). Collectively, this data suggests that miR-328 regulates bacterial phagocytosis in neutrophils and macrophages at least partly through the control of cell surface bacterial binding proteins. To mechanistically extend our observations, we looked at some important events that occurred following phagocytosis. Using confocal microscopy, we showed that pre-treatment of macrophages with ant-328 increased expression of the lysosomal enzyme, Cathepsin D (increased red fluorescence in cytoplasm), following NTHi infection compared to Scr control treated macrophages (Fig 2F). Activation of the respiratory oxidative burst and the production of ROS is another crucial event in bacterial killing by lysosomes in phagocytes. Thus, we determined whether miR-328 plays a role in activating these killing pathways in macrophages and neutrophils. Following NTHi infection, pre-treatment of macrophages with ant-328 significantly increased both the number of cells producing ROS and the intensity of its production per cell compared to Scr treated controls (Fig 2G). Likewise, the inhibition of miR-328 in neutrophils produced similar results (S3E Fig). Importantly, in neutrophils ROS production was increased by ant-328 treatment even in the absence of NTHi, although to a lower extent, suggesting that the effect on these bacterial killing pathways was partly independent of phagocytosis (S8A and S8B Fig). Interestingly, the increased killing was not associated with pro-inflammatory cell activation as the major pro-inflammatory cytokines produced by infection, IL-6 and TNF-α, were not altered by ant-328 treatment of macrophages (S9A and S9B Fig). Thus, miR-328 plays a very specific role in the microbicidal activity of these innate immune cells. To investigate the signalling pathways involved in the regulation of miR-328 expression, we used specific inhibitors to block the activation of p38, JNK, and ERK MAPK following NTHi infection. Exposure of macrophages to NTHi or vehicle (DMSO) plus NTHi led to a significant 2-fold reduction in miR-328 expression (Fig 2H). However, pre-treatment of macrophages with a p38 inhibitor, doramapimod, or a JNK inhibitor, SP600125, prior to NTHi infection, completely blocked the down-regulation of miR-328 expression (Fig 2H). In contrast, the ERK inhibitor, U0126, had no effect. This suggests that NTHi down regulates miR-328 at least in part through p38 and JNK signalling pathways. We confirmed that NTHi exposure to primary macrophages activates p38 (S10A Fig) and JNK (S10B Fig) but not ERK (S10C Fig) signalling. We then determined whether other miRNAs that were also identified from the array data (S1 Fig) as having increased or decreased expression in the lungs of mice following NTHi infection, could also play a role in bacterial clearance. Antagomir-mediated inhibition of miR-21-3p and miR-223 (increased expression following bacterial infection), or miR-376c (decreased expression), and miR-21 (control, no change in expression) had no effects on NTHi clearance by neutrophils in vitro (S11 Fig). Collectively, these data suggest that NTHi activates the p38 and JNK signalling pathways, which regulates the cellular levels of miR-328. Inhibition of miR-328 with ant-328 suggests that this microRNA regulates phagocytosis and killing of bacteria by macrophages and neutrophils in vitro. To specifically assess the role of miR-328 in macrophage- and neutrophil-mediated bacterial clearance in vivo we conducted adoptive transfer experiments. Macrophages or neutrophils were isolated from naïve mice and then treated with ant-328 or Scr control for 12 h before i. t. transfer into recipient naïve BALB/c mice (Fig 3A and 3C). To monitor the effectiveness of transfer, cells were pre-stained with CFSE. Mice were then inoculated with NTHi. Both CFSE-labelled macrophages (S12B and S12C Fig) and neutrophils (S12E and S12F Fig) entered the lungs and remained there for at least 24 and 2 h, respectively. There were no differences in the numbers of these cells between ant-328- and Scr-treated controls. Total bacterial load in BAL fluid and lung homogenates was measured by plating and colony counting. Importantly, mice that received ant-328 treated macrophages (Fig 3A and 3B) or neutrophils (Fig 3C and 3D) showed significantly improved clearance of NTHi from the lungs compared to Scr control treatment. However, there were no statistical differences in total inflammatory cell infiltrates in BAL fluid (S13A and S13F Fig). Mice that received ant-328 treated macrophages had similar total number of macrophages (S13B Fig) but a significantly reduced number of neutrophils (S13C Fig). It is likely that the reduction in neutrophil numbers is due to increased clearance of bacteria by ant-328 which leads to reduction in the early inflammatory events that promote neutrophil influx. The adoptive transfer of macrophages is likely to promote the removal of apoptotic neutrophils as well. Mice that received ant-328 treated neutrophils had no difference in macrophages (S13G Fig) or neutrophils numbers (S13H Fig). The concentrations of pro-inflammatory cytokines, IL-6 and TNF-α, in BAL fluid were equivalent in all experiments (S13D–S13E and S13I–S13J Fig). These data suggest that the effect of ant-328 on bacterial clearance was mediated directly by phagocytes and not by increased inflammatory cell recruitment or cytokine production. We next assessed whether inhibition of miR-328 improved NTHi clearance in vivo, which would indicate whether the use of ant-328 would be a potential therapeutic option for respiratory bacterial infection. Naïve BALB/c mice were inoculated with NTHi, and 6 h later were treated i. t. with ant-328 or Scr control (Fig 3E). Inhibition of miR-328 function by ant-328 was confirmed during infection (Fig 3F). NTHi clearance from the lungs was significantly enhanced (4-fold) by ant-328 treatment compared to controls (Fig 3G). Again total (Fig 3H) and neutrophils infiltration (S13L Fig) in the BAL and the levels of the pro-inflammatory cytokines, IL-6 and TNF-α, were not significantly altered (S13M–S13N Fig). In contrast, there was a small but significant increase in the number of macrophages recruited to the lungs by ant-328 treatment following NTHi infection (S13K Fig). We then tested whether the small changes in inflammatory cell numbers in the BAL in some experiments was due to the direct effect of ant-328, or indirectly due to a secondary effect induced by alterations in NTHi clearance after ant-328 treatment. In order to assess this we treated naïve mice with Scr or ant-328 and measured inflammatory cells and cytokines in the BAL. We observed that there was no effect of ant-328 on BAL cell recruitment or pro-inflammatory cytokine production in non-infected mice (S14A–S14D Fig). The mainstay treatment of many respiratory diseases is the prolonged use of corticosteroids, however this leads to immunosuppression and increased risk of hospitalisation for pneumonia, especially in COPD patients [23]. This has been shown in rodent and human studies to result from the suppression of phagocytosis of bacteria by macrophages [24–26]. We next determined whether miR-328 was involved in dexamethasone-mediated suppression of immunity and bacterial clearance, and if inhibition of this miRNA could enhance bacterial clearance in these immune suppressed mice. Mice were administered dexamethasone i. p. daily for 3 days and were then inoculated with NTHi (scheme Fig 4). Dexamethasone treatment did not alter miR-328 expression, suggesting that this miRNA was not directly involved in mediating the effects of corticosteroids (Fig 4A). As expected, dexamethasone pre-treated mice inoculated with NTHi had substantially increased bacterial load in their lungs during infection (Fig 4B). Interestingly, pre-treatment also increased the inflammatory cell infiltrate in the lungs. This is likely due to the increased bacterial load, which results in a greater proinflammatory response leading to the recruitment of more inflammatory cells (Fig 4C). Ant-328 treatment inhibited miR-328 in both dexamethasone and vehicle-treated mice infected with NTHi to similar levels (Fig 4D). Inhibition of miR-328 in dexamethasone pre-treated mice reversed the increased bacteria load compared to controls (Fig 4E). Importantly, bacterial levels were reduced to below those in non-immune-suppressed mice (vehicle+Scr control), and to equivalent levels compared to non-immune suppressed mice treated with ant-328 (vehicle+ant-328). These data indicate that the inhibition of miR-328 was as effective at clearing bacteria in both immune competent and immune suppressed environments. Again the number of infiltrating inflammatory cells in the BAL was only increased by dexamethasone treatment and there was no difference between Scr control and the ant-328 treated groups (Fig 4F). NTHi is commonly isolated from the airways of COPD patients and is linked to exacerbations of disease [20]. Thus, we next investigated if inhibition of miR-328 improved bacterial clearance in our model of cigarette smoke-induced emphysema. Mice were exposed for cigarette smoke for 8 weeks and were then inoculated with NTHi followed by antagomir treatment (scheme Fig 5). Exposure of the lung to ant-328 resulted in significant reduction in the levels of miR-328 (Fig 5A). Again, treatment with ant-328 during infection enhanced NTHi clearance in cigarette smoke-exposed mice (Fig 5B). Total numbers of cellular infiltrates were similar between ant-328 treated and Scr control groups (Fig 5C). In the same model, pulmonary function was measured using the forced oscillation technique. Inhibition of miR-328 function also resulted in a decrease in pulmonary compliance (Fig 5D) and an increase in pulmonary elastance (Fig 5E). Inhibition also nearly completely ablated the increase in muc-5ac expression induced by smoke-exposure with NTHi infection (Fig 5F). MiR-328 is conserved across species, with an identical sequence in mice and humans. Therefore, we investigated if miR-328 plays a similar role in human macrophages and neutrophils infected with NTHi in vitro. Neutrophils were purified from healthy adult blood and macrophages were differentiated in culture from monocytes derived from the PBMC fraction. Similar to the results observed with murine cells, inhibition of miR-328 significantly increased bacterial uptake by human macrophages (Fig 6A) and neutrophils (Fig 6B) in vitro. miRNAs are known to be involved in regulating innate immune responses. Expression is induced by various stimuli such as TLR pathways [27,28] or pro-inflammatory cytokines [29,30] and they can act as both positive and negative feedback signals to control these pathways. Furthermore, host pathogen interactions and activation of the immune system has been shown to alter, and be altered by, miRNA expression in various infection models [31–35]. In this study we show that within 24 h of NTHi infection the levels of miRNA in the lung are rapidly altered with expression of individual miRNAs being both increased and decreased. Alterations in the level of expression are strongly associated with bacterial clearance and recruitment of innate immune cells into the lung. Notably from profiling data, miR-328, a highly conserved miRNA, was significantly down-regulated by infection and its basal level are amongst the highest of all down-regulated miRNA, suggesting a key role in the host response to NTHi infection. Similarly, infection of primary lung macrophages by NTHi also downregulated miR-328 expression in vitro. miR-328 is involved in cancer [36–38], autoimmunity [39], and neuronal disease [40]. Here we demonstrate a new role in regulating innate immune cell function, wherein suppression of miR-328 improves bacterial clearance in the lungs. Inhibition of this miRNA in either macrophages or neutrophils from both mice and humans in vitro increases uptake (of live and heat-killed NTHi) and decreases survival of bacteria. This indicates that miR-328 regulates phagocytosis and killing of bacteria. Bacterial clearance is also dependent on ROS production [41]. In chronic granulomatous disease, phagocytic activity is normal, however, there is an inability of phagocytes to form ROS after phagocytosis, which leads to increased bacterial survival [42]. Here we show that increased production of ROS in ant-328 treated phagocytes, indicating that this bacterial killing mechanism is also regulated by miR-328. Oxygen-independent bacterial killing mechanisms (such as Cathepsin D) also operate following phagocytosis although they usually coincide with oxygen-dependent lysosomal killing. Cathepsin D is a lysosomal cationic protease that disrupts bacterial membranes, and increases susceptibility of gram-negative bacteria to lysis by lysozyme [43]. We observed increased cathepsin D expression in ant-328 treated macrophages suggesting that bacterial killing was also enhanced using an oxygen independent pathway. Thus miR-328 plays a critical role in the microbial host defence mechanism of innate immune cells by augmenting phagocytosis, the production of ROS and microbicidal activity. Although inhibition with an antagomir dramatically promoted NTHi clearance, application of a mimetic (increasing miR-328 levels) failed to delay clearance of NTHi. We speculate that this is because the levels of endogenous miR-328 are very high, hence masking the effect of mimetic treatment. High levels of miR-328 are linked to macrophage homeostasis and thus it is only when levels are dramatically decreased by antagomir treatment that the cell becomes activated to clear bacteria. In cancer, miR-328 expression is regulated through the ERK1/2 pathway [38]. Here we found that p38 and JNK MAPK is activated by NTHi infection, and inhibition of these MAPK signalling prior to bacterial infection prevents the down-regulation of miR-328 levels. This suggests that NTHi suppresses miR-328 expression via activation of the p38 and JNK MAPK pathway. We did not observe a functional role for ERK. These experiments were performed in vitro with macrophages. Thus, it is possible that miR-328 expression could be regulated by other pathways in a more complex in vivo environment. Adoptive transfer of miR-328-depleted macrophages or neutrophils increased bacterial clearance in the lung, further supporting our in vitro observations and directly demonstrating that inhibition of miR-328 in these cells amplifies their ability to clear respiratory infections. We next explored the potential of ant-328 as a therapeutic treatment for bacterial lung infection by directly administering the inhibitor to the lung after NTHi inoculation. Inhibition of miR-328 with ant-328 substantially increased bacterial clearance suggesting that targeting this miRNA could be potentially used as a new approach to anti-microbial therapy. Although we detected increased ROS production in vitro following ant-328 treatment, which is often linked to lung injury and activation of pro-inflammatory pathways, we did not observe any increase in inflammation as measured by cellular infiltration and the production of pro-inflammatory cytokines in the lung. One potential explanation is that the increased ROS production that occurs with more rapid bacterial clearance following ant-328 treatment does not reach a threshold level or period of production that is required to elicit deleterious inflammatory changes in the lung. Thus, importantly, treatment with ant-328 in vivo enhanced bacterial clearance without a consequent negative impact on lung inflammation (cellular recruitment or proinflammatory mediator production) or pathology. In chronic lung disease, corticosteroid treatment can lead to immunosuppression and increase the risk of pneumonia [23]. In our current study pre-treatment with dexamethasone prior to NTHi infection significantly impaired bacterial clearance from the lungs. Previous studies have shown that dexamethasone inhibits macrophage phagocytosis in vitro while treatment in vivo suppresses clearance of bacteria [24,25]. Importantly, dexamethasone administration did not alter miR-328 expression levels despite the increased bacterial load in the lungs. Notably, ant-328 treatment of dexamethasone immune-suppressed mice brought about a dramatic reduction in bacterial load, to levels below that seen in immune competent mice. This data demonstrates that although dexamethasone treatment inhibits phagocytosis along with other inflammatory anti-bacterial pathways, this effect is overcome by treatment with ant-328. These results suggest that it may be possible to use ant-328 in conjunction with dexamethasone in the treatment of chronic lung diseases in order to better control bacterial infection. This would be of particular interest in the case of antibiotic resistant bacterial strains where conventional therapies tend to fail and in patients receiving steroid therapy where infections are difficult to control. Bacterial colonisation and infection are commonly associated with COPD and exacerbation of disease, and NTHi is one of the most frequently isolated strains of bacteria [20]. The reason underlying this may be that alveolar macrophages from COPD patients exposed to cigarette smoke extract are less efficient at phagocytosing NTHi [44,45]. Using a cigarette smoked-induced model of experimental COPD [46], we demonstrated that inhibiting miR-328 brought about a 3-fold increase in bacterial clearance. A definitive characteristic of COPD is the loss of elastic recoil in the lung and increased lower airway remodelling including mucous cell hyperplasia [47]. Following infection with NTHi, treatment with ant-328 significantly improved both elastic recoil and suppressed muc-5ac expression in the lungs of chronically cigarette exposed mice. The exact mechanism of how this occurs remains to be elucidated but it may involve altered surface tension in lungs by impairing mucus production [48]. In a clinical study on acute exacerbation of COPD, half of the sputum samples from patients enrolled tested positive for bacterial growth. Approximately 50% of the strains isolated were H. influenzae and Moraxella catarrhalis, of which the vast majority were resistant to penicillin [49]. MiR-328 has a highly conserved sequence between mice and humans. Inhibition of miR-328 function in human macrophages and neutrophils using ant-328 also resulted in increased bacterial phagocytosis, indicating that our studies are potentially translatable into anti-microbial innate host defence pathways in human cells. Our study identifies a potential alternative approach to the treatment of pathogenic microbial infections and bacterial induced exacerbation of chronic lung disease. Although the specific targets of miR-328 remain to be elucidated it is likely that this miRNA targets transcripts involved in regulating phagocytosis and controlling the microbicidal activity of the macrophages and neutrophils. Importantly, ant-328 enhances bacterial killing by specifically augmenting bacterial clearance pathways without further promoting a proinflammatory environment that may be deleterious to lung tissue. Targeting miRNA would be of particular interest to enhance therapeutic outcomes in patients suffering from disorders such as COPD, cystic fibrosis and asthma where innate host defences may be compromised due to steroid therapy or underlying disease mechanisms. Innate immune defects also occur in HIV and transplantation patients. In HIV patients, phagocytosis and respiratory burst are reduced in monocytes and neutrophils [50]. After solid organ transplantation ~50% of deaths are due to infection and during the first four months following bone marrow transplantation, phagocytosis by alveolar macrophage phagocytosis is impaired [51,52]. Although speculative, targeting miR-328 may be more broadly applicable and lead to improvement of innate immune cell function in these immune-deficient patients. While the employment of direct inhibitors of miRNA pathways needs to be approached with caution, we would envisage acute application of antagomir to treat infection or exacerbations induced by acute infections. It will be important, however, to study the chronic impact of altering miR-328 function in murine models, and subsequently in human disease. To date we have not observed toxic effects of antagomirs in chronic models of asthma [53]. In a similar approach, treatment of chronically infected chimpanzees with a locked nucleic acid–modified oligonucleotide (SPC3649) complementary to miR-122 induced long-lasting suppression of HCV viremia, with no evidence of viral resistance or side effects in the treated animals [54]. However, continued research into the amazing regulatory powers of these small non-coding RNA is required to fully determine their potential for the treatment of disease and their fundamental role in maintaining health. This study is the first to identify a role for miR-328 in regulating bacterial infection in the lung and provides proof-of-principle data that targeting specific host miRNAs could lead to future therapeutics to combat multi-drug resistant bacteria and infection in chronic lung diseases and immune-compromised patients. The animal protocols used were conducted in accordance with the NSW, Australia Animal Research legislation. The experimental protocols 987, A-2009-152, A-2009-141 have been reviewed and approved by University of Newcastle Animal Ethics Committee. All adult subjects provided their informed written consent to participate in the study. All human studies were conducted in accordance with approval from the University of Newcastles Human Research Ethics Committee. Specific pathogen-free adult male BALB/c mice were used throughout, and were obtained from the University of Newcastle central animal house. NTHi biotype II (NTHi-289) was obtained, prepared and used as previously described [55,56]. Bacteria were prepared prior to infection by growing overnight on chocolate agar plates (Oxoid, Australia) at 37°C in an atmosphere of 5% CO2. The following day, the bacteria were harvested, washed and diluted in sterile PBS and CFU calculated by spectrophotometric analysis. Live NTHi were used in NTHi infection both in vivo and in vitro. Heat killed bacteria were generated by incubating the live NTHi at 70°C for 30min. Mice were anaesthetised (12. 5 mg/kg Alfaxan, Jurox, NSW, Australia) intravenously via the tail vein and innoculated intratracheally (i. t) using a cathether (Terumo, Sureflo Hospital Supplies of Australia) with 5x105 or 5x106 CFU of live NTHi in 30 μl of PBS. Following NTHi inoculation, mice were sacrificed at various time points and BAL fluid and lung homogenate (homogenised in 1 ml sterile PBS (Gibco, Invitrogen, New Zealand) ) were used to determine bacterial growth. Serial dilutions were loaded onto chocolate agar plates (Oxoid, Australia) and incubated overnight at 37°C in an atmosphere of 5% CO2. After 16 h, bacterial colonies for NTHi were counted. BAL cells were enumerated as described previously [57]. Total RNA was isolated using TRI Reagent (Ambion). miRNA expression was validated using TaqMan miRNA assays (Applied Biosystem) according to the manufacturer’s protocol and normalised to the housekeeping small RNA gene sno-202. Macrophages in cell culture were detached in citric saline (0. 135 M KCL, 0. 015 M Na citrate) while neutrophils were removed by rinsing. Cells were spun down and resuspended in FACS buffer (2% fetal calf serum (FCS; Sigma-Aldrich) in PBS) before incubation with 10μM dihydroethidium (DHE; Sigma-Aldrich) for 30 minutes at 37°C. Fluorescence intensity was measured by flow cytometry with 488nm excitation and 610nm emission. Immunofluorescence assays were performed as described previously [56]. Macrophages were labelled with 5μM of CFSE (Molecular Probes) for 10 minutes at 37°C. Staining was quenched with fetal calf serum (FCS) for 1 minute, and cells were washed and resuspended in PBS. Cells were fixed in 1% (w/v) paraformaldehyde in PBS buffer for 10 min at room temperature (RT), blocked and permeabilized with 0. 2% (v/v) Triton X-100 and 5% normal goat serum for 1 h at RT. Cathepsin D (CTSD) expression were stained using rabbit monoclonal Ab to CTSD (Abcam) overnight at 4°C. Cells were then incubated with Cy3-conjugated goat anti-rabbit IgG (GE Healthcare) for 45 minutes at RT and nucleus was counter-stained with DAPI (Sigma-Aldrich). Images were analyzed using Olympus FluoView FV1000 confocal microscope. Macrophages were pre-treated 5μM doramapimod (inhibit p38 MAPK) (LC lab), 50μM SP600125 (inhibit JNK) (LC lab), 50μM U0126 (inhibit ERK) (LC lab) for 30 minutes, prior to NTHi infection for 8 h. DMSO was used as a vehicle control. Mice were exposed to cigarette smoke (in the form of 12 cigarettes) in a closed chamber twice a day, 5 days per week for a duration of 8 weeks as previously described [46]. Mice were injected intraperitoneally (i. p.) with 3mg/kg of dexamethasone (DEX; Sigma-Aldrich) for 3 consecutive days before they were inoculated with (5 x 105 CFU / mouse) with NTHi. The complementary antagomir strand of the miR-328-3p sequence obtained from miRbase was synthesised from Sigma-Aldrich with the following chemical modifications 5’mA. *. mC. *. mG. mG. mA. mA. mG. mG. mG. mC. mA. mG. mA. mG. mA. mG. mG. mG. mC. *. mC. *. mA. *. mG. *. 3’-Chol where “m” represents 2’-OMe-modified phosphoramidites, “*” represents phosphorothioate linkages, and “-Chol” was hydroxyprolinol-linked cholesterol to allow permeation of cell membranes. The scrambled control antagomir sequence was 5' mU. *. mC. *. mA. mC. mA. mA. mC. mC. mU. mC. mC. mU. mA. mG. mA. mA. mA. mG. mA. *. mG. *. mU. *. mA. *. 3' -Chol. Mice were treated in vivo with 50μg of antagomir intratracheally (see NTHi infection: methods), and macrophages and neutrophils were treated with antagomir at 50μg/ml 12 h before assays in vitro. Results are presented as means ± standard errors of the means (SEM). Results were analysed using one way ANOVA plus Bonferroni post-test for statistical significance or Student' s t-test plus Mann-Whitney test. Analysis was performed with Prism v4. 0 (GraphPad Software, CA. USA). p< 0. 05 was considered to represent a significant difference. Additional methods were described in S1 Methods.

MicroRNAs regulate pathogen recognition pathways by modulating translation. In the immune system, miRNAs have been identified as important regulators of gene expression programs, which regulate differentiation, growth and function of innate and adaptive immune cells. Using miRNA microarray, we demonstrated that lung miRNAs were differentially expressed following non-typeable Haemophilus Influenzae (NTHi) infection in mice. To study the role of a specific miRNA in macrophages, we used antagomir (chemically modified single-stranded RNA analogues, complementary to the target miRNA) to block miRNA function. Interestingly, inhibition of microRNA-328 in mouse and human macrophages increases microbicidal activity by amplifying phagocytosis and production of reactive oxygen species. Inhibition of mR-328 in the lung enhanced bacterial clearance in mouse models of immunosuppression and emphysema. Our study provides proof of principle that miRNA pathways can be targeted in the lung and offer a potential new anti-microbial approach for the treatment of respiratory infection.

Abstract Introduction Results Discussion Materials and Methods

2015

Antagonism of miR-328 Increases the Antimicrobial Function of Macrophages and Neutrophils and Rapid Clearance of Non-typeable Haemophilus Influenzae (NTHi) from Infected Lung

Seminal fluid proteins transferred from males to females during copulation are required for full fertility and can exert dramatic effects on female physiology and behavior. In Drosophila melanogaster, the seminal protein sex peptide (SP) affects mated females by increasing egg production and decreasing receptivity to courtship. These behavioral changes persist for several days because SP binds to sperm that are stored in the female. SP is then gradually released, allowing it to interact with its female-expressed receptor. The binding of SP to sperm requires five additional seminal proteins, which act together in a network. Hundreds of uncharacterized male and female proteins have been identified in this species, but individually screening each protein for network function would present a logistical challenge. To prioritize the screening of these proteins for involvement in the SP network, we used a comparative genomic method to identify candidate proteins whose evolutionary rates across the Drosophila phylogeny co-vary with those of the SP network proteins. Subsequent functional testing of 18 co-varying candidates by RNA interference identified three male seminal proteins and three female reproductive tract proteins that are each required for the long-term persistence of SP responses in females. Molecular genetic analysis showed the three new male proteins are required for the transfer of other network proteins to females and for SP to become bound to sperm that are stored in mated females. The three female proteins, in contrast, act downstream of SP binding and sperm storage. These findings expand the number of seminal proteins required for SP' s actions in the female and show that multiple female proteins are necessary for the SP response. Furthermore, our functional analyses demonstrate that evolutionary rate covariation is a valuable predictive tool for identifying candidate members of interacting protein networks. Sexual reproduction is a fundamental biological process by which many eukaryotic organisms transmit their genetic material to the next generation. While the end result of a successful mating is the fusion of the gametes, other molecular interactions must occur to allow this fusion. In internally fertilizing animals, males transfer to females not only sperm, but also a suite of seminal fluid proteins (Sfps) that are essential for reproductive success. Across diverse taxa, Sfps are required for: the mobilization of sperm and their storage within the female; increasing the reproductive capacity of the female; affecting the outcome of sperm competition between multiple males; and, facilitating the union of the gametes [reviewed in 1]. In insects, Sfps also alter female behaviors and physiology [2]. Effects of Sfps can be caused by interactions between specific Sfps, between Sfps and proteins on the sperm, and between Sfps and proteins native to the female reproductive tract. Thus, characterizing the functions and interactions of Sfps is important for understanding how the sexes together ensure the successful production of progeny. Post-mating changes in physiology and behavior induced by Sfps have been extensively characterized in Drosophila melanogaster [2], [3]. In response to the receipt of Sfps, females produce, ovulate and lay eggs [4]–[6]; store sperm in specialized storage organs [7]–[10]; show altered immune responses [11], [12]; undergo changes in sleeping, feeding and excretion behavior [13]–[16]; and, become refractory to male courtship [17], [18]. Several of these behavioral changes – egg production, sperm storage and release, and refractoriness to remating – persist in females for several days after mating and have thus been termed the long-term response [19]–[21]. The proximate cause of these changes is a short (36 amino acid) seminal protein called sex peptide (SP) [17], [18]. While most Sfps are no longer detectable in females several hours after mating [22], SP persists in females for days by binding to stored sperm [19]. Gradually, the C-terminal portion of the peptide is proteolytically cleaved to release it from sperm into the female reproductive tract [19]. This C-terminal portion of SP can then signal through its receptor, sex peptide receptor (SPR), which prolongs at least some behavioral changes in the female [23]–[26]. Indeed, SP cleavage is required for the protein to affect female behavior for more than one day [19] and for sperm to be released efficiently from storage [27]. We have previously used RNA interference (RNAi) or gene knockout lines to test 32 Sfps for function in the SP-mediated long-term response [4], [7], [10], [20], [28], [29]. These studies identified five proteins that are required for SP to function over the long term in mated females: two C-type lectins, CG1652 and CG1656; a serine protease homolog, CG9997; a cysteine-rich secretory protein, CG17575; and, a serine protease, seminase (CG10586). These proteins act in a network in which each member is required for SP to become bound to sperm [21], [28]. Loss of any network protein causes an early resumption of female receptivity to remating and a decrease in long-term fecundity. Such loss also impairs the release of sperm from the seminal receptacle in the days following mating [27]. Specific members of the network act interdependently on one another. For example, males that do not produce CG9997 are unable to transfer CG1652 and CG1656 to the female, while CG1652 and CG1656 are required to slow the rate at which CG9997 is processed in the female. Thus, while SP-SPR signaling is the proximate cause of the female post-mating response, several additional Sfps are required for this signaling to persist over the long term. We refer to this set of seven proteins as the SP network. While genomic and proteomic analyses in D. melanogaster have identified hundreds of proteins from sperm [30], [31], seminal fluid [32]–[35], and the female sperm storage organs [36]–[40], we know of few examples of how these proteins interact to cause the dramatic post-mating phenotypes observed in females [21], [26], [28]. Biochemical approaches to identify interacting proteins are challenging due to the small amount of protein per fly, and exhaustive genetic screening of each known reproductive protein would be laborious. Here, we demonstrate a successful effort to prioritize male and female proteins for functional testing by examining covariation in their rates of evolution among species. Evolutionary Rate Covariation (ERC) is a new metric that bioinformatically infers functional relationships between proteins based solely on their evolutionary rates across an array of species [41]. ERC operates from the hypothesis that functionally related proteins will experience correlated rate changes, because forces governing protein evolutionary rate are expected to influence entire pathways simultaneously. Evolutionary rate depends on several factors including a protein' s expression level, its essentiality, and its interactions with other proteins [42]–[49]. Pathway-wide fluctuation in each of these factors has been associated with correlated rate changes (i. e. , ERC) between functionally related proteins [41], [50]–[53]. In practice, an ERC value is calculated by computing the correlation between the rates of change of two proteins across all branches of a phylogeny. ERC values range from 1 to −1 for a perfect positive or negative correlation, respectively, with the genome-wide ERC distribution between all protein pairs centered at zero [41]. Functionally related pairs of proteins have been observed to have more positive ERC values in taxa as diverse as eubacteria, fungi, invertebrates and mammals [41], [50], [51], [54]–[58]. This finding holds for proteins that share physical or genetic interactions and proteins that are found in common complexes or metabolic pathways [41], [59]. Generally, a high ERC value is best interpreted as a potential functional link, which could have resulted from a common evolutionary force acting on both proteins. Accordingly, we can infer that proteins with correlated rates may be functionally related. ERC and related methods have primarily been used to study proteins that are already known to interact functionally or physically; the use of such methods for functional prediction is only now starting to emerge [60]. We tested the utility of applying ERC prospectively by examining proteins required for Drosophila SP function. Because proper function of the SP network is essential for fertility, we reasoned that members of this network could have experienced shared evolutionary selective pressures over time and might thus show patterns of ERC across the phylogeny of sequenced Drosophila species [61]. To test this hypothesis, we created an ERC dataset specific to Drosophila. This analysis revealed significant levels of ERC between known members of the SP network. We then screened for new members of this network by searching for elevated ERC between known network proteins and sets of uncharacterized Sfps and female reproductive proteins. RNAi tests of 18 top candidates revealed three female and three male proteins required for network function. Through molecular genetic analysis, we placed five of these proteins into specific positions in the SP network, and we observed that the steps in the network in which these new proteins act are largely consistent with their evolutionary correlations. Our results demonstrate that signatures of ERC can be used prospectively to predict members of a protein network, suggesting that this method may be broadly applicable for identifying novel protein interactions. We first calculated Evolutionary Rate Covariation (ERC) values for all pairs of orthologous proteins (reproductive and otherwise) from 12 Drosophila species. Briefly, we assembled orthologous protein sequences for each gene from each species for which they were available, resulting in 11,100 multiple alignments. For each pair of alignments, we calculated the correlation coefficient between their branch-specific evolutionary rates (see Methods and Figure S1). The resulting ERC values ranged from −1 to 1 and reflect the degree to which evolutionary rates correlate for any particular pair of proteins. Typically, ERC values between functionally related protein pairs are elevated compared to unrelated pairs [55]. We observed this same pattern for the seven previously known members of the Drosophila SP network. ERC values calculated for all possible pairs of these seven proteins had a mean of 0. 3115, compared to the proteome-wide mean of 0. 0019 (Figure S1A). The highly significant elevation between SP network proteins (permutation p = 0. 000154) suggests that ERC could be used to predict additional SP network proteins. However, since proteins that are expressed at similar levels or in similar patterns can also show correlated evolution [43], we also tested whether reproductive proteins as a class had elevated ERC values. To do so, we examined a set of 664 proteins found in seminal fluid, sperm, or female sperm storage organs (see Methods and Figure S1A; we refer to these proteins below as “reproductive” but note that some are also expressed in non-reproductive tissues and could thus have other functions). The mean ERC value between all reproductive proteins was 0. 0326, a highly significant elevation for sets of this size (permutation p<0. 0001). This elevation could be driven by direct functional relationships and/or more indirect relationships such as expression patterns [41]. To control for this elevation in ERC across all reproductive proteins when evaluating correlations between individual pairs of proteins, we factored out the broad relationship between them. To do so, we recalculated ERC using only the 664 reproductive proteins to estimate the background rate of evolution, instead of all 11,100 proteins (see Methods and Figure S1B). After this adjustment, the mean pairwise ERC between all proteins in the reproductive set fell to 0. 0047. By contrast, the mean correlation between the seven known SP network proteins remained significantly elevated (mean = 0. 2806; permutation p = 0. 001002). These results suggest that while shared patterns of expression or function can cause a significant increase in ERC, a much stronger signal is shared by the specific set of proteins that act together in the SP network. Several of the strongest pairwise correlations between known members of the SP network were found between proteins with recognized genetic interactions. For example, males that do not produce network protein CG9997 are unable to transfer CG1652 and CG1656 to females during mating [21]. These pairs of proteins show ERC values in the top 5 percent of all pairwise correlations (CG9997-CG1652: r = 0. 62, empirical p = 0. 03; CG9997-CG1656: r = 0. 62, empirical p = 0. 03; Figure 1). In other instances, we did not observe strong correlations between proteins that might be expected to coevolve, such as SP and SPR. However, this particular lack of correlation may be explained by the fact that SPR has additional, non-reproductive ligands besides SP [62], [63], which may constrain its evolution. Nonetheless, the overall signature of correlated evolution throughout the SP network, the high proportion of positive pairwise correlations in the group (i. e. , 16 of the 21 pairwise correlations in Table 1 are positive), and the significant correlations between specific group members suggest that members of the SP network show significant levels of evolutionary rate covariation. Since we detected positive evolutionary correlations between known SP network proteins, we applied the ERC method prospectively to identify new candidate network members. For this analysis, we calculated pairwise correlations using the reproductive protein data set described above, and we focused specifically on correlations between the known SP network proteins and the 434 proteins that comprised the sets of secreted Sfps and proteins present in the female reproductive tract. To identify candidates, we queried each of five network proteins (CG1652, CG1656, CG9997, CG17575 and SP) against the 434 Sfp and female proteins. SPR was not used as a query because it has additional ligands that do not appear to function in reproduction [62], [63]. Thus, SPR may need to maintain interactions with multiple proteins, which may dampen signals of correlated evolution with any single interacting partner. Seminase was excluded because unambiguous orthologs were found in only five species, which would cause low statistical power. We found 111 proteins (55 Sfps, 56 female proteins) that showed a significant correlation (p<0. 05) with at least one of the five network proteins. From this group, we selected 21 candidates for further testing, each of which showed a significant (p<0. 05) level of ERC with multiple SP network proteins and/or a highly significant (p<0. 01) level of ERC with at least one network protein (Table 1). We tested each candidate in Table 1 by using RNAi to knockdown expression of the gene in the appropriate sex; five of the 21 candidates showed no evidence of knockdown by RT-PCR and were excluded from further analysis. For the remaining 16 candidates, we screened for genes whose knockdown caused a significant increase in female remating receptivity four days after an initial mating. Of the 16 candidates that were at least partially knocked down by RNAi, five showed highly significant effects on 4-day remating receptivity (Table 1). Knockdown of the remaining 11 candidates caused no significant increase in female receptivity. This latter result could be due in some cases to insufficient knockdown or to functional redundancy with other Sfps or female proteins. Alternatively, these proteins may not function in the SP network. Of the positive candidates, three genes (CG14061, CG30488 and CG12558) are expressed specifically in the male accessory glands [64]; at least two of them (CG14061 and CG30488) encode proteins that are transferred to females as Sfps at mating [33]. The other two positive candidates, CG3239 and CG5630, are each expressed in the female' s spermathecae, as well as in other non-reproductive locations [64]. CG5630 is also expressed in the female' s seminal receptacle [39]. One striking feature of several of the new candidate network genes was their genomic positioning next to previously known SP network genes (Table S2). This pattern was previously observed for the SP network lectins, CG1652 and CG1656, which are believed to have arisen from an ancient gene duplication event [33], [34]. We found that three additional pairs of network genes (CG9997 and CG14061, CG17575 and CG30488, and CG3239 and SPR) are also located in tandem with one another. For two of these pairs, the tandemly-located genes encode proteins in the same biochemical category (CG9997 and CG14061 each encode predicted serine protease homologs, and CG17575 and CG30488 each encode predicted CRISPs), but in contrast to the situation with the lectins CG1652 and CG1656, we do not find unambiguous evidence that either the protease or the CRISP cluster arose by tandem gene duplication. However, regardless of each cluster' s origin, it is possible that such genomic clustering enables the co-regulation of genes that function in a common pathway [65]. In the CG17575/CG30488 cluster, we found a third annotated gene that encodes a seminal fluid protein of the same predicted functional class as the other cluster members: CG30486, which encodes a predicted CRISP. Similarly, we observed a known Sfp gene encoding a predicted serine protease homolog, CG34295, immediately upstream of CG12558. While neither CG30486 nor CG34295 was identified by our ERC analysis, we hypothesized that their shared locations with known or candidate SP network members could indicate their involvement in the SP network. However, when each of these additional genes was knocked down individually, we observed no effect on female remating receptivity 4 days after mating (Table 2). Thus, either these neighboring genes are uninvolved in the SP network, or they function in the network in a completely redundant role. Alternatively, their degrees of knockdown may have been insufficient to produce a phenotype. We also asked whether signatures of ERC between these new candidates and the rest of the large sets of seminal fluid or female proteins might identify additional network proteins (Figure S1C). To this end, we used RNAi to test two additional female genes that showed highly significant ERC levels with at least one new candidate protein (Table 2). One of these genes, epidermal stripes and patches (Esp), showed a highly significant effect on female remating receptivity. Taken together with the results above, these data suggest that ERC has strong sensitivity to detect new candidate members of the SP network. To confirm that the receptivity and fertility effects we observed in the above RNAi experiments were not due to RNAi off-target effects and/or insertions of RNAi-triggering constructs into essential genes, we first used UP-TORR [66] to analyze each line' s RNAi-triggering sequence against all current D. melanogaster gene annotations. No off-target transcripts were predicted for any RNAi construct used. We then performed receptivity and long-term fertility assays (see Methods and below) on additional RNAi lines, where available, that controlled for either the site of the UAS-RNAi construct insertion (for CG5630 and Esp) or both the insertion site and the hairpin sequence used to trigger RNAi (for CG30488 and CG3239). (No additional RNAi lines exist for CG14061 or CG12558). These tests (summarized in Table S3) confirmed the receptivity and fertility phenotypes seen with the first lines tested for CG30488 and Esp. Likewise, knockdown of CG5630 by a second hairpin showed a strong effect on fertility and a marginally significant effect on receptivity. Knockdown of CG3239 by a second hairpin also replicated a strong effect on fertility, but showed no significant effect on receptivity. However, RT-PCR revealed that with this hairpin, CG3239 transcript levels were only partially knocked down, which could explain the less severe phenotype. Because of the high degree of replication, results reported below come from experiments performed on the original lines (details of which are described in Table S1). To evaluate whether each of these six genes was required only for extended female non-receptivity, we next tested each positive candidate for effects on remating receptivity at 1 day after an initial mating. As shown in Table 3, in no case did knockdown of a candidate gene cause an increase in short-term receptivity. Thus, rather than having general effects on female post-mating behavior, each candidate is required specifically for the long-term loss of female receptivity to remating. This phenotype is consistent with a malfunction in the SP network [20], [21]. In females mated to SP network knockdown males, SP transferred at mating but not bound to sperm is sufficient for full fertility and non-receptivity 1 day after mating. However, if SP cannot bind to sperm, it is no longer detected in the reproductive tract by 4 days after mating [19]. We reasoned that if these six positive candidates affect the function of the SP network, they should also affect long-term fertility, which requires the long-term storage and utilization of SP [17], [18], [20], [26], [28]. Consistent with a role in the SP network, each new protein was required for full fertility over the course of a 10-day assay (Figure 2). Males knocked down for CG14061, CG30488 or CG12558 induced normal levels of egg-laying and progeny production in females for the first day after mating, but these measures declined relative to controls as early as the second day after mating. Females knocked down for CG5630 or Esp showed the same pattern of normal fertility on day 1 after mating, but reduced fertility in the following days. Females knocked down for CG3239 had significantly reduced egg-laying and progeny production even on the first day after mating, mimicking the effects of knocking down SPR (Figure 2, Figure S2). These knockdown females then continued to have lower egg and progeny production throughout the assay. We further observed that knockdown of any male gene or of the female gene Esp had no significant effect on egg-hatchability, while knockdown of the remaining female genes caused hatchability to be significantly lower (Figures S3, S4). This effect was most pronounced in CG3239 knockdown females, and much less severe in CG5630 and SPR knockdown females. Effects on hatchability were unlikely to be due primarily to reduced viability of offspring inheriting both the UAS-RNAi construct and the GAL4 driver (see Text S1). Thus, each of these six candidates identified by ERC is required for both the long-term loss of remating receptivity and the long-term maintenance of fertility. In our subsequent results and discussion, we adopt new names for the previously unnamed genes: male-expressed genes are named after lunar modules used in the Apollo space program (CG14061: aquarius; CG30488: antares; CG12558: intrepid), and female-expressed genes are named after sites on the moon at which Apollo missions landed (CG3239: fra mauro; CG5630: hadley). The new male genes encode proteins predicted to belong to functional classes often found in insect and mammalian seminal fluid [33], [34], [67]–[69] and already represented in the SP network. Like CG9997, aquarius and intrepid encode serine protease homologs [70]; like CG17575, antares encodes a cysteine-rich secretory protein. In females, fra mauro encodes a protein that contains a partial, predicted neprilysin protease domain. Neprilysins are a class of protease that preferentially cleave prohormones and neuropeptides and are important for male and female fertility in mammals [71]–[73] and Drosophila (J. Sitnik et al. submitted). Neither annotated isoform of fra mauro is predicted by SignalP [74] to be secreted or extracellular, raising the question of how this protein could interact with SP network proteins. Inspection of the 5′ untranslated region of fra mauro revealed the presence of a potential alternative initiation codon, which is followed by a region predicted by SignalP to encode a functional secretion signal sequence. RT-PCR analysis on female cDNA found that a product could be amplified when a forward primer is placed in this region (data not shown), raising the possibility that an alternative isoform of the protein may be secreted and thus more accessible to other network proteins. In addition, we found this alternative start codon and secretion signal to be conserved in at least 11 of 12 Drosophila species analyzed (the D. willistoni genome sequence contains a sequencing gap in this region), which provides strong evidence that this secreted protein isoform is functionally important (Figure S5). The hadley protein is predicted to be secreted, but its potential functional class remains unknown, as neither conserved domain searching [75] nor three-dimensional structural modeling [76] could identify a conserved protein domain. The Esp gene was initially identified as a target of homeotic genes [77] but is, otherwise, poorly characterized. While the Esp protein is not predicted to be secreted, it shows homology to transmembrane sulfate transporters. In adults, Esp is expressed predominantly in the spermathecae [64], with additional expression reported in the seminal receptacle [39]. We next sought to position these six new proteins in the SP network. To do so, we first used Western blotting to test whether SP was successfully stored over the long-term in mates of knockdown males or in knockdown females. In wild-type matings, SP is readily detectable from dissected female seminal receptacles (SRs) 4 days after a mating. However, knockdown of any of the known SP network proteins eliminates this retention [21], [28]. We observed that wild-type females mated to males knocked down for aquarius, antares or intrepid showed little or no SP at 4 days after mating (Figure 3). These reduced levels of SP were not due to less SP having been transferred at mating (see Figure 4). These results suggested that male proteins aquarius, antares and intrepid are each required for network function at a step upstream of SP binding sperm in the SR. By contrast, when wild-type males were mated to fra mauro, hadley or Esp knockdown females, normal levels of SP were observed at 4 days after mating (Figure 3). Thus, these female proteins may be necessary for the utilization of SP after it becomes stored in the SR or may be required for proper SP-SPR signaling. To further determine where the new male proteins fit into the network, we examined the production of the known SP network proteins in males knocked down for aquarius, antares or intrepid (Figure 4). In all cases, we observed no difference in the production of SP, CG1652, CG1656, CG9997 and CG17575 between knockdown and control males (Figure 4; compare lanes for knockdown and control males). We then tested whether knockdown males could transfer these proteins to females and examined their processing in female reproductive tracts. Males knocked down for intrepid transferred all proteins at equivalent levels to controls, and females mated to these males showed normal CG9997 processing [21] in their reproductive tracts. Males knocked down for aquarius or antares transferred normal levels of SP, CG9997 and CG17575, but much lower levels of CG1652 and CG1656 (Figure 4; compare lanes for females mated to aquarius or antares knockdown or control males). Consistent with the absence of these proteins in females after mating [21], the post-mating processing of CG9997 was also disrupted, with mates of knockdown males showing an increased level of the 36-kDa form of CG9997 relative to the 45-kDa form of this protein. We also examined the production and transfer of seminase and observed no differences between knockdown and control flies for each gene (data not shown). Because SP is required for the release of sperm from storage [27], we examined sperm storage and retention in the SRs of females mated to males knocked down for each of these genes (Figure 5). At 2 hours after mating, sperm from antares and intrepid males were present in the SR at equivalent levels to controls, while sperm from aquarius males were present at slightly lower levels. However, by 10 days after mating, mates of control males had largely depleted their stores of sperm in the SR, while mates of males knocked down for any of the three genes showed significantly higher numbers of sperm. Taken together with the lack of SP retention (see Figure 3), these data confirm that male proteins aquarius, antares and intrepid are each required for SP to become bound to sperm. Disruption of this binding, in turn, inhibits the ability of sperm to be released from the seminal receptacle. This inability to release sperm from storage likely contributes to the reduction in long-term fertility when each of these male genes is knocked down (Figure 2). Taken together, our results allow us to place aquarius, antares, fra mauro, hadley and Esp into the SP network (Figure 6A). The male proteins aquarius and antares act at the same step of the network as CG9997, as each of these proteins is required for the transfer of CG1652 and CG1656. The female proteins fra mauro, hadley and Esp appear to act at the downstream end of the network, after SP has bound to sperm. At present, we are unable to position intrepid within the network, though its effect on SP retention (Figure 3) suggests that it acts upstream of SP-SPR signaling. When comparing the positioning of these six new proteins in the network to their patterns of ERC with the previous known seven network proteins (Figure 6B), we observed that the new male proteins showed their strongest correlations with the upstream players of the network. In particular, each new male protein showed a significant correlation with CG9997, which functions in the same step of the network (CG1652/CG1656 transfer) as aquarius and antares. At the downstream end of the pathway, two of the new female proteins showed their strongest correlations with downstream players in the network, including SPR, which is consistent with their potential functions. Thus, the patterns of ERC observed between new and established network proteins are consistent with the steps in the network in which these new proteins are found to act. Our results suggest that ERC successfully prioritized a large set of proteins for detailed functional testing; the observed success rate was six positive hits out of 18 candidates tested, and this rate could be higher if genetic redundancies or insufficient knockdown prevented positive results for some candidates. This rate likely represents a significant enrichment of network genes because if the same success rate were applied to the full list of 434 reproductive proteins, it would imply that there are 145 long-term mating response genes waiting to be discovered in that list alone. Although this is a formal possibility, this number seems high. Importantly, ERC allowed us to explore new functional classes of protein from the female reproductive tract. Previous studies [20], [28] chose male-expressed candidates based on molecular classes that were known to function in sperm storage and fertilization. In contrast, ERC directed us to proteins that unlikely would have been selected for screening, as fra mauro was not annotated to be extracellular and hadley had no predicted functional class. We can also prescribe a strategy to improve ERC analysis by retrospectively analyzing the positive candidates. Very strong correlations (p<0. 01) tested positive more often, so future applications of this method could focus on single, strong correlations rather than those proteins that correlate more weakly (p<0. 05) with multiple network members. Finally, we note that several reproductive proteins showed strong signals of ERC with the SP network but were not quickly testable because RNAi lines were not available. In cases like these, emerging technologies such as the CRISPR/Cas9 system that is now being optimized for Drosophila [78], [79] may in the future enable null mutants to be generated, which could potentially expand the SP network further. By expanding the SP network to include new proteins from both sexes, our results provide a more complete picture of how SP controls female post-mating responses. Until now, SPR was the only known female regulator of SP action [26], but our results show that fra mauro, hadley and Esp are also necessary for sperm-bound SP to exert its long-term effects on females. In addition to their expression in the spermathecae, each of these female genes is expressed in regions other than the female reproductive tract [64]. SPR follows the same pattern: it is expressed in several reproductive regions [26], including the spermathecae, and elsewhere in the adult female. However, only six SPR-expressing neurons in the reproductive tract are required for the SP response [23]–[25]. It is also interesting to compare the fertility phenotypes for fra mauro, hadley, Esp and SPR knockdown females (Figure 2). Knockdown of fra mauro or SPR causes both a long-term fertility deficit and an immediate reduction in egg-laying in the first 24 hours after mating. In contrast, hadley or Esp knockdown females show normal fertility on day 1, but then have reduced fertility over the following days. Assuming that the extent of gene knockdown was sufficient to reveal null-like phenotypes, one possible model to explain these results could be that fra mauro is necessary to facilitate SP-SPR signaling, while hadley and Esp are necessary for the efficient release of SP from stored sperm. SP-SPR signaling is required for full fertility at all time points after mating (Figure S2 and [26]), but impaired release of SP from sperm affects fertility only after day 1 [19]. Another possibility is that fra mauro is required to coordinate temporally the release of sperm from storage when eggs are ovulated and ready to be fertilized. Furthermore, while knockdown of fra mauro, hadley and SPR each caused a reduction in egg hatchability, the magnitude of this effect was by far the greatest for fra mauro (Figures S3, S4). Thus, in addition to laying significantly fewer eggs than controls (Figure 2), fra mauro females also experience far lower egg-to-adult viability. Finally, it is interesting to observe that Esp is a predicted sulfate transporter. In mammalian systems, anion concentration in the female reproductive tract is critical for proper sperm function and fertility [80]. In Drosophila, it is possible that attenuation of extracellular levels of anions such as sulfate in the sperm storage organs affects Sfp-sperm binding, sperm storage, SP release, or another process required for SP network function. Two observations suggest that interactions between SP network proteins may begin in the male. First, CG9997, aquarius and antares are each required for lectins CG1652 and CG1656 to be transferred efficiently to females [21] (Figure 4). It is possible that one or more of the former proteins may bind to either lectin protein as Sfps transit the male reproductive tract during mating. Such binding could protect the lectins from proteolysis or modification. For instance, CG9997 and aquarius both encode serine protease homologs that are predicted to have inactivating mutations in their active sites [70]. It has been speculated that such inactive proteases could act as competitive inhibitors of proteolytic processing by binding to processing targets, rendering them less accessible to the numerous active protease in the seminal fluid [81]. Second, it is presently unclear whether intrepid is transferred at mating, as previous proteomic experiments have not detected this protein in mated females [33]. While intrepid may be transferred but poorly detectable in mated females (e. g. , due to low abundance or rapid degradation), it may, alternatively, act in males to modify or activate another network protein (s). Processing of Sfps within males is observed in other cases. For example, the Drosophila seminal metalloprotease CG11864 is processed in the male reproductive tract during transfer to females, and this processing is required for CG11864 to mediate the processing of additional Sfps in the female reproductive tract [28], [82] (B. LaFlamme, F. Avila et al. , submitted). In nematodes, interactions between a protease, TRY-5, and a protease inhibitor, SWM-1, regulate the activation of sperm during transit through the male reproductive tract [83]–[85]. Thus, it will be interesting to determine whether any members of the SP network are the agents or targets of processing within the male reproductive tract. If network proteins are modified while still in the male, this process may be regulated spatially and temporally by the sequestration of interacting components into distinct compartments of the reproductive tract, including the ejaculatory bulb [86] and vesicles found in secondary cells of the accessory gland [87], [88]. Such compartmentalization could ensure that interacting proteins do not encounter each other until the appropriate time during or after mating. Our results, combined with previous work [20], [21], [26], [89], suggest that at least 13 proteins participate in the SP-mediated post-mating response in female Drosophila melanogaster. How did this complex network arise, and how have its members evolved? Orthologs of the sex peptide receptor (SPR) are found in diverse insect taxa, including mosquitoes, silkworms and moths, and these receptors are responsive to stimulation by D. melanogaster SP [26], [90]. However, SP has not been identified outside of Diptera; a putative SP ortholog was identified by bioinformatics in Anopheles [91], but the short length of SP makes it difficult to detect orthologs in other species, including some drosophilids. Furthermore, the female post-mating responses of insects with SPR orthologs often differ substantially from those of the melanogaster group of Drosophila. For example, D. mojavensis females re-mate more readily than D. melanogaster females [92], and while A. gambiae females become unreceptive to further courtship after a single mating, this behavioral change does not require the transfer of sperm [93]. Within the genus Drosophila, other members of the network show different levels of evolutionary conservation. We identified orthologs of CG1652, CG1656, CG9997 and CG17575 in 11 of 12 sequenced Drosophila species (all but the most distant species, D. grimshawi). Most of the new proteins we identified share this broad distribution throughout the genus. Hadley and fra mauro are found in all 12 species, but appear not to have orthologs in sequenced mosquito species (data not shown). Aquarius and antares show the same species distribution as CG1652, CG1656, CG9997 and CG17575. Esp orthologs are found in only ten species, but these include one member of the more distantly related Drosophila clade, D. mojavensis, suggesting an older origin for this protein. In contrast, intrepid and seminase appear to have evolved more recently, with orthologs detectable only in the Sophophora clade. Orthologs of intrepid were found in 9 of 12 species (all but D. virilis, mojavensis and grimshawi), while seminase orthologs were detected only in D. melanogaster-D. ananassae. Taken together, these varying degrees of evolutionary conservation suggest that the SP network, as it presently functions in D. melanogaster, may have evolved in pieces over time. Indeed, the emergence of the full SP network correlates with changes in remating rate. Frequent mating (daily or more than once per day) was inferred to be the ancestral condition for drosophilids, while less frequent mating is derived and appears in those species (D. melanogaster through D. pseudoobscura) that have all or nearly all of the SP pathway components [94]. Some reproductive proteins of many species have evolved under positive selection [95]–[97]. One proposed explanation for this pattern suggests that males and females may experience sexual conflict over some aspect of reproduction (e. g. , the rate of female remating). Substantial evidence suggests that sexual conflict occurs in D. melanogaster [98]–[100] and is mediated by SP [101]. At the molecular level, the result of sexual conflict could be continual coevolution between male and female protein sequences. Population genetic studies have detected evidence of recent selective sweeps on SP [102] and CG9997 [103], but most other members of the network appear well conserved [33]. One possible explanation centers on the observation that SPR is sensitive to multiple ligands [26], [62], [63], which may constrain its ability to coevolve with SP and thus reduce the requirement for constant coevolution. It will also be instructive to examine the molecular evolution of all network members across the Drosophila phylogeny and to determine whether any have experienced bursts of positive selection on the same phylogenetic lineages, as might be predicted for proteins showing patterns of ERC [50]. We have shown that signatures of evolutionary rate covariation can be used prospectively to identify new members of a protein network. In the context of the Drosophila SP pathway, this genomic approach allowed us to efficiently screen hundreds of known reproductive proteins so as to prioritize candidates for functional analysis, thereby identifying new long-term mating response proteins from both males and females. Interestingly, male and female proteins appear to participate in distinct sections of the SP network, and this separation was reflected in their signatures of correlated evolution. We believe that the ERC approach will be broadly applicable to identifying new members of other protein networks in any taxa for which comparative genomic data are available. We used a combination of published proteomic and transcriptomic data sets and genome-wide expression data to create three sets of reproductive genes used in the analysis: seminal fluid proteins (Sfps), female reproductive tract proteins, and sperm proteins. The first set consisted of 208 genes encoding Sfps that had been identified by mass spectrometry in the reproductive tracts of mated females [32], [33] or predicted secreted proteins from the male accessory gland [34]. The second set included 226 genes expressed in the female sperm storage organs. This set included the D. melanogaster orthologs of EST sequences identified from the spermathecae of D. simulans [36], [38] and EST sequences identified from the seminal receptacle of D. melanogaster [39]. We removed from these sets annotated housekeeping genes (e. g. , ribosomal and mitochondrial proteins) since they were unlikely to interact with proteins in the SP network. Because EST sequencing may not sample all relevant genes, we then supplemented these genes with genes identified in FlyAtlas [64] to be predominantly expressed in the spermathecae (the only female sperm storage organ for which genome-wide expression data are available). The third set included 322 genes that encode proteins in the D. melanogaster sperm proteome [30], [31] and that were found in FlyAtlas to be predominantly expressed in the testis. This filtering was performed to enrich for proteins likely to function specifically in reproduction, since proteins involved in additional biological processes may interact with several partners and thus show dampened signals of ERC. While we used all three sets of genes (756 genes in total) for optimizing the ERC method (see below), we focused our further functional tests on ERC candidates identified from the seminal fluid and sperm storage organ gene sets (434 in total). We identified orthologous genes from 12 Drosophila species using a combination of high-throughput and manual searching. Protein amino acid sequences were produced by the Drosophila 12 Genomes project and downloaded from FlyBase (http: //flybase. org) [61]. The species were: Drosophila melanogaster, sechellia, simulans, yakuba, erecta, ananassae, pseudoobscura, persimilis, willistoni, grimshawi, virilis, and mojavensis. Orthologs were identified using InParanoid, and the resulting groups were aligned by MUSCLE [104], [105]. Many alignments were missing species either due to evolutionary loss or missed gene annotation. To increase the number of species and thereby improve our power, we manually searched for unannotated genes in the 11 non-melanogaster species using a combination of tBLASTn and BLAT. This effort added 81 previously unannotated sequences to a total of 31 alignments. To perform ERC analysis, we first calculated the amount of amino acid divergence for each branch in the species tree for each of the 11,100 orthologous protein alignments produced above; this was done using ‘aaml’ of the PAML package [106]. Next, raw branch lengths were transformed into rates of evolution relative to the expected branch length. This projection operation, introduced by Sato et al. [58], removes the inherent correlation of all proteins due to the underlying species tree and improves the power of ERC to resolve functionally related protein pairs from unrelated pairs [55], [58]. Finally, we used these corrected branch-specific rates to calculate the correlations for all pairs of proteins, resulting in a proteome-by-proteome matrix of correlation coefficients, termed the ERC matrix. To limit the effect of outlier points, we limited all rates to 2 standard deviations from the mean. In spite of our efforts (above) to improve species coverage, most alignments were missing at least one species. We set a minimum species threshold at 5, so species representation ranged from 5 to 12. This heterogeneity required us to create a flexible system to compare ERC results between different sets of species. A table of relative rates (projection operation, above) was produced for each unique set of species shared between protein pairs, resulting in 1,815 projections. Importantly, the distribution of ERC values varied depending on the particular set of species employed. For example, the variance of ERC values is consistently larger for smaller numbers of species (Figure S6). To correct for these effects we converted every observed ERC value in to an empirical p-value based on the observed distribution of ERC values for that particular set of species. The comparison of p-values allowed us to compare ERC results across all protein pairs. Hence, we report all ERC results as p-values ranging from 0 to 1, where a lower value indicates stronger evidence for rate correlation. Significance testing for elevated ERC values in a set of proteins was performed using a proteome-wide permutation test (Figure S1A). The mean ERC value observed between all pairs in the tested set, such as the SP network, was compared to the mean ERC values of 10,000 sets of the same number of proteins randomly chosen from the entire proteome. A p-value for the tested set was computed as the proportion of random sets that had a mean ERC value equal to or greater than the tested set of proteins. Randomly chosen ERC values were taken from the same species-matched projections as in the observed set, which controlled for variation in ERC distributions due to different sets of species present in those genes. The “reproductive protein only” analysis (Figure S1B–C) was performed as above, except that analysis was limited to the 756 Sfps, female proteins, and sperm proteins described above. We further limited this set to the 664 proteins that had detectable orthologs in at least 5 species. Significance testing for single pairs and for sets of proteins was performed as above, through empirical p-values. Calculations of pairwise correlations between pairs of known network proteins and between known network proteins and members of the sets of Sfps and female proteins were performed using this reproductive protein set. To knock down expression of candidate genes, we used a variety of RNAi lines and drivers. Most lines were second-generation (KK) RNAi lines provided by the Vienna Drosophila RNAi Center (www. vdrc. at) [107]; several others were either provided by the Transgenic RNAi Project (TRiP; Harvard University) [108] or constructed in house using the pVALIUM20 vector [109], [110] provided by the TRiP. When possible, we used the tubulin-GAL4 driver to knockdown genes ubiquitously, but in some cases knockdown with this driver caused lethality. When ubiquitous knockdown of a male-expressed Sfp gene caused lethality, we first attempted to use the prd-GAL4 driver [111] to knockdown expression in the accessory glands. However, we observed phenotypes consistent with SP network malfunction when this driver was crossed to a control background strain that does not induce RNAi. Thus, we instead used the ovulin-GAL4 driver [17] to knock down male Sfp genes. To knockdown female genes expressed in the spermathecae, we used the Send1-GAL4 driver [112], sometimes in combination with a UAS-Dicer2 sequence to enhance RNA interference. The RNAi line numbers, specific crosses and genetic controls used are given in Tables S1 and S3. All flies were reared on a 12 hr/12 hr light-dark cycle. Most crosses were performed at room temperature (22°C±1°); some were instead performed at 25° to attempt to induce greater knockdown. We determined the degree of knockdown by using RT-PCR [20], [28] to measure the expression level of each RNAi-targeted gene in knockdown flies and their respective controls, using amplification of the RpL32 transcript as a positive control (see Protocol S1 for further details). For tubulin-GAL4 knockdown, we analyzed RNA isolated from whole flies; for tissue-specific knockdown, we analyzed RNA isolated from dissected reproductive tracts. We qualitatively scored the degree of knockdown as “complete/near complete, ” “partial, ” or “no detectable knockdown”, and we chose for functional analyses only those genes (16 of 21 tested) that showed at least partial knockdown. Figure S7 shows knockdown levels for all positive candidates. For several days after an initial mating, females are reluctant to remate in a one-hour, single-pair test, but only if the SP network is functioning properly [19], [20]. Thus, we initially screened each candidate gene for its effects on a female' s willingness to remate within 1 hour, 4 days after an initial mating, using previously described methods [20]. Positive candidates were then evaluated by the same assay for remating receptivity at 1 day after mating, and for fertility, fecundity and egg hatchability over 10 days after an initial mating. These assays were performed according to previously described methods, with minor modifications. For more detail, see Protocol S1. While all RNAi lines used above were designed to specifically minimize off-target effects [107], [108], we also confirmed that the phenotypes we observed were due specifically to the knockdown of the intended target. We first confirmed that all RNAi-triggering constructs had no predicted off-target effects against the most current D. melanogaster gene annotations [66]. We then tested an additional RNAi line for all genes for which such a line was available (antares, fra mauro, hadley and Esp). These tests controlled for either the insertion site of the RNAi-triggering construct or both the insertion site and the sequence of the RNAi-triggering construct, depending on which type of additional line was available. Details of these lines are given in Table S3. Finally, we note that our rate of positive hits in our screen (33 percent; 6 out of 18 ERC-identified candidates) is dramatically higher than previous estimates of RNAi effects on cell viability (maximum rate: 2. 2 percent, including both true positive effects and potential off-targets) [113]. Thus, our results are unlikely to be due to off-target effects or general effects on cell viability. To examine the production, transfer and processing of known SP network proteins in flies knocked down for a newly identified candidate, we performed Western blot experiments using available antibodies to SP, CG1652, CG1656, CG9997 and CG17575 as previously described [21]. For each positive candidate, we first tested whether SP was retained on sperm over the long term by dissecting 13–20 lower female reproductive tracts for each treatment at 4 days after the start of mating (ASM). While the number of female reproductive tracts per lane across experiments varied within this range, pairs of samples being compared never differed by more than 2 tracts. Extracted proteins were run on 15% acrylamide gels, transferred to membranes, and then probed for SP and alpha-tubulin (as a loading control) as previously described. For candidates that caused a reduction of SP levels in females at 4 days ASM, we then evaluated the production, processing and transfer of the known network proteins by testing for their presence in male reproductive tracts and in mated females at 1 hr ASM. Proteins were separated on 10. 6% acrylamide gels and then transferred and probed for as described previously. Approximately 0. 5–1 male reproductive tract equivalents and 2–4 lower female reproductive tract equivalents were loaded in each lane. While the number of female reproductive tract equivalents per lane varied between blots for different SP network proteins, comparisons between knockdown and control flies for any given protein were performed with an equal number of reproductive tracts in each lane. As a loading control for each blot, we primarily used alpha-tubulin. In cases where CG1652 and CG1656 co-migrated with alpha-tubulin, we also examined a consistently observed cross-reactive band.

Reproduction requires more than a sperm and an egg. In animals with internal fertilization, other proteins in the seminal fluid and the female are essential for full fertility. Although hundreds of such reproductive proteins are known, our ability to understand how they interact remains limited. In this study, we investigated whether shared patterns of protein sequence evolution were predictive of functional interactions by focusing on a small network of proteins that control fertility and female post-mating behavior in the fruit fly, Drosophila melanogaster. We first showed that the six proteins already known to act in this network display correlated patterns of evolution across the Drosophila phylogeny. We then screened hundreds of otherwise uncharacterized male and female reproductive proteins and identified those with patterns of evolution most similar to those of the known network proteins. We tested each of these candidate genes and found six new network members that are each required for long-term fertility. Using molecular genetics, we also observed that the steps in the network at which these new proteins act are consistent with their strongest evolutionary correlations. Our results suggest that patterns of coevolution may be broadly useful for predicting protein interactions in a variety of biological processes.

Abstract Introduction Results Discussion Methods

animal models drosophila melanogaster model organisms genetics comparative genomics biology genomics evolutionary biology genomic evolution

2014

Evolutionary Rate Covariation Identifies New Members of a Protein Network Required for Drosophila melanogaster Female Post-Mating Responses

Parasitic zoonoses (PZs) pose a significant but often neglected threat to public health, especially in developing countries. In order to obtain a better understanding of their health impact, summary measures of population health may be calculated, such as the Disability-Adjusted Life Year (DALY). However, the data required to calculate such measures are often not readily available for these diseases, which may lead to a vicious circle of under-recognition and under-funding. We examined the burden of PZs in Nepal through a systematic review of online and offline data sources. PZs were classified qualitatively according to endemicity, and where possible a quantitative burden assessment was conducted in terms of the annual number of incident cases, deaths and DALYs. Between 2000 and 2012, the highest annual burden was imposed by neurocysticercosis and congenital toxoplasmosis (14,268 DALYs [95% Credibility Interval (CrI): 5450–27,694] and 9255 DALYs [95% CrI: 6135–13,292], respectively), followed by cystic echinococcosis (251 DALYs [95% CrI: 105–458]). Nepal is probably endemic for trichinellosis, toxocarosis, diphyllobothriosis, foodborne trematodosis, taeniosis, and zoonotic intestinal helminthic and protozoal infections, but insufficient data were available to quantify their health impact. Sporadic cases of alveolar echinococcosis, angiostrongylosis, capillariosis, dirofilariosis, gnathostomosis, sparganosis and cutaneous leishmaniosis may occur. In settings with limited surveillance capacity, it is possible to quantify the health impact of PZs and other neglected diseases, thereby interrupting the vicious circle of neglect. In Nepal, we found that several PZs are endemic and are imposing a significant burden to public health, higher than that of malaria, and comparable to that of HIV/AIDS. However, several critical data gaps remain. Enhanced surveillance for the endemic PZs identified in this study would enable additional burden estimates, and a more complete picture of the impact of these diseases. Various parasites infecting humans depend on vertebrate animals to complete their life cycle. Humans most commonly become infected with these zoonotic parasites through consumption of infected hosts or through fecal-oral contamination. The results of these infections may vary from asymptomatic carriership to long-term morbidity and even death. Although data are still scarce, it is clear that these parasitic zoonoses (PZs) present a significant burden for public health, particularly in poor and marginalized communities [1], [2]. Moreover, PZs can lead to significant economic losses, both directly, through their adverse effects on human and animal health, and indirectly, through control measures required in the food production chain [3], [4]. Estimates of the impact of diseases on public health, generally referred to as burden of disease, may be valuable inputs for decision makers when setting policy priorities and monitoring intervention programs. In Nepal, it is now recognized that health sector needs should be prioritized, and that disease burden should be considered as one of the bases for this prioritization [5]. However, disease burden estimates are not readily available. While the World Health Organization and the Global Burden of Disease (GBD) initiative have generated such estimates for Nepal, these were largely based on regional extrapolations, and, more importantly, included only a limited number of PZs [6], [7]. If disease burden estimates are to be used for priority setting, an incomplete assessment of the burden of PZs may lead to a vicious circle of under-recognition, a wrong ranking of priorities and under-funding for research, prevention and control programs [8]. To address this issue, a disease burden assessment of PZs was conducted in Nepal. Ideally, the primary data sources for such studies would be official surveillance data and death registers. In Nepal, however, these data sources have limited value in terms of PZs. The official passive surveillance system of the Government of Nepal, the Health Management Information System (HMIS), has been reported to suffer from inconsistencies, incomplete reporting, and under-reporting from mainly central-level and private hospitals [9], [10]. Active surveillance systems are in place, but only target certain vaccine-preventable diseases, and not PZs. Death registration is reported to have a completeness rate of 32% [11]. We therefore opted for a more comprehensive approach, based on a systematic review of all possible secondary data sources related to PZs in Nepal from 1990 to 2012. This comprehensive review allowed us to identify endemic and possibly endemic PZs, and, subsequently, to quantify the disease burden of those PZs for which sufficient quantitative data were available. The twenty PZs considered in this study are listed in Table 1. This selection is based on a recent review of the world-wide socioeconomic burden of PZs [1] and a review of emerging food-borne parasites [12], as many PZs may be classified as being food-borne. Seven of the considered PZs also belong to the group of neglected tropical diseases, i. e. , leishmaniosis, cystic and alveolar echinococcosis, cysticercosis, food-borne trematodosis, schistosomosis and soil-transmitted helminthosis [13], [14]. Direct and indirect evidence on the occurrence of the considered PZs was located through a systematic search of national and international peer-reviewed and grey literature. Direct evidence was defined as any data on prevalence, incidence or mortality of the PZ in humans. Indirect evidence was defined as occurrence of the concerned parasite in animal hosts or in the environment (e. g. , water, soil). If no direct or indirect evidence could be identified from Nepal (further referred to as “local” evidence), recent case reports were sought from (North) India, Nepal' s largest neighbor with whom it shares an open border in the west, south and east, and from the Tibet Autonomous Region, which borders Nepal in the north (Figure 1). For each PZ, we constructed a search phrase consisting of the key word “Nepal” and any element of a list containing the name of the PZ, possible synonyms, and the name (s) of the causative parasite (s) (Table S1-1 in Supporting Information S1). Manuscript titles were retrieved through searching PubMed, Web of Science, WHO Global Health Library, Asia Journals OnLine (AsiaJOL) and MedInd. If available, the major Nepalese journals were additionally searched through their websites (Table S1-2 in Supporting Information S1). In addition, the thesis libraries of Tribhuvan University (Kathmandu, Nepal) and the Institute of Animal Agriculture Sciences (Rampur, Chitwan district) were manually explored to find relevant manuscripts. Dissertations were also collected from the website of the Veterinary Public Health master course jointly organized by Chiang Mai University (Thailand) and Freie Universität Berlin (Germany), as this program has a regular intake of Nepalese students. No dissertations were sought from countries neighboring Nepal, as we did not have prior knowledge of masters courses organized in these countries with a regular intake of Nepalese students. In a second step, the retrieved titles were screened for eligibility by applying a set of predefined criteria to the titles and, if possible, to the abstracts and full texts. Only papers published in 1990 or later were considered eligible, and no restrictions were placed on the language of publication. For the qualitative assessment, documents were only excluded if they did not relate to the PZ in question, or if they did not pertain to Nepal or Nepalese patients. For the quantitative assessment, additional restrictions were put on the year of publication (between 2000 and 2012), the study setting and population (Nepalese patients infected in Nepal), and the type of information (quantitative, thus excluding case reports and case series). Finally, additional titles were sought for using forward and backward reference searches (so-called “snowballing”). In the forward reference search, the titles eligible for the qualitative assessment were entered in Google Scholar (http: //scholar. google. com/) to obtain a list of articles citing the former. The latter were then screened using the same criteria as used in the initial searches. In the backward reference search, the reference lists of the initially retrieved eligible documents were hand-searched and the same criteria were applied. The forward and backward searches were repeated until no more new information could be retrieved. Figure 2 presents a generic flow diagram of this applied search strategy. Relevant data on study setting, diagnostic methods and study results were extracted from all eligible articles, and entered in spread sheet documents for further use. This initial assessment aimed at classifying the considered PZs according to their presumed endemicity status and data availability. To this end, we defined four categories: Additionally, information regarding the zoonotic nature of potentially zoonotic parasites was considered, with respect to alternative (dominant) anthroponotic transmission. Where possible, the prevalence of each PZ classified as “probably endemic & quantifiable” was modeled using a random effects meta-analysis in a Bayesian framework. In this model, it is assumed that the number of positive samples xi in each study results from a binomial distribution with sample size ni and a study-specific true prevalence θi, which is in its turn the result of an overall true prevalence π and a random study effect. The study effect is assumed to be normally distributed with mean zero and variance τ2. The prior distribution of τ2 is Gamma with scale and shape parameter equal to 1, while a Normal distribution with mean 0 and precision 0. 001 was used as prior for the logit-transformed true prevalence. Markov chain Monte Carlo methods are used to fit the model. More information on the meta-analysis model is provided in Supporting Information S2. If data allowed, the health impact of the concerning PZs was also quantified as the number of incident cases, deaths and DALYs. The DALY metric is a summary measure of public health, widely used in disease burden assessments and cost-effectiveness analyses [6], [7]. DALYs represent the overall number of healthy life years lost due to morbidity and mortality, hereby facilitating comparisons between diseases, and between countries and regions. The standard DALY formulas are: DALY = YLD + YLL YLD = Number of cases * Duration * Disability Weight YLL = Number of deaths * Life expectancy at age of death Calculation of DALYs was done using the standard formulas, and implemented in a fully stochastic framework using the DALY Calculator in R [15]. Supporting Information S3 presents the disease models and input distributions used for assessing the burden of the concerned PZs. We calculated undiscounted and unweighted DALYs, based on the Coale-Demeny model life table West, as our base case scenario. However, in order to enhance comparability of our estimates to estimates made by other authors, we performed scenario analyses by varying the time discount rate from 0% to 3%, by including age weighting, and by using the life expectancy table developed for the GBD 2010 study [7]. These different scenarios were denoted by DALY{K, r}, with K equal to 0 for unweighted DALYs and equal to 1 for age-weighted DALYs, and with r the time discount rate. For all scenarios, results were calculated at the population level (i. e. , absolute number of DALYs per year) and at the individual level (i. e. , relative number of DALYs per symptomatic case). Incident cases, deaths and DALYs were calculated for reference year 2006, i. e. , the midpoint of the eligible publication period, 2000–2012. The total population size for 2006 was calculated as the mean of the population sizes estimated in the 2001 and 2011 censuses. The age and sex distribution of the 2006 population was derived from the 2006 Nepal Demographic and Health Survey [16]. Table 2 presents the resulting population sizes used in the calculations. The data collection activities required for this study were approved by the ethical review board of the Nepal Health Research Council (Ramshahpath, Kathmandu, Nepal) and of the Ghent University Hospital (Ghent, Belgium; registration number B670201111932). For all twenty considered PZs, we identified 267 unique peer-reviewed documents and 50 unique dissertations. All identified documents were published in English. Table 3 summarizes the results of the systematic review for each considered PZ. Table 4 presents the results of the qualitative classification of PZs. Out of the twenty considered PZs, only Anisakidae infection, zoonotic sleeping sickness (trypanosomosis) and zoonotic schistosomosis were classified as probably not endemic as no direct or indirect evidence was found. Seven PZs were classified as potentially endemic, i. e. , alveolar echinococcosis, angiostrongylosis, capillariosis, dirofilariosis, gnathostomosis, sparganosis and cutaneous leishmaniosis. The ten remaining PZs were considered probably endemic, and the burden of three of these, neurocysticercosis, congenital (but not acquired) toxoplasmosis and cystic echinococcosis, could be fully quantified in terms of incident cases, deaths and DALYs. As disease burden estimates are of increasing importance for policy making and evaluation, the need for such estimates becomes eminent. In the late 1990s, the World Bank commissioned a comprehensive analysis of health care delivery in Nepal. Several recommendations were made for the further development of the Nepalese health sector, one of which was the establishment of priorities [168]. These recommendations were carried forward in the development of the Nepal Health Sector Programmes (NHSP), short-term strategic frameworks for the further development of the health sector. Since then, disease burden is recognized as one of the bases for setting program priorities [5]. However, when routine surveillance systems are performing poorly and baseline epidemiological studies are rare, these estimates are not readily available [169]. In this paper, we present the first comprehensive systematic review of the burden of PZs in Nepal. Information was sought from the international and national peer-reviewed scientific literature, and an important source of information was found in dissertations. The information found allowed qualitative assessment of the twenty PZs considered. However, quantitative estimates of prevalence or disease burden were possible for only a few. Nepal is considered endemic for at least ten PZs, and might be endemic for seven others. Most of these diseases probably only have a small public health impact. However, neurocysticercosis and congenital toxoplasmosis are likely to impose an important burden to public health. Indeed, if we compare with the three “major” infectious diseases, we see that the estimated burden due to major clinical manifestations of three PZs, with in total 0. 57 DALY{1,0. 03} per 1000 people, is higher than that of the WHO 2004 GBD estimate for malaria (0. 05 DALY{1,0. 03} per 1000), comparable to that for HIV/AIDS (0. 74 DALY{1,0. 03} per 1000), but substantially lower than that for tuberculosis (5. 45 DALY{1,0. 03} per 1000) [6]. These comparisons suggest that greater attention for PZs in Nepal is warranted. Toxoplasmosis is for instance not reported in any official Nepalese data collection system, and cysticercosis and toxoplasmosis were not considered in the WHO 2004 GBD update [6]. As a result, the incidence of congenital toxoplasmosis remains a critical data gap, and considerable uncertainties remain regarding the epilepsy prevalence and proportion of neurocysticercosis-associated epilepsy. Data on the zoonotic potential of intestinal helminths and protozoa and their health effects are lacking, although these infections may represent a considerable additional health burden. In our study, certain methodological choices were made with as a consequence certain limitations. First, instead of applying strict inclusion/exclusion criteria, we aimed at collecting as much relevant information as possible. Inherently, this leads to large heterogeneity in the collected quantitative data. As a result, our burden estimates have large uncertainty intervals, making it for instance impossible to statistically distinguish the burden of neurocysticercosis and congenital toxoplasmosis. For congenital toxoplasmosis, as no direct evidence was available, we estimated the incidence based on a single age-specific seroprevalence study. Clearly, this puts an important constraint on the representativeness of our resulting burden estimate. Direct evidence on the incidence of congenital toxoplasmosis (e. g. , through serological studies on newborns), preferably obtained through a multi-center study, is therefore needed to confirm our burden estimate. Second, uncertainty was introduced by the selection and valuation of the clinical outcomes for the three diseases. We based our disease models on published studies [170]–[173], but note that other authors applied alternative ones. For instance, Bhattarai et al. [174] also included severe headaches in their assessment of the burden of neurocysticercosis in Mexico, whereas this was deemed infeasible in our study. Likewise, the disability weights assigned to the different included clinical outcomes were derived from earlier studies [170]–[173], in order to enhance comparability with those studies. Nevertheless, other studies, including the GBD studies, are less transparent about their applied disease models and disability weights, impeding unambiguous comparisons. For instance, the GBD 2010 study estimated the number of DALY{0,0} in Nepal to be 667 [141–2073] for “echinococcosis” and 4220 [2786–6022] for cysticercosis [175]. Given a lack of knowledge on the disease models, disability weights and data behind these estimates, it is difficult to assess the reason for any differences or similarities with our estimates. Toxoplasmosis and other PZs appear to be absent from the GBD 2010 study. Third, subjective methodological choices regarding the calculation of DALYs may lead to further uncertainty. We tried to deal with this source of uncertainty by calculating DALYs under different common sets of normative assumptions, i. e. , no discounting and age weighting, 3% time discounting and no age weighting, 3% time discounting and age weighting; and by calculating DALYs based on both the Coale-Demeny model life table West and the new GBD 2010 life expectancy table. As expected, time discounting led to smaller burden estimates. The difference between both life tables was minimal. In addition, this study focused on the population burden of PZs. Some PZs, however, might have an important individual burden, even though their population burden is negligible or small. Likewise, the burden suffered by specific sub-populations (e. g. , caste or ethnic groups), might be much higher than average population burden [85]. Finally, due to a lack of time and resources, we had to place restrictions on the nature of the diseases to be studied, and on the nature of the burden estimates to be generated. Indeed, this study only focuses on the burden of parasitic zoonoses. However, as means for interventions are poor, future integrated control should be packaged by, for instance, simultaneously controlling cystic echinococcosis, brucellosis and rabies. This analysis should therefore be extended to the burden of bacterial and viral zoonoses in Nepal. By quantifying the burden in terms of incidence, mortality and DALYs, we also focused on the health impact of the concerned diseases. Some PZs might have an important economic impact, for example in terms of livestock health, or might reduce psycho-social wellbeing in a way not captured by the applied metrics. Truly evidence-informed priority setting and decision making should take in account all these aspects of disease burden, implying that our estimates should be complemented by others. Despite these limitations, this study has identified the most important PZs for Nepal, as far as existing data allows. The quantitative estimates of disease burden for three of these diseases suggest that PZs deserve greater attention and more intensive surveillance. As population and disease transmission dynamics change over time, disease burden changes dynamically as well. Therefore, the presented results should be updated regularly, and this exercise should be extended to other groups of neglected diseases or even to a full national burden of disease study. We therefore hope that this study will stimulate further research, so that the overall human health burden in Nepal can be better characterized. In the long term, however, continued efforts to improve surveillance and database system at the local level should enable truly monitoring of disease burden over time.

Various parasites that infect humans require animals in some stage of their life cycle. Infection with these so-called zoonotic parasites may vary from asymptomatic carriership to long-term morbidity and even death. Although data are still scarce, it is clear that parasitic zoonoses (PZs) present a significant burden for public health, particularly in poor and marginalized communities. So far, however, there has been relatively little attention to this group of diseases, causing various PZs to be labeled neglected tropical diseases. In this study, the authors reviewed a large variety of data sources to study the relevance and importance of PZs in Nepal. It was found that a large number of PZs are present in Nepal and are imposing an impact higher than that of malaria and comparable to that of HIV/AIDS. These results therefore suggest that PZs deserve greater attention and more intensive surveillance. Furthermore, this study has shown that even in settings with limited surveillance capacity, it is possible to quantify the impact of neglected diseases and, consequently, to break the vicious circle of neglect.

Abstract Introduction Materials and Methods Results Discussion

2014

The Burden of Parasitic Zoonoses in Nepal: A Systematic Review

Leishmaniases are neglected parasitic diseases in spite of the major burden they inflict on public health. The identification of novel drugs and targets constitutes a research priority. For that purpose we used Leishmania infantum initiation factor 4A (LieIF), an essential translation initiation factor that belongs to the DEAD-box proteins family, as a potential drug target. We modeled its structure and identified two potential binding sites. A virtual screening of a diverse chemical library was performed for both sites. The results were analyzed with an in-house version of the Self-Organizing Maps algorithm combined with multiple filters, which led to the selection of 305 molecules. Effects of these molecules on the ATPase activity of LieIF permitted the identification of a promising hit (208) having a half maximal inhibitory concentration (IC50) of 150 ± 15 μM for 1 μM of protein. Ten chemical analogues of compound 208 were identified and two additional inhibitors were selected (20 and 48). These compounds inhibited the mammalian eIF4I with IC50 values within the same range. All three hits affected the viability of the extra-cellular form of L. infantum parasites with IC50 values at low micromolar concentrations. These molecules showed non-significant toxicity toward THP-1 macrophages. Furthermore, their anti-leishmanial activity was validated with experimental assays on L. infantum intramacrophage amastigotes showing IC50 values lower than 4. 2 μM. Selected compounds exhibited selectivity indexes between 19 to 38, which reflects their potential as promising anti-Leishmania molecules. Leishmaniases are neglected diseases caused by multiple protozoan parasite species of the genus Leishmania. These vector-borne diseases affect more than 98 countries that are mainly developing countries with limited public health resources. Additionally, more than 300 million people are at risk of being infected. [1] Three different clinical forms are described: Cutaneous Leishmaniasis (CL), Muco-Cutaneous Leishmaniasis (MCL) and the most severe Visceral Leishmaniasis (VL), which is fatal if left untreated. Each year, 1. 5 to 2 million cases are reported, among which 0. 5 million are cases of VL that cause 40,000 deaths per year. [1] Currently available control measures are mainly based on diagnosis, patient treatment and vector control. Mainstay therapy is based on the use of pentavalent antimonials. [2,3] Commonly used second-line drugs are miltefosine, amphotericin B, liposomal amphotericin B and paromomycin. All these treatments are given by injections, except for miltefosine that is administered orally. [4] They require long treatment courses, are toxic and costly, and have adverse effects. [3,5, 6] The identification of novel drug targets, therapeutic molecules or immune modulators that enhance the response to treatment constitute research priorities, particularly against the fatal VL mainly caused by L. donovani and L. infantum. [7] Different criteria help to define potential drug targets: these include expression in relevant life stages, unique genetic or biochemical properties including essentiality, druggability and structural features that allow selection of inhibitors, and assayability. [8] Different targets are being investigated. [8] We focus our work on the Leishmania infantum translation initiation factor 4A (LieIF). [9] Translation factors play key roles in the cell and they are considered as relevant drug targets in cancers. In particular, the translation initiation factor eIF4A, [10–14] the prototype of the DEAD box proteins (DBPs) family, is considered a potent target. [15,16] It plays a pivotal role in the translation initiation complex eIF4F as an essential enzyme. [17,18] In Leishmania infantum, experimental evidence assigned an eIF4A-like functional role to a protein called LieIF that is encoded by LinJ. 01. 0790/LinJ. 01. 0800 genes mapping to chromosome 1 (LmjF. 01. 0770/LmjF. 01. 0780 in L. major). [9,19,20] In previous investigations, we have shown that LieIF has the ability to bind to eIF4G of yeast in vitro, and that it has a dominant-negative phenotype when expressed in yeast, resulting in growth reduction. [9] LieIF is expressed in both parasite stages [21–25], and it is involved in kinases signaling cascades of the amastigotes leading to its phosphorylation on THR135. [25] Interestingly, it has been identified among proteins of the secretome and exosomes of infectious promastigotes of L. infantum [26,27] and of L. donovani. [28] In addition, LieIF is a vaccine subunit that exerts a natural Th1-type adjuvant property. [29,30] LieIF is an RNA-dependent ATPase and an ATP-dependent RNA helicase. [9] It shares 53% identity with the mammalian eIF4AI (DDX2A) and 57% identity with eIF4AIII (DDX48), both of which belong to the RNA helicase family of the DEAD-box proteins (DBPs). [31,32] The amino acids (AAs) sequence of LieIF contains the eleven characteristic motifs of the DBPs. The fact that LieIF is an RNA-dependent ATPase and an ATP-dependent RNA helicase confirms that it belongs to the DBPs family. [9] DBPs are RNA helicases associated with all processes involving RNA, from transcription, translation, splicing, RNA modification to RNA degradation. [33] They share a ∼400 residue-long core region containing the characteristic motifs of the DBPs (Fig 1), and non-conserved flanking regions. Moreover, they all present a dumbbell 3D shape, consisting of two RecA-like domains connected by a flexible linker having variable size and sequence (Fig 2). [34] The N-terminal domain of the conserved core, also called domain 1, contains motifs Q, I, Ia, GG, Ib, II and III. The C-terminal domain, called domain 2, contains motifs IV, QxxR, V and VI. These motifs are important for the biological activities of the DBPs. It has been demonstrated that: (i) motifs I and II from domain 1 are implicated in ATP binding and hydrolysis; (ii) motif III couples ATP and RNA binding and therefore indirectly affects the unwinding activity; and (iii) motifs Ia, GG and Ib in domain 1, and motifs IV, QxxR and V in domain 2 are implicated in RNA binding. [33–41] Domain 2 also is involved in ATP binding through motifs V and VI. Noticeably, ATP binding and hydrolysis, as well as RNA binding and unwinding, are highly dependent on a complex and not yet fully elucidated network of interactions between both domains. Inter-domain interactions occur mainly when a DBP binds the substrate, which results in a compact closed conformation. In this structure, most of the conserved motifs are located in the cleft formed between the two domains, and they largely interact with Mg2+, ATP and RNA ligands. [37–41] RNA bound to the DBP contacts both domains and stabilizes the closed conformation. Inversely, in the absence of ATP or RNA substrates, no inter-domain interactions are observed and an open conformation is adopted. In this case, the two domains present different relative orientations [37,42–44] due to the linker flexibility. The conformational change (open to closed) occurs when both ATP and RNA are bound. [45,46] Once the RNA is unwound, ADP and inorganic phosphate are released and the open conformation is adopted. A new catalytic cycle can then take place. Comparisons of LieIF with yeast eIF4A, a functional homolog of mammalian eIF4A, showed similar enzymatic activities, as one might expect for enzymes involved in the same process. [9] Still, significant differences in the biochemical properties, such as their affinity for ATP and RNA, were observed. [9] Notably, LieIF appears to have a higher affinity for RNA than the yeast protein. LieIF has less affinity (higher Km) for ATP and a higher kcat value for the ATPase activity. It has similar affinity for both ADP and ATP, while in yeast and human eIF4A the affinities for ADP are about three times higher. It also has a broader optimum range of divalent cation (Mg2+) concentrations for its ATPase reaction. Thus, there were clear differences in the properties between LieIF and yeast eIF4A (and presumably the human protein) that in principle can be used to selectively target the LieIF protein. [9] At the primary sequence level, the two proteins show the highest divergence in their N-terminal parts. Deleting the most divergent 25 N-terminal residues that are outside the conserved core abolishes the dominant-negative phenotype that LieIF exerts when expressed in yeast. [9] Nevertheless, neither LieIF nor mouse eIF4AI complemented for the loss of the eIF4A-encoding genes in yeast. [9,47] This suggests the existence of organism-specific interactions between protein partners in the cell that are mediated through the N-terminal sequences. [9] Moreover, it has been proposed that Vasa, a related DEAD-box protein from Drosophila, regulates ATP binding using residues located in the non-conserved amino-terminal sequence. [39] Thus, the importance of LieIF in translation initiation and its significantly distinctive features constituted strong arguments to consider this protein as a potential drug target. Moreover, highly conserved proteins implicated in vital processes are recognized as potential targets for drug discovery. [48,49] In this work, we present a computational approach for the in silico selection of novel small molecules targeting LieIF followed by a biochemical screening for inhibiting its ATPase activity, and we present evidence for the biological effects of LieIF inhibitors on both L. infantum promastigotes and intracellular amastigotes. We used available structure information on the DBPs from the Protein Data Bank (PDB) [50,51] to build 3D models of LieIF through a comparative modeling approach. We generated open and closed conformation models. We validated their stereochemical quality and their stability in molecular dynamic (MD) simulations. MD trajectories were then used to identify relevant cavities, and two potential binding pockets were selected on the open conformation of LieIF. Virtual screenings (VS) were performed with these pockets and a filtering protocol was set for each pocket using Self-Organizing Maps (SOMs) as a clustering technique. Other chemical, energy-based and geometrical filters were used to select a final set of molecules. We then assessed the effects of these molecules on the ATPase activity of LieIF and its mammalian homolog DDX2A (eIF4AIMus). The most promising hit, 208, was used as a bait to search and select for 10 chemical analogues that were tested for a potential inhibiting effect on the LieIF ATPase activity and on promastigote viability. The inhibitors affected the promastigotes viability, did not present toxic effects on mammalian cells and reduced the number of amastigotes in the infected cells. This study is a first step towards the validation of LieIF as a potential drug target and identifies chemically related compounds as promising prototypes of novel leishmanicidal compounds. We used an in-house implementation of the Self-Organizing Maps (SOM) algorithm first introduced by Kohonen [63] to analyze the ligand docking poses upon the different virtual screenings that we present below. We trained a 2D periodic map, (Ωij) 0≤i≤I, 0≤j≤J, with n input vectors containing the Euclidean distances between the Cα of each amino acid defining the targeted pocket and the center of mass of each of the n docking poses. The map dimensions I and J were set to 50. The map was initialized randomly with a uniform distribution preserving the range of values composing the input vectors. The training process was composed of cycles. In each cycle, each input vector was presented once in random order and the map was updated after each presentation. Two phases, similar to that presented by Bouvier [64], were pursued. In the first phase ϕ = 1, two training cycles were performed with constant radius and learning rate equal to 36 and 0. 5, respectively. In the second phase, three cycles were performed. The radius and the learning rate decrease exponentially from 36 to 1 and from 0. 5 to 0, respectively. The decay constant of the exponential, λϕ, was equal to the total number of iteration per phase divided by 10. An efficient way to visualize the SOM map was the unified distance matrix, called the U-matrix. [65] It contained the mean euclidean distance of the map neurons to their respective 26 neighbors. Clusters containing neurons with U-values lower than a given cutoff (in Å) could be defined. They are referred to as clusters with high homogeneity in the present manuscript. Virtual screenings (VS) of the French Academic Compound Library, called “Chimiothèque Nationale” (CN) [66,67] were performed on the identified pockets. The version of the database used in this work contained 43407 chemical compounds. All possible stereoisomers were generated for each compound using Corina. [68,69] This gave 95493 Mol2 records for docking. The receptor and ligands were prepared for docking with Chimera: [70,71] Gasteiger charges were added to receptor and ligands, and hydrogen atoms were added to the receptor. The grid tool from Dock6. 0 [72,73] (UCSF Dock) was used to generate the energy grid. The spheres defining the docking space within the targeted pockets were generated with sphgen. Default parameters were used for the docking with Dock 6. 4. Twenty poses were generated, when possible, and recorded with their grid-based scores. AutoDock vina 1. 1. 2 [74] (ADvina) also was used in this work. It required input files of the receptor and the ligands in pdbqt format. We used the Open Babel converter [75] to generate them from the Mol2 files. Default parameters were used to generate 20 docking poses, when possible, for each chemical compound. The most promising hit, identified through virtual and biochemical screenings, and all compounds constituting the Zinc database [79] were decomposed using the circular Morgan Fingerprints [80,81] as implemented in RDkit. [82] Fingerprints were calculated by decomposition of the compounds into substructures with a user-defined radius limit (set to 2 atoms). A unique integer identifier was assigned to each substructure. Fingerprints consisted in a vector of the substructure identifiers and their number of occurrences in the corresponding compound. To calculate similarity between these fingerprints, the Jaccard similarity criterion was extended to integer values. For that, each integer variable of value, n, is replaced by a series of booleans of which the first n are set to true, the rest being false. Hence, for descriptor vectors A and B, restricted for computation purpose to the Nid integers for which Ai or Bi is/are non-zero, the similarity can be calculated as follows: J (A, B) = ∑ i = 1 N i d m i n (A i, B i) ∑ i = 1 N i d m a x (A i, B i) (1) Extended Jaccard similarity was calculated between the bait (compound 208) and each compound of the Zinc database to rank them from most to least structurally homologous. Among stereoisomers, the compound presenting the closest stereochemistry was selected. Molecules selected in silico were purchased from the corresponding chemists through the French Academic Compound Library System [66], for the minimal costs of their shipping. Minimal quantities necessary for preliminary ATPase assays could be obtained. Compounds 6-α/β-aminocholestanol (208) and 6-α-aminocholestanol (20) were provided by Université de Caen de Basse-Normandie, Centre d’Études et de Recherche sur le Médicament de Normandie (CERMN), UFR des Sciences Pharmaceutiques, under the references MR26628 and MR26620, respectively. Compound 6-ketocholestanol (48) was purchased from Sigma-Aldrich (St. Louis, MO, USA) under the reference K1250. The other 8 compounds also were purchased from Sigma-Aldrich. Stock solutions of all compounds were prepared at 10 mM in dimethyl sulfoxide (DMSO; Sigma-Aldrich). We used comparative modeling to generate 3D models of LieIF in two different states. Ten templates were used for the open form (ligand-free), and seven templates were used for the closed form (substrate-bound; Table 1). Eight out of ten ligand-free templates and four out of seven substrate-bound ones had identity rates (IR) between the protein target and its templates above the twilight region, as needed for robust model construction (i. e. , IR ≥ 30%, Table 1). [89–91] Templates having IR< 30% were nevertheless kept for diversity sake. LieIF models were built and assessed for their robustness. Ten models were generated for each state and their Ramachandran plots were assessed (see S1 Table for a summary) in order to help us select the most reliable model for each state. For the ligand-free models, we chose model N°4 presenting 95. 3% residues within the favored region (98% expected); 3. 0% residues within the allowed region (2. 0% expected) and 1. 7% residues within the outlier region (0. 0% expected). For the substrate-bound, model N°3 was considered as the best structure with 97. 5% of the residues within the favored region, 2. 0% within the allowed region and only 0. 5% within the outlier. These two structures were selected and will be referred to as apo-LieIF and holo-LieIF, respectively (Fig 2). They presented different conformations as expected; apo-LieIF resembled the open conformation of the DBPs and holo-LieIF was quite compact representing the closed conformation (see their Ramachandran plots in S1 (a) and S1 (b) Fig, respectively). Both apo-LieIF and holo-LieIF presented unstructured termini. These regions were described in many DBPs as intrinsically disordered. [36] On some of the unbound models of LieIF, the N-terminal sequence tended to fold into the inter-domain cleft. Hence, we considered their structures as unreliable, since they were obtained through a comparative modeling but lacked reliable alignments with the templates’ termini. In fact, many templates had been truncated (Table 1) for experimental reasons, and high divergence was observed between LieIF models and their templates at the N- and/or C-terminal regions. Thus, the termini structures were removed as a conservative measure prior to MD simulations and VS calculations. Twenty-four N-terminal and seven C-terminal residues were removed. Among them, 4 out of a total of 12 and 1 out of 8 were in the allowed region of the Ramachandran plots of apo-LieIF and holo-LieIF, respectively. For the outliers, these figures were 3 out of 7 and 0 out of 2, respectively. Truncated structures of LieIF [AA 25-396] were denoted apo-LieIFtrunc and holo-LieIFtrunc. The resulting percentages of favored, allowed, outliers amino acids were 97. 4,1. 5,1. 1%, and 99. 6,0. 4,0. 0% for apo-LieIFtrunc and holo-LieIFtrunc, respectively. To verify the stability of the structures and to probe their local relaxations, we ran MD simulations for apo-LieIFtrunc and holo-LieIFtrunc during 2 ns. As we did not mean to study longer-term protein motion, relatively short simulation times were chosen. For comparison purposes, we performed similar calculations for the mammalian eIF4AI using chain A of the PDB entry 3EIQ (3EIQ_A). Holo-LieIFtrunc RMSD varied within 0. 5 and 2. 0 Å, a fairly stationary evolution through time, thus indicative of its stability. In contrast, apo-LieIFtrunc displayed RMSD within 0. 5 and 4. 5 Å, presenting increasing values through the trajectory (S2 Fig). These somewhat larger variations could seem wide, but in fact they depict higher flexibility of the protein in its unbound form due to the presence of a flexible linker between the two fairly independent domains (Fig 2; apo-LieIFtrunc). In this latter case, the trajectory would embed higher conformational diversity. Noticeably, similar variation could be observed with the crystal structure of the mammalian protein (3EIQ_A), leading us to consider apo-LieIFtrunc as a more relevant state of the protein as compared to holo-LieIFtrunc (S2 Fig). Substrates of eIF4AI bind within the inter-domain cleft, which makes the active site definition too fuzzy and large for accurate docking simulations. Thus, we used the MD trajectories of both states of LieIF and 3EIQ_A to search for pockets that may have functional relevance for the parasite protein, but no equivalent on the mammalian counterpart. The compact structure of holo-LieIFtrunc presented small fluctuations, and no interesting cavities could be detected (S3 Fig). Conversely, apo-LieIFtrunc presented multiple cavities including the inter-domain cleft. A clustering step enabled us to identify the cavities consistently present during the trajectory (2ns). Two pockets were manually selected based on our knowledge on DBPs. They were both present on snapshot N°19 of the MD trajectory of apo-LieIFtrunc, and they have no equivalent on the mammalian protein (S3 Fig). This particular structure was considered for further analysis, and it will be denoted as apo-LieIFtrunc/MD. The first pocket was located at the beginning of the truncated protein (Fig 2). It was constituted by 17 residues [AA 27-33,38,41-42,46-50,53-54]. It had a volume of 132 Å3, and it will be referred to as P1. Although it had a small size, it was selected as a potential druggable pocket of LieIF for two reasons. First, P1 was spatially close to the divergent 25 N-terminal residues of LieIF that were responsible for the dominant negative phenotype of LieIF in yeast that leads to growth impairment. [9] Hence, it has a significant potential to create a binding site for Leishmania-specific compounds. The second reason was the fact that it contained the Q-motif [AA 45-53], which is an adenine recognition element with features common to all ATP-dependent helicases, and it may play the additional role of regulating ATP binding. [33,36] Moreover, the Q-motif was at the interface between the two domains constituting LieIF. Inter-domain interactions are known to be important in ATP binding and hydrolysis in many DBPs. [39,92] A small molecule that binds to P1 may interfere with inter-domain interactions by steric hindrance or by impeding conformational changes necessary for ATP binding and/or hydrolysis. These elements suggested that P1 could be a suitable specific inhibitory binding site in association with the non-conserved amino terminus of LieIF. The second pocket, referred to as P2, also was located on domain 1 (Fig 2) and was constituted by 33 residues [AA 104,107-108,111-112,115,125-137,139-140,142-144,154,156-157,160-161,164-167]. Thus, P2 contains residues from: (i) motif Ia [AA 102-107], (ii) motif Ib [AA 154-158], (iii) the variable loop containing the GG doublet [AA 133-134] and the THR135 residue, which is the only phosphorylation site known for L. infantum LieIF [93], and (iv) a sequence insertion that is specific to LieIF [AA 125-127]. The conserved motifs Ia, Ib and the doublet GG are implicated in RNA binding in the DBPs. [15,37,39,40,94] The sequence insertion that is unique to LieIF may increase the probability of identifying selective molecules, as has been the case for other Leishmania proteins. [95] In fact, residues [S125, K126, F127] of this insertion constitute a part of a long loop connecting the α-helices containing motifs Ia and Ib. Residues corresponding to this loop in other DBPs have different secondary structures depending on the protein. For example, DDX2A (3EIQ), yeast eIF4A (1FUU), DDX48 (2XB2) and Vasa (2DB3) present two β-sheets separated by an α-helix, while DDX19 (3G0H, 3FMO) and yeast DBP5 (2KBE) present only two β-sheets separated by a long variable loop at this particular region. Thus relevant differences were observed on the structural and sequence levels for different DBPs around residues corresponding to P2 on LieIF. In addition, the tunnel-shape of P2 and its size (364 Å3) were among the properties that led to its selection as a promising and potentially specific druggable pocket. In order to select for potential LieIF ATPase inhibitors, we proceeded to a virtual screening of the CN library targeting pockets P1 and P2. Dock was used for docking calculations. Then, a clustering step using Self-Organizing Maps (SOMs) was performed on the VS results. This step permitted us to identify clusters of consensual docking poses. To reduce the number of compounds within those clusters, we used other filters according to the pocket and based on drug likeness properties, low (favorable) energy of interaction with the protein, pose geometry or chemical diversity. For pocket P1, only 19013 compounds were successfully docked out of the initial set (95494 compounds). The SOM analysis revealed a map with two low U-valued clusters presenting low (favorable) docking scores (S4 (a) and S4 (b) Fig). These clusters contained 2921 compounds that were mainly small with low molecular weights. This is essentially due to the relatively small size of the pocket (132 Å3). Since oral drugs against VL are highly recommended, we filtered them according to the Lipinski “Rule of Five” and to a geometrical sieve which reduced the set down to 783. We clustered them according to their chemical structures and ranked them according to their docking scores within each chemical cluster. From each cluster, the two best-scored molecules were selected, when more than one compound occurred. So, we retained a selection of 131 consensual, chemically-diverse drug-like molecules, correctly docked inside the pocket and well scored (according to Dock grid-based scores). For pocket P2, we also docked the CN molecules using Dock. The SOM analysis revealed three homogeneous clusters with low docking scores (S4 (c) and S4 (d) Fig) that contained 12408 compounds (SET1). This represented a larger number of docked molecules as compared to P1, and it is due to the large size of this pocket. As P2 contained the THR135, a phosphorylated site of an amastigote version of the LieIF protein, we investigated the effect of such a post-translational modification on the docking of these molecules. We docked the molecules of SET1 on the phosphorylated form of LieIF and only 6712 (∼ 54%) molecules were successfully docked (SET2). A shift to positive docking scores was observed compared to the docking on the non-phosphorylated P2 (S5 Fig). This indicated a global negative impact of the phosphate group on the docking results. SET2 contained a large number of molecules to be tested in vitro. In order to optimize the chance to select relevant molecules interacting with both forms of the protein, we performed a third docking calculation targeting the non-phosphorylated P2. A second docking algorithm was used (ADvina) to screen the CN. It uses a different searching algorithm and a different scoring function compared to Dock. This would permit us to perform a selection with no algorithm-related bias. The SOM analysis revealed a map with three homogeneous clusters (S4 (e) and S4 (f) Fig) containing 12298 compounds (SET3). The intersection between SET2 and SET3 contained 155 molecules. Through the geometric filter, we eliminated 11 molecules. Thus, the remaining 144 compounds constituted a set of molecules with consensual docking poses according to two different searching algorithms and presenting good docking scores according to two different scoring functions. For the sake of diversity, two additional sets of molecules not included in the intersection (SET2 ∩ SET3) were constituted. Fifteen molecules exclusively docked with Dock, with low, favorable docking scores and passing the geometrical filter were selected (SET1 but not SET2 or SET3). Similarly, fifteen molecules docked exclusively with ADvina with low docking scores and good poses were selected (SET3 but not SET1 or SET2). Thus, 174 molecules were selected for P2 as potential hits. This final set contained 144 consensually docked molecules through the three VSs and 30 molecules chosen for their best docking scores (Dock or ADvina scores). Finally, 305 molecules (screened against P1 and P2) were selected and purchased at the French Academic Compound Library [66,67] for experimental validation. In order to select the compounds that will experimentally inhibit the ATPase activity of LieIF, we established screening assays where both the purified recombinant LieIF and eIF4AIMus were tested for their ATPase activity in the presence of commercially-available, total yeast RNA with a colorimetric assay based on molybdate Malachite Green that measures the free phosphate released. [9] We performed the screens monitoring the ATPase efficiency in the presence of 500 μM of the compounds in 96 well plates in three independent experiments. We used this concentration in order to enhance the chances of observing inhibition or stimulation of the compounds because we used a relatively high protein concentration in the assays (around 1 μM); high protein concentrations were needed because of the relatively weak RNA-dependent ATPase activities of eIF4A-like proteins. [36] Through the biochemical screen, we calculated the percentage of inhibition of each compound and we detected four signals of inhibition of LieIF corresponding to structurally unrelated molecules; two docked on P1 and two on P2. We show the results for one representative screening plate (Fig 3), where the Z’-score of 0. 76 confirmed the quality of the screen. Due to the lack of sufficient amounts of three of the compounds, only compound 208 was further characterized. We also obtained higher amounts of this compound from the corresponding chemists to be able to proceed with further enzymatic and biological experiments. Compound 208, an epimeric mixture (α/β: 84%/16%) called 6-α/β- aminocholestanol (Fig 4 (a) ) [96], was identified within the set of compounds that successfully docked on pocket P2 by both Dock and ADvina. In order to characterize the effect of the compound on the ATPase activity, we performed time courses for the ATPase activity of LieIF and eIF4AIMus (∼ 1μM) at different concentrations of the compound in the 0 to 1 mM range, in the presence of 1 mM ATP and saturating concentrations of RNA. The amount of ATP hydrolyzed for both proteins increased in a time-dependent manner and the corresponding ATPase reaction rates for each compound concentration were determined and plotted (Fig 4 (b-i) ). The relative reaction velocity of the ATPase activity decreased in the presence of increasing amounts of the compound in a dose-dependent manner. The IC50 values were interpolated for the inhibition of the ATPase activity of 1 μM of LieIF and eIF4AIMus, and we obtained IC50 values of 150 ± 15 μM and 115 ± 25 μM, respectively (Fig 4 (b-i) ). For comparison, the Km reported for ATP in similar reactions were higher than our IC50 values (350 ± 120 μM for LieIF and 250 ± 90 μM for yeast eIF4A). [9] Next, as a proof of concept, we identified chemical analogues of compound 208 to test their effect on LieIF. Nine commercially available analogues of 208 were identified and purchased (Sigma Aldrich, S6 Fig). Moreover, the 6-α-aminocholestanol (20) could be obtained from the chemists that provided us with the 208 epimeric mixture (S6 Fig). All ten molecules were screened at 500 μM for their effects on the ATPase activity of LieIF and eIF4AIMus. Two molecules (20 and 48) demonstrated inhibition of the ATPase activity of LieIF and eIF4AIMus (Fig 4 (b-ii) and 4 (b-iii) ). Structures of compounds 20 (6-α-aminocholestanol) and 48 (6-ketocholestanol) are shown in Fig 4 (a). Therefore, we further characterized these molecules as we did for 208 using time course experiments testing different compound concentrations in the 0-1 mM range: 0,100,200,400 and 600 μM for 20; and 0,200,400,600,800 and 1000 μM for 48. We also determined relative reaction velocities and interpolated IC50 values. Compound 20 showed a comparable activity to 208 with IC50 values of 160 ± 25 μM and 185 ± 25 μM for LieIF and eIF4AIMus respectively (Fig 4 (b-ii) ). On the other hand, compound 48 showed a lower activity. It inhibited the ATPase activity of 1 μM of LieIF and eIF4AIMus with IC50 values up to 1 mM (Fig 4 (b-iii) ). All compounds presented different kinetic properties according to the proteins (Fig 4). The least effective compound (6-ketocholestanol) presents a ketone group replacing the amino group on carbon C6 as compared to 208 and 20 (Fig 4 (a) ). Interestingly, all 8 non-active analogues lacked this group at this possition. Even a nitro group could not ensure activity (S6 Fig). Thus, the amino group appeared important for the inhibitory activity. In order to gain insights into the potential binding modes and affinities of the three hits on LieIF, we performed further docking calculations targeting their plausible binding site, pocket P2. Epimers of the mixture 208 were considered separately, as the 6-α-aminocholestanol, represented by compound 20, and the 6-β-aminocholestanol. Docking scores and estimated Ki values were obtained for each hit. Different binding modes were obtained for the three molecules (S7 Fig). As a special interest in the amino/ketone group on carbon C6 arose through the ATPase assays, we investigated its potential interactions with the pocket residues. Compound 20 appeared to be the most potent compound according to its estimated Ki (221. 6 nM) and the free energy of its binding (-9. 1 kcal/mol) to LieIF (Table 2). Noticeably, the amino group of compound 20 interacted with the phosphorylation site of LieIF (T135) through H-bonds (Table 2, Fig 5), while the ketone group of compound 48 established hydrophobic interactions with T135 (Fig 5). All three hits also were docked on phos-LieIF and exhibited either higher or comparable docking scores and Ki estimations to those we obtained with the non-phosphorylated form of LieIF. Docking poses on both forms of LieIF presented significantly different interactions between the compounds and the protein residues, suggesting an important impact of the phosphate group on T135 on the pocket properties, which would directly impact the binding of the inhibitors. This confirmed our initial interest in the phosphorylated form of the pocket P2 as a differentiating target in the virtual screening. The best interactions were predicted with the non-phosphorylated form of LieIF, and the highest affinity with compound 20 (6-α-aminocholestanol). Many residues interacting with compound 20 were non-conserved in the mammalian eIF4AI. We performed the same docking calculations targeting a site equivalent to P2 on the mammalian eIF4AI structure, but pockets were different/absent in the site region (S3 Fig). A significant shift to higher docking scores was obtained (Table 2) as compared to LieIF. As no equivalent pocket was detected on the mammalian eIF4AI and taking into consideration the qualitative differences between the kinetics obtained with LieIF and eIF4AIMus (Fig 4), we hypothesized that the binding modes and sites of the inhibitors differ between both proteins. In order to confirm that LieIF inhibitors also had an effect on the parasite viability, we assessed the effect of compound 208 and its 10 analogues on the viability of L. infantum promastigotes at the stationary phase using an MTT assay after a 24h exposure. Only the three hits, already identified as inhibitors of the ATPase activity of LieIF, affected the promastigote viability in a dose-dependent manner (S8 Fig). IC50 values of 4. 1 μM, 3. 6 μM and 39. 1 μM were obtained for compounds 208,20 and 48, respectively (Fig 6 (a) ). The remaining eight compounds tested at different concentrations within the range of 0-100 μM did not show inhibitory effects on parasite viability. The results observed at 100 μM with these compounds are reported (S2 Table). The next step was to assess the effect of different concentrations of the three compounds on macrophages derived from THP-1 cells by PMA activation, as commonly used for drug testing, [97] by using an MTT assay. All three compounds showed no significant toxicity on the THP-1 macrophages (S8 Fig). The viability of the macrophages treated with each compound was around 90–100% at the concentration corresponding to the IC50 on the promastigotes (S8 Fig). The CC50 values were determined by interpolation as 43. 4 μM, 35. 2 μM and 81. 4 μM for 208,20 and 48, respectively. A positive correlation was observed between these CC50 values and the IC50 values obtained for the promastigotes viability. Their selectivity was thus illustrated by selectivity indexes that varied accordingly from 10. 6 to 2. 1 (Table 3). These selectivity values illustrated different effects on the parasite and host cell. So, we further evaluated the effect of the three compounds on the intracellular amastigote forms of L. infantum parasites. Upon L. infantum infection, THP-1 cell-derived macrophages were incubated in the presence of different compound concentrations (1 μM, 3 μM and 5 μM for 20 and 208; 3 μM, 5 μM and 25 μM for compound 48) for an additional 24h. The number of intracellular amastigotes and infected cells were then counted. An infection index was also calculated that integrated both parameters (Fig 6 (d) ). Untreated control cells harbored a rate of 65. 1% of infected cells and a mean amastigote number per cell of 5. 5, which corresponds to an infection index of 355. 6. In the presence of the compounds, the number of intracellular amastigotes drastically decreased upon the 24h exposure at the different concentrations tested, in a dose-dependent manner. All three compounds also had an effect on the number of infected cells in a dose-dependent manner (S3 Table), which thus reflected on the infection index. There was also a positive correlation between the numbers of infected cells and intracellular amastigotes. Based on the amastigote numbers, we determined the IC50 values of the three hits (0. 9–4. 2 μM, Table 3), which also were positively correlated to the values measured for the promastigotes and in the toxicity assays. Thus, selectivity indexes (SI) measured for the amastigotes were 31. 7,37. 5 and 19. 3 for compound 208,20 and 48, respectively. Based on all results, the 6-α-aminocholestanol appeared as the more potent hit notably with a selectivity index of 37. 5. The identification of novel drugs or targets constitutes a research priority for the treatment of leishmaniases. [7] Different criteria need to be fulfilled in order to validate a target including its absence in the host cell or the occurrence of substantial differences between the host and Leishmania proteins, its essentiality (demonstrated genetically or chemically), its expression in relevant stages (amastigotes in case of Leishmania), the presence of small molecules binding cavities, and its assayability for high throughput screening assays. [8,98,99] By these criteria, a list of relevant enzymes involved in metabolism, pathways or other cellular mechanisms have received attention, ranging from parasite-specific proteins to highly conserved proteins that have unique structural features impacting the protein function. [8,95] One such example is the elongation factor 1-α, which was shown to be a relevant target despite its 82% identity with the mammalian orthologue. [95] Inhibitors selective to the Leishmania protein could be identified that targeted a structural feature unique to the parasite protein. [95] This also demonstrated the feasibility of targeting highly conserved proteins and the relevance of using virtual screenings as a cost effective approach in identifying novel inhibitors and leishmanicidal molecules. The L. infantum translation initiation factor eIF4A (LieIF) was selected in this study by taking into account a range of evidence that hypothesized that it could be a novel candidate target. The study aimed at selecting inhibitors of this protein that subsequently affect Leishmania parasites viability by using a virtual screening process for the identification of compounds interacting with the protein combined with biochemical screening and with the biological characterization of the inhibitors. Computational approaches have the advantage of reducing the number of compounds to be screened in in vitro assays and thereby the costs of chemicals and the global screening procedure. Screening methods depend on the targets and pockets, the compound library and docking method used and on the strategy for effective selection of docked compounds. Of the 305 compounds selected by the strategy adopted here, four inhibitory signals were detected, corresponding to structurally different molecules. Only one compound was further characterized and confirmed as an ATPase inhibitor of LieIF and eIF4AIMus, and was used as a basis for identifying additional active analogues. This original description of LieIF inhibitor series brings further evidence on the druggability of eIF4A proteins. [49] EIF4A is the prototype of the DEAD box protein family, where members present a characteristic structural fold with the occurrence of 11 conserved motifs involved in the biochemical activity of these proteins. [16,33] There is 50-53% identity across yeast, human and Leishmania eIF4As. Availability of a range of crystal structures of human and yeast eIF4A and other DEAD box proteins in the presence or absence of different ligands, [34,37,38,40] facilitated comparative modeling of the Leishmania protein. The structure models of LieIF, presented herein, had the characteristic dumbbell shape in both bound and unbound states. Noticeably, the presence of druggable pockets in the NH2-terminus domain, identified as specific to LieIF versus the human eIF4AI, pointed to the relevance of primary sequence diversity. Biochemical analysis of LieIF highlighted significant differences in reactions’ requirements and substrate affinities between LieIF and yeast protein. [9] These differences extend to the mammalian eIF4A, as confirmed here, and infer different enzymatic properties of the eIF4A orthologues. The results obtained with the three compounds (208,20 and 48) also indicated different kinetic properties between LieIF and eIF4AIMus. In fact, LieIF does not complement for the loss of yeast eIF4A in spite of its ability to bind in vitro to yeast eIF4G, the molecular scaffold of the eIF4F complex. In contrast, it does exert a dominant-negative phenotype in yeast resulting in growth reduction, indicating a non-productive interaction with the translation machinery. Importantly, deletion of the 25 NH2 residues of LieIF abolishes the dominant-negative phenotype and yields normal growth, yet without allowing complementation. This suggests significantly different molecular mechanisms and interactions across species. [9] In line with such observations, sequence divergence across species is more important in the NH2-terminal part of the protein including a Leishmania-specific insertion within a poorly conserved region (Fig 1). [9] This insertion is included in the P2 pocket, the putative RNA binding site against which we selected the compound 208. LieIF has interdependent ATPase and RNA helicase activities. Notably, we confirmed its assayability and established a simple RNA-dependent ATPase assay that uses the malachite green to measure the amount of Pi released. [9] Herein, it was adapted to fit 96-well microtiter plates, and statistical evaluation provided robust Z’-scores (> 0. 5) indicating that the assay is reliable. The screening assay used 500 μM compound concentrations, justified by the high protein amount engaged (1 μM) as eIF4A activities in vitro are poor. [9,100] Under these conditions, compound 208 selected against the P2 pocket showed efficient ATPase inhibition (90%), and presented IC50 values lower than the Km value for ATP. Actually, IC50 measures depend on reaction conditions notably the amount of protein (1 μM) and substrate (1 mM) engaged in the reaction, so here it corresponds to 150-fold excess over the protein. To our surprise, the compound also reacted with eIF4AIMus with a comparable IC50 value (115 ± 25 μM), but with different kinetic properties. One explanation could be the occurrence of a different binding site on the mammalian eIF4AI, as supported by the modeling and docking results. The present work leaves open questions on the inhibition mode and the interaction of the compound with the proteins. Work is in progress to address these questions. To ascertain the interest of 208,10 structurally related analogues were selected and tested on both proteins. Two compounds, 20 and 48, inhibited eIF4AIMus and LieIF with different efficiencies. None of the other eight compounds were shown to be active against the two proteins. Similar IC50 values were obtained for each protein but as seen with 208 the kinetic properties were different according to the protein. Importantly, none of the eight inactive compounds had an effect on Leishmania promastigotes, the extracellular form of the parasite. However, all ATPase inhibitors (208,20 and 48) negatively impacted viability of the promastigotes with low IC50 values (4. 1,3. 6 and 39. 1 μM, respectively) that were positively correlated with those determined for LieIF ATPase assays (150,160 μM and > 1mM, respectively). The three compounds presented CC50 values on macrophage cells that reflected the potency of the ATPase inhibitors but at a higher concentration range than on the parasite (43. 4,35. 2 and 81. 4 μM, respectively), indicating a more potent effect on the parasite than on the host cells. These compounds also similarly reacted on the intracellular amastigotes (1. 4,0. 9 and 4. 2 μM respectively) and demonstrated even better selectivity indexes (37. 5–19. 3) than with the promastigotes (10. 6–2. 1) as expected in drug screening campaigns. [101] The difference between the IC50 values on the enzymes (150 - > 1000 μM) and those on the parasite (< 1–40 μM) could be explained by the fact that the amounts of protein used in the assay are far above physiological concentrations determined in Leishmania. [19] In addition, the activity is measured on the proteins as single units but eIF4A is a member of a multimeric complex. Notably, it is well known that the activity of eIF4A is enhanced by cofactors, and it can reach 20-fold increase upon association with its partner proteins, such as the components of the pre-initiation complex [102–104] or even under molecular crowding. [105] In addition, the study did not investigate effects on the RNA helicase, the other enzymatic activity ensured by these proteins. So, the effect of these compounds could be more pronounced or more selective on this activity. Our hits consisted of amine and ketone cholestanol scaffolds. Compound 208 is an epimeric mixture of 6-α/β-aminocholestanol. Compound 20 is the α-epimer and compound 48 is the 6-ketocholestanol. Far less effective, compound 48 presents a ketone group replacing the amino group on carbon C6 on 208 and 20 (Fig 4 (a) ). The amino group is also absent on the eight analogues inactive on the ATPase activity. Even the addition of reactive chemical groups at the same position could not ensure activity (S6 Fig). This indicates the importance of the amino group in the protein-inhibitor interactions. Docking of the three hits on P2 supported this hypothesis, and it is in line with experimental results predicting better and more efficient interactions of the 6-α-aminocholestanol with P2 residues, as compared to the β-epimer or the 6-ketocholestanol. These aminocholestanols were described as anti-fungal molecules that reduced yeast growth at low micromolar concentrations (∼ 31 μM), presumably by targeting ergosterol synthesis. [96,106] Sterol derivatives that interfere with ergosterol biosynthesis, and presenting a chemical relatedness to our hits, were described for their leishmanicidal activities. [107] Noticeably, sterol derivatives such as the 7-α/β-aminocholesterol reduced by 59% the number of intracellular L. donovani (another VL agent) amastigotes at 1. 94 μM concentration, but it demonstrated a low selectivity index (∼ 3). [107] Its structure presents a double bond on the second ring, which confers a local planar 3D shape to this molecule, compared to our hits, in addition to a displacement of the amino group on carbon C7. This study also hypothesized that this aminocholesterol could target ergosterols biosynthesis, but with no experimental evidence provided. [107] Our results do not permit us to exclude interactions with other targets, but there is clear biochemical evidence for the interaction of the cholestanol-based inhibitors (virtually selected without prior reference to literature) with LieIF, and there is a positive correlation between the potencies of enzyme inhibition and leishmanicidal effects of the three molecules. These compounds also bear a distant similarity to hippuristanol, a selective inhibitor of the mammalian eIF4A, [108] thought to act as an allosteric inhibitor of RNA binding in the C-terminal domain of eIF4A. [108–110] It inhibits eIF4A helicase activity by blocking the protein in the closed conformation, [111] and it is unable to affect the activity of other DBPs like human DDX19 and DDX52. [108] No evidence is available on its effect on LieIF or on Leishmania. This or other eIF4A inhibitors will need to be tested on LieIF and their cidal effects assessed on Leishmania. The role of eIF4A proteins is pivotal as an essential enzyme of the eIF4F translation initiation complex. [17,18] Its essentiality has been genetically confirmed in yeast, [31] mammals [112] and in Trypanosoma brucei, another kinetoplastid parasite. [19] However, as RNAi is not applicable in the Leishmania subgenus, and the gene is organized as a cluster of two identical tandem copies on the likely polyploid chromosome 1, [9,113] genetic confirmation of its essentiality is difficult. With the advent of CRISP-Cas9 technology, strategies may be deployed to confirm the essentiality of LieIF in L. infantum [114,115] and to assess the biological relevance of the interactions of the inhibitors with LieIF as has been shown for rocaglates and eIF4AI. [116] This study constitutes a first step towards validation of LieIF as a drug target. It delivers novel eIF4A inhibitors. As shown here, the 6-α-aminocholestanol with IC50 value lower than 1 μM on intracellular amastigotes, little toxicity and a selectivity index higher than 20, constitutes a promising anti-Leishmania molecule that deserves further investigation.

Leishmaniases constitute a group of neglected parasitic diseases that inflict major burden on public health. Novel drugs and targets need to be identified since current therapies have adverse side effects. Herein, we focused on Leishmania infantum translation initiation factor 4A (LieIF), as a potential drug target. LieIF, a pivotal enzyme in the translation machinery, is also implicated in host-pathogen interactions. We modeled its 3D structure and identified two pockets, which were used in virtual screenings of a chemical compound library. Therefore, we selected and purchased 305 compounds. We established a reliable ATPase screening assay to test the molecules against the enzymatic activity of LieIF and its mammalian homologue. A promising hit was retained and further characterized. It inhibited both proteins but showed different kinetic properties. It was used as a basis to identify similar analogues and two additional inhibitors were identified. All three hits reduced the viability of the extracellular promastigote form of the parasite, but they had no significant cytotoxic effects on host cells. They also affected the viability of the intracellular amastigote form and reduced the macrophage infection. This selectivity is very promising and indicates that these inhibitors would constitute an avenue to develop strategies to fight leishmaniases.

Abstract Introduction Materials and methods Results Discussion

blood cells medicine and health sciences immune cells enzymes immunology enzymology microbiology parasitic diseases protozoan life cycles parasitic protozoans phosphatases developmental biology protozoans leishmania rna helicases sequence motif analysis promastigotes research and analysis methods sequence analysis white blood cells bioinformatics proteins animal cells life cycles amastigotes adenosine triphosphatase biochemistry leishmania infantum eukaryota helicases cell biology database and informatics methods biology and life sciences protozoology cellular types macrophages organisms

2018

Identification of novel leishmanicidal molecules by virtual and biochemical screenings targeting Leishmania eukaryotic translation initiation factor 4A

Biofilms are communities of bacteria that grow encased in an extracellular matrix that often contains proteins. The spatial organization and the molecular interactions between matrix scaffold proteins remain in most cases largely unknown. Here, we report that Bap protein of Staphylococcus aureus self-assembles into functional amyloid aggregates to build the biofilm matrix in response to environmental conditions. Specifically, Bap is processed and fragments containing at least the N-terminus of the protein become aggregation-prone and self-assemble into amyloid-like structures under acidic pHs and low concentrations of calcium. The molten globule-like state of Bap fragments is stabilized upon binding of the cation, hindering its self-assembly into amyloid fibers. These findings define a dual function for Bap, first as a sensor and then as a scaffold protein to promote biofilm development under specific environmental conditions. Since the pH-driven multicellular behavior mediated by Bap occurs in coagulase-negative staphylococci and many other bacteria exploit Bap-like proteins to build a biofilm matrix, the mechanism of amyloid-like aggregation described here may be widespread among pathogenic bacteria. Biofilm formation is universal for all bacteria. The molecular mechanisms governing this process vary among bacteria, but they all culminate in the synthesis of an extracellular matrix. The composition of the extracellular matrix is complex and variable, even within the same bacterial species when environmental conditions are altered [1,2]. However, one common principle is that the matrix scaffold is built from exopolysaccharide or proteins, which eventually can be interwoven with extracellular genomic DNA [3–5]. The reasons underlying the election of a polysaccharide or protein-based biofilm matrix are not well understood, but an increasing number of studies indicate that proteinaceous scaffolds are more common than previously anticipated. Proteins anchored to the bacterial cell surface can assemble the matrix scaffold through homophilic interactions between identical molecules expressed on neighboring cells or through heterophilic interactions with other surface proteins or with non-proteinaceous cell wall structures [6,7]. Members of this group of proteins include autotransporter adhesins [8–11], carbohydrate-binding proteins [12–14], and cell-wall anchored proteins covalently linked to the peptidoglycan (CWA) [2,15–21]. Another strategy by which proteins can contribute to the formation of the matrix scaffold is through polymerization into functional amyloid fibers. Secreted proteins can assemble to form insoluble fibers with a characteristic cross-β-strand structure, where the β-sheets run perpendicular to the fibril axis [22]. Once polymerized, amyloid fibers constitute a strong platform able to mediate interactions between the neighboring cells and surfaces [23–26]. Examples of amyloid fibers mediating biofilm development include curli pili present in Enterobacteriaceae [27,28], FapC in Pseudomonas fluorescens [29], TasA in Bacillus subtilis [30], the aggregative flexible pili named MTP in the pathogen Mycobacterium tuberculosis [31,32] and phenol soluble modulins (PSMs) in Staphylococcus aureus [33]. Biofilm associated proteins (Bap) are high molecular weight multi-domain proteins, characterized by a repetitive structure and localized at the cell surface [34]. The first member of this family of proteins was identified in a mobile pathogenicity island (SaPIbov2) present in some strains of S. aureus. So far, the bap gene has been identified in mastitis-derived staphylococcal species, but has never been found in S. aureus human isolates. However, bap orthologous genes are present in the core genome of several coagulase-negative staphylococcal species that belong to the human commensal microbiota such as S. saprophyticus ATCC15305 (Accession number GCA_000010125. 1), S. epidermidis (GCA_000759555. 1) and S. warneri SG1 (GCA_000332735. 1) [35]. Bap promotes the initial attachment to inert surfaces and cell-to-cell interactions through a mechanism that is independent of exopolysaccharide [21,36]. During infection, Bap facilitates the persistence in the mammary gland by enhancing adhesion to epithelial cells and prevents cellular internalization through the binding to GP96 host receptor, which interferes with the FnBPs mediated invasion pathway [37,38]. Overall these results indicated that Bap plays a dual function: on the one hand, mediating bacterial-bacterial interactions and on the other, bacterial-host interactions. However, the molecular mechanisms by which Bap performs these functions and the region of the protein involved in each process remain unexplored. In this report, we investigated the mechanistic basis by which Bap proteins promote the formation of the biofilm matrix scaffold. Our results have shown that Bap is constitutively expressed along the growth curve and processed. The resulting fragments, which likely contain mainly the N-terminal region of the protein, form insoluble amyloid–like aggregates when the pH of the media becomes acidic and the concentration of calcium is low. If calcium concentration increases, metal-coordinated Bap adopts a more stable conformation as shown by thermal denaturation monitored by intrinsic fluorescence, nuclear magnetic resonance (NMR), proteinase K digestion and analytical ultracentrifugation. As a consequence, the N-terminal region is unable to self-assemble and to mediate intercellular aggregation and biofilm formation. Furthermore, we show that biofilm assembly by Bap orthologs also depends on the critical N-terminal domain suggesting that the mechanism of biofilm assemblage is conserved in staphylococci. In view of these results, we propose that Bap plays dual role in the bacterial physiology, acting as a sensor and promoting biofilm formation, a configuration that has not hitherto been described for any component involved in biofilm formation. To investigate the molecular mechanisms underlying Bap-mediated staphylococcal biofilm development, we monitored the expression of Bap in rich liquid media (LB-glu) along the growth curve using native and denaturing gel electrophoresis. Western immunoblotting under denaturing conditions revealed the presence of Bap from early stages of growth until the population entered stationary phase (OD600nm = 5). From that point, the levels of Bap decreased significantly in denaturing gels, whereas a band of high molecular weight appeared in the native gels suggesting that Bap formed aggregates when bacteria entered stationary growth phase (Fig 1A). When S. aureus is grown in a media containing glucose, the entry in stationary phase is accompanied by a decrease in pH due to the accumulation of acidic byproducts from glucose fermentation [39]. We therefore investigated whether Bap aggregation and Bap mediated biofilm development were related and occurred in response to changes in the media pH. To investigate this hypothesis, bacteria were grown in LB-glu, where the pH levels dropped below 5 when bacteria reached stationary phase (OD600nm = 5) and in LB without glucose, where the pH remained neutral all along the growth curve (Fig 1B). Western immunoblotting revealed that Bap failed to form protein aggregates when bacteria grew in LB (Fig 1C). Moreover, results showed a strong correlation between Bap protein aggregates and Bap mediated biofilm formation since bacteria grown in LB-glu (pH<5) formed bacterial clumps and strong biofilms in microtiter plates whereas in LB media (neutral pH) Bap failed to promote bacterial clumping and biofilm development (Fig 1D and S1A Fig). To further corroborate the effect of pH on aggregation of Bap-positive strains we evaluated cell clumping of S. aureus V329 and ∆bap grown in LB medium acidified with 0. 1 M HCl to pH 4. 5. After an overnight incubation V329 wild type strain clearly showed a biofilm adhered to the microtiter plate and bacterial clumps at the bottom of the tube, while ∆bap strain did not (S2A and S2B Fig). Moreover, the two strains were also grown in LB-glu and, after an overnight incubation the medium was replaced by LB to evaluate the possible disassembly of bacterial aggregates. Indeed, no bacterial clumps and no biofilm were observed after media were exchanged indicating that the process of interbacterial interaction mediated by Bap is reversible when pH arises (S2C and S2D Fig). Together, these results suggest that acidification of the growth media promotes Bap aggregation and biofilm development. If Bap is engaged in homophilic interactions during biofilm development, Bap aggregates should be composed primarily of the Bap protein. In contrast, if Bap mediates heterophilic interactions with other surface proteins, Bap aggregates should also contain additional proteins. To distinguish between these possibilities, we determined the protein content of the Bap aggregates by recovering the insoluble protein material retained within the wells of the stacking gel from preparations of S. aureus grown in LB-glu and LB and analyzing their identity by mass spectrometry (MS). MS analysis of the material retained in gel pockets from preparations of S. aureus grown in LB-glu identified peptides that corresponded mostly to the Bap protein strongly suggesting that this polypeptide is the main constituent of the aggregates (S3 Table). Apart from Bap and some ribosomal proteins, the vast majority of the other identified peptides corresponded to proteins that were also detected by MS analysis from preparations of S. aureus grown in LB medium, where no presence of Bap was observed (S3 Table). We also discarded the possibility that additional matrix molecules such as PNAG or eDNA could be involved in the formation of the Bap aggregates since neither Bap insoluble aggregates nor Bap-mediated biofilms were affected by the treatment with dispersin B (DspB) and DNase I (S3A, S3B and S3C Fig). MS analysis also showed that the identified peptides covered the N-terminal sequence of mature Bap almost completely (amino acid 49 to 819), but only a single short peptide from C repeats region (Fig 2A and 2B) was observed. These results suggest that the insoluble aggregates likely contain Bap fragments that at least include N-terminal region. To further investigate the mechanism of Bap proteolytic cleavage, we performed western immunoblotting of cell wall extracts of S. aureus V329 grown in LB-glu and LB culture conditions. Results revealed that Bap was proteolytically processed in both media (Fig 2D), but the cleavage products obtained in LB were unable to form high molecular weight aggregates (Fig 2C). Western immunoblotting of surface proteins from cells grown in LB-glu extracted at different points of the growth curve showed the presence of degradation bands that increased in number and intensity as bacteria grew (S4A Fig). Mass spectrometry analysis of the largest processed band confirmed that it corresponded to a degradation product of Bap containing at least the N-terminal region of the protein (S4B Fig). Interestingly, when bacteria entered stationary phase (OD600nm≈4, pH<5) bands corresponding to full-length and the resulting processed fragments of Bap disappeared from the gel and insoluble aggregates recognized by anti-Bap antibody were readily detectable (S4B Fig). With the aim to identify the extracellular proteases responsible for the proteolytic processing of Bap, we constructed mutants in 3 extracellular proteases: a serine protease (V8 protease; SspA), a cysteine protease (SspB) and its specific inhibitor (SspC) and a metalloprotease (aureolysin; Aur). The resulting protease-deficient strains showed similar Bap cleavage patterns and formed cell clumps and biofilm at levels similar to wild-type strain (S4 Fig, left panels). Besides, addition to the culture media of protease inhibitors such as α2-macroglobulin, E-64 (cysteine protease inhibitor), PMSF (serine protease inhibitor) and the inhibitor Staphostatin A (ScpB) that specifically targets the extracellular cysteine protease ScpA, did not interfere with Bap-mediated aggregation and biofilm development (S4 Fig, right panels). Taking together these findings suggest that Bap is processed either by spontaneous cleavage or by the activity of a protease different from the ones tested here, or perhaps by the action of more than one protease. The resulting processed products are more aggregation prone and form the high molecular weight aggregates under acidic conditions. These latter observations lead us to consider that the N-terminal region of Bap may be sufficient to promote biofilm development. To assess this hypothesis, we generated chimeric proteins comprising different regions of Bap tagged with the 3xFLAG amino acid sequence and linked to the R domain of the clumping factor A (R-ClfA) containing the LPXDG motif (Fig 3A). Variants of Bap comprising domain A (Bap_A, amino acid residues 49 to 361), domain B (Bap_B, amino acid residues 362 to 819), or domain A and B (Bap_AB, amino acid residues 49 to 819) were cloned in pCN51 vector under the control of the Pcad-cadC promoter and expressed in S. aureus ∆bap. The expression of the whole Bap or the chimeric Bap proteins on the bacterial cell wall was verified by western-blot and immunofluorescence using strains deleted in their spa gene (V329 Δspa and ΔbapΔspa) to avoid unspecific antibody labeling of protein A through its union to the Fc fraction of immunoglobulins (S5 Fig). S. aureus producing Bap_AB or Bap_B formed huge cell-to-cell aggregates (Fig 3B) and robust biofilms on polystyrene (Fig 3C and S1B Fig) or on a glass surface under flow culture conditions (Fig 3D). In contrast, no cell clusters and biofilm development were found in S. aureus Bap_A and ClfA strains. The observation that domain B of Bap is sufficient to induce biofilm phenotype suggests that Bap_B functionality could be affected by the pH. Similarly to the Bap full-length protein, Bap_B formed high molecular weight aggregates when a culture of ∆bap strain expressing Bap_B reached stationary phase (Fig 3F). Accordingly, this strain formed biofilm when it was grown in LB-glu (pH<5), but not in LB (Fig 3E and S1C Fig), and showed bacterial clumping and biofilm formation when grown in LB acidified with 0. 1 M HCl (S2A and S2B Fig). Bacterial aggregates formed by ∆bap expressing Bap_B chimeric protein in LB-glu (pH<5) were disassembled when the medium was exchanged for LB (pH>7) (S2C Fig). Next, we explored whether Bap_B was sufficient to confer cell-to-cell interactions to naturally bap deficient strains: S. aureus MW2, S. aureus Newman and S. carnosus TM300. As shown in Fig 3G and S1D Fig, expression of Bap_B in these strains conferred strong bacterial clumping capacity after an overnight incubation in LB-glu. Taking together, these results indicated that the B domain of Bap (amino acids 362 to 819) is sufficient to bestow multicellular behavior under acidic culture conditions, similarly to the entire Bap protein. To get insights about the molecular mechanisms by which the N-terminal region of Bap mediates cell-to-cell interactions, we used a purified recombinant protein comprising exclusively the B region of Bap (rBap_B). Purified rBap_B formed a visible ring of protein adhered to the walls of the tube when incubated at acidic pH in a grade of pH from 3. 6 until 5 (S6A Fig). Interestingly, the process was reversible and rBap_B aggregates dissociated completely when the pH was raised to 7 (Fig 4A). To validate the functionality of rBap_B, we analyzed the capacity of rBap_B to restore bacterial clumping phenotype of S. aureus ∆bap. Exogenous addition of rBap_B protein (2 μM) induced bacterial clumping only when S. aureus ∆bap was grown under acidic culture conditions (Fig 4B). Next, we performed a biophysical characterization of the rBap_B domain. First, we determined the relative size of the aggregates by dynamic light scattering (DLS). The graphic in Fig 4C illustrates the size characterization (hydrodynamic radius R) of rBap_B in solution at different pH. In the table, the correlation between the diffusion coefficient (D) of each population and its corresponding radius is shown. It can be observed that pH 4. 4 is the condition at which rBap_B protein presented the highest hydrodynamic radius and the lowest D value (peak 2), as expected for aggregated particles that move slower that smaller particles, with a polydispersity percentage below 15% characteristic of monodispersed samples (Fig 4C). At pH 3, rBap_B presented protein populations with intermediate R values. At neutral pH the obtained peak showed a D that once substituted in the Svedberg equation, together with the previously obtained experimental sedimentation coefficient, buffer density and partial specific volume, corresponded to the monomer of the protein (Fig 4C). Far-UV circular dichroism spectra (CD) of rBap_B showed a moderate increase in β-sheet structure (+5%) when the pH was acidified, at the expense of the predominant non-regular secondary structure (Fig 4D and S4 Table). Next, we analyzed more in depth these β-sheet-rich rBap_B aggregates formed at acidic pH. We examined the amide I region of the Attenuated Total Reflectance–Fourier Transform Infrared spectroscopy (ATR-FTIR) spectrum (1700–1600 cm-1) of rBap_B ring assemblies. This region corresponds to the absorption of the carbonyl peptide bond group of the protein main chain and is a sensitive marker of the protein secondary structure. Deconvolution of the FTIR-absorbance spectra allowed us to assign the individual secondary structure elements and their relative contribution to the main absorbance signal. The FTIR spectra of rBap_B aggregates was dominated by β-sheet/β-turn components contributing >80% to the signal. In particular, the strong bands at 1628 and 1694 cm-1 were consistent with the presence of amyloid-like intermolecular β-sheet structure (Fig 5A). To assess whether the prevalent intermolecular β-sheet in the rings formed by rBap_B aggregates was amyloid-like in nature, we evaluated the binding of the aggregates to the amyloid diagnostic dyes Thioflavin-T (ThT), Congo Red (CR) and ProteoStat. The presence of rBap_B aggregates induced a 25-fold increase in ThT maximum fluorescence emission (Fig 5B). Interestingly, when fresh rBap_B was incubated at pH 4. 5 for 5 min it bound readily to ThT in a concentration dependent manner, indicating a fast assembly of rBap_B into ThT positive structures (S7A Fig). In contrast, no change in ThT fluorescence was observed when the protein was incubated at pH 7. 0, independent of the protein concentration assayed (S7B Fig). The fast assembly of rBap_B at pH 4. 5 was also evident from the strong increase in light scattering relative to the signal obtained at pH 7. 0. (S7C Fig). These early assemblies displayed a strong binding to the dye bis-ANS, evidencing the presence of hydrophobic patches exposed to solvent, which potentially might recruit rBap_B monomers into the aggregates and/or contact other cellular molecules through non-polar interactions (S7D Fig). In agreement with an amyloid-like conformation the absorbance of CR and its spectrum maximum red-shifted in the presence of rBap_B ring aggregates (Fig 5C). The absorbance of ProteoStat, a novel fluorescent dye able to stain specifically amyloid aggregates in vivo [40], showed a 20-fold increase in its fluorescence maximum at 550 nm (Fig 5D). Altogether, these data strongly suggest that the intermolecular β-sheet structures formed upon aggregation of rBap_B at pH 4. 5 posses an amyloid-like conformation. In amyloid-like aggregates, short sequence fragments usually promote and guide the formation of amyloid-like structures and become embedded in the inner core of the cross-β structure [41–43]. To identify the likely amyloidogenic regions in the series of Bap_B peptides previously identified by MS in the biofilm we used computational algorithms: AGGRESCAN [44], PASTA [45], WALTZ [46] and ZipperDB [47]. The predictions converged to indicate two Bap_B short sequence stretches as potentially amyloidogenic: TVGNIISNAG named as peptide I (aa 487 to 496), and GIFSYS named as peptide II (aa 579 to 584) (Fig 2B). We synthetized the two peptides and incubated them at 10 μM at pH 4. 5. Both peptide solutions formed an evident gel (S8 Fig), a property shared by many amyloidogenic peptides [48] as well as biofilm matrices [49]. Analysis of the structure of the two gels by transmission electron microscopy (TEM) indicated that they comprise fibrils with a typical amyloid morphology (Fig 6A and 6B), and bound to ThT with high affinity (Fig 5E). Taken together these data indicate that Bap_B contains at least two short regions with high amyloidogenic propensity. Analysis of these peptides using the RosettaDesign program [50] implemented in ZipperDB [51] rendered average interaction energies of -25. 0 and -25. 9 kcal/mol for peptide I and peptide II and shape complementarities between strands of 0. 87 and 0. 81, respectively. These parameters are compatible with these peptides being able to form steric-zippers that might contribute to N-terminal Bap amyloid assembly. Of course, we cannot discard that the presence of several additional or alternative short amino acid stretches with amyloidogenic tendencies in the sequence of B-domain of Bap would be required for the assembly of the complete domain into aggregated structures at acidic pH. We next determined by transmission electron microscopy the presence of amyloid fibers. Electron microscopy analysis of the aggregates formed by purified rBap_B at pH 4. 5, first revealed the presence of isolated fibers and fibers entangled in larger electron dense aggregates (Fig 6C). Similar fibers were detected in the surface of S. aureus Δbap when bacteria were grown in the presence of exogenously added rBap_B under acidic culture conditions (Fig 6D). We further analyzed the presence of fibers in wild-type S. aureus V329 grown under biofilm forming conditions. Consistent with all the findings obtained for the rBap_B domain, S. aureus V329 contained fibers (Fig 6E and 6F) that specifically reacted with gold-labelled anti-Bap antibody (Fig 6G and 6H). Finally, we determined the presence of amyloid fibers in the Bap-mediated biofilm by staining the extracellular matrix of S. aureus V329 grown in LB-glu medium with ProteoStat, a dye specific of amyloid fibers (S9A Fig) [40]. We also extracted from a gel native pocket the insoluble aggregated material of S. aureus V329 strain and stained it with ProteoStat. As shown in S9B Fig, the dye stained the protein aggregates formed by S. aureus V329 but not the Δbap strain. Also, we tested the effect of the compound (-) -epigallocatechine gallate (EGCG), known to exert an anti-amyloidogenic effect in the case of proteins involved in neurodegenerative diseases [52], on biofilm development by S. aureus V329 wild type strain. As a control, we used S. aureus 15981 strain that forms a PNAG-dependent biofilm. ECGC significantly disassembled biofilm formed by S. aureus V329, but not by 15981, at all concentrations tested (P<0. 001, n = 5) (S9C Fig). These results ratify the amyloidogenic nature of Bap assemblies and their role in biofilm formation. In order to investigate the biological relevance of Bap-dependent biofilm formation related to the amyloidogenic properties of domain B, we have analyzed the capacity of S. aureus V329 strain grown in LB-glu and LB media to adhere to bovine mammary epithelial (MAC-T) cell line. The results revealed that V329 strain adhered more efficiently (P<0. 01) to epithelial cells when bacteria were grown in LB-glu compared to LB (Fig 7A). Accordingly, the corresponding S. aureus Δbap mutant strain showed similar capacity to adhere to epithelial cells when grown in LB and LB-glu media. These results suggest that fibers formation would improve S. aureus adhesion to host cells. Then, we performed an experiment to evaluate the colonization ability of S. aureus V329 wild type and Δbap mutant using a mouse foreign body infection model. The reasoning is that S. aureus V329 wild type should have higher capacity to colonize and persist on catheters than the Δbap strain in the case that Bap mediates biofilm development in this specific environment. To test this hypothesis, sterile catheters were implanted and inoculated into the mice with 107 CFU of S. aureus V329 and Δbap strains grown overnight in LB-glu at 37°C. Enumeration of S. aureus cells attached to the catheters 4 days after infection showed slight but not significant differences between S. aureus V329 wild type and the Δbap mutant strains (Fig 7B). However, at 10 days post-infection, the number of recovered bacteria was significantly higher for the wild type strain (CFU 5. 8 x 105) compared to the bap mutant (P<0. 05) (Fig 7B). These results suggest that Bap-mediated biofilm is important for the persistence of S. aureus through an infection process and, since in S. aureus V329 biofilm development depends on Bap amyloid fibers, this would imply a key role of these structures in the colonization of indwelling medical devices in vivo. However, further studies using different mutant strains that express Bap proteins incapable of aggregate (mutated in the major amyloid sequence stretches required for fibrillation, or mutated in its N-terminal domain) are required. Bap-mediated multicellular behavior is inhibited in the presence of millimolar concentrations of calcium bound to the EF-hand domains present in the region B of Bap [36]. The question arises as how calcium and pH environmental signals reconcile to regulate Bap-mediated biofilm formation. To address this question, we investigated the aggregation kinetics of Bap when S. aureus V329 and ∆bap producing Bap_B were grown in LB-glu supplemented with 20 mM of CaCl2. The presence of calcium inhibited aggregation of the Bap protein (S10A and S10B Fig, left panels), as well as biofilm formation and bacterial clumping (S10C Fig) despite acidification of the growth media. On the other hand, the wild type V329 and the Δbap Bap_B strains mutated in their EF-hand 2–3 calcium binding motifs (ΔEF and Bap_B_ΔEF respectively) showed no disruption of either biofilm phenotype (S10C Fig) or protein aggregation (S10A and S10B Fig, right panels) in the presence of calcium. Moreover, S. aureus V329 strain was grown in LB-glu and, after an overnight incubation the medium was replaced by LB-glu containing 20 mM CaCl2 to evaluate the possible disassembly of formed biofilm. No disaggregation was observed after the medium was exchanged suggesting that Ca2+ inhibitory effect on Bap functionality might be relevant in steps prior to amyloid self-polymerization process (S2D Fig). Next, we evaluated the effect of Ca2+ on bacterial aggregation of Δbap strain induced by rBap_B under acidic culture conditions. We observed that interbacterial interactions did not occur when Δbap mutant strain exogenously complemented with rBap_B was incubated in the presence of calcium (Fig 8A). We also tested the effect of calcium on Bap amyloid formation by analyzing in vitro aggregation of rBap_B into ThT positive amyloid-like structures in the presence of the cation. Results showed that calcium significantly inhibited the formation of amyloid-like aggregates (Fig 8B). To determine whether the inhibitory effect of Bap-amyloid aggregation induced by calcium is due to a change in protein structure upon binding to the cation we used several biophysical approaches. For this, it is important to clarify that because fast aggregation of rBap_B at pH 4. 5 even at low protein concentration (0. 01 mg/ml) makes difficult the characterization of the conformational properties of the soluble monomers at this pH, we analyzed the biophysical properties of the Bap_B domain in the presence of calcium at neutral pH. First, 1 and 10 mM CaCl2 are sufficient to induce a concentration dependent increase in the ellipticity of the far-UV CD spectra of rBap_B (Fig 8C). Deconvolution of the spectra in the absence and in the presence of 1 mM Ca2+ indicated that the protein displays very similar secondary structure content in these conditions. The spectrum is dominated in both cases by disordered conformations, although a small reduction in the overall β-sheet content could be observed in the presence of the Ca2+ (the 10 mM CaCl2 spectrum could not be deconvoluted due to the strong increase in HT voltage below 200 nm in this condition). We next decided to monitor by near-UV CD and intrinsic fluorescence the overall tertiary structure of rBap_B in the presence or absence of Ca2+. Despite no significant impact of Ca2+ on the environment of rBap_B aromatic residues could be observed by near-UV CD (S11B Fig), the cation promotes a detectable increase in intrinsic fluorescence emission (S11C Fig), suggesting rearrangements in the tertiary context of the protein. To confirm the existence of a change in the aromatic residues environment, we performed thermal denaturation in the absence and in the presence of 1 and 10 mM CaCl2. Results indicated that the temperature at which the protein loses half of its intrinsic fluorescence, augmented in the presence of increasing concentrations of calcium (1 and 10 mM), suggesting that the cation exerts a global stabilizing effect on Bap conformation (Fig 8D). Further techniques support this idea. Analytical ultracentrifugation analysis revealed that, in the presence of calcium the rBap_B monomer exhibited a significantly higher sedimentation coefficient (s (20, w) ∼3. 4 versus ∼3. 0). Additionally, rBap_B protein showed a frictional ratio f/f0 = 1. 39, compatible with a slightly elongated protein, while in the absence of the cation the protein showed a frictional ration of 1. 66 indicating a more elongated and moderately asymmetric protein shape (Figs 8E and S12A). Size exclusion chromatography analysis supported this by revealing that rBap_B is eluted with retard in the presence of calcium (S12C Fig). Additionally, when we performed 1D-NMR of rBap_B in the presence or absence of Ca2+, we observed an increase in the number of peaks corresponding to the methyl (0. 5 ppm) and amide (8. 5 ppm) regions of the spectrum (S12D Fig). The broader line-widths observed in these regions in the absence of Ca2+ are in concordance with a molten globule that is semi stable and fluctuates between several conformations. The sharpening of the peaks when Ca2+ is present would then be indicative of protein ordering into a more stable state with a smaller hydrodynamic radius (S12D Fig). Finally, analysis of Bap accessibility to proteolytic degradation in the presence of calcium showed that rBap_B was readily hydrolyzed by proteinase K in the absence of calcium, whereas it was protected from proteinase K activity when the cation was present (Fig 8F). Together all these results are consistent with the idea that Bap protein adopts a transient molten globule-like state in the absence of calcium prior to amyloid formation that is stabilized upon calcium binding thus impeding amyloid assembly due to tertiary rearrangements of Bap conformation. Although orthologs of Bap exist in many coagulase-negative staphylococci, homology in region B is variable [35] (S13 Fig and S5 Table). Thus, we wondered whether Bap orthologs could also mediate multicellular behaviour by the generation of amyloid-like aggregates. We selected Bap_B of S. saprophyticus (Bap_Bsapro) as it shares an intermediate percentage of identity with Bap_B of S. aureus (58% identity over the entire length of the B domain). Expression of a chimeric protein containing Bap_Bsapro linked to R-ClfA in S. aureus ∆bap∆spa (S5 Fig), induced bacterial clumping under acidic culture conditions, but not under basic conditions (Fig 9A and 9B). Consistent with the presence of EF-hand domains, Bap_Bsapro was sensitive to the presence of calcium in the media (Fig 9B). As previously shown for rBap_B, purified rBap_Bsapro formed precipitated protein aggregates in acidic phosphate-citrate buffer (pH 4. 5) that reversibly disassembled after raising the pH to neutral (S6B Fig). Together these results indicate that Bap_B domain of S. saprophyticus mediates multicellular behavior under acidic culture conditions, analogous to Bap_B domain of S. aureus. Biophysical characterization of the rings formed by rBap_Bsapro at pH 4. 5 indicated that they possess clear amyloid-like features, displaying strong binding to ThT (Fig 9C), ProteoStat (Fig 9D) and CR (Fig 9E). As for rBap_B, light scattering and bis-ANS binding assays demonstrated that rBap_Bsapro self-assembled rapidly into aggregates displaying exposed hydrophobic clusters at pH 4. 5, whereas it remained soluble at pH 7. 0 (S14 Fig). Indeed, when fresh rBap_Bsapro was incubated at pH 4. 5 for 24 h, the presence of fibrillar structures became apparent (Fig 9F). Finally, we extended the analysis of the amyloid-forming propensity to Bap_B domains of S. simiae, S. xylosus, S. epidermidis and S. simulans (S5 Table). For that, we used the curli-dependent amyloid generator (C-DAG) system that provides a simple cell-based method to test particular target proteins for their amyloid-forming propensity [53]. The presence of extracellular amyloid aggregates was detected by analyzing the capacity of the strains to bind Congo Red dye (CR). Interestingly, all the Bap_B domain orthologs expressed in C-DAG system were able to bind CR, whereas Bap_A domain of S. aureus did not (Fig 9G). Together, these data indicate that Bap orthologs also utilize amyloid assembly as a molecular mechanism to induce multicellular behaviour. There is a growing recognition that proteins play an important role building biofilm matrix scaffold. To fulfill this function these proteins need to provide stable intercellular connections and at least in some cases, also mediate adhesion to the surface. In this report, we have shown that Bap forms extracellular amyloid-like fibers that assist in building the biofilm matrix in S. aureus. Bap shares structural and functional properties with SasG and Aap proteins of S. aureus and S. epidermidis respectively, implicated in cell-to-cell accumulation and adhesion to epithelial cells [38,54–56]. However, the mechanims of action of SasG and Aap are completely different to the one reported here for Bap. All three proteins undergo a limited proteolytic cleavage of the N-terminal domain that induces biofilm formation [15,57]. The mechanism underlying this processing is different among the three proteins. SasG is known to suffer spontaneous cleavage at labile bonds in its B domain, since protease inhibitors added to the growth medium, as well as strains deficient in each known extracellular and membrane-bound proteases, had no effect on the pattern of SasG processing [15]. In the case of Aap, endogenous and also exogenous host-derived proteases are the responsible for protein cleavage, and addition of α2-macroglobulin to the growth medium specifically led to the loss of cell clumping and biofilm formation of S. epidermidis [57]. We failed to identify a staphylococcal protease responsible for Bap cleavage, because protease mutants (∆aur, ∆sspA and ∆sspBC) and protease inhibitors (α2-macroglobulin, E64, ScpB and PMSF) did not change the proteolytic pattern of Bap (S4 Fig). However, the possibility that a protease different from the ones tested cannot be discarded and requires further study. In the case of SasG and Aap, the N-terminal domain is removed by proteolysis allowing the C-terminal region containing the G5 domains to promote zinc-dependent self-association of opposing molecules [6,15,57,58]. In contrast, it is the N-terminal region of Bap that is released to the extracellular media and self-assembles into amyloid-like fibers, whilst part of the C-terminal repeats region remains anchored to the membrane. Several pieces of experimental evidence support the amyloid-like properties of the Bap_B domain aggregates. First, far-UV CD spectra reveal a modest switch in secondary structure of Bap from disordered to β-sheet, as the pH becomes more acid. Second, FTIR spectrum of Bap aggregates is dominated by β-sheet/β-turn secondary structure. Third, rBap_B binds to the amyloid diagnostic dyes Thioflavin-T, Congo Red and Proteostat and forms aggregates with fibrillar morphology when observed by electron microscopy. Finally, rBap_B contains short-sequence stretches with significant amyloidogenic potency that together with other unknown sequence stretches could contribute to the fibrillogenesis of Bap fragments. The self-assembly of rBap_B at acidic pH is a fast process, where hydrophobic interactions appear to play an important role, at least at the early stages of the reaction. Genuine bacterial functional amyloids utilize sophisticated machineries that direct the polymerization of amyloid fibers outside the cell. For instance, curli (csgACB-csgDEFG), Fap (fapA-F) in Pseudomonas strain UK4, chaplins (chpA-H) in Streptomyces coelicolor and TasA (tapA-sipW-tasA) in Bacillus are expressed together with accessory proteins involved in secretion, nucleation, and assembly of the amyloid subunit [29,30,59–61]. Bap appears to follow a more simplistic model of amyloid auto-aggregation, which does not require the expression of accessory proteins. In this respect, amyloidogenic behavior of Bap could be similar to the mechanism conducted by the surface protein antigen I/II (adhesin P1) of Streptococcus mutans [62,63]. What are the underlying reasons for the conversion of a cell wall anchored protein like Bap into an amyloid fiber? Our results suggest that this strategy allows Bap to play a dual role during biofilm development (Fig 10). Initially, Bap is secreted and covalently anchored to the cell wall. Then, Bap is processed or non-enzymatically cleaved releasing fragments containing the N-terminal region to the media that remain soluble at neutral pH. If the pH of the environment decreases, the N-terminal domain of Bap would transition from its partially ordered native state to an aggregation-prone conformation that would facilitate polymerization into amyloidogenic fibrillar structures. The presence of calcium drastically influences the multicellular behavior promoted by Bap. From a biophysical perspective, the binding of calcium probably to the EF-hand domains of the protein, stabilizes its initial molten globule-like state, likely sequestering the functional N-terminal fragments released from Bap cleavage, and consequently impairing their self-assembly into amyloid structures (Fig 10). In eukaryotes, there are several examples of proteins involved in aggregation disorders, whose capacity to form multimeric aggregates depend on changes in protein folding caused by binding of metal ions [64]. This is the case of S100A6, an amyloid protein largely expressed in patients with Amyotrophic Lateral Sclerosis (ALS) disease. When S100A6 binds calcium, it suffers a remodeling of the surface electrostatics and hydrophobic patch exposure at the aggregation hotspot inhibiting protein self-assembly into amyloid fibrils [65]. In the case of bacteria, Ca2+ bound to α-haemolysin secreted by pathogenic E. coli, makes the protein more compact, stabilizing its structure and making it less prone to oligomerization [66]. In a similar way, the binding of Ca2+ to Bap causes tertiary rearrangements that increase the stability of the intermediate molten globule-like state of Bap in solution and thus decrease its aggregation behavior. This ultimately prevents cellular interactions and biofilm formation in S. aureus. Amyloid structures are especially well suited for assembling the biofilm matrix scaffold, as polymerization can occur in the extracellular media in the absence of energy. Furthermore, the amyloid structure provides high stability and inherent resistance against protease digestion and denaturation [28,67]. The pH at which the B domain of Bap shows aggregation activity (pH∼5, early stationary phase) is very close to the isoelectric point (pI∼4. 61), where lack of a net charge facilitates interactions between protein molecules, making protein self-assembly more likely. Indeed, a large number of globular and non-globular proteins, including the pathogenic amyloid β peptide and α-synuclein have been shown to display maximum amyloid propensity when they approach their pI, indicating that the solubility of a polypeptide chain is a major factor that determines its conversion to the amyloid state [68]. Because Bap assembly can be reversed when pH is restored to neutrality, it is not difficult to imagine that Bap is able to withstand pH fluctuations, adapting its function by switching from aggregated to soluble forms (and vice versa). The ability to fluctuate between soluble and amyloid-like states has been shown to underlie key physiological processes like processing bodies and stress granules formation [69], the cellular response to DNA breakage [70] or the integrity of the cytoskeleton [71]. In S. aureus this mechanism may have a relevant physiological effect during infectious processes, in which local acidosis usually arises from the accumulation of acidic products as a result of an inflammatory response [72], and bacterial metabolism. Also, this pH-driven Bap-mediated bacterial aggregation mechanism would be physiologically significant for those Staphylococcus species expressing Bap homologous proteins that are capable of colonize human host niches displaying mildly acidic conditions (e. g. , skin, anterior nares, vagina, urinary tract and mouth) [73,74], as in the case of S. epidermidis (skin, vagina during prepubertal phase), S. saprophyticus (urinary tract and vagina) and S. warneri (skin, nasal cavity, urinary tract). In a similar way, Foulston et al [39] demonstrated that many cytoplasmic proteins reversibly associated with the cell surface in response to pH, acting as a biofilm scaffold matrix in S. aureus. Regarding calcium ion, its effect on Bap functionality might serve to explain how changes in Ca2+ levels during the stages of the lactation cycle affect intramammary infections caused by S. aureus. Bap displays low binding affinity to calcium, thus medium-to-high millimolar concentrations of the cation are required to saturate Ca2+-binding sites in the protein. Normally, the concentration of free Ca2+ in mammalian blood is strictly maintained between 1. 1–1. 3 mM [75,76]}. However, the total Ca2+ concentration in milk is higher, around 1. 2 mg/liter (~30 mM), being one third of this total amount free in serum (~11 mM) [77]. Thus, Ca2+ levels present in the milk during the lactation period are sufficient to inhibit Bap-mediated biofilm development. On the contrary, the low Ca2+ concentration conditions that occur in the udder during the dry period allow the formation of Bap-mediated biofilms and the establishment of long-term persistent infections on the mammary gland epithelium [36]. One question that remains open from this study is how Bap amyloid fibers interact with the bacterial surface to induce cell-to-cell aggregation? In S. coelicolor, it has been proposed that covalently linked cell wall chaplin variants ChpA–C contribute to anchoring the fibers to the cell surface [59]. Following the same reasoning, one would expect that Bap amyloid fibers might interact with the C-terminal domain of Bap that remains covalently anchored to the cell wall. However, the finding that extracellular addition of rBap_B to bap deficient S. aureus strains (Fig 4B) and also to L. monocytogenes and E. faecalis (S15 Fig) promoted intercellular adhesion and biofilm formation makes this possibility very unlikely. Finally, our results indicated that Bap orthologs share similar molecular mechanisms as Bap for mediating biofilm development. S. saprophyticus is a notable human uropathogen [78]. The pH of the urinary tract varies between 4. 5 and 7 [74], representing an environment in which Bap, through the formation of amyloid aggregates and together with urease, UafA [78] and other virulence factors [79] could play an important role in the survival and uropathogenesis of S. saprophyticus. Except for S. saprophyticus and S. simiae, the rest of CNS strains used in this study were mostly isolated from mastitis of lactating dairy cows (cultured from milk samples), a physiological situation in which the presence of calcium, as previously explained, can actually play a relevant role in regulating the functionality of Bap. The diversity of yet unknown factors that could affect Bap amyloid behaviour among different bacterial species is worthy of further study. The bacterial strains and plasmids used in this study are listed in S1 Table. Oligonucleotides were synthesized by StabVida (Caparica—Portugal) (S2 Table). Enzymes for DNA manipulation were supplied by Thermo Scientific and were used according to manufacturer’s recommendations. Staphylococcal strains and E. coli and were grown in Luria-Bertani (LB) broth or in LB agar (Pronadisa). Media were supplemented when appropriate with 10 μl/ml erythromycin, 100 μg/ml ampicillin, 0. 25% wt/vol glucose, 1 μM CdCl2,15 mM CaCl2 and 1. 25 mM EDTA. Plasmid DNA was isolated from E. coli strains using a Qiagen plasmid miniprep kit (BioRad), according to the manufacturer’s protocol. Plasmids were transformed into staphylococci by electroporation, using a previously described procedure [21]. Deletion mutants were generated via allelic replacement using the vector pMAD as described previously [80]. The signal peptide (SP) and the different N-terminal domains of the bap gene (AB, B, B∆EF and A) were amplified from S. aureus V329 and V329 ∆EF [36]. To amplify Bap_AB fragment we used primers Bapori-1mB and Bap-63cK (S2 Table). To obtain Bap_B and Bap_B ∆EF regions we first amplified the signal peptide sequence using primers Bapori-1mB and Bap-65c and second, the region B with primers Bap-66m and Bap-63cK. An overlapping PCR was performed with primers Bapori-1mB and Bap-63cK to get a single fragment. To obtain Bap_A region, two fragments were amplified using primers Bapori-1mB and BapB1 (comprising signal peptide sequence), and BapB2 and BapB3K (comprising A domain). A second overlapping PCR was performed with primers Bapori-1mB and BapB3K in order to obtain a single fragment. To obtain Bap_B region of S. saprophyticus, we first amplified the signal peptide of Bap from S. aureus V329 using primers Bapori-1mB and SPbap-sapro-Rv and second, the B-region from S. saprophyticus B20080011225 using primers bapB-sapro-Fw and Sapbap-KpnI-Rv. An overlapping PCR was performed with primers Bapori-1mB and Sapbap-KpnI-Rv to obtain a single fragment. The entire ClfA fragment used as a control for Bap chimeras, was developed by amplifying clfA gene from S. aureus Newman using primers ClfA-9mB and ClfA-7cE. To allow anchoring of amplified bap domains to the bacterial cell wall, the R region of clumping factor A gene containing an LPXTG motif was amplified from S. aureus Newman strain using primer K-3xF-ClfA containing a flag tag and a recognition sequence for KpnI, and primer ClfA-7cE with a recognition sequence for EcoRI. The KpnI/EcoRI-restricted R-clfA was ligated with KpnI/EcoRI-restricted pCN51 vector [81]. The resulting construct was then digested with BamHI and KpnI to insert the previously amplified domains of bap gene. The final pCN51 plasmid constructs thus contained different parts of bap gene fused to a flag tag followed by the C-terminal R domain of clumping factor A gene, expressed under the activity of a cadmium inducible promoter. To obtain E. coli strains for curli-dependent amyloid generator (C-DAG) system, we PCR amplified from purified genomic DNA (i) region A of bap from S. aureus (primers CDAG BAP_A-Fw and CDAG BAP_A-Rv) and (ii) region B of bap from several staphylococcal species: S. aureus (primers cdag-B-NotI-Fw and cdag-B-XhoI-Rv, S2 Table), S. saprophyticus (primers BAPsapro-cdag-Fw and BAPsapro-cdag-Rv), S. simiae (primers BAPsimiae-cdag-Fw and BAPsimiae-cdag-Rv), S. epidermidis (primers epider-CDAG-Fw and epider-CDAG-Rv), S. simulans (primers simulans-CDAG-Fw and simulans-CDAG-Rv), and S. xylosus (primers xylosus-CDAG-Fw and xyosus-CDAG-Rv). The NotI/XhoI-restricted bapA and bapB fragments were ligated with NotI/XhoI-restricted pEXPORTXhoI plasmid (S1 Table). This vector was obtained by replacing XbaI recognition sequence of the original pEXPORT plasmid [53] for that of XhoI using QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies) and primers pVS72-XhoI-5 and pVS72-XhoI-3. The final pEXPORT constructs were transformed in E. coli VS39 strain. Induction of protein production and presence of amyloid-like material was assessed on solid medium containing 10 μg/ml Congo Red by evaluating colony-color phenotype, as previously described [53]. To generate the deletion in the aur, sspA, sspB genes coding for proteases present in S. aureus, and a deletion in the spa gene coding for surface protein A, we used the pJP437, pJP438, pJP439 [82] and pMADspaAD [20] plasmids which contained two fused fragments of 500 bp each that flanked the left and the right sequence of aur, sspA, sspBC and spa genes, respectively. Plasmids were transformed in V329 or Δbap strains by electroporation. Homologous recombination experiments were performed as described [80]. V329 Δaur, ΔsspA and ΔsspBC (this last strain was also deleted in the cysteine protease inhibitor SspC, which is the last gene of the operon that codifies for SspA and SspB) strains were verified using primers ssp-20cN/ssp-17mS, ssp-24cN/ssp-21mS and aur-Fw/aur-Rv (S2 Table). V329 Δspa and ΔbapΔspa strains were verified using primers spaF/spaE (S2 Table). Biofilm formation assay in microtiter wells was performed as described [83]. Briefly, strains were grown overnight at 37°C and then diluted 1: 40 in the corresponding media supplemented when required with antibiotics, 20 mM CaCl2, or proteases inhibitors (2 U/ml α-macroglobulin, 2 mM cysteine protease inhibitor E64,10 μM PMSF and 250 nM ScpB). Cell suspension was used to inoculate sterile 96-well polystyrene microtiter plates (Thermo Scientific). After 24 hours of incubation at 37°C wells were gently rinsed two times with water, dried and stained with 0. 1% of crystal violet for a few minutes. When desired, crystal violet adhered at the bottom of the wells was resuspended with 200 μl of a solution of ethanol: acetone (80: 20 vol/vol) and quantified using a Multiskan EX microplate photometer (Thermo Scientific) with a 595 nm filter. For biofilm disassembly assays, cells were grown in LB-glu at 37°C on polystyrene microtiter plates. Once formed, adhered biofilm were treated with dispersant agents (0. 4 μg/ml Dispersin B, 0. 4 μg/ml DNase I, and 20,100 and 200 μM EGCG) for 2 hours at 37°C. Alternatively, old LB-glu media were extracted and replaced for new LB, LB-glu, and LB-glu + 20 mM CaCl2, and incubated overnight at 37°C. Finally, treated and non-treated biofilms adhered to polystyrene wells were macroscopically determined and quantified as previously described. Biofilm formation under flow conditions was analyzed using microfermenters (Pasteur Institute’s laboratory of Fermentation) with a continuos flow 40 ml/h of LB-glu and constant aeration with sterile pressed air (0. 3 bar) [84]. Medium was supplemented with 10 μg/ml erythromycin and 1 μM CdCl2 when required. Each microfermentator was inoculated with 108 bacteria from an overnight culture of the corresponding strain. Biofilm development was recorded with a Fugifilm FinePix S5800 digital camera. Aggregation phenotype in cell suspension was determined as described before [10]. Cells were grown overnight in the corresponding medium (TSB-glu, LB-glu or LB acidified with 0,1 M HCl) at 37°C, shaking at 200 rpm and were examined macroscopically for the presence or absence of aggregates (intercellular adhesion). For bacterial clumping reversion assay, bacteria were grown overnight in LB-glu at 37°C, 200 rpm. Cultures were subsequently centrifuged and LB-glu medium was replaced for LB. After after 6 and 18 h incubation at 37°C, 200 rpm, bacterial aggregation at the bottom of the tube was evaluated for each strain and pictures were taken with a FUJIFILM FinePix S5800 digital camera. To quantify bacterial aggregation, the OD600nm at the top of the culture tubes (approximately 1 cm from the surface) was measured as an estimation of non-settled bacteria (planktonic cells) present in the culture after an overnight incubation at 37°C, 200 rpm. The experiment was independently repeated three times, and data were analyzed with the Mann-Whitney test. For immunofluorescence, cells were grown overnight in the corresponding tested conditions and fixed with 3% paraformaldehyde (SIGMA) for 5 minutes. 200 μl of fixed bacteria were placed on coverslips and incubated for 30 minutes. After several washes with PBS, cells were saturated with PBS-0. 5% BSA, and finally stained with anti-Bap or anti-Flag (Sigma) antibodies diluted 1: 1000. Alexa Fluor 488-conjugated goat anti-rabbit (Invitrogen) diluted 1: 200 was used as a secondary antibody and DAPI diluted 1: 200 was used to label nuclei. For ProteoStat staining of amyloid material in-vivo, cells were grown overnight in LB and LB-glu, at 37°C, in polystyrene 24-wells plates. Adhered biofilm was resuspended and fixed with 3% paraformaldehyde (SIGMA) for 5 minutes. Bacteria were washed several times with 1X PBS, and then incubated for 30 minutes, at room temperature and in darkness with ProteoStat Mix buffer (1X Assay Buffer, 1 μl ProteoStat®, 2 μl Hoechst). Bacteria were washed twice with 1X PBS. All preparations were observed with an Axioskop 2 plus epifluorescence microscope (Zeiss) and images were acquired and analyzed with EZ-C1 software (Nikon). For Transmission electron microscopy (TEM), cells were grown overnight in the corresponding tested conditions, washed twice with phosphate-buffered saline (PBS) and then fixed with paraformaldehyde 2% (SIGMA) for 1 hour at room temperature. Formvar/carbon-coated nickel grids were deposited on a drop of fixed sample during five minutes and rinsed three times with phosphate-buffered saline (PBS). Negative staining was performed using 2% uranyl acetate (Agar Scientific, Stansted, UK). Observations were made with a JEOL 1011 transmission electron microscope. For Bap immunogold labelling, grids coated with the sample were washed and incubated for 45 minutes on a drop of PBS containing 1: 10 antibody against BapB. After washing with PBS, grids were incubated 45 minutes with gold-conjugated (10nm) goat-anti-rabbit secondary antibody (Aurion, Wageningen, Netherlands). Grids were stained with uranyl acetate as described above. Region B of Bap (amino acids 361–819) was PCR amplified from purified S. aureus V329 genomic DNA using high fidelity Phusion DNA Polymerase (Thermoscientific) and primers bapB1-LIC-Fw and bapB1-LIC-Rv (S2 Table) designed for use in the LIC cloning system. The resulting 1377 bp fragment was cloned in pET46-Ek/LIC vector (Novagen). B-region of Bap from S. saprophyticus B20080011225 was PCR amplified from its purified genomic DNA using primers bapB-sapro-LIC-Fw and bapB-sapro-LIC-Rv. The resulting 1311 bp fragment was cloned in pET46-Ek/LIC vector (Novagen). Both fusions resulted in Bap_B constructs containing an N-terminal hexahistidine tag (rBap_B and rBap_Bsapro). Overnight cultures of Escherichia coli BL21 DE3 containing Bap_B expression plasmid were diluted 1: 100 and grown to an OD600nm of 0. 6. Isopropyl B-D-thiogalactopyranoside (IPTG) was added to a final concentration of 0,1 mM and the cultures were shaken at 20°C overnight. After centrifugation, pellets were resuspended, sonicated and centrifuged. Supernatants were filtered (0,45 μm) and rBap_B protein purified by Ni affinity chromatography using HisGraviTrap gravity-flow columns (GE Healthcare). To achieve the highest purity, size exclusion chromatography was applied with a HiLoad 16/600 Superdex 200 pg column (GE Healthcare). The concentration of the purified protein was determined by the Bicinchoninic Acid (BCA) Protein Assay (Pierce, Thermo Scientific) using BSA as a standard. Rabbit polyclonal antibodies raised against purified rBap_B protein were supplied by Abyntek Biopharma S. L. (Spain). Antibodies were subsequently immunoabsorbed and purified using NAb Spin Kit (Thermoscientific). To test extracellular complementation, bacteria were grown in LB, LB-glu or LB-glu + 20 mM CaCl2 media mixed with 2 μM-purified rBap_B shaking at 200 rpm at 37°C. Aggregation phenotype in cell suspension was determined and quantified as described above. To determine the exact pHs at which rBap_B is capable to form aggregates, 2 μM of the protein was incubated in phosphate-citrate buffer at pH ranging from 2. 0 until 8. Protein aggregates were macroscopically determined and pictures were taken with a FUJIFILM FinePix S5800 digital camera. For aggregates reversion assay of 2 μM assembled rBap_B and rBap_Bsapro protein, the phosphate-citrate buffer at pH 4. 5 was removed and exchanged for phosphate-citrate buffer at pH 7. After an overnight incubation at 37°C and 200 rpm, dissolution of rBap_B and rBap_Bsapro aggregates was macroscopically determined and pictures were taken with a FUJIFILM FinePix S5800 digital camera. Overnight cultures of S. aureus strains were diluted 1: 100 and grown in LB-glu or LB supplemented with the corresponding antibiotic, 20 mM CaCl2 and 1 μM CdCl2 when necessary, at 37° C and 200 rpm. Samples were obtained at different point of the growth curve. For protease inhibition assays, diluted S. aureus V329 cultures were supplemented with proteases inhibitors (2 U/ml α2-macroglobulin, 2 mM cysteine protease inhibitor E64,10 μM PMSF and 250 nM ScpB) and grown until OD≈0. 7. Cells were harvested, washed and finally resuspended in 100 μl of PBS buffer containing 30% raffinose (Sigma), 5 μl of lysostaphin 1 mg/ml (Sigma) and 2 μl of DNase 1mg/ml (Sigma). After 2 hours of incubation at 37° C, cells were centrifuged. The supernatants from surface protein extracts were recovered and analyzed by SDS-PAGE or Native gels. For SDS-PAGE, 1 volume of Laemmli buffer was added to the samples and boiled for 5 minutes. 10 μg of protein was used for SDS-PAGE analysis (7,5% separation gel; 5% stacking gel). For Native gels, surface protein extracts were mix 1: 2 with native sample buffer (BioRad). Proteins were separated in Criterion XT Tris-acetate gels and Tris/glycine running buffer (BioRad). For Western blot analysis, protein extracts were blotted onto Hybond-ECL nitrocellulose membranes (Amersham Biosciences). Anti-Bap purified antibody and monoclonal anti-Flag M2-Peroxidase (HRP) antibody (Sigma) were diluted 1: 20. 000 and 1: 1000, respectively, with PBS-Tween 5% skim-milk. Alkaline phosphatase-conjugated goat anti-rabbit Immunoglobulin G (Thermo Scientific) diluted 1: 5000 in PBS-Tween 5% skim-milk was used as a secondary antibody for Bap detection and the subsequent chemiluminescence reaction was recorded (Chemiluminescent Substrate Thermo Scientific). S. aureus V329 strain was grown in LB and LB-glu media, at 37°C, 200 rpm. After an overnight incubation, cells were harvested, washed and finally resuspended in 100 μl of PBS buffer containing 30% raffinose (Sigma), 5 μl of lysostaphin 1 mg/ml (Sigma) and 2 μl of DNase 1mg/ml (Sigma). After 2 hours of incubation at 37° C, cells were centrifuged. Supernatants were mix 1: 2 with native sample buffer (BioRad). Proteins were separated in Criterion XT Tris-acetate gels using Tris/glycine running buffer (BioRad). The material retained in the wells of the native gels was excised, washed three times in ddH20, and digested in-gel with 250 ng of trypsin (Sequencing grade modified Trypsin-Promega) in 50 mM ammonium bicarbonate for 16 h at 37°C, after a denaturation step with DTT (10 mM, 30 min, 40°C) and an alkylation step with Iodoacetamide (25mM, 30 min, room temperature). The resulting peptides were extracted with 1% formic acid, 50% acetonitrile and evaporated to dryness prior to LC-MSMS analysis. For each digested sample, a total volume of 5 μl of tryptic peptides was injected with a flow rate of 300 nL/min in a nanoLC Ultra1D plus (Eksigent). A trap column Acclaim PepMap100 (100 μm x 2 cm; C18,2 μm, 100 Å) and an analytical column Acclaim PepMap RSLC (75 μm x 15 cm, C18,5 μm, 100 Å) from Thermo Scientific were used following the next gradient: 0–1 min (5% B), 1–50 min (5–40% B), 50–51 min (40–98% B), 51–55 min (98% B), 55–56 min (5% B), 56–75 min (5% B). (Buffer B: 100% acetonitrile, 0. 1% formic acid, Buffer A: 0. 1% formic acid). MS analysis was performed on a Q-TRAP 5500 system (ABSciex) with a NanoSpray® III ion source (ABSciex) using Rolling Collision Energy in positive mode. MS/MS data acquisition was performed using Analyst 1. 5. 2 (AB Sciex) and submitted to Protein Pilot software (ABSciex) against UniprotKB/Swiss-Prot database (restricted to “Staphylococcus”) and then against a specifically restricted database for BAP protein from Staphylococcus aureus, using the Paragon™ Algorithm and the pre-established search parameters for 5500 QTRAP. Adherence experiments were performed as described previously [85]. Briefly, prior to use, wells were seeded with 0. 3 x 106 MAC-T cells in 6-well tissue culture plates. Once cells were confluent (1. 2 x 106) the culture medium was removed and cells were washed with DMEM plus 10% heat-inactivated fetal bovine serum. Overnight bacterial cultures were mixed vigorously and added to the monolayer cells in a multiplicity of infection of 10 in DMEM. Incubation was carried out 1 hour at 37°C in 5% CO2. To remove non-adherent bacteria, cells were washed three times with sterile PBS. Eukaryotic cells were lysed with 0. 1% Triton X-100. Before plating extracts were mixed vigorously by vortexing. The number of adherent bacteria were determined by serial dilution and plating. Experiments were performed in triplicate. A mouse foreign body infection model was used to determine the role of Bap aggregates in the pathogenesis of S. aureus. Groups of 6 CD1 mice were used. A 3-cm segment of intravenous catheter (24G1”, B. Braun) was aseptically implanted into the subcutaneous interscapular space. Each group of six mice was inoculated with 1 x 107 CFU of either S. aureus V329 or Δbap mutant previosly grown overnight in LB-glu at 37°C. Twelve animals were euthanatized by cervical dislocation on days 4 or 10 post-infection. The catheter was aseptically removed, placed in a sterile microcentrifuge tube containing 1 ml of PBS, and vortexed at high speed for 3 min. Samples were serially diluted and plated onto TSA plates for enumeration of viable staphylococci. All animal studies were reviewed and approved by the Comité de Ética, Experimentación Animal y Bioseguridad, of the Universidad Pública de Navarra (approved protocol PI-019/12). Work was carried out at the Instituto de Agrobiotecnología building under the principles and guidelines described in European Directive 86/609/EEC for the protection of animals used for experimental purposes. Proteolysis of rBap_B (1 mg/ml) was performed at 37°C in the presence or absence of 50 mM CaCl2. The protein was incubated with 80 μg/ml Proteinase-K (SIGMA) for 0,15,30,45 minutes and the reaction was stopped by the addition of 5 mM PMSF. Degradation pattern was analyzed by SDS-PAGE (12%) followed by western immunoblotting with anti-Bap purified primary antibody (1: 10. 000) and alkaline phosphatase-conjugated goat anti-rabbit Immunoglobulin G (1: 5000) (Thermo Scientific) as a secondary antibody. Thioflavin-T (ThT) binding was analyzed for 0. 1 mg/ml aggregated rBap_B and rBap_Bsapro in the presence of 25 μM ThT, 25°C, pH 4. 5. ThT binding was also measured for rBap_B at different concentrations (0. 01,0. 018,0. 027 and 0. 036 mg/ml) in the presence of 25 μM ThT, 25°C, at pH 4. 5 and pH 7. Fluorescence emission spectra were recorded from 460 to 600 nm with an excitation wavelength of 440 nm, using a slit width of 5 nm for excitation and emission in a Jasco FP-8200 spectrophotometer (Jasco corporation, Japan) at 25°C. Each trace represents the average of 5 accumulated spectra. Aggregation kinetics of 0. 01 mg/ml rBap_B protein in phosphate-citrate buffer at pH 4. 5, pH 4. 5 plus CaCl2 and pH 7. 0 were recorded for 1000 s under agitation (800 rpm) at 25°C, in the presence of 25 μM ThT. The kinetic traces were measured exciting at 440 nm and emission was recorded at 475 nm, slit widths of 5 nm were used for excitation and emission in a Jasco FP8200 spectrophotometer (Jasco corporation, Japan). ThT fluorescence spectra were recorded at the end of the experiment. Congo red (CR) interaction with 0. 1 mg/ml aggregated rBap_B and rBap_Bsapro at pH 4. 5 was tested using a Cary-400 UV/Vis spectrophotometer at 25°C. After 5 minutes of equilibration, the absorbance spectra were recorded from 400 to 700 nm. Each trace represents the average of 5 accumulated spectra. Fluorescence emission of 0. 1 mg/ml assembled rBap_B and rBap_Bsapro stained with ProteoStat was measured on a Jasco FP-8200 fluorescence spectrophotometer (Jasco corporation, Japan) at 25°C. The emission spectra were recorded between 500 and 650 nm wavelength and the samples excited at 484 nm. Slit widths for excitation and emission spectra were 5 nm. The spectra were obtained from the average of 5 consecutive scans. ATR FTIR spectroscopy analyses of rBap_B aggregates formed in phosphate-citrate buffer pH 4. 5 were performed with a Bruker Tensor 27 FTIR Spectrometer (Bruker Optics Inc.) with a Golden Gate MKII ATR accessory. Spectrum acquisitions consisted of 16 independent scans, measured at a resolution of 2 cm-1 within the 1800–1500 cm-1 range. Spectra were acquired, background subtracted, baseline corrected and normalized using the OPUS MIR Tensor 27 software. Second derivatives of the spectra were used to determine the frequencies at which the different spectral components were located. All FTIR spectra were fitted to overlapping Gaussian curves using PeakFit package software (Systat Software) and the maximum and the area of each Gaussian were calculated. Samples of 0. 1 mg/ml rBap_B and rBap_Bsapro soluble proteins (phosphate-citrate buffer at pH 7) or protein aggregates (phosphate-citrate buffere at pH 4. 5) were prepared in solutions containing 10 μM of Bis-ANS and analyzed immediately on a Jasco FP-8200 fluorescence spectrophotometer (Jasco corporation, Japan) at 25°C. The samples were excited at 370 nm and emission measured between 400 and 600 nm with slit widths of 5 nm. The spectra were obtained from the average of 5 consecutive scans. Far-UV CD spectra were measured in a Jasco-710 (Jasco, Japan) or in a Chirascan (Applied Photophysics) spectropolarimeter thermostated at 25°C. rBap_B at concentrations ranging from 0. 2 to 1. 5 mg/ml was measured in 10 mM MOPS either with 1mM CaCl2,10 mM CaCl2 or 100 mM CaCl2, or alternatively 10 mM NaCl or 100 mM NaCl and 10 mM EDTA at pH 7. 0/7. 5. For measurements at acidic pH, rBap_B (6 mg/ml) in 10 mM NaPO4 pH 7. 0,50 mM (NH4) 2SO4 was diluted 30-fold to 0. 2 mg/ml into 100 mM NaPO4,10 mM EDTA at pH 4. 4. Spectra were recorded from 260 to 190 nm, at 0. 2 nm intervals, 1 nm bandwidth, and a scan speed of 50 nm/min. Twenty accumulations were averaged for each spectrum. Deconvolution of the data were performed using the Dichroweb server implementing the CDSSTR algorithm with reference set 7 [86,87]. Near-UV CD spectra were recorded in a Jasco-710 spectropolarimeter (Jasco, Japan) thermostated at 25°C, from 260 to 320 nm with a 1 nm bandwidth, and a scan speed of 50 nm/min in 10 mM MOPS pH7. 0 with 1 mM CaCl2,10 mM CaCl2, or 10 mM NaCl 10 mM EDTA. Tryptophan intrinsic fluorescence was measured at 25°C on a Varian Cary Eclipse spectrofluorometer using an excitation wavelength of 280 nm and recording the emission from 300 to 400 nm. Five averaged spectra were acquired and slit widths were typically 5 nm for excitation and emission. Protein concentration was 1. 5 mg/ml in 10 mM MOPS either with 100mM CaCl2 or 100 mM NaCl, 10 mM EDTA at pH 7. 5. Thermal denaturation was monitored in a Jasco FP-8200 fluorescence spectrophotometer (Jasco, Japan) The samples were excited at 280 nm and the emission was recorded at 350 nm, using slit widths of 5 nm for excitation and emission. The emission was registered each 0. 25 K with a heating rate of 0. 5 K/min. Static light scattering of 0. 1 mg/ml rBap_B and rBap_Bsapro in phosphate-citrate buffer at pH 4. 5 and pH 7 was recorded on a Jasco FP-8200 spectrofluorometer (Jasco corporation, Japan). Five accumulative spectra were registered with excitation at 330 nm and emission between 320 and 340 nm. Slit widths of 5 nm for excitation and emission were used. Dynamic light scattering data of 1 mg/ml rBap_B protein in phosphate-citrate buffer at pH 3,4. 4 and 7 were obtained with a DynaPro DLS reader (Wyatt Technology, Germany) using an 825 nm wavelength laser and analyzed with Dynamics V6 software. Hydrodynamic radium (nm), polidispersity percentage and diffusion coefficient (cm2/s) of each population observed at the different pH values were obtained. The size exclusion chromatography experiment was performed using a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) connected to an AKTAprimeTM Plus chromatography system (GE Healthcare). A 500 μl portion of rBap_B was loaded onto the gel filtration column equilibrated in MOPS buffer (10 mM MOPS, 100 mM NaCl, pH = 7,5) with 100 mM CaCl2 or 10 mM EDTA and eluted with one column volume (124 ml) at a flow rate of 1 ml/min. Recorded data were analyzed using PrimeView software (GE Healtcare). All AUC experiments were carried out at 20°C in the presence of 100 mM of CaCl2 and 10 mM EDTA, on a Beckman XL-I analytical centrifuge using absorbance optics. Sedimentation velocity was performed for rBap_B at three different concentrations (1,2 and 3 mg/ml) at 48,000 rpm overnight and the data were analyzed using SedFit 14. 7g [88]. Sedimentation equilibrium runs were performed for rBap_B (loading concentrations of 1,2 and 3 mg/ml) at speeds of 13,000 and 8,500 rpm and analyzed using HeteroAnalysis 1. 1. 44. One-dimensional proton NMR experiments were performed at 30°C on 350 μM Bap_B in a buffer containing 10 mM MOPS, 100 mM CaCl2,10% D2O or 10 mM MOPS, 100 mM NaCl, 10 mM EDTA, 10% D2O. Spectra were processed within TopSpin (Bruker). The predicted peptides GIFSYS and TVGNIISNAG were obtained from CASLO ApS (Lyngby, Denmark) with high purity (99. 88% and 98. 29% respectively). Peptide stock solutions at 1 mM were prepared by dissolving into citrate buffer. Samples were immediately sonicated for 10 min to dissemble preformed nuclei and centrifuged (5 min at 16,100g) to deposit insoluble material. Peptide solutions were incubated at room temperature (25°C) for four weeks and amyloid properties were evaluated as described above. The statistical analysis was performed with the GraphPad Prism 5 program. A nonparametric Mann-Whitney test was used to assess significant differences in biofilm formation and bacterial aggregation capacity, as well as for analysis of experimental infection. The differences in bacterial aggregation after exogenous complementation with rBap_B protein were determined using the unpaired Student’s t test.

Major components of the biofilm matrix scaffold are proteins that assemble to create a unified structure that maintain bacteria attached to each other and to surfaces. We provide evidence that a surface protein present in several staphylococcal species forms functional amyloid aggregates to build the biofilm matrix in response to specific environmental conditions. Under low Ca2+ concentrations and acidic pH, Bap is processed and forms insoluble aggregates with amyloidogenic properties. When the Ca2+ concentration increases, metal-coordinated Bap adopts a structurally more stable conformation and as a consequence, the N-terminal region is unable to assemble into amyloid aggregates. The control of Bap cleavage and assembly helps to regulate biofilm matrix development as a function of environmental changes.

Abstract Introduction Results Discussion Materials and Methods

bacteriology biofilms cell walls medicine and health sciences fluorescence pathology and laboratory medicine enzymes pathogens microbiology enzymology staphylococcus aureus electromagnetic radiation cellular structures and organelles bacteria bacterial pathogens staphylococcus saprophyticus staphylococcus medical microbiology proteins extracellular matrix microbial pathogens luminescence physics biochemistry bacterial biofilms cell biology biology and life sciences proteases physical sciences organisms amyloid proteins

2016

Staphylococcal Bap Proteins Build Amyloid Scaffold Biofilm Matrices in Response to Environmental Signals

Guiding axon growth cones towards their targets is a fundamental process that occurs in a developing nervous system. Several major signaling systems are involved in axon-guidance, and disruption of these systems causes axon-guidance defects. However, the specific role of the environment in which axons navigate in regulating axon-guidance has not been examined in detail. In Drosophila, the ventral nerve cord is divided into segments, and half-segments and the precursor neuroblasts are formed in rows and columns in individual half-segments. The row-wise expression of segment-polarity genes within the neuroectoderm provides the initial row-wise identity to neuroblasts. Here, we show that in embryos mutant for the gene midline, which encodes a T-box DNA binding protein, row-2 neuroblasts and their neuroectoderm adopt a row-5 identity. This reiteration of row-5 ultimately creates a non-permissive zone or a barrier, which prevents the extension of interneuronal longitudinal tracts along their normal anterior-posterior path. While we do not know the nature of the barrier, the axon tracts either stall when they reach this region or project across the midline or towards the periphery along this zone. Previously, we had shown that midline ensures ancestry-dependent fate specification in a neuronal lineage. These results provide the molecular basis for the axon guidance defects in midline mutants and the significance of proper specification of the environment to axon-guidance. These results also reveal the importance of segmental polarity in guiding axons from one segment to the next, and a link between establishment of broad segmental identity and axon guidance. In the Drosophila nerve cord, about 20 longitudinal axon tracts on either side of the midline, each consisting of axons from several neurons, connect different segments with one another. Several direct players in axon guidance have been identified. For example, previous studies have shown that mutations in two signaling pathways, the ligand Slit (Sli) and its receptors Roundabouts (Robo, Robo2 and Robo3) and the ligand Netrin (Net) and its receptor Frazzled (Fra; the vertebrate homologue is known as Deleted in Colorectal cancers or DCC) disrupt the precise positioning of these tracts by altering their growth cone guidance [1]–[7]. Whenever the Slit system is disrupted, longitudinal axon tracts inappropriately cross the midline [1], whereas with the disruption of the Net-Fra system, which primarily mediates the attraction of commissural tracts to facilitate their midline crossing [4]–[6], a large number of commissural growth cones fail to cross the midline [4], [5], [8]. There is a second set of players not linked to the direct players such as Slit-Robo or Net-Fra, but cause axon guidance defects when disrupted. In these mutants, the pioneering axon growth cones fail, either due to the absence of the neurons themselves or due to a mis-specification of their identity. As a result, follower neurons fail to properly project their growth cones along the correct trajectories. For instance, when the pioneering neurons pCC or vMP2 are either ablated [9] or mis-specified [10], the follower axon tracts cross the midline, ignoring the guidance cues mediated by Slit and Robo [10]. It is obvious that the environment in which growth cones travel would have an impact on axon guidance. However, it is not clear in what specific way the environment in which axons travel influence axon guidance or how specific the influence would be on axon guidance. The environment is defined by cells, which express guidance determinants on their surface or release cues into the extracellular matrix. Segmentation genes, in particular segment polarity genes, broadly define the environment in which axons travel by specifying cellular identity, which then by expressing specific genes regulate guidance of specific growth cones. Segment polarity genes are expressed in rows and columns within the nerve cord and mutational analysis indicates that they specify the initial NB identity along the rows and columns [11]–[15]. For instance, row 5 identity is set mainly by the expression of Wg and Gsb (all row 5 cells express these genes), whereas row 4 is determined by the expression of Patched (Ptc) in row 4, Wg in row 5, and the absence of expression of Gsb in row 4 [reviewed in ref. 15]. Loss of function for these genes alters the identity of NBs along the entire rows. Thus, loss of function for Ptc changes row 4 into row 5, loss of Gsb changes row 5 into row 4, and loss of Wg alters rows 5,6 and 4 identities (non-cell autonomous function of Wg also confers row identity to adjacent rows) [11]–[15]. Their expression persists in successive divisions of NBs, even as NB-specific expression of transcription factors changes following each division of a NB [16], [17]. Loss of function for these genes also cause axon guidance defects [18]. However, we do not know if the axon guidance defects in segmentation mutants are due to mis-specification of a pioneering neuronal identity, or broad changes in the environment in which axons travel (or both). Given that growth cones interact with the environmental niche along their path, it is reasonable to suppose that broad changes in the local environment can affect axon pathways. However, separating neuronal identity from changes in the environment in influencing axon guidance has been experimentally difficult. We have been studying a gene called midline (mid), which belongs to a class of transcription factors known as T-box binding (Tbx) proteins. Tbx proteins are highly conserved among metazoans and are defined by the presence of a T-box domain, a 180–230 amino acid DNA-binding domain. Tbx proteins bind to a T-Box element (TBE), a 20-bp degenerate palindromic sequence [19]. However, TBEs are highly variable in sequence, number and distribution within promoters and Tbx proteins diverge significantly in their sequence preference [20]. Tbx proteins are known to repress transcription [21]. Moreover, mutations in Tbx genes can be haploinsufficient, i. e. developmental processes are sensitive to the levels of some Tbx proteins. For example, upper limb malformation and congenital heart defects in Holt-Oram syndrome are due to haploinsufficiency for TBX5 [22], [23]. Haploinsufficiency for mouse brachyury and human TBX3 and TBX1 genes causes dominant phenotypes such as short tails/tailless, Ulnar-Mammary syndrome and DiGeorge syndrome, respectively [23], [24]. In Drosophila, loss of function for mid (also known as lost in space or los, or extra) was initially shown to cause cuticle defects in the midline region of the embryo, thus the name midline [25]. Subsequently, it has been shown that mid mutants also cause heart defects [26], defects in the lateral chordotonal axons, and shorter and defasciculated dorsally routed axons in the peripheral nervous system (PNS) [5]. We recently showed that Mid ensures ancestry-dependent fate specification of a GMC, i. e, fate of a GMC is changed without affecting the parent NB identity, thus overriding the GMC' s ancestry [20]. Thus, in mid mutants, a GMC from an unrelated NB (we have named it the M-lineage, M for Mid) changes into GMC-1 (also known as GMC4-2a) of the RP2/sib lineage without altering the parent NB identity. Also, this occurs several hours after the window of time in which the bona fide GMC-1 of the RP2/sib cells is formed. A subsequent study by Liu et al. [27] reported that ectopic expression of mid in salivary gland can ectopically induce expression of robo, slit, Netrin and frazzled. The implication is that Mid regulates axon guidance via regulation of these guidance genes and that the axon guidance defects observed in mid loss of function mutants are due to loss of expression of these guidance genes. However, the regulation of these genes by Mid in salivary glands, where none of these axon guidance genes including mid are normally expressed, is of no functional significance. Mid must regulate these genes in the nerve cord to be of relevance. Moreover, the mostly non-overlapping expression patterns of mid, robo and slit in the developing CNS, save a few cells in the lateral region of the nerve cord as reported by Liu et al [27] does not make sense with a functional direct transcriptional regulatory role for Mid on these genes during axon guidance. We sought to explore these issues, including the possibility of an indirect regulation of axon guidance genes by mid, with the aim to understand the molecular basis for the guidance defects in mid mutants. We show here that the primary axon guidance defect in mid mutant embryos is stalling of axon tracts midway between the posterior commissure (PC) and the anterior commissure (AC) of the next segment, with tracts often crossing the midline, or projecting peripherally outward, perpendicular to the midline. This defect is due to the transformation of row 2 NBs and their precursor neuroectodermal (NE) cells, which are located midway between the PC and the AC of the next segment, into row 5 cells. Row 5 is normally located at the level of the PC and defines the parasegmental boundary (PSB). The fact that axon tracts stall or project across the midline or towards the periphery precisely along this transformed row, indicates that these newly re-specified row 5 cells creates an unsuitable or inhibitory niche for these pioneering axons to navigate along the midline. These results argue that the role of Mid in regulating axon guidance is indirect and via proper specification of row identity within the nerve cord. Our results also show that Mid does not regulate transcription of frazzled, sli or robo, directly or indirectly, in cells where their expression matters. These results provide novel insight into how segmentation or row identity facilitate axon guidance later in neurogenesis and distinguishes how broad environmental identities, as opposed to individual neuronal identity, influence axon guidance. Previous results have indicated that embryos mutant for mid show axon guidance defects [5]. We sought to examine in detail the axon guidance defects in mid mutants in the embryonic CNS during development and compare those defects to the defects at corresponding developmental stages in slit and robo mutants. As shown in Fig. 1, embryos of different developmental stages were stained for Fasciclin II (Fas II) positive axon pathways using an antibody against Fas II. Fas II staining of ∼9 hours old embryos reveals the nascent medial tract, which is closest to the midline and is pioneered by the growth cone from pCC (arrows in Fig. 1A, wild type). In ∼9 hours old mid deficiency embryos the pCC growth cones were the same as in wild type, projecting slightly outward and then parallel to the midline (Fig. 1D, arrows). However, in ∼9 hours old slit mutant embryos, the pCC growth cones projected directly towards the midline (arrows in Fig. 1G). In ∼9 hours old robo mutant embryos, the pCC growth cones also projected towards the midline, although the defects were less severe than in slit embryos (Fig. 1J, arrows). By 10 hours of development, the growth cones from pCC in mid embryos were projecting outward and away from the midline as if they had encountered an inhibitory zone (Fig. 1E, arrows, compare with wild type, Fig. 1B), whereas in ∼10 hours old slit mutant embryos, the growth cones from pCC were fasciculated with each other at the midline (Fig. 1H). By ∼14 hours of development in mid embryos, the three different Fas II tracts, the medial tract, the intermediate tract and the lateral tract, all run parallel to the midline, could be seen with similar spacing between each other and from the midline, as in wild type (Fig. 1F to 1C). However, as shown in Fig. 1F, in ∼14 hours old mid mutant embryos we could observe tracts inappropriately projecting outward (thick arrow), breaks or missing tracts along the longitudinal axis (arrowhead) and crossing the midline (midline arrow). We could also observe stalled growth cones forming a blob of axon tracts along the nerve cord (Fig. 1C, star, see also Table 1). While in ∼14 hours old slit mutant embryos the three tracts were all collapsed at the midline (Fig. 1I), in robo mutants, the medial tract was mostly fused at the midline, with the other two tracts more or less normal (Fig. 1K) (the partial penetrance of the guidance defects is due to redundancy with Robo2 and Robo3 receptors) [2], [3], . The frequency of various guidance defects in mid, slit and robo mutants are presented in Table 1. These results indicate that axon guidance defects in mid mutants are significantly different from axon guidance defects in slit and robo mutants. If Mid regulates axon guidance via regulating slit and robo, the guidance phenotypes in all the three mutants should fall more or less into the same general category. Our above results show that this is not the case and argues against the possibility that Mid regulates slit and robo and that the axon guidance defects in mid mutants are due loss of function for these axon guidance genes. We next sought to determine the growth cone projections from vMP2, dMP2, MP1, pCC and aCC neurons in mid mutant embryos using more selective markers. We chose to examine the growth cones from these neurons since these neurons send out pioneering growth cones. For example, the anteriorly projecting growth cones from vMP2 and pCC pioneer the medial Fas II tract to meet the homologous axons from the next anterior segment [9], [10]. Similarly, the posteriorly projecting growth cones from MP1 and dMP2 pioneer the lateral Fas II tract to meet up with the homologous axons from the next posterior segment. Therefore, first we stained mutant embryos with a monoclonal antibody 22C10, which is raised against MAPIB. In a ∼10 hours old embryo, vMP2 (Fig. 2A) projects its growth cone anteriorly (arrow), while dMP2 projects posteriorly (arrow) (Fig. 2A). By ∼11. 5 hours of development, 22C10 antibody staining revealed a fasciculated, more mature medial tract (Fig. 2B, upper arrow) and lateral tract (lower arrow), as well as several other axon pathways including the motor pathway of the aCC and RP2 neurons, both fasciculated together to form the intersegmental nerve bundle before exiting the CNS (smaller arrow). In mid mutant embryos, both vMP2 and dMP2 neurons are normally formed, but we observed two key defects in their projection pattern: the growth cones often projected away in a posterior-lateral pathway similar to and/or sometimes part of the aCC-pioneered intersegmental nerve bundle (Fig. 2C, top, left arrow with star). The projections were either stalled or projected away like a motor pathway (Fig. 2C, D, E, arrow and arrow with a star). These aberrant projection patterns suggest that these growth cones have come upon a non-permissive barrier or a zone of repulsion and cannot travel in their normal path. They either stall and or project away. We further examined the projection pattern from vMP2 by expressing mCD8-GFP (mCD8 targets GFP to membrane) using the achaete (ac) -GAL4 driver. While in the wild type the axon tract from vMP2 is projected along the midline (Fig. 2F, arrow), in the mutant the projection is diverted away and perpendicular to the midline in a pathway towards the periphery, often exiting the nerve cord (Fig. 2G, arrow with star). We next examined the projection from MP1 by expressing tau-GFP (tau directs GFP to microtubules) using the sim-GAL4 driver. While in the wild type the axon tract from MP1 is projected along the midline (Fig. 2H arrow), in the mutant the projection is diverted towards the periphery, perpendicular to the midline (Fig. 2I, J, arrow with star). This aberrant projection defect in MP1 was highly penetrant and severe. We next examined the growth cone projection from pCC by expressing UAS-tau-lacZ transgene in pCC neuron using the RN2-GAL4 driver. This driver drives the tau-lacZ in pCC neuron (Fig. 2K, L; it also drives in aCC and RP2, Fig. 2M, N). As shown in Fig. 2K, in the wild type the pCC projects its axon anteriorly along the midline (arrow). However, in the mutant, the projection is diverted away towards the periphery perpendicular to the midline (Fig. 2L, arrow with star). We also examined the two motor pathways from neurons aCC and RP2, but did not observe any defects in their pathfinding (Fig. 2N). These results indicate that the defects are mostly confined to axon tracts from interneurons. These defects are unlikely due to a negative effect on axon growth, instead, the projections appear to encounter a barrier in their normal path and travel in an aberrant path as defined by this barrier. The above results show that the axon guidance defects in mid are fundamentally different from those in slit or robo mutant embryos. However, given the recent report that Mid ectopically regulates sli and robo transcription in salivary glands [27], we sought to examine mid mutant embryos for the expression of these genes in cells where they are normally expressed. If one of the functions of Mid in wild type is to regulate expression of slit and robo genes, a significant reduction in the levels of Sli and Robo proteins should be observed in their respective domains in loss of function mid mutant embryos. First, we stained mid mutant embryos with a Slit antibody. As shown in Fig. 3 (A, C and E) in wild type, Slit is present at high levels in midline glial cells where the axon tracts of AC and PC cross the midline. It is also present in commissural and longitudinal tracts due to movement of Slit from the midline to the axon tracts via the commissural tracts [7]. We examined the two alleles of mid (mid1 and los1) (Fig. 3B and D) and the mid H15 deficiency (which removes both mid and its sister gene H15) (Fig. 3F). We have reported previously [20] that the mid1 allele has a stop codon at amino acid (aa) position 128 (this allele is likely the strongest loss of function mid mutant allele) and los1 has a 22 base-pair deletion resulting in a deletion of 7 aa at position 321, as well as a frame shift leading to a stop codon at aa 350 (thus, in addition to the truncation the mutant protein in this allele has 28 amino acids that are entirely different from the wild type gene; this has the potential to cause gain of function/neomorphic effects in addition to loss of function effects). The Slit protein level was not significantly affected in homozygous mid1 allele nor was it affected in the homozygous mid H15 deficiency embryos (Fig. 3B and 3F); a marginal reduction in the levels of Slit protein was observed in los1 embryos in the PC region (Fig. 3D). Whether this is due to a los1-specific gain of function effect or a background effect is not known. We further examined if the levels of the Slit protein is affected in younger stage embryos from mid1, los1, and the mid H15 deficiency. However, no reduction in the levels of the Slit protein was observed in these alleles (data not shown). To quantify the level of Slit between wild type and the mutant embryos, we performed Western analysis of Slit in the mid H15 deficiency embryos. The results reproducibly showed only a marginal reduction in the amount of Slit (Fig. 3G). One possibility for this slight reduction in Slit protein levels is that Liu et al [27] had reported that there is an overlapping expression of Mid and Slit in a few neurons located laterally within the nerve cord. It is possible that Mid regulates slit expression in these cells and that the slight reduction on Westerns reflect this regulation. Alternatively, the slight reduction in the levels as seen in Western blots is due to indirect effect of loss of function for mid and H15 genes, such as mis-specification of relevant neurons/glia. Since Mid is a transcription factor, we next sought to determine if the transcription of the slit gene is affected in mid mutant embryos by performing whole mount slit RNA in situ. If Mid regulates slit transcription at least in the PC region, where mid is expressed, we should observe loss of slit transcription in these midline cells in mid mutant embryos. However, as shown in Fig. 3J, K and L), no such effect on the transcription of the slit gene by loss of function for mid was observed. We next examined the expression of Robo in mid1, mid H15 deficiency, and in embryos transheterozygous for the mid H15 deficiency and mid1 alleles using an antibody against Robo (Robo levels were also examined in los1 allele, see later section). As shown in Fig. 4A, in wild type Robo is expressed in longitudinal pathways and is also present very weakly in AC and PC (due to incomplete down-regulation of Robo by a Commissureless protein-mediated process in commissural tracts [1]). In mid mutant embryonic CNS, the levels of Robo was not affected in any significant way (Fig. 4B); the lack of Robo staining in tracts (arrows, Fig. 4B) is due to the absence of axon tracts themselves. We also examined the expression of Robo in mid H15 deficiency embryos by Western analysis, which indicated a slight reduction in the levels of the Robo protein relative to wild type (Fig. 4G). This reduction is likely due to a secondary effect originating from the breaks in axon tracts or loss of Robo-expressing cells [due to identity changes, see ref. 20] as opposed to a direct Mid regulation of robo. We also examined the transcription of sli, robo and frazzled (fra) in mid H15 deficiency embryos using the sensitive qRT-PCR method. As shown in Fig. 5, no significant differences were detected in the transcription of any of these genes in mid H15 mutant embryos compared to wild type. These qPCR results were reproducible using three different samples of embryo RNA preparations prepared separately in three different days, and qPCR was done in triplicates for each of the samples (the averages with standard errors were shown in Fig. 5). These results suggest that Mid has no role in the transcription of these genes during neurogenesis (note that there is no maternal contribution of mid). Finally, Liu et al [27] had suggested that mid and fra show transheterozygous genetic interaction since they found that embryos transheterozygous for mid and fra have strong axon guidance defects. We re-examined if the two mutations show such an interaction by staining mid/+, fra/+ embryos from mid/CyO and fra/CyO with Fas II and BP102 antibodies. However, we did not observe any axon guidance defects in these embryos. Sometimes, balancer-bearing parents generate a few embryos that show axon guidance (or other) defects. We have previously named this ‘balancer-induced parental effect’ [7], [20]. This effect can also be suppressive. Therefore, we generated transheterozygous embryos from mid and fra parents that do not carry any balancers (mid/+ and fra/+). The transheterozygous embryos from this cross also did not have any axon guidance defects (Table 1). We did not find any axon guidance defects in embryos transheterozygous for the mid H15 deficiency and fra as well (Table 1). Similarly, no transheterozygous interaction between slit and mid was observed (Table 1). Therefore, we conclude that no transheterozygous interaction occurs between mid and fra or between slit and mid. We next sought to determine the molecular basis for the axon guidance defects in mid mutant embryos. Our results show that in mid mutant embryos some of the interneuronal pathways that normally project along the midline stall between PC and AC of the next segment and then get redirected across the midline or away towards the periphery, perpendicular to the midline (there are variations to this phenotype but the spectrum of such variations are all within this category). NBs are formed in waves (S1–S5) and in rows (1–7) under the control of neurogenic and proneural genes. Previous studies have shown that many of the segmentation genes, especially segment polarity genes, are expressed row-wise in NE and NB cells. These genes play a crucial role in the row-wise specification of NB identity [reviewed in ref. 15]. To determine if the row-wise cellular identity within the nerve cord is altered in mid mutants, which might underlie the inhibitory zone and the associated guidance phenotype, we sought to examine the expression of some of the segment polarity genes. First, we examined mid mutant embryos for the expression of Wingless (Wg or W in Fig. 6K) and Gooseberry (Gsb and G in Fig. 6K). In wild type, Wg is present in row 5 NBs and the corresponding NE cells (Fig. 6A, B, K, see also Bhat, 1998). In mid mutant embryos, row 5 NB or NE expression of Wg was not affected, however, we observed ectopic expression of Wg in row 2 NBs and the corresponding NE cells (Fig. 6C–H). This ectopic expression was often stronger in alternate segments (see Fig. 6C, D, E). We note that the extent of ectopic expression of Wg was variable from segment to segment. For example, we found hemisegments or segments in which large patches of cells in the region between row 5 and row 7 (of the preceding segment) expressing ectopic Wg (Fig. 6G and H), which can also explain some of the variations in the guidance defects. Nonetheless, these results indicate that cells in row 2 behave as if they were row 5 cells. That this mis-expression occurs during segmentation is also indicated by the cuticle defects seen in mid mutant embryos, with missing denticle belts in the corresponding region (see Text S1 and Fig. S1). Consistent with the above interpretation of Wg results, Gsb expression was also mis-expressed in mid mutant embryos. In wild type, Gsb is expressed in rows 5,6 and one NB in row 7 (NB7-1). In mid mutant embryos, while the normal expression of Gsb in rows 5,6 and NB7-1 was not affected, we observed ectopic expression of Gsb in the same cells expressing ectopic Wg (Fig. 6E). However, unlike the ectopic Wg stripe, which was always present in the mutant embryos at detectable levels, the ectopic expression of Gsb in the stripe was often incomplete and at times undetectable. Occasionally, we observed strong ectopic Gsb expression corresponding to both row 5 and row 6 cells suggesting that in mid mutants in addition to row 2 cells changing into row 5 cells, some row 3 cells may change into row 6. Though infrequent, in such segments it appears there is a reiteration of row 5 and 6 (rows 1,5, 6,4, 5,6, 7) to varying degrees within the nerve cord in mid mutant embryos. We next stained the mutant embryos for the expression of Sloppy-paired (Slp). We decided to examine Slp expression since in wild type Slp is expressed in rows 4 and 5 [Fig. 6I; see also ref. 13] and a change in Slp expression in mid mutants would allow us to confirm the results from the Wg and Gsb staining. This would also help us determine if cells corresponding to row 4 have changed to some other row of cells. In mid mutant or deficiency embryos, we observed ectopic expression of Slp in cells corresponding to row 2 cells (possibly some cells from row 3) (Fig. 6J). However, the ectopic expression of Slp was stronger in those segments where ectopic Wg was also strongly expressed. Again, the ectopic Slp expression was incomplete compared to ectopic Wg. Nonetheless, these results show that multiple row 5-specific segmentation genes are expressed in row 2 cells in mid mutant embryonic CNS. Our previous results have shown that Mid is strongly expressed in row 7 and row 1 cells as well as in corresponding midline cells [20]. Since the expression of key genes can change quickly from division to division in NBs, and is highly time-sensitive [16], [17], we re-examined wild type embryos with an antibody against Mid. As shown in Fig. 6L–O, we found that Mid is indeed expressed at low levels in a large number of NBs, including in rows 2,3 and 4 (perhaps also in one NB in row 5). Except for the strong expression in row 7 and row 1, which remained unchanged during neurogenesis, the expression pattern of Mid in other NBs changed as neurogenesis proceeded (Fig. 6L–O). If we stain wild type Drosophila embryos with a monoclonal antibody BP102, we can clearly visualize commissural architecture with the longitudinal axon tracts (LC) and the anterior and posterior commissures (AC and PC; see Fig. 7A, green and Fig. 7B). Unlike the Fas II or other markers examined in the preceding sections, which are all directed against a small number of axon tracts, BP102 recognizes many more CNS axons and provides a more complete picture of axon tracts within the nerve cord. Therefore, we double stained embryos with BP102 and an antibody against Even-skipped (Eve) to determine the position of certain Eve-positive neurons (and therefore their parent NBs) in relation to the commissural architecture. Eve staining shows that an RP2 neuro, which is generated by NB4-2 (a row 4 NB), is located at the inner armpit of AC [Fig. 7A; see also ref. 28], affirming the position of row 4 NBs at the level of the posterior border of AC. While RP2 undergoes a complex migration within the nerve cord during development, ultimately it settles down in the same row where its parent NB is formed [28]. Similarly, the Eve-positive aCC/pCC neurons are located at the inner armpit of PC (Fig. 7A), which are generated by NB1-1, thus, fixing the location of row 1 to the posterior border of PC. Unlike the RP2, aCC/pCC neurons do not undergo much migration and stay in the same row 1 [28]. Thus, although NBs generate numerous progeny and there is both germ band retraction and condensation of the nerve cord, the relative position of commissures at a later point in neurogenesis to early NB rows remain more or less stable. We next stained embryos from different alleles of mid (los1, mid1, mid H15 df and los1/mid, H15 df) with BP102 (Fig. 7C–H) to visualize the commissural architecture in mutant embryos (we did not double stain mutant embryos with BP102 and Eve in order to be able to flatten the nerve cord to fully visualize the commissural architecture and also to maximize the number of mutant embryos examined; the double staining shown in Fig. 7A was done to determine the relative position of AC and PC to rows of NBs). As shown in Fig. 7C–H, significant disruption of the commissural tracts was observed in all the mutant alleles of mid. We could clearly observe blobs of tracts or tracts projecting laterally (black arrow in 7C, D–G) at the level of AC, breaks in LC, as well as mis-projection of commissures between two adjacent neuromeres, creating a criss-cross phenotype (panel 7E; this criss-cross phenotype was observed in other alleles/combinations as well, data not shown). These tracts appear to encounter a non-permissive region for LC projection at row 2/3 (which lies just above AC). Furthermore, the posterior commissural (PC) tracts are reduced to only a few axon pathways in nearly all commissures, indicating a loss of axon tracts in PC. This may be a secondary effect of stalling of axon tracts in preceding neuromeres, thus, reducing the number of axon tracts that cross the midline through PC. The anterior commissural (AC) tracts were also affected but to a lesser degree. In general, in all mid alleles, more than 80% of the hemisegments had longitudinal tracts stalled at the AC level, resulting in breaks above AC. However, it appears that los1 has slightly more severe overall CNS defects compared to other alleles or the deficiency and this appears to be the case in embryos transheterozygous for los1/Df as well. This may be consistent with the possibility that this specific allele has some gain of function effects given the molecular lesion in the gene in this allele [20]. Nevertheless, the defects were similar in all alleles. Although Robo is present at very low levels in commissural tracts [Robo is down–regulated in commissures, see ref. 1], the Robo-staining pattern closely resembles that of BP102, minus the strong commissural staining of BP102 (Fig. 7I). With Robo staining of mid mutants, we could observe that the longitudinal axon tracts stall at the AC level in all mid mutants (see Fig. 4 also). This corresponds to row 3 NBs in wild type, which is just before row 2 (re-specified as row 5 in the mutant). It appears that, when longitudinal axon tracts encounter the re-specified row 5 cells in mid mutants, they stop and simply congregate at this position, forming a blob (Fig. 7J, see also C–H, indicated by star). This is also evident by the breaks in the continuity of longitudinal tracts (Fig. 7J, arrowhead). These defects are consistent with the presence of a region or a barrier above AC that is not permissive to longitudinal axon projection. These results argue that loss of function for mid alters the identity of rows of NE and NB cells. By the time neurons begin to project their growth cones, this change of row identity creates a zone which is either non-permissive or lacks signals for growth cones to continue in their usual path. Thus, these growth cones either collapse on to themselves or project laterally outward, or in some segments/hemisegments cross the midline in this region (see Fig. 2 also). Our results show that Mid does not regulate the expression of slit, robo or fra genes in the CNS. Consistent with this, the axon guidance defects in mid are distinct from the defects in slit or robo mutants. We confirm this by several different ways: immunostaining, RNA whole mount in situ, Western analysis, qPCR and genetic interaction studies. A previous study has suggested that Mid regulates sli, robo and fra [27]. They based their conclusion on finding a strong transheterozygous genetic interaction between mid and sli, and mid and fra, detected using BP102 staining of embryos that are transheterozygous for mid1 and sliGA20 and mid1 and fra3. Furthermore, they reported that levels of fra mRNA and Fra and Robo proteins in mid mutant embryos were down regulated, and that this can be completely rescued by expressing mid using elev-GAL4 driver [27]. They also reported that ectopic expression of mid in salivary glands induces expression of robo and slit. We have not found most of these effects reported by Liu et al [27]. For example, we failed to observe any genetic transheterozygous interactions between mid and sli mutants (Table 1). Transheterozygous interactions are rare given the negative evolutionary impact of such interactions to survival, but when observed, it is usually with mutations in receptor-ligand pairs, or with gain of function/neomorphic situations). We used stronger allelic combinations than the ones used by Liu et al [27] with mid and sli. For mid, we used not only mid1, but also a deficiency that removes both mid and its sister gene H15, as well as los1. For slit, we used sli2, which is the strongest loss of function allele and genetically behaves as a null. Furthermore, we also failed to observe any such transheterozygous interactions between mid and fra (Table 1). Secondly, we found that while the ectopic expression of mid in salivary glands induced robo expression as was reported by Liu et al [27], no such induction was observed with slit (Text S1 and Fig. S2). One should also consider the fact that Mid, Robo (and Slit) have mostly non-overlapping domains of expression in the CNS, therefore, the direct regulation of robo by Mid in the salivary gland has little relevance in the CNS or CNS development mediated by Robo or Mid. Our results indeed bears this out. Not only the axon guidance defects are different between mid and slit or robo, the transcription of robo, slit or fra are also unaffected in mid mutants (Figs. 3–5, Table 1). There was some reduction in the levels of Slit in los1 allele in the midline in the PC region. But, the molecular lesion in los1 is complex and might have some allele-specific gain of function effects that alters cellular identity or function of the corresponding midline glia to mediate reduction in the Slit level in this region. Since no such reduction in the levels are seen in other mid alleles and more importantly the transcription of slit is unaffected in the deficiency that removes mid (and H15), we think that the slight reduction in the levels of Slit in los1 is allele-specific. In the case of robo, the promoter has three TBEs. With three sites, Mid is more likely to be able to induce robo in an ectopic site. However, within the CNS, we did not find any significant loss of Robo expression in mid mutant embryos by immunohistochemistry (either in los1 or mid H15 deficiency embryos) or by Western analysis (Fig. 4) or robo transcription by qPCR (Fig. 5). A slight reduction in the levels of Robo seen in Westerns is likely due to a secondary effect of loss of tracts and perhaps loss of some of the Robo-expressing neurons perhaps due to identity changes [19]. The reason for the significant reduction in the expression of Robo in mid observed by Liu et al [27] is not clear. We think that this may be due to some technical reasons such as variability from embryo to embryo to fixing and staining. Because of this possibility, we follow a simple rule: in this case, we focused on mid mutant embryos that had strong guidance defects to determine if such mutant embryos also had a strong or weak expression of Robo and Slit. We found that embryos with strong guidance defects also had strong Robo (or Slit) expression. Thus, we avoided selecting sub-stained mutant embryos and comparing them to optimally stained wild type embryos. Finally, Liu et al [27] reported that there is overlapping expression of Mid and Slit in a small number of cells located laterally within the nerve cord. It is possible that Mid regulates slit expression in these cells, however, the contribution of Slit or such a regulation of slit to the overall axon guidance mediated by Slit is not clear and likely very minimal, if there is any. We have also not examined if mid affects netrin gene expression. Our work shows that in mid mutants, the majority of axon growth cones of the longitudinal tracts stall and club together at the level of AC, creating a blob of axons, thus leading to interruptions between neuromeres. Interestingly, some of the tracts project outward towards the periphery or inward across the midline (see Table 1). This outward projection route is quite revealing: the projection path is mostly perpendicular to the midline and just below the transformed row of cells. The transformed row of cells corresponds to the region right above the AC or where the tracts stall (Figs. 6 and 7). The most consistent change is seen with row 2 cells, changing into row 5 cells. How does these changes relate to wild type? In wild type, row 5 cells normally separate one neuromere from the next and also define PSB. Thus, the change from row 2 to row 5 in mid must be creating an environment that either lacks the necessary permissive/attractive cues or possess cues that are inhibitory to the projection of these axon tracts, causing the tracts to stall. For example, in wild type row 5 cells are located between pCC and vMP2, the two axons that pioneer the Fas II-positive medial tract. The growth cone from vMP2 in wild type only marginally encounters row 5 cells but does not necessarily traverse row 5. This is due to the fact that vMP2 is located in row 5 and the growth cone from a vMP2 stops at row 5 region and fasciculates with the vMP2 of the next hemisegment. However, it does encounter row 2 cells midway through the projection path. In mid mutants, since row 2 cells change to row 5, creating a region that vMP2 growth cone is perhaps normally programmed to stop. For the proper guidance of medial tract, normal projection of vMP2 and pCC is necessary and loss of either of the two pioneer neurons causes aberrant medial tract guidance [9], [10]. Therefore, it seems likely that vMP2 stalls and the pCC projection, along with several other follower projections, also stalls; or some of the tracts project across the midline or away towards the periphery. In fact, these abnormal projection patterns, especially towards the periphery appear to be guided by the newly created barrier. This situation is also the same for MP1 or dMP2. What is the mechanism within the re-specified row of cells that eventually mediates the block for axon projection? The re-specified rows of cells would have a whole set of new (row 5-specific) genetic programs that may simply not conducive to longitudinal tracts. Additionally, the role of Ephrin pathway in axon guidance may be relevant here. The Drosophila Ephrin (Eph), which is a transmembrane protein, is shown to prevent interneuronal axons from exiting the Drosophila embryonic CNS [29]; some of the interneuronal pathways in mid mutant exit the nerve cord (Fig. 2G). Ephrin/Eph signaling is via cell-to-cell contact and depends on the clustering of Eph receptors and their ligands [30]–[33]. This multimerization activates the kinase activity of the receptor and leads to the phosphorylation of the receptor within the cytoplasm-exposed tail region and the binding of downstream effectors [34]. This triggers the depolymerization of actin in growth cones, modifying the Integrin-based cell adhesion [29], [35]. The CNS-exiting phenotype of interneuronal pathways in mid mutants suggests a possible de-regulation of the Eph-pathway. But, it may also be that changes in Eph or similar cell-adhesion mechanism mediate the formation of the barrier and exiting of some of the interneuronal pathways from the CNS. Our previous results show that Mid acts as a transcriptional repressor of gsb-n [20]. However, in mid mutants the transformation of row 2 into row 5 also activates Gsb expression (Fig. 6). The ectopic activation of Gsb in these cells in mid, however, is not a direct de-repression of gsb, but an indirect consequence of the transformation of cell identity from row 2 (a Gsb-negative row of cells) to row 5, a Gsb-positive row. Finally, our results provide clear evidence that segmentation genes can regulate axon guidance via broadly defining cellular identity, creating a permissive and non-permissive boundaries or niche. We also emphasize that extrapolating expression relationships to functional relevance from induction in ectopic sites, in vitro and tissue culture experiments, bioinformatics or other similar in vitro studies carry inherent risks and should be done with caution. mid mutant alleles used were mid1, mid2 and los1. We also used a deficiency that removes both mid and H15 genes (mid H15df; BL# 7498: breakpoints: 25D5-25E6). The other lines used were: sli2, robo4, robo-deficiency [Df (2R) BSC787, breakpoints: 58F4-59B1; BL#27359], UAS-mid, sim-GAL4, sgs-3-GAL4 (to induce mid in the salivary gland), ac-GAL4, UAS-mCD8-GFP, RN2-GAL4 (eve-GAL4) and UAS-tau-lacZ. For wild type, we used Oregon R flies. All the mutant lines were balanced using GFP-bearing balancer chromosomes to facilitate identification of the mutant genotype. Transgenic UAS-mid fly lines containing one or two copies of the UAS-mid were previously generated in the lab [20]. The transgenic flies were crossed to sim-GAL4. Embryo collection was done overnight at 28°C. The embryos were fixed, divided into three portions and stained separately with antibodies against Mid, Robo and Slit. ac-GAL4 driver (BL#: 8715) and the UAS-mCD8-GFP (BL# 41803) were introduced into the mid H15 deficiency background and the embryos were stained for GFP and Odd-skipped. ac-GAL4 drives the UAS-mCD8-GFP in vMP2 and dMP2 and their axons. Odd-skipped is expressed in dMP2 and MP1. RN2-GAL4 (portion of the eve promoter that drives expression in aCC/pCC and RP2 neurons) and UAS-tau-lacZ were introduced into mid H15 deficiency background and the embryos were stained for LacZ. sim-GAL4 and UAS-tau-GFP introduced to mid H15 deficiency background and stained for GFP and Eve (Eve to identify the mutant). Tau and MCD8 targets GFP to axon tracts. The embryo collection, fixation and immunostaining were performed according to the standard procedures. The following antibodies were used: anti-Sli C (1∶20, DHSB), anti-Robo (1∶5, DHSB), anti-Mid [1∶50, generated in the lab, see ref. 20], anti- Fas II (1∶5, DHSB), 22C10 (1∶1, DHSB), anti-Wg (1∶5), anti-Gsb (1∶3), anti-Slp (1∶400), anti-GFP (1∶300), anti-Odd (1∶500), anti-Eve (1∶2000), anti-Lac Z (1∶500). For color visualization, either AP-conjugated or HRP-conjugated secondary antibodies were used. For double staining, secondary antibodies conjugated with AlexaFluor488 and AlexaFluor635 were used. Whole-mount RNA in situ hybridization for sli expression was done following the standard procedure using a digoxigenin-labeled slit probe, synthesized by PCR. Cuticle preparation was done as per standard procedure by fixing embryos and dissolving organic embryonic material on slides using Hoyer' s solution at 65°C for 24 hours. For western blot analysis, 30 embryos were collected (homozygous mutant embryos were identified by the lack of GFP expression under microscope), homogenized in 37. 5 µL lysis buffer (0. 15 M NaCl, 0. 02 M Tris pH, 7. 5,0. 001M EDTA, 0. 001 M MgCl2,1% Triton-X-100 and PIC) and kept on ice for 10 minutes. The lysed protein is centrifuged for 5 minutes at 13,000 rpm, the supernatant is collected and diluted with 12. 5 µL 4× Laemelli sample buffer. The protein sample is boiled in water for 10 minutes and kept in 4°C for 10 minutes. Equal amount of lysed protein 20 µL (15 embryos per lane) was loaded on to a 4–12% SDS-PAGE gel. The separated proteins were transferred to a Nitrocellulose membrane. The efficiency of transfer was determined by Ponceau S staining. The membrane is blocked in 5% milk for 2 hours at room temperature, and incubated with primary antibodies (anti-Slit N 1∶50000 or anti-Robo 1∶40) overnight at 4°C and washed with PBST (PBS+0. 02% Tween 20). The Nitrocellulose membrane was then incubated with HRP-conjugated secondary antibody (anti-Rabbit 1∶20000 or anti-mouse 1∶20000) for 2 hours at room temperature and washed with the washing buffer. Proteins were detected by the chemi-luminiscent ECL reaction method (Thermo Scientific). The autoradiographs were scanned and intensities of bands were analyzed using the software AlphaEaseFC. Anti-Tubulin antibody (Abcam, 1∶4000) was used for determining loading control. Embryos from Oregon R (wild type) and mid H15 deficiency lines were collected and aged for 12–14 hours. They were dechorionated in 50% bleach and washed with water. Approximately 150 embryos were selected under microscope for each sample. Total RNA isolation from these embryos were performed using the RNaqueous Kit (Ambion). The isolated RNA was DNase treated and quantified using Nanodrop Spectrophotometer (Nanodrop Technologies) and qualified by analysis on RNA Nanochip using Agilent 2100 Bioanalyzer (Agilent Technologies). Synthesis of cDNA was performed with 1 µg of total RNA in a 20 µL reaction using the Taqman Reverse Transcription Reagents Kit (ABI). Reaction conditions were as follows: 25°C, 10 minutes, 48°C, 30 minutes and 95°C, 5 minutes. Primers for real-time PCR were designed and made by the Molecular Genomic Core facility at UTMB. Real-time PCR were done using 1. 0 µL of cDNA in a total volume of 20 µL using the Faststart Universal SYBR green Master Mix (Roche, #04913850001). RpL32 was used as endogenous control. All PCR assays were performed in the ABI Prism 7500 Sequence Detection System and the conditions were as follows: 50°C, 2 min, 95°C, 10 min, 40 cycles of 95°C, 15 sec and 60°C, 1 min. Primers used: slit: Forward: 5′-GCGTTATGCCCGGTTCC-3′, Reverse: 5′ TCCACAACGTGCCGCTC-3′); robo: Forward: 5-CAGCATTAGTCTTCGTTGGGC-3, Reverse: 5-AATCCAACCAGTTTGCAGATTC-3); fra: Forward: 5-AGACCCCAGAGCATCCTTATG -3, Reverse: 5-TCTTTAGAGGATGGCCACGC-3. The qRT-PCR was done on three seperate embryo collections for each genotype and in triplicates for each collection.

During nervous system development, once formed from neuroblasts, neurons grow axons in the direction of their synaptic partners. Genetic forces guide these axon growth cones towards the target. This is known as axon guidance or pathfinding. There are a number of proteins that regulate axon-pathfinding. The well-known examples are the Slit and its receptor Roundabout, and Netrin and its receptor Frazzled. The Drosophila embryo and the nervous system are divided into segments by segmentation genes. Mutations in segmentation genes affect axon guidance, although how they do so is not well understood. In our work described here, we show that the T-box protein Midline prevents mis-specification of neuroblast rows, in particular, it prevents row 2 from becoming row 5. Thus, in midline mutants, row 2 changes into row 5, ultimately creating a non-permissive barrier that prevents axons from following their defined path. Thus, axons stop and diverge when they reach this barrier. Our results show how mutations in segmentation genes can affect axon guidance and how significant the environment is for axon-pathfinding. Our work is also a cautionary reminder that guidance defects need to be interpreted with care and can arise due to a variety of other defects.

Abstract Introduction Results Discussion Materials and Methods

2013

The Midline Protein Regulates Axon Guidance by Blocking the Reiteration of Neuroblast Rows within the Drosophila Ventral Nerve Cord

Multipotent neural crest (NC) progenitors generate an astonishing array of derivatives, including neuronal, skeletal components and pigment cells (chromatophores), but the molecular mechanisms allowing balanced selection of each fate remain unknown. In zebrafish, melanocytes, iridophores and xanthophores, the three chromatophore lineages, are thought to share progenitors and so lend themselves to investigating the complex gene regulatory networks (GRNs) underlying fate segregation of NC progenitors. Although the core GRN governing melanocyte specification has been previously established, those guiding iridophore and xanthophore development remain elusive. Here we focus on the iridophore GRN, where mutant phenotypes identify the transcription factors Sox10, Tfec and Mitfa and the receptor tyrosine kinase, Ltk, as key players. Here we present expression data, as well as loss and gain of function results, guiding the derivation of an initial iridophore specification GRN. Moreover, we use an iterative process of mathematical modelling, supplemented with a Monte Carlo screening algorithm suited to the qualitative nature of the experimental data, to allow for rigorous predictive exploration of the GRN dynamics. Predictions were experimentally evaluated and testable hypotheses were derived to construct an improved version of the GRN, which we showed produced outputs consistent with experimentally observed gene expression dynamics. Our study reveals multiple important regulatory features, notably a sox10-dependent positive feedback loop between tfec and ltk driving iridophore specification; the molecular basis of sox10 maintenance throughout iridophore development; and the cooperation between sox10 and tfec in driving expression of pnp4a, a key differentiation gene. We also assess a candidate repressor of mitfa, a melanocyte-specific target of sox10. Surprisingly, our data challenge the reported role of Foxd3, an established mitfa repressor, in iridophore regulation. Our study builds upon our previous systems biology approach, by incorporating physiologically-relevant parameter values and rigorous evaluation of parameter values within a qualitative data framework, to establish for the first time the core GRN guiding specification of the iridophore lineage. Despite decades of work, we still have only a superficial idea of how stem cells generate their distinct derivatives. This question becomes more acute if we consider that these fate choices are often made in a physically constrained environment (e. g. a stem cell niche), suggesting that fate-specification by environmental signals may be only part of the mechanism. Neural crest cells (NCCs) are a multipotent embryonic cell-type, sharing many properties with stem cells and indeed being retained as adult neural crest stem cells in various niches [1]. They are an important model for understanding the genetics of stem cell fate choice, since they generate a fascinating diversity of derivative cell-types, including many peripheral neurons, all peripheral glia, various skeletogenic cells, and pigment cells [2–4]. The latter are restricted to melanocytes in mammals, but are much more diverse in the anamniotes, such as fish [5–7]. In the well-studied zebrafish, there are three distinct types of pigment cells, namely black melanocytes, iridescent iridophores and yellow xanthophores, and in medaka, these three are supplemented by white leucophores. It is a long-standing, although largely untested, proposal that all pigment cells (or chromatophores) share a common origin from a neural crest (NC) derived, partially-restricted pigment cell progenitor, a chromatoblast [8], [9]. This, in conjunction with the inherent genetic tractability of these cell types, makes study of pigment cell development from the NC an exciting ‘model within a model’ for the genetics underlying stem cell fate choice. It is generally assumed that NC fate specification follows a progressive fate restriction model, with early, fully multipotent NCCs giving rise to individual pigment cell fates via a series of partially-restricted intermediates, and with fate choice consisting of a series of binary choices until a single fate is adopted [10], [11]. This view is crystallised in the iconic Waddington landscape model of stem cell development [12]. Consistent with this view, aside from the chromatoblast, these partially-restricted intermediates for pigment cells have been suggested to include bipotent Schwann cell precursors, capable of forming melanocytes as well as Schwann cells [13], [14], bipotent melanoiridoblasts [15], and bipotent xantholeucoblasts [16], [17]. Underpinning the observed fate choices are gene regulatory networks (GRNs) with the emergent property of distinct, stable states of gene expression, each corresponding to the molecular signature of a specific derivative cell-type. To understand stem cell fate choice, it is crucial to identify the key components of these GRNs and their regulatory logic. For pigment cell development, genetics has identified a small set of genes crucial for the control of lineage specification and differentiation [5], [18]–[20]. Integrating studies of these key mutants focused on identifying the core melanocyte GRN. Melanocyte specification centres on expression and maintenance of Microphthalmia-related transcription factor a (Mitfa), a bHLH-Leu Zipper transcription factor that functions as a master regulator of melanocyte development [19], [21], [22]. Initial expression of mitfa depends upon the Sry-related HMG-box 10 (Sox10), a transcription factor shown to directly regulate mitfa expression, cooperating with Wnt signalling [21], [23], [24]. Sox10 plays a similar role in specification of both xanthophores and iridophores as well [25], [26]. In the case of iridophores, as well as Sox10, the receptor tyrosine kinase Leukocyte tyrosine kinase (Ltk) plays a crucial role, with loss of function mutants lacking embryonic and adult iridophores, and constitutively activated Ltk signalling driving NCCs to adopt an iridophore fate [9], [27], [28]. We have shown that ltk expression appears to show two phases, one in early NC development which we propose represents a multipotent, chromatoblast-like progenitor, and a second in the definitive iridophore lineage [9]. Importantly, mitfa mutants show an intriguing increase in iridophores accompanying the absence of melanocytes, suggesting a close relationship between these two fates, and interpreted as revealing a shared bipotent progenitor, a melanoiridoblast [15], [19]. Mitfa belongs to a subfamily of related transcription factors containing the Transcription Factor E factors; one of these, tfec, is expressed in early NCCs, but later in a pattern strikingly reminiscent of iridophores, and is a strong candidate for a master regulator of iridophore development ([29] and Petratou et al. , in prep.). Finally, the Forkhead box D3 transcription factor (Foxd3) has been proposed to repress mitfa expression, consistent with the suggested role of FOXD3 in repressing MITF in other models [30], [31], thus biasing bipotent melanoiridoblast progenitors towards an iridophore fate [15], [32]. Although endothelin receptor Ba (Ednrba) shows an expression pattern that marks iridophore development, ednrba mutants show no discernible embryonic iridophore phenotype, although they do show loss of iridophores in adults [33]. Finally, pnp4a has been identified as a useful differentiation marker for the iridophore lineage [15]. However, these key genetic insights have yet to be integrated into a comprehensive GRN of pigment cell progenitors, the analysis of which might lead to understanding of how the NC generates each cell-type, and in appropriate ratios. As a first step in this, we have identified a core GRN for melanocyte fate specification [22]. As the number of components of a GRN increase, the standard network diagrams used to depict them become increasingly difficult to interpret using intuition alone. Importantly, therefore, we used an iterative cycle of experimental observations and mathematical modelling to more rigorously assess the GRN as we developed this core model. Using a similar approach, we have subsequently integrated the biphasic role of Wnt signaling in melanocyte development [24]. As a next step towards developing an integrated GRN for pigment cell fate-specification in zebrafish, we here extend the combined use of experimental genetics and mathematical modelling to develop a core GRN for iridophore specification. Many of the experimentally identified key genes in iridophore development, specifically ltk, tfec, sox10 and mitfa, show multiphasic expression in the NC and so we begin by outlining a working definition of the phases of iridophore specification from early, fully multipotent NCCs to differentiated iridophores. We then use this framework to allow careful interpretation of the highly dynamic gene expression patterns of the key genes in both wild-type (WT) and appropriate mutant embryos to assess the regulatory relationships between them. We refine the mathematical modelling approach developed to analyse the melanocyte GRN [22], using a literature search to limit parameter space to a reasonable physiological range, and Monte Carlo simulations to assess the robust predictions of GRN models throughout that parameter space. We emphasize that our Monte Carlo approach is particularly suitable in all those cases when quantitative data are not available, but rather qualitative behaviours are known. Supplemented by this approach for model selection, we then use our systems biology framework as a tool to rigorously evaluate a set of related models, refining and expanding them to define the first core GRN for iridophore development in zebrafish. This study was performed with the approval of the University of Bath ethics committee and in full accordance with the Animals (Scientific Procedures) Act 1986, under Home Office Project Licenses 30/2937 and P87C67227. Embryos were obtained from natural crosses. Staging was performed according to Kimmel et al. [34]. Unless stated otherwise, we used the WIK stock for experiments in WTs, and the following mutant lines: sox10t3 [26], mitfaw2[19], ltkty82 [9] and tfecba6 (Petratou et al. , in prep.). The tfecba6 allele shows recessive loss of function, and was generated via CRISPR/Cas9 directed mutagenesis. It corresponds to the deletion of 6 nucleotides from the 7th exon of the gene. This deletion of two amino acids is predicted to interrupt critical spacing in the second alpha-helix of the dimerization domain [35]. Phenotypically, homozygotes show a phenotype (nearly complete loss of iridophores and failure to inflate the swim-bladder) identical to those resulting from frameshift mutations in the DNA binding domain (Petratou et al. , in prep.). Embryos were obtained by incrossing heterozygous carriers for each mutant allele, with WT siblings were used as controls. Detailed information on the preparation of materials and the protocols for performing chromogenic whole mount in situ hybridisation (WISH) as well as multiplex fluorescent RNAscope can be found in Petratou et al. [36]. Probes used for chromogenic WISH were sox10 [26], foxd3 [37], ltk [9], pnp4a [15], mitfa [19]) and tfec (NM_001030105. 2). To generate the tfec probe, cDNA prepared from total RNA extracted from 72 hpf zebrafish embryos was amplified with the following primers: forward 5’-AGCCAACAATCACGACAGTG-3’ and reverse 5’-CCAATAGAAACGGGAGGTCA-3’. The product was cloned into pCR II-BluntTOPO vector (Invitrogen) and the orientation assessed by sequencing. The plasmid was linearised with PstI restriction enzyme (NEB) and in vitro transcription was with the SP6 polymerase of the DIG labelling kit (Roche; Cat# 11175025910). For multiplex RNAscope, the following probes were used: ltk (ACD; Cat No. 444641), tfec (ACD; Cat No. 444701), mitfa (ACD; Cat No. 444651), sox10 (ACD; Cat No. 444691) and foxd3 (ACD; Cat No. 444681). Embryos were imaged using an upright compound Imager 2 microscope (Zeiss). WISH samples were imaged under transmitted light, with an Axiocam 506 colour camera (Zeiss). RNAscope samples were imaged with dsRed, YFP and DAPI filters (supplied by Zeiss), using the Orca Flash 4. 0 V2 camera (Zeiss) and Apotome. 2 (Zeiss). Images were processed using the ZEN software (Zeiss), the FIJI package and Adobe Photoshop CS6. We note that mutant and WT embryos subjected to WISH were morphologically indistinguishable after fixation and were usually processed together. The Pearson’s chi-squared test [38–40] was used to test the null hypothesis that in a sample of mixed WT, heterozygous and homozygous mutant embryos, observed alternative gene expression patterns correspond to the expected Mendelian ratios (75% of embryos are expected to show WT phenotype and 25% to potentially show altered gene expression). For this test, degrees of freedom = 1. The chi-squared table [41] was used to calculate the probability that the number of observed embryos with an alternative expression phenotype was consistent with the expected number of homozygous mutants. For p-value > 0. 1 the null hypothesis was accepted. For p-value < 0. 1 it was assumed that alternative phenotypes in our samples were due to effects independent of the mutant genotype. RNAscope results were derived from two independent experimental repeats. From each repeat, between 3 and 5 embryos were examined. For each experiment, the number of embryos (2 or more representative individuals for each stage) used to score cells is indicated in S3 Table. For overexpression assays, 50–70 pg of purified mRNA diluted in sterile water were injected in each WT (WIK) one-cell stage embryo using standard methods [19]. Capped mRNA was prepared from plasmid templates using the SP6 mMessage mMachine kit (Ambion) for the overexpression of GFP, Sox10WT/Sox10m618 [26] and MitfaWT/Mitfab692 [19], [42]. For TfecWT/Tfecba6 overexpression, in vitro capped and polyadenylated mRNA was prepared using the mMessage mMachine T7 Ultra transcription kit (Ambion). Total RNA was isolated using TRI reagent (Sigma) from dissected trunks of 10–15 72 hpf WT (WIK) or homozygous tfecba6/ba6 embryos and WT or mutant cDNA was generated using the SuperScript III First Strand Synthesis Supermix kit (Invitrogen). The tfec coding sequence (ENSDART00000164766. 1) was amplified from the cDNA templates using the following primers: forward 5’-AGCGAGATCCTCCTGCTTCG-3’, reverse 5’-ATTCTGAGAGTGCGGTCCAG-3’. The T7 promoter was fused to the 5’ end of the amplicons through additional PCR amplification using the same reverse primer and the following forward: 5’-TAATACGACTCACTATAGGGAGAAGCGAGATCCTCCTGCTTCG-3’. The resulting amplicons were used as templates for in vitro transcription. Total RNA was extracted using TRI reagent (Sigma) from 8 embryos per sample at 6 hours post-injection. cDNA was synthesised from 1 μg of total RNA using the iScript cDNA synthesis kit (Biorad; Cat# 1708890). qRT-PCR was performed in duplicate using Fast SYBR Green Master Mix (Applied Biosystems; Cat# 4385617) and the StepOnePlus Real-Time PCR System (Thermo Scientific; Cat# 4376600). For normalisation, we used expression of the housekeeping gene rlp13 (primers ready-made from Primerdesign Ltd.). Primers for tfec transcript detection: forward 5’-GGAGCTTGGATTGCATGGAG-3’, reverse 5’-TTGATCAGCACCGTACACCT-3’. Primers for pnp4a transcript detection: forward 5’-TGGATGCAGTTGGAATGAGT-3’, reverse 5’-TTGACAGTCTCGTTGTCCTCA-3’. The ΔΔCt method [43] was used to evaluate relative changes in pnp4a expression, whereas absolute levels of tfec transcript were assessed using a standard curve. The null hypothesis that there was no change in the level of gene expression between control samples (injected with GFP or with null transcripts) and overexpression samples was rejected if p-value < 0. 05 using a two-sample t-test without assuming equal variances. Unpaired, two-tailed t-tests were performed using Microsoft Excel. Modelling of gene interactions was done using the approach presented in Greenhill et al. (2011). For each model, gene expression dynamics were described using a system of ordinary differential equations (ODEs; see S1 Text). The equations were solved numerically using the ode45 solver in MatLab software. Solving these equations returns the average concentration of gene output, measured in nM, across a homogeneous cell population. The Monte Carlo sampling algorithm used for randomising the constant parameters and subsequently scoring model outputs was run on MatLab. By random uniform logarithmic draw, we let all parameters vary in the range between a multiple 1/3. 5 and 3. 5. For each random draw, the system of ODEs of interest is solved and the resulting gene output dynamics are scored for biological relevance. For the scoring criteria refer to the results section, and for the mathematical functions used to calculate the scoring measure (S) for each system of equations, with randomly assigned parameter sets, see S2 Text. To interpret the expression dynamics of genes of interest during iridophore development in wild-type (WT) embryos, as well as changes of expression patterns in different loss of function contexts, it was crucial to distinguish cell populations comprising the different stages of iridophore development. Gene expression in the zebrafish NC is highly dynamic, reflecting both the rapid fate specification and differentiation of NC derived lineages in zebrafish and the multiphasic expression patterns of many key genes. For example, in previous studies of the ltk marker, we have proposed at least three phases of expression, one in multipotent premigratory progenitors, and two representing iridoblasts and differentiated iridophores respectively [9]. Tfec was first identified as a Mitf-related bHLH-ZIP transcription factor expressed in premigratory NC and later in a pattern reminiscent of iridophores [29]. We have recently shown that Tfec is crucial for fate specification of the iridophore lineage, and that tfec expression labels both early iridoblasts and differentiated iridophores (Petratou et al, in prep.). Building on these previous studies, we assessed the spatio-temporal locations of presumed iridophore progenitors at key stages of embryogenesis by examining tfec expression in whole-mount embryos. WISH on single embryos at 72 hpf confirmed that tfec is a definitive marker of differentiated iridophores (Fig 1A), similar to the established iridophore lineage marker, ltk [9]. Moreover, double labelling of ltk and tfec expression using multiplexed fluorescent RNAscope revealed that tfec is expressed throughout iridophore development (Fig 1C and 1C’; S3 Table). Furthermore, tfec transcripts were first seen in premigratory NC at very early stages, considerably before ltk (Fig 1B). Consequently, we interpret tfec expression as an excellent marker of ‘iridophore potential’ during zebrafish embryogenesis and use it here to produce a working classification of the stages of iridophore development. We note explicitly that this classification is intended to provide a framework for interpretation of mutant phenotypes, and that assessment of tfec alone cannot provide insight into the multipotency of cells at any specific stage. At 18 hours post fertilisation (hpf), premigratory NCCs reside along the dorsal trunk and tail (Fig 1B). Towards the posterior tail, these precursors are characterised by WISH as expressing markers such as sox9b, sox10, snai1b and foxd3 [9], [26], [44] and likely correspond to fully multipotent early NCCs (eNCCs). At the same stage, more anteriorly (i. e. posterior trunk and rest of tail), sox9b, snai1b and foxd3 are downregulated, while sox10 is retained. Interestingly, tfec is expressed in premigratory NC cells throughout the trunk and anterior tail (Fig 1B), in a manner similar to sox10, even though fate-mapping of premigratory NCCs shows that only a relatively small subset of NCCs will generate iridoblasts [26]. At this stage neither the melanoblast marker mitfa nor two other early iridoblast markers ltk and pnp4a were detectable by WISH in tfec-positive NCCs of the trunk (Fig 1B) [9], [44]. However, mitfa and ltk are activated widely by 22 hpf [9], [19], with pnp4a following soon after [15]. We consider that these premigratory cells of the Anterior Region of the Posterior Trunk (ARPT; Fig 1B) which express tfec, but no longer foxd3, and which only later detectably upregulate other pigment markers, are multipotent iridophore progenitors corresponding to the proposed partially restricted pigment cell progenitor [9]. Given the spatiotemporal gradient of development that is so pronounced during the stages of NC development, throughout this paper we will largely focus on a readily-defined anatomical zone, the ARPT, when considering expression patterns at different stages, thus minimising the heterogeneity of the examined population of cells. The ARPT lies above the anterior yolk sac extension (YSE) (approximately the region of somites 9–11; bracketed in Fig 1B). At 24 hpf, NCCs of the trunk have entered the medial and lateral migratory pathways, whereas less developed NCCs of the tail remain in premigratory positions. Although chromogenic WISH revealed maintenance of strong tfec expression in the premigratory NC domain, the majority of trunk NCCs located dorsal to the neural tube in the ARPT downregulate tfec, presumably due to cells becoming specified towards alternative lineages (Fig 1D). In this region, prominent tfec expression was retained in a small subset of precursors scattered over the spinal cord (Fig 1D), as well as in presumed iridoblasts migrating through the medial pathway (Fig 1Ei). tfec was only weakly detectable in cells entering the lateral migratory pathway, consistent with the medial migration pathway bias for iridoblasts noted previously [45] (Fig 1Eii). To further characterise the tfec-positive medially migrating cells, we used multiplexed fluorescent RNAscope to determine co-expression of tfec with the melanocyte lineage marker, mitfa (Fig 1I and 1I’; S3 Table). In all stages we scored dorsally located and medially migrating tfec+ cells of the trunk for mitfa expression. Importantly, the majority of tfec-expressing cells on the medial migration pathway were found to co-express mitfa, leading us to distinguish these as iridophore progenitors which retain at least bipotency. In the context of this paper, we will designate such tfec-expressing cells on the medial pathway as specified iridoblasts (but equally they could also be considered specified melanoblasts, using the definition of [46]). Numerous cells on the medial pathway were positive for mitfa, but displayed very weak or completely lacked expression of tfec (Fig 1I). We interpret these as having lost, or being on the way towards losing, iridophore potential, and are likely definitive melanoblasts. At 24 hpf, for quantitation cells were only scored if the tfec expression signal was elevated (>3 spots of fluorescence surrounding the nuclei) compared to the widespread low level (1–2 spots of fluorescence) expression displayed by numerous cells of the premigratory domain. This threshold was set in order to minimise inclusion of NC-derived cells which were in the process of downregulating tfec expression, thus focusing our analysis on cells most likely to be adopting the iridophore fate. At this stage, 10. 3% of the tfec+ cells scored lacked mitfa. Although conceivable that this subpopulation might reflect an alternative pathway to generate iridophores, one that does not require an mitfa+ progenitor, we suspect this small number of cells simply reflects technical limitations of the RNAscope technique. By 30 hpf, tfec transcript is present in scattered non-melanised cells along the dorsal posterior trunk and in medially migrating cells along the posterior trunk and anterior tail. Bilaterally patterned tfec+ premigratory NCCs were only detectable in the posterior tail (Fig 1F). In the APRT, co-expression analyses via RNAscope revealed that expression of tfec and mitfa has now resolved to be non-overlapping in the majority of cells (76. 9%) (Fig 1I and 1I’; S3 Table), which corresponds to a statistically significant drop in the proportion of cells co-expressing the two genes. We distinguish these cells as definitive iridoblasts, whether in the dorsal position characteristic of differentiated iridophores at later stages, or migrating on the medial migration pathway. Finally, we assessed tfec expression at 48 hpf, a stage when in live embryos differentiating, light-reflecting iridophores are distinguishable, interspersed along the dorsal, ventral and yolk sac stripes, occupying the lateral patches and overlying the eye. tfec expression was detectable by chromogenic WISH in all of these positions (Fig 1H). Co-expression analyses using RNAscope showed discrete expression of tfec and mitfa (Fig 1I and 1I’; S3 Table), confirming that tfec expression at this stage definitively marked differentiating iridophores. The small percentage (16. 6%) of tfec+; mitfa+ cells is most likely an artefact, due to the very close proximity and overlap between iridophores and melanophores along the dorsal stripe. This can result in expression from one cell being falsely assigned to an adjacent one. In summary, in addition to the fully multipotent early NCC state (eNCC), these marker studies clearly distinguished four sequential phases of iridophore development in WT zebrafish embryos: 1) premigratory NCCs presenting with widespread expression of tfec, but which have downregulated eNCC markers (for example, foxd3), interpreted as broadly multipotent pigment cell progenitors (chromatoblasts, Cbl); 2) scattered cells strongly maintaining tfec expression dorsal to spinal cord and on migration pathways, but also expressing mitfa and so interpreted as at least bipotent iridoblast progenitors (specified iridoblasts, ib (sp) ); 3) scattered undifferentiated cells in iridophore positions, showing rounded morphology and prominent expression of iridophore markers, but not mitfa (definitive iridoblasts, ib (df) ); and 4) discrete iridophore marker-expressing (and reflective in live fish) cells in characteristic definitive iridophore pattern (mature iridophores, iph). We note that ib (sp) and ib (df) can only be distinguished in double WISH, and so will not be strictly distinguishable in many experiments, although from the above discussion it can be seen that at 30 hpf most such cells in the ARPT will be ib (df). This characterisation provides a framework for assessment of expression patterns of other genes and for the interpretation of expression patterns seen in mutant embryos. Although well-known as a key factor in iridophore specification and a key marker of multipotent NCCs [9], [26], [47], sox10 expression has yet to be characterised in the iridophore lineage. We used both WISH and RNAscope to investigate the transcriptional dynamics of sox10 during iridophore development. We imaged iridophores of live embryos at 72 hpf using reflected light, and subsequently detected sox10 transcript in individual fish using chromogenic WISH. sox10 expression was readily detected in all iridophores (e. g. in each of the dorsal, ventral and yolk sac stripes; Fig 2A–2D). We then employed RNAscope to assess whether sox10 expression was maintained throughout all stages of iridophore specification, as opposed to its becoming re-activated in differentiated cells. We found that cells expressing the iridophore lineage marker, ltk, consistently co-expressed sox10 at each of 24 hpf, 30 hpf and 48 hpf (Fig 2E–2T). Therefore, just like tfec, sox10 expression in premigratory multipotent NCCs (Fig 1B) is maintained throughout fate restriction to ib (df) and their subsequent differentiation as iridophores. Previous studies with ltk have concluded that iridophore specification fails and that pigment cell progenitors are trapped in a multipotent progenitor state (Cbl) in sox10 mutants [9], [26]. We re-assessed this proposed role of sox10 using both loss and gain of function assays. In sox10 mutants, both at 24 hpf and at 30 hpf (Petratou et al. , in prep; Fig 3A and 3B), tfec expression is prominently retained in the premigratory NCC domain, which unlike in WT siblings extends anteriorly throughout the embryo. These observations indicate that Sox10 function is not required for establishment of the tfec-positive multipotent progenitor. Importantly, tfec transcripts were undetectable in ventrally migrating iridophore progenitors, indicating a requirement for sox10 to maintain tfec expression in a subset of cells (ib (sp) and ib (df) ). By 48 hpf, we could not detect tfec expression in sox10 mutant embryos (Fig 3E and 3F), consistent with apoptotic elimination of NC derivatives which fail to become specified, including progenitors for all chromatophore lineages [26]. To test the sufficiency of Sox10 for expression of tfec, we overexpressed WT Sox10 or a null mutant version [22], [26] by injection of mRNA into single cell stage WT embryos and assayed absolute tfec transcript levels using qRT-PCR at 6 hours post-injection. Our data clearly showed that functional Sox10, but not the null version of Sox10, ectopically activated expression of endogenous tfec (Fig 3O). Together these data show that Sox10 is not essential for initial activation of tfec in early NCCs, but it is required for maintenance of tfec expression as multipotent progenitors become specified towards an iridophore fate, i. e. for iridoblast fate specification. We next asked what roles Tfec and Ltk played in the iridophore GRN. Importantly, at both 18 and 24 hpf, we were unable to distinguish differences in tfec expression between ltk mutants and their WT siblings (S1 Table). Specifically, the premigratory NCC domain as well as specified iridoblasts in the posterior dorsal trunk and on the medial migration pathway of the trunk were unaffected in all examined embryos. Thus, tfec is activated in NCCs and is maintained at early stages of iridoblast specification, independently of Ltk activity. Nevertheless, from 30 hpf we observed a statistically significant decrease in the number of ib (df) located in the dorsal posterior trunk of ltk mutants (Fig 3A, 3C and 3M), and by 48 hpf no cells expressing tfec were identifiable in the ib (df) positions of the embryonic trunk in these mutants (Fig 3E and 3G). Study of ltk expression in tfec mutants by chromogenic WISH suggests that Tfec function is crucial for ltk expression from the earliest stages onwards. At 24 hpf, approximately 25% of assessed embryos completely lacked ltk expression, with the exception of very rare escaper cells (Petratou et al. , in prep.). This phenotype remained clearly identifiable at 30 hpf and persisted until at least 48 hpf (Fig 3I–3L). Moreover, examination of tfec expression in tfec mutant embryos, readily distinguishable owing to lack of melanin pigment in the RPE (Petratou et al. , in prep; Fig 3A and 3D insets), revealed a subtle but consistent reduction in the numbers of tfec-positive ib (df) in mutants from 30 hpf (Fig 3A, 3D and 3M). Specifically, tfec mutants displayed a 35% and a 45% decrease in the number of tfec-expressing cells along the migratory pathways and in the developing lateral patches respectively, compared to WT siblings. Similarly, at 48 hpf, tfec mutant embryos (identified by the clear eye phenotype; Fig 3E and 3H insets) showed reductions in tfec expressing cells in the dorsal stripe and the ventral stripe to 58% and 45% of those in WT siblings (Fig 3E, 3H and 3N). Although the remaining tfec-expressing cells show a distribution consistent with their being iph, we note that they lack both ltk expression and visible reflective platelets and hence cannot correspond to ib (df), which are positive for ltk expression, nor to mature iridophores. We speculate that these cells represent an interesting state in which iridoblasts are trapped in a very early stage of their development, where tfec expression, but not other markers, continue to be maintained. Taken together, our data strongly support the model that Tfec and Ltk function in a positive-feedback loop to maintain each other in specified iridoblasts, although tfec can be activated in this cell type independently of Ltk function. The gene pnp4a has been defined as an iridophore lineage marker, although it is expressed rather widely in NCCs and long before iridophores differentiate [15]. RNA-seq analysis of gene expression in purified iridophores and melanocytes has shown that it is expressed at high levels not only in differentiated iridophores, but also, albeit at lower levels, in melanocytes [48]. We used chromogenic WISH studies to assess pnp4a expression in iridophore development and in various key mutants. We first examined the WT expression pattern of pnp4a, and compared it to that of the iridophore and melanocyte lineage markers tfec and mitfa, respectively, at 24 hpf, 30 hpf and 48 hpf (Fig 4A–4I). At 24 hpf, it is notable that the expression pattern of pnp4a strikingly resembled that of mitfa, rather than that of tfec (Fig 4A, 4D and 4G). Specifically, we see clusters of cells just posterior to the otic vesicle and numerous cells on the medial migration pathway in the expression patterns of both mitfa and pnp4a, although both these regions are only sparsely positive for tfec (Fig 4A, 4D and 4G and insets). By 30 hpf, this pattern is still detectable, and indeed now pnp4a transcripts are detectable in differentiating melanocytes clustered behind the otic vesicle, as well as in melanised cells of the head, in a pattern similar to mitfa, but not tfec, expression (Fig 4B, 4E and 4H and insets). In addition, at this stage, the pattern of tfec and pnp4a expression in both the dorsal posterior trunk as well as overlying the RPE showed strong similarities; mitfa transcript is absent from the latter region (Fig 4B, 4E and 4H). By 48 hpf, the pattern of pnp4a strikingly resembled that of tfec, with both transcripts detected in iph positions, consistent with previously reported data [15], whereas mitfa was expressed in melanised cells of the head and of the dorsal, lateral and ventral stripes (Fig 4C, 4F and 4I). Considered together, our data suggest that while at later stages pnp4a is a definitive marker of differentiated iridophores, initially it is expressed widely in specified and differentiating melanoblasts. This suggested that pnp4a expression might be regulated by both Tfec and Mitfa. We began by investigating pnp4a expression in tfec mutants and WT siblings. At 30 hpf, WTs showed prominent pnp4a expression along the dorsal and ventral posterior trunk and the migratory pathways across the trunk and tail (Fig 4J). In contrast, tfec mutants displayed partial loss of pnp4a-positive cells (Fig 4L). Specifically, compared to WT siblings, tfec mutants showed decreased numbers of cells (expressing relatively low levels of pnp4a) along the dorsal trunk, and have comparatively few cells both on the migration pathway and in the ventral trunk, mostly more anterior. This partial reduction was also observed at 24 hpf (Petratou et al. , in prep.), and principally affected cells in the ventral trunk and premigratory NC. By 48 hpf, pnp4a expression in iph locations was eliminated in tfec mutants (Fig 4N and 4P; Fig 3N). Thus, pnp4a expression in iridophores and in ib (sp) is dependent upon Tfec, whereas the persistence of pnp4a expression in a subset of developing NC derivatives until 30 hpf suggested that pnp4a expression also depends on additional inputs. Due to the striking similarity of their expression patterns, we investigated a possible interaction between mitfa and pnp4a. Interestingly, pnp4a-expression was nearly eliminated in homozygous mitfa mutants, compared to their WT siblings at both 24 hpf (S1A–S1B’ Fig) and at 30 hpf (Fig 4J and 4M). Whereas in the former stage very few pnp4a expressing cells persisted along the posterior trunk of homozygous mutants (S1A–S1B’ Fig), in the latter a distinct group of dorsally located cells patterned in a ib (df) -like manner along the posterior trunk and anterior tail region were retained. In contrast, medially migrating cells were almost absent (Fig 4M). These results indicated that mitfa is an important regulator of pnp4a in premigratory and migrating NC cells, but that pnp4a activation in ib (df) is not affected. Thus, pnp4a expression appears to switch from Mitfa to Tfec-dependency during the transition from Cbl to ib (df), and to be detectable transiently in all melanoblasts and early differentiating melanocytes. We then asked whether Mitfa or Tfec were alone sufficient to drive pnp4a expression. We overexpressed each transcription factor, or a null mutant variant, in 1-cell stage WT embryos and measured pnp4a expression at 6 hours post-injection by qRT-PCR (Fig 4Q). As a negative control, we injected GFP RNA, allowing us to measure the relative (fold) change of expression between samples injected with GFP mRNA, compared to those injected with RNAs encoding WT or mutant. As expected, neither mutant Mitfa nor mutant Tfec altered pnp4a transcript levels, compared to overexpression of GFP. Interestingly, introducing WT Mitfa led to a statistically significant 4-fold increase (Fig 4Q). Surprisingly, however, overexpression of WT Tfec did not result in a statistically significant ectopic activation of pnp4a (Fig 4Q), suggesting that Mitfa, but not Tfec, is sufficient in this ectopic context to upregulate pnp4a. Loss of function studies using sox10 mutant embryos revealed that pnp4a was completely absent (Fig 4K), and that the gene remained inactive throughout the investigated developmental time-course (Fig 4O). We concluded that sox10 function was directly or indirectly required for all aspects of pnp4a expression, including the tfec-dependent pnp4a upregulation to occur. More broadly, we conclude that pnp4a regulation is more complex than previously assumed and that it should not be considered a definitive marker of the iridophore lineage at early stages. Bringing the above interactions together, we propose a preliminary iridophore GRN, comprising model A (Fig 5A–5C). We use solid lines to describe known direct interactions and dashed lines to indicate interactions where their nature is unknown. Sox10 has been previously shown to bind directly to the promoter of mitfa in zebrafish, and to activate its expression [21], we thus include that interaction. Furthermore, Ltk relies on intracellular cascades and effector transcription factors to activate gene expression, thus its input is always indirect. However, for the remainder of the interactions it remains unclear whether Sox10, Tfec or Mitfa bind directly to the promoters of downstream genes. We observe that in tfec and ltk mutants, sox10, tfec and ltk are all absent from the iridophore lineage from the ib (sp) stage onwards. Due to the nature of the sox10-dependent Tfec/Ltk positive feedback loop, and its central role in gene regulation within the GRN, loss of function experiments similar to those used to derive the aforementioned interactions cannot clearly indicate which of the candidate genes (Tfec, Ltk or Sox10 itself) is responsible for sox10 maintenance in iridophores. We propose three distinct variants of Model A, distinguished by the mechanism of sox10 maintenance in iridoblasts and iridophores (Models A1, A2 and A3 (Fig 5A–5C) ). In A1, sox10 maintenance occurs through an autoregulatory positive feedback loop (Fig 5A), which maintains sox10 expression from premigratory progenitors (Fig 1B). In A2, the presumed iridophore master regulator, tfec, is directly or indirectly responsible for sox10 activation in the context of iridoblasts (Fig 5B). Finally, model A3 proposes that sox10 maintenance is dependent upon Ltk signalling, independent of Ltk’s action in maintenance of Tfec expression (Fig 5C). These three Model A variants all share positive feedback loops between Sox10, Tfec and Ltk, and whilst biologically distinct, it is not obvious intuitively how they could be distinguished without detailed investigation of transcriptional regulatory mechanisms. However, like others, we have previously demonstrated the value of simple predictive mathematical modelling of GRNs in developing a robust understanding of their biological implications [22]. Thus, we utilised mathematical modelling of each model A variant, to assess more rigorously whether they could be distinguished. If so, we wished to identify the model offering highest predictive power, i. e. the network that was best able to recapitulate the experimentally observed gene expression dynamics in the iridophore lineage (S3 Fig). We generated systems of ordinary differential equations (ODEs) describing the interactions in each of the proposed networks (see S1 Text). The changes in the expression of each gene over time were determined using an ODE which incorporated all activatory and repressive influences from other members of the network, as well as a term for degradation of the gene’s protein product. The model aimed to capture the average output (nominally as protein product, assuming direct correlation with transcript production) of each gene in a homogeneous group of cells at a given time. The necessary parameter values characterizing the regulatory dynamics (mRNA maximum production rates (g), protein degradation rates (d), dissociation constants for transcription factors binding (K) ) were chosen following exploration of existing literature to identify physiologically relevant values (see S1 Text; S2 Table). We solved the systems of ODEs numerically in MatLab. To define initial conditions (here at t = 18 hours), we chose the population of premigratory NCCs occupying the dorsal ARPT at 18 hpf. Using chromogenic WISH, sox10 and tfec, but not ltk, mitfa or pnp4a transcripts were detectable in this population of cells (Fig 1B), allowing us to approximate the initial conditions for our simulations. As time proceeds, the simulations were tested for their ability to broadly replicate the changes in gene expression of tfec-expressing cells as they transition through the stages of ib (sp), ib (df) and then iph. In vivo, differentiating iridophores are observable from 42 hpf and prominent by 48 hpf. To account for inaccuracies in our default parameter sets (see S1 Text; S2 Table), we allowed computational simulations to progress until 60 hpf, thus helping to ensure that any biologically meaningful steady state could be successfully reached. Specifically, in the WT context, it was crucial that Sox10 and Tfec stay upregulated in mature pigment cells, i. e. reach a positive steady-state. Similarly, Ltk and Pnp4a concentrations should increase and reach a positive steady-state. Mitfa levels should initially rise rapidly, reflecting the widespread expression of mitfa in all pigment cell progenitors [32], but should then drop to a distinctly lower level at differentiation stages (this work). We note that the Mitfa concentration is not required to attain zero, but simply to drop to a lower steady-state value; given the expectation that in situ hybridisation techniques have a ‘detection threshold’, we consider that this final lower value would reflect expression levels undetectable by our detection methods in differentiated iridophores, although they would still be measurable by microarray in pooled isolated iridophores [48]. As a further test of each of Models A1, A2 and A3, we used MatLab to predict gene expression changes in the context of different mutant scenarios, when function of Sox10, Ltk or Tfec were individually ablated in silico (Fig 5). In the sox10 mutant context we asked that Sox10, Tfec and Ltk acquire positive values, as expression has been identified in trapped chromatoblasts ([9], [26] and this work), however at no point do either Mitfa or Pnp4a become upregulated. Loss of Ltk function was required to predict initial rise of Ltk, Tfec and Pnp4a concentrations (at approximately 24–30 hpf), followed by gradual downregulation to undetectable levels. Similarly, loss of Tfec function should be accompanied by a peak and subsequent decline of Tfec, Sox10 and Pnp4a concentrations within 30–50 hpf. Ltk should never become detectable in the developing iridophore population in this context. These simulations showed that regardless of the interaction underlying sox10 maintenance, in the WT context all iridophore markers were appropriately upregulated in the course of iridophore development, consistent with biological observations. In all three models, expression of the melanocyte marker, mitfa, a direct target of Sox10 [21], was predicted to be upregulated and then maintained in the iridophore lineage (Fig 5D–5F). These predictions are in contrast to previously published experimental data, showing that maintenance of mitfa is restricted to melanocytes [21], [22], although lower level mitfa expression has been detected by RNA-seq in differentiated iridophores [48]. We used RNAscope to assess directly the relative changes in mitfa expression in the iridophore lineage. Even with this technique, notably more sensitive than conventional chromogenic WISH, we confirmed that co-expression of mitfa with the iridoblast marker, tfec, occurred at 24 hpf, but such overlap was not detectable at 30 hpf and at 48 hpf (Fig 1I). Thus, all three models failed to correctly predict the expected initial peak, followed by downregulation, of mitfa expression in the iridophore lineage. These observations are readily explained by the absence of a mechanism for repression of melanocyte fate in our model; we explore this later. For all three versions of model A, simulation of loss of Sox10 function appropriately predicted maintenance of tfec (Fig 3B) and of ltk [9], consistent with observations that tfec+; ltk+ progenitors remain trapped in the dorsal trunk and tail. Likewise, they appropriately predict the failure to upregulate both mitfa [21] and pnp4a (Fig 4K). However, it has been previously shown that dorsally trapped progenitors continue to express sox10 upon loss of Sox10 function [26], a feature predicted successfully by models A2 and A3, but not by A1. Similarly, computational implementation of tfec loss of function revealed that model A1 did not generate biologically accurate predictions, whereas models A2 and A3 performed better. Specifically, Model A1 with simulated loss of Tfec function did not result in the experimentally observed lack of ltk and gradual downregulation of both tfec and pnp4a expression (Fig 3D; Fig 4L). In models A2 and A3, ltk expression was correctly predicted to remain undetectable throughout iridophore specification and differentiation, while tfec and pnp4a were gradually diminished. Finally, in silico inhibition of Ltk signalling in models A1 and A2 failed to predict the experimentally observed initial activation, followed by downregulation, of ltk [9], tfec (Fig 3C) and pnp4a (S1C–S1F Fig) expression as iridoblasts differentiate into iridophores. Model A3, however, successfully predicted gradual elimination of iridophore marker gene expression in the lineage, more accurately reflecting the current experimental observations. Based on the above observations, we conclude that Model A3 has the highest degree of predictive power using the default parameter set. These parameters were chosen based on ranges indicated from the literature as physiologically relevant, nevertheless the exact values were assigned somewhat arbitrarily. We, therefore, conducted an unbiased assessment of whether the experimentally set output requirements, as outlined above, could be achieved using alternative sets of parameter values in any of the models. To that effect, we designed a Monte Carlo algorithm able to randomly assign parameters drawn from a pre-assigned range, spanning two orders of magnitude from 5x lower than the physiological mean value to 5x higher than that value. For each model, the outputs for each of 20,000 combinations of parameters were scored computationally by a suitably designed scoring function, according to our set of qualitative criteria (see S2 Text), which took into account that only qualitative expectations of gene regulatory dynamics could be tested. The scoring function for a given model output is the multiplication product of the individual scoring measures for each of the gene expression curves. These individual functions could be binary, adopting either 0 or 1 values, if the assessed feature is absolutely required for an output to be considered as biologically relevant (for example successful upregulation of iridogenic genes). Alternatively, individual scores may be quantitative (for example the score of the efficiency of the Mitfa ‘rise and drop’ behaviour), meaning different curve behaviours would result in relatively higher or lower score values (S). This feature implies that the highest score (Smax) achieved by each model may be used an indicator of its ability to produce outputs that closely match experimental observations. Importantly, S values act as relative ranking tools, comparing the capabilities of our models, but note that the exact values bear no biological significance. Furthermore, the frequency by which acceptable and high scores are achieved is crucial in identifying models robustly predicting the experimentally observed gene expression dynamics. In the principal component analysis (PCA) plots (Fig 5G–5I), the frequency of high scores in each model is visualised by the density of dark blue, purple and magenta spots. The three principal components depicted represent linear combinations, each pointing to the direction of maximal variance, with respect to their score-weighted position vectors. PCA was used to visualise the frequency of scores within parameter space and to compare the three models’ respective capacities to reproduce experimental observations (Fig 5G–5I). Interestingly, all outputs derived from randomly assigning parameter values in the set of equations representing model A1 failed to predict crucial aspects of the biology, thus consistently achieving zero scores (Fig 5G). Model A2 was found able to predict those features correctly, although only limited subsets of parameters achieved admissible outputs (Fig 5H). Model A3 performed similarly to Model A2, except that it generated predictions broadly consistent with the known biology for a wider range of parameter combinations (Fig 5I). Although the analysis of the model A variants identified a more favourable model (A3) for most aspects, none of the model A alternatives were able to reproduce the expected Mitfa dynamics (i. e. sufficient downregulation of Mitfa in differentiating iridophores), while simultaneously maintaining relatively high outputs of iridogenic gene products (Fig 5; S6A–S6C Fig). We used model A3 as a starting point to improve this aspect of the iridophore GRN. Since repression of mitfa in the iridophore lineage has not thus far been investigated, we asked which interactions would be able to produce appropriate outputs when mathematically implemented. After testing predictions of alternative models with our default parameter set (S4 Fig), we concluded that upregulation in the iridophore lineage of an unknown mitfa repressor, which we termed factor R, was crucial. The resulting Model B (Fig 6A) incorporated Tfec-dependent activation of factor R, which our implementations suggested should be absent in the multipotent progenitors of the ARPT at t = 18 hours. Manually adjusting the parameters in the system of ODEs describing Model B revealed that the experimentally determined rise and drop of mitfa expression in our group of cells could be achieved using the default parameter set (Fig 6B), and even enriched with alternative parameter values, within the determined physiologically relevant range (S6E and S6G Fig). Random assignment of parameters and algorithmic scoring of respective outputs (see S1 Text) indicated that, of all the tested models, model B best reflected experimental observations regarding gene expression dynamics. Specifically, compared to models A1-A3, a broader range of model B trials achieved high scores, with absolute values higher than those attainable through models A1-A3 (Fig 5; Fig 6). Notably, model B (derived from model A3) consistently scored higher than a designated alternative model B (2), which was derived by introducing factor R into model A2 (S6F and S6G Fig). We considered Foxd3 as a candidate for factor R, as the transcriptional regulator has been previously implicated in mitfa repression [15]. Our modelling predicted that, in the WT context, factor R should be expressed in undetectable levels in the ARPT at 18 hpf, and should then be robustly upregulated in all developing iridophores, before reaching a stable plateau at differentiation. We tested these predictions for foxd3 using RNAscope. Surprisingly, we detected only low levels of foxd3 transcript, and these in only half of ltk-positive iridophore lineage cells at each of 24 hpf, 30 hpf, 36 hpf and 48 hpf (S5 Fig; S3 Table), making it unlikely to be the key factor repressing mitfa expression in the iridophore lineage. As a further test, we examined foxd3 mutants and WT siblings by WISH at 24 hpf, to determine whether absence of Foxd3 activity resulted in derepression of mitfa in cells on the medial pathway (ib (sp) and later stages of iridophore differentiation). Contrary to this prediction, numbers of mitfa positive cells were found to be somewhat reduced in this region of homozygous mutants compared to WT siblings (S5 Fig). These observations were inconsistent with the hypothesis that Foxd3 mediated mitfa repression during iridophore lineage differentiation. Hence, we conclude that whilst previous data indicates a role for foxd3 in pigment cell development, it is unlikely to perform the role of factor R in iridophore GRN. In previous work we used an iterative process of experimental genetics and mathematical modelling to develop a robust core GRN for the zebrafish melanocyte. Here, we extend that approach to establish a core GRN for the iridophore, a second pigment cell-type that shows a close developmental genetic relationship with the melanocyte, and which has been proposed to derive from a shared bipotent progenitor [15]. In the course of our experimental analysis, it soon became clear that iridophore-related genes often showed multiphasic expression, being detectable in differentiated iridophores, but also in much earlier, even premigratory stages of NC development. We had first noted this in our study of ltk expression [9], but here we showed that tfec, pnp4a and sox10 behave similarly. This same phenomenon has been documented, but not emphasised, in the case of melanocyte development, with mitfa being expressed initially in almost all NC cells [32], but it is less clear whether other melanocyte-specific genes present with similar biphasic expression. Our use here of the RNAscope assay, readily allowing highly sensitive detection and quantitation of co-expression, reveals that early markers of fate specification of different cell-types (e. g. mitfa and tfec) may be initially co-expressed. This reflects the distinction between fate specification, when a cell is beginning to show characteristics of a specific lineage, and commitment, when it has stably adopted that fate at the expense of alternative ones. These considerations, plus the standard limitation that we usually examine only one or two markers at once, resulted in us attempting to standardise our assessment of gene expression patterns, taking account of not only marker expression and levels of expression, but also cell location and cell morphology. This led to an explicit working model of stages in iridophore development from early NCCs (Fig 1 and Fig 7). This model is broadly consistent with the current progressive fate restriction model of NC development. However, the experimental restrictions noted above mean that we can, at best, assess minimal levels of potency: where we see overlap of expression of key genes for different fates, we interpret this as reflecting the cell having potential for at least these fates. Analysis of ltk expression in sox10 mutants ([9] and Nikaido et al. , in prep.), interpreted in the light of our detailed studies of the mutant phenotype, including single cell fate-mapping of NC ([26] and Subkhankulova et al. , in prep.), led us to propose that premigratory NCCs in the trunk and tail go through a multipotent pigment cell progenitor phase, that we refer to here as the Cbl phase (Fig 7). Combining the RNAscope and WISH data presented here with those studies and our similar observations for other markers, provides support for this interpretation, since premigratory cells expressing tfec or ltk do not express markers of fully multipotent eNCCs (e. g. foxd3, but also snai1b and sox9b). We further distinguish two phases to this Cbl stage, with cells initially expressing tfec, but not ltk, mitfa nor pnp4a (which we designate early Cbl cells), before rapidly turning on all these genes (becoming late Cbl cells; Fig 7). A crucial step is then the establishment of a positive feedback loop between Ltk and Tfec, which drives maintenance of the iridophore specification state. Our findings build on the established role of Ltk signalling, linking it to a key transcription factor for iridophore fate specification, Tfec. Tfec is a close homologue of Mitfa, so it is intriguing that it seems to have a similarly central role in iridophore fate specification as does Mitfa in melanocyte development. However, Tfec is expressed much earlier than mitfa and ltk in NC development, being detected widely in early NCCs [29], yet this is unaffected in a sox10 mutant. We hypothesise that here other factors present in the early NCC (but downregulated in differentiating iridoblasts) act redundantly with Sox10, but this will require experimental testing. Such redundant regulation has previously been reported in early NCC GRNs. For instance, NC-specific expression of sox9a, sox9b and sox10 has been shown to depend on both foxd3 and tfap2a function in zebrafish embryos [49]. Although such upstream regulation of tfec in early NCCs remains to be elucidated, we here provide experimental evidence that Ltk signalling is required to maintain Tfec in a subset of cells, setting them aside as ib (sp). However, it is important to note that these cells are initially co-expressing mitfa, consistent with their being at least bipotent progenitors of both melanocytes and iridophores. Intriguingly, we show that one early function of Tfec is to activate ltk expression in premigratory NCCs, in what we consider to be a first step in chromatophore fate restriction (Fig 7), distinguishing the multipotent Cbl from the fully multipotent eNCC. Based on our loss of function data presented here, we identify a Tfec/Ltk positive feedback loop as a key feature of the core GRN for iridophore fate choice. Importantly, this work highlights the ongoing role of Sox10 in iridophore development. In melanocytes, Sox10 acts together with Wnt signalling to establish mitfa expression, but sox10 is then strongly downregulated, and indeed maintenance of expression is thought to promote multipotency and delay differentiation [22]. In the iridophore lineage, loss of sox10 function results in failure of iridophore fate specification, suggesting that there are strong parallels between the genetic mechanisms of melanocyte and iridophore fate specification. Surprisingly, in contrast to the regulatory dynamics taking place in the melanocyte lineage, our data indicate an ongoing role for sox10 over the course of iridophore development, with expression of the gene being readily detectable by RNAscope in all stages of specified iridoblasts and by WISH in mature iridophores. It will be interesting therefore to explore the molecular basis for repression of alternative fate choices that allows sox10 expression (which is strongly associated with NCC multipotency; [47], [50–52]) and iridophore fate commitment to proceed hand-in-hand. Our experimental data showed that sox10 expression needed to be maintained for tfec to remain expressed in the specified iridophore lineage (ib (sp), ib (df) and iph). We considered an alternative interpretation of our loss of function results, that sox10 is required for tfec+ iridoblast migration, rather than for maintenance of tfec expression per se in this cell type. However, we consider this less likely in light of our sox10 gain of function data, which strongly suggest the ability of Sox10 to upregulate tfec expression. Nevertheless, an additional role of sox10 in iridoblast migration remains to be tested. Our demonstration that sox10 mutants show persistent and extensive tfec expression in premigratory NCCs could be interpreted as indicating a role for Sox10 in repression of tfec. However, we consider direct repression of tfec by Sox10 unlikely, since, as we have shown here, 1) there is consistent co-expression of the two factors in multipotent Cbls and in iridophores and 2) overexpression of Sox10 drives transcriptional activation of tfec. Instead, we propose a more parsimonious explanation, that in the Cbl stages tfec expression is established and maintained in a Sox10-independent manner, but that as these cells become specified to most lineages, tfec expression is downregulated; the exception is that those cells that become specified to the iridoblast lineage maintain and indeed upregulate tfec as a key part of that specification process. In sox10 mutants, we suggest that NCCs become trapped in the Cbl state, since specification to all non-ectomesenchymal fates is blocked [9], [26], [52]. An important question to be addressed in future work is what are the factors that indirectly repress tfec expression, downstream of Sox10-dependent specification of non-iridophore fates? Indeed, our work highlights the previously overlooked significance of gene repression as part of the fate specification mechanism. For example, in the differentiation of melanocytes from Cbl cells, tfec, ltk, as well as sox10, all have to be downregulated, in what we assume will be an Mitfa-dependent manner. Our data made clear the need for maintenance of sox10 expression in iridoblasts, but the close regulatory relationship between the Ltk-Tfec feedback loop and Sox10 made it difficult a priori to distinguish three variants: 1) sox10 autoregulation, or 2) input from the Ltk-Tfec loop through Tfec (or a downstream target of Tfec), or 3) through Ltk independent of Tfec. Intuitively, the impact of these three distinct modes is difficult to decipher, so an unexpected outcome of the mathematical modelling was the realisation that the behaviour of the GRN was quite different under these models. Our simulations, supported by unbiased random sampling via a Monte Carlo approach, clearly suggested that one model (Model A3) was superior to the alternatives, in that it most readily and robustly led to a predicted pattern of gene expression most closely mimicking that observed experimentally. This nicely illustrates the benefits of simple predictive mathematical modelling in rigorous assessment of GRNs. Our observations also revealed an unexpected complexity to the regulation, and thus the likely role, of pnp4a in pigment cell development. Although the gene has been considered a definitive marker of the iridophore lineage [15], our study reveals complex regulation of pnp4a in premigratory and migrating NCCs, by both Tfec and Mitfa, as well as by Sox10. During these stages of fate specification and early differentiation pnp4a is best considered a marker of both specified melanoblasts and specified iridoblasts (which as we have shown likely include many shared cells), although we also confirm that at later stages by WISH at least it is a definitive marker of the iridophore lineage. This gene encodes purine nucleoside phosphorylase 4a, an enzyme converting guanosine mono-phosphate to guanine [48]. We note that in medaka the guanineless/pnp4a gene mutant phenotype is a pronounced reduction of iridophore reflectivity, consistent with its proposed enzymatic role in generating high concentrations of guanine in iridophores to allow reflecting platelet formation [53]. The gene’s role in melanoblasts remains unclear. One significant innovation in our implementation of the mathematical modelling in this study over its use in the Greenhill study, is in our assessment of parameter values. One well-known problem with mathematical modelling is that as GRNs increase in complexity the number of parameters increases rapidly. In most in vivo systems, the absolute values of these parameters cannot be easily measured, leading to considerable uncertainty about the resulting simulations and their validity. To overcome this general drawback here, we restricted the values of all parameters to physiologically relevant ranges, based upon published measurements, choosing values that we consider to be best estimates based upon the same or similar molecular interactions. Furthermore, we employed an unbiased Monte Carlo sampling to ask explicitly how significant changes to those parameters might be for the output of the models. For example, we asked whether randomly varying the originally assigned ‘sensible’ parameter values, implementation of which had resulted in a particular model’s predictions broadly matching experimental observations, might render the subsequent modelling outputs radically different and thus the model less convincing. Alternatively, we aimed to confirm that any model that gave inaccurate predictions with the originally assigned parameter set remained incapable of predicting the experimentally observed gene regulatory dynamics even with different parameter sets. We emphasize that the method presented here is particularly suitable to all modelling attempts where experimental data is qualitative in nature and limited to few developmental time-points due to technical restrictions. In this sense, it addresses an important drawback, typical in Systems Biology model reconstructions, when both the topology of the network and the parameters are unknown. In standard fitting procedures the topology is assumed, and the optimal parameter set is chosen so as to minimise the discrepancy between the theoretical predictions and the (quantitative) experimental data. This approach of course fails when a quantitative dataset is not available. The method described here aims to cover those situations where only a limited, non-quantitative set of data on dynamical behaviours is available, and attempts to assess different network topologies in a broad spectrum of parameter values. By augmenting the model GRN with the presented scoring functions, our Monte Carlo screening algorithm allowed us to rigorously explore the proposed model variants and to compare their predictive powers under a broad set of physiologically-relevant parameter values. This approach renders the process of either validating or refuting model variants considerably more objective. The assessment of our GRN by mathematical modelling revealed a key feature, one that will be crucial when integrating the melanocyte and iridophore GRNs, namely the factor (factor R) repressing mitfa, and thus melanocyte fate, in the iridoblasts. A series of published experimental observations, including the foxd3 mutant phenotype, with partial loss of iridophores, partially rescued in foxd3; mitfa double mutants [15], had led to the proposal that FoxD3 might have such a role. Our modelling indicates that factor R is required throughout iridophore development, including into differentiation phases. As a test of the suitability of foxd3 for this role, we assessed expression in the iridophore lineage throughout a developmental time-course using both conventional WISH, as well as RNAscope. Previous analyses using a foxd3: gfp transgenic line have suggested that Foxd3 is expressed in mature iridophores [32]. Here we used RNAscope to assess co-expression of foxd3 with the iridophore lineage marker ltk. This technique is both highly sensitive and quantitative, thus much better suited for sensitive detection of co-expression, with the added advantage that rapid turnover of mRNA (in contrast to slow degradation of GFP) allows for more precise assessment of regulation of gene expression. To our surprise, although co-expression of foxd3 and ltk is readily detected in 24–48 hpf larvae, overlap is seen in only half of the detected ltk-expressing cells at any of these stages. The criteria derived by our modelling with regard to the expression dynamics of factor R prompted us to conclude that mitfa repression cannot be fully explained by FoxD3 activity, although we cannot rule out this known transcriptional repressor [30]–[32], [54] making a partial contribution. Our work here, plus that from other groups, provides some indication that there may be two sub-populations of iridophores. Specifically, we show that foxd3 is only expressed in half of the lineage cells throughout the process of specification, broadly consistent with the published loss of function phenotype in which only a proportion of iridophores are missing [55]. Furthermore, we report a subset of tfec-positive cells, distributed in a pattern similar to that of mature iridophores, which persist along the dorsal trunk of tfec mutants. Finally, mitfa mutants at 30 hpf appear to maintain the same, or even increased numbers of pnp4a positive cells along the dorsal trunk, but to show a consistent lack of cells in the migratory pathways compared to their WT siblings, indicating that the former subgroup is independent of, while the latter requires mitfa function. Whilst the idea of discrete iridophore sub-types is an exciting one, the data are not sufficient at present to make this interpretation compelling; for example, foxd3 expression in differentiating iridophores may simply be close to a detection threshold and hence incompletely detected, and if fewer iridophores are produced these may preferentially occupy locations at the premigratory position. Furthermore, a recent study suggests a new role for FoxD3, namely as a pioneer factor in neural crest development [56], which is consistent with the observed delay in mitfa expression and reduced numbers of iridophores. Nevertheless, future work should aim at better characterisation of iridophore subpopulations, for instance by extensive marker co-expression analyses. In summary, we have produced the first core GRN for the zebrafish iridophore, incorporating all known major players. This work now forms the basis for integration with our core GRN for the zebrafish melanocyte [22] in order to begin to see how integration of these GRNs is achieved in the pigment cell precursors enabling melanocyte versus iridophore fate choice. Whilst here we have focused on experimental genetics approaches and integrated mathematical modelling, a complementary approach looking at NC-specific histone marks would be informative, for example revealing active regulatory elements in iridophore lineage cells. However, a major priority will be to decode the mechanism repressing melanocyte (and potential other) fates in the iridophore lineage, with identification of factor R a crucial first step. Finally, continuous development of co-expression detection strategies [57] will soon allow for simultaneously identifying an increasing number of marker genes, thus providing insight into the true potencies of partially restricted progenitors in vivo.

Multipotent neural crest (NC) progenitors generate an astonishing array of derivatives, including neuronal, skeletal components and pigment cells, but the molecular mechanisms allowing balanced selection of each fate remain unknown. In zebrafish, melanocytes, iridophores and xanthophores, the three chromatophore lineages, are thought to share progenitors and so lend themselves to investigating the complex gene regulatory networks (GRNs) underlying fate segregation of NC progenitors. Although the core GRN governing melanocyte specification has been previously established, those guiding iridophore and xanthophore development remain elusive. Here we present expression data, as well as loss and gain of function results, guiding the derivation of a core iridophore specification GRN. Moreover, we use a process of mathematical modelling and rigorous computational exploration of the GRN to predict gene expression dynamics, assessing them by criteria suited to the qualitative nature of our current understanding of iridophore development. Predictions were experimentally evaluated and testable hypotheses were derived to construct an improved version of the GRN, which we showed produced outputs consistent with experimentally observed gene expression dynamics. The core iridophore GRN defined here is a key stepping stone towards exploring how chromatophore fate decisions are made in multipotent NC progenitors.

Abstract Introduction Materials and methods Results Discussion

medicine and health sciences fish vertebrates pigments animals cell differentiation epithelial cells simulation and modeling animal models osteichthyes developmental biology model organisms stem cells materials science experimental organism systems chromatophores embryos mathematical modeling research and analysis methods embryology cell potency animal cells melanocytes animal studies gene expression biological tissue zebrafish eukaryota cell biology anatomy genetics epithelium biology and life sciences cellular types multipotency physical sciences materials organisms

2018

A systems biology approach uncovers the core gene regulatory network governing iridophore fate choice from the neural crest

One key problem in precision genome editing is the unpredictable plurality of sequence outcomes at the site of targeted DNA double stranded breaks (DSBs). This is due to the typical activation of the versatile Non-homologous End Joining (NHEJ) pathway. Such unpredictability limits the utility of somatic gene editing for applications including gene therapy and functional genomics. For germline editing work, the accurate reproduction of the identical alleles using NHEJ is a labor intensive process. In this study, we propose Microhomology-mediated End Joining (MMEJ) as a viable solution for improving somatic sequence homogeneity in vivo, capable of generating a single predictable allele at high rates (56% ~ 86% of the entire mutant allele pool). Using a combined dataset from zebrafish (Danio rerio) in vivo and human HeLa cell in vitro, we identified specific contextual sequence determinants surrounding genomic DSBs for robust MMEJ pathway activation. We then applied our observation to prospectively design MMEJ-inducing sgRNAs against a variety of proof-of-principle genes and demonstrated high levels of mutant allele homogeneity. MMEJ-based DNA repair at these target loci successfully generated F0 mutant zebrafish embryos and larvae that faithfully recapitulated previously reported, recessive, loss-of-function phenotypes. We also tested the generalizability of our approach in cultured human cells. Finally, we provide a novel algorithm, MENTHU (http: //genesculpt. org/menthu/), for improved and facile prediction of candidate MMEJ loci. We believe that this MMEJ-centric approach will have a broader impact on genome engineering and its applications. For example, whereas somatic mosaicism hinders efficient recreation of knockout mutant allele at base pair resolution via the standard NHEJ-based approach, we demonstrate that F0 founders transmitted the identical MMEJ allele of interest at high rates. Most importantly, the ability to directly dictate the reading frame of an endogenous target will have important implications for gene therapy applications in human genetic diseases. Programmable nucleases such as TALEN (Transcription Activator-like Effector Nuclease) and CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) systems have enabled a new era of scientific research [1,2]. Instead of relying on knock-down models or expensively outsourced knock out lines, laboratories across the world now have tools with which to generate indels (insertions and deletions) of varying sizes on the gene (s) of interest. However, DNA Double-strand Break (DSB) repairs largely result in diverse sequence outcomes owing to the unpredictable nature of the most commonly used Non-homologous End Joining (NHEJ) pathway [3,4] (Fig 1). This significantly confounds experimental readouts as knock-out cell lines often harbor more than just one desired frameshift mutation. In the case of model organisms such as zebrafish (Danio rerio), the F0 founders are genetically mosaic, warranting a complex and time-consuming series of outcrossing to establish molecularly defined lines before any biological questions can be addressed [5,6]. In contrast to NHEJ, the MMEJ (Microhomology-mediated End Joining) DNA repair pathway utilizes a pair of locally available direct sequence repeats on both sides of a DSB that are apposed, annealed and extended [7–10]. As such, DSB repair outcomes are highly stereotyped (Fig 1), resulting in deletion of the intervening sequence as well as one of the repeats. Consequentially, there is an increasing interest in utilizing MMEJ for precision genome engineering applications [11–14]. To date, however, effective harnessing of this pathway remains challenging due to the paucity of genetic and mechanistic understanding [8]. Bae, et al. [14] developed a sequence-based scoring system to estimate the frequency of MMEJ-associated deletions induced by DSBs in human cells. While this improved the predictability of MMEJ activation, the DSB repair outcomes tended to consist of a heterogeneous population of multiple MMEJ alleles. In this study, we sought to improve upon the existing algorithm with the goal of developing tools to more reliably predict target loci that would be predisposed to generate a more homogeneous mutant allele population through MMEJ. We demonstrate the feasibility and utility of such reagent design on the molecular level (i. e. , DNA repair outcomes) and on the physiological level (i. e. , F0 phenotype). We further demonstrate that our approach can be applied to generating highly homogeneous MMEJ alleles in cultured human cells, suggesting our findings may be broadly translatable to multiple model systems. We believe our approach can inform and benefit applications such as rapid phenotype-genotype correlation in F0 animals, with an eye toward applications in human gene therapy and facilitation of resource sharing & recreation of various cell and animal lines on a global scale. Prior works examining MMEJ activation in vertebrate organisms primarily focused on in vitro models [8–10,14–18]. Initial analyses using a targeted knockin strategy suggested that MMEJ was operational in the zebrafish embryo, though the efficiency of these MMEJ outcomes was rather modest [13]. Importantly, while previous studies reported incidental identification of several zebrafish genomic loci that repaired preferentially through MMEJ when using programmable nucleases [19,20], no consortium–small or large–of genomic loci that repair primarily through NHEJ vs MMEJ has been compiled. To this end, we examined the repair outcomes of previously designed TALEN and CRISPR-Cas9 genomic reagents (S1 Table). The plurality of custom enzymes induced diverse sequence outcomes, consistent with the idea that NHEJ is being used as the primary DNA repair pathway at these loci. However, a few reagents induced sequence outcomes satisfying the following criteria, suggesting that MMEJ was the preferred pathway: 1) most predominant mutant allele is the top predicted allele by the Bae, et al algorithm [14], 2) most predominant mutant allele comprises ≥ 50% of the total mutant allele population, and 3) mutagenic efficiency > 20%. For the purpose of this study, a programmable nuclease satisfying all these criteria is referred to as a Predominant MMEJ Allele (PreMA) reagent. Three sticky-end generating TALEN (chrd, mitfa #4 & surf1) and two blunt-end generating CRISPR-Cas9 (surf1 & tyr #2) reagents fell into this category (S1 Table, Fig 2A, Fig 3A). Injecting the chrd TALEN pair (37. 5 pg/arm) resulted in characteristic chrd loss of function phenotypes: Intermediate-Cell-Mass expansion and a smaller head by 1 day post-fertilization [21] (1 dpf; Fig 2B). Median penetrance for Moderate and Severe phenotypes was 15. 8% and 20. 0%, respectively (Fig 2B, S2 Table). Strong MMEJ activation by this TALEN pair was confirmed by subcloning analysis (Fig 2A) – 16/32 recovered mutant reads corresponded to the top predicted 7 bp deletion allele. Similarly, perturbing tyr gene with a CRISPR-Cas9 reagent recapitulated a previously reported, loss of melanin production phenotype, observable by 2 dpf [22] (Fig 3B). Ribonucleoprotein (RNP) delivery at the dose of 300 pg tyr #2 sgRNA and 660 pg Cas9 resulted in Moderate and Severe loss of pigmentation phenotypes in 22. 7% and 50. 0% of embryos respectively (Fig 3B, S2 Table). Subcloning analysis showed 21/24 (88%; Fig 3A) of resulting alleles contained a 4 bp deletion consistent with a strong MMEJ activation by this CRISPR-Cas9. Together with the chrd TALEN results, these data support that MMEJ can be an effective repair pathway in F0 embryos at some genomic loci, irrespective of programmable nucleases used. A subset of these zebrafish reagents described above was prospectively designed using the Bae, et al. algorithm (S1 Table). This algorithm calculates the strength of each pair of microhomology arms (i. e. , Pattern Score) according to the length and GC content of each pair, as well as the length of the intervening sequence. The additive sum of all the possible Pattern Scores is then returned as Microhomology Score. This latter score was found to have positive correlation with the rate of MMEJ activation in HeLa cells [14]. All fourteen prospectively designed reagents had a Microhomology Score of at least 4000 –a median score found on human BRCA1 gene. However, only four of these reagents induced majority MMEJ outcomes as judged by the Microhomology Fraction (S1 Table, S1 Note). We therefore retrospectively analyzed the repair outcomes of these reagents to identify additional factor (s) that would enhance predictability of MMEJ induction. Intriguingly, when the pattern score values clustered closely to one another (i. e. , a flatter Slope Value as calculated according to S2 Note), this was indicative of an unfavorable target for MMEJ activation in zebrafish embryos. Conversely, loci at which Pattern Scores dropped precipitously (i. e. , a steeper Slope Value) were good candidates of MMEJ activation in vivo (p = 0. 0048; S1 Fig). Based on these observations, we hypothesized that locally available microhomology pairs are in direct competition with one another such that overabundance of these pairs is a negative predictor of MMEJ activation. In other words, MMEJ activation is more favorable at loci with one or two predominant microhomology pair (s) (Low Competition loci) rather than many strong microhomology pairs (High Competition loci). To determine whether the zebrafish-based hypothesis was generalizable to human cells (HeLa), we re-analyzed the deep sequencing dataset used to generate the Bae, et al. algorithm [14]. Available results from 90 genomic loci were sorted alphabetically by the names of target genes then divided into two groups: first 50 and the remaining 40. The first group was then used for a retrospective, correlative analysis while the latter was used for an analysis compatible with a prospective study design. Outcomes from the first 50 targets showed a correlation similar to that observed in zebrafish; higher Microhomology Fractions generally correlated with low Slope Values (p = 0. 00001; S2A Fig). This correlation was lost when microhomology arms of 2 bp were included in the analysis (p = 0. 2644; S2B Fig); accordingly, microhomology arms of less than 3 bp were excluded from subsequent analyses. The remaining 40 targets were then binned into High, Medium and Low Competition groups based on quartile distribution of the Slope Value (S2C Fig). In agreement with our Competition Hypothesis, the median Microhomology Fraction was significantly higher in the Low Competition group than in the High Competition group (0. 300 vs 0. 105, p = 0. 011; S2D Fig). Based on this Competition Hypothesis, we designed 20 Low Competition sgRNA targets across 9 genes and analyzed the DSB repair outcomes (S3 Table). Slope Values smaller than -40 was used as the cut-off for Low Competition, as 3 out of 4 previously designed zebrafish targets produced majority MMEJ outcomes in this range (S1 Table and S1 Fig). For initial assessments, we used TIDE (Tracking Indels by DEcomposition) analysis–a chromatogram analyzing tool that estimates proportions of length varying mutant alleles present in a pool of mixed alleles [23]–which revealed that 5 of these sgRNAs against 3 genes (mtg1, tdgf1, ttn. 2 #1, ttn. 2 #2, and ttn. 2 N2B #1) were in the PreMA class. These results were subsequently confirmed by subcloning analyses (S3 Table). Perturbation of tdgf1 (alternatively known as One-eyed Pinhead) causes aberrant, “pinhead” morphology and cyclopia as judged by reduced forebrain protrusion by 1 dpf [24] (Fig 4B). RNP injections of CRISPR-Cas9 at the dose of 300 pg sgRNA and 660 pg Cas9 resulted in highly homogeneous DSB repair outcomes, generating the top-predicted 4bp allele in 28 of 39 clones analyzed (Fig 4A). Aberrant head morphology alone was classified as Weak whereas that in combination with varying degrees of forebrain protrusion was classified as Moderate or Severe phenotypes. Median penetrance for Moderate and Severe morphology was 21. 8% and 11. 4% (Fig 4B, S2 Table), consistent with the subcloning results. We next explored whether these PreMA reagents are useful for recapitulating a more subtle phenotype beyond aberrant gross morphologies observed in the tdgf1 mutants. Splice blockade at the N2B exon of ttn. 2 gene by a synthetic morpholino oligonucleotide was previously reported to reduce the cardiac contractility by ~70% on 2 dpf [25], phenocopying the pickwickm171 mutation [26]. RNP delivery at the dose of 300 pg ttn. 2 N2B #1 sgRNA + 660 pg Cas9 resulted in reduction of the shortening fraction to a comparable degree (Fig 5B). Importantly, RNP delivery of NHEJ-inducing ttn. 2 N2B #2 sgRNA at the same dose only resulted in a more attenuated phenotype, despite it targeting the same exon and having comparable activity (Fig 5; S4 Table). Due to the high editing efficiency, animals injected with these doses of ttn. 2 N2B #1 RNP were not viable in post larval phases. For this reason, animals injected at the lower dose of 75 pg sgRNA + 165 pg Cas9 protein were raised to adulthood. Two F0 founders were successfully outcrossed to wildtype zebrafish. Heterozygous offspring were identified using the dsDNA heteroduplex-cleaving Surveyor assay [27], and the transmission of the top predicted 5 bp deletion allele was confirmed from both founders by subcloning analyses (S3 Fig). We also designed an sgRNA against exon 13 of ttn. 2 (ttn. 2 #2 sgRNA), expected to produce a 12 bp deletion allele as a proof-of-principle for in-frame gene correction (Fig 6A). RNP delivery at the dose of 300 pg sgRNA + 660 pg Cas9 resulted in the induction of this 12 bp deletion allele in 72. 7% of the clones. While the injected animals presented with mild cardiac edema evident by 2 dpf (median rate: 50. 0%; Fig 6B, S2 Table), unlike the N2B #1 sgRNA CRISPR-Cas9 injected animals, these were viable to adult age. These data implicate the utility of PreMA reagents for various applications that require precision gene editing. However, sgRNA design based on the Competition Hypothesis only yielded 5 PreMA reagents out of 20 that were tested (S3 Table, S3 Note). While this represented an improvement over the initial approach solely relying on the Microhomology Score (1 out of 14; S1 Table), we sought to further fine-tune the predictability for the PreMA targets. To this end, we pooled the results from all the programmable nucleases described above (S1 and S3 Tables) and seven Medium ~ High Competition sgRNAs designed as controls based on the Competition Hypothesis (S4 Table). In so doing, we noted that PreMA outcomes were only observed if the two arms of the top predicted microhomology were separated by no more than 5 bp. We subsequently identified the second parameter: high ratio (≥ 1. 5) of the Pattern Scores between the top predicted and second predicted MMEJ alleles for a given locus (Fig 7). Seven out of eight reagents that satisfied both of these parameters were PreMA. Of the nine reagents that satisfied the first parameter but not the second, two were PreMA. All the other thirty reagents that failed to meet the first parameter failed to induce the top predicted MMEJ allele strongly. Most importantly, all the failed cases (i. e. , incorrect predictions according to the original Competition Hypothesis) can be explained using our revised approach (Competition Hypothesis V2; Fig 7C). The Version 2 also captured three PreMA reagents that would have been missed by the original Competition Hypothesis alone, and one PreMA reagent that would have been missed by the Microhomology Score alone. Similar trends were observed using independently collected, previously published deep sequencing dataset from zebrafish [28] and HeLa cells [14] (S4 Fig). To test the generalizability of our findings, we prospectively designed 11 sgRNAs against the human genome (S5 Table) and delivered as RNPs to HEK293T cells. Of the 5 active guides cutting above 20% efficiency, DSBs induced by GJB2 #1 and #2 guides resulted in more homogeneous repair outcomes (Fig 8A and 8B) than any of the 92 guides tested by Bae, et al (S4B Fig) [14]. DSBs at AAVS1 #2 and MYO7A #3, on the other hand, repaired primarily through 1bp indels, consistent with the report by Bae, et al using HeLa cells. Intriguingly, the second most prevalent class of repair at these loci was the top predicted MMEJ allele (Fig 8C and 8D), as identified by subcloning analyses. We thus conclude that the specific trigger for efficient MMEJ-activation may be conserved in vertebrate organisms, albeit with nuances that are yet to be elucidated. The broad potential utility of this updated PreMA Algorithm for MMEJ prediction led us to develop a web-based automated analysis tool called MENTHU (http: //genesculpt. org/menthu/). The tool can also be downloaded and installed on a local computer (www. github. com/Dobbs-Lab/menthu/). MENTHU accepts a user-specified DNA sequence and targeting scheme as input, and outputs recommended CRISPR gRNA target sites that are predicted to result in PreMA type outcomes. We validated the accuracy and functionality of MENTHU against select gRNA sites used in this study using whole exonic sequences as inputs (S6 Table); importantly, the software identified novel PreMA candidate loci against surf1 and tdgf1 where only Group 3 gRNA loci had been found by previous methods. Finally, we conducted a preliminary assessment to examine the prevalence of PreMA loci and found roughly 10% prevalence of such loci among all possible NGG PAM on human CSF2 as well as zebrafish tp53 genes (S7 Table). To date, precision genome engineering is limited by the ability to predictably, efficiently, and reproducibly induce the identical sequence alterations in each and every cell. Here, we demonstrate the feasibility and utility of creating allelic consistency by an MMEJ-centric approach for designing programmable nucleases. While the precise cellular components of the molecular machinery involved in MMEJ remain incompletely understood [8], we provide evidence that we can enrich for MMEJ events by strictly sequence-based queries. We also demonstrate that MMEJ predominant repairs do not operate at the cost of overall mutagenic efficiency; median edit efficiency for PreMA reagents was 91. 4% in zebrafish. As genetically unaltered wildtype zebrafish were used throughout the study, we have no reason to believe that NHEJ should have failed at any tested loci. This is in contrast to the proposal that MMEJ is a back-up pathway to NHEJ [7,8, 16,17,29]. Our findings, on the other hand, are compatible with a previous report wherein MMEJ-specific factors such as PolQ are abundantly expressed in embryonic zebrafish [20]. Interestingly, maternally zygotic PolQ mutant embryos failed to repair DSB at two out of three MMEJ loci, leading to premature deaths [20]. The third locus–which preferentially used a 2 bp microhomology and exhibited more heterogeneous DSB repair outcomes–was able to be repaired at a measurable rate, though significantly less so than in WT embryos. Thus NHEJ and MMEJ may be non-competing, parallel processes with unique triggers. Based on the data presented here, we speculate that there is a reaction-limiting factor for MMEJ that is involved in identifying compatible microhomology pairs on both sides of the DNA double stranded break. In the case of abundantly available local microhomology pairs, this factor may fail to localize to a single suitable pair, thus rejecting the MMEJ activation. As end-resection is required for MMEJ and not for NHEJ [9,17,18], this yet identified factor may be the deciding factor for committing DSB repair through one End Joining pathway to another. This view is similar to a recent report wherein CtIP/Artemis dependent limited end resection was a key trigger for a slow-kinetic Lig1/3 independent NHEJ event that frequently utilized Microhomology to repair a reporter plasmid [30]. In our analysis, the primary driver of this decision making process is the proximity of 2 microhomology arms, further aided by the lack of competing microhomology arms. Successful deployment of the PreMA reagents makes it possible to directly dictate the reading frame or to do in-frame gene manipulations on endogenous targets. Even assuming a somewhat modest outcome of 50% edit efficiency in which 50% of the mutant allele pool is of the desirable allele, more than 10% of the cell population will be homozygous for this desired allele. Conversely, many real-life gene editing applications would require only one of the diploid copies to be corrected. In these settings under the same assumptions, just 11 viable cells are needed to achieve 95% confidence for establishing the right clone, bringing the idea of precision molecular surgery closer to reality. Our present study expands upon the current state-of-art understanding for MMEJ activation and demonstrates the ability to prospectively design robustly active PreMA reagents in-vivo. We also provide evidence that this 2-component approach may be broadly applicable beyond zebrafish; testing of the true generalizability of our approach will be facilitated by our web-based application, MENTHU (http: //genesculpt. org/menthu/). Importantly, MENTHU allows users to flexibly define a PAM sequence and the cut site (in nts from PAM) so as to accommodate potential future variants of the CRISPR system. Active investigations are underway to accommodate alternative or more lax PAM requirements, such as the case with xCas9–a recently described variant of Cas9 that may function efficiently on an NG PAM [31]. As MMEJ-based loci are inherently restricted to genomic locations that leverage endogenous sequence contexts, availability of more flexible programmable nucleases will become the key for broadening the utility of PreMA reagents. We provide strong evidence to support the utility of the MMEJ-centric approach beyond phenotype-genotype correlations in F0 animals. We envision this approach to be useful for: 1) studying the effects of homozygous gene knock-out in culture cells (as opposed to more common, compound heterozygous loss-of-function cell lines), 2) rapid small molecule screening in F0 animals as a complimentary approach to studying in germline mutant animals, 3) globally sharing and reproducing gene knock-out cell and animal lines, 4) pathway dissection for MMEJ, and finally, 5) human gene therapy. The animal studies were conducted following guidelines and standard procedures established by the Mayo Clinic Institutional Animal Care and Use Committee (Mayo IACUC). The Mayo IACUC approved all protocols involving live vertebrate animals (A23107, A 21710 and A34513). For the purpose of this study, microhomology is defined as any endogenous direct sequence repeats of ≥ 3 bp surrounding a DSB site. ≤ 2 bp direct sequence repeats were not considered sufficient substrates of MMEJ activation based on our initial analyses of the DSB repair outcomes by previously designed programmable nucleases. Correlation for Microhomology Fraction vs the Slope Value was tangentially stronger when only ≥ 3 bp arms were considered (r2 = 0. 382 vs r2 = 0. 353; S1 Fig) in zebrafish, whereas the correlation was lost when 2 bp arms were considered in HeLa cells (r2 = 0. 339 vs r2 = 0. 034; S2 Fig). All zebrafish (Danio rerio) were maintained in accordance with protocols approved by the Institutional Animal Care and Use Committee at Mayo Clinic. Zebrafish pairwise breeding was set up one day before microinjections and dividers were removed the following morning. Following microinjections, the fertilized eggs were transferred to Petri dishes with E3 media [5 mM NaCl, 0. 17 mM KCl, 0. 33 mM CaCl2, and 0. 33 mM MgSO4 at pH 7. 4] and incubated at 28. 5 °C. All subsequent assays were conducted on fish less than 3 dpf, with the exception of assessing for germline transmission. In this case, injected founders were raised to adulthood per the standard zebrafish husbandry protocol. All of the oligonucleotides used for this study were purchased from IDT (San Jose, CA). Upon arrival, they were reconstituted into 100μM suspensions in 1x TE and stored at -20 °C until use. pT7-gRNA was a gift from Wenbiao Chen (Addgene plasmid # 46759). Given that the minimum requirement for the T7 promoter is a single 5’ G, the GG start on this vector was mutagenized via site-directed mutagenesis (SDM) to accommodate GA, GC, GT starts, using Forward and Reverse primers given (S8 Table). Platinum Pfx DNA Polymerase (Invitrogen 11708013. Carlsbad, CA) was used for 20 cycles of PCR amplification with the Tm of 60 °C and extension time of 3 minutes. DpnI (NEB R0176. Ipswich, MA) was subsequently added to reaction prior to transforming DH5α cells. The target sequence was cloned in as previously described, with the exception of conducting oligo annealing and T4 ligation (NEB M0202. Ipswich, MA) in 2 separate steps. In each case, transformed cells were cultured with Carbenicillin, and plasmids were purified with Plasmid Mini Kit (Qiagen 12123. Hilden, Germany). TALEN constructs were generated using the FusX kit (Addgene # 1000000063) as previously described [32]. In short, RCIscript-GoldyTALEN was linearized with BsmBI (NEB R0580. Ipswich, MA) along with 6 triplet RVD (Repeat-Variable Diresidue) plasmids. Subsequently, they were ligated together in one reaction by a modified Golden-Gate Assembly. Blue-White colony screening with X-Gal/IPTG, colony PCR and finally pDNA sequencing were done to ascertain the correct assembly. pT3TS-nCas9n (a gift from Wenbiao Chen: Addgene plasmid # 46757) was linearized with XbaI (NEB R0145. Ipswich, MA), whereas TALEN constructs were linearized with SacI-HF (NEB R3156. Ipswich, MA) and sgRNA vector with BamHI-HF (NEB R3136. Ipswich, MA). Tyr sgRNA #2 –a construct made in the Essner Lab–was linearized with HindIII (NEB R0104. Ipswich, MA). RNA was made using T3 mMessage mMachine kit (Ambion AM1348. Foster City, CA) or HiScribe T7 High Yield RNA synthesis kit (NEB E2040. Ipswich, MA) according to manufacturer’s protocols with the addition of RNA Secure to the reaction (Ambion AM7010. Foster City, CA). To purify RNA, phenol-chloroform extraction was performed using Acid Phenol, Chloroform, and MaXtract High Density Tubes (Qiagen 129046. Hilden, Germany). RNA was then precipitated with Isopropanol at -20 °C, pelleted, air dried and resuspended into nuclease free water. The quality and quantity of RNA were ascertained by using a Nanodrop spectrophotometer and running aliquot on agarose gel. Each batch of RNA was aliquoted into small single use tubes and stored at -80 °C until the morning of microinjections. sgRNA was thawed on ice in the morning of microinjections. This was then diluted to the concentration of 300 ng/μL in Duplex Buffer [100 mM KCH3COO, 30 mM HEPES at pH 7. 5]. Appropriate folding of sgRNA was induced by heating it to 95 °C for 5 minutes and gradually cooling the solution to room temperature. Equal volumes of sgRNA and 0. 66 mg/mL Alt-R S. p. Cas9 Nuclease 3NLS (IDT 1074181. San Jose, CA) in Cas9 Working Buffer [20 mM HEPES, 100 mM NaCl, 5 mM MgCl2,0. 1 mM EDTA at pH 6. 5] were mixed and incubated at 37 °C for 10 minutes. RNP solutions were subsequently kept on ice until immediately before use. RNA was thawed on ice in the morning of microinjections. TALEN mRNA was diluted to working concentrations in the range of 12. 5 ng/μL to 100 ng/μL in Danieau solution [58 mM NaCl, 0. 7 mM KCl, 0. 4 mM MgSO4,0. 6 mM Ca (NO3) 2,5. 0 mM HEPES at pH 7. 6]. sgRNA and nCas9n mRNA were mixed and diluted to the final concentrations of 150 ng/μL and 100 ng/μL, respectively, in Danieau solution. These were all kept on wet ice until immediately before use. Microinjections were carried out as previously described [33]. In short, 1-cell stage fertilized embryos were harvested and aligned on an agarose plate with E3 media. In the case of CRISPR-Cas9 reagents, either 1 or 2 nL was delivered to the cell. In the case of TALEN reagents, 1 ~ 3 nL was delivered to the yolk mass. They were then transferred to Petri dishes in E3 media for incubation at 28. 5 °C. Dead and/or nonviable embryos were counted and removed each subsequent morning. Each experiment was conducted in at least a technical triplicate and a biological duplicate. Detailed outcomes are provided in S4 Table. Gross phenotypes were scored visually on either 1 dpf or 2 dpf using a standard dissecting microscope. Subsequently, representative pictures were taken with Lightsheet Z. 1 (Zeiss 2583000135. Oberkochen, Germany). Shortening Fractions were scored as previously reported [34]. In short, live 2 dpf larvae were immobilized and positioned in 3% methylcellulose. An Amscope camera (MU1403. Irvine, CA) mounted on a Leica Microscope (M165. Wetzlar, Germany) was used to capture a 15 second clip of the beating heart at 66 fps. These clips were subsequently used to measure the distance of the long axis along the ventricle at maximum dilation and maximum contraction using ImageJ software [35]. Shortening Fraction was calculated as below: ShorteningFraction=100* (1−DistanceatMaximumShorteningDistanceatMaximumDilation) Shortening Fractions from 5 cycles were averaged for each animal. Typically, 8 uninjected wildtype fish and 8 injected fish were randomly collected without prior screening for phenotype. Chorion was predigested with 1 mg/mL Pronase at room temperature as needed. 1 ~ 3 dpf animals were then sacrificed for individual DNA extractions in 100 mM NaOH for 15 minutes at 95 °C. Equal volumes of 8 fish DNA from the same condition were then mixed and used as templates for PCR with either MyTaq (Bioline BIO-21108. London, UK), Phusion (NEB M0530. Ipswich, MA), or KOD (EMD Millipore 71085. Burlington, MA) polymerases per manufacturer’s protocols. The PCR amplicon was resolved on agarose gel, gel extracted with either Monarch DNA Gel Extraction Kit (NEB T1020. Ipswich, MA) or QiaEx II Gel Extraction Kit (Qiagen 20021. Hilden, Germany), and subsequently sent out for sequencing. The chromatograms from both uninjected and injected amplicons were used for TIDE analysis [23]. Alternatively, purified amplicons were used for subcloning analysis with either Topo-TA Cloning Kit (Thermo Fisher Scientific 451641. Waltham, MA) or StrataClone PCR Cloning Kit (Agilent 240205. Santa Clara, CA) per manufacturer’s protocols. Resultant white to pale blue colonies by Blue-White screening were subjected to colony PCR with M13F and R primers, using MyTaq polymerase. Once successful amplification was confirmed on agarose gel, these amplicons were sent out for sequencing either with M13F, M13R or endogenous gene target primers. RNP containing N2B sgRNA #1 was prepared at 4x diluted dose as described above. Following microinjections, viable fish were raised to sexual maturity. Both F0 founders we attempted to out cross successfully mated and produced viable embryos. DNA was extracted from all viable embryos on 1 dpf, and individual DNA was used as template for PCR amplification using MyTaq Polymerase. Once the thermocycling ran to completion, the amplicons were melted by heating to 95 °C and re-annealed by a gradual step-wise cooling. Surveyor assay [27] was conducted per the manufacturer’s protocol (IDT 706025. San Jose, CA), and the results were analyzed by resolving the post-digest amplicons on agarose gel. Amplicons from 4 heterozygous offspring each were subcloned, and 5 colonies each were sent for Sanger Sequencing to confirm successful transmission of the 5 bp deletion allele. For zebrafish dataset, sgRNA screen SRA files were obtained from NCBI’s Short Read Archive (Accession: PRJNA245510) [28]. These files were converted to the fastq format with fastq-dump command using—split-spot function under SRA Toolkit (NCBI. Bethesda, MD). The fastq files were then uploaded onto Cas-Analyzer (http: //www. rgenome. net/cas-analyzer/) and analyzed with Comparison range of 25 ~ 40 and Minimum frequency of 1 [36]. Following number of reads were recorded: total, total mutant, total top predicted allele. A top predicted allele was allowed to be included so long as the read contained no more than 2 polymorphisms on the analysis window AND the polymorphisms did not fall on the microhomology arms. Subsequently, the calculated mutagenic efficiency was plotted against the reported efficiency (r2 = 0. 306). Of 122 targets designed by Gangnon, et al, following were excluded to arrive to the 34 targets that were used for analysis presented in S4 Fig Panel A: non-NGG targets (36 loci), targets that did not align to WT consensus sequence (GRCz11; 8 loci), targets with total recovered read counts less than 1% of expected (7 loci), high rate of permutation outside of the target site (1 locus), targets that did not have good agreements between calculated and reported (i. e. , fell beyond 99% Confidence Interval; 10 loci), targets that had less than 5% calculated AND reported mutagenic efficiencies (26 loci). The HeLa cell dataset [14] was obtained from Dr. Kim in the form of excel spread sheet with aligned sequence outputs +/- 25 bp of the predicted cut site. Following number of reads were recorded: total, total mutant, total top predicted allele with 2 bp microhomology, and total top predicted allele with 3 bp or longer microhomology. As with zebrafish dataset, top predicted allele was allowed to be included so long as the read contained no more than 2 polymorphisms on the analysis window AND the polymorphisms did not fall on the microhomology arms. Of the 92 targets, following were removed to arrive to 74 targets that were used for analyses presented in S2 Fig and S4 Fig Panel B: targets with total recovered read counts less than 1% of expected (2 loci), and targets that had less than 20% mutagenic efficiency (16 loci). There were no targets with non-NGG PAM, no alignment against consensus sequence, nor a high rate of permutation outside of the predicted cut site. HEK293T cell line was purchased from ATCC (Manassas, VA) and maintained in DMEM (Invitrogen. Carlsbad, CA) with 10% Fetal Bovine Serum (Sigma. St. Louis, MO). DAPI stain was used to check for mycoplasma contamination. RNP transfection was conducted as follows in a 48-well format using Lipofectamine CRISPRMAX reagent (Invitrogen CMAX00015. Carlsbad, CA). In vitro transcribed sgRNA was diluted to 2 μM concentration in Duplex Buffer. Secondary structure was induced by heating it to 95 °C for 5 minutes and gradually cooling it to room temperature. 3. 0 μL of sgRNA was then complexed with 3. 0 μL of 2 μM Alt-R S. p. Cas9 Nuclease V3 (IDT 1081058. San Jose, CA) in 42. 8 μL OPTI-MEM (Life Technologies. Carlsbad, CA) and 1. 2 μL Cas9 Plus Reagent. This mixture was incubated for 5 minutes at 25 °C. 2. 4 μL of CRISPRMAX reagent and 47. 6 μL OPTI-MEM was then added to the RNP, transferred to empty wells, and further incubated for 20 minutes at 25 °C. 200 μL cell suspension at 400,000 cells / mL in complete medium were subsequently added to each well. The dosing of RNP was consistent for all targets except for both GJB2 targets wherein 1 μM each of sgRNA and Cas9 protein was used. HEK293T cells were harvested 24 hour post transfection for gDNA extraction using DNeasy Blood & Tissue Kit (Qiagen 69506. Hilden, Germany). 20 ng of gDNA was used as a template for PCR with KOD polymerase per manufacturer’s protocol. The PCR amplicon was resolved on agarose gel, gel extracted with Monarch DNA Gel Extraction Kit and subsequently sent out for sequencing. The chromatograms from both uninjected and injected amplicons were used for TIDE analysis [23]. For sgRNAs that showed > 20% activity by TIDE, single A overhang was added to the 3’ end of purified amplicons by incubating them with MyTaq polymerase at 72 °C for 15 minutes. They were then used for subcloning analysis with StrataClone PCR Cloning Kit. 96 resultant white to pale blue colonies by Blue-White screening were subjected to colony PCR with endogenous primers using MyTaq polymerase. Once successful amplification was confirmed on agarose gel, these amplicons were subjected to T7E1 assay [27]. Briefly, 2. 5 μL each of colony PCR amplicon and wildtype amplicon were heteroduplexed in 1x NEB 2. 0 Buffer (25 μL). This was incubated for 15 minutes at 37 °C with 0. 5 μL T7 Endonuclease I (NEB m3020. Ipswich, MA) and 4. 5 μL dH2O. The digested amplicon was resolved on 2% agarose gel. Number of colony PCR-positive clones and digest positive clones are reported in S5 Table. Some of the digest positive clones were then sent for sequencing to ascertain the nature of mutation. 2 targets (CSF2 #1 and MYO7A #4) that did not meet the 20% edit efficiency cutoff nonetheless produced statistically significant aberrant sequence peaks by TIDE analysis (p < 0. 001). Summary outcomes for Top MH Fraction calculation based on estimated alleleic prevalence is given in S5 Table. We developed a software tool, MENTHU (MMEJ kNockout Target Heuristic Utility), to automate calculations required to implement the 2-component PreMA strategy: 1) identification of top predicted microhomology arms separated by ≤ 5 bp of intervening sequence, and 2) identification of “low competition” target sites (i. e. , with a #1-ranked to #2-ranked Pattern Score ratio ≥ 1. 5). We designed MENTHU to first compute two of same sequence-based parameters (Pattern Score and Microhomology Score) used in the algorithm of Bae, et al. , (which are computed online by the RGEN online tool, http: //www. rgenome. net) To do so, we used R [37] to re-implement and modify the original Python source code provided in S3 Fig of the original publication [14]. The MENTHU webserver operates under R version 3. 4. 1 and RShiny [38] v1. 0. 5. The MENTHU code was built through RStudio [39] v1. 1. 442. Details regarding specific R package versions, complete documentation and a full downloadable version of MENTHU for local installation are provided at www. github. com/Dobbs-Lab/menthu/. MENTHU v2. 0 can be freely accessed online at http: //genesculpt. org/menthu/. To preliminarily assess the abundance of PreMA loci, MENTHU was locally run to screen the sequences of two genes: human colony stimulating factor 2 (CSF2; Gene ID– 1437) and zebrafish tumor protein p53 (tp53; Gene ID– 30590). MENTHU was run twice on each gene: exonic target screen and whole gene target screen. A custom R script was used to mine the MENTHU results in a. csv format to determine both the amounts of total targetable sites by spCas9 (i. e. , total number of unique cut sites with NGG PAM on either strand) and the subset of those predicted to be PreMA. All of the statistical analyses were carried out using JMP software (SAS Institute. Cary, NC). In all instances, p-values were calculated assuming non-Gaussian Distributions. Wilcoxon Each Pair calculation was used for multiple group comparisons with adjusted p-values.

New gene editing tools precisely break DNA at pre-defined genomic locations, but cells repair these lesions using diverse pathways that often lead to unpredictable outcomes in the resulting DNA sequences. This sequence diversity in gene editing outcomes represents an important obstacle to the application of this technology for human therapies. Using a vertebrate animal as a model system, we provide strong evidence that we can overcome this obstacle by selectively directing DNA repair of double-stranded breaks through a lesser-described pathway termed Microhomology-mediated End Joining (MMEJ). Unlike other, better-understood pathways, MMEJ uses recurring short sequence patterns surrounding the site of DNA breakage. This enables the prediction of repair outcomes with improved accuracy. Importantly, we also show that preferential activation of MMEJ is compatible with effective gene editing. Finally, we provide a simple algorithm and software for designing DNA-breaking reagents that have high chance of activating the MMEJ pathway. We believe that the MMEJ-centric approach to be broadly applicable for a variety of gene editing applications both within the laboratory and for human therapies.

Abstract Introduction Results Discussion Materials and methods

biotechnology genome engineering fish engineering and technology nucleases enzymes synthetic biology dna-binding proteins vertebrates cloning enzymology microhomology-mediated end joining animals synthetic bioengineering animal models osteichthyes model organisms experimental organism systems dna molecular biology techniques synthetic genomics research and analysis methods bioengineering synthetic genome editing artificial gene amplification and extension proteins molecular biology genetic loci subcloning talens biochemistry zebrafish eukaryota hydrolases nucleic acids polymerase chain reaction genetics biology and life sciences dna repair non-homologous end joining organisms

2018

Robust activation of microhomology-mediated end joining for precision gene editing applications

Brain metastasis (BM) is a major complication of lung adenocarcinoma (LAD). An investigation of the pathogenic mechanisms of BM, as well as the identification of appropriate molecular markers, is necessary. The aim of this study was to determine the expression patterns of microRNAs (miRNAs) in LAD with BM, and to investigate the biological role of these miRNAs during tumorigenesis. miRNA array profiles were used to identify BM-associated miRNAs. These miRNAs were independently validated in 155 LAD patients. Several in vivo and in vitro assays were performed to verify the effects of miRNAs on BM. We identified six miRNAs differentially expressed in patients with BM as compared to patients with BM. Of these, miR-4270 and miR-423-3p were further investigated. miR-4270 and miR-423-3p directly targeted MMP19 and P21, respectively, to influence cell viability, migration, and colony formation in vitro. miR-4270 downregulation and miR-423-3p upregulation was associated with an increased risk of BM in LAD patients. Thus, our results suggested that miR-4270 and miR-423-3p might play an important role in BM pathogenesis in LAD patients, and that these miRNAs might be useful prognostic and clinical treatment targets. MicroRNAs (miRNAs) are small, (18–23 nt), single-stranded noncoding RNAs [1]. miRNAs play critical roles in almost all important cellular processes, including tumorigenesis [2]. Compared to long non-coding RNAs (lncRNAs; >200 nt), miRNAs are more stable in vivo [1–2]. As miRNAs are widely distributed in various bodily fluids, such as spit and serum, miRNA-based tumor detection techniques may have great clinical value. Mature miRNAs combine with the 3' -UTR sequences of target genes, forming RNA-induced silence compounds; these compounds regulate gene transcription or degrade cytoplasmic mRNAs, thereby affecting protein synthesis [1–3]. Several differentially expressed miRNAs in malignant human tumors have been identified [4−6]. In general, a single miRNA can target and silence a series of target genes, granting miRNAs extensive control of various cellar processes, including proliferation, apoptosis, and tumor metastasis [7−9]. Emerging evidence indicates that miRNAs may play an important role in cancer pathogenesis [8. 9]. Besides functioning as signal molecules in tumor tissues, exome-encoded miRNAs are also secreted in bodily fluids [10,11]. Due to the stability of miRNAs in vivo, miRNAs are promising candidate biomarkers for human cancers [12]. Lung cancer is one of the most aggressive malignant cancers worldwide [13]. Non-small cell lung cancer (NSCLC), the main pathological subtype of lung cancer, includes two major histologic subtypes: lung adenocarcinoma (LAD) and squamous cell carcinoma [14,15]. Despite advancements in clinical management, the overall survival of NSCLC patients remains poor, with a five-year survival rate of less than 15% [14,15]. Metastasis is the main cause of NSCLC-associated death. In addition to the lungs, NSCLC tends to metastasize in the brain, bones, and liver [16]. Brain metastasis (BM) is a frequent complication of LAD; the incidence of locally-advanced LAD with BM is as high as 30–50% [15–17]. BM is often associated with severe neurologic and cognitive difficulties, as well as disappointing survival rates [18]. Radiotherapy is the standard treatment for NSCLC-associated BM, but long-term survival remains low, with a median survival time of about 2. 4–4. 8 months [17,18]. Thus, it is critically important to improve patient stratification. Although the patient stratification process is greatly is greatly simplified by molecular markers, the identification of new molecular markers remains difficult. Therefore, to improve NSCLC management, it is important to identify biomarkers that enable the accurate detection of early alterations in the molecular characteristics of BM-associated tumors. Several miRNAs useful for NSCLC tumor classification, metastasis prediction, and patient prognosis have already been described [19−21]. However, it is still important and useful to identify reliable predictive miRNA markers for BM NSCLC [22]. BM occurs significantly more frequently in stage IIIA and IIIB (N2) LAD patients than in squamous patients [23,24]. Indeed, certain tumor alterations, linked to BM risk by pertinent biomarkers, could be integrated into clinical decisions as prognostic indicators, facilitating the development of personalized treatments and follow-up plans. Therefore, we focused on the regulatory roles of miRNAs in the cellular activities that impacted LAD metastasis. We also explored the use of miRNAs as biomarkers and in therapy. To identify the miRNAs associated with BM LAD, we compared miRNA expression profiles in 32 LAD patients with BM to those of 55 patients without BM (NBM) (Table 1). These 86 patients were considered the" discovery group." Fluorescent signals were normalized using the median center gene tool in Cluster 3. 0, and evaluated using the significance analysis of microarrays (SAM) method, with a false discovery rate (FDR) threshold of 0, a fold-change of ≥2 or ≤0. 5, and a P-value <0. 05. We identified six miRNAs that were differentially expressed in the BM LAD tissues as compared to the NBM tissues (Fig 1A, S1 Data). Five of these miRNAs (miR-214, miR-423-5p, miR-210, miR-193a-5p, and miR-423-3p) were significantly upregulated in the BM group, while one miRNA (miR-4270) was downregulated. We validated these results in 68 BM LAD samples using quantitative real-time polymerase chain reaction (qRT-PCR). These 60 samples were considered the" validation group." The qRT-PCR results for the validation group were consistent with the analysis of the discovery group: the same six miRNAs were identified as differentially expressed (Fig 1B; Table 1). These results suggested that only six miRNA markers might be necessary to effectively identify BM LAD in Chinese patients. For subsequent analyses, all samples were split into high- and low-expression groups based on the median fluorescence signal values or the qRT-PCR results for the six differentially expressed miRNAs. Univariate cox regression analyses of the discovery dataset, the validation dataset, and both datasets combined indicated that the upregulation of miR-214, miR-423-5p, miR-210, miR-193a-5p, and miR-423-3p, as well as the downregulation of miR-4270, were associated with an increased risk of BM in LAD (Fig 1C; S1A and S1B Fig; S1 Table). The area under the curve (AUC) values, which reflected the predictive accuracy of the six differentially expressed miRNAs, were 0. 926,0. 870, and 0. 902 for the discovery, validation, and combined groups, respectively (Fig 1D). Thus, these miRNAs accurately predicted BM in LAD patients. The sensitivity, specificity, and false discovery rate of the six micro-RNAs combined were better than those of the upregulated miRNAs alone or the downregulated miRNA alone when predicting BM with a multivariate cox regression analysis (S2–S4 Tables). As little is known about the roles of miR-4270 and miR-423-3p in tumor invasion and, these miRNAs were selected for further investigation [25]. To investigate whether miR-4270 and miR-423-3p were suitable biomarkers of BM in LAD patients, we first analyzed miR-4270 and miR-423-3p expression levels in LAD samples. miR-4270 expression levels in NBM LAD patients were significantly higher those of BM LAD patients in all three groups (the discovery group, the validation group, and the combined group; Table 2). miR-4270 expression levels were not correlated with age, gender, T stage, histologic grade, or lymph node ratio (Table 2). Across all parameters, multivariate Cox proportional hazard regression analyses suggested that miR-4270 upregulation led to significantly positive prognostic factors, and was independent of BM in all three groups (Table 3). However, miR-423-3p upregulation was significantly correlated with the number of brain metastases (Table 4). miR-423-3p expression levels were also not correlated with age, gender, T stage, histologic grade, or lymph node ratio (Table 4). Multivariate cox proportional hazard regression analysis suggested that miR-423-3p upregulation was linked to poor survival in BM LAD patients, and was an independent prognostic factor for overall survival (OS) of LAD patients (Table 5). We next analyzed the molecular functions of the differentially expressed miRNAs during BM pathogenesis in a lung adenocarcinoma mouse model. Two LAD cell lines, ANIP-973, expressing a relatively high level of miR-4270 (Fig 2A), and NCI H1299, expressing a relatively low level of miR-423-3p (Fig 3A), were selected for further experiments. A stable miR-4270-knockdown ANIP-973 cell line (Fig 2B) and a stable miR-423-3p-overexpression H1299 cell line (Fig 3B) were produced for in vivo experiments. Subcutaneous injection of ANIP-973 and H1299 cells into BALB/c mice produced transplanted tumors within one week. Tumor volumes were measured weekly, and mice were euthanized after six weeks. Both miR-4270 knockdown and miR-423-3p overexpression increased the growth of LAD tumors in vivo (Fig 2C; Fig 3C). Immunohistochemistry (IHC) results indicated that both the downregulation of miR-4270 and the upregulation of miR-423-3p upregulated the cell proliferation marker Ki-67. (Fig 2D; Fig 3D). We then injected 106 luciferase-labeled cells into the tail veins of nude mice. Mice were euthanatized after six weeks. Luciferase activity was used to evaluate the tumor burden in all mouse organs. The lung, liver, adrenal gland, and bone metastasis burdens were significantly higher in the mice injected with miR-4270-knockdown cells or with miR-423-3p-overexpression cells, as compared with the control group (Figs 2E and 3E). As expected, the tail-vein injection of either miR-4270-knockdown ANIP-973 cells or miR-423-3p-overexpression H1299 cells significantly increased BM risk (Figs 2E and 3E). These results suggested that miR-4270 and miR-423-3p play important roles in the LAD growth and metastasis in vivo, especially with respect to BM progression. To further explore the biological functions of miR-4270 and miR-423-3p, we transferred miR-4270 and miR-423-3p mimics into LAD cells. qRT-PCR results indicated that miR-4270 and miR-423-3p expression levels increased significantly in cells transfected with miRNA mimics (Fig 4A; Fig 4E). Transfection of H157 and H1299 cells with the miR-4270 mimic for 48,72, and 96 hours significantly inhibited cell viability (Fig 4B). Cell colony formation and migration ability were also suppressed after transfection with a miR-4270 mimic (Fig 4C and 4D; S2A Fig). Transfection of H157 and H1299 cells with the miR-423-3p mimic had the opposite effect: overexpression of miR-423-3p not only enhanced the cell viability, but also increased colony formation and cell migration (Fig 4F–4H; S2B Fig). We transfected LAD cells with miR-4270 and miR-423-3p inhibitors to validate our mimic-transfection results. As expected, the inhibition of miR-4270 expression enhanced the malignant phenotype of LAD in vitro, increasing cell growth, colony formation, and cell migration (Fig 5A–5D; S2C Fig). In contrast, the inhibition of miR-423-3p expression in LAD cells significantly suppressed the malignant phenotype (Fig 5E–5H; S2D Fig). These results indicated that miR-4270 inhibited the proliferation, migration, and invasion of LAD cells in vitro, while miR-423-3 promoted these malignant activities. To identify proteins regulated by miR-4270 and miR-423-3p, we first searched the publicly available databases TargetScan, PicTar, and miRanda. We found that 3’-UTR of one of the significantly differentially expressed genes, MMP-19, included a sequence complementing miR-4270. Similarly, a sequence complementing miR-423-3p appeared in the 3’-UTR of another differentially expressed gene, P21. Based on this result, as well as our transcriptome analysis, MMP-19 and P21 were selected as potential miRNA targets and further analyzed (Fig 6A and 6B). Transfection with miR-4270 and miR-423-3p mimics decreased the mRNA and protein expression levels of MMP-19 and P21, respectively (Fig 6C and 6D). Conversely, the mRNA and protein expression levels of MMP19 and P21 were significantly increased after transfection with inhibitors of miR-4270 and miR-423-3p, respectively (Fig 6E and 6F). We next used a dual-luciferase reporter assay investigate the regulatory effects of miR-4270 on MMP-19, and of miR-423-3p on P21. The overexpression of miR-4270 in H157 cells suppressed the activity of the luciferase reporter fused to the 3’-UTR of wild-type MMP19, but not the activity of the luciferase reporter fused to the 3’-UTR of mutant MMP19. Similarly, miR-423-3p suppressed the activity of the luciferase reporter fused to the 3’-UTR of wild-type P21, but not the activity of the luciferase reporter fused to the 3’-UTR of mutant P21 (Fig 6G and 6H). Moreover, MMP-19 was more strongly expressed in upregulated miR-4270 xenograft tumors than in controls expressing miR-4270 normally (Fig 6I). In contrast, P21 was more weakly expressed in miR-423-3p-overexpressed cell-derived tumors than in tumors expressing normal levels of miR-423-3p (Fig 6J). Thus, both our in vitro and in vivo results suggested that MMP19 was a direct target of miR-4270, and that P21 was a direct target of miR-423-3p. Rescue experiments were performed to confirm that MMP-19 and P21 were the functional targets of miR-4270 and miR-423-3p, respectively. H157 cells were transfected with a miR-4270 mimic, a control plasmid, or a plasmid expressing MMP-19. The downregulation of MMP-19 and the reduced cell migration caused by miR-4270 overexpression were markedly reversed after transfection with the MMP-19 plasmid (Fig 7A and 7B). To further explore the clinical significance of MMP-19 expression, we used IHC analysis to evaluate differences in MMP-19 expression among identical formalin-fixed paraffin-embedded (FFPE) tissue types from the 68 LAD patients. We found that MMP-19 expression was significantly higher in tissues from BM LAD patients, as compared to NBM LAD patients (Fig 7C; S5 Table). Moreover, MMP-19 expression was positively correlated with histologic grade (S5 Table). We also performed a Spearman correlation coefficient analysis to measure the association between miR-4270 expression and MMP-19 expression in LAD tissues. miR-4270 expression levels were inversely correlated with MMP-19 expression levels (Fig 7D). This was consistent with the observation that P21 mRNA and endogenous protein expression levels in the H1299 cell line decreased after mimic transfection, and were rescued after transfection with the P21 expression plasmid (Fig 7E). Thus, the migration and invasion induced by mimic transfection were reversed by transfection with the P21 expression constructs (Fig 7F). P21 staining was negative or weak in NBM LAD tissue, but positive in BM LAD tissue (Fig 7G; S5 Table). Furthermore, P21 expression was negatively correlated with the LAD lymph node metastasis ratio (S5 Table). Finally, Spearman analysis indicated that miR-423-3p expression levels were inversely correlated with P21 levels in the 68 LAD specimens (Fig 7H). BM is becoming increasingly prevalent as systemic diseases become better controlled, because large molecules, such as antibodies, are used to treat primary cancers [26,27]. Due to the blood-brain barrier (BBB) and the unique brain microenvironment, novel approaches are required to treat BMs [18,26]. The ability to identify patients at risk of BM might lead to novel prophylactic interventions, mitigating morbidity and mortality. However, previous efforts to reveal the etiopathogenesis of NSCLC BM, and to predict which NSCLC patients will develop BM [26], have been mostly unsuccessful. Recently, miRNAs have emerged as a prominent class of gene regulators [1,2]. Bioinformatics analyses have suggested that 30% of all mammalian protein-coding genes may be regulated by miRNAs [3]. Due to the multimodal downstream signaling effect, miRNAs may be useful for human cancer prognoses as well as treatments [4,5]. Indeed, some studies have investigated the mechanistic roles of miRNAs in the BM phenotype [22,23]. Several differentially expressed miRNAs have previously been identified in NSCLC patients (e. g. miR-1, miR-137, and Let-7a) [28–29]. However, few studies have examined the association between miRNA expression and BM risk in NSCLC patients. Although some BM-associated miRNAs (e. g. miR-328 and miR-326) in NSCLC patients using miRNA arrays [22,30], sample sizes in these previous studies were relatively small (n = 12). Here, we used miRNA array profiling and real-time PCR validation to identify miRNAs associated with BM in a relatively large sample (n = 155). Notably, most of the differentially expressed miRNAs identified in this study differed from those previously reported [22]. Of these miRNAs, five (miR-193a-5p, miR-210, miR-214, miR-423-3p, and miR-423-5p) were significantly upregulated in BM LAD as compared to NBM LAD, and one miRNA (miR-4270) was significantly downregulated. In addition, the mature products of miR-4270 and miR-423-3p were significantly differentially expressed in the BM LAD tissues as compared to the NBM LAD tissues. miR-4270 expression levels in BM LAD patients were significantly lower than in NBM LAD. However, higher levels of miR-423-3p expression were significantly correlated with increased brain metastases. Multivariate cox proportional hazard regression analysis of each of these parameters indicated that low miR-4270 expression and high miR-423-3p expression were significantly unfavorable prognostic factors for LAD, independent of BM. Therefore, the downregulation of miR-4270 and the upregulation of miR-423-3p in BM-positive patients suggested that these miRNAs might be involved in ‘‘brain-seeking” metastatic potential [31,32]. That is, miR-4270 and miR-423-3p may be associated with the pathogenesis of BM from primary LAD, and are thus potential biomarkers. Here, miR-4270 downregulation and miR-423-3p upregulation were associated with the development of BM from primary LAD. Intriguingly, previous studies have shown that miR-4270 is upregulated in breast cancer and the peritoneal metastasis of gastric cancer [33,34], indicating that different levels of miR-4270 expression might be associated with different cancers. In addition, miR-423-3p downregulation has been reported during NSCLC lymph node metastasis [35], suggesting that different levels of miR-423-3p expression might be associated with different stages of NSCLC metastasis. miR-423-3p might be a useful biomarker of the difference between hereditary and non-hereditary breast cancers [36]. miR-423-3p might also contribute to LAD tumor progression through similar signaling pathways to those as observed in laryngeal hepatocellular carcinomas [37,38]. The role of miRNAs in the biology of BMs has been established in previous studies of various primary tumor types [39–42]. In lung cancer, MiRNA-378 promotes brain metastases in NSCLC by upregulating MMP-7, MMP-9, and VEGF and downregulating Sufu; these genes are critically involved in angiogenesis and extracellular matrix invasion [43]. miRNA-145 downregulation leads to LAD progression, and promotes BM formation [44]. Similarly, changes in miR-1258 expression are directly correlated with the upregulation of heparanase, a prometastatic enzyme found in BM breast cancer cells known to degrade heparan sulfate chains; this degradation affects the cytoskeleton and renders cells more capable of crossing the BBB [45]. Our clinical analyses and mouse models both suggested that miR-4270 and miR-423-3p not only play an important role in BM pathogenesis, but are potential RNA markers of NSCLC patients at a high risk for BM. The miRNAs identified herein are potential diagnostic markers and targets for BM NSCLC drug therapy. Our cell viability, migration, and colony formation assays indicated that both the downregulation of miR-4270 and the upregulation of miR-423-3p increased cell proliferation and migration. The molecular functions of the putative target genes of miR-4270 and miR-423-3p (MMP19 and P21, respectively) have been well characterized [46−52]. Here, the correlations between miRNA and gene expression in the LAD patients supported the modulation of MMP-19 and P21 by miR-4270 and miR-423-3p, respectively. MMP-19 and P21 expression levels differed between BM LAD patients and NBM LAD patients. To our knowledge, this is the first study to demonstrate that MMP19 and P21 are involved in the BM pathogenesis in LAD patients. However, with the exception of p21 and MMP19, little in known about the proteins associated with the signaling mechanisms that mediate BM. In future work, we therefore aim to investigate the relationship between other signaling pathways and brain metastasis. In summary, our miRNA array screening identified several miRNAs associated with BM in LAD. Downregulation of miR-4270 or upregulation of miR-423-3p significantly increased cell proliferation, colony formation, and cell migration in vitro. miR-4270 and miR-423-3p increased the risk of BM in mouse models by targeting MMP19 and P21, respectively. Our results suggested that miR-4270 and miR-423-3p are potential diagnostic markers and drug targets, which may improve predictions of BM risk as well as the clinical treatment of LAD patients. We enrolled 155 LAD patients in this study. In all cases, LAD was histologically-confirmed between 2003 and 2008 at the Cancer Hospital, Chinese Academy of Medical Sciences (Beijing, China). Of these patients, 62 developed BM. The remaining 93 were classified as NBM. All patients underwent surgical resection followed by adjuvant therapy, according to the standard of care. Formalin-fixed paraffin-embedded (FFPE) specimens were collected from each patient. Two pathologists independently evaluated the histologic tumor type, tumor grade, and tumor percentage using hematoxylin and eosin (H&E) -stained specimens. The clinical characteristics of the patients are summarized in Table 1. Written informed consent was obtained from each patient for the use of their biological materials. This study was approved by the Institutional Review Board of the Cancer Hospital of the Chinese Academy of Medical Sciences. Human LAD cell lines A549, NCI-H1299, NCI-H157, ANIP-973, and GLC-82 were obtained from the Cell Culture Center of Peking Union Medical College (Beijing, China) and the Typical Culture Cell Bank of the Chinese Academy of Sciences (Shanghai, China). DNA typing of the H157 cells indicated that this cell line was an exact match to the cell line in the cell bank. The NCI-H157 cells corresponded to the CRL-5802 cells in the DSMZ database; multiple alleles were not detected in this cell line. The STRs in the H157 cells were compared with the STRs in all cell lines from the ATCC, DSMZ, JCRB, and RIKEN databases. However, the H157 cells could not be matched with any cell line in any databank. Human embryonic kidney (HEK) 293T cells were purchased from the ATCC (Manassas, VA). Human LAD cell lines were cultured in RPMI-1640 medium, and HEK 293T cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS; Gibco-BRL, Grand Island, NY). Both sets of cells were maintained in a humidified atmosphere with 5% CO2 at 37°C. All endogenous mature miRNA mimics and inhibitors were purchased from RiboBio (Guangzhou, China). miRNA mimics, control RNAs, inhibitors, control inhibitors, and plasmids were transfected using Lipofectamine 2000 (Invitrogen), following the manufacturer’s protocols. Total RNA was extracted from cells and tissues using TRIzol reagent (Invitrogen). First-strand cDNA was synthesized by using MMLV reverse transcriptase (Promega), following the manufacturer’s instructions. PCRs were performed using Taq polymerase (Takara) with the specific primers for MMP-19 (forward: 5’-CTTCAGCAGCTACCCCAAAC-3’; reverse: 5’- CCGTACCTGAGGGAGTGGTA-3’; P21 (forward: 5’- TTAGCAGCGGAACAAGGAGT-3’; reverse: 5’-GCCGAGAGAAAACAGTCCAG-3’;), and GAPDH (as the internal control; forward: 5’-TCTCTG CTCCTCCTGTTC-3’; reverse: 5’-GGTTGAGCACAGGGTACTTTATTGA-3’;). miRNAs were detected using stem-loop primers purchased from Ribobio (Guangzhou, China), following the manufacturer’s instructions. GAPDH and U6 small nucleolar RNAs were used for normalization. qRT-PCRs were performed with the QuantiTect SYBR Green PCR Kit (Takara) on a StepOne Real-Time PCR System (Applied Biosystems). Relative expression levels were calculated by using the 2−ΔΔCt method in Biorad CFX Manager v3. 1. The plasmids pDonR223-MMP-19 and pDonR223-P21, carrying human MMP-19 and P21, respectively, were purchased from Axybio Bio-Tech Co. , Ltd. (Changsha, China). The complete coding sequence of human MMP-19 (EMBL accession no. BC050368;) and P21 (EMBL accession no. BC001935) were amplified from the pDonR223-MMP-19 and pDonR223-P21 plasmids, respectively. The PCR products of MMP-19 and P21 were digested separately with the pEGFP-N1 plasmid using Xho I and Hind III. Fragments were purified and ligated with T4 DNA ligase. The ligated products were transformed into TOP10 competent cells, and positive clones were named either pEGFP-N1-MMP-19 or pEGFP-N1-P21. Total RNAs were analyzed with μParaflo MicroRNA microarray assays (LC Sciences). The array probes were designed based on miRBase v10. 0 (www. mirbase. org). All procedures were performed as stipulated on the LC Sciences website (www. lc-bio. com). RNA samples were also sent to CapitalBio Corp. miRNA microarray experiments were performed using GeneChip miRNA 1. 0. All procedures were performed as stipulated on the CapitalBio website (www. capitalbio. com). Sequences were clustered using Cluster 3. 0, and visualized with TreeView. Bioinformatics analysis was performed with PicTar (pictar. mdc-berlin. de]), miRanda (www. microrna. org), and TargetScan (www. targetscan. org). Fold change and P-value were calculated for each miRNA. For the MMP-19 and P21 3′-UTR luciferase assays, we constructed dual-luciferase vectors (pmiR-RB-REPORT dual-luciferase vector) containing either wild-type or mutant miR-4270 or miR-423-3p binding sites in the 3′-UTRs of MMP-19 or P21, respectively. Mutant binding sites were constructed by substituting four nucleotides in the seed region. NCI-H157 cells were co-transfected with the luciferase reporter gene constructs and miR-4270, and NCI-H1299 cells were co-transfected with the luciferase reporter gene constructs and miR-423-3p. Both cells lines were also co-transected with the luciferase reporter gene constructs and negative control oligonucleotides (miR-NC). After 48 h, cells were lysed and luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega). A plasmid containing an expression cassette for Renilla luciferase, pRenilla, was co-transfected and used to normalize the firefly luciferase values expressed by the luciferase reporter gene constructs. Cell proliferation was evaluated using 3- (4,5-dimethylthiazol-2-yl) -5- (3-carboxymethoxyphenyl) -2- (4-sulfophenyl) -2H-tetazolium, following the manufacturer’s instructions. Briefly, an optimal density (5 × 103 cells/mL) of cells in 200 μL of culture medium was added to each well in 96-well culture plates. After 0–96 h of culture, 20 μL of MTS solution was added to each well, and plates were incubated at 37°C for 2 h. The optical density of each sample was immediately measured at 570 nm using a microplate reader (BioRad). We seeded 1×103 cells were seeded into 100-mm culture plates, and incubated the plates at 37°C under 5% CO2 for 2 weeks. Plates were cultured in duplicate. The colonies which total numbers of individuals are more than 50 were counted. After washing with pre-cooled PBS, cultures were fixed with pre-cooled methanol for 20 min, and then stained with crystal violet for 15 min. The invasion assay was performed using 24-well Millicell hanging cell culture inserts, consisting of an 8-μm PET membrane (Millipore) coated with BD Matrigel Basement Membrane Matrix. The invasion assay was performed following. Briefly, cells were trypsinized and resuspended in 1640 medium without FBS. The, 5 × 104 cells were added to the upper chamber of each well, while 500 μL of 1640 medium containing 20% FBS was added to the lower chamber. After incubation for 12 h at 37°C, cells on the upper membrane surface were removed by careful wiping with a cotton swab, and the filters were fixed by treatment with methanol for 20 min. Filters were then stained with a 0. 1% crystal violet solution for 20 min. Invasive cells adhering to the undersurface of the filter were counted (using five high-power fields) with an inverted microscope. The migration assay was identical to the invasion assay, except that no Matrigel was used. Cells were harvested and protein extracts were obtained by lysing the cells with lysis buffer [1% NP-40,250 mm NaCl, 50 mM Tris-HCl, 10 mM EDTA, and 1 mM DTT, supplemented with a complete protease inhibitor tablet (Sigma, Inc) ]. Equal amounts of protein were electrophoresed on 10% SDS–PAGE gels and then transferred to PVDF membranes. After blocking with 5% bovine serum albumin (BSA), membranes were probed with anti-MMP-19, anti-P21 (Abcam), and anti-β-actin (Santa Cruz Biotechnology, Inc.), followed by incubation with a horseradish peroxidase-conjugated secondary antibody [goat-anti-mouse IgG (1: 2,000) and goat-anti-rabbit IgG (1: 3,000) ]. Proteins were visualized with Image Reader LAS 4000 (Fujifilm) and analyzed with Multi Gauge v3. 2. Recombinant lentiviral vectors expressing low levels of miRNA-4270 (miRNA-4270-knockdown), high levels of miRNA-423-3p (miRNA-423-3p-overexpression), or an irrelevant sequence were purchased from Hanbio Biotechnology (Shanghai, China). In addition to the lentiviral expression vectors, we used luciferase and puromycin reporter genes, driven by the EF1α promoter, to indicate infection efficiency. To construct lentiviral vectors, the precursor sequence for miRNA-4270 (pre-mir4270), sponge miRNA-423-3p, and the irrelevant sequence (negative control) were inserted into pHBLV-U6-MCS-EF1α-Luc-T2A-puromycin lentiviral vectors. The recombinant lentiviruses were packaged via the co-transfection of HEK-293T cells with pSPAX2 and pMD2. G using LipoFiter reagent. Supernatants with lentiviral particles were harvested 48 and 72 h after transfection and filtered through 0. 45-μm cellulose acetate filters (Millipore, USA). Recombinant lentiviruses were concentrated with ultracentrifugation. To establish stable cell lines, ANIP-973 and NCI-H1299 cells were transducted with lentiviruses (MOI of ~5) in the presence of 5 μg/mL polybrene. Each supernatant was removed after 24 h and replaced with fresh complete culture medium. Infection efficiency was validated with qRT-PCR 96 h after infection. Cells were incubated with 2 μg/mL puromycin for 2 weeks. All animals used in these experiments were treated humanely, in compliance with the “Guide for the Care and Use of Laboratory Animals”, the Institute of Laboratory Animal Resources, National Institutes of Health, and according to the Animal Experiment Guidelines of Samsung Biomedical Research Institute. The effects of miR-4270 and miR-423-3p on the tumorigenic and metastatic potentials of LAD cells were analyzed in orthotopic and systemic metastasis in vivo models. To construct the orthotopic model, 4–6-week-old BALB/c nude mice were subcutaneously injected in the right hip with 1×106 transfected cells. To construct the experimental metastasis in vivo model, transfected cancer cells (1×106 cells in 100 μL of HBSS) were directly injected into the tail vein. Six weeks later, the tumor colonies in the subcutaneous tissues were observed using H&E staining and histological examination. Bioluminescence images were collected to assess the growth and metastasis of the implanted tumor cells. To quantify the in vivo bioluminescence signals, mice were anesthetized with isoflurane before in vivo imaging, and D-luciferin solution (in vivo imaging solutions, PerkinElmer; 150 mg/kg in PBS) was intravenously injected into both the orthotopic and systemic xenograft mouse models. Bioluminescence images were acquired with the IVIS Spectrum imaging system (PerkinElmer) 2–5 min after injection. Captured images were quantified using the Living Image software package (Perkin Elmer/Caliper Life Sciences), by measuring the photon flux (photons/s/cm2/steradian) within a region of interest (ROI) drawn around the bioluminescence signal. Immunohistochemical staining was performed on 4 μm-thick slices following the two-step EnVision procedure described for the Dako REAL EnVision Detection System. Slides were incubated with one of three primary antibodies: Ki-67 (1: 50, Santa Cruz), MMP-19 (1: 50, Abcam) and P21 (1: 50, Abcam). After incubation, sliders were re-incubated with the HRP-labeled secondary antibody, and then visualized with diaminobenzidine. The expression levels of MMP-19 in the cytoplasm and P21 in the cytoplasm or nucleus were calculated as the average percentage of positive cells times the intensity of the positive cells under five randomly selected high-power fields. Scores were assigned to the obtained percentages as follows: <5% (0), 5–25% (1), 25–50% (2), 50–75% (3), and >75% (4). For intensity, scores were assigned as follows: no staining (0), light brown (1), brown (2), and dark brown (3). For MMP-19 and P21, scores of 0 and ≥1 were defined as negative and positive, respectively. All measurement data were expressed as the means ± SD, and all error bars represent the standard deviation of the mean. Student’s t tests, χ2 tests, and repeated measures ANOVAs were used to determine statistical significance. The cumulative incidence of BM was estimated with the Kaplan-Meier method, and the differences between the two groups were analyzed with the logrank test. Cox regressions (proportional hazards model) were used for multivariate analysis of prognostic factors. Receiver operating characteristic (ROC) curves and the area under the ROC curve were used to assess the accuracy of the miRNA-based BM predictions. All statistical tests were two-sided. P < 0. 05 was considered statistically significant. Statistical analyses were performed with SPSS v16. 0 (SPSS Inc. USA).

Brain metastasis (BM) is a major complication of lung carcinoma. Here, we aimed to identify the key miRNAs involved in BM lung cancer. We first profiled miRNA expression in 32 tissues from NSCLC patients with BM and 55 tissues from NSCLC patients without BM. We independently validated our results in 68 additional tissues from NSCLC patients. Based on our results, we identified a panel of miRNAs that distinguish BM lung adenocarcinomas from non-BM We report here for the first time that either miR-4270 downregulation or miR-423-3p upregulation significantly increased cell proliferation, colony formation, and migration in vitro. miR-4270 and miR-423-3p increased the risk of BM in mouse models by targeting MMP19 and P21, respectively. Our results suggested that miR-4270 and miR-423-3p might be useful markers of BM in lung adenocarcinoma.

Abstract Introduction Results Discussion Materials and methods

transfection medicine and health sciences luciferase diagnostic radiology natural antisense transcripts gene regulation enzymes cancers and neoplasms enzymology basic cancer research in vivo imaging oncology neurological tumors micrornas molecular biology techniques research and analysis methods brain metastasis imaging techniques proteins lung and intrathoracic tumors gene expression oxidoreductases bone imaging molecular biology biochemistry rna radiology and imaging metastasis diagnostic medicine nucleic acids neurology genetics biology and life sciences non-small cell lung cancer non-coding rna

2019

Molecular predictors of brain metastasis-related microRNAs in lung adenocarcinoma

Tissue-specific gene expression plays a fundamental role in metazoan biology and is an important aspect of many complex diseases. Nevertheless, an organism-wide map of tissue-specific expression remains elusive due to difficulty in obtaining these data experimentally. Here, we leveraged existing whole-animal Caenorhabditis elegans microarray data representing diverse conditions and developmental stages to generate accurate predictions of tissue-specific gene expression and experimentally validated these predictions. These patterns of tissue-specific expression are more accurate than existing high-throughput experimental studies for nearly all tissues; they also complement existing experiments by addressing tissue-specific expression present at particular developmental stages and in small tissues. We used these predictions to address several experimentally challenging questions, including the identification of tissue-specific transcriptional motifs and the discovery of potential miRNA regulation specific to particular tissues. We also investigate the role of tissue context in gene function through tissue-specific functional interaction networks. To our knowledge, this is the first study producing high-accuracy predictions of tissue-specific expression and interactions for a metazoan organism based on whole-animal data. Tissue-specific gene expression is a fundamental aspect of multicellular biology, underlying the development, function, and maintenance of diverse cell types within an organism. Accounting for tissue-specific expression is a precursor to any systems-level understanding of metazoan organismal development and function and large-scale studies of spatio-temporal gene expression both at the single-gene and whole-genome level have been performed in several organisms [1]–[5]. Additionally, tissue specificity is an important aspect of many complex diseases; notable examples of tissue interactions associated with disease include stroma-tumor interactions in cancer [6] and tissue-specific effects of insulin signaling in diabetes [7]. Although several experimental techniques have been developed to identify tissue-specific gene expression signatures, both at the single-gene and whole-genome level, our current knowledge of tissue-specific expression is incomplete. The model organism Caenorhabditis elegans provides a good framework for the study of tissue-specific expression. Its invariant cell lineage allows single-cell resolution of tissue-specific expression patterns through a variety of experimental techniques [5], [8]. In situ hybridizations of the entire transcriptome are in progress [9], and GFP-promoter tagging has been applied on a large scale [8], [10], [11]; as a result, the expression of approximately 3500 genes has been studied at the single-gene level [12], providing a “gold standard” for gene expression. Additionally, several methods have been developed to isolate mRNA samples enriched for a specific tissue or cell type, allowing global analysis using microarrays or SAGE [13]–[22]. Despite the variety of techniques available and the number of studies performed thus far, our understanding of tissue-specific expression in C. elegans is not yet complete; most genes have not been analyzed at the single-gene level, nor under diverse conditions and developmental stages. Additionally, each of the individual techniques for measuring tissue-specific expression suffers from drawbacks. GFP-promoter constructs, though they present the most accurate method amenable to high-throughput analysis, may incompletely capture endogenous expression or may fail to express well, a problem that is particularly severe in the germ line due to silencing [23]. Directed microarray studies, while powerful, depend on the ability to isolate mRNA from a particular tissue, since dissection is not possible in most cases, and methods to achieve this each have disadvantages: studies using mutants may report non-endogenous expression; embryonic cell sorting misses expression that only occurs in later stages of development, as post-embryonic cell sorting is not yet feasible; and poly-A binding studies depend on the ability to introduce the binding protein construct into and extract the protein out of the tissue of interest [21]. Thus, the ability to directly study the expression specificity of each gene across tissues, especially small tissues, and ideally to also account for the effects of development and environmental conditions, remains challenging. Here we present a computational method that leverages existing experimental information to expand and improve our knowledge of tissue-specific expression. Using data from whole-animal microarrays, we accurately predict tissue-specific expression in all major tissues and even for several tissues that comprise only a few cells. Our approach not only outperforms directed high-throughput studies in all but one case, but also captures information that complements existing experiments, for example, by uncovering tissue-specific expression that is only seen under specific conditions. To confirm our predictions, we experimentally verified the expression of several genes. We have made our predictions available through a dynamic web-based interface at http: //function. princeton. edu/worm_tissue to enable hypothesis generation and further experimental follow up by the community. Using this accurate large-scale, tissue-specific information, we perform further computational analyses, such as prediction of transcriptional regulatory motifs specific to understudied tissues as well as tissue-specific miRNA target regulation. In addition, we extended our algorithm to produce tissue-specific functional interaction networks that provide a framework for discovering protein function specific to particular tissues. Our ability to uncover tissue-specific information should allow higher-detail analysis of expression and further hypothesis testing to identify expression changes that are important for biological function. We compiled a large compendium of C. elegans microarray data (comprised of 916 experiments from 53 datasets). A few (16) of these microarray studies address tissue-specific expression, but most studies examined changes in gene expression in the animal as a whole (see supplementary website at http: //function. princeton. edu/worm_tissue for a list of microarray experiments used). Using a rank-based statistic, we evaluated the level of under- or over-expression of genes associated with each tissue in a given microarray experiment against a “gold standard” of 2872 genes known to be expressed in a particular tissue. Our gold standard is composed of information derived from single gene studies such as promoter-GFP tagging, antibody staining, and in situ hybridizations (WormBase), which we hand curated to account for tissue naming synonyms. The gold standard also includes the 1872 promoter –GFP fusions from the C. elegans Tissue Expression Consortium [10], [11], [24]. Importantly, the gold standard is completely independent from the microarray or SAGE gene expression data in our compendium. This gold standard of tissue-specific gene expression allowed us to identify substantial tissue bias in the transcriptional responses of microarray experiments. We quantified over or under-expression of tissue-specific gene sets using a rank-based statistic (Figure 1A). Despite the fact that only a small number of studies isolated specific tissues, we found that tissue-specific signals can be observed in many whole-animal experiments. For example, analysis of two developmental time courses [24] revealed dramatic tissue-specific temporal patterns that reflect developmental timing; as might be expected because neurons are born in early larval stages, earlier developmental stages are enriched for neuronal transcripts, while later stages are enriched for germ line transcripts, correlating with the development of reproductive tissues and the onset of reproduction (Figure 1B). We can also quantify a number of previously uncharacterized tissue-specific responses. For example, motor and sensory neurons have distinct developmental profiles (Figure 1B) and, in contrast to other non-reproductive somatic tissues, intestinal expression steadily increases with developmental stage. Tissue-specific responses can also be observed when experimental treatments are applied to animals in the same developmental stage. For example, our analysis of tissue-specific signals in a whole-animal microarray study of unfolded protein response [25] revealed that various mutations in UPR pathway genes have different effects on tissue-specific expression (Figure 1C). Consistent with previous studies [26], [27], we observed that an ire-1 mutation has a strong effect on epithelial tissues such as the intestine and the excretory cell, as genes expressed in those tissues are significantly down-regulated as a result of ire-1 mutation. On the other hand, an atf-6 mutation causes a decrease in neuronal transcripts suggesting greater reliance on the atf-6 branch of the UPR in neurons. Distinct tissue-specific profiles can be observed for other treatments as well. Thus, our analysis demonstrated that we can identify both known and novel tissue-specific expression information from existing gene expression microarray experiments. The previous examples suggest that substantial information about tissue-specific expression can be gained by a directed analysis of whole-animal microarray data. As such, we applied a state-of-the-art machine learning algorithm, support vector machines (SVM) [28], to build a predictive model of tissue-specific microarray profiles. Intuitively, SVM automatically identifies expression patterns in our compendium whose combination maximally separates genes expressed in a particular tissue (e. g. , neurons) from other (e. g. , non-neuronal) genes. This classifier can locate hidden tissue-specific expression patterns that are scattered through only a few experiments in the compendium and might come from diverse types of studies. By contrast, clustering methods (e. g. standard hierarchical clustering [29] or the C. elegans TopoMap [30]), while clearly important for functional data exploration, cannot detect these signals at resolution sufficient for prediction of tissue-specific expression (see Table S1 for comparison between correlation and SVM). Using the SVM classifier to predict tissue-specific gene expression based on the microarray compendium, we achieved a high degree of accuracy, outperforming directed microarray-based studies of tissue-specific expression in most cases. Our evaluation is based on the standard cross-validation technique, where only a fraction of the genes with known expression is used for building the classifier while the rest is held out for evaluation. Our predictions reach a precision of 90% for all of the major tissues of the worm (intestine, hypodermis, muscle, neurons, and pharynx) except germ line (Figure 2A). It is likely that germ-line performance is substantially underestimated, since the expression of many of the genes in the gold standard was investigated using promoter-GFP fusions, which are often germ-line silenced [23]. We also evaluated performance of other tissue-enriched gene lists acquired from directed microarray experiments against the same gold standard, using the processed lists from each publication. In all but one case, our approach outperforms these studies, predicting more genes at higher accuracy (Figure 2A). The single exception is the neuronal gene list published by Von Stetina et al. [22] that correctly recalls 384 of our gold standard neuronal genes with 89% accuracy, somewhat above what we are able to predict at the same recall. Our method accurately predicts tissue-specific gene expression even from whole-animal microarray data alone. When we exclude the 16 studies that directly address tissue-specific expression, our prediction accuracy remains high; in some cases it is even unchanged (Figure 2A, “tissue data excluded”). In particular, even for the intestine, for which there are a number of high-quality directed studies [17], [18], our prediction accuracy is not decreased when we use only whole-animal data. Functional analysis of the top tissue-specific predictions (GO enrichment analysis) demonstrates that many of the genes we predict to express in specific tissues have functions consistent with that tissue. For example, predictions for germ line expression were enriched for cell cycle-related GO terms, those for muscle included “muscle contraction” and “respiration”, intestine included terms related to digestion and metabolism such as “fatty acid biosynthetic process”, neuron predictions were associated with “synaptic transmission” and “memory”, and hypodermis-expressed predictions included enrichment for terms related to molting and cuticle components. The pharynx is a complex organ that is comprised of muscle, structural and gland cells and genes predicted to express in the pharynx are enriched for diverse functions related to cytoskeleton, cuticle components, and secretion. (See supplementary website for all GO enrichment results.) While techniques for isolating tissue-specific mRNA are steadily improving, it remains a particular challenge to examine the expression of genes in smaller tissues. Therefore, it is of particular interest to be able to predict expression in tissues that are comprised of only a few cells. Using our approach, we were able make high-quality predictions for many tissues where biochemical methods have yet not been successfully applied. While we do not achieve the high level of precision we observe in major tissues (which is expected, as far fewer genes are reported to express in the smaller tissues, making new candidates significantly more difficult to identify), we were able to identify genes that are significantly enriched for expression in the small tissue of interest when compared to the genomic background (Figure 2B). For example, among the genes in our gold standard, only 1 in 10 express in the vulva. However, we were able to correctly recall 30% of all vulval genes with a precision of 20% percent, a two-fold improvement above the genomic background rate, and likely an under-estimate as our GFP-based gold standard is far more incomplete for these small tissues than for larger tissues. Among the genes that scored highly in vulval predictions is dgn-1, a homolog of human dystroglycan. The gene was not a top prediction for any major tissue except pharynx, suggesting that it is not widely expressed. Among small tissues, dgn-1 was predicted to express in the uterus, distal tip cells, and the excretory cell in addition to the vulva. While dgn-1 was not included in our gold standard, expression in these tissues, including expression in pharyngeal epithelia, has been confirmed recently [31]. Additionally, this gene has been shown to be functionally important for the development of the vulva and the excretory cell [31], [32], in contrast to its vertebrate homolog, which functions in muscle. Among other small tissues, we were also able to make reasonable predictions for the excretory cell, the spermatheca, the uterus, ceolomocytes, and distal tip cells. In many cases the predicted genes have annotations that are consistent with the function of the tissue. For example, our distal tip cell predictions are enriched for many GO terms including “cell migration, ” “protein localization, ” and several “cellular component” terms associated with exocytosis. These GO associations appear reasonable, as distal tip cells are two highly polarized cells that lead gonad migration during development. Secretion from these cells is known to play an active role in gonad migration [33], and the cells' morphology (as visualized by EM) is indicative of active endo/exocytosis [34]. In addition, the top 200 distal-tip cell predictions significantly (p<10−2, hyper-geometric test) overlap with the list of genes associated with distal tip cell migration phenotypes compiled in a recent RNAi study [35]. Thus, our results demonstrate that even small tissues that are challenging to isolate experimentally have distinct expression profiles within whole-animal microarray data. Our ability to make such predictions will likely improve as new gene expression experiments are added to the compendium. We experimentally verified tissue-specific expression of six top genes with previously unreported tissue-specific predictions by creating transgenic lines carrying promoter-GFP constructs (Figure 3). Three of these genes were predicted to express in hypodermis. We chose to focus on hypodermis since, to our knowledge, no large-scale study investigating hypodermal expression has been reported. Promoter-GFP constructs of two of the predicted hypodermal genes, K08B12. 1 and F58H1. 2, were most prominently expressed in the hypodermis at earlier stages (Figure 3A and 3B and Figure S4). The third gene, F55H12. 4, showed strongest hypodermal expression during L4 and adult stages (Figure 3C). We also verified the expression of genes that we predicted to be expressed in muscle (C29F5. 1, Figure 3D), intestine (F13D12. 6, Figure 3E), and neurons (gnrr-1, Figure 3F). The tissue specific expression of gnrr-1, a homolog of the human gonadotropin releasing receptor, was previously studied using antibody staining [36]. While our algorithm predicted with high confidence that gnrr-1 expresses in neurons, neuronal expression was not reported in that study, and the gene was not included in the Von Stetina et al. list of neuronally-enriched genes [22]. Nevertheless, our promoter-GFP (Pgnnr-1: : gfp) construct expressed primarily in head neurons and ventral cord neurons (Figure 3D), validating our prediction. It is likely that the protein product of gnrr-1 is heavily post-translationally modified, as species of multiple molecular weights are observed [36]. Thus, it is possible that differences in such modification explain the discrepancy between our gene expression results and the previous antibody staining experiment, due to epitope differences. Furthermore, gnrr-1 is strongly over-expressed in L1 and L2 larval stages in multiple developmental microarray time courses, which is the pattern observed for many neuronal genes Our ability to make high-quality predictions also provided potential insights regarding the transcriptional regulation associated with the tissue-specific expression signal in whole animal data. We used a motif-finding program, FIRE [37], to identify motifs that are overrepresented in the upstream regions of our top-scoring predictions for each of the major tissues (Figure 4). While no genome-wide study of hypodermal expression has been published thus far, we were able to use our predictions to uncover motifs that are promising candidates for regulators of hypodermal transcription. A GATA-like motif was enriched among our top hypodermal predictions. This is consistent with previous studies showing that GATA transcription factors are essential for hypodermal cell specification, and that a GATA consensus sequence is required for hypodermal expression [38], [39]. In addition, we have identified a motif that is similar to the binding site for the CF1/USP-like nuclear hormone receptor that affects molting and developmental transitions in insects [40]. An intriguing possibility is that this motif and a functional USP homolog are involved in the nematode molting process as well, despite the fact that no direct USP homologs have been detected in the genome [41]. Using our germ-line predictions, we recovered an E2F-like motif (ETF). The C. elegans homolog of mammalian E2F, efl-1, is expressed exclusively in the germ line and is involved in oogenesis, regulating the expression of genes whose promoters contain the E2F binding motif [42]. Another motif, TAC. GTA, was also strongly represented among germ-line predictions. We could not detect a clear match to any known transcription factor consensus sequence, but a similar motif was previously discovered in a C. elegans-C. briggsae sequence comparison [43]. A GATA-like motif was also overrepresented among intestine predictions. GATA transcription factors are known to regulate expression of intestinal genes [44], and this motif is very similar to those reported by previous whole genome intestinal expression studies [17], [18] and aging studies [45], [46]. Our pharynx prediction yielded the largest number of motifs of any tissue. One of the motifs represents a possible match to the pha-4 consensus ([T[AG]TT[TG][AG][TC] [15]) though other motifs did not resemble any known binding sites (see Table S2 for a complete list of motifs). Surprisingly, there was a shortage of neuronally-overrepresented motifs. In fact, the most significant result for neurons was instead motif avoidance. This is consistent with the hypothesis, supported by many experimental observations in C. elegans (see for example [47], [48]) that neuronal differentiation is a “ground state” that is superseded in non-neuronal cells. The identification of global tissue-specific expression patterns allows us to address biological questions that are difficult to address experimentally, such as the question of tissue bias in microRNA targets. Non-coding microRNAs have emerged as critical developmental regulators, and are predicted to regulate the expression of a large fraction of all mammalian genes [49], [50]. Specific miRNAs direct development in particular tissues [51], [52], yet experimental identification of miRNA targets in individual tissues remains difficult. This is in part because expression of miRNA targets may be unchanged if translational inhibition, as opposed to mRNA degradation, is involved. Moreover, the ability to identify all targets for all miRNAs simultaneously is still more challenging. Previous studies using human data have detected cell type-specific signatures among miRNA targets [53]. To address this problem in C. elegans, we leveraged our predictions of tissue-specific expression to investigate tissue bias, as measured by a rank-based statistic, among a list of likely C. elegans miRNA targets predicted by Miranda [54], TargetScan [49], and PicTar [55]. While many miRNAs had no detectable tissue bias among their targets, a subset showed significant tissue preference or tissue avoidance (see Figure S3 for all microRNAs-tissue interactions). In particular, robust tissue avoidance for three microRNAs was detected in all three sets of target predictions (Figure 5). The miR-124 mammalian homolog is known to induce neuronal differentiation [52]. Our analysis demonstrates that its predicted targets are depleted for neuronal genes, while enriched for genes specific to other somatic tissues; these results suggest that its function is conserved in C. elegans. miR-2 showed a pattern of neuronal depletion similar to mir-124' s pattern, implying that it is also involved in neuronal differentiation; this is consistent with the exclusively neuronal pattern of GFP expressed from the miR-2 promoter [56]. The mir-71 target set, on the other hand, is significantly depleted for intestinal genes but enriched for genes expressed in muscle, hypodermis and pharynx. In contrast to miR-2, the anatomical expression of miR-71 appears to be ubiquitous, suggesting that tissue-centric target analysis provides complementary information that is not captured by expression studies. We have been able to leverage diverse microarray data to predict tissue-specific expression, including for genes expressed in more than one tissue. However, many genes that are expressed in several tissues (or ubiquitously) perform different functions in different cellular contexts. A natural way to explore such functional roles is through functional interaction networks, which connect genes that participate in the same biological process, an approach that has been used by us and others to examine functional roles of proteins on whole-genome scale [32], [57]. In contrast to previous approaches, in the case of tissue-specific functional networks, a network for a given set of genes may vary depending on the tissue of interest, as the same set of gene products may not perform the same function or share the same physical or other interactions in different tissues. We have developed an SVM-based algorithm to predict tissue-specific functional networks from our compendium of C. elegans transcriptional data. Although simple expression correlation has often been used to investigate gene function on a global (non tissue-specific) level, our analysis above (and in Figure S1) demonstrates that a single global correlation computation is unable to distinguish between tissue-specific effects. On the other hand, the observation that whole-animal microarrays may contain a strong tissue specific signal suggests that it is possible to assess the tissue-dependent functional roles of genes given the right analytic approach. Thus, we have developed a network generation algorithm in which certain experiments are trusted more or less depending on the extent to which they reflect a particular tissue-specific functional signal. Similarly to previous network integrations, we define a gold standard of functional interactions that is then used to determine how data is combined into a network. However, in contrast to previous studies [32], [58], we define several tissue-specific gold standards, one for each tissue, and we use an SVM rather than a Bayesian formulation to combine microarray data. An advantage of the SVM for this problem is that SVMs have the ability to adjust weights of individual experiments while Bayesian integration typically assigns weights to whole datasets. In the case of the C. elegans compendium, the ability to treat each experiment individually is crucial for prediction of tissue-specific networks, as a single dataset can contain experiments that are informative for different tissues. For example, within a single developmental time course (see Figure 1B), early larval stages are informative of neurons, when neuronal cells are overrepresented, while the adult stage is highly informative of germ-line. Using an SVM-based approach, we are able to integrate microarray data into different tissue-specific functional interaction networks. Such networks link genes that are likely to participate in the same process within a specific tissue context and contain information that may otherwise be overwhelmed in a global view of co-expression. As an example, we considered exc-7, an RNA-binding protein that is involved in the formation of the excretory canal, but that also plays a role in neuronal development, affecting cholinergic synaptic transmission [59]. Several of the interaction partners present in its neuron-specific interaction network support our understanding of exc-7 neuronal function (Figure 6A): hmr-1 is required for the outgrowth of some motor neurons [60]; unc-38 is an acetylcholine receptor [12]; and the mammalian homolog of abl-1 is involved in post-synaptic acetylcholine receptor clustering [61]. Another partner, rhgf-1, a RhoGEF, is known to regulate neurotransmitter release at the neuromuscular junction [62]. Our network results also suggest an interaction between exc-7 and smg-1, a key component of the nonsense-mediated mRNA decay pathway, and spk-1, which is involved in mRNA splicing [12]. The presence of RNA processing genes among the interaction partners is potentially related to exc-7' s RNA-binding function. A standard correlation computation produces an entirely different, non-neuron-specific set of genes associated with exc-7, including aquaporins and a gene involved in excretory cell formation (Figure S2). Our technique, on the other hand, automatically identifies a subset of microarray experiments with strong neuronal signals, and thus we are able to uncover neuron-specific functional interactions that are not immediately visible in a global correlation network. Apart from finding tissue-specific interactions that can be lost in a global view, as in the above example, tissue-specific networks have the potential to tease out how the same gene may perform different functions within different tissue contexts. The C. elegans homolog of Ras, let-60, is a canonical example of a ubiquitously-expressed gene that participates in diverse processes. For example, let-60 promotes progression through meiosis during oogenesis [63] and affects olfaction in neurons [64]. To explore these tissue-specific functions of this gene, we queried our germ line and neuronal networks with let-60. Two of the genes in the neuronal network are involved in chemosensation (Figure 6B): rgs-3 is a regulator of G-protein signaling required for normal response to a variety of sensory stimuli [65], and ckk-1 is a CaM kinase kinase that regulates the expression of chemosensory receptor genes [66]. Other neighbors in the network are involved in further aspects of neuronal function: zaf-1, syd-9, and sad-1 function in synapse development, and egl-19 is a calcium channel that contributes to fate specification in olfactory neurons [67]. By contrast, the germ line let-60 network is comprised of an entirely different set of genes that are consistent with let-60' s function in meiosis: cej-1 (cpg-1) is required for proper meiotic chromosome segregation [68], and zyg-11 is part of a ubiquitin-ligase complex that promotes meiotic anaphase II [69]. Other interactors are likely to participate in related processes: zen-4 is a kinesin protein that localizes to midzone microtubules [70]; kbp-1 localizes to kinetochores [71]; and both rfc-4 and pos-1 affect a large number of events in the oocyte to embryo transitions [72]. Our networks focus on interaction information within a tissue-specific context, providing a framework for generating precise hypotheses about tissue-specific gene functions that can help direct follow-up experiments. We have developed a computational method that accurately predicts tissue-specific expression based on expression profiles of primarily whole-animal microarrays. We show that strong tissue biases can be observed in data from microarray experiments, despite the fact that most C. elegans microarray experiments isolate mRNA from the whole animal, with the resulting expression values representing a population average of many cell types. With our SVM classifier, we were able to leverage these signals in existing whole-animal microarrays to produce predictions of tissue-specific gene expression and generate networks of tissue-specific functional interactions. In addition to achieving accuracy higher than most directed microarray studies, our algorithm captures information about tissue-specific expression that is complementary to standard approaches. Microarray experiments analyzing tissue-specific expression are able to discover tissue-specific genes based on the difference in mRNA levels, a method that is ultimately sensitive to total mRNA abundance. Our method instead relies on co-expression with known tissue-specific genes in some informative condition, and thus identifies tissue-specific expression even for genes that have very low levels of expression in any one experiment. As we analyze microarray experiments from a variety of conditions, our approach can uncover genes expressed in a particular tissue in a condition-dependent manner which may be difficult to directly detect experimentally. For example, a promoter-GFP tagging study reported expression of ins-7 exclusively in neurons [73], while our method predicts expression in both neuron and intestine. In fact, a recent study has shown that ins-7 is indeed expressed in the intestine at a low level, with expression increasing significantly in aging animals and under conditions of high insulin signaling [74]. The earlier GFP study focused on young wild-type adults and thus did not identify this age-related expression. Thus, our method provides a valuable tool for study of tissue-specific expression that is relatively unbiased, as it does not rely on mRNA abundance directly and can leverage existing whole-animal compendia that provide a variety of developmental stages and conditions represented in these collections. From a more general perspective, our method extracts tissue-specific expression and interaction information from large compendia of diverse microarray studies. Even in the case of larger animals where it may be feasible to perform microarray studies on dissected tissues, the underlying samples are nevertheless typically comprised of multiple cell types; a method to predict gene expression in tissue subtypes will be applicable to other organisms, limited only by the existence of an appropriate “gold standard” gene expression set. Our results demonstrate that sample heterogeneity, when appropriately analyzed, can provide valuable information regarding cell-type specific gene expression and function. Tissue localization data was retrieved from WormBase 170 [12] and parsed in a semi-automated way. Since a variety of terms are used to describe the same tissue and/or organ, we hand-compiled a table of tissue synonyms. In addition we applied some hierarchical propagation to tissue labels, such as assigning specific neurons to their neuron class (sensory, motor, interneurons). A majority of genes were reported to express in multiple tissues and each gene was considered a positive example for all tissue where it was found to express. This data includes all 1,872 genes investigated by the C. elegans Tissue Expression Consortium [8], [24] as well as expression patterns from smaller scale experiments, for a total of 2872 genes in the gold standard. These data did not include any large-scale expression studies (microarray or SAGE), and was limited to single-gene GFP or in situ experiments. We collected microarray data from 53 publications (see Supplementary website for complete list). The microarray values from a single publication were considered a coherent dataset and processed together. Data for single-channel platforms was transformed by dividing every gene value by its average over the dataset and taking the log of the result. All missing values were imputed using the KNN impute algorithm [75] (k = 10). For input to SVM learning the gene values within a single dataset were normalized to mean 0 and variance 1 before all datasets were concatenated. Since the SVM algorithm does not accommodate missing values, genes that were present in some datasets but not others were assigned a value of 0 when absent. For each tissue we used our gold standard to assign genes with known expression into 2 classes (tissue expressed and not tissue expressed). We the used the two classes and the microarray expression values to calculate an AUC score and the associated probability. The probabilities were used to correct the results for multiple hypothesis testing at a false discovery rate of 0. 05. Single gene predictions were made using linear support vector machines (SVM). Given a set of genes known to be expressed in a particular tissue, the SVM identifies specific patterns of gene expression in a subset of experiments that differentiates these genes from those not expressed in the tissue. We performed 5-fold cross validation and optimized the parameters for maximal precision at 30% recall (fraction of genes in the gold standard correctly recalled) for major tissues and 10% recall for small tissues. SVMs are a maximal margin classifier that optimizes classification performance on the training set while maximizing model generalizing power by maximizing the distance of the nearest correctly classified examples to the separating plane. If and define the plane that separates the positive and negative examples, are the vectors of microarray data, are the training labels, and denote the degree of misclassification for each example, the SVM problem is to minimizesubject to. The constant is empirically optimized to achieve the best performance at classifying new examples. Genes were selected based on the following criteria: top prediction scores that are specific to a single tissue, no previously reported tissue-specific localization to that tissue, and absence of any tissue-bias that could be inferred from sequence information alone. In particular, we avoided all collagen-related genes predicted to express in hypodermis due to ease of prediction of this particular tissue-specific expression from sequence. In addition, we specifically selected gnrr-1 because of the discrepancy between our predictions (made with a top prediction score) and previously published results ([36]). Based on the above criteria we picked 14 genes, for which we obtained 9 lines; 6 of these fluoresced and these 6 are all shown in Figure 3. The GFP-promoter constructs were made using the Gateway system with the unc-119 rescue plasmid pDestDD03 and promoter clones from the C. elegans promoterome [10]. The resulting constructs were bombarded into unc-119 (ed3) mutants. Motif discovery was performed for each tissue separately. For a single tissue, all the genes that were present in our microarray compendium were assigned a cluster number of 1 if they were in the top 500 predicted genes and a cluster number of 0 otherwise. This cluster assignment was used as input to the FIRE algorithm. Kmer length was set to 9 and default values were used for all other parameters. To generate the tissue-specific interaction standard we first generated a global functional interaction standard using a combination of GO, KEGG, and Textpresso-curated interactions [12], [76]. We then defined a set of tissue-specific interactions by cross-referencing with our gold standard of tissue expression used for single gene expression prediction. A tissue-specific interaction was defined as a pair of genes that were co-annotated to a specific GO term (see Supplementary methods) and were also both found to express in a particular tissue in our expression gold standard. The negative set was composed of positive interactions from other major tissues as well as random pairs of GO annotated genes. The classification problem is then to differentiate interactions specific to a particular tissue from interactions in other tissues as well as non-interacting gene pairs. The algorithm computes a weighted sum of single experiment similarity measures. Since the expression values are normalized to have mean 0 and variance 1, single experiment similarity measures are thus single terms within a per-dataset Pearson correlation. The contribution of expression data to the final value is thuswhere and represent the expression values of genes and in experiment and is the weight assigned to that experiment by the SVM classifier. (See Text S1 for a detailed description).

In animals, a crucial facet of any gene' s function is the tissue or cell type in which that gene is expressed and the proteins that it interacts with in that cell. However, genome-wide identification of expression across the multitude of tissues of varying size and complexity is difficult to achieve experimentally. In this paper, we show that microararray data collected from whole animals can be analyzed to yield high-quality predictions of tissue-specific expression. These predictions are of better or comparable accuracy to tissue-specific expression determined from high-throughput experiments. Our results provide a global view of tissue-specific expression in Caenorhabditis elegans, allowing us to address the question of how expression patterns are regulated and to analyze how the functions of genes that are expressed in several tissues are influenced by the cellular context.

Abstract Introduction Results Discussion Methods

computational biology/sequence motif analysis genetics and genomics/functional genomics genetics and genomics/gene expression computational biology/transcriptional regulation developmental biology/aging

2009

Global Prediction of Tissue-Specific Gene Expression and Context-Dependent Gene Networks in Caenorhabditis elegans

Dengue disease is currently a major health problem in Indonesia and affects all provinces in the country, including Semarang Municipality, Central Java province. While dengue is endemic in this region, only limited data on the disease epidemiology is available. To understand the dynamics of dengue in Semarang, we conducted clinical, virological, and demographical surveillance of dengue in Semarang and its surrounding regions in 2012. Dengue cases were detected in both urban and rural areas located in various geographical features, including the coastal and highland areas. During an eight months' study, a total of 120 febrile patients were recruited, of which 66 were serologically confirmed for dengue infection using IgG/IgM ELISA and/or NS1 tests. The cases occurred both in dry and wet seasons. Majority of patients were under 10 years old. Most patients were diagnosed as dengue hemorrhagic fever, followed by dengue shock syndrome and dengue fever. Serotyping was performed in 31 patients, and we observed the co-circulation of all four dengue virus (DENV) serotypes. When the serotypes were correlated with the severity of the disease, no direct correlation was observed. Phylogenetic analysis of DENV based on Envelope gene sequence revealed the circulation of DENV-2 Cosmopolitan genotype and DENV-3 Genotype I. A striking finding was observed for DENV-1, in which we found the co-circulation of Genotype I with an old Genotype II. The Genotype II was represented by a virus strain that has a very slow mutation rate and is very closely related to the DENV strain from Thailand, isolated in 1964 and never reported in other countries in the last three decades. Moreover, this virus was discovered in a cool highland area with an elevation of 1,001 meters above the sea level. The discovery of this old DENV strain may suggest the silent circulation of old virus strains in Indonesia. Dengue is one of the most important arthropod-borne viral diseases with large global burden. The disease is caused by dengue virus (DENV), a member of Flaviviridae family, with four distinct serotypes (DENV-1, -2, -3, and -4) circulating in tropical and subtropical regions in the world. DENV is transmitted to human by Aedes mosquitoes as vector [1]. Dengue clinical manifestations vary from asymptomatic or mild flu-like syndrome known as classic Dengue Fever (DF) to more severe form known as Dengue Hemorrhagic Fever (DHF) and the potentially fatal Dengue Shock Syndrome (DSS) [2]. DENV genome consists of ∼10. 7 kb single-stranded positive-sense RNA genome encoding 3 structural (C, prM/M, E) and 7 non-structural (NS1, NS2A, NS2B, NS3, NS4A, NS4B, NS5) proteins [3]. Similar to other RNA viruses, DENV possess diverse genetic characteristics as shown by the presence of various genotypes within serotypes [4]. Dengue was first reported in Indonesia in 1968 in Jakarta and Surabaya [5]. Up to now, dengue afflicts all the 33 provinces of the vast Indonesian archipelago [6] and become a public health problem annually while periodic major outbreaks occurred such as those in 1998 [7] and 2004 [8]. Nearly 60% of Indonesian people reside in Java island where most of them living in urban areas of big cities where dengue is a problem. However, it has been reported that the disease has also influenced people living in rural areas which probably due to intense people movement [6]. Semarang municipality is a region located in Central Java that is routinely affected by the disease. The region contributes 1. 15% of Central Java province with 373. 7 km2 of areas, divided into coastal and inland areas with various topographical features. The city was inhabited by more than 1. 5 million residents. Semarang is listed as top 5 of population number in Central Java with population density of 4,133 per km2. In the year of 2011, Semarang region has reported 1,303 dengue cases with 10 fatalities (Profil Kesehatan Kota Semarang 2011). Despite of annual dengue incidence in the region, no detail data of the epidemiology of the disease is present. To fully understand the dengue disease in Semarang municipality, we performed comprehensive dengue surveillance study in Semarang regions, including the Semarang district and Salatiga City, investigating the clinical, virological, and demographical aspects of the disease. The clinical and demographical data of patients were recorded, and the geographical distribution was studied by monitoring the dengue cases both in urban/coastal and rural/highland areas. DENVs were isolated from patients' sera, their serotypes were determined, and their genetic aspects were studied by using phylogenetic and comparative genomic analyses. Correlation between clinical manifestation and virological aspects was also studied. Ethical clearance for this study was obtained from the Dr. Kariadi hospital and Diponegoro University Medical Research Ethics Committees. Dengue-suspected patients from hospitals and primary health care centers were invited to participate in the study and enrolled after written informed consents were obtained from all participants. For minors/children participants, written informed consents were sought from their parent/legal guardians. The study was performed in three regions around Semarang municipality, the capital city of Central Java province, Indonesia during December 2011 until July 2012. The study sites were encompassing both coastal and urban area (Semarang municipality) and rural/highland areas (Semarang district and Salatiga City) with altitude ranged from 0–1,500 meters above sea level (masl). Patients recruitments were conducted at Dr. Kariadi hospital, Dr. Adhyatma hospital, Semarang City hospital, Roemani Muhammadiyah Semarang hospital, Ungaran hospital, Ambarawa hospital, Salatiga City hospital, and primary health care centers (Puskesmas), namely Sumowono and Kedungmundu. The geographic coordinates of each patient were recorded using handheld GPS Garmin 72H and mapped (Figure 1). Sera from dengue-suspected patients were subjected to serology tests. Serological analysis was performed in sample collection sites using Panbio Dengue Duo Cassette (Alere, Brisbane, Australia) to detect anti-dengue IgG and IgM. Serology test results were confirmed using Panbio Dengue Duo ELISA (Alere) in a laboratory setting, which was also used to determine the primary/secondary infection status of patients. Detection of DENV NS1 antigen was done using Panbio Dengue Early Rapid kit (Alere). Clinical symptoms were documented and laboratory tests including thrombocyte count and hematocrit level were performed as a routine procedure in the hospital. Dengue classification was based on WHO/TDR 2009 guidelines [9]. RT-PCR confirmations were performed to detect the presence of DENV in 66 NS1 and/or IgM-positive samples. Virus RNA was extracted from serum samples using QiaAmp Viral RNA Mini kit (Qiagen, Hilden, Germany) according to manufacturer' s instructions. DENV nucleic acid detection and serotyping were performed using two steps conventional RT-PCR according to protocol previously described by Lanciotti, et al. [10]. Detection and serotyping results were confirmed by quantitative real-time RT-PCR using Simplexa Dengue Molecular Assay performed in 3M Integrated Cycler machine (Focus Diagnostic, Cypress, CA). Detail method of the Simplexa Dengue Molecular Assay was as described by the manufacturer. To determine the positivity of the samples, the cycle threshold (Ct) cut off value of 38 (instead of 40) was used to ensure the strict detection and serotyping of the DENV. Serum samples with serological or RT-PCR-positive result were subjected to a maximum of two passages of inoculation in C6/36 (Aedes albopictus, mid gut) cell line [11]. Briefly, monolayer of cells was inoculated with 200 µl of sera in 2 ml of 1× RPMI medium supplemented with 2% of Fetal Bovine Serum (FBS), 2 mM of l-glutamine, 100 U/ml of Penicillin, and 100 µg/ml of Streptomycin (all from Gibco-Life Technologies, Carlsbad, CA). Flasks were incubated for 1 hour at 28°C to allow virus attachment. Following the incubation period, inoculation medium was discarded and the medium was replenished with 3 ml of fresh medium. Infected cells were incubated at 28°C for up to 14 days. DENV genotyping was performed using Envelope (E) gene sequence (1,485 nt in length). DENV RNA was reverse-transcribed into cDNA using Superscript III reverse transcriptase (RT) (Invitrogen-Life Technologies) and DENV-specific primers described elsewhere [12]. The resulting cDNA was then used as template for PCR amplification using Pfu Turbo Polymerase (Stratagene-Agilent Technologies, La Jolla, CA). PCR products were purified from 0. 8% agarose gel using QIAquick gel extraction kit (Qiagen) and used in cycle sequencing reaction performed using 6 overlapping primers [12] from both strands and BigDye Dideoxy Terminator sequencing kits v3. 1 (Applied Biosystems-Life Technologies), following manufacturer' s instructions. Purified DNA was subjected to capillary sequencing performed on 3130xl Genetic Analyzer (Applied Biosystems) at the Eijkman Institute sequencing facility. Primers used in genotyping were described elsewhere [12]. Sequence reads were assembled using SeqScape v. 2. 5 software (Applied Biosystems) with additional manual adjustment performed when manual inspection of the assembly showed some discrepancies. The E protein gene sequences obtained in this study have been deposited in GenBank with accession number KC589008–KC589013 (Supplementary Table S1). For genotype classification, we grouped the isolate sequences with the relevant reference sequences based on classifications by Goncalvez et al. [13], Twiddy et al. [14], and Lanciotti et al. [15] for DENV-1, -2 and -3, respectively. MrBayes was used to construct Bayesian inference phylogenetic trees with mixed model across GTR space model and gamma rates for one million generations with 4 chains, sampled every 1,000 iterations. For evolution studies, we downloaded all publicly available DNA sequences of DENV-1, -2 and -3 from NCBI GenBank as of 12 December 2012. Sample sequences were combined with the downloaded GenBank sequences according to sample' s serotypes to create dataset for each dengue serotype. Sequence clustering was performed on each dataset using USEARCH [16]. Multiple alignments resulted from sequence clustering from each cluster containing sample sequences were trimmed to obtain only the alignment representing E protein segment. Sequences without E protein segment were removed from the alignment. We built phylogenetic tree using FastTree [17] for fast approximation with GTR and gamma rate evolution model. We selected 20 closest public sequences from each isolate sequence based on the patristic distance of the FastTree' s phylogenetic trees. The multiple alignment of the selected sequences along with the sequence samples were used for phylogenetic reconstruction using Bayesian MCMC method as implemented in BEAST v 1. 7. 2 [18] using GTR+Γ4 model with codon model, relaxed uncorrelated lognormal molecular clock and Bayesian skyline prior, with 60 million generations and sampled for every 1000th iteration. Statistical analysis was performed using SPSS Statistics software version 11. 5 (SPSS Inc. , Chicago, IL). Pearson chi-square test was used to correlate the clinical manifestations and DENV serotypes. One-way ANOVA test was used to compare groups of laboratory tests results and DENV serotypes. A probability value of p<0. 05 was considered statistically significant. One hundred and twenty febrile patients were recruited during the course of the study. The dengue cases were equally distributed throughout the month of January–July 2012 with the most cases observed in May (Figure 2A). Of 120 patients, 66 (55%) patients were serologically positive for dengue as determined by IgM/IgG and/or NS1 tests. RT-PCR confirmations were performed to detect the presence of virus on those 66 serum samples and 31 (47%) samples were dengue positive. Of the 66 serologically positive dengue cases, most cases (n = 51 or 77%) were secondary infections, as determined by IgM and IgG ELISA according to kit' s manual. The male to female ratio was 1. 0 with an average age of 15. 98±12. 16 (CI 95% 12. 99–18. 98). Cases occurred predominantly in children aged below 10 year old (42%) (Figure 2B). Among serologically-positive patients, most of them are school children/student (41. 7%), followed by working adult (33. 3%) and toddler/pre-school children/unemployed adults (25%) (data not shown). Of the 31 RT-PCR-confirmed dengue cases, serotyping revealed the predominance of DENV-1 (35. 5%), followed by DENV-2 (12. 9%), DENV-3 (12. 9%), and DENV-4 (9. 7%). We also observed the presence of multiple infections of different serotypes (29%), of which DENV-1 involved in most of the multiple infections (Figure 3A). These multiple infections were confirmed by Simplexa Dengue Molecular Assay. Dengue vector distribution is influenced by the local temperature where the mosquitoes circulating. To determine whether topological aspect give account to the dengue cases in Semarang, we performed dengue surveillance encompassing areas with various topological and geographical features. We obtained dengue cases in either coastal or highland areas. Most cases were found in areas with the altitude of 0–250 masl. However, a single case was also found in area with the altitude of 1,001 masl, in which the local temperature in daytime was approximately 24–26°C (Figure 1 and 3B). All dengue-confirmed patients underwent clinical examination and laboratory tests at enrollment. The severity of their clinical manifestations was categorized based on WHO 2009 guidelines grading criteria. Of the 66 dengue-confirmed patients, most of them (n = 56 or 84%) were classified as DHF. To assess whether there is correlation between serotype and clinical manifestation, we compiled the patients' data and summarized them in Table 1. Of the 31 cases confirmed by RT-PCR detection, 23 (74%) cases were DHF, and 8 (26%) cases were DSS (Table 1). We did not find any statistically significant correlation between clinical manifestations and medical laboratory examination results of the patients with the infecting DENV serotypes. In order to study the circulating DENV genotypes in Semarang, we performed genotyping analysis based on E gene sequences. The DENV E gene sequences were aligned with reference sequences to generate genotype classifications in each serotype. The resulting phylogenetic trees for the genotype grouping are described in Supplementary Figures. Of 11 isolates that were serotyped as DENV-1, three viruses were successfully PCR-amplified for their E genes after a single passage in C6/36 cell line. Phylogenetic analysis revealed the grouping of those isolates into two different genotypes, Genotype I (SMG-SE058 and SMG-SE059) and Genotype II (SMG-SE003) based on Goncalvez classification [13] (Supplementary Figure S1). The grouping of Semarang DENV-1 into Genotype II is a new information given this genotype never been found in Indonesia previously. Meanwhile, the Genotype I of DENV-1 is quite commonly found in Indonesia [12], [19]. We further analyzed the evolutionary history and rate of the DENV-1 from Semarang. Of 20 closely-related sequences of each Semarang isolates from published sequences in GenBank, a total of 56 non-redundant, unique sequences were used for further analysis consisting of mainly isolates from Singapore, China, and imported cases in Taiwan. Because the lack of public sequences for Genotype II, the 20 closely-related sequences for the Genotype II isolate (SMG-SE003) included other genotypes such as Genotype I, III and IV. As shown in Figure 4, this particular isolate was closely related to DENV-1 isolated in Thailand in 1964 [20], which was used as a strain for vaccine development [21]. The mean evolutionary rate of DENV-1 as calculated by BEAST was 2. 72×10−4 subs/site/year [95% Highest Posterior Density/HPD: 1. 24–4. 54×10−4]. For DENV-2, we managed to genotype one isolate which was grouped into Cosmopolitan genotype according to Twiddy classification [14] (Supplementary Figure S2). This genotype is commonly found in Indonesia and surrounding countries. The collected 20 closely-related sequences of the Semarang isolate consisted of sequences from Indonesia and neighboring countries such as Singapore, Malaysia and Brunei, as well as other Asian countries such as China, Taiwan, and Vietnam. The Semarang DENV-2 isolate clustered together with Jakarta DENV-2 isolated in 2004 as indicated in Figure 5. The mean rate of evolution of E-protein in this data set was 11. 66×10−4 subs/site/year [95% HPD: 6. 04–17. 68×10−4]. We also successfully genotyped two isolates of DENV-3 from Semarang, which were grouped into Genotype I based on Lanciotti classification [15] (Supplementary Figure S3). Figure 6 showed that the majority of 20 non-redundant, unique closely-related sequences from these isolates were Indonesian isolates, either collected previously in other regions in Indonesia such as Yogyakarta (in 1988), Palembang (in 1998) [7], and Jakarta (in 2004) [12] or collected as imported cases such as from Queensland, Australia (of Bali origin in 2010) [22] and from Taiwan. This suggests the endemicity of this genotype in the area. Other isolates from Asian countries such as Singapore, Malaysia and China were also included. The mean evolutionary rate of DENV-3 was 7. 24×10−4 subs/site/year [95% HPD: 5. 08–9. 50×10−4]. In this study, we also detected three DENV-4 infections (Figure 3A), however, we were not able to PCR-amplify the E gene from the isolates to be used for genotyping. We described here the clinical, virological, and demographical features of dengue in Semarang, the capital city of Central Java province, and its surrounding regions. The study is somewhat unique as it involved the survey of those features in various regions with different topography, encompassing the coastal, urban/inner city and rural/inland areas, as well as highland areas with the elevation of more than 1,000 m above sea level. Our surveillance was conducted from December 2011 until July 2012. This duration encompasses the rainy and dry season periods, as in Indonesia the wet season is commonly occurred during October–April while the dry season occurred in April–October. The dengue cases were equally detected throughout January to July with the recorded peak in May. Data from Semarang Meteorology Bureau indicated that May to June 2012 was the transition months between wet and dry seasons (data not shown); therefore the high dengue incidence in this month is understandable. The dengue mosquito vectors might still actively breed soon after the decrease of rainfall, since heavy rainfall may wash out the breeding sites while lower rain intensity will maintain the breeding sites. A total of 120 dengue-suspected patients were recruited in this study. Of this, 66 (55%) patients were serologically confirmed for dengue infection, suggesting that dengue places a considerable burden in the community. Molecular detection revealed the presence of DENV in 31 (47%) patients' sera. Results from serotyping identified the presence of all DENV serotypes in Semarang, with DENV-1 as the predominant serotype, followed by DENV-2, -3, and -4 (Figure 3A). Using a real-time quantitative RT-PCR detection system with a strict standard for detecting the presence of DENV genomes, we observed the presence of multiple DENV infections in nine (29%) out of 31 samples, with DENV-1 involved in most of the multiple infections. This finding further supports the hyper-endemicity of the disease and the predominance of DENV-1. The DENV-1 predominance is currently in place in other cities in Indonesia including in Surabaya [19] and Makassar (Sasmono et al, 2013, submitted for publication). However, historical data of DENV serotype distribution in Indonesia reported the predominance of DENV-2 and -3 in several cities [6], [12]. As there has been a very limited data on dengue serotype distribution in Semarang, we were not able to conclude whether the current serotype replaced the previous serotype predominance. On the clinical aspect of dengue in Semarang, we documented the clinical symptoms and medical laboratory tests results and grouped the clinical manifestation according to WHO 2009 guideline. In our study, we observed the occurrence of DF, DHF and DSS in dengue-confirmed patients involved in this study. Most of the cases were manifested as DHF, followed by DSS and DF. This finding is understandable as the surveillance was conducted in either health care center or hospital. Based on serological data, most patients (77%) were secondary infection. This indicates sustained disease intensity over a number of years and the endemicity of dengue in the region. There have been studies reporting the association of DENV serotype with clinical manifestation [23]–[25], in which particular serotypes have been correlated with the severity of the disease. To understand the role of each serotype in influencing the clinical outcomes of the disease, we compared the clinical findings of 31 serotyped dengue cases against each serotype. As shown in the Table 1, we did not observe any direct correlation of particular serotype with the disease manifestations. However, we are aware that the small sample number obtained and the unequal distribution of serotypes in this study may not be an ideal basis to draw a conclusion. In term of geographical feature, this study revealed that the most dengue cases were found in urban area of Semarang City, an area with an elevation of 1–250 masl. The high density population in the area plus the hot and humid weather of the city may give account for the successful transmission of DENV through its Aedes sp mosquito vectors. Dengue cases were also reported in Salatiga, a city with a colder weather than Semarang with elevation of 750–850 masl. Probably the most striking finding was the occurrence of one dengue case in Sumowono, a village that is located in the mount Ungaran in the Semarang district. The house where the patient resides has an elevation of 1,001 masl with the average daytime temperature of 24–26°C. To further investigate this rare case (no report of dengue cases in the village in the last 5 years), we conducted vector surveillance and found Aedes larvae in used tires and outdoor water containers within the radius of 20–50 m from the house. The infected patient was diagnosed as DHF with thrombocyte count of 51,000/mm3 and presented common symptoms of dengue such as fever, headache, nausea, loss of appetite, positive tourniquet test, lethargy, and sleeplessness. The patient was fully recovered. There is possibility that the dengue infection occurred outside the village but this was negligible, as the patient, a 50 y. o. housewife, was very rarely travelling outside the village because she has been semi-paralyzed due to stroke attack. This finding indicates the virus was transmitted by local Aedes mosquito vector and thus suggests the ability of this vector to adapt and circulate in area with higher altitude and colder temperature. Previously, there was a report of dengue outbreak at an area with altitude of 1,700 masl in Guerrero State, Mexico, in 1988 [26]. Therefore the presence of this vector in high altitude area is not impossible. Nevertheless, the occurrence of the dengue cases in this highland area, to the best of our knowledge, represents the first report in Indonesia. A more detail data on the vector surveillance in the study area will be described elsewhere (Fahri et al, 2013, unpublished results). The genotype of DENV circulating in Semarang was determined by phylogenetic analysis of the E gene of the DENV. For the DENV-1, based on classification by Goncalvez [13], we observed the presence of Genotype I circulating in the region. The SMG-SE058 isolate was clustered with the Singaporean samples isolated in 2008 [27], and had TMRCA (time to most recent common ancestor) around year 1999. The SMG-SE059 isolate was clustered together with Taiwan isolate (0705aTw) in 2007 originated from Indonesia as stated imported case [28], and Korean isolate (DenKor-11) in 2008 from a traveler who visited Indonesia. The TMRCA for this clade is around year 2002. This indicates that strains from these DENV-1 Genotype I clades are likely to have been circulating in Semarang more than a decade. This genotype is currently predominant and common in Indonesia and has been reported to replace the previously predominant Genotype IV [12], [19]. In this study, we did not find the Genotype IV in Semarang area, which may be present but not sampled and genotyped. The other genotype of DENV-1 that was discovered was the Genotype II, based on Goncalvez classification [13]. This is a novel DENV-1 genotype in Indonesia, as it has never been reported before. This genotype also has never been spotted in other countries in the last three decades. This genotype was isolated from the patient resides in Sumowono village (elevation 1,001 masl) described above. From all publicly available DENV-1 sequences in GenBank, this isolate was found to be very closely related with DENV-1 strain 16007 isolated from patient in Thailand having dengue hemorrhagic fever and shock in 1964 [20], as well as a virus strain which has been undergone serial passages from the parental 16007 isolate (strain 16007 (PDK-13) ) produced during vaccine development [21]. The evolution rate of Indonesian SMG-SE003 was very slow (6. 92×10−5 mutation/site/year) compared to the general rate of DENV-1 in this analysis (2. 72×10−4 mutation/site/year), as indicated by the blue branch line in Figure 4. The high mutation rate of strain 16007 (PDK-13), indicated by bright red branch line, was most likely the effect of serial passages process. Another interesting finding related to this strain was that the dengue NS1 antigen test performed in patient' s serum was negative, which may raise question if this virus escaped from detection by NS1 diagnostic test. However, the IgM/IgG ELISA confirmed the dengue infection for this sample. Further study is needed to fully understand this finding. On the origin of this Genotype II strain in Indonesia, one may suspect a possibility of DENV contamination from lab strains during the process of culture, PCR and sequencing. However, this is very unlikely since our lab has never handled and manipulated DENV-1 strains from outside Indonesia (except DENV-1 reference strains Nauru-Western Pacific and Hawaii-USA which do not belong to this genotype). Therefore, this finding reflects the reappearance of an old and unique strain of DENV-1 in Central Java, Indonesia. Considering that SMG-SE003 isolate is closely related to a strain from Thailand isolated in 1964 and has very slow mutation rate, there is a possibility that either this strain has been actually present for a long time and is maintained in nature in a low circulation and infection rate but only sampled now, or that this strain has been dormant and recently emerged into circulation. We are not sure whether this strain will behave like the Thailand strain which caused DHF if it actively re-infecting humans. We are also not sure if it will be spreading and causing epidemic in the region, but given the presence of Aedes vector breeding sites in the proximity of isolate origin, the spreading of this strain is possible. Further surveillance is needed to monitor the activity of this strain. The genotype of DENV-2 circulating in Semarang region was the Cosmopolitan genotype. This genotype is quite common in the region and is widely circulated in India, South East Asia, Africa, the Middle East, and Australia [14]. In Indonesia, this genotype has been circulating since a long time ago and causing outbreaks in 1998 and 2004 [12]. The phylogenetic analysis of the SMG-SE001 indicated that the isolate share common ancestor with some isolates circulating in Taiwan, as imported cases from Indonesia [28], [29], and Guangzhou, China. Semarang is one of the regions in Indonesia that have been supplying workers for countries such as Taiwan; hence dengue cases found in those particular countries might be brought by those workers. Semarang is also one of the regions with high population of Chinese-descendant, with frequent direct flights between Semarang and Guangzhou, China. The genotypes of DENV-3 isolates from Semarang were clustered in genotype I, which is also common in South East Asia regions. The DENV-3 isolates from Semarang had common ancestor with the isolates from imported cases in Taiwan, which is similar situation as DENV-2 isolate. In particular, the cluster containing these strains had been circulating in Indonesia for more than a decade. The mean evolution rates of the DENV in this study were within the range observed by other studies, which were in the range of 4. 6–11. 6×10−4 subs/site/year [30]. With the exception of DENV-1 Genotype II isolate SMG-SE003, all rates of individual isolates were within the boundary of the 95% HPD of each corresponding serotypes. This is another indication that this Genotype II isolate warrants a further investigation. In this study, we were unable to obtain any DENV-4 sequences from Semarang cases. This might be attributed to low viral titers which can only be detected by real-time RT-PCR method, but not sufficient enough for conventional PCR amplification for sequencing purposes. We are aware that there are methods that utilize shorter fragment of E gene that could be used in determining the genotype of our DENV-4, and we might apply this in the future to better understand the genetic aspects of DENV-4 in Semarang. In conclusion, we have described the clinical, virological, and demographical features of dengue in Semarang in which all serotypes are circulating and highlighted the presence of an old genotype of DENV-1. We also observed the occurrence of dengue in area with high altitude. Altogether, the study suggests the importance of continuous virus surveillance in dengue endemic regions such as Indonesia to better understand the dynamic of the disease.

We studied dengue disease in Semarang municipality, Central Java, one of the endemic regions in Indonesia. The disease occurred in wide geographical regions which include urban, rural, coastal, and highland areas. All four dengue virus serotypes were found. The infecting serotypes were not associated with disease severities. We also determined the genotype of the circulating viruses. One of the interesting findings was the presence of an old genotype of DENV-1 which has never been reported in the last three decades, which may suggest the silent circulation of this particular genotype in Semarang. These findings offer the first information of the clinical, virological and demographical aspects of the dengue disease in Semarang, Indonesia.

Abstract Introduction Materials and Methods Results Discussion

medicine infectious diseases molecular epidemiology epidemiology dengue fever neglected tropical diseases dengue biology viral diseases evolutionary biology genomic evolution

2013

Molecular Surveillance of Dengue in Semarang, Indonesia Revealed the Circulation of an Old Genotype of Dengue Virus Serotype-1

We consider the problem of genetic association testing of a binary trait in a sample that contains related individuals, where we adjust for relevant covariates and allow for missing data. We propose CERAMIC, an estimating equation approach that can be viewed as a hybrid of logistic regression and linear mixed-effects model (LMM) approaches. CERAMIC extends the recently proposed CARAT method to allow samples with related individuals and to incorporate partially missing data. In simulations, we show that CERAMIC outperforms existing LMM and generalized LMM approaches, maintaining high power and correct type 1 error across a wider range of scenarios. CERAMIC results in a particularly large power increase over existing methods when the sample includes related individuals with some missing data (e. g. , when some individuals with phenotype and covariate information have missing genotype), because CERAMIC is able to make use of the relationship information to incorporate partially missing data in the analysis while correcting for dependence. Because CERAMIC is based on a retrospective analysis, it is robust to misspecification of the phenotype model, resulting in better control of type 1 error and higher power than that of prospective methods, such as GMMAT, when the phenotype model is misspecified. CERAMIC is computationally efficient for genomewide analysis in samples of related individuals of almost any configuration, including small families, unrelated individuals and even large, complex pedigrees. We apply CERAMIC to data on type 2 diabetes (T2D) from the Framingham Heart Study. In a genome scan, 9 of the 10 smallest CERAMIC p-values occur in or near either known T2D susceptibility loci or plausible candidates, verifying that CERAMIC is able to home in on the important loci in a genome scan. In genetic association analysis of a binary trait, such as presence or absence of a disease, it can be useful to properly control for relevant covariates. Inclusion of environmental risk factors in the analysis has the potential to increase statistical power by reducing phenotypic noise, and adjustment for confounding factors can provide some protection against spurious association [1,2]. For a sample of independent individuals, logistic regression provides a natural approach. However, it is common for association studies to contain some related individuals. This can arise in isolated populations, or by chance in large samples in outbred populations, or, for example, when families collected for linkage studies are included in association analysis. In the presence of related individuals, it is important to model the familial correlation in the sample appropriately to avoid inflation of the significance of association [3] and to improve power [4]. Furthermore, if missing data occur in samples with related individuals (for instance, when some individuals are phenotyped but not genotyped or vice versa), additional power can often be obtained by appropriately incorporating partial information [5]. It is of interest to develop an association mapping method for binary traits that addresses the aforementioned challenges, yet has computational time complexity feasible for current scales of genome-wide association studies (GWASs). The classical transmission disequilibrium test (TDT) [6] and the FBAT method [7] are applicable only to family-based designs. For general case-control designs with distantly-related individuals, methods [8–10] based on linear mixed models (LMM) are commonly used in recent literature. However, such methods are specifically designed for quantitative traits, and, although they can be applied to case-control data, they can suffer from power loss due to failure to capture the binary nature of the outcome. More recent work [11,12] specifically addresses the analysis of ascertained case-control samples with low levels of relatedness. To model binary traits in samples with high levels of relatedness, several authors [13–15] have proposed methods based on the generalized linear mixed model (GLMM) statistical framework, which extends the classical logistic regression model by including genetic random effects on the logit scale. Of these, only GMMAT [15] has a computational speed that is feasible for large data sets. For samples with unknown population structure, Jiang et al. [16] recently proposed CARAT, a binary-trait genome-wide association testing approach, which adjusts for relevant covariates on a logistic, instead of linear, scale, and which incorporates an additive polygenic effect within a computationally efficient quasi-likelihood framework. However, none of the above methods exploit information contained in partially missing data. MQLS [4] is a binary-trait association testing method that allows related individuals in the sample and that makes full use of the relationship information in order to incorporate partially missing data, but it does not adjust for covariates or for additive polygenic effects. MASTOR, [17] a more recent association method for samples with related individuals, is able to use information in partially missing data by applying a retrospective analysis to a LMM. Like other LMM approaches, it is designed for quantitative traits, although it can be applied to binary traits. We consider a somewhat different setting from that addressed by CARAT and GMMAT. We assume a sample that includes at least some closely-related individuals, for whom the pedigree structure is assumed known. We also want to allow for the possibility of incorporating partially missing data, which is not possible with CARAT or GMMAT. We propose CERAMIC (Case-control Efficient Related-individual Association Mapping Incorporating Covariates), a mixed-model, binary-trait association testing method, which adjusts for relevant covariates and efficiently incorporates missing data to enhance power. The mean and covariance structure of CERAMIC is tailored for binary traits, and it would therefore be expected to outperform LMM methods such as EMMAX, [8] GEMMA [10] and MASTOR. By using a quasi-likelihood framework, CERAMIC achieves computational efficiency for genome-wide analysis. When assessing significance, we modify the genotypic model used by CARAT to allow for possible correlation between the genotype and the covariates, which is of particular relevance when, for example, ancestry-informative covariates are included to control for population stratification. As a further development over CARAT and a further advantage over prospective LMM and GLMM methods (e. g. GMMAT, EMMAX, GEMMA), CERAMIC effectively exploits partially missing data to improve power by incorporating data on individuals with missing genotypes who have a genotyped relative. Our method can also be considered a generalization of MQLS to allow adjustment for covariates and additive polygenic effects. Unlike the family-based, covariate-adjusted TDT test, CERAMIC can be applied to completely general samples of related and unrelated individuals, provided the genealogy is known. Major features of CERAMIC are summarized as follows: (1) it is applicable to essentially arbitrary combinations of related and unrelated individuals, including small outbred pedigrees and unrelated individuals, as well as large, complex inbred pedigrees; (2) it incorporates information on individuals with partially missing data while correctly accounting for dependence; (3) it corrects binary phenotypes for both covariates and additive polygenic effects using a model that exploits the binary nature of the trait; and (4) it is computationally feasible for current association studies. For comparison, we have developed two other retrospective tests that also have features (1), (2) and (4) above, but which do not model additive polygenic effects: MQLS-LOG, which uses a logistic approach to adjust binary phenotypes only for covariates, not for polygenic effects, and MQLS-LIN, which uses an ordinary least-squares regression approach to do the same. We conduct simulation studies to assess the type 1 error and power of our methods, CERAMIC, MQLS-LOG and MQLS-LIN, and to compare them to previously reported methods, MASTOR, EMMAX, GEMMA, CARAT and GMMAT. Finally, we apply CERAMIC to the analysis of type 2 diabetes (T2D) data from the Framingham Heart Study (FHS). CARAT [16] was developed for binary trait association mapping in samples that are possibly subject to unknown population structure, assuming that genome-wide data are available. CARAT is based on a quasi-likelihood model in which only the conditional mean and variance of Y given genotype and covariate information are specified. The conditional mean is given by E (Y i | X, G) = μ i, with log μ i 1 - μ i = X i β + G i γ, i = 1, ⋯, n, (1) where β is a k-dimensional vector of unknown covariate effects, and γ is the unknown scalar association parameter. The conditional variance is given by Ω: = Var (Y | X, G) = Γ 1 / 2 Σ Γ 1 / 2, with Σ = ξ Φ + (1 - ξ) I, (2) where Γ is an n × n diagonal matrix with ith diagonal element equal to μi (1 − μi), I is an n × n identity matrix, Φ is a genetic relationship matrix, and 0 ≤ ξ ≤ 1 is an unknown scalar parameter measuring the relative importance of additive polygenic vs. i. i. d. error variance in explaining trait variability. The appearance of the terms Γ1/2 in the formula for Ω in Eq (2) ensures that, for an outbred individual i, the relationship between the conditional mean and variance given by Var (Yi|X, G) = μi (1 − μi) holds, as is necessary for a binary random variable. In CARAT, Φ is taken to be an empirical genetic relationship matrix calculated from genome-wide data. The unknown parameter for CARAT is (γ, β, ξ), where, for association analysis, γ is the main parameter of interest, while (β, ξ) is typically considered a nuisance parameter. To detect association between the phenotype and the SNP, H0: γ = 0 is tested against H1: γ ≠ 0. To form the CARAT test statistic, we first obtain the null estimate, (0, β ^ 0, ξ ^ 0) of the parameter (γ, β, ξ) by iteratively solving a system of estimating equations under the constraint γ = 0. The quasi-score estimating equation for β is given by X T Γ 1 / 2 Σ - 1 Γ - 1 / 2 (Y - μ) = 0, (3) where μ = μ (γ, β) = (μ1, ⋯, μn) T. The estimating equation for ξ is given by (Y - μ) T Γ - 1 / 2 Σ - 1 (Φ - I) Σ - 1 Γ - 1 / 2 (Y - μ) = trace (Σ - 1 (Φ - I) ). (4) Then (β ^ 0, ξ ^ 0) is defined to be the solution to the system consisting of Eqs (3) and (4), with γ constrained to be 0. The CARAT test is based on the quasi-score statistic, U 0 ≔ G T Γ ^ 0 1 / 2 Σ ^ 0 - 1 Γ ^ 0 - 1 / 2 (Y - μ ^ 0), (5) where μ ^ 0, Σ ^ 0 and Γ ^ 0 are μ, Σ and Γ evaluated at (γ, β, ξ) = (0, β ^ 0, ξ ^ 0). (For additional details on quasi-score tests and their use in statistical genetics see references [18] and [19]). To assess the significance of the association test, CARAT takes a retrospective approach, in which it is assumed that, under the null hypothesis of no association, E 0 (G | X, Y) = 2 p 1 n and Var 0 (G | X, Y) = σ g 2 Φ, (6) where 0 ≤ p ≤ 1 is the unknown allele frequency of the variant of interest, σ g 2 > 0 is an unknown parameter, and the subscript “0” indicates that expectation and variance are taken under the null hypothesis. CARAT estimates σ g 2 using σ ˜ g 2 = 2 p ^ (1 − p ^), where p ^ = 0. 5 × G ‾ is the sample average estimator of p. The CARAT test statistic can then be defined by CARAT ≔ (Z T G) 2 σ ˜ g 2 · Z T Φ Z, where Z = Γ ^ 0 1 / 2 Σ ^ 0 - 1 Γ ^ 0 - 1 / 2 (Y - μ ^ 0). (7) Significance of association is assessed by comparing the test statistic to a χ 1 2 random variable. In addition to testing for association, one can also obtain an estimate of the full parameter (γ, β, ξ) by solving a system consisting of Eqs (3) and (4) and the following quasi-score estimating equation for γ: G T Γ 1 / 2 Σ - 1 Γ - 1 / 2 (Y - μ) = 0. (8) The system consisting of Eqs (3), (4) and (8) can be solved iteratively to obtain the estimated parameter vector. Consider a sample of n individuals who are arbitrarily related, with the pedigree information assumed to be known. (Note that individuals who are unrelated to anyone else in the sample are also allowed.) Let the kinship matrix derived from the pedigree information be given by Φ = 1 + h 1 2 ϕ 1,2 … 2 ϕ 1, n 2 ϕ 2,1 1 + h 2 … 2 ϕ 2, n ⋮ … … ⋮ 2 ϕ n, 1 2 ϕ n, 2 … 1 + h n, (9) where ϕi, j is the kinship coefficient between individuals i and j, and hi is the inbreeding coefficient of individual i. To model the binary trait variable Y, we adopt a quasi-likelihood framework similar to that of CARAT, in which the mean structure is given by Eq (1). In the conditional variance of Y given X and G, shown in Eq (2), we could use either the empirical genetic relationship matrix as in CARAT or the pedigree-based kinship matrix defined in Eq (9) as the genetic relationship matrix, Φ. As we will show in the next subsection, we use the pedigree-based kinship matrix as part of our approach to extract additional information for association from partially missing genotype data on related individuals, so it is convenient (although not necessary) to also use the pedigree-based kinship matrix for Φ in Eq (2). This would also eliminate the need for genome-wide data. In the complete data case (i. e. , when X, G, and Y are fully observed), to detect association between the binary phenotype and the genetic variant, we first obtain the quasi-score statistic defined as in Eq (5), but with the empirical genetic relationship matrix replaced by the kinship matrix calculated from the pedigree. Three commonly-used approaches for assessment of significance of an association test in genetic analysis can be described as (1) prospective, in which the conditional distribution of the phenotype given genotype and covariates is considered, (2) retrospective, in which the conditional distribution of the genotype given phenotype and covariates is considered, and (3) permutation-based. For CERAMIC, we take a retrospective approach similar to that of CARAT. One advantage of the retrospective approach is that it is robust to misspecification of the phenotype model, because correct type 1 error of CERAMIC relies only on the null conditional mean and variance of the vector of genotypes. Another advantage is that it allows for a natural way to incorporate information on individuals with missing genotypes, as described in the next subsection. For the retrospective analysis, we make the following modeling assumptions about the distribution of G conditional on Y and X under the null hypothesis: E 0 (G | Y, X) = X α, and (10) Var 0 (G | Y, X) = σ G 2 Φ, (11) where α is an unknown k-dimensional vector of coefficients, σ G 2 > 0 is an unknown parameter, and Φ is the known kinship matrix defined in Eq (9). The null mean assumption in Eq (10) says that, under the null hypothesis of no association between genotype and phenotype, the genotype is permitted to be linearly related to the covariates, or it can be unrelated to the covariates. The possibility that G could be linearly related to the covariates is particularly relevant, for example, when ancestry vectors are used as covariates to account for population structure (e. g. M. P. Conomos, A. P. Reiner, M. S. McPeek and T. A. Thornton, under review). Then the null mean assumption allows, e. g. , for different sub-populations to have different allele frequencies. The null variance assumption in Eq (11) is a version of the standard variance relationship that holds, for example, under Mendelian inheritance in a single population. Here, however, we do not require σ G 2 = 2 p (1 − p), where p is the allele frequency at the variant of interest, which would hold under Hardy-Weinberg equilibrium. Instead we use a more robust variance estimator [17] given by σ ^ G 2 = 1 n - k G T P G, (12) where P = Φ−1 − Φ−1 X (XT Φ−1 X) −1 XT Φ−1. Note that σ ^ G 2 is equivalent to the residual mean square error for the generalized linear regression of G on X, with covariance matrix proportional to Φ. The CERAMIC test statistic in the complete data case is then given by CERAMIC c = U 0 2 Var ^ 0 (U 0 | Y, X) = (Z T G) 2 σ ^ G 2 Z T Φ Z, (13) where Z is defined in Eq (7) and is referred to as the vector of transformed null phenotypic residuals. The subscript “c” on CERAMICc stands for “complete data. ” Under suitable regularity conditions, significance of association could then be assessed by comparing CERAMICc to a χ 1 2 random variable. In addition to testing for association, one can also obtain an estimate of the full parameter (γ, β, ξ) by iteratively solving the system consisting of Eqs (3), (4) and (8). Let (γ ^, β ^, ξ ^) denote the estimator obtained as the solution of this system, and let Γ ^ and Σ ^ denote Γ and Σ, respectively, evaluated at (γ ^, β ^, ξ ^). Then the estimated asymptotic covariance matrix for (γ ^, β ^ T) T is given by Cov ^ ( (γ ^, β ^ T) T) = (X ˜ T Γ ^ 1 / 2 Σ ^ - 1 Γ ^ 1 / 2 X ˜) - 1, (14) where we define X ˜ = (G, X). The approximate standard errors for the elements of (γ ^, β ^ T) are obtained as the square roots of the corresponding diagonal elements of the matrix in Eq (14). We note that the validity of Eq (14) as an estimator of the covariance relies on the validity of the modeling assumptions of Eqs (1) and (2) and on the sample size, n, being large. Calculation of the standard error for the estimator, ξ ^, of the variance parameter would require higher-order (i. e. , third and fourth) moment assumptions on Y, so it is not available in our approach, in which we need only specify the mean and variance structure of the phenotype. First, we describe our notation with regard to missing data. Let N denote the full set of n sampled individuals. For a given genetic variant to be tested, let R ⊂ N denote the subset of individuals with non-missing genotype at that variant, and let r = |R| denote the number of such individuals. We define GR to be the r × 1 sub-vector of G that contains the genotypes for the individuals in set R, and we define ΦR to be the r × r submatrix of Φ consisting of the rows and columns corresponding to the individuals in set R. We can partition the set R into two disjoint subsets, U and V, where U denotes the subset of individuals with non-missing genotype at the tested variant who also have complete phenotype and covariate data, and V denotes the subset of individuals with non-missing genotype at the tested variant who are missing either the phenotype or one or more covariates. We let u = |U| and v = |V|, and we have R = U ∪ V, U ∩ V = ∅ and r = u + v. We let S denote the set consisting of the remaining s = n − r individuals not in the set R, i. e. , S = N ∩ Rc denotes the subset of individuals who have missing genotype data at the variant of interest. There are certain categories of individuals who do not make a contribution to our association analysis, and we assume that these have already been deleted from the set S (and from N). To be retained in S, individuals with missing genotype are required to have non-missing phenotype and covariate information, and, in addition, to satisfy at least one of the following two conditions: (1) the individual has a genotyped relative; or (2) the individual is in the same pedigree with an individual with non-missing phenotype and covariates who either has non-missing genotype or has a relative with non-missing genotype at the tested variant. Let W = U ∪ S, so W is the set of w = u + s individuals remaining in N who have complete phenotype and covariate information. Notice that the sets N, R, U, V, S, and W can, in principle, vary across tested variants that have different patterns of genotypic missingness. This point is discussed in more detail in the subsection Some computational considerations for CERAMIC. To form the CERAMIC test statistic in the case of partially missing data, we propose to use the genotype data for the individuals in the set R combined with the phenotype and covariate data for the individuals in the set W. As in the case of complete data, we first obtain an estimator of the phenotypic nuisance parameter, (β, ξ), under the null hypothesis, H0: γ = 0. This estimator, which we call (β ^ W 0, ξ ^ W 0), is obtained by solving Eqs (3) and (4) with γ set to 0, where all vectors and matrices in Eqs (3) and (4) are restricted to contain only those individuals in set W. We then let ZW denote the vector of transformed null phenotypic residuals for the set W, where Z W = Γ ^ W 1 / 2 Σ ^ W - 1 Γ ^ W - 1 / 2 (Y W - μ ^ W), (15) where Γ ^ W, Σ ^ W, and μ ^ W are Γ, Σ, and μ, respectively, restricted to the individuals in set W and evaluated at (γ, β, ξ) = (0, β ^ W 0, ξ ^ W 0), and YW is the vector Y restricted to the individuals in set W. We define the CERAMIC statistic with partially missing data to be CERAMIC = (F T G R) 2 Var ^ 0 (F T G R | Y, X) = (F T G R) 2 σ ̌ G 2 F T Φ R F = (Z W T Φ R W T M G R) 2 σ ̌ G 2 Z W T Φ R W T M Φ R M Φ R W Z W, (16) where F = MΦRW ZW is a vector of length r that incorporates phenotype, covariate, and pedigree information, in which M = Φ R − 1 − Φ R − 1 1 r (1 r T Φ R − 1 1 r) − 1 1 r T Φ R − 1, where 1r is a vector of length r with every element equal to 1, and ΦRW denotes the submatrix of Φ with rows corresponding to the individuals in R and columns corresponding to the individuals in W, i. e. , the (i, j) th element of ΦRW is 2ϕij, where ϕij is the kinship coefficient between the ith individual in set R and the jth individual in set W. The variance estimator [17], σ ̌ G 2, is just the variance estimator, σ ^ G 2, of Eq (15) restricted to the set, Q, of individuals in N who have non-missing genotypes and covariates but may or may not have observed phenotypes (U ⊂ Q ⊂ R), q = |Q|, i. e. σ ̌ G 2 = 1 q - k G Q T P Q G Q, (17) where P G = Φ Q − 1 − Φ Q − 1 X Q (X Q T Φ Q − 1 X Q) − 1 X Q T Φ Q − 1, and GQ, XQ, and ΦQ are G, X, and Φ, respectively, restricted to the individuals in set Q. With complete data, i. e. , N = R = U = W and S = V = ∅, the CERAMIC statistic of Eq (16) reduces to the CERAMICc statistic of Eq (13) (see S1 Text for details). Under assumptions described in the next subsection, the CERAMIC statistic follows an asymptotic χ 1 2 distribution under the null hypothesis. One possible interpretation of the CERAMIC statistic in Eq (16) is that it uses best linear unbiased prediction to impute missing genotypes based on relatives’ genotypes, while downweighting predictions with low information level, correcting for imputation error and correcting for extra correlation due to imputation. This can be seen [5] by rewriting Eq (16) in terms of the best linear unbiased predictor (BLUP) of the missing genotypes for the individuals in set S. More generally, we can let G ^ W = 1 w p ^ + Φ W R Φ R - 1 (G R - 1 r p ^) = [ 1 w (1 r Φ R - 1 1 r) - 1 1 r T Φ R - 1 + Φ W R M ] G R (18) denote the BLUP of GW, where GW is the vector G restricted to the individuals in set W, 1w is a vector of length w with every element equal to 1, and p ^ = (1 r Φ R − 1 1 r) − 1 1 r T Φ R − 1 G R is the best linear unbiased estimator [20] of the allele frequency, p, of the variant of interest. Consider G ^ W i, the ith element of G ^ W. If individual i is in set U, then G ^ W i can be shown [5] to be the observed genotype of individual i, while if individual i is in set S, then G ^ W i is the BLUP of the unobserved genotype of individual i. In other words, if we reorder the individuals in set W so that the individuals in set U come first in the list and the individuals in set S follow, then we can write G ^ W = G U 1 s p ^ + Φ S R Φ R - 1 (G R - 1 r p ^) = G U G ^ S, (19) where GU is the vector G restricted to the individuals in set U, and G ^ S is the BLUP for the missing genotypes of the individuals in set S. The CERAMIC statistic can then be rewritten (see S2 Text) in terms of the BLUP imputed genotypes, as CERAMIC = (Z W T G ^ W) 2 Var ^ 0 (Z W T G ^ W | Y, X). (20) With retrospective modeling in which the conditional variance is assessed with respect to genotypes, the additional uncertainty and dependence due to genotype imputation is directly accounted for. Alternatively, CERAMIC can be interpreted as a quasi-score test derived from a retrospective mean model [4,17], though we do not detail this interpretation here. To obtain the asymptotic null distribution for CERAMIC, we slightly modify the null mean assumption in Eq (10) and assume E 0 (G ^ W | Y, X) = X W α, where XW is the matrix X restricted to the individuals in set W and α is a k-dimensional vector of unknown coefficients. Then CERAMIC follows an asymptotic χ 1 2 distribution under the null hypothesis under regularity conditions [21]. The accuracy, in finite samples, of the χ 1 2 approximation to the null distribution is assessed in Results. In order to carry out the iterative solution of the system of estimating equations given by Eqs (3) and (4) (or by Eqs (3), (4) and (8) when γ is to be estimated as well), we need to obtain the inverse of the n × n matrix Σ = ξΦ + (1 − ξ) I for different values of ξ. We reduce the computational burden in two ways: (1) Σ is inverted block-wise where each block of Σ corresponds to a pedigree; and (2) a single spectral decomposition, Φ = AJAT (where A is an orthogonal matrix, and J is diagonal), is used to compute the inverse of Σ for different values of ξ, because Σ−1 = A (ξJ + (1 − ξ) I) −1AT, where ξJ + (1 − ξ) I is a diagonal matrix [22]. The calculation of the transformed null phenotypic residual vector, ZW, depends on the set W, which is a function of the genotypic missingness pattern for the genetic variant being tested. In a GWAS, different SNPs often have different genotypic missingness patterns, so this could imply that in the worst case scenario, ZW would need to be computed separately for each SNP in the genome. One possible way to avoid this would be to compute ZW only once per genome scan based on all individuals with non-missing phenotype and covariate information (or based on some other fixed subset of individuals) and use the same ZW for association testing with respect to all SNPs across the genome. Another approach, which is the one we actually take, would be to estimate the variance component (VC) parameter, ξ, once per genome using all individuals with non-missing phenotype and covariate information, and then for each SNP, compute only the regression parameter, β, by solving Eq (3), with γ = 0. In this way, we need only solve Eq (3) to obtain ZW separately for each SNP, which drastically reduces the computational burden. In addition to CERAMIC, we also developed two other alternative approaches to association testing for binary traits in related individuals. These two methods, which we call MQLS-LOG and MQLS-LIN, both have the following features: (1) they incorporate covariates; (2) they are generalizations of the MQLS test [4]; (3) they are retrospective and handle missing data in the same way that CERAMIC does; and (4) they do not involve estimation of an additive polygenic component of variance. The main difference between them is that MQLS-LOG has a logistic mean structure while MQLS-LIN has a linear mean structure. In the remainder of this subsection, we give the details of these two tests, and in the Results section, we compare them, in terms of type 1 error and power, to CERAMIC and to three previously proposed tests, MASTOR, EMMAX, and GLOGS. The MQLS-LOG and MQLS-LIN test statistics can each be constructed from the CERAMIC test statistic of Eqs (16) and (20) by replacing the transformed null phenotypic residual vector, ZW, by some other type of residual vector that is a function of (XW, YW). To obtain the MQLS-LOG test statistic from the CERAMIC test statistic, we replace ZW by the vector of residuals from the logistic regression model, where this model is given by Y i | X W ∼ Bernoulli (p i), independently, for i ∈ W, with log p i 1 - p i = X i T β. (21) Let β ˜ be the maximum likelihood estimator for β under the model of Eq (21), and let ϵ be the resulting null phenotypic residual vector, defined to have ith element ϵ i = Y i − p ^ i, for i ∈ W, where p ^ i is given by log p ^ i 1 − p ^ i = X i T β ˜. Then the MQLS-LOG test statistic is given by MQLS-LOG = (ϵ T G ^ W) 2 Var ^ 0 (ϵ T G ^ W | Y, X) = (ϵ T Φ R W T M G R) 2 σ ̌ G 2 ϵ T Φ R W T M Φ R M Φ R W ϵ, (22) where G ^ W is defined in Eq (18), and where we have used the fact that ϵT1w = 0 for logistic regression. To obtain the MQLS-LIN test statistic from the CERAMIC test statistic, we replace ZW by the vector of residuals from the ordinary linear regression model, where this model is given by E (Y W | X W) = X W T β and Var (Y W | X W) = σ 2 I. (23) Let β ̌ = (X W T X W) − 1 X W T Y W denote the ordinary least squares estimator for β under model (23), and let e = Y W − X W β ̌ be the resulting null phenotypic residual vector. Then the MQLS-LIN test statistic is given by MQLS-LIN = (e T G ^ W) 2 Var ^ 0 (e T G ^ W | Y, X) = (e T Φ R W T M G R) 2 σ ̌ G 2 e T Φ R W T M Φ R M Φ R W e. (24) Under the same assumptions as for CERAMIC, the MQLS-LOG and MQLS-LIN test statistics both have χ 1 2 asymptotic null distributions. In simulations, we assess the type 1 error and power of the three methods we propose, CERAMIC, MQLS-LOG, and MQLS-LIN, and we compare them to five previously-proposed methods, EMMAX [8], GEMMA [10], MASTOR [17], GMMAT [15] and CARAT [16]. Table 1 summarizes some of the major features of the methods that are particularly relevant to the type 1 error and power studies. In the simulations, we provide CERAMIC, GMMAT and GEMMA with the pedigree-based kinship matrix, while for CARAT and EMMAX, we use an empirical kinship matrix based on 10,000 independently simulated SNPs with their MAFs randomly drawn from the uniform distribution on the interval between 0. 05 and 0. 45. We simulate genotype, covariate and phenotype data for a sample that includes some unrelated individuals and some individuals in three-generation pedigrees (see Fig 1). For each individual, four covariates are simulated: age, sex, height, and an i. i. d. normal covariate. The individuals in pedigrees are each assigned to one of three generations based on their position in the pedigree: first (i. e. , grandparent) generation, second (i. e. , parent) generation, and third (i. e. , offspring) generation. Among the sampled unrelated individuals, 50% are randomly assigned to the first generation, 25% to the second generation, and 25% to the third generation. The age of an individual is generated according to the generation the individual belongs to. For an individual in the first generation, the age is simulated according to a uniform distribution on the set of integers from 78 to 88, i. e. , uniform on {78,79, ⋯, 88}. An individual in the second generation has his or her age generated from a uniform distribution on the set of integers {48,49, ⋯, 58}, and for an individual in the third generation, we use a uniform distribution on the set {18,19, ⋯, 28}. Ages for different individuals are generated independently, regardless of their familial relationships. Let X (2) denote the column vector of age values for a simulated sample. Sampled individuals from three-generation pedigrees have their sex pattern fixed as shown in Fig 1. Among the unrelated individuals, half are randomly assigned to be males and half females. Let X (3) denote the column vector of sex values for a simulated sample. Height is simulated as a heritable trait that exhibits correlation among family members and depends on age and sex. Let X (4) denote the column vector of height values for a simulated sample. The model for height is multivariate normal, given by X (4) | X (2), X (3) ∼ MVN (ν (X (2), X (3) ), σ h a 2 Φ + σ h e 2 I), (25) where σ h a 2 = 36 represents additive polygenic variance for an outbred individual, and σ h e 2 = 13 represents i. i. d. error variance, resulting in narrow-sense heritability ≈ 73%. The mean height vector, ν (X (2), X (3) ), has entry 172. 5 for a male with age ≥ 65 and entry 176. 5 for a male with age < 65. For a female with age ≥ 65, the entry in the mean height vector, ν (X (2), X (3) ), is 160. 2, while for a female with age < 65, the mean height is set to be 163. 2. Let X (5) denote the column vector of the values of the i. i. d. normal covariate for a simulated sample. The entries of X (5) are generated from the N (8,9) distribution, independently of all other covariates. Let X = (1, X (2), ⋯, X (5) ) be the covariate matrix consisting of an intercept and the four covariates described above. Two types of trait model are considered in our simulation studies. One is a mixed-effects logistic regression model, which has the following form: Y i | G, X, u ∼ Bernoulli (p i), independently, with p i = h (f (G i) + X i T β + u i), (26) for i = 1, …, n, where X i T is the ith row of X, u = (u1, ⋯, un) T is a vector of additive polygenic effects, Gi = (Gi1, Gi2) represents the genotypes at causal SNPs 1 and 2 for individual i, and f (Gi) is a function of Gi that represents the combined genetic effect, on the phenotype, of causal SNPs 1 and 2. The coefficient vector β is chosen to satisfy two conditions: (1) the mean of the covariate effects on the logit scale, E (X i T β) = 0, and (2) the variance of the covariate effects on the logit scale, V a r (X i T β), achieves a specified level that depends on the simulation setting. The additive polygenic effects, u, independent of covariates, have a multivariate normal distribution with mean 0 and covariance matrix σ a 2 Φ, where σ a 2 varies across simulation settings. Let θ a = V a r (u i) / V a r (X i T β + u i) = σ a 2 / (V a r (X i T β) + σ a 2) denote the fraction of V a r (X i T β + u i) that is due to additive polygenic effects, and let θc = 1 − θa denote the fraction due to covariate effects. We fix V a r (X i T β) + σ a 2 = 100 and let σ a 2 take possible values 0,20,40,60,80, and 100, so that (θa, θc) takes possible values (0,1), (. 2, . 8), (. 4, . 6), (. 6, . 4), (. 8, . 2) and (0,1), representing a range of the relative importance of additive genetic effects vs. covariate effects. Note that the model also results in Bernoulli error in the phenotype that is conditionally independent across individuals and that accounts for ∼ 20% of total phenotypic variance in our simulation scenarios (where this value is obtained by simulation). Causal SNPs 1 and 2, whose genotypes for individual i are encoded in Gi, are unlinked with minor allele frequencies (MAFs). 1 and. 2 respectively, and they are generated independently of the covariates and additive polygenic effects. They act on the phenotype epistatically: an individual with at least one copy of the minor allele at causal SNP 1 and at least one copy of the minor allele at causal SNP 2 has mean penetrance E (Yi|G) ≈. 15; an individual with a genotype not satisfying that condition has mean penetrance E (Yi|G) ≈. 05. (Note that a target mean penetrance can be achieved by setting f (Gi) to be an appropriate value, obtained in a simulation-based approach.) The resulting prevalence is E (Yi) ≈. 057. The other type of model we consider is a liability threshold model, Y i = 1 if L i ≥ λ (G i) 0 if L i < λ (G i), with L i = X i T β + u i + ϵ i, (27) where Li is the underlying liability for individual i, and λ (Gi) represents the individual’s liability threshold, beyond which the disease is activated, as a function of individual i’s genotypes at causal SNPs 1 and 2. The liability Li consists of three components: X i T β, the covariate effects, ui, the random additive polygenic effects, and ϵi, which represents measurement error or environmental effects assumed to be acting independently across individuals. The Xi’s and the ui’s have the same distributions as described above. The error terms ϵ1, ⋯, ϵn are i. i. d. N (0, σ e 2) and are independent of (X, u). We fix the total liability variance V a r (L i) = V a r (X i T β) + σ a 2 + σ e 2 to be 100, and the error variance σ e 2 to be 20, so that the liability variance due to additive polygenic effects and covariates, V a r (X i T β) + σ a 2 = 80. Let π a = V a r (u i) / V a r (L i) = σ a 2 / (V a r (X i T β) + σ a 2 + σ e 2) represent the fraction of total liability variance due to additive polygenic effects, and let π c = V a r (X i T β) / (V a r (X i T β) + σ a 2 + σ e 2) =. 8 − π a represent the fraction due to covariate effects, while the fraction due to the independent error is fixed at. 2. We choose different values for β and σ a 2 to allow (πa, πc) to take on the possible values (0, . 8), (. 2, . 6), (. 4, . 4), (. 6, . 2), and (. 8,0), representing a range of the relative importance of additive genetic vs. covariate effects. Gi, individual i’s genotypes at causal SNPs 1 and 2, has the same distribution as for the mixed-effects logistic regression model, and in each setting, the values of λ (Gi) are chosen so that an individual with at least one copy of the minor allele at causal SNP 1 and at least one copy of the minor allele at causal SNP 2 has mean penetrance E (Yi|Gi) ≈. 15, while an individual with a genotype not satisfying that condition has mean penetrance ≈. 05. The resulting prevalence is again ≈. 057. In addition, we consider two variations on the liability threshold model. In the first variation, we model the effects of shared environment by incorporating a sibship random effect that accounts for 10% of the total liability variance. In that case, we set the error variance to also account for 10% of the total liability variance, and the additive polygenic and covariate effects are as described in the previous paragraph. In the second variation on the liability threshold model, we modify the threshold values, λ (Gi), so that the prevalence is reduced to. 01. In simulations, we consider both the case when there is complete genotype, phenotype and covariate data as well as cases with missing data. In the complete data case, we simulate genotype, covariate and phenotype data, according to one of the binary trait models described above, for either a 600-person sample (consisting of 30 families, with each family having the 16-person pedigree shown in Fig 1, and an additional 120 unrelated individuals) or a 1000-person sample (consisting of 50 families and an additional 200 unrelated individuals). In some scenarios, families are ascertained conditional on containing at least four affected individuals, while unrelated individuals are sampled at random from the population (call this ascertainment setting A), while in other scenarios, families are ascertained as in setting A while unrelated individuals are sampled in a 1: 1 case-control ratio (call this ascertainment setting B). All sampled individuals are assumed to have non-missing covariates, phenotypes and genotypes. For scenarios with missing data, we simulate genotype, covariate and phenotype data for either a 1,200-person sample (consisting of 60 families, with each family having the 16-person pedigree shown in Fig 1, and an additional 240 unrelated individuals), or a 2,000-person sample (consisting of 100 families and an additional 400 unrelated individuals). In addition, for the run time assessments, we also simulate samples of size 103,4 × 103,6 × 103,8 × 103 and 104, each consisting of 80% related and 20% unrelated individuals. We use ascertainment settings A and B as described above. For each individual in a family, his or her phenotype and covariates are assumed to be all non-missing with probability. 8 (and are assumed to be all missing with probability. 2), independently across individuals and families, and his or her genotype at the tested locus is assumed to be non-missing if and only if at least one of the following two conditions holds: (1) the individual has non-missing phenotype and is affected, or (2) at least half of the individual’s first degree relatives who have non-missing phenotypes are affected. Among the unrelated individuals, all their phenotypes and covariates are assumed to be missing, and all their genotypes are assumed to be non-missing, i. e. , they are included as controls of unknown phenotype. The FHS is a multicohort, longitudinal study whose primary objective is to identify the risk factors and characteristics responsible for cardiovascular disease. The goal of our data analysis is to identify SNPs that are associated with T2D. Our use of the FHS data was approved by the Institutional Review Board of the Biological Sciences Division of the University of Chicago. The FHS sample consists of unrelated individuals as well as individuals from multigeneration pedigrees. For cohort 1 (i. e. , original cohort), we use phenotype and covariate information from 27 clinical exams, for cohort 2 (i. e. , offspring cohort), we use information from 7 clinical exams, and for cohort 3 (i. e. , generation three cohort) we use information from 1 clinical exam. We determine the T2D phenotype status in a similar way to that in a previous work [23]. For individuals in cohort 1, we use data from exams 1–27 to label their T2D status as follows: individuals who have at least one exam with nonfasting blood glucose (BG) level ≥ 200mg/dl or who were under treatment for diabetes, where the measurement or treatment occurred between the ages of 35 and 75 years, are classified as “affected. ” For individuals who have all exams with nonfasting BG<200 mg/dl and have never taken any treatment by the time of the last exam, a phenotype label “unaffected” is given if the individual has age ≥ 70 years at the time of the last exam, while the “unknown” label is given otherwise. We use data from exams 1–7 and exam 1 to determine the phenotype status for individuals in cohorts 2 and 3, respectively. Phenotype is coded as follows for both cohorts: individuals who have at least one exam with fasting plasma glucose (FPG) level ≥ 126mg/dl or who were under treatment for diabetes, where the measurement or treatment occurred between the ages of 35 and 75 years, are classified as “affected. ” For individuals who have all exams with FPG<126 mg/dl and have never taken any treatment by the time of the last exam, a phenotype label “unaffected” is given if the individual has age ≥ 70 years at the time of the last exam while “unknown” is given otherwise. Sex and body mass index (BMI) are included in our analysis as covariates, where we use the mean of an individual’s available BMI values from all clinical exams that the individual participated in. Note that onset age is not included as a covariate in our analysis, because it is not well-defined for an unaffected individual. In addition, the age of an individual at the time of the last exam and cohort ID are artificially correlated to the phenotype status in the restricted sample with known phenotypes, and therefore neither should be added as a covariate. Among the 9240 study individuals for whom Affymetrix 500K genotype data are available, we exclude individuals who have either (1) completeness (the proportion of markers with successful genotype calls) ≤ 96%, or (2) empirical self-kinship coefficient Φ ^ i i ≥ 1. 05. In addition, we exclude 298 individuals whose off-diagonal empirical kinship coefficient values are not consistent with the given pedigree information. Of the 8080 individuals retained in the analysis, 639 are not related to anyone else in the data set with the remaining 7441 related through 840 pedigrees. 6042 individuals have either missing phenotype or missing covariate information, 625 are affected with non-missing covariates, and 1413 are unaffected with non-missing covariates. We exclude from our analysis SNPs that have (1) call rate ≤ 96%, or (2) Mendelian error rate > 2%, or (3) MAF < 1%, which results in a total of 368,802 SNPs retained in the analysis. Furthermore, following Wu and McPeek (submitted), we note that individuals in the original cohort appear to have on average lower genotype quality (lower completeness and higher empirical self-kinship values) than those in the other two cohorts. To prevent spurious association potentially caused by poor genotype quality in cohort 1, for each SNP, we test for an allele frequency difference between cohort 1 and the other cohorts combined. If the allele frequency difference is significant at level 10−7, the SNP is removed from our study. Under this screening procedure, we exclude an additional 1,032 SNPs, resulting in a final set of 367,770 SNPs to include in the analysis. To assess the type 1 error of CERAMIC, MQLS-LOG and MQLS-LIN, we perform simulations as described in Methods. In each simulation scenario, phenotypes are generated according to either the mixed effects logistic regression model of Eq (26) with (θa, θc) = (. 6, . 4) (referred to in Table 2 as “Logistic”) or the liability threshold model of Eq (27) with (πa, πc) = (. 4, . 4) (referred to in Table 2 as “Liability”). Association is tested with a SNP that is neither linked nor associated with the trait, with MAF set to be either. 1 or. 2. In every scenario, we simulate 1,200 individuals with missing data using ascertainment setting A as described in Methods. In each simulation scenario, we consider one of two approaches to analyzing the data, either (1) individuals with partially missing data are included in the analysis for every statistic (denoted by “All” in Table 2), or else (2) individuals with partially missing data are dropped from the analysis for every statistic (denoted by “MX” for “missing excluded” in Table 2). Table 2 shows that in every case, the empirical type 1 error is not significantly different from the nominal, verifying the correct type 1 error of CERAMIC, MQLS-LOG and MQLS-LIN in these scenarios. S1 and S2 Tables contain additional type 1 error results for scenarios that include effects of shared environment on the trait and more stringent ascertainment on the unrelated individuals in a sample (ascertainment setting B) or more stringent ascertainment due to reduced prevalence (value. 01). The number of individuals in a sample ranges from 600 to 2,000, depending on the scenario. In every scenario, the type 1 error remains correct for CERAMIC, MQLS-LOG and MQLS-LIN. We perform additional simulations in which we compare the type 1 error rate of GEMMA to those of CERAMIC, MQLS-LOG and MQLS-LIN. Phenotypes are generated according to the liability threshold model of Eq (27) with various settings of (πa, πc). In every scenario, we simulate 1,200 individuals with missing data and use ascertainment setting A. Association is tested with an unlinked, unassociated SNP with MAF. 2. S3 Table shows that in every scenario, the type 1 error of GEMMA is significantly inflated when the mean genotype value is imputed for the missing genotypes, and it is significantly deflated when the missing genotypes are removed. In contrast, the type 1 error of CERAMIC, MQLS-LOG and MQLS-LIN is correct in all scenarios. Because we observe uncontrolled type 1 error for GEMMA in these settings, we do not consider GEMMA further in our simulations. In an association study, the correct trait model is generally unknown. In particular, it may not be known which covariates should be included in the model. In Table 3, we compare the type 1 error of CERAMIC, MQLS-LOG and GMMAT in the situation in which the relevant covariates are inadvertantly left out of the fitted model. The results show that in almost every scenario, the type 1 error of GMMAT is compromised (either significantly inflated or significantly deflated) when the trait model is misspecified. In contrast, CERAMIC and MQLS-LOG retain correct type 1 error when the trait model is misspecified. This likely reflects the fact that retrospective methods tend to be much more robust to phenotype model misspecification than prospective methods are. We compare the power of CERAMIC to that of MQLS-LOG, MQLS-LIN, MASTOR, EMMAX, GMMAT and CARAT in various simulated scenarios with missing data and ascertainment, as described in Methods. GMMAT offers two options to deal with missing genotype data: the missing genotypes can either be removed or replaced by the estimated mean genotype value. In practice, we found that the two options give identical results, so the simulation results we report for GMMAT apply to both options. Panel A in Fig 2 gives the empirical power results for various settings of the mixed effects logistic regression model with missing data, while Panel B in Fig 2 gives the results for the liability threshold model with missing data. In both cases, ascertainment setting A is used. Numerical results for these power simulations can be found in S4 and S6 Tables. S12 and S13 Tables give power results for additional missing data scenarios that include effects of shared environment on the trait and more stringent ascertainment on the unrelated individuals in a sample (ascertainment setting B) or more stringent ascertainment due to reduced prevalence (value. 01). From Panels A and B in Fig 2, it is clear that in every scenario, in terms of power, CERAMIC either outperforms, or has equivalent performance to, the best of the other methods, regardless of the relative strength of covariates and additive polygenic effects on the trait. In particular, CERAMIC has dramatically higher power than the previously-proposed binary trait methods GMMAT and CARAT. This result also holds in the partially missing data scenarios in S12 and S13 Tables. In Panels A and B of Fig 2, a major feature distinguishing the power of the methods is that those methods that make sophisticated use of missing data (CERAMIC, MQLS-LOG, MASTOR and MQLS-LIN) substantially outperform those that do not (CARAT, EMMAX and GMMAT). Within each of these two groups, when covariate effects are large relative to additive polygenic effects (θa ≤. 4 or πa ≤. 2), the methods that fit a logistic mean structure outperform the methods that fit a linear mean structure, i. e. , CERAMIC and MQLS-LOG outperform MASTOR and MQLS-LIN, while CARAT and GMMAT outperfom EMMAX. For the mixed-effects logistic regression trait model, this is not surprising, because the simulated model also has a logistic mean structure. However, it is notable that, among methods that treat missing data in the same way, the methods that fit a logistic mean structure also outperform those that fit a linear structure in the case of the liability threshold model, in which the simulated model does not have a logistic mean structure. This improvement may be due to the flexibility afforded by the nonlinearity of a logistic mean structure. Through the power comparison of CERAMIC to MQLS-LOG and that of MASTOR to MQLS-LIN, we can see that fitting the additive VC (as in MASTOR and CERAMIC) does not harm power in any scenario, and it improves power when the additive polygenic effects are large relative to covariate effects (θa ≥. 6, or πa ≥. 8). When additive polygenic effects are large, GMMAT has the lowest power of all methods. This might reflect known limitations [24] of the penalized quasi-likelihood approach, which is used by GMMAT. We perform additional simulations to compare the ability of the three binary trait methods, CERAMIC, GMMAT and CARAT, to recover power from partially missing genotype information when mean values are plugged in for missing genotype values in both GMMAT and CARAT. We simulate under the liability threshold trait model with (πa, πc) = (40,40) with 1,200 individuals in each simulated replicate, under acertainment setting A with missing data. Because the data are simulated, the missing genotype values are actually available, so we can determine what the power would have been for each of the three methods had the genotype data not been missing. This power is represented in the leftmost set of three bars in Fig 3, labeled “Complete Genotype Data. ” This is compared to the power when the individuals with missing genotype data are removed from the input files before the methods are run, and the power when individuals with missing genotype data remain in the input files and CERAMIC is run in the usual way, while GMMAT and CARAT are run with the mean genotype value plugged in for the missing genotypes. From Fig 3, we can see that in this setting, CERAMIC is able to recover virtually all of the power of the complete genotype data by using the information from the partially missing genotype data. In contrast, the strategy of imputing the mean genotype value for missing genotype data in GMMAT or CARAT results in power that is not significantly different from that obtained by throwing those individuals out of the analysis. This demonstrates that appropriate handling of missing data can result in a substantial power advantage compared to a simple strategy such as imputing the mean genotype value or discarding missing values. Because the trait model typically cannot be known with certainty, it is always possible that relevant covariates may be left out of the fitted model. Panels C and D of Fig 2 show results from the same simulated scenarios as those of Panels A and B, respectively, but for the situations in which the fitted model excludes the relevant covariates. Numerical results for these power simulations can be found in S5 and S7 Tables. In Fig 2, from a comparison of Panels A and B to Panels C and D, we can see that, for all methods, adjusting for covariates improves power in the settings in which covariates play a role in explaining the phenotypic variation (θa < 1 or πa <. 8) and does not compromise power in the other cases. In Panels C and D, it can be seen that CERAMIC has the highest power in all settings. In contrast, GMMAT is severely underpowered, having the lowest power (or power not significantly different from the lowest) for all settings. Among the three methods that do not correct for missing data (CARAT, EMMAX and GMMAT), CARAT, which is a retrospective method, has the highest power for all the settings in which the model is misspecified (θa ≤. 8 in Panel C and πa ≤. 6 in Panel D), likely reflecting the greater robustness to trait model misspecification of the retrospective methods. The power difference between the methods that correct for missing data (CERAMIC and MQLS-LOG) and the others is quite large (up to 6-fold), particularly in the settings in which covariates play an important role. S8, S9, S10 and S11 Tables give power for scenarios analogous to those in Fig 2 but with complete data instead of partially missing data. S12 and S13 Tables give power results for additional complete-data scenarios that include effects of shared environment on the trait and more stringent ascertainment on the unrelated individuals in a sample (ascertainment setting B) or more stringent ascertainment due to reduced prevalence (value. 01). With complete data, when all relevant covariates are included in the model, the three binary trait methods have approximately equal power in all scenarios. However, when relevant covariates are excluded from the model, the power of the prospective method GMMAT is lower than that of the retrospective methods CERAMIC and CARAT. The MQLS-LOG method, which has a logistic mean structure but does not include an additive polygenic VC has power approximately equal to that of the most powerful binary trait methods (CERAMIC and CARAT) when the additive polygenic variance is low, but loses power when the additive polygenic variance accounts for a high proportion of the trait variance. The EMMAX and MASTOR methods, which have an additive polygenic VC but linear instead of logistic mean structure, have power approximately equal to that of the most powerful methods (CERAMIC, CARAT and GMMAT) when covariates do not play an imporant role in the trait model and additive polygenic variance does. Compared to the retrospective method MASTOR, the prospective method EMMAX loses power when relevant covariates are omitted from the fitted model in the settings in which covariates play an important role in the trait model (θa ≤. 2 or πa ≤. 2). For EMMAX, in particular, this could be explained in more detail by the fact that EMMAX assesses the variation of the test statistic based on phenotypic variance (i. e. , phenotypes are treated as random in EMMAX), and failure to adjust for covariates would lead to inflation of the estimated phenotypic variance, and hence, to a reduction in power, whereas the retrospective methods such as MASTOR and CERAMIC assess variation based on genotypic variance (i. e. , genotypes are treated as random), so are robust to power loss arising from misspecification of the phenotype model. Regardless of whether or not covariates are adjusted for, CERAMIC has higher power than EMMAX when covariate effects are large relative to additive polygenic effects (θa ≤. 2 or πa ≤. 2), and maintains similar power to EMMAX in other scenarios. For the analysis of T2D data from the FHS (sample size 8080 individuals with 367,770 SNPs after quality control), we restrict consideration to methods that did not experience type 1 error problems in our simulations. We compare CERAMIC, MASTOR, EMMAX, MQLS-LOG and MQLS-LIN. Tables 4 and 5 report the estimates, with standard errors, of the regression parameters and VCs obtained by CERAMIC and MASTOR (which use different null phenotypic models). The Q-Q plots (not presented) for the genome scan p-values from all five methods do not exhibit any evidence of inflation, and their genomic control inflation factors [25] are all below 1. 01. Table 6 presents the p-values for the SNPs with the strongest association signals with T2D, i. e. , the SNPs for which at least one of CERAMIC, MASTOR and EMMAX gives a p-value < 2 × 10−5. The two SNPs with the smallest p-values, rs4506565 and rs7901695, are in an intron of TCF7L2 (MIM 602228), which has been extensively reported to have strong association with T2D [26–28]. Among the other 5 genes listed in the table, TLE1 (MIM 600189) has previously been reported and replicated as a T2D susceptibility locus [29,30], and GALNT9 (MIM 606251) is the left flanking gene of SNP rs10747083 previously found to be significantly associated with fasting glucose [31]. DLGAP1 has previously been associated with serum insulin-like growth factor-binding protein 3 (IGFBP-3) levels [32]. DLGAP1 has also been previously associated [33] with levels of cardiac troponin T measured by a highly sensitive assay (hs-cTnT), where hs-cTnT has been found [34] to be associated with diabetes mellitus in patients with stable coronary artery disease. PALLD has previously been associated [35] with aspartate aminotransferase (AST) level, where elevated AST level has shown evidence of possible association with risk of T2D [36]. From Table 6, we observe that among the three tests that account for additive polygenic effects, i. e. CERAMIC, MASTOR and EMMAX, EMMAX almost always gives the largest p-values, while CERAMIC often yields the smallest p-values. CERAMIC as well as MQLS-LOG and MQLS-LIN are implemented in the CERAMIC softare, which will be made freely available at http: //www. stat. uchicago. edu/~mcpeek/software/index. html. We report run times for CERAMIC in simulations and in analysis of the FHS data set. All runs are completed using only one core (at 3. 5GHz) of Intel Xeon CPU E5-2637 v3. In CERAMIC, the time-limiting step is the incorporation of missing data, which depends strongly on the sizes of the individual pedigrees making up the sample because the missing data incorporation is based only on genotype and phenotype information from the close relatives of the individuals with missing genotype. Therefore, because incorporation of missing data is the slowest step, for fixed size of the families making up the sample, the computation could be expected to scale approximately linearly in sample size. We report run times on simulated data sets with varying sample sizes, where each sample consists of 20% unrelated individuals and 80% related individuals in families of the type in Fig 1, tested at 50000 SNPs. For sample sizes of 1 × 103,2 × 103,4 × 103,6 × 103,8 × 103 and 1 × 104, we obtain run times of 4. 9,7. 8,16. 2,23. 6,32. 8 and 39. 0 minutes, respectively. These run times are plotted in S1 Fig, from which it is clear that the run time is indeed approximately linear for a fixed family size. For the FHS data set, which contains 8080 individuals, it takes approximately 4. 2 hours to perform a scan of 367,770 SNPs with phenotypic residuals computed once per genome screen, and approximately 6. 1 hours if phenotypic residuals are computed separately for each SNP. This time is greater than would be needed for a sample of 8080 individuals of the type in our simulations because Framingham includes several families that each have hundreds of individuals, for whom the missing data step is much more time consuming. However, even in this case, the computation time would be expected to scale approximately linearly as the sample size increased, for a fixed family complexity. With complete data, the computations could be substantially sped up by taking a different algorithmic approach, similar to those used in LMM methods. However, such an approach is not optimal for partially missing data. In all cases, the computations are easily parallelized across tested variants. For genetic association mapping of binary traits in samples with related individuals, we have developed a new method, CERAMIC, which incorporates pedigree and covariate information and effectively handles partially missing data. CERAMIC is applicable to samples that contain essentially arbitrary combinations of related and unrelated individuals. CERAMIC can be viewed as a hybrid of logistic regression and LMM approaches. Like LMM methods, CERAMIC incorporates an additive component of variance and can accommodate related individuals in a computationally feasible way. Like logistic regression methods, CERAMIC uses a logistic function to model the effects of covariates on a binary trait, and it accounts for the dependence of the variance on the mean (i. e. Bernoulli variance). As a result, CERAMIC is able to gain power, over LMM methods, for association mapping of binary traits. In addition to adjusting for covariates, CERAMIC can increase power by incorporating partially missing data. CERAMIC is based on a set of estimating equations, and we take a retrospective approach to assessment of significance of the test statistic, which provides a way to more easily incorporate partially missing data and also leads to robustness of the method to misspecification of the phenotype model. CERAMIC is implemented in freely-available software and is computationally feasible for current genome-wide association studies. In simulations, we demonstrate that CERAMIC outperforms previously-proposed binary-trait methods GMMAT and CARAT in scenarios with partially missing data, with CERAMIC giving large increases in power over the other two methods in many scenarios. CERAMIC also outperforms GMMAT when the trait model is misspecified, with both large increases in power and also improved type 1 error control over GMMAT. When there are no missing data and the correct set of covariates is included in the fitted model, the three methods have approximately equal power. We show that the sophisticated handling of partially missing data in CERAMIC can recover a large portion of the power of complete data. In contrast, imputation of the mean genotype value for missing genotype data in GMMAT or CARAT does a poor job, recovering almost no power. In a range of simulated scenarios with different types of trait models, various levels of relative importance of covariate effects and additive polygenic effects within a trait model, and either complete data or partially missing data, CERAMIC outperforms or performs as well as the best-performing of the other methods considered, MQLS-LOG, MQLS-LIN, MASTOR, EMMAX, GMMAT and CARAT. In addition, we have demonstrated that, when covariates play a major role in the trait model and relevant covariates are included in the fitted model, the methods that incorporate a logistic mean structure tend to perform better than those that incorporate a linear mean structure, even when the underlying trait model is not logistic, but instead follows a liability threshhold model. When additive polygenic effects play a major role in the trait model, the methods that include an additive polygenic VC tend to have higher power than the other methods. When data are partially missing among related individuals, the retrospective methods that incorporate sophisticated missing data handling (CERAMIC, MQLS-LOG, MQLS-LIN and MASTOR) boost power by exploiting information contained in partially missing data. We apply our methods to analysis of T2D in the FHS data, where we replicate association with two previously-identified T2D susceptibility loci TCF7L2 [26–28], and TLE1 [29,30]. In fact, of the 10 smallest CERAMIC p-values in our genomewide analysis, 9 occur in or near either known T2D susceptibility loci or plausible candidates (6 loci in total), verifying that CERAMIC is able to home in on the important loci in a genome scan. In genetic association studies, it can be of interest to estimate the association parameter γ, for example, as a way to quantify the strength and direction of association and/or to build a predictive phenotype model. With complete data, γ can be estimated by γ ^, where (γ ^, β ^, ξ ^) is the solution of the system consisting of Eqs (3), (4) and (8), with the standard error of γ ^ given by the square root of the first diagonal element of Eq (14). With partially missing data, if our primary aim is to estimate γ, rather than to test for association, then we first need to make a careful choice of the set of individuals to be included in the estimation. We would naturally include all the individuals in U, the set of individuals with complete data, and we can also include a subset S′ ⊂ S, where S′ is a set of individuals who have non-missing phenotype and covariate information, and whose genotypes can be informatively estimated from genotyped relatives. We can then set W′ = U ∪ S′ and use Eq 19, with W′ and S′ substituted for W and S, to obtain the BLUP vector G ^ W ′. Then γ can be estimated by solving the system of Eqs (3), (4) and (8), where all vectors and matrices are restricted to contain only those individuals in set W′ and where G ^ W ′ is substituted for G. In the presence of substantial amounts of missing data, the choice of the set S′ could potentially impact the properties of the resulting estimator of γ. While including in the estimation some ungenotyped individuals whose genotypes can be informatively estimated could improve the precision of the estimator by drawing information from genotyped relatives, inclusion of ungenotyped individuals on whose genotypes the data provide relatively low information could bias the estimate of γ. This is in contrast to the testing problem, in which including such individuals would simply provide a relatively low amount of additional power. In deriving phenotypic residuals, we have used an estimating equation framework to estimate the regression coefficients and VCs. This framework, although built specifically for binary traits, can be generalized to traits with a general exponential family distribution, e. g. , count phenotypes distributed as Poisson. For such traits, we propose a general mean and variance structure for testing the null hypothesis, E (Y i | X, G) = μ i (β) = g - 1 (x i T β + G i γ), (28) Var 0 (Y | X, G) = diag (V (μ 1), ⋯, V (μ n) ) Σ diag (V (μ 1), ⋯, V (μ n) ), (29) where g (x) and V (μ) are the link and variance functions, respectively, chosen for the given exponential-family distribution [37]. For binary traits, typical choices for g (x) and V (μ) are the logit function (i. e. log (x/ (1 − x) ) ) and Bernoulli variance μ (1 − μ), while for Poisson traits, typical choices would be g (x) = log (x) and V (μ) = μ. Furthermore, the correlation matrix Σ can be extended to include more VCs (e. g. , dominance variance) by assuming Σ = ξ 1 Φ 1 + ⋯ + ξ k Φ k + (1 − ∑ 1 k ξ i) I. A system of estimating equations can be constructed in a similar way and solved in a recursive fashion.

Case-control association testing has proven to be useful for identification of genetic variants that affect susceptibility to disease. One can expect to gain power for detecting such variants by including relevant covariates in the analysis, by accounting for any relatedness of sampled individuals, and by making use of partial information in the data. For analysis of continuously-varying traits, variations on linear mixed-model (LMM) approaches have proven effective at achieving some of these goals. However, for case-control or binary trait mapping, there remain significant challenges. Direct application of LMM approaches to binary traits suffers from power loss when covariate effects are strong, and existing generalized LMM approaches can perform poorly in the presence of trait model misspecification and partially missing data. We propose CERAMIC, a method for binary trait mapping, which is computationally feasible for large genome-wide studies, and which gains power over previous approaches by improved trait modeling, retrospective assessment of significance, accounting for sample structure, and making use of partially missing data. We illustrate this approach in genome-wide association mapping of type 2 diabetes in data from the Framingham Heart Study.

Abstract Introduction Materials and Methods Results Discussion

genome-wide association studies medicine and health sciences cardiovascular anatomy variant genotypes genetic mapping simulation and modeling diabetes mellitus endocrine disorders mathematics statistics (mathematics) materials science test statistics genome analysis type 2 diabetes materials by structure research and analysis methods genomics mathematical and statistical techniques endocrinology metabolic disorders anatomy phenotypes heredity genetics biology and life sciences physical sciences computational biology ceramics statistical methods heart human genetics

2016

CERAMIC: Case-Control Association Testing in Samples with Related Individuals, Based on Retrospective Mixed Model Analysis with Adjustment for Covariates

In line with the key role of methionine in protein biosynthesis initiation and many cellular processes most microorganisms have evolved mechanisms to synthesize methionine de novo. Here we demonstrate that, in the bacterial pathogen Staphylococcus aureus, a rare combination of stringent response-controlled CodY activity, T-box riboswitch and mRNA decay mechanisms regulate the synthesis and stability of methionine biosynthesis metICFE-mdh mRNA. In contrast to other Bacillales which employ S-box riboswitches to control methionine biosynthesis, the S. aureus metICFE-mdh mRNA is preceded by a 5′-untranslated met leader RNA harboring a T-box riboswitch. Interestingly, this T-box riboswitch is revealed to specifically interact with uncharged initiator formylmethionyl-tRNA (tRNAifMet) while binding of elongator tRNAMet proved to be weak, suggesting a putative additional function of the system in translation initiation control. met leader RNA/metICFE-mdh operon expression is under the control of the repressor CodY which binds upstream of the met leader RNA promoter. As part of the metabolic emergency circuit of the stringent response, methionine depletion activates RelA-dependent (p) ppGpp alarmone synthesis, releasing CodY from its binding site and thereby activating the met leader promoter. Our data further suggest that subsequent steps in metICFE-mdh transcription are tightly controlled by the 5′ met leader-associated T-box riboswitch which mediates premature transcription termination when methionine is present. If methionine supply is limited, and hence tRNAifMet becomes uncharged, full-length met leader/metICFE-mdh mRNA is transcribed which is rapidly degraded by nucleases involving RNase J2. Together, the data demonstrate that staphylococci have evolved special mechanisms to prevent the accumulation of excess methionine. We hypothesize that this strict control might reflect the limited metabolic capacities of staphylococci to reuse methionine as, other than Bacillus, staphylococci lack both the methionine salvage and polyamine synthesis pathways. Thus, methionine metabolism might represent a metabolic Achilles' heel making the pathway an interesting target for future anti-staphylococcal drug development. Staphylococci are important skin and mucosa commensals but also major human pathogens. The most pathogenic species Staphylococcus aureus causes a wide range of diseases and, together with coagulase-negative staphylococci (CoNS), accounts for approximately 30 per cent of all hospital-acquired infections [1]. The development of antibiotic resistance in staphylococci increasingly limits therapeutic options and is a matter of major concern [2]. In recent years, studies into staphylococcal metabolism and its possible links to bacterial virulence have become a major focus of research but basic metabolic pathways remained largely unexploited in the development of new antibiotic drugs [3]. In this study, we investigate the regulation of methionine biosynthesis in staphylococci. Methionine and its chemical derivatives have important functions in the cell. For example, (formyl-) methionine is the universal N-terminal amino acid of nearly all proteins and therefore plays an eminent role in the initiation of protein biosynthesis. Moreover, the methionine derivative S-adenosylmethionine (SAM) serves as a methyl group donor in a variety of cellular processes and is the precursor molecule in polyamine synthesis [4]. Many microorganisms are able to synthesize methionine de novo and staphylococci employ the trans-sulfuration pathway to generate methionine [5]. Most bacteria from the order Bacillales are thought to control this pathway by SAM-binding S-box riboswitches [5], [6], [7]. Interestingly, in silico analysis predicts the presence of a T-box riboswitch in the 5′-untranslated region of the methionine biosynthesis operon (metICFE-mdh operon) in staphylococci [5], [6], [8], suggesting the use of alternative mechanisms to regulate methionine synthesis. T-box riboswitches are transcriptional control systems which have been extensively studied in Bacillus subtilis and other Firmicutes (reviewed in [6]). Their function is controlled by specific interactions and differential binding to charged and uncharged cognate tRNA, respectively, thus providing a means to “sense” the amino acid concentration in the cell [9]. T-box leader RNA/tRNA interaction essentially occurs at two sites: (i) the tRNA anticodon basepairs with the specifier-loop domain of the T-box leader RNA ensuring specific binding of the respective T-box element with its cognate tRNA; (ii) the free 3′-CCA end of an uncharged tRNA binds to the T-box motif, thereby triggering the formation and stabilization of an antiterminator which enables transcription of downstream genes [9]. In this study, we characterized interactions of the metICFE-mdh leader RNA with methionyl-tRNAs and demonstrate that they represent a functional T-box riboswitch that preferentially binds to initiator formylmethionyl-tRNA (tRNAifMet). We further show that, in staphylococci, T-box control of methionine biosynthesis has a key role in a complex regulatory network that also involves stringent response-mediated CodY regulation and RNA decay to tightly control this pathway. The methionine biosynthesis genes (metI, metC, metF, metE, mdh (metal-dependent hydrolase) ) are organized in an operon-like structure and are annotated as SACOL0431 - SACOL0427 in S. aureus COL and as NWMN_0351 - NWM_0347 in S. aureus Newman, respectively, with metF being named metH in the latter strain (Figure 1A). Northern blot analysis using a double-stranded DNA probe confirmed the expression of a stable transcript of approximately 400 nucleotides (nt) from the intergenic region (IGR) upstream of metI (Figure 1B, left panel). Hybridization employing in vitro-transcribed RNA probes revealed that the orientation of the transcript was identical to that of the metICFE-mdh operon (Figure 1B, middle and right panels). 5′- and 3′-RACE experiments identified a single transcription start site and a transcript length of 439 nt (Figure 1C). Sequence analysis of various S. aureus and S. epidermidis strains demonstrated that the region is highly conserved (SI Figure S1) but lacks ribosomal binding sites and open reading frames, suggesting specific (non-coding) functions of the transcript, for example, as a 5′-untranslated region (5′-UTR) of the metICFE-mdh RNA. Interestingly, a putative binding site for the repressor protein CodY [10], [11] could be identified next to the 5′-UTR promoter region (Figure 1C). Most striking was the presence of a highly conserved canonical T-box sequence motif (5′-AAGGUGGUACCGCG-3′) which partially overlapped with a strong Rho-independent transcription termination signal in the 3′-portion of the transcript (Figure 1C). Overall, the transcript, which was named met leader RNA, harbored all the characteristics of previously characterized T-box riboswitches from Bacillus subtilis, and sequence alignments with Bacillus T-box systems led to a putative structural model of the Staphylococcus met leader RNA (SI Figure S2). First, we tested if the Staphylococcus met leader RNA interacts specifically with methionyl-tRNAs (Figure 2A). met leader RNA was in vitro transcribed in the presence of the appropriate radioactively labeled tRNA species. Binding between radioactively labeled tRNA and in vitro-transcribed met leader RNA was determined by non-denaturing polyacrylamide gel electrophoresis and autoradiography. Staphylococci genomes harbor four methionyl-tRNA gene loci, two of which (tRNA-Met-1 and -2) being identical and representing the initiator tRNAifMet. Binding studies using tRNAifMet, tRNAMet3 and tRNAMet4, respectively, with free 3′-CCA ends revealed that the met leader RNA interacted strongly with tRNAifMet while interactions with tRNAMet3 and tRNAMet4 proved to be weak (Figure 2B). tRNAifMet binding to met leader RNA increased linearly within a 5-fold molar range (Figure 2C). In contrast, binding was abolished when the 3′-end of tRNAifMet included one additional cytosine, mimicking a charged tRNA molecule (3′-CCAC; AdC in Figure 2D). Also, no interaction was detectable in the presence of cysteinyl-tRNA, regardless of whether 3′-CCA or 3′-CCAC was present at the 3′ end (Figure 2C, D). In classical T-box riboswitches, tRNA/leader RNA interaction is mediated by the T-box motif which forms a bulge that facilitates basepairing interactions with the tRNA 3′-CCA and supports antiterminator formation [9] (Figure 2A). We therefore studied whether the predicted T-box motif participates in tRNAifMet binding by generating a series of met leader RNAs carrying T-box mutations (SI Table S2, Figure 3A). tRNAifMet binding was clearly diminished in mutants SC2 and SC5, which are both likely to lack the putative T-box bulge for tRNA 3′-CCA interaction (Figure 3A). Also, in SC8, a single U to A exchange at position 363 was sufficient to reduce tRNAifMet binding, whereas other mutations within the putative T-box bulge, i. e. SC3, SC4, SC6 and SC7, enhanced tRNAifMet interactions with the met leader RNA (Figure 3B). Finally, alteration of a methionine-specific codon AUG to cysteine UGC in the putative specifier box of the met leader RNA in SC1 did not affect tRNAifMet binding efficiency (Figure 3B), nor did it confer tRNACys binding activity (SI Figure S3) suggesting that other components of this T-box system confer specificity to initiator tRNAifMet binding. Taken together, these data suggest that the met leader RNA upstream of the staphylococcal metICFE-mdh operon harbors a canonical T-box riboswitch that specifically binds uncharged initiator tRNAifMet. Through interaction with either uncharged tRNAs (antiterminator formation) or charged tRNAs (terminator formation), T-box riboswitches indirectly sense amino acid levels in the bacterial cell [6], [9]. To determine whether or not met leader RNA/metICFE-mdh transcription is sensitive to methionine availability, S. aureus strain Newman was grown in chemically defined medium (CDM) in the presence or absence of methionine. RNA was isolated in the early exponential (E1), exponential (E2) and early stationary (S) growth phase and analyzed by Northern hybridization using met leader RNA- and metI-specific DNA probes, respectively (Figure 4A, B). In the presence of methionine, basal met leader RNA transcription could be detected. Methionine starvation induced the transcription of this RNA, especially during exponential growth (Figure 4A). In contrast, the metI mRNA signal was not detectable in the presence of methionine (Figure 4B). Upon methionine deprivation, however, metI mRNA transcription was activated with the strongest expression detected during exponential growth (Figure 4B). Interestingly, metI-specific Northern probing did not reveal a distinct fragment representing the full-length metICFE-mdh mRNA (Figure 4B). Instead, an RNA smear was detected in repeated experiments. As the RNA quality and integrity had been confirmed prior to the experiments, the Northern blot data suggest rapid degradation of the methionine starvation-induced metICFE-mdh transcript. CodY is a global transcription repressor that controls the expression of a variety of genes in S. aureus, many of which being involved in amino acid biosynthesis and transport [10], [11]. Detection of a consensus sequence for CodY binding upstream of the met leader RNA and the recent identification of the methionine biosynthesis genes as direct CodY targets in S. aureus [10] prompted us to study the role of this factor for met leader/metICFE-mdh transcription in more detail. A S. aureus codY deletion mutant was grown in CDM with and without methionine and analyzed for met leader and metICFE-mdh transcription by Northern blot hybridization. In the absence of methionine, a stronger met leader RNA signal was detected in both the wildtype and the codY mutant (Figure 4A). However, compared to the wildtype, the codY mutant showed a generally enhanced met leader transcription, suggesting that met leader RNA expression was de-repressed in the absence of CodY irrespective of whether methionine was present or not (Figure 4A). In contrast, downstream metICFE-mdh operon expression remained sensitive to varying concentrations of methionine and was only activated both in the wildtype and the codY mutant in the absence of methionine (Figure 4B). In many bacteria, nutrient limitation triggers the so-called stringent response to appropriately adjust gene expression patterns. Stringent response is characterized by the rapid synthesis of the alarmone (p) ppGpp involving bifunctional RelA/SpoT synthetases/hydrolases (RSHs) and affecting/modulating many cellular functions. Recently, a link between CodY and the stringent response of S. aureus has been demonstrated [12]. The CodY repressor function depends on its two effector molecules GTP and branched-chain amino acids (BCAA, valine, leucine, isoleucine) which enhance synergistically the affinity of CodY for its DNA targets [13], [14]. RSH-mediated (p) ppGpp synthesis lowers the GTP levels in the cell and eventually facilitates release of CodY from its DNA targets [12]. In a next set of experiments, we sought to identify possible regulatory links between methionine deficiency, stringent response and CodY. For this purpose, met leader RNA/metICFE-mdh transcription upon methionine depletion was studied in a rsh mutant carrying a deletion of the (p) ppGpp synthetase domain in strain S. aureus Newman [12]. As a marker for stringent response-controlled genes, brnQ-1, which encodes a CodY-repressed BCAA permease, was included in the analysis. In the S. aureus wildtype, methionine depletion led to brnQ-1 induction along with met leader RNA and metICFE-mdh expression (Figure 5A). In contrast, in the rsh mutant, induction of both brnQ-1 and met leader/metICFE-mdh transcription were significantly reduced upon methionine starvation in comparison to the wildtype, suggesting that RSH-mediated (p) ppGpp synthesis may be required for efficient activation of the system. Deletion of codY resulted in a generally higher basal transcription of both brnQ-1 and met leader RNA in the presence of methionine, whereas metICFE-mdh transcription remained tightly controlled and switched off under these conditions (Figure 5A). In the codY mutant, brnQ-1 expression was de-repressed and not further inducible by methionine deprivation, suggesting that CodY is responsible for the methionine-dependent brnQ-1 induction observed in the wildtype (Figure 5A). The experiments described in Figure 4 suggest that metICFE-mdh mRNA is subject to rapid degradation. To investigate the possible involvement of specific RNases in this process, the stability of met leader RNA and metICFE-mdh was analyzed in S. aureus mutants that were deficient in RNase J2 and RNase III activity, respectively. For this purpose, de novo RNA synthesis was interrupted by the addition of the RNA polymerase inhibitor rifampicin to the cultures. Total RNA was isolated at different time points and subjected to Northern blot analysis (Figure 5B). Comparison of the wildtype and the RNase-deficient mutants revealed that the metICFE-mdh transcript was more stable in the RNase J2 mutant, suggesting that RNase J2 may be involved in metICFE-mdh degradation. In contrast, no significant effect on met leader RNA stability was detectable in the RNase J2 mutant (Figure 5B, upper panel). The RNase III mutant exhibited a slightly enhanced stability both of the met leader and metICFE-mdh RNA indicating a possible function of RNase III in met leader RNA and metICFE-mdh decay (Figure 5B). In this study, we show that methionine biosynthesis control in S. aureus involves a T-box riboswitch. While the conservation of this T-box in staphylococci was predicted previously using bioinformatic tools [5], [6], [8], we now provide, for the first time, direct experimental proof for a specific interaction of the predicted T-box with initiator tRNAifMet. While methionyl-tRNA-specific T-box riboswitches (met-T-box) are rare among Bacillales they are more common in Lactobacillales. In both orders they are associated with methionine metabolism or transport (Table 1). Methionyl-tRNAs (tRNAMet) are encoded by four distinct gene loci in the genomes of S. aureus and S. epidermidis. Two of them are identical and represent the initiator tRNAifMet, while the other two tRNAMet loci differ in their nucleotide sequence from each other and from the tRNAifMet. Surprisingly, we found a clear preference for interaction of the met leader RNA T-box with the initiator tRNAifMet (Figure 2B). In prokaryotes, the first N-terminal methionine of newly synthesized proteins is N-formylated and, hence, N-formylmethionine (fMet) is indispensable for protein translation initiation and bacterial growth. fMet is carried to the ribosomal translation initiation complex by tRNAifMet which differs structurally from the elongator tRNAMet used for the incorporation of methionine residues into the growing polypeptide chain. Although all tRNAMet are charged with methionine by (the same) methionyl-tRNA synthetase, it is only tRNAifMet that is specifically recognized by the methionyl-tRNA-formyltransferase which then mediates N-formylation of methionine to produce fMet. At present, it is not clear how the observed specificity of the met leader RNA T-box for tRNAifMet is accomplished. An involvement of the putative specifier box in the 5′-region of the met leader RNA seems unlikely because all methionyl-tRNAs use the same anticodon, suggesting that other regions of the met leader RNA interact with structures that are unique to the initiator tRNAifMet. In line with this hypothesis, our data showed that tRNAifMet binding by the met leader RNA was not affected when the specifier box in the met leader was substituted with a cysteine-specific codon (Figure 3B). Also, these nucleotide replacements were insufficient to confer tRNACys binding activity (SI Figure S3). Our observation that the T-box riboswitch, shown in this study to be a key regulator of methionine biosynthesis in S. aureus, preferentially binds tRNAifMet points to an elegant mechanism by which protein translation initiation efficiency could both be sensed and, if necessary, adjusted by modulating fMet supply. It will be interesting to investigate if this tRNAifMet preference also applies to other met-T-box riboswitches that control the expression of genes not directly involved in methionine biosynthesis. Also, potential metabolic implications of the use of T-box-controlled fMet supply in staphylococci versus S-box-controlled methionine biosynthesis in other bacteria remain to be studied. The data obtained in this study lead us to propose that a hierarchical regulatory network controls methionine biosynthesis in S. aureus, most likely, to minimize unnecessary de novo methionine biosynthesis. Centerpiece of this regulation turns out to be the tRNAifMet-specific T-box riboswitch located in the 5′-met leader that precedes the coding regions of the metICFE-mdh mRNA. Another important player is the global repressor CodY which drives met leader RNA transcription and links the system to the metabolic emergency circuit of the bacterial stringent response. Finally, staphylococcal RNases were implicated in this network by degrading both metICFE-mdh mRNA and met leader RNA, which may be considered a form of posttranscriptional control of metICFE-mdh gene expression. Figure 6 summarizes our major findings and suggests a model for the control of methionine biosynthesis in staphylococci. While stringent response-mediated CodY release and subsequent met leader RNA transcription are sensitive to general amino acid availability and the energy status of the cell, the T-box riboswitch is highly selective and ensures that downstream metICFE-mdh mRNA transcription only occurs if methionine concentration is low. The experiments also indicate that lack of methionine alone is sufficient to trigger the stringent response and, as a consequence, the release of CodY, thus securing efficient met leader RNA/metICFE-mdh transcription when needed (Figure 6). Interestingly, the regulatory cascade identified in this study seems to represent an exception rather than common rule. Thus, database searches of B. subtilis and S. aureus genomes revealed that combinations of CodY with T-box riboswitches are restricted to methionine and tryptophan biosynthesis in S. aureus and branched-chain amino acid (BCAA) biosynthesis in B. subtilis, respectively [15], [16] (Table 1). Methionine biosynthesis is a costly process consuming ATP and other resources. Riboswitches are generally regarded as fast and tight regulatory systems that do not depend on protein factors which, in many cases, react more slowly and/or are subject to complex regulation of protein expression, activation and degradation [5], [17]. Our study suggests that, in addition to regulation at the transcriptional level, S. aureus employs RNA decay mechanisms to quickly remove newly transcribed metICFE-mdh mRNA from the system, thus further limiting the risk of sustained (over) expression of genes involved in methionine synthesis. Although the complete mechanism and enzymes involved in the process still need to be established, the data give a first hint that RNase J2 participates in metICFE-mdh degradation (Figure 5B). This is in contrast to the situation in B. subtilis where mRNAs of the methionine biosynthesis genes, polyamine synthesis as well as the methionine salvage pathway (see below) were found to be not degraded by RNase J1/J2 [18]. The data further suggest that RNase III might be involved in met leader RNA degradation (Figure 5B). This observation is consistent with previously published data showing that RNase III co-immunoprecipitates with met leader RNA and targets also other S. aureus riboswitches for degradation [19]. Taken together, the data lead us to suggest that RNA decay is another mechanism involved in the control of methionine synthesis in staphylococci that merits further future investigation. T-box control of methionine biosynthesis genes in staphylococci is an exception among Bacillales which usually regulate this pathway by S-adenosylmethionine (SAM) -binding S-box riboswitches [5]. Apart from protein synthesis, most microorganisms use methionine to produce SAM which plays a central role in many cellular functions [4]. First, SAM serves as a methyl group donor for nucleic acid and protein methylation. Products of the methylation reaction are detoxified and recycled to homocysteine which is then reused for methionine/SAM synthesis. Second, SAM is used, following decarboxylation, to form polyamines. The remaining 5′-methylthioadenosine moiety is again metabolized to methionine by enzymes of the methionine salvage pathway. Comparative genomics using the KEGG database (http: //www. genome. jp/kegg/) and experimental research [20], [21], [22] suggest that, unlike Bacillus, staphylococci may have only limited capacity to reuse or redirect methionine to other pathways because they lack both the methionine salvage and polyamine synthesis pathways (Table 1). Therefore, synthesis and recycling of SAM may be the only possibility to make use of excess methionine, implying that a more stringent control of de novo biosynthesis may be required in staphylococci. Interestingly, the Lactobacillales, which preferentially control their methionine biosynthesis genes by T-box riboswitches [5], also appear to lack polyamine synthesis and methionine salvage genes (Table 1). Based on these observations, we hypothesize that the lack of both polyamine synthesis and methionine salvage might favor control by a T-box rather than a S-box riboswitch, the major advantage being that the T-box riboswitch is able to sense methionine supply directly and then react immediately by switching off transcription of methionine biosynthesis genes, whereas S-box regulated systems would require an additional step (i. e. SAM synthesis) to produce the effector molecule required to stop methionine production. Alternatively, microorganisms that produce polyamines may need a larger SAM pool as precursor for the synthesis of these important compounds. Therefore, S-box control of methionine biosynthesis might be more effective in these organisms to ensure a constant SAM supply. More experimental work is needed to further substantiate these hypotheses. The general life style of staphylococci provides easy access to methionine sources from the respective host they colonize or infect and, therefore, suggests that methionine supply may not be a limiting factor under normal conditions. The data presented in this paper provide first insight into the regulation of methionine synthesis gene expression in staphylococci and, more importantly, show that staphylococci have evolved special mechanisms to tightly restrict de novo methionine biosynthesis. It is tempting to speculate that overproduction (rather than lack) of methionine may be critical to staphylococci and, thus, this strict control of methionine de novo synthesis would not only save resources and energy but also meet the requirement to prevent methionine accumulation. While excess methionine biosynthesis might be critical for staphylococci under most conditions, de novo methionine biosynthesis becomes crucial for growth and survival when methionine supply is limited, for example, when entering the host during infection [23] or under specific external stress conditions like antibiotic and antimicrobial peptide exposure [24], [25]. It is therefore conceivable that both the efficient activation and shut-off of methionine biosynthesis might represent a metabolic “Achilles' heel” for staphylococci. Interestingly, successful inhibition of the S. aureus methionyl-tRNA synthetase by an experimental compound has already provided evidence that methionyl-tRNA metabolism is a suitable anti-staphylococcal target [26]. Also, structure-based drug design recently resulted in the identification of lead compounds that specifically interact with T-box structures in vitro, indicating that RNA-based drug targeting is a promising new avenue in medicinal chemistry [27], [28], [29]. The data presented in this paper may open perspectives for specifically targeting methionine metabolism and protein translation initiation in future efforts to develop novel Staphylococcus-specific antibiotics. Total RNA was treated with recombinant DNase I (Ambion) and RNA quality was checked with the Agilent 2100 Bioanalyzer (Agilent Technologies). The 5′/3′ RACE Kit, 2nd Generation (Roche) was used for determination of transcript ends by synthesis of first-strand cDNA with reverse transcriptase (Fermentas) and the oligonucleotides listed in Table S4 (Text S1). The met leader RNA sequence of S. aureus COL was amplified by PCR with oligonucleotides T7-F_met-sRNA and R_met-sRNA (Tables S5, Text S1). The product was inserted into the pGEM-T Easy vector (Promega) yielding plasmid pGEMmetCOL. For site directed mutagenesis, the oligonucleotides listed in Table S5 were used in PCR reactions with pGEMmetCOL as a template followed by DpnI treatment prior to transformation of the PCR products into Escherichia coli DH5α cells. The rescued plasmids were sequenced and constructs SC1 to SC8 (Table S2) with the appropriate mutations were used as met leader RNA templates in tRNA binding assays. tRNA templates were generated by PCR from genomic DNA of S. aureus COL and oligonucleotides listed in Table S4 (Text S1). Four picomol DNA template were subsequently used for IVT in a 20 µl reaction volume consisting of T7 transcription buffer, 20 U RNase inhibitor, 20 U T7 RNA polymerase (Fermentas), 0. 5 mM ATP, UTP, GTP, 12 µM CTP and 9 mM GMP as described [30]. [α32P]-CTP was added and the reaction was incubated at 37°C for six hours. For met leader RNA in vitro transcription, PCR templates were either generated from plasmid pGEMmetCOL or constructs SC1 to SC8 (Table S2) with oligonucleotides T7-F_met-sRNA and R_met-sRNA (Table S5). The products were used in IVT/tRNA binding assays in the presence of pre-formed tRNAs: 10 µl reaction volume consisted of T7 transcription buffer, 10 U RNase inhibitor, 6 U T7 RNA polymerase (Fermentas) and 0. 5 mM NTPs. DNA templates of the met leader RNA and of pre-formed tRNA were adjusted to end concentrations of 8 nM and 50 nM, respectively, and the mix was incubated at 37°C for two hours. Samples were immediately separated on a non-denaturing 6% (w/v) polyacrylamide gel by electrophoresis at 4°C. Visualization was attained with the PhosphoImager (Fujifilm FLA-7000) and quantification of bands was achieved with the software Multi Gauge V2. 2. Bacteria from overnight cultures were diluted in 100-ml flasks in 40 ml CDM medium with methionine to an initial optical density at 600 nm (OD600) of 0. 05 and grown with shaking at 220 rpm at 37°C to an OD of 0. 5. The cultures were filtered over a 0. 22 mm filter applying vacuum, washed twice with sterile phosphate buffered saline (PBS) and bacteria were resuspended in 15 ml CDM medium without methionine and grown for another 60 minutes in a 30-ml tube. Then rifampicin (500 µg ml−1) was added to the cultures. Before (0) and after 0. 5,2, 5,10 and 60 minutes of rifampicin exposure, RNA was isolated and Northern blot analyses were performed as described in the Supporting Information (Text S1).

Prokaryote metabolism is key for our understanding of bacterial virulence and pathogenesis and it is also an area with huge opportunity to identify novel targets for antibiotic drugs. Here, we have addressed the so far poorly characterized regulation of methionine biosynthesis in S. aureus. We demonstrate that methionine biosynthesis control in staphylococci significantly differs from that predicted for other Bacillales. Notably, involvement of a T-box instead of an S-box riboswitch separates staphylococci from other bacteria in the order. We provide, for the first time, direct experimental proof for an interaction of a methionyl-tRNA-specific T-box with its cognate tRNA, and the identification of initiator tRNAifMet as the specific binding partner is an unexpected finding whose exact function in Staphylococcus metabolism remains to be established. The data further suggest that in staphylococci a range of regulatory elements are integrated to form a hierarchical network that elegantly limits costly (excess) methionine biosynthesis and, at the same time, reliably ensures production of the amino acid in a highly selective manner. Our findings open a perspective to exploit methionine biosynthesis and especially its T-box-mediated control as putative target (s) for the development of future anti-staphylococcal therapeutics.

Abstract Introduction Results Discussion Materials and Methods

bacteriology bacterial physiology microbial metabolism gene expression gene regulation genetics molecular genetics staphylococci biology microbiology bacterial pathogens gram positive

2013

Methionine Biosynthesis in Staphylococcus aureus Is Tightly Controlled by a Hierarchical Network Involving an Initiator tRNA-Specific T-box Riboswitch

A growing number of solved protein structures display an elongated structural domain, denoted here as alpha-rod, composed of stacked pairs of anti-parallel alpha-helices. Alpha-rods are flexible and expose a large surface, which makes them suitable for protein interaction. Although most likely originating by tandem duplication of a two-helix unit, their detection using sequence similarity between repeats is poor. Here, we show that alpha-rod repeats can be detected using a neural network. The network detects more repeats than are identified by domain databases using multiple profiles, with a low level of false positives (<10%). We identify alpha-rod repeats in approximately 0. 4% of proteins in eukaryotic genomes. We then investigate the results for all human proteins, identifying alpha-rod repeats for the first time in six protein families, including proteins STAG1-3, SERAC1, and PSMD1-2 & 5. We also characterize a short version of these repeats in eight protein families of Archaeal, Bacterial, and Fungal species. Finally, we demonstrate the utility of these predictions in directing experimental work to demarcate three alpha-rods in huntingtin, a protein mutated in Huntington' s disease. Using yeast two hybrid analysis and an immunoprecipitation technique, we show that the huntingtin fragments containing alpha-rods associate with each other. This is the first definition of domains in huntingtin and the first validation of predicted interactions between fragments of huntingtin, which sets up directions toward functional characterization of this protein. An implementation of the repeat detection algorithm is available as a Web server with a simple graphical output: http: //www. ogic. ca/projects/ard. This can be further visualized using BiasViz, a graphic tool for representation of multiple sequence alignments. Tandems of repeated protein sequences forming structural domains occur in at least 3% of proteins in eukaryotic organisms [1]. Characterization of these repeats by sequence similarity is sometimes difficult as weak evolutionary constraints cause rapid sequence divergence [2]. In particular, repeats including two alpha helices packed together then stacked to form a flexible rod (denoted here alpha-rod) belong to this category (see an example in Figure 1). Some of these alpha-rod repeats have been defined in terms of sequence similarity and are widespread in multiple protein families: HEAT [3], [4], Armadillo [5] and HAT [6]. Others are evident in just one protein family, for example the PFTA repeats [7]. Some, however, bear no statistically significant sequence similarity and may not have originated from sequence duplication (for example, the all-helical VHS domain in Drosophila melanogaster Hrs protein [8], or the subunit H of Saccharomyces cerevisiae vacuolar ATP synthase [9]). This divergence complicates the detection of alpha-rod repeats by methods based on sequence similarity. For example, profile-based methods used in the protein domain databases PFAM [10] and SMART [11] detect only two of the 14 HEAT repeats of human AP-2 complex subunit beta-1 (Figure 1), and might fail to detect any repeats in other alpha-rod containing sequences. Despite the heterogeneity of alpha-rod repeats, they have common features (discussed in [4]): length of about 40 amino acids, anti-parallel alpha-helices, and constraints given by the packing of consecutive repeats. This suggests that alpha-rod repeats are a protein structural feature that obeys some physical constraints irrespective of their evolutionary origin and particular sequence. Coiled coils and transmembrane alpha-helices are other examples of such structural features. Statistical methods have been used to predict coiled coils [12] and transmembrane alpha-helices [13] with excellent reliability, using algorithms that learn to recognize these features from amino acid sequences. In particular, back-propagation neural networks [14] have been used with success to predict secondary structure [15], [16], transmembrane alpha-helices [17], and protein residue solvent accessibility [18]. We hypothesized that a back-propagation neural network could be better suited than homology based methods for the detection of different types of alpha-rod repeats, if trained in an appropriate set of sequences containing these repeats. The last ten years have seen the resolution of a sufficient number of protein 3D structures of sequences with alpha-rod repeats to provide a useful training set for such predictions. The parameters of the method were optimized using the analysis of proteins of known structure. We found that hits above a score of 0. 8 were reliable, especially when the protein had several of them in the appropriate periodicity. Identification of a sequence as containing an alpha-rod was optimal when requiring at least three hits above a score of 0. 8 with a minimum spacing of 30 amino acids between hits and a maximum of 135. Further details can be found in the supplementary Text S1. A total of 87 sequences were selected with this threshold, which can be grouped in 12 protein families of which 8 were not homologous to those used in the training set (Table S2 in Text S1). Since these examples correspond to proteins of known structure, it was easy to visually verify that of those eight families seven were true positives and only one constituted a false positive. Homology of these proteins to the ones used in the training is extremely low or statistically non-significant. Therefore, we concluded that the network was useful in expanding our current knowledge of the occurrences of these repeats and we set to demonstrate this. For simplicity we will denote our methodology as ARD (Alpha-rod Repeat Detection) henceforth. To illustrate the coverage of the method we analyzed the complete protein sets from a series of fully sequenced organisms. The threshold tested in the analysis of PDB was used to select positive sequences. The results of the analysis are in Table 1. The fractions of alpha-rod repeat proteins are around 0. 4% for the nine eukaryotic genomes and lower (0. 05%–0. 21%) in the three prokaryotic organisms tested. No correlation was found between proteome size and fraction of positives. Using ARD we were able to detect protein sequences that PFAM [10] and SMART [11] do not detect or that they detect with multiple profiles (PFAM: Arm, HEAT_PBS and HEAT; SMART: ARM, EZ_HEAT and HEAT). Many of these were not described in the literature. To illustrate the ability of ARD to identify new results we will focus on families with at least one human gene. To illustrate how the method covers various profiles used by SMART and PFAM we will examine results on families with HEAT repeats of the PBS type from fungi, bacteria, and archaea. Finally, we illustrate an experimental application of the method to dissect domains in huntingtin, the protein mutated in Huntington' s disease, for which little is known regarding its structure and function. A total of 86 human proteins were found to contain alpha-rod repeats, which we grouped in 52 families on the basis of their sequence similarity. Of those families, at least 16 have not been yet described to contain alpha-rod repeats in the literature, with 9 undetected by both the SMART and PFAM domain detection web tools (see Table 2). In particular, six families have neither literature nor database repeat assignment; for these, we could verify the repeats using a manually tuned iterative PSIBLAST sequence search [19] of the region with repeats, which showed significant similarity to alpha-rod repeat regions in other protein families. Four of these families encode proteins of unknown function: Serac1, C8orf73, C17orf66, and KIAA0423 (and homolog LOC23116). A fifth family has three members in humans, the stromal antigens 1,2 and 3 (STAG1-3), subunits of the cohesin complex, which mediates cohesion between sister chromatids [20]. In particular, the phosphorylation of STAG2 is essential for cohesin dissociation during prophase and prometaphase [21]. This family has two homologs in Xenopus (demonstrated to form part of two different cohesion complexes [22]), the plant Arabidopsis thaliana (Scc3, needed for the orientation of the kinetochores during meiosis [23]) and yeast (Irr1/Scc3, involved in cell wall integrity [24]). The analysis of the family suggests that their sequences are composed of alpha-rod repeats (Figure 2 and Figure S3A in Text S1). The sixth novel assignment case is the PSMD family (proteasome 26S subunit, non-ATPase) members 1, and 2, and 5. PFAM/SMART identify these as containing repeats of the Proteasome/cyclosome (PC_rep), originally predicted to be composed of a beta strand and a alpha helix [25]. However, ARD predicts 5 repeats which overlap with those. Secondary structure predictions (using JPRED3 [26]) and homology to alpha-rod repeats proposed for PSMD1 yeast homolog Sen3/RPN2 [27] clearly suggest that these are alpha-rod repeats, and that the current PC_rep motif used by PFAM/SMART cuts one of the helices in half. This suggests that the PFAM/SMART domain definition should be revised. Another family for which a redefinition of the PFAM/SMART profile may be required is RRP12, homolog to the yeast Ribosomal RNA processing 12, identified as HEAT-repeat containing, Ran binding, and required for the nuclear export of both the 40S and 60S ribosomal subunits in yeast [28]. SMART and PFAM identify only one HEAT repeat in the human sequence because other repeats overlap with domain NUC173, defined as present in several nucleolar proteins [29], whereas ARD identifies 9 repeats. Three other families remain undetected by PFAM and SMART profiles but have been described to contain alpha-rod repeats in separate publications: these are the MRO (Maestro), which expresses a nucleolar protein of unknown function during male mouse gonad development [30], FRAP1/mTOR, which we described as repeat containing in the first publication defining the HEAT repeats [3] (Figure 2 and Figure S3B in Text S1), and NIPBL (the homolog to Drosophila Nipped-B) related to sister chromatid cohesion yeast proteins Scc2 and Mist4 [31]. For ten other gene families, PFAM and SMART suggest the presence of the repeats but their coverage is more limited than that of ARD and this evidence remains unreported in the literature. This is the case of STK36/FU (the homolog to Drosophila fused, a mediator of sensitivity to PARP [32]), INTS4 (integrator complex subunit 4, which associates with the C-terminal domain of RNA polymerase II large subunit [33]), and of eight hypothetical proteins: C1orf175, LOC165186, HEATR2, HEATR4, HEATR6, KIAA1468, RTDR1 (deleted in rhabdoid tumour), and TMCO7 (which interacts with MACF1, the microtubule-actin crosslinking factor 1 according to a two-hybrid screening [34]). The combination of ARD analyses of the human protein homologs in other organisms, secondary structure prediction and definition of regions of amino acid composition bias facilitates the definition of the boundaries of domains composed of repeats sometimes reused in different domain architectures. Here we present three examples. We found that the LOC165186 and KIAA0423 hypothetical human proteins (mentioned above) define two families whose structured sequence is likely alpha-rods; these two proteins share a C-terminal domain possibly made of more than 10 repeats (Figure 2 and Figure S3C in Text S1). LOC165186, conserved in mammals, has an additional N-terminal composition biased region of around 500 amino acids, whereas KIAA0423, conserved down to worms, has an extra N-terminal domain of alpha-rod repeats connected to the C-terminal repeat domain by a middle linker that is enlarged in the chordate sequences. Human CKAP5/TOG (cytoskeleton associated protein 5), a component of the centrosome that is required for spindle pole assembly [35], has similar-length homologs in mammals, frog, and fly. Analysis of the family identifies five alpha-rods of six repeats each in these sequences and a C-terminal non-repeat containing domain (Figure 2 and Figure S3D in Text S1). The worm homologs are shorter since they have only three of the repeat domains. The structure of one of those domains in Caenorhabditis elegans zyg9 was solved and confirmed the presence of an alpha-rod of six repeats [36]. The CLASP family proteins are microtubule-associated proteins, conserved in animals, fungi, and plants [37]. In humans, there are two homologs, hCLASP1 and hCLASP2, which, similar to CKAP5, associate with the ends of growing microtubules to participate in mitotic spindle formation [38]. Their multiple sequence alignment with homologs suggests that they are formed by four alpha-rods (Figure 2 and Figure S3E in Text S1), also noted in [38]. Other genes previously identified in the literature and by SMART/PFAM are: TBCD (tubulin folding cofactor D) reported by [31]; PSME4/PA200, identified as containing 18 HEAT-like repeats in [39]; BTAF1 (RNA polymerase II, B-TFIID transcription factor-associated, 170 kDa) whose homolog in yeast, Mot1, was noted by [31]; MMS19, involved in nucleotide excision repair and transcription, noted by [40]; huntingtin [3]; both subunits of non-SMC condensin II complex D3 and G2, noted by [31]; and PDS5B/APRIN, a chromatin regulator in hormonal differentiation [41], whose homolog Spo76 in Sordaria macrospore was noted by [31]. The existence of two cases where the evidence of repeats originates from low resolution electron microscopy images deserves special mention. SF3B1 (splicing factor 3b, subunit 1) is proposed to have 22 repeats according to the structure obtained by single-particle electron cryomicroscopy at a resolution of less than 10 angstroms of its complex with splicing factor 3a (SF3B14/P14) where it is shown to coil around SF3B14 [42]. The low resolution electron microscopy structure of the yeast complex of mTOR with KOG1 suggests that KOG1 has a middle alpha-rod domain [41]. We can confirm through ARD analysis that both SF3B1 and KOG1 have alpha-rods in the regions suggested. As noted in the section on analysis of PDB, armadillo repeats are not well detected by ARD and generally PFAM and SMART are as good or better than ARD in recognizing them (for example, for JUP and ARMC8). However, two genes are detected by ARD that are covered by one single PFAM armadillo match and no SMART matches: these are HSPBP1 (hsp70-interacting protein) whose solved 3D structure indicates four armadillo repeats [43] and newly identified RTRD1, for which we detect 3 and 6 repeats, respectively. Finally, of all 52 protein families with human genes we recognized just three false positives: PACS2 (phosphofurin acidic cluster sorting protein 2), OBSCN (obscurin, cytoskeletal calmodulin and titin-interacting RhoGEF), and P2RY9 (purinergic receptor P2Y, G-protein coupled, 8). This was determined by lack of further evidence (no homology to regions with repeats in other families, incompatible secondary structure predictions) combined with a small number of hits in the human sequence, in homologs in other species, or by the overlap of those hits with other domains. In the results of fungal and prokaryotic sequences, we noted a number of cases where the repeats identified for the sequences selected were so similar that it was possible to align most of the repeats by hand in stark contrast to the very divergent examples noted above. We illustrate these with 8 examples, which are not related by homology (see Table S3 in Text S1). Their high percentage of inter-repeat sequence identity is indicative of very recent events of duplication occurring independently in these eight examples. Secondary structure prediction suggests that the structure of the repeat is composed of two helices of ∼10 residues, with a middle loop of three, and an outer loop of ∼10 residues, for a total length of 31–35 aa. Although most of the repeats were identified by SMART and PFAM (EZ_HEAT and HEAT_PBS profiles, respectively), not all repeat instances were marked and some were detected with the alternative HEAT profile. In contrast, ARD identified all obvious repetitions and some additional borderline ones. Orthologs of these eight examples were identified in related taxa (Table S3 in Text S1). The puzzling question remains of why or how these eight apparently unrelated families arose and converged to these short alpha-rod repeats. Whether there are common mechanisms for the duplication and selection of these repeats and for their functions is, at the moment, unclear. The human protein huntingtin is involved in Huntington' s disease. Its function remains unclear [44]. In 1995 we described that huntingtin contains HEAT repeats [3] but their identification was restricted to 10 units covering ∼400 scattered amino acids out of a total sequence length of 3144 amino acids. Since then, no other characteristic structural features have been described for this protein, which complicates its description in terms of separate domains with independent folds and functions. As a result no 3D structure of any fragment of this protein has been yet solved, and although interacting partners of this protein have been found they are mostly restricted to the N-terminal 500 amino acids of the protein [45]. Here, we applied the methodology described above to define alpha-rods in huntingtin and subsequently tested the validity of our predictions experimentally. Initially, we produced an alignment of human huntingtin with a representative set of homologous sequences from the database (provided as supplementary Dataset S2). For this we used not only sequences from protein databases but also sequences derived from ESTs and from genomic fragments. We identified for the first time the existence of huntingtin homologs in worms (nematoda genus Caenorhabditis, and annelida Capitella sp.), amoebae (Naegleria fowleri and Dictyostelium discoideum), sea anemone Nematostella vectensis, and choanoflagellate Monosiga brevicollis, notably expanding the scope of this family. We did not find homologs of huntingtin in fungi. The analysis of human huntingtin by ARD suggests six matches but other low scoring hits are consistently present in homologs. Comparison to biased regions sharply defines two N-terminal domains of six and seven repeats (H1 from amino acid 114 to 413 and H2 from 672 to 969) and suggests the existence of a C-terminal domain of seven repeats (H3 from 2667 to 2938) (Figure 2 and Figure S3F in Text S1). Iterative sequence searches using PSIBLAST with these regions indicated homology to HEAT repeats in otherwise unrelated proteins in the 2nd or 3rd iterations. Consistently, sequence analysis suggested a HEAT-repeat fold (using SVMfold [46]), and threading suggested that those regions adopt a HEAT-repeat fold with high likelihood (using GenTHREADER [47]). The comparative protein structure modeling tool TASSER-Lite [48] produced an alpha-rod for H1 and H2, but an alpha-beta barrel for H3 (incompatible with the predicted secondary structure of the region using JPRED3 [26]). Given secondary structure predictions and scattered matches it is tempting to speculate that other alpha-rods exist outside of the H1, H2, and H3 domains. However, we were unable to obtain consistent results using PSIBLAST or threading for fragments outside these regions. To test our predictions, we produced huntingtin fragments spanning the complete sequence of the protein but separating the predicted alpha-rods into different fragments (Figure 3A) in order to study intra-molecular domain interactions in huntingtin by yeast two hybrid (Y2H) assays (see Methods). Our rationale is that only well defined domains will fold and produce interactions, whereas wrongly defined domains will either not interact or produce nonspecific interactions. We found that the huntingtin fragment Htt507-1230 with the H2 domain self-associates in the Y2H assays. In addition, interactions between Htt507-1230 and Htt1-506Q23 (H1 domain) as well as with the fragment Htt2721-3144 (H3 domain) were observed (Figure 3B). No other interactions were observed. The results obtained with the Y2H assays were also confirmed in mammalian cells using a modified version of the LUMIER method (luminescence-based mammalian interactome mapping technology, [49]). Protein A (PA) -Renilla luciferase- and Firefly-V5 luciferase (Luc) -tagged huntingtin fusion proteins were co-expressed in HEK293 cells and were assessed for the expression of the fusion proteins by immunoblotting and luciferase assays (Figure 3C and 3D). The PA-Renilla-tagged fusion protein is then immunoprecipitated from the soluble cell extracts with IgG coated Dynal magnetic beads. After washing, binding of the Firefly-V5 Luc-tagged fusion protein is quantified by measuring the firefly luciferase activity in a luminescence plate reader. As shown in Figure 3D, interactions between the huntingtin fragments Htt1-506Q23 and Htt507-1230, Htt507-1230 and Htt507-1230, Htt507-1230 and Htt2721-3144 were observed with the assays. Taken together, these experimental results give the first evidence of domains in huntingtin that mediate potential intra- as well as inter-molecular huntingtin interactions. One of many plausible structural assemblies of huntingtin' s domains that are consistent with our results and with those in the literature is discussed in Figure 4. We have developed and applied a neural network for the prediction of alpha-rod repeats. Analysis of the results suggests that it discovers more repeat-containing proteins and repeats per protein than sequence similarity based methods using manually curated profiles, which were previously the best method to detect these repeats. We estimate a level of false positives below 10%: 1 in 12 families in the analysis of PDB (approximately 8%), 3 in 52 families in the analysis of human genes (below 6%). The level of false negatives could be eventually reduced by expanding the training set after new structures of sequences with alpha-rod repeats are solved, but one must be cautious about this to avoid over-prediction. Here, we preferred to train the neural network with a conservative set of known structures to demonstrate that they allow detection of recently identified cases. We consider it very encouraging that the network learned from a small number of examples and generalized to recognize repeats not used in the training, e. g. the shorter PBS lyase repeats, or those found for the first time in six human protein families. Most of the repeats detected correspond to HEAT, PBS, and Armadillo. Whereas the network effectively detected a number of unrelated alpha-rod repeat types, it failed to detect the HAT repeats [6]. Although their length is similar, their structural arrangement in highly parallel helices [50] and the conservation of aromatic residues [51] make them significantly different from HEAT and Armadillo repeats explaining why they cannot be detected by our method. The performance of PFAM, SMART and ARD in predicting each type of alpha-rod repeats in sequences deposited in the PDB database is summarized in Table 3. ARD outperforms PFAM and SMART in the detection of HEAT and PBS repeats but underperforms in the detection of Armadillo repeats (although it identifies some proteins with Armadillo repeats that escape detection by both PFAM and SMART, see Table S2 in Text S1). The proteins in PDB that are currently annotated with HAT repeat regions are detected exclusively by SMART. The lack of a common evolutionary origin for all repeats forming alpha-rods indicates that some specific constraints drive convergent evolution to repeatedly rediscover these repeats as a common solution to a general functional need: protein–protein interactions. Structures of alpha-rods suggest that they are extremely flexible and this allows the ensemble to coil around their target as a boa constrictor would do with its prey. A good example is given by the structure of Exportin Cse1p in complex with Kap60p and RanGTP, where both Cse1p and Kap60p are alpha-rods which wrap around each other, and Cse1p wraps around RanGTP [52]. The necessity to coil around proteins possibly explains why the length of these repeats varies between 30 and 45 amino acids. Shorter repeats might not produce enough interactions between the units to form the rod; consequently the rod would not be stable enough and would unfold too easily. Longer repeats might not produce a rod flexible enough to coil around typical protein targets of diameters in the range of 30 to 50 angstroms. The current data from protein structures and the predictions of protein domains for proteins with alpha-rods (See Table S2 in Text S1) does not suggest the co-occurrence of alpha-rods with other protein domains. We think that this constitutes further evidence that alpha-rods can be used pretty much to bind any protein as needed. Neuwald and Hirano identified in [31] several novel HEAT-repeat containing proteins with functions related to chromosomal organization and microtubule interaction. In agreement with this, here we have identified many alpha-rod repeat containing sequences with related functions, notably direct tubulin binding. A well characterized example is the TOG domain (an alpha-rod of HEAT repeats), which binds tubulin heterodimers to assist addition of tubulin to the plus-end of microtubules [53]; the crystal structure of the TOG domain in Caenorhabditis elegans Zyg9 suggests how this interaction may happen through intra-repeat turns [36]. There is evidence of other microtubule-interacting sequences with alpha-rod repeats: yeast Stu2p binds tubulin [36], clathrin-coated vesicles are assembled along microtubules [54], the protein phosphatase 2A (PP2A) binds to microtubules [55], armadillo-repeat containing sperm antigen 6 (Spag6) colocalizes with microtubules [56] (its homolog in Chlamydomonas reinhardtii is PF16, involved in protein–protein interactions required for microtubule stability and flagellar motility [57]), huntingtin association with microtubules was initially found in vitro [58] and then with the beta subunit of tubulin in vivo [59]. A particular case is the plant specific family Tortifolia1/TOR1/SPR2, first characterized in Arabidopsis thaliana as microtubule-associated protein and containing HEAT repeats [60]. Its N-terminal HEAT repeat domain has been proven to bind to tubulin [61]. Our analysis suggests that this domain possibly contains seven repeats and is distantly related to the CLASP family (data not shown). Several non-plant protozoan sequences (in amoeba Dictyostelium discoideum, and in ciliates Paramecium tetraurelia strain d4-2 and Tetrahymena thermophila SB210) are more similar to the plant family than to distantly related metazoan members hinting at a complex evolution for this family, possibly involving horizontal transfer events between plants and protozoa (data not shown). Other proteins with alpha-rod repeats not known to be directly involved in interaction with microtubules or tubulin have broadly associated functions: excess importin-beta blocks kinetochore-associated microtubule formation and enhances centrosome-associated microtubule formation [62], STAG/Scc3 localizes to the spindle poles during mitosis and interacts with NuMA, a spindle pole-associated factor required for mitotic spindle organization [60]. This evidence further confirms a general function of eukaryotic alpha-rods in the organization of cellular structure, chromosome segregation, vesicular transport, and control of cell division by protein–protein interactions that tend to involve the microtubules if not tubulin subunits directly. We demonstrated how to combine information from homologous proteins and secondary structure predictions for a better definition of domains of repeats. We used this approach to define three domains of alpha-rod repeats in human huntingtin: H1 between positions 114–413, H2 between 672–969, and H3 between 2667–2938 (Figure 3A). The definition of these three domains correlates well with previous definitions of cleavage sites in huntingtin. In striatum of brains from patients of Huntington' s disease a 40–50 kDa N-terminal and a C-terminal 30–50 kDa fragment are observed [63], which would include H1 and H3, respectively. In addition, several caspase cleavage sites have been verified for huntingtin in positions 513,552 and 586 [64], which fall in between predicted H1 and H2 alpha-rods. Using our predictions, we verified for the first time interactions between domains of human huntingtin. These involve three domains of HEAT-repeats. Interactions between domains composed of HEAT-repeats are known. For example, several of the subunits of the AP1 clathrin adaptor core are an alpha-rod of HEAT-repeats and interact with each other [65]. We observed the self-association of one of the huntingtin fragments containing a HEAT-repeat domain. This suggests the possibility that huntingtin homodimerizes through inter-molecular association of this domain, in agreement with previous reports [66]. Homodimerization through interaction of domains with HEAT repeats has been suggested for the DNA-PKc/Ku70/Ku80 complex [67]. The interaction of these domains implies their folding in functional units that correspond to the boundaries we have defined. These results are the first demonstration of domains in huntingtin. This opens avenues for further research into the structure and function of this large protein, which had been hampered until now by its lack of definition in terms of structural units. It is now possible to study the interaction of huntingtin with other proteins on a per domain basis. We have provided a way forward for the description of these elusive repeats that will facilitate the characterization of domains, structures, and eventually functions of a large number of proteins, possibly up to 0. 5% of the proteomes of eukaryotic organisms. Further work is needed to expand the scope of the method, for example to detect HAT repeats and conceivably other as-yet undiscovered alpha-rod repeats. To facilitate the use of the method we have made it available at http: //www. ogic. ca/projects/ard. Results of the analysis of protein families can be studied together using ARD in combination with secondary structure predictions via an updated version of our BiasViz multiple sequence alignment viewer (http: //biasviz. sourceforge. net). We used a neural network of feed-forward type with three layers of neurons [14]. Inputs were obtained by scanning the sequence with a 39 amino acid window. The encoding procedure converts the sequence into a binary string where each amino acid is codified by the binary pattern. The length of the entry layer is 39 times 20, where 20 is the number of possible amino acids. One hidden layer with three neurons is used for connecting the inputs with the output layer containing one neuron predicting whether the window is on a repeat or not (e. g. takes real values from 0. 1 to 0. 9 where the larger values indicates the larger probability of the repeat detection). This architecture was found to be optimal in terms of recall and precision on the training set and computation time required for training and evaluation. Further details of algorithm and training procedure are available in the supplementary Text S1. DNA fragments coding for huntingtin fragments separating predicted domains of alpha-rod repeats were generated by PCR amplification using pAC1-HD plasmid as template. PCR reactions contained, in a 50 µl volume, ∼50 ng plasmid DNA, 15 pmol primer oligonucleotides, 20 mM TRIS-HCl pH 8. 8,2. 5 mM MgCl2,50 mM KCl, 10 mM 2-mercaptoethanol and 2. 5 U Pwo DNA polymerase (Sigma). Fragments were amplified in 30 cycles with the following profile: 60 s denaturation at 94°C followed by 120 s annealing at 45–65°C and 120 s extension at 72°C. Amplified DNA products were isolated from 1. 2% agarose gel and recombined into GATEWAY compatible pDONR221 plasmid (Invitrogen), thus creating the desired entry DNA plasmids. The identity of all PCR products was verified by DNA sequencing. The sequences of the oligonucleotide primers used to generate huntingtin fragments are available at the supplementary Text S1. Recombination of entry vectors with pACT-DM and pBTM116_D9 plasmids was used to create prey and bait plasmid constructs for Y2H interaction mating, respectively. Recombination of different DNA fragments was checked by BsrGI restriction. DNA sequences encoding the huntingtin fragments Htt1-506Q23, Htt507-1230, Htt1223-1941, Htt1934-2666, Htt2536-3144 and Htt2721-3144 were sub-cloned into DNA binding domain (baits) and activation domain (preys) Y2H plasmids using GATEWAY technology (Invitrogen) and a matrix of individual MATa and MATalpha yeast strains was generated for systematic interaction mating [68]. Then, yeast strains expressing bait and prey proteins were mixed in 96-well microtiter plates and diploid yeast strains were formed on YPD agar plates. Y2H interactions were scored by the frequency of appearance on the SDIV agar plates and β-galactosidase activity in SDII and SDIV nylon membranes, respectively. Growth in SDII-agar was monitored as a mating control. Human embryonic kidney HEK293 cells were seeded in 96-well plates and cultured in Dulbecco' s modified Eagle' s medium supplemented with 10% fetal bovine serum at 37°C and 5% CO2. Co-transfection of plasmids was done using Lipofectamine 2000 (Invitrogen) following the manufacturer' s protocol. The analyses were performed after 48 hours of transfection. For immunoblotting and LUMIER assay, cells were lysed at 4°C for 40 min in 100 µl lysis buffer containing 50 mM HEPES-KOH pH = 7. 4,150 mM NaCl, 0. 1% NP40,1. 5 mM MgCl2,1 mM EDTA, 1 mM DTT, 75 Unit/ml Benzonase (Merck) in the presence of protease inhibitor cocktail (Roche Diagnostic). The expression of the constructs was analyzed by Western blot using antibodies against V5-epitope (Invitrogen) and Protein-A (Sigma), while equal protein loading with anti-tubulin antibodies (Figure 3C). For LUMIER assay two vectors were generated based on pCDNA3. 1 (+) (Clontech). For the pPAReni-DM the following cassette was cloned between the BamHI and XbaI sites: Kozak sequence, a double protein A epitope, Renilla Luciferase and the ccdB cassette with flanking R1 and R2 att-sites. For the pFireV5-DM vector the following cassette was cloned between the BamHI and XbaI sites: firefly Luciferase, V5 epitope and the ccdB cassette with flanking R1 and R2 att-sites. (Sequences of cloned inserts are in Supplementary Table S4 in Text S1). Pairs of PA-Renilla and firefly-V5-tagged huntingtin-fragment fusion proteins were co-expressed in HEK293 cells. Cell extracts were prepared and assessed for the expression of the fusion proteins by immunoblotting and luciferase assays. Protein complexes were isolated from 70 µl cell extracts using 5 µl IgG-coated Dynal magnetic beads (Dynabeads M-280 Sheep anti-Rabbit IgG), subsequently washed with 100 µl PBS, and the binding of the firefly-V5-tagged fusion huntingtin fragment (Co-IP) to the PA-Renilla-tagged fusion huntingtin fragment protein was quantified by measuring the firefly luciferase activity in a luminescence plate reader (TECAN Infinite M200). Renilla activity was also measured as a control for PA-Renilla constructs expression and binding (IP, data not shown). Luciferase activity was measured using the Dual-Glo Luciferase Assay System (Promega) and a luminescence plate reader (TECAN Infinite M200). Each experiment was performed as triplicate transfection.

Many proteins have an elongated structural domain formed by a stack of alpha helices (alpha-rod), often found to interact with other proteins. The identification of an alpha-rod in a protein can therefore tell something about both the function and the structure of that protein. Though alpha-rods can be readily identified from the structure of proteins, for the vast majority of known proteins this is unavailable, and we have to use their amino acid sequence. Because alpha-rods have highly variable sequences, commonly used methods of domain identification by sequence similarity have difficulty detecting them. However, alpha-rods do have specific patterns of amino acid properties along their sequences, so we used a computational method based on a neural network to learn these patterns. We illustrate how this method finds novel instances of the domain in proteins from a wide range of organisms. We performed detailed analysis of huntingtin, the protein mutated in Huntington' s chorea, a neurodegenerative disease. The function of huntingtin remains a mystery partially due to the lack of knowledge about its structure. Therefore, we defined three alpha-rods in this protein and experimentally verified how they interact with each other, a novel result that opens new avenues for huntingtin research.

Abstract Introduction Results Discussion Methods

molecular biology/bioinformatics computational biology/protein structure prediction computational biology/macromolecular sequence analysis biochemistry/macromolecular assemblies and machines

2009

Detection of Alpha-Rod Protein Repeats Using a Neural Network and Application to Huntingtin

RNase H enzymes promote genetic stability by degrading aberrant RNA∶DNA hybrids and by removing ribonucleotide monophosphates (rNMPs) that are present in duplex DNA. Here, we report that loss of RNase H2 in yeast is associated with mutations that extend identity between the arms of imperfect inverted repeats (quasi-palindromes or QPs), a mutation type generally attributed to a template switch during DNA synthesis. QP events were detected using frameshift-reversion assays and were only observed under conditions of high transcription. In striking contrast to transcription-associated short deletions that also are detected by these assays, QP events do not require Top1 activity. QP mutation rates are strongly affected by the direction of DNA replication and, in contrast to their elevation in the absence of RNase H2, are reduced when RNase H1 is additionally eliminated. Finally, transcription-associated QP events are limited by components of the nucleotide excision repair pathway and are promoted by translesion synthesis DNA polymerases. We suggest that QP mutations reflect either a transcription-associated perturbation of Okazaki-fragment processing, or the use of a nascent transcript to resume replication following a transcription-replication conflict. RNA∶DNA hybrids exist as normal intermediates during the cellular transactions of transcription and DNA replication. During transcription, a small segment of the nascent RNA is transiently base-paired with the template DNA strand as part of a transcription bubble. More extensive and stable hybrids between RNA transcripts and the DNA template (an R-loop), however, can form under certain conditions (reviewed in [1], [2], [3]). Impediments to the transcription process such as disruption of co-transcriptional mRNA packaging or a failure to remove negative superhelical stress in the transcribed region favor the formation of R-loops. RNA∶DNA hybrids also form during genome replication, especially during synthesis of the lagging strand, which occurs discontinuously as a series of ∼200 nt Okazaki fragments (reviewed in [4], [5]). Each Okazaki fragment is initiated by a complex of DNA polymerase α (Pol α) and primase, which together synthesize primers comprised of ∼10 ribonucleotide monophosphates followed by ∼20 deoxyribonucleotide monophosphates (rNMPs and dNMPs, respectively). Following a polymerase switch to Pol δ, the primary lagging-strand polymerase, primers are extended to complete Okazaki-fragment synthesis. Finally, RNA primers are removed by Pol δ-mediated strand displacement coupled with flap processing. In addition to forming in association with transcription and replication, RNA∶DNA hybrids can also arise from the stochastic incorporation of rNMPs by DNA polymerases. Replicative DNA polymerases are highly selective in discriminating between rNTP and dNTP substrates, but nevertheless utilize rNTPs at a low level during DNA synthesis. In Saccharomyces cerevisiae, it has been estimated that Pol δ and Pol ε incorporate 1 rNMP for every 1,250 or 5,000 dNMPs, respectively [6]. In combination with Pol α, which has lower sugar discrimination than either Pol δ or Pol ε, there likely are >10,000 rNMPs incorporated during each round of haploid genome duplication. Though this generally is assumed to result in single rNMPs embedded in otherwise duplex DNA, tandem rNMPs could also be incorporated. In mammals, rNMPs are found at more than 1,000,000 sites per genome, suggesting that rNMPs are the most abundant type of endogenous DNA “lesion” [7]. RNA∶DNA hybrids are a serious threat to genome stability and are removed by the RNase H class of enzymes, which specifically degrade the RNA component (reviewed in [8]). RNase H1 is a single-subunit enzyme; it requires at least four contiguous rNMPs for cleavage of RNA∶DNA hybrids in vitro [9], [10] and is essential for the replication of mitochondrial DNA [11]. RNase H2 is a three-subunit enzyme that is highly expressed in proliferative mammalian tissues [7] and, in addition to its ability to process R-loops and Okazaki fragments, has the unique ability to incise at a single rNMP within otherwise duplex DNA [10], [12], [13]. Persistent R-loops interfere with replication during mitosis and meiosis, leading to accumulation of DNA breaks, especially at highly transcribed and/or repetitive sequences (reviewed in [1]). A failure to remove lagging-strand primers and complete Okazaki-fragment maturation likewise results in hyper-recombination, chromosomal instability, hyper-sensitivity to DNA damaging agents and reduced cell viability [14]. In the absence of RNase H2, accumulation of single (or a small number of) rNMPs in genomic DNA also leads to cell-cycle defects and chromosomal instability [7], [15]. Though neither RNase H1 nor H2 is essential in yeast, both are required for mammalian cell viability and embryonic development [7], [11]. Significantly, hypomorphic alleles of genes encoding RNase H2 subunits are associated with the hereditary, early-onset neuro-inflammatory disease Aicardi-Goutières syndrome [16]. In addition to gross chromosomal effects, persistent rNMPs in DNA elevate mutagenesis. There is a dramatic increase in deletions occurring within short tandem repeats in yeast strains deficient in RNase H2 [17], [18], [19], and a similar mutation signature is found in highly transcribed sequences [20]. The deletions in both cases require activity of the yeast Type 1B topoisomerase (Top1), an enzyme that relieves transcription-associated torsional stress in DNA (reviewed in [21]). The first step in a Top1-catalyzed reaction is generation of a transient single-strand break in the DNA backbone, with the enzyme linking itself to the 3′ end of the break via a covalent phosphotyrosyl bond. Following rotation of the broken strand around the intact strand, the Top1-generated nick is resealed and the enzyme is released in a self-catalytic reaction with the hydroxyl group on the 5′ side of the break. When the initial Top1 cleavage occurs at an rNMP, however, the 2′ hydroxyl group of the ribose can attack the phosphotyrosyl bond. This results in release of Top1 from DNA and leaves behind a nick flanked by a 2′, 3′ cyclic phosphate and a 5′-OH [22]. We have proposed that when Top1 cleaves at an rNMP, either further processing of the dirty ends thus created or a subsequent cleavage-ligation cycle by Top1 results in short deletions [19]. The significance of the tandem repeat in the deletion process is that it provides an opportunity for strand misalignment, which converts a gapped intermediate into a more efficiently ligated nick. That rNMP-dependent deletions reflect Top1 incision at a single rNMP was recently confirmed using a mutant RNase H2 that retains only the ability to nick at tandem rNMPs [10]. In the current study, we describe a new type of transcription-associated mutation that is greatly elevated in RNase H2-deficient yeast strains. The relevant mutations occur at imperfect inverted repeats or quasi-palindromes and, in striking contrast to deletions in tandem repeats, do not require Top1 activity. Genetic analyses demonstrate a strong effect of the direction of replication fork movement as well as roles for translesion synthesis DNA polymerases and RNase H1 in generating these novel, Top1-independent events. We suggest that, in absence of functional RNase H2, RNase H1 generates potentially mutagenic roadblocks at sites of incompletely processed RNA∶DNA hybrids. In order to obtain a variety of rNMP-dependent mutations, the reversion of three frameshift alleles that query an ∼150 bp sequence was monitored: lys2ΔBgl, lys2ΔA746 and lys2ΔA746, NR. The lys2ΔBgl allele was constructed by filling in BglII-generated 5′ overhangs to generate a 4-bp duplication, the equivalent of a +1 frameshift mutation [23]. Reversion to lysine prototrophy requires acquisition of a compensatory net −1 frameshift mutation within a window defined by stop codons in the alternative reading frames. The lys2ΔA746 allele contains an engineered 1-bp deletion in the same region of LYS2 and requires a compensatory net +1 frameshift mutation for reversion [24]. Finally, the lys2ΔA746, NR allele was derived from the lys2ΔA746 allele by disrupting mononucleotide runs in the reversion window that were longer than 3 bp [25]. For the lys2ΔA746 and lys2ΔA746, NR alleles, transcription was driven by either the endogenous LYS2 promoter or by a tetracycline/doxycycline-repressible promoter (pTET) to achieve low- or high-transcription conditions, respectively. For the lys2ΔBgl allele, the pTET promoter was fully active in the absence of doxycycline or was repressed to a low level by addition of doxycycline to the growth medium. Loss of RNase H2 under low-transcription conditions is associated with only small changes in mutation rate, but can be accompanied by striking changes in mutation spectra [17], [26]. Mutation types that are specifically enriched in the absence of Rnh201, the catalytic subunit of RNase H2, will be designated here as rNMP-dependent mutations. In the lys2ΔBgl assay, RNase H2 loss is associated with reversion via deletion of one copy of an (AGCT) 2 tandem repeat [26]. We previously demonstrated that the rate of this rNMP-dependent, 4-bp deletion (highlighted gray in Figure 1A) is greatly elevated under high-transcription conditions and that it requires the presence of Top1 [18], [19]. Specifically under high-transcription conditions, disruption of RNase H2 also resulted in a 100-fold increase in “complex” events, which are defined as mutations with multiple changes. These events occurred at a discrete hotspot (highlighted yellow in Figure 1A; see Table 1 for rates) and, in striking contrast to 4-bp deletions, their rates were not affected by Top1 loss (Table 1). Figure 1A shows sequence changes in four types of rNMP-dependent complex mutations at the hotspot detected in the lys2ΔBgl assay, each of which replaces all or part of a common GGCC sequence with one or more thymines. Significantly, this region contains two 4-bp inverted repeats (IRs) that together comprise a larger, imperfect IR or quasi-palindrome (QP). On the nontranscribed (coding) strand, the upstream segment of the QP is 5′ TTGGccTTTT (the lowercase letters are spacers between the smaller IRs), and the downstream segment is 5′-AAAACCAA; the two arms of the QP are separated by 3 bp. Each of the complex mutations generated a perfect IR that ranges in size from 5–8 bp, with the most frequent event generating an 8-bp palindrome. Hereafter, the transcription- and rNMP-dependent mutations originally designated as complex events will be referred to as QP mutations. In the lys2ΔA746 assay, RNH201 deletion elevated the reversion rate 1. 8-fold under low-transcription conditions (Table 1; [19]) and 2-bp deletions within a 6A run comprised ∼30% of reversion events (highlighted gray in Figure 1B). Under high-transcription conditions, 2-bp deletions at the 6A hotspot were prominent even in the presence of RNase H2, and were elevated an additional 20-fold in an rnh201Δ background. Like the 4-bp deletions detected in the lys2ΔBgl assay, 2-bp deletions in the 6A run were Top1 dependent. Also observed specifically under high-transcription conditions in the rnh201Δ background were QP mutations that localized to the same region as those in the lys2ΔBgl assay (highlighted yellow in Figure 1B). As in the lys2ΔBgl assay, QP mutation rates in the lys2ΔA746 assay were not reduced when TOP1 was deleted from the rnh201Δ background. In the case of the lys2ΔA746, NR allele, deletion of RNH201 under low-transcription conditions was accompanied by the appearance of a novel 2-bp deletion hotspot in which TC was deleted from the imperfect dinucleotide repeat, TGTCTG (events are highlighted gray in Figure 1C; see Table 1). The TGTCTG repeat is unique to the lys2ΔA746, NR allele and was generated during elimination of a 5T run (TTTTT changed to TgTcT) that overlaps the QP mutation hotspot in the lys2ΔA746 and lys2ΔBgl assays. The TC deletions were elevated an additional 100-fold by high transcription, and like the short deletions in perfect tandem repeats, were dependent on Top1. In addition to TC deletions in the TGTCTG repeat, there were two types of rNMP-dependent +1 frameshift events at the same location, and these were observed only under high-transcription conditions (highlighted yellow in Figure 1C). One involved insertion of a T into the imperfect tandem repeat (TGTCTG changed to TGTCtTG), while the other was comprised of the same T insertion plus a nearby G to A transition (TGTCTG changed to TaTCtTG). Although the primary sequence of the lys2ΔA746, NR allele is slightly different from that of the lys2ΔBgl and lys2ΔA746 alleles, a similar pattern of converting a QP to a perfect palindrome was evident. In this case, the arms of the QP are 5′-TCTGGAT and 5′-ATCCAaGA (the lowercase letter interrupts the palindrome); addition of a T (complementary to the disrupting “a” in the downstream arm) with or without an accompanying base change generates a perfect 8- or 9-bp palindrome, respectively (Figure 1C). As observed in the lys2ΔBgl and lys2ΔA746 assays, QP mutations detected in the lys2ΔA746, NR assay were not affected by Top1 loss (Table 1). In summary, each frameshift reversion assay identified two distinct types of rNMP- and transcription-dependent mutations: Top1-dependent deletions in tandem repeats and Top1-independent events in quasi-palindromes. The potential in vivo substrates of RNase H2 include rNMPs mis-incorporated into duplex DNA [17], the RNA primers at the 5′ ends of Okazaki fragments ([27], [28] and reviewed in [5]) and the RNA component of R-loops [8]. Though RNase H1 can neither incise at single rNMPs embedded in DNA [10] nor substitute for RNase H2 in a reconstituted ribonucleotide excision repair assay [13], it does share the other two types of RNA∶DNA substrates with RNase H2. To ascertain the type of RNA∶DNA hybrid that is relevant to QP mutations, we examined the effect of eliminating RNase H1 in addition to RNase H2. If only one or a few rNMPs is relevant to QP mutations, then disruption of RNase H1 in an rnh201Δ background would be expected to have no effect on these events. If it is multiple rNMPs that are relevant, however, then QP mutations should be further elevated upon additional loss of RNase H1. Unexpectedly, the rates of QP mutations in each frameshift-reversion assay were reduced at least 10-fold in the rnh201Δ rnh1Δ double- relative to the rnh201Δ single-mutant background (Figure 2A). As expected, the rates of Top1-dependent deletions at the tandem-repeat hotspots were not affected by additional deletion of RNH1 in an rnh201Δ background (Figure 2B). The opposing effects of RNase H1 and RNase H2 loss on QP mutations (reducing and enhancing these events, respectively) indicate that extensive RNA∶DNA hybrids, as well the few rNMPs that may remain after their processing, are relevant. Studies using mutant forms of Pol2 that have altered rNTP discrimination suggest that most rNMPs in yeast genomic DNA are incorporated by Pol ε, the leading-strand polymerase. To explore whether Pol ε might be the source of the rNMPs that initiate transcription-associated QP mutations, we introduced the rNTP-restrictive Pol2-M644L protein [6] into an rnh201Δ top1Δ background. Expression of the Pol2-M644L protein had no effect on the rates of QP mutations in either the lys2ΔA746, NR or lys2ΔA746 assay (Tables S1, S2). These data suggest that the rNMPs that trigger QP mutations do not reflect aberrant incorporation by Pol ε, and likely are derived either from Okazaki fragment primers or from R-loops. QP mutations have been proposed to occur via a template switch between the two arms of the QP during DNA synthesis (for reviews, see [29], [30]). In principle, this can involve either an intra- or an inter-molecular switch, each of which is illustrated in Figure 3 for the most common QP event observed in the lys2ΔBgl assay. Intra-molecular switches are thought to occur primarily during lagging-strand synthesis, where the single-strand nature of the template strand may contribute to hairpin formation and promote the initial template switch [31]. In the case of inter-molecular switches, data suggest that the nascent leading strand switches to the lagging-strand template, a switch that similarly may be facilitated by the extensive tracts of single-strand DNA formed between Okazaki fragments [32]. In either case, there are two switches: the first to an alternative template that is used to direct very limited DNA synthesis, and a second switch back to the original template. Because the rate of a given QP mutation is often affected by the direction of DNA replication, it has been suggested that the corresponding intermediate is generated primarily during either leading-or lagging-strand synthesis [32], [33]. In all of the strains used here, the resident LYS2 locus on Chromosome II was deleted and lys2 frameshift alleles were positioned on Chromosome III near the early-firing replication origin ARS306 [34]. When in the SAME orientation, which is the orientation used to generate the data in Table 1 and Figure 1, the lys2 allele is oriented so that transcription from pTET is co-directional with the direction of replication fork movement from ARS306. In this orientation, the non-transcribed strand (NTS) of lys2 is the lagging-strand template during DNA replication (Figure 3A). When in the OPPO orientation, the transcription and replication forks converge and the NTS strand is the leading-strand template (Figure 3B). To determine whether the direction of replication fork movement affects rNMP-dependent QP mutations, we compared their rates in SAME and OPPO strains. The direction of replication had a profound effect on QP mutation rates in each assay; when lys2 alleles were in the SAME orientation, rates were at least 10-fold higher than when in the OPPO orientation (Table 2). As reported previously, the direction of DNA replication had no effect on overall levels of transcription-associated mutagenesis in a WT background, and only a subtle (at most 2-fold) effect on Top1-dependent deletions in tandem repeats (Table 2). The sensitivity of QP mutations to the direction of DNA replication is consistent with strand switching that is triggered preferentially during either leading- or lagging-strand synthesis. Indeed, QP mutations in the E. coli genome are greatly elevated in the absence of mismatch repair (MMR), consistent with temporal connection of these events to replication fork movement [35]. To assess whether transcription-associated QP mutations in yeast are similarly limited by the MMR system, we examined reversion of the lys2ΔA746 and lys2ΔA746, NR alleles in an MMR-deficient (mlh1Δ) background. Loss of Mlh1 had, at most, a 1. 5-fold effect on QP mutations (Tables S1 and S2), suggesting that the majority of the underlying strand switches are unlikely to occur directly at the replication fork. Transcription-associated mutations that arise in the presence of RNase H2 are elevated when excision repair pathways are disabled and are dependent on the Pol ζ translesion synthesis (TLS) DNA polymerase, suggesting that most arise during the error-prone bypass of DNA damage [34], [36]. To ascertain whether DNA damage is relevant to rNMP-dependent QP mutations, we examined the effect of disabling the nucleotide excision repair (NER) pathway. Rad1, the homolog of mammalian XPF, forms a stable complex with Rad10. Rad1–Rad10 cleaves at the junction between double- and single-strand DNA, and makes one of the incisions that initiates removal of a damage-containing oligonucleotide during NER (reviewed in [37]). This complex additionally has a flap endonuclease activity that removes 3′ single-strand tails that arise during recombination in yeast [38] and has been implicated in repair of hairpin structures formed by CAG trinucleotide repeats [39]. Deletion of RAD1 in an rnh201Δ top1Δ background was associated with an ∼10-fold increase in QP mutations in the each of the frameshift reversion assays (Figure 4A). The rates of Top1-dependent 2–4 bp deletions were not significantly affected by Rad1 loss (Tables S4, S5, S6), however, confirming a fundamentally different mechanism for generating rNMP-dependent QP versus short-deletion events. To discern whether the effect of Rad1 loss on QP mutations reflects the role of Rad1–Rad10 during NER, we examined the effect of disrupting Rad14, the homolog of mammalian XPA. Rad14 is required for the damage recognition step of NER and its loss eliminates NER in yeast, but not other functions of the Rad1–Rad10 complex. For the QP mutations identified in the lys2ΔA746 and lys2ΔBgl assays, there was no discernable difference between the rates of these events in the rnh201Δ top1Δ rad14Δ versus rnh201Δ top1Δ rad1Δ strains (Figure 4A). In the lys2ΔA746, NR assay, however, the rate of QP mutations in the rnh201Δ top1Δ rad14Δ background was only 2. 2-fold higher than in the rnh201Δ top1Δ background, significantly less than the 9-fold increase observed in the rnh201Δ top1Δ rad1Δ background. The QP events identified in the lys2ΔA746 and lys2ΔBgl assays are thus limited by NER, while those detected in the lys2ΔA746, NR assay reflect primarily an NER-independent role of Rad1–Rad10. Prior analyses of transcription-associated mutagenesis in the lys2ΔA746 reversion assay suggest that most events reflect DNA-damage bypass by the Pol ζ TLS DNA polymerase [34], [40]. To examine whether rNMP-dependent QP mutations similarly require Pol ζ activity, we eliminated the Rev3 catalytic subunit of the enzyme in rnh201Δ top1Δ strains. In an rnh201Δ top1Δ rev3Δ background, the rate of QP mutations in the lys2ΔA746, NR or lys2ΔBgl assay was reduced ∼2-fold, while that in the lys2ΔA746 assay was not significantly affected (Figure 4B). Given the small effect of Pol ζ loss, we examined whether Pol η, the other major TLS polymerase in yeast, might be relevant to rNMP-dependent QP mutations. In the lys2ΔA746, NR assay, deletion of the Pol η-encoding RAD30 gene was accompanied by a 50-fold reduction in the QP mutation rate. In the lys2ΔA746 and lys2ΔBgl assays, rates of QP mutations rates were reduced only ∼3-fold in a rad30Δ background. QP mutations detected using the lys2ΔA746 allele, however, were completely eliminated upon loss of both Pol ζ and Pol η (rnh201Δ top1Δ rev3Δ rad30Δ mutant; reversion of the lys2ΔBgl allele was not examined in this background). Altogether, the data indicate that generation of rNMP-dependent QP mutations requires the participation of Pol ζ and/or Pol η, with specific strand-switch events relying on individual TLS polymerases to different extents. Again, this is in contrast to the rates of rNMP- and Top1-dependent deletions in tandem repeats, which are not affected by loss of TLS polymerases. We previously reported two types of mutations that are specifically elevated by high levels of transcription in yeast. First, removal of uracil that replaces thymine in transcriptionally active DNA creates abasic sites that are bypassed in a mutagenic manner by Pol ζ, resulting in elevated TA to GC transversions [36], [40]. Second, small deletions in tandem repeats reflect Top1 activity, which is important for removing transcription-associated supercoils [19], [20]. Here, we report a third transcription-associated mutation signature comprised of complex sequence changes that extend the complementarity between two arms of a quasi-palindrome (QP). Transcription-associated QP mutations are greatly elevated in an RNase H2-defective (rnh201Δ) background and, therefore, are associated with the persistence of some type of RNA∶DNA hybrid. Though both transcription-associated short deletions and QP mutations are elevated in an rnh201Δ background, other genetic requirements for these two types of events are distinctly different (summarized in Table 3). Most notably, rNMP-dependent QP mutations are affected by the direction of replication-fork movement and require RNase H1 rather than Top1 activity. In addition, only QP mutations are limited by NER and are dependent on TLS polymerases. Complex sequence changes at QPs have been most extensively characterized in E. coli (reviewed in [30]), but also have been identified in bacteriophage T4 [41] and are found among TP53 gene mutations in human cancer cells [42]. A template-switch model for QP mutations was first proposed to explain unusual frameshift mutations in the yeast iso-1-cytochrome c gene [43], and we suggest that a similar mechanism is responsible for transcription-associated QP mutations. More recently, QP mutations in yeast have been shown to occur during the repair of DNA double-strand breaks via homologous recombination [44], [45]. Homologous recombination cannot be the source of the transcription-associated QP events observed here, however, as these events are not affected by loss of Rad52 (see Tables S1, S2), a protein essential for recombination in yeast (reviewed in [46]). In bacterial studies, a strong effect of the direction of replication fork movement on a given QP mutation has been interpreted as evidence that the event primarily reflects either an intra-strand switch during lagging-strand synthesis or a transient inter-strand switch of a nascent leading strand to the lagging-strand template (see Figure 3). In either case, it is assumed that the switch is facilitated by the inherent single-strand character of the lagging-strand template and is precipitated by a block/pause during DNA synthesis. For example, there is a positive correlation between intrinsic DNA polymerase pause sites in vitro and QP-associated mutation hotspots in phage T4 [47]. In addition, QP-mediated mutations, but not simple base substitutions or frameshifts, are strongly stimulated by the nucleoside analog 3′-azidothymidine (AZT) in E. coli, which causes chain termination and replication stalling when incorporated by DNA polymerases [48]. Because transcription-associated QP mutation rates were much higher in the SAME than OPPO orientation, these events most likely reflect a problem encountered during lagging-strand synthesis. As will elaborated further below, the sensitivity of QP mutations to the direction of replication-fork movement may additionally reflect an orientation-dependent transcription-replication conflict. In spite of the strong effect of replication direction on transcription-associated QP mutations, these events were not further stimulated by MMR loss and thus are unlikely to occur directly at the fork. Though this is in contrast to the removal of some types of QP mutation intermediates by MMR in E. coli [35], we note a similar MMR independence in the specific case of transcription-associated QP mutations [49]. One possible explanation for the MMR independence of transcription-associated QP mutations is that they arise during the filling of gaps that are left behind the fork when processive DNA synthesis is disrupted. Requirements of the yeast TLS polymerases for transcription-associated QP mutations (Figure 4B) would be consistent with such a post-replicative, gap-filling process. Indeed, it has been shown that other types of mutation intermediates introduced by Pol ζ are refractory to correction by the MMR machinery, and hence presumably arise outside the context of a replication fork [25]. Roles of the TLS polymerases during the formation of QP mutations could be in the relatively nonprocessive synthesis required for successive template switches and/or in extending mispaired primer-template termini that arise as a result of a template switch. Pol ζ, for example, is particularly good at extending primer-template mispairs [50]. In contrast to their strong stimulation upon RNase H2 loss, transcription-associated QP mutations are dependent on the presence of RNase H1. RNase H1 requires at least 3–4 contiguous rNMPs to incise the RNA component of RNA∶DNA hybrids, and we consider three possible sources of such hybrids. First, a DNA polymerase could mis-insert consecutive rNMPs during DNA synthesis. Though the major source of stochastic rNMP incorporation into the yeast genome appears to be due to Pol ε [6], expression of a mutant form of Pol ε with a reduced propensity to incorporate rNMPs did not significantly lower the rate of QP mutations. An alternative possibility that we cannot exclude is that one of the yeast TLS polymerases, each of which contributes to QP mutations (Figure 4), might be the source of the consecutive rNMPs that RNase H1 acts on. E. coli Pol V, for example, which is in the same polymerase family as yeast Pol η, is characterized by low discrimination between dNTPs and rNTPs [51]. A second source of more extended RNA∶DNA hybrids is the RNA primers used to initiate Okazaki fragments during lagging-strand synthesis. The major pathways for removing RNA primers appear to involve either the successive cleavage of short, Pol δ-displaced 5′ flaps by the Fen1 flap endonuclease or the cleavage of longer strand-displacement flaps by the combined action of Dna2 and Fen1 [5]. This does not exclude, however, an important role for RNases H1 and H2 in Okazaki primer removal in some circumstances, and a specific case relevant to QP mutations might be under high-transcription conditions. Finally, R-loops that form during transcription provide a third source of RNA∶DNA hybrids that are potential substrates of RNase H1. The opposing roles of RNase H1 and H2 in the generation of QP mutations (a reduction and increase upon deletion of RNH1 and RNH201, respectively) suggest that RNase H1 likely degrades most of the RNA component of the relevant RNA∶DNA hybrid, leaving a small number of rNMPs that can only be removed by RNase H2. It is then the persistence of one or a few rNMPs that is assumed to perturb lagging-strand synthesis and eventually precipitate the intra-strand template switch that leads to QP mutations. We note that the type of template switch proposed here would only allow a small number of blocking rNMPs to be bypassed. Longer RNA-DNA hybrids presumably lead to a different type of instability that cannot be detected in our reversion assays. A final feature of the transcription-associated QP mutations described here is their elevation upon loss of the NER pathway, suggesting that removal of some type of helical distortion or secondary structure limits these events. Helical distortion could reflect DNA damage or the presence of consecutive rNMPs in duplex DNA [52], while the hairpin formed during an intra-strand template switch might be a relevant secondary structure. In addition to a general role for NER in preventing QP mutations, an NER-independent role for Rad1 was evident in the lys2ΔA746, NR assay. We speculate that the NER independence reflects the well-known role of Rad1–Rad10 in removing 3′ flaps [38], which may be analogous to the role of 3′ to 5′ single-strand exonucleases in preventing QP mutations in E. coli [35]. Whether other endo- or exo-nucleases are relevant to QP mutations in yeast remains to be explored. Based on the genetic requirements of transcription-associated QP mutations, we entertain two different models for how these events might arise in yeast (Figure 5). Central to both models is an involvement of an RNA∶DNA hybrid and an effect of the direction of replication-fork movement. In the first model, replication stalling and subsequent strand misalignment within a QP reflects a deficiency in Okazaki fragment maturation in the absence of RNase H2 (Figure 5A). This would be consistent with the presumptive origin of QP mutations during lagging-strand synthesis and with the apparent site-specificity of these mutations, which could reflect the position of an Okazaki fragment-priming site. There are, for example, additional QPs in the reversion window monitored that should be able to template complex mutations, but corresponding QP mutations were not detected at these positions. The involvement of NER could reflect processing of a hairpin intermediate formed during the template switch, while an NER-independent role of Rad1–10 could reflect removal of unpaired 3′ ends during an attempted template switch [35]. A prediction of this model is that QP mutations should be exacerbated by additional perturbation in Okazaki fragment synthesis and/or processing. A second model (Figure 5B) is suggested by in vitro studies of replication-transcription conflicts using bacterial proteins, in which co-directional collisions disrupt both DNA and RNA synthesis. Significantly, the transcript can be used as a primer to re-start leading-strand replication [53]. Recent studies in E. coli have corroborated these findings and suggest that co-directional collisions with backtracked/arrested RNA polymerase are particularly problematic [54]. Following replication re-start, a failure to properly process the primer would lead to transcript-derived rNMPs in the lagging-strand template at the next round of replication. rNMP-provoked replication stalling could trigger formation of an rNMP-containing gap that would need to be filled post-replicatively, leading to the observed QP mutations. For yeast DNA polymerases, the efficiency of rNMP bypass varies depending on the sequence context [55], and this might account for the site-specificity of events. With regard to the role of NER in limiting QP mutations, pausing/stalling of the transcription machinery at damage that is normally removed by NER would be expected to increase co-directional collisions. More frequent collisions should also be promoted by mutations that slow the elongation rate of RNA polymerase or that eliminate anti-backtracking mechanisms. This model could be tested by examining the effect of eliminating transcription elongation factor TFIIS, which also reactivates backtracked RNA polymerase [56]. Ribonucleotides are likely the most prevalent, non-canonical component of genomic DNA and can lead to catastrophic events in the absence of efficient removal [7]. In addition to Top1-generated strand breaks at rNMPs, which lead to short deletions [18] and replication stress [57], the QP mutations described here underscore the importance of efficient rNMP removal by RNase H2 for the maintenance of genome integrity. The more extensive palindromes generated by QP mutations can fold into stable secondary structures such as hairpins and cruciforms, which would further increase the risk of genome instability. Additional studies will be required to understand the molecular mechanism whereby transcription perturbs the processing and/or promotes formation of RNA∶DNA hybrids that lead to QP mutations in yeast. Yeast strains were grown at 30°C in non-selective YEP medium (1% yeast extract, 2% peptone; 2% agar for plates) supplemented with 2% dextrose (YEPD) or with 2% each of glycerol and ethanol (YEPGE). pTET was maximally activated under these growth conditions, and was repressed by addition of doxycycline (Sigma) to 2 µg/ml. Selective growth was on synthetic, 2% dextrose (SD) medium supplemented with all but one nutrient. Selection of hygromycin- or nourseothricin-resistant transformants was on YEPD plates containing 300 µg/ml hygromycin (Mediatech) or 100 µg/ml nourseothricin (Axxora), respectively. Yeast strains were derived from YPH45 (MATα ura3-52 ade2-101oc trp1Δ1). Deletion of the LYS2 locus and introduction of the pTET-lys2ΔA746, pTET-lys2ΔBgl, or pTET-lys2ΔA746, NR allele near ARS306 on Chromosome III was previously described [18], [20], [34]. The pol2-M644L mutation was introduced by two-step allele replacement using plasmid p173 [17]. Genes were deleted using PCR-generated cassettes containing a selectable marker flanked by ∼60 bp of homology. As appropriate, marker genes flanked by loxP sites were deleted using a Cre-expressing plasmid [58]. One-ml YEPGE cultures were inoculated with 250,000 cells and grown for 3 days. Appropriate dilutions were plated on YEPD or SD-Lys plates to determine the total number of cells or the number of Lys+ revertants, respectively, in each culture. Data from 10–24 cultures were used to calculate each mutation rate using the method of the median [59], and corresponding 95% confidence intervals were determined as described previously [60]. To construct mutation spectra, we isolated independent Lys+ revertants from 0. 3-ml YEPGE cultures inoculated from single colonies. Following isolation of genomic DNA, the relevant region of LYS2 was PCR amplified and sequenced by the Duke University DNA Analysis Facility.

Mutation rates are correlated with the level of gene expression in budding yeast, demonstrating a link between transcription and stability of the underlying DNA template. In the current work, we describe a novel type of transcription-associated mutation that converts imperfect inverted repeats (quasi-palindromes or QPs) to perfect inverted repeats. Using appropriate mutation reporters, we demonstrate that QP mutations are strongly affected by the direction of DNA replication and have distinctive genetic requirements. Most notably, rates of transcription-associated QP events are regulated by the RNase H class of enzymes, which are specialized to process the RNA component of RNA∶DNA hybrids. The source of the RNA∶DNA hybrids that initiate QP mutations is unclear, but could reflect transcripts that remain stably base-paired with the DNA template, or aberrant processing of the RNA primers normally used to initiate DNA synthesis. These studies further expand the diverse ways that transcription affects the mutation landscape, and establish a novel way that RNA∶DNA hybrids can contribute to genetic instability. The high conservation of basic DNA-related metabolic processes suggests that results in yeast will be broadly applicable in higher eukaryotes.

Abstract Introduction Results Discussion Materials and Methods

2013

RNA∶DNA Hybrids Initiate Quasi-Palindrome-Associated Mutations in Highly Transcribed Yeast DNA

Toll-like receptors (TLRs) are important regulators of the innate immune response to pathogens, including Mycobacterium leprae, which is recognized by TLR1/2 heterodimers. We previously identified a transmembrane domain polymorphism, TLR1_T1805G, that encodes an isoleucine to serine substitution and is associated with impaired signaling. We hypothesized that this TLR1 SNP regulates the innate immune response and susceptibility to leprosy. In HEK293 cells transfected with the 1805T or 1805G variant and stimulated with extracts of M. leprae, NF-κB activity was impaired in cells with the 1805G polymorphism. We next stimulated PBMCs from individuals with different genotypes for this SNP and found that 1805GG individuals had significantly reduced cytokine responses to both whole irradiated M. leprae and cell wall extracts. To investigate whether TLR1 variation is associated with clinical presentations of leprosy or leprosy immune reactions, we examined 933 Nepalese leprosy patients, including 238 with reversal reaction (RR), an immune reaction characterized by a Th1 T cell cytokine response. We found that the 1805G allele was associated with protection from RR with an odds ratio (OR) of 0. 51 (95% CI 0. 29–0. 87, p = 0. 01). Individuals with 1805 genotypes GG or TG also had a reduced risk of RR in comparison to genotype TT with an OR of 0. 55 (95% CI 0. 31–0. 97, p = 0. 04). To our knowledge, this is the first association of TLR1 with a Th1-mediated immune response. Our findings suggest that TLR1 deficiency influences adaptive immunity during leprosy infection to affect clinical manifestations such as nerve damage and disability. Mycobacterium tuberculosis (MTb) has established latent infection in one-third of the world' s population and, among those with progressive disease, causes about 2 million deaths per year [1]. Mycobacterium leprae (ML), a related organism, is the etiologic agent of leprosy, an ancient scourge that still causes illness in several regions of the world [2]. Both MTb and ML produce a spectrum of illness in their hosts, yet, aside from frank immunodeficiency, the host factors that underlie the various clinical manifestations of MTb and ML infection are largely unknown. One possible explanation for this diversity of outcomes is common, subclinical variation in host defense genes. Several lines of evidence suggest that genetic factors influence susceptibility to leprosy and other Mycobacteria [1]–[5]. Rare individuals with primary immunodeficiency syndromes are highly susceptible to certain mycobacterial species due to Mendelian disorders associated with highly penetrant phenotypes [2]. However, in most individuals, susceptibility to mycobacterial infection is associated with complex inheritance patterns that are determined by the combined effects of variation across many genes, with a modest contribution from each polymorphism. Evidence that commonly inherited gene variants influence susceptibility to mycobacterial infection comes from twin studies, genome-wide linkage studies, and candidate gene association studies [2]. Studies of leprosy infection in twins have shown a three-fold greater concordance for type of leprosy disease in monozygotic compared to dizygotic twins [6]. Genome-wide linkage studies have identified two single nucleotide polymorphisms (SNPs) in the shared promoter region of the PARK2 and the PARCG gene, several HLA-DR2 alleles, and a non-HLA region near chromosome 10p13 that are associated with leprosy or leprosy subtypes [2], [7], [8]. Candidate gene association studies have also shown associations between leprosy and polymorphisms in several genes, including lymphotoxin-a (LTA) [9], the vitamin D receptor [10], TNF-α [11], laminin-2 [12], and mannose binding lectin [13], [14]. Human infection with M. leprae presents a unique opportunity to link innate and adaptive immune responses to host genetic factors. Leprosy' s divergent clinical forms reflect two distinct immune responses to the same pathogen. Lepromatous leprosy (defined as polar lepromatous (LL) or borderline lepromatous (BL) ) is characterized by a Th2 immune response and poor containment of the infection. At the opposite pole, tuberculoid leprosy (defined as polar tuberculoid (TT) or borderline tuberculoid (BT) ) features a Th1 cytokine response, vigorous T cell responses to ML antigen, and containment of the infection in well-formed granulomas [15], [16]. Reversal reactions (RR) represent the sudden activation of a Th1 inflammatory response to ML antigens. They often occur after the initiation of treatment in patients towards the lepromatous pole of the leprosy spectrum (LL, BL, or borderline borderline (BB) categories) and reflect a switch from a Th2-predominant cytokine response toward a Th1-predominant response [15], [16]. Risk factors for RR intrinsic to the host include age [17] and gene variants, although the latter have not been intensively investigated [18], [19]. Toll-like receptors (TLRs) are a family of highly conserved, type 1 transmembrane proteins that orchestrate the innate immune response to microbial motifs, also known as pathogen associated molecular patterns (PAMPs) [20]–[22]. The TLR pathway regulates the innate immune response to mycobacteria through several TLRs, including TLR1,2, 4,6, and 9 [23]–[26]. TLR2 (Online Mendelian Inheritancce in Man (OMIM): 603028), as a heterodimer with TLR1 (OMIM: 601194) or TLR6 (OMIM: 605403), mediates recognition of several mycobacterial motifs, including lipopeptides, the 19 kDa protein, lipoarabinomannan (LAM) [1], [27], [28]. Interaction of these ligands with the extracellular domain of TLRs leads to activation of a signaling pathway, which results in expression of chemokines and cytokines [20]. Functional work by many investigators has shown that TLR2 is a critical mediator of the innate immune response to ML and MTb [29], [30]. In addition, several TLR2 polymorphisms have been reported to be associated with susceptibility to MTb [31]–[33]. By contrast, very little is known about the effect of TLR1 variation on the innate response to mycobacteria or clinical susceptibility to mycobacterial disease. It also remains controversial whether and how the innate immune response mediated by any individual TLR shapes adaptive immunity [34]–[37]. We recently characterized a non synonymous SNP, T1805G (I602S), in the transmembrane domain of TLR1 that regulates signaling in response to PAM3, a synthetic ligand of TLR1 [38]. Johnson et al. also found that this polymorphism was associated with decreased signaling as well as protection from leprosy in Turkey [39]. Intriguingly, it appears that the TLR1 signaling defect is due to a complete absence of TLR1 on the surface of monocytes in GG individuals [39]. Here, we investigate an association of this SNP with different clinical forms of leprosy in Nepal and examine the effect of this SNP on leukocyte signaling in response to ML stimulation. RPMI Medium 1640, L-glutamine, penicillin-streptomycin, and DMEM were from GIBCO/Invitrogen (Carlsbad, CA). Ultrapure lipopolysaccharide (LPS) was from Salmonella minnesota R595 (List Biological Labs, Inc.). Lipopeptides PAM2Cys-SKKKK (diacylated, PAM2) and PAM3Cys-SKKK (triacylated, PAM3) were from EMC Microcollections (Tuebingen, Germany). Macrophage-activating Lipopeptide-2 S-[2,3-bis (Palmityloxy) - (2R) -propyl-cysteinyl-GNNDESNISFKEK] (diacylated lipopeptide from Mycoplasma fermentens, Malp-2) was obtained from Alexis Biochemicals (Lausen, Switzerland). M. leprae reagents were obtained from J. Spencer (Colorado State University) through NIH, NIAID Contract No. NO1-AI-25469, entitled “Leprosy Research Support. ” HEK293 cells (ATCC#CRL-1573) were grown in DMEM (GIBCO cat. no. #11995), supplemented with 10% fetal bovine serum, 10 units/ml penicillin, and 10 µg/ml streptomycin. Study participants in Seattle were healthy adults with no known history of unusual susceptibility to infections [40]. Study participants in Nepal included 933 leprosy patients referred for treatment at Anandaban Hospital in Katmandu, Nepal and later recruited to a study of genetic factors influencing susceptibility to reactional episodes in leprosy. The study population comprised more than 8 different ethnic and religious groups that included Brahmin (25. 6%), Chhetri (22. 3%), Tamang (14. 3%), Newar (7. 3%), Magar (5. 4%), Muslim (3. 3%), Sarki (3. 5%), and Kami (2. 7%), with 15. 5% having unrecorded ethnicity A diagnosis of leprosy and determination of leprosy type was made by clinical symptoms, skin smears and biopsy reports. Assignment of leprosy category followed the Ridley/Jopling classification scheme [41]. Each patient had a minimum of three years of regular clinic follow-up prior to recruitment. In accord with guidelines of the US Department of Health and Human Services, protocols were approved by the Nepal Health Research Council, the University of Washington, the University of Medicine and Dentistry of New Jersey, and the Western Institutional Review Board. Written informed consent was obtained from all patients or from their relatives if the patient could not provide consent. DNA from subjects in Nepal was obtained by extraction from whole blood using Nucleon BACC2 Genomic DNA (Amersham Lifesciences) and Roche High-Pure PCR template preparation extraction kits. DNA from subjects in Seattle was extracted from whole blood using QIAamp DNA Blood Midi kits (Qiagen, Valenica, CA). Genotyping was carried out with a MassARRAY technique (Sequenom) as previously described [42], [43]. For functional studies, the coding region of TLR1 was amplified from genomic DNA and cloned into the pEF6/V5-His-TOPO vector (Invitrogen, Carlsbad, CA) as previously described [38]. To obtain the polymorphic variants of TLR1, a whole plasmid PCR strategy with mutant primers was used as previously described [30]. PBMCs were derived from whole blood separated by centrifugation on a Ficoll-Hypaque gradient, plated at a density of 1×105 cells per well in 96-well plates in RPMI (supplemented with 10% fetal bovine serum), and incubated overnight. PBMC cytokine assays were then performed by stimulating with various TLR ligands or extracts of M. leprae for 18 hours. Each sample was assayed in triplicate. Cytokine levels were determined with a sandwich ELISA technique (Duoset, R&D Systems, Minneapolis, MN) or with a multiplex kit for the luminex platform (Human Fluorokine MAP Base Kit, Panel A, R&D systems, Minneapolis, MN). Levels of contaminating LPS as determined by the chromogenic Limulus amebocyte lysate test (Cambrex, MD) were 0. 05–0. 27, and 0. 03–0. 19 endotoxin units/ml, in wells incubated with whole irradiated ML (ML) and ML cell wall (MLcw), respectively, depending on the dose of reagent used. These values correspond to 4. 5–27. 2 and 3. 04–19. 4 pg/ml of endotoxin, respectively, in wells treated with whole irradiated ML and ML cw. All wells that received ML reagents were additionally treated with polymyxin B at a concentration of 10 µg/ml. HEK293 cells were transfected with Polyfect (Qiagen, Hilden, Germany) per the manufacturer' s instructions with 2–5×104 cells per well in a 96-well plate with pRL-TK (to control for transfection efficiency), ELAM-luciferase (NF-κB reporter), one of two TLR1 variants, TLR2, and CD14. After an overnight transfection, cells were stimulated with TLR ligands or extracts of ML for 4–6 hours, and then lysed and processed for luciferase readings per the manufacturer' s instructions for the Dual Luciferase Reporter Assay System (Promega, Madison, WI). Univariate analysis was performed for categorical variables with a Chi-Square test; Fisher' s exact test was used when the number of samples in a group was less than 5. The Mann-Whitney U-test was used to make comparisons of the cytokine production between groups, as small sample sizes precluded an assumption of normal distribution. Student' s t-test was used to compare results in the luciferase assay. Two-sided testing was used for all comparisons to evaluate statistical significance. A P value (p) of ≤0. 05 was considered significant. Statistics were calculated with Prism version 4. 03software. For genetic analysis, allelic, genotypic, and haplotypic frequencies were compared between groups. Haplotypes were constructed with an Expectation/Maximization (EM) algorithm with the program HAPIPF in IC Stata (version 10. 0) [44]. Except for minor deviations, the observed allelic frequencies of SNPs were consistent with expected frequencies under Hardy-Weinberg equilibrium. We investigated the effect of SNP T1805G (I602S) on NF-κB responses to ML in HEK293 cells transfected with 1805T (602I) or 1805G (602S), firefly luciferase conjugated to an NF-κB promoter (ELAM), TLR2, CD14, and Renilla luciferase conjugated to thymidine kinase to control for transfection efficiency [38]. Cells were then stimulated with media, ML extracts or TLR ligands: PAM3, a ligand for the TLR2/1 heterodimer or Malp-2, a ligand for the TLR2/6 heterodimer (Fig 1). HEK293 cells transfected with TLR2+1805T and stimulated with 50 µg/ml of whole, irradiated ML had significantly greater NF-κB activity than HEK293 cells transfected with TLR2 alone (600. 7 vs. 159. 8 relative luciferase units (RLU), p = 0. 001), or cells transfected with TLR2+1805G (600. 7 vs. 157. 1 RLU, p = 0. 000004) (Fig 1). Responses to ML were dose-dependent in cells transfected with either TLR1 variant and the signaling difference between 1805T and 1805G transfectants persisted over a range of doses (comparison for ML 5 µg/ml: 511. 9 vs. 75. 8, p<0. 0001; comparison for ML 250 µg/ml: 673. 7 vs. 262. 8, p = 0. 0005). We then investigated whether T1805G influenced signaling in response to MLcw, which contains lipopeptide moieties known to stimulate through TLR2/1 [29]. TLR2+1805T-transfected cells stimulated with 1 or 10 µg/ml of MLcw had significantly greater NF-κB activity than cells transfected with TLR2+1805G (Fig 1, comparison for MLcw 1 µg/ml: 444. 4 vs. 53. 8 RLU, p = 0. 00005; comparison for MLcw 10 µg/ml: 562. 8 vs. 238. 1 RLU, p = 0. 004). As a control, we also compared baseline signalling activity and response to tri-acylated lipopeptide (PAM3) in the two 1805 variants. Consistent with previous observations [38], the 1805T variant, when co-transfected with TLR2, mediated greater constitutive NF-κB activity compared to TLR2 alone (Fig 1, stimulation with media alone: 329. 7 vs. 3. 9 RLU, p = 0. 000003). The TLR2+1805T transfectants were also readily distinguished from the TLR2+1805G transfectants by a significantly higher level of basal signaling (329. 7 vs. 24. 2 RLU, p = 0. 000002). In addition, TLR2+1805T-transfected cells stimulated with PAM3 had significantly greater NF-κB activity compared to cells transfected with TLR2+1805G (841. 4 vs. 383. 7 RLU, p = 0. 0009). In contrast, responses to Malp-2 did not significantly differ between the two variants (783. 1 vs. 650. 6 RLU, p = 0. 37). Together, these results suggest that TLR1 variant 1805G leads to impaired innate immune responses to ML because of a defect in basal signaling of the TLR2/1 heterodimer. We next examined whether TLR1_T1805G regulates innate immune responses to ML in human primary immune cells. We obtained PBMC from whole blood from 28 healthy individuals whose genotypes for TLR1_T1805G had previously been determined [38]. The PBMCs were then stimulated with whole, irradiated ML, MLcw, and a variety of TLR ligands, including PAM3, PAM2, and LPS. We compared the high (1805TT) and medium (1805TG) responding genotypes with the low responding genotype (1805GG) (Fig 2). When stimulated with MLcw, PBMCs from 1805TT or 1805TG (1805TT/TG) donors showed significantly greater IL-6 responses compared to 1805GG PBMCs (Fig 2B, for MLcw at 2 µg/ml: 5,950 vs. 2,198 pg/ml, p = 0. 0005; for MLcw at 10 µg/ml: 6,115 vs. 3,320 pg/ml, p = 0. 0076). Similarly, responses to whole, irradiated ML were significantly higher in 1805TT/TG PBMCs compared to 1805GG PBMCs (Fig 2B, for whole, irradiated ML at 20 µg/ml: 3,310 vs. 1,649 pg/ml, p = 0. 0005; for whole, irradiated ML at 100 µg/ml: 6,183 vs. 2,246 pg/ml, p = 0. 0017). As previously observed, after stimulation with PAM3, significantly higher levels of IL-6 were seen in PBMCs heterozygous or homozygous for 1805T compared to1805GG PBMCs (Fig 2A, for TT/TG genotypes vs. GG genotypes stimulated with 75 µg/mL PAM3: 3,966 vs. 1,491 pg/ml, p = 0. 0007). Stimulation with LPS and PAM2, ligands with specificity for TLR4 and TLR2/6, respectively, produced no significant differences in IL-6 production between the two groups (Fig 2A). We also assessed the levels of other cytokines important in the monocyte immune response to mycobacteria and found that IL-1β production in PBMCs from 1805TT/TG donors stimulated with PAM3 or whole irradiated ML was significantly higher than in 1805GG PBMCs (Fig 3A). IL-1β levels did not differ between the two groups when PBMCs were stimulated with PAM2 or with LPS (TLR2/6 and TLR4 ligands, respectively) controls. Production of TNF-α, similarly, was significantly higher in 1805TT/TG PBMCs stimulated with PAM3, whole irradiated ML, or MLcw compared to 1805GG PBMCs (Fig 3B), but did not differ between the two groups after stimulation with Pam2 or LPS (TLR2/6 and TLR4 controls) (Fig 3B). Interestingly, there were no differences in IL-1β levels between TT/TG and GG groups after PBMC stimulation with MLcw, in contrast to the pattern seen with other cytokines (Fig 2B, 3A). Although recently published data suggests that TLR1_T1805G is associated with susceptibility to leprosy, associations with different types of leprosy or immunologic reactions have not been previously examined. To determine whether the TLR1_T1805G polymorphism was associated with different forms of leprosy or leprosy immune reactions, 933 patients from Anandaban Hospital in Kathmandu, Nepal were enrolled in a retrospective study. Of this total, 581 had polar lepromatous (LL), borderline leprosy (BL) or borderline borderline (BB) and 343 had tuberculoid leprosy (including borderline tuberculoid (BT) and polar tuberculoid (TT) ). A total of 344 patients experienced immune reactions during 3 years of regular visits to a leprosy clinic, of whom 238 had RR and 108 had erythema nodosum leprosum (ENL) and 2 had both reactions. The baseline characteristics of this population are described in Table 1. When individuals with lepromatous leprosy were compared to those with tuberculoid leprosy, there were no significant associations between SNP T1805G and either form of leprosy, either at the allelic or genotypic level of analysis (Table 2). However, there was a trend toward an association of the TG or GG genotypes with lepromatous leprosy (OR [odds ratio] 4. 76,95% CI [95% confidence interval] 0. 58–38. 87, p = 0. 11) in comparison to tuberculoid leprosy that did not reach significance. We also genotyped five additional TLR1 SNPs that are contained in common TLR1 haplotypes. We did not find associations of any of these SNPs with leprosy type (Table 2). Analysis of TLR1 haplotypes generated from these six SNPs similarly yielded no association with leprosy type (data not shown). We next investigated whether T1805G (I602S) might be associated with ENL or RR, a Th1-mediated immune event clinically manifested by inflamed skin lesions, fever, and neuritis. There was no association of T1805G or any other TLR1 polymorphisms with ENL. In contrast, the 1805G allele was associated with a reduced risk of developing RR in comparison to the 1805T allele (Table 3). The allele frequency of 1805G was 3. 9% in those with RR versus 7. 4% in those without (unadjusted OR 0. 51,95% CI 0. 29–0. 87, p = 0. 01). The distribution of the genotype frequencies was also significantly different with a p value of 0. 05 (Table 3). We next examined whether this association was affected by population admixture. This cohort contains representatives of more than 8 different ethnic groups with the majority belonging to one of four groups (Brahmin, Chhetri, Tamang, and Newar). There were no significant differences in frequencies of leprosy type or immunologic reactions among the different ethnic groups (Table 1). We performed a multivariate logistic regression, adjusting for ethnicity, and found that the odds ratio remained significant (OR 0. 52,95% CI 0. 30–0. 92, p = 0. 03). We also adjusted for ethnicity, sex and age as a continuous variable, and again found that the odds ratio remained significant (OR 0. 54,95% CI, 0. 30–0. 96, p = 0. 04). Previously, we found that TG individuals are intermediate between TT and GG individuals in responses to TLR1 stimulation [38]. We therefore investigated the influence of TG heterozygotes with a recessive (assumes T is recessive to G and compares TT versus TG/GG frequencies) or a dominant model (assumes T is dominant over G and compares TT/TG versus GG frequencies). In the recessive model, we found that the 1805 TG/GG genotypes were associated with a lower likelihood of RR compared to the TT genotype (OR 0. 55,95% CI 0. 31–0. 97, p = 0. 04). In the dominant model, the GG genotype was associated with a non-significant reduction in risk when compared to the TT/TG genotypes (OR 0. 15,95%CI 0. 01–2. 62, p = 0. 12). We next examined other TLR1 polymorphisms to determine whether there were any additional associations between individual SNPs or haplotypes and RR. SNP rs5743592, located in intron 2 adjacent to the 5′ UTR of TLR1, was found to be associated with a modestly increased risk of RR (Table 3, OR for allelic comparison: 1. 29,95% CI 1. 02–1. 64, p = 0. 04). When haplotypes of 6 TLR1 SNPs were examined (Table 4), one haplotype, TATTAG, was associated with protection from RR (OR 0. 55,95% CI 0. 31–0. 97, p = 0. 05). None of the other five haplotypes had any association with RR. Haplotype TATTAG was the only haplotype occurring with a frequency greater than 1% that contained the 1805G allele. Lastly, we examined whether haplotypes formed from SNPs rs5743592 and 1805 were associated with altered risk of RR. The haplotype associations were consistent with the individual effect of each SNP on the risk RR (Table 4), without any additive or synergistic effects. Together, these genetic data demonstrate that TLR1 SNP 1805G is associated with protection from RR. In this manuscript, we demonstrate that a human TLR1 SNP regulates the innate immune response to ML and is associated with protection from RR. This is the first study to describe an association of a TLR1 SNP with a Th1-mediated adaptive immune response. One weakness of our study is the low frequency of the 1805G variant in Nepal, which limited our power to detect associations with leprosy type. However, the relevance of the 1805G SNP to leprosy pathogenesis is supported by work from Johnson and colleagues, who recently reported an association of this variant with protection from leprosy in a Turkish cohort [39]. Although these authors do not mention whether or not T1805G was associated with different forms of leprosy in Turkey or with immune reactions, this may be due to the small size of their patient cohort (57 individuals). Genetic association studies that utilize cohorts of multiple ethnicities are also open to the criticism that associations are due to the effects of population admixture rather than the variant of interest. However, when we adjusted for ethnic composition of the comparision groups, we still found that the 1805G variant was associated with significant protection against RR. There are several possible mechanisms by which TLR1 might affect the pathogenesis of RR. At the cellular level, TLR1 might exert its influence through control of innate immune functions, such as the capacity of dendritic cells (DCs) and macrophages to control bacillary replication. Alternatively, or in addition, SNP 1805 may regulate DC maturation and/or antigen presentation and thereby influence the activation and maintenance of T cell responses to M. leprae antigens. Interestingly, recent work by other investigators demonstrates that the differentiation of monocytes into mature antigen presenting cells (APCs) is mediated by TLR signaling [35], [36]. For example, the 19 kDa protein of MTb signals through TLR2/1 to downregulate MHC class I and II antigen presentation by macrophages, leading to impaired T cell activation [45]. We have previously shown that ML also exerts an inhibitory effect on APC activation and maturation, through an as-yet unidentified mechanism [46]. In LL patients who are clinically stable, macrophages within LL lesions contain numerous bacilli that are seemingly resistant to host killing [47], [48]. However, this inhibition of phagocyte function seems to be overturned during RR. When patients undergo RR, the bacilli within these macrophages are rapidly cleared. This clearance coincides with the influx into the lesion of CD1b+ DC, which activate M. leprae-specific T cells and thereby promote intracellular killing. Importantly, the generation of CD1b+ DCs appears to be dependent on signaling through TLR2/1 [48]. Here, we show that individuals carrying the 1805G SNP are protected against reversal reactions. Our data suggests that TLR1 may be an important regulator of these effects on DCs. At the molecular level, the defect in 1805G signaling is likely due to a failure to express or retain TLR1 on the cell surface. Johnson recently showed that monocytes from 1805GG individuals completely lack surface TLR1, although total levels of this receptor are normal [39]. In Nepal, we observed a trend toward an association of the 1805G variant with lepromatous rather than tuberculoid leprosy. This trend, although not statistically significant, is consistent with impaired Th1 immunity, which is required for the tuberculoid form of the disease. An association of 1805G with lepromatous leprosy could explain the intriguing earlier observation by Krutzik and coworkers, who examined TT and LL lesions and were unable to detect any TLR1 staining in LL lesions [29]. Collectively, these findings suggest that TLR1 biology is different in lepromatous leprosy than in other forms of the disease. The absence of membrane-inserted TLR1 in 1805GG individuals may be associated with a Th2 immune response that arises by default in the absence of robust Th1 cytokine responses. This Th2 bias may in turn permit continued replication of the M. leprae bacillus and result in the clinical phenotype of LL. In our initial characterization of TLR1_T1805G, we found that this polymorphism is present in up to 76% of Caucasian Americans and is associated with a defect in innate responses to bacterial lipopeptide [38]. Worldwide, the T1805G polymorphism has variable frequency across ethnic groups. In Turkey, the allele frequency of this variant is 43% [39], while among African Americans and Vietnamese individuals, it has a frequency of 25% and 1%, respectively [38]. A broad array of pathogens are sensed by TLR2, and consequently by TLR2/1 or TLR2/6 heterodimers. These microorganisms include gram-positive and gram-negative bacteria, fungi, parasites, and mycobacteria [49]. Given the association of TLR1 1805G with Th1-mediated immune events, this SNP may influence the pathogenesis of any number of inflammatory conditions, including chronic mycobacterial infection, autoimmune disorders, sepsis and allergic reactions.

Mycobacterium leprae (ML) causes a disabling and stigmatizing disease that is characterized by distinct immune responses. ML produces a spectrum of illness in humans, and several lines of evidence indicate that host genetic factors influence susceptibility and clinical manifestations. Leprosy can occur as the lepromatous or tuberculoid forms, which are associated with different clinical manifestations, histopathology, T cell cytokine profiles, and bacterial burden in affected sites. Leprosy is also associated with unique immunologic reactions, such as reversal reaction, which is characterized by the rapid development of a Th1 T cell cytokine response that can cause substantial morbidity. We and others recently discovered a common human polymorphism in TLR1 (T1805G, I602S) that regulates cytokine production in response to lipopeptide stimulation, influences the cellular innate immune response to Mycobacteria, is associated with altered localization, and is present in 50% of individuals worldwide. Here, we show that in humans the 1805G variant does not mediate an inflammatory response to ML in vitro and that this polymorphism is associated with protection from reversal reaction. These data suggest that a common variant of TLR1 is associated with altered adaptive immune responses to ML as well as clinical outcome.

Abstract Introduction Methods Results Discussion

infectious diseases/neglected tropical diseases immunology/innate immunity genetics and genomics/genetics of the immune system

2008

Human TLR1 Deficiency Is Associated with Impaired Mycobacterial Signaling and Protection from Leprosy Reversal Reaction

Muscle contractions are generated by cyclical interactions of myosin heads with actin filaments to form the actomyosin complex. To simulate actomyosin complex stable states, mathematical models usually define an energy landscape with a corresponding number of wells. The jumps between these wells are defined through rate constants. Almost all previous models assign these wells an infinite sharpness by imposing a relatively simple expression for the detailed balance, i. e. , the ratio of the rate constants depends exponentially on the sole myosin elastic energy. Physically, this assumption corresponds to neglecting thermal fluctuations in the actomyosin complex stable states. By comparing three mathematical models, we examine the extent to which this hypothesis affects muscle model predictions at the single cross-bridge, single fiber, and organ levels in a ceteris paribus analysis. We show that including fluctuations in stable states allows the lever arm of the myosin to easily and dynamically explore all possible minima in the energy landscape, generating several backward and forward jumps between states during the lifetime of the actomyosin complex, whereas the infinitely sharp minima case is characterized by fewer jumps between states. Moreover, the analysis predicts that thermal fluctuations enable a more efficient contraction mechanism, in which a higher force is sustained by fewer attached cross-bridges. Intracellular forces and motions are generated by cyclic interactions between myosin and actin filaments. Myosin–actin interactions govern many important phenomena, such as molecular transport and muscle contraction. In the currently accepted theory, the globular portion of myosin firmly attaches to the actin filament until an ATP molecule, which fuels the molecular motors, binds to the myosin catalytic domain, releasing the myosin from the actin. The ATP is then hydrolyzed, allowing the myosin to weakly interact with the actin without generating any force. As the interaction strengthens and the hydrolization products are released, the myosin head can generate forces or motions that modify its configuration into one or more relatively stable states. When a new ATP molecule binds to the myosin molecule, it detaches from the actin and the cycle repeats. The driving force of actin-attached myosin is sourced from its total potential energy Et. The total actomyosin potential energy may be decomposed into two components. The first is generated by the chemical force field acting at the actin and myosin interface during their strong interaction, when the actomyosin complex is formed. We call this the biochemical energy Ec. The second component is generated by the mechanical force field associated with the myosin protein internal deformation; we call this the mechanical, or elastic, energy Ee. The balance between these two components defines the relative stability of the stable states of the attached myosin head [1]. Mathematical models of muscle contraction based on single actomyosin interactions have been extensively analyzed in the literature since the pioneering work of Huxley in 1957 [2]. Several experimental studies have been addressed and different basic hypotheses have since been proposed, because different definitions of these energy landscapes lead to different mathematical predictions of the muscle contraction mechanism. Although the general framework is well established, some hypotheses are still under debate. We do not review the basic hypotheses of these different approaches, as this paper focuses on a particular property of the biochemical energy that is overlooked among the great majority of mathematical models, namely the sharpness of the actomyosin potential wells. We will demonstrate the extent to which this property affects the model outcomes. Theoretical treatments of the rate constant dependence on the total free energy shape were extensively studied in the mid-1970s by Hill and co-workers [3–7]. In particular, in [5], the authors explicitly linked the rate constants between attached states with the total reaction free energy surface given by the sum of the reaction free energy (our Ec) and the structural free energy (our Ee). In that paper, a graphical description of the flexibility of the attached myosin motor on the total free energy was given as a contour map of the mechanical x and chemical s variables, referred to as the reaction coordinates. Our bidimensional representation in Fig 1 corresponds to the intersection of that energy with the reaction path x (s), the path of the minimum potential energy [5], which explicitly relates the two variables. However, to the best of our knowledge, this is the first time that this property has been quantitatively investigated by introducing thermal fluctuations within the minima and comparison with the commonly used hypothesis of infinitely sharp minima. Ee is associated with the deformation of the elastic component of the myosin protein, and is described as a convex energy. We refer to muscle myosin II isoforms, although other myosin types can be treated similarly. The elastic energy Ee has been characterized by applying small force or length perturbations to myofibers [1,8, 9] or to one/several molecules [10–13]. Although the results vary somewhat among experiments, the mechanical energy component can be completely characterized by determining the myosin elastic stiffness. The location of the flexible part of the myosin molecule, which has not been clarified to date, is not required in this analysis. Two common features of Ec among different models are: (i) non-convexity, which means that two or more minima, corresponding to actomyosin stable states, are present, and (ii) a bias toward the minimum corresponding to the post-power-stroke state (Fig 1, top-left). Cryo-electron microscopy and X-ray crystallography have confirmed at least two stable conformations of the actomyosin complex, each corresponding to a minimum biochemical energy [14–16]. The observed conformations are usually interpreted by the lever arm model, in which rotation of the long light chain domain of the attached myosin is driven by the energy released by ATP hydrolysis (however, see our Conclusions for a different interpretation). Depending on the external conditions, the long light chain acts as a lever, generating tension or motion. Between these two conformational states, researchers have hypothesized that there exists one or more stable states to explain some macroscopic behavior of the muscle fibers [1,17–20]. Experimental studies of single myosin molecules have suggested the presence of two power-stroke steps [12]. The most common approach to modeling muscle contractions consists of separating the myosin population into different stable states within the cross-bridge cycle and allowing jumps between states, characterized by forward (kf) and backward (kb) rate constants. A detailed balance requires that the ratio of the two constants is equal to the ratio of the probability distributions of the two states at equilibrium. Each probability distribution is a function of the stretch in the myosin elastic element at the moment of the state transition. Almost all previous models have assumed that this function has a relatively simple exponential dependence of the form kf/kb = exp (β (−ΔEe) ) exp (β (−ΔEc) ), where ΔE is the difference between the two energies in the local minima, and β = 1/ (κbT) or β = (κbT) −1, where κb is the Boltzmann constant and T denotes the absolute temperature. As shown in the Material and Methods section, this equation is only valid under the hypothesis of infinitely sharp minima in Ec. This simplification was also used by Hill and coworkers in the quantitative application of Hill’s theoretical formalism to muscle models [3,6, 7] as well as on later applications of their formalism to different muscle behavior (see for instance [22] and [23]). Although they considered a finite width within stable states, they did not include the effects of the width between stable states. In other words, they used the simplification that the mechanical coordinate x was independent of the reaction coordinate s. This hypothesis has sometimes been made explicit, as in the seminal paper of Huxley and Simmons [1] or in the work of Smith and Sleep [24], but is more often assumed implicitly. This approach collapses all the information on the biochemical energy landscape between the minima to a single constant value (the energy barrier), assuming the effect of its shape is negligible with respect to other biochemical energy parameters such as the number of stable minima n, the distance between the minima d, and the energy drop ΔEc caused by the ATP bias (see SI). Physically, the infinitely sharp minimum hypothesis corresponds to a fixed angle of the lever arm in each stable configuration. Thus, thermal fluctuations within the stable states are neglected in the majority of mathematical models, in the sense that they only formally allow for the state transitions, but their effects are collapsed into a single constant value (the unloaded rate). Thermal fluctuations within stable states can be simulated by defining an actomyosin energy landscape with stable minima of finite curvature (Fig 1, top-left). In this paper, we analyze whether and to what extent the width of the actomyosin energy minima in mathematical models may affect the predicted muscle mechanics at three levels: single myosin molecules, single fibers, and macroscopic muscle. To this end, we compare three shapes of the actomyosin biochemical energy in a ceteris paribus approach. Maintaining constant values for the actomyosin energy and other parameters characterizing the actomyosin cycle, we find that changing only the width of the energy minima alters the predicted dynamics of the myosin head in the attached state. If the minima are wide, the myosin lever arm can more easily modify its state, moving within the energy landscape. This micro-behavior affects the behavior of single fibers, causing a lower number of attached cross-bridges to generate higher tension in the fiber. Finally, to analyze whether such a difference is preserved at the macroscopic scale, where the huge number of cycling cross-bridges may average out the effects, we apply three models in a finite-element ventricle model [21]. We show that this behavior is preserved at the macroscopic scale of heart contraction, and improves the blood pumping performance. As the minima widen, the predicted ejection fraction increases, relaxation becomes faster, and the heart’s efficiency improves significantly. Although our analysis can be applied to all myosin isoforms, the mechanism is probably more important for cooperative motors such as muscle myosin. In analyzing the problem from single molecules to macroscopic muscle, we focus on the human cardiac isoform, but our goal is to generalize the relationship between the shape of the biochemical component of the actomyosin energy and the actomyosin dynamics. Hence, to fit the experimental data at each level, we select parameter values that replicate the physiological behavior of the heartbeat, but compare the single-fiber simulations with experimental data from frog skeletal muscle, as these have been intensively reported in the literature (unlike cardiac myosin data). Each myosin head follows the cross-bridge cycle shown in Fig 1 (top-right), with one detached state D, one weakly attached, non-force-generating state W, and an attached state divided in a pre-power stroke S0 and two post-power stroke states S1 and S2 (two-step power stroke). We use the common simplification that one single parameter x can describe the position of the myosin head with respect to its relaxed position on the thick backbone at x = 0, in a model where the myosin protein is attached to the thick filament through an elastic element of stiffness k. We specify that the first minimum (pre-power stroke) of Ec is at x0 (ta), the stretch in the myosin elastic element at the moment of attachment ta. At any time t during the actomyosin complex lifetime, we can compute the stretch in the pre-power stroke state x0 (t) as a function of the relative sliding of thin and thick filaments z (t) through x0 (t) = x (ta) + z (t) − z (ta). kx0 is the tension of the elastic element in the pre-power stroke state, and is then a function of z (t). The elastic element is characterized by an asymmetric elasticity [10] with stiffness k+ = 2 pN/nm for x > 0 (numerical values are listed in S1 Table). Let us consider for the moment only the attached states. We first explicitly derive the dependence of the detailed balance condition from the two components of the total energy, Ec and Ee. We can write the total energy of each myosin in x as E t (x, x 0) = E c (x, x 0) + 1 2 k x 2. The stationary distribution of the probability density then becomes ps (x, x0) = Nexp (−βEt (x, x0) ), where N is the normalization constant. For the sake of simplicity, assume that Ec has only two minima, at x = x0 and x = x0 + d, where d is the power stroke step, and a maximum in b (the reasoning can be extended to the actual case used in this paper, with three minima). At the stationary state, we can define the probability of being in the first minimum as: π 1 (x 0) = ∫ - ∞ b p s (x, x 0) d x (1) and that of being in the second minimum as: π 2 (x 0) = ∫ b ∞ p s (x, x 0) d x (2) The detailed balance condition requires that kf/kb = π2/π1. The previous equation cannot be solved without an analytic expression for Ec, so kf and kb, and their ratio cannot be calculated. A very common hypothesis used in almost all mathematical models of muscle mechanics is that the minima in the chemical energy are infinitely sharp, so the exponential of −βEc (x, x0) can be approximated by a delta function whose integral has a non-zero value only at the minima and is zero elsewhere. Under this assumption, the integrals in the detailed balance condition can be explicitly solved leading to kf/kb = exp (−β (Et (x0 + d, x0) − Et (x0, x0) ) ), which can be factorized into the two separate components of the energy. Any model (e. g. , [1,18,24–27]) that uses such a simplified dependence of the detailed balance on the strain of the elastic element in the pre-power stroke state (x0) is based on such a hypothesis. In this paper, we develop the theoretical formalism introduced by Hill [5] to account for the thermal fluctuations during the transition between attached stable states, and explore their influence on the model predictions. We consider a two-step power stroke corresponding to three minima in the biochemical energy. Each step of the power stroke (i. e. , the distance between the Ec minima) is d = 4. 5 nm. To appreciate the effect of infinitely sharp minima on the model outcome, we construct three scenarios with different actomyosin energy shapes. Two scenarios (SRI and SRII) are characterized by infinitely sharp minima, but with two different dependencies on the elastic tension. In the third scenario (SL), all of the minima have finite curvature (wide minima). SRI is the classical Huxley and Simmons model [1], in which the backward rates k10 and k21 are assumed to be constant and independent of the tension in the elastic element. The forward rates k01 and k12 are then defined so that their ratios to the backward rates satisfy the detailed balance. As noted in [24], SRI corresponds to a flat potential energy with a large repulsive energy barrier protecting the post-power stroke minima at x0 < 0 (Fig 1). In SRII, the minima are again infinitely sharp, but the rate constants are computed through the flat potential energy hypothesis by applying the Kramers–Smoluchovski (KS) approximation: k i j (x 0) = β η ∫ m a x l m a x r e β (- E t (x, x 0) ) d x ∫ m i n i m i n j e β (E t (x, x 0) ) d x - 1 (3) The first integral is computed between the left and right maxima (maxl and maxr) around minimum i; the second is computed between minimum i and minimum j (mini and minj); for a detailed description, see [28]. As described before, in the infinitely sharp minimum hypothesis, the variation of ΔEe inside the minimum can be ignored in the first integral, leading to the approximate solution: ∫ m a x l m a x r e β (- E t (x, x 0) ) d x ≈ k 0 e - β (k d (x 0 + d / 2) ) (4) where k0 is a constant related to the reaction rate in the unloaded case. The second integral of Eq (3) is analytically solved in [24]. Scenarios SRI and SRII require constants kb and k0, respectively, which are associated with the change of state in the unloaded myosin. These constant values are defined after the description of the third scenario. In the third scenario (SL), the central region of the biochemical energy landscape is defined as: E c (x, x 0) = H sin (2 π (x - x 0) / d + α d) + F A T P (x - x 0) (5) where the last term simulates the (linear) bias of the biochemical energy generated by the ATP energy release, such that E c S 0 > E c S 1 > E c S 2. αd is a constant angle that fixes the first minimum in x = x0. The energy barrier in the sinusoidal function is chosen as H = 6κb T0 at T0 = 310. 15K. We can then approximate the minimum in Ec as a parabola with stiffness kc = H (2π/d) 2 = 48 pN/nm. In this scenario, the detailed balance condition would require a numerical integration at each value of x0. Instead, the more detailed shape of Ec allows for a different approach. In the numerical simulations of SL, each myosin head was modeled as a material point in over-damped dynamics, experiencing viscous resistance through its drag coefficient η = 70 pNns/nm [29]. This leads to a system of Langevin equations of motion, given by: η x ˙ i = - ω i (t) E c ′ (x i, x 0 i) - E e ′ (x i) + η κ b T Γ (t) (6) for the i−th myosin. ω assumes a value of one or zero when the myosin is in the attached or detached state, respectively, and Γ (t) is a random term characterized as white noise by <Γ (t) > = 0 and <Γ (t1) Γ (t2) > = 2δ (t1 − t2). The equations are coupled through the dependence of x 0 i on the relative position of thin and thick filaments z. This system was solved by an implicit method (see SI) that exploits the physical properties of the components acting on the myosin dynamics without requiring rate constants. The approach is inapplicable to scenarios SRI and SRII, because the time step is related to the characteristic time of the problem, which is inversely proportional to the sharpness of the minima. In these models, we compute the rate constants at different x0 using the formulas proposed in [1] for SRI and in [24] for SRII, as described previously. As mentioned earlier, SRI and SRII require two unloaded rate constants (one for each power stroke minimum). To maintain homogeneity when comparing the scenarios, we define the constants such that the unloaded rates in SRI and SRII match the KS approximation of the unloaded energy landscape in SL. In each scenario, the positions of all single molecules were tracked through the numerical simulations, and the total tension was obtained by summing the tensions in the elastic elements of the attached myosin heads. Finally, to complete the cross-bridge cycle, we must describe the transitions among the strongly attached states and the detached (D) and weakly attached (W) states. The W–S and S–D transitions follow the classical H57 hypothesis [2]; that is, the attachment rate is non-zero only in the stretched configuration of the elastic element (x > 0), where it increases linearly with x up to xlim. The detachment rate, kSD in Fig 1, also increases linearly with x > 0, and attains a very high value when x < 0. This hypothesis, related to the Brownian search and catch mechanism, has been observed in Myosin V [30], and has been extensively used in muscle modeling under different modifications. If the absolute value of x exceeds 15 nm, mechanical dislodging occurs. The transition from D to W and vice-versa is governed by two rate constants, kDW and kWD in Fig 1, which are independent of the tension in the elastic element, but do depend on the calcium concentration and on the sarcomere geometry, as described in the following. To extend the comparison of the different hypotheses beyond a single molecule, we include them in a bidimensional sarcomere geometry, enabling a quantitative simulation of several experimental protocols. We then use these simulations to compare the different outcomes of the three models in a ceteris paribus approach to investigate the consequences of accounting for thermal fluctuations within each stable state. The single half-sarcomere depicted in Fig 1 (bottom-left) is based on the real geometry used. Each thick myosin filament has NXB myosin heads attached to it, and these can interact with actin filaments. The half-sarcomere is composed of NFil parallel thick filaments with an actin filament interposed. Actin filaments are constrained to have the same value of z through the Z-line. z = 0 refers to the optimal length of the sarcomere. Both filaments are rigid. The actin filaments have troponin-tropomyosin units attached, which tune the D–W interaction according to the Ca concentration (this effect is more important in the heart level model, because the single sarcomere model is always tetanized in the presence of a high Ca concentration). The D–W interaction is also regulated by the cooperativity effect (an increase in actin affinity for myosin with attached neighbor heads) and thin filaments overlapping due to over-compression. The three mechanisms are described in SI, and in [21]. Using the parameters described above (which are identical in the three actomyosin energy scenarios), we simulate the tension generated in the filaments under various experimental protocols, namely isometric contraction, isotonic shortening, and the length-step protocol [1]. Most parameters remained unchanged throughout the simulations reported in this paper. The exception is the temperature, which (to match the experimental data) was set to 37°C in the beating heart analysis and to 4°C in the single-molecule and single-fiber analyses. The temperature effect was modeled through the Q10 parameter in the attachment and detachment rate constants [31], and was intrinsically included in the KS approximation of the power stroke rate constants. To test the influence of the different hypotheses when thousands of myosin heads are working in unison under physiological conditions, we coupled the sarcomere models described in the previous subsection with the finite-element ventricle model, a complete description of which is given in [21]. In each finite element, we included the tension generated by the three different sarcomere models, scaled over the area of each finite element. In doing so, we made the assumption that the single half-sarcomere is representative of the tension generated by the whole muscle cell (homogeneity of contraction). The numerically simulated Langevin approach is unsuitable for a heart simulator. Even in the implicit scheme, the characteristic time scales of the model limit the maximum time step to a few nanoseconds. To simulate several heart beats with thousands of elements, we used the KS approximation of the wide minima scenario (see Eq (3) ). The two forward and two backward rates, k01, k10, k12, and k21, were computed at 102 discrete intervals of the stretching of the elastic component in the interval −20 nm ≤ x0 ≤ 20 nm. Moreover, in both the detached and attached states within each minima, the position of each myosin head was randomly selected from the corresponding probability distribution (see SI), reproducing the effect of thermal fluctuations on the actual stretching level of the elastic element. To test the new approach, we compared it with the SL and SRII scenarios at the single cross-bridge and whole-fiber levels. The approximation agreed with the Langevin approach (see S3 and S4 Figs), especially in the comparison with SRII. In the numerical experiments, the ventricle wall was discretized into 45,256 tetrahedral elements (Fig 1, bottom-right). The macroscopic contraction force along the prescribed fiber orientation in each element was provided by 12 embedded filament pair models. The macroscopic contraction force was specified such that the sum of the work done by each actomyosin complex equaled the work done by the contraction during an infinitesimal deformation of the continuum. The shortening velocity of the sarcomere was dz/dt = − (SL0/2) × dλ/dt, where SL0/2 is the initial half-sarcomere length and λ is the magnitude of stretch along the fiber orientation. With this construction, we could explore, at the organ level, the physiological meanings and impacts of different energy landscapes in the three scenarios. The first effect of including thermal fluctuations in the actomyosin complex is directly appreciated at the single-molecule level. Recall that the power-stroke parameters are the same in all three scenarios, and that each scenario differs only in the shape of the minima in the energy landscape. The number and separation distance of the minima, myosin stiffness, and drag coefficient are the same. As previously described, we also set the rate constants of the unloaded myosin to be the same in all three cases. This is equivalent to imposing the same energy barriers between the minima. Nonetheless, the dynamics of the actomyosin complex during the same attachment–detachment event differed substantially among the scenarios. Fig 2 shows representative traces in the three scenarios. The simulations reproduce an isometric contraction (z (t) = 0). After attachment in the pre-power stroke state, the lever arm must overcome an energy barrier to generate the power stroke. This barrier depends on both the biochemical and mechanical parts of the actomyosin energy. In SL, the actomyosin state changes several times during the actomyosin lifetime, even in an isometric contraction. In contrast, the SRI and SRII scenarios generate only one or a few power stroke events. With infinitely sharp minima, the power stroke exhibits an on–off behavior. The myosin attaches in the pre-power stroke state, generating low or zero force, and then undergoes a single transition to a completely different state. This new state generates a higher force, pulling the thin filament as far as possible before one ATP arrives. The ATP detaches the head and reprimes the lever arm of the myosin. The wider minima generate a more dynamic framework. In this case, the power stroke is no longer a single event; instead, the actomyosin complex exists as a family of states. After attaching in the pre-power stroke, the myosin head can easily explore the energy landscape before or after the release of the hydrolization products. In some sense, the wider minima case corresponds to a loose coupling of the mechanical event and release of the ATP hydrolization products. The lever arm can more easily move between stable states in the presence or absence of Pi and ADP [32]. In contrast, the sharp minima are more closely related to tight coupling, with one (or few) transition (s) per attachment–detachment event. This different behavior can be explained by superposing the biochemical and mechanical energy landscapes of the moving particle in the SL and SR scenarios (Fig 1, top-left). With the parabolic approximation described above, we can compute the shift of each minimum Δd (shown in Fig 1, top-left) as a function of kc and x0: δd= (1−1k/kc+1) (x0¯) (7) with x ¯ 0 = x 0 + n d, where n = 0,1, 2 for S0, S1, or S2, respectively. The power stroke step then depends on the tension of the elastic element in SL, whereas in SRII (or SRI) with kc = ∞, the power stroke step has a constant distance d, and Δd = 0. A similar effect applies to the maximum in the energy barrier, which is shifted toward a position of lower mechanical energy for wider minima, reducing the energetic barrier (Fig 1). Thus, under the wider minimum hypothesis, the dynamics of the myosin head cause the transition rate to increase at higher tensions, despite being the same in the unloaded condition. Obviously, the differences between the scenarios are related to the relative rates of the attachment–detachment events and the power stroke event. Nonetheless, with the parameter values used to generate Fig 2, we can quantitatively fit the microscopic model to the macroscopic experimental observations, as shown in the following subsections. As mentioned above, widening the minima shifts the minima of the post-power strokes S1 and S2 toward the minimum of S0 and decreases the energy barrier, favoring state transitions. Moreover, the power stroke step d (x0) and the total force decrease. One may wonder how the altered dynamics at the single cross-bridge level affect the macroscopic muscle contraction. To answer this question at the fiber level, we introduce the three scenarios (SRI, SRII, SL) into a single sarcomere model. Moreover, we assume, for simplicity, that all sarcomeres aligned in series and parallel within a single fiber behave in the same way; therefore, we can simulate the total tension generated by the fiber. We also simulate the rising phase of an isometric contraction and two experimental setups: the force clamp and the length step. The former experiment analyzes the constant velocity of contraction under different constant loads applied to the fiber; the latter analyzes the force recovery after a small, fast change in the fiber length. During an isometric contraction, the wider minima generate a higher force with fewer attached cross-bridges (Fig 3). This improved performance is attributable to the higher average population in the second minimum with respect to the total number of attached cross-bridges STOT in SL (43% of STOT, see S1 Table) than in SRII (29%) and SRI (12%). The reduced number of attached cross-bridges is a result of the hypothesis of faster detachment from the more stretched elastic component. The wider minima also accelerate the attachment–detachment cycle by increasing the contraction velocity against a constant force. For any given force, SL achieves a higher contraction velocity than SRI and SRII (Fig 4). With the selected relative values of the parameters, the maximum velocity is comparable to the experimental data. However, the parameter values were identical in the three scenarios, so the faster velocity resulted solely from the altered energy landscape. Let us assume that the half-sarcomere model is in mechanical equilibrium with the constant external force F pulling the Z-line. For simplicity, we assume that the myosin elastic element has linear elasticity with the force constant k. Then, we have: k ∑ i δ s i (t) (x i (t) - x 0 i (t) ) + (x 0 i (t a i) - z (t) + z (t a i) ) = F (8) Here, the myosin proteins are indexed with i, and δs is equal to 1 if the myosin is attached and 0 otherwise. Let us estimate the increment Δz given by the state transitions within the strongly attached state (changes of x i (t) - x 0 i (t) ) in the time interval [t: t + Δt]. For simplicity, we assume that there are no transitions between the strongly attached state and the other states for any myosin. Then, we have: k ∑ i δ s i (t) (Δ (x i - x 0 i) - Δ z) = 0 (9) where Δ (x i - x 0 i) is the strain increment given by the power strokes in [t: Δt]. The above equation can be rewritten as Δ z Δ t = 1 Δ t ∑ i δ s i (t) Δ x i ∑ i δ s i (t) = Δ X P S / Δ t ∑ i δ s i (t) (10) where ΔXPS/Δt is the generation rate of the power strokes in the hypothesis that no attachment or detachment occurs in [t: t + Δt]. If we take into account the strain at the elastic part given by attachment and lost by detachment, we can rewrite the numerator in Eq (10) as (ΔXA + ΔXPS − ΔXD) /Δt, where ΔXA and ΔXD are the magnitude of strain given by the attachments and lost by the detachments, respectively, in [t: t + Δt]. The above equation means that the sliding velocity varies inversely with the number of attached myosin heads for the same generation rate of power strokes. This explains the superiority of SL for achieving faster shortening. One might question whether the difference in myosin dynamics would persist after modifying the unloaded rate constants in SRI and SRII, which are here chosen to equalize that in SL. This modification would accelerate the transitions in the attached state, and also modify the fast recovery of the tension after a small length change. When an isometrically contracting muscle is subjected to a small, rapid length change, the tension almost instantaneously changes from its isometric value T0 to T1, then partially recovers to its original value, reaching a level T2 (see SI) at a rate r that reflects the power stroke velocity. T1, T2, and r depend on the length change δ. Fig 5 (upper panel) plots the tension recovery rate computed by exponentially fitting the tension vs. time curve. The recovery rate better matches the experimental results in SL than in the other scenarios with their chosen constants. Increasing the value of k0 in SRII (Eq (4) ) such that the recovery rate approaches that of SL at δ = 5. 6 nm results in overestimates and underestimates of the recovery rate itself at higher and lower values of δ, respectively (Fig 5, lower panel). This result shows that the choice of the constant parameter alone cannot justify the differences in the previous analysis. Moreover, the shape of the r (δ) curve supports the initial choice of k0. At the single-sarcomere level, the cross-bridges interact with each other through rigid actin and myosin filaments. In the whole fiber, the geometry is simplified by considering the sarcomeres to have uniform behavior. Whether these effects are preserved in a more physiological situation with several half-sarcomeres interacting in series and in a more complicated geometry is not a trivial question. To explore this idea further, we incorporated the three single-fiber models into a finite-element ventricle model. As shown in Fig 6, the previous differences are maintained at the macroscopic level of the heart. This result is non-trivial, because our analysis was made ceteris paribus, that is, we altered only the shape of the biochemical energy within and between the minima. All other parameters, such as the number of minima, the distance and energy drop between minima, constants of the attachment–detachment process, and geometric parameters, were unchanged. Importantly, thermal fluctuations inside the energy minima are usually neglected in mathematical models of muscle mechanics, which typically adopt the infinitely sharp minima hypothesis. Widening the minima increases the left ventricular pressure, leading to a higher ejection fraction. The relaxation is also accelerated, probably because of faster detachment (predicted at the single fiber level). The wider minima improve the general performance of the ventricle model, although SRII also generates physiological behavior. With the same parameter values, the SRI scenario shows limited ability to replicate a healthy heart. The higher pressure and ejection fraction in SL than in SRII are again linked to the lower percentage of attached cross-bridges. During the contracting phase, S2 is more populated in SL than in SRII. On the contrary, S0 and S1 are more populated in SR than in SL (Fig 6 (central panel) and Fig 7). It is worth noting that the simulated percentages of attached cross-bridges agree with the values inferred in [35]. In the simulations, only 20% of the simultaneously attached myosin heads generated the proper stroke volume. Finally, the contraction efficiency (defined in SI) is higher in the wider minima case of SL (28. 7%) than in SRII (25. 8%) and especially in SRI (16. 1%). We quantitatively analyzed the dynamics of the cross-bridges during one heart beat in three different sarcomeres, placed at the endocardium, the epicardium, and between these two regions. Although the number of attachment and detachment events was quite similar in the three scenarios, the number of backward jumps was much higher in SL than in SRII and SRI. In SL, each attached cross-bridge made approximately five backward (followed by forward) jumps between the minima. In the sharp minima case, less than one-third of the cross-bridges demonstrated such dynamic behavior. By varying the shape of the minima and the intervening maxima of the potential energy of the actomyosin complex, this study explored the influence of the sharpness of the minima on the predicted actomyosin dynamics in three different mathematical models. We compared three scenarios, one corresponding to the classical Huxley and Simmons hypothesis [1], one following the analysis introduced in [24], and one introduced in [36]. The differences among the scenarios were compared at the single molecule, sarcomere, and whole-organ levels. Mathematically, the first two scenarios are characterized by infinitely sharp minima, whereas the third introduces a finite stiffness into the minima. In physical terms, the actomyosin complex in the first two scenarios is allowed to jump between stable states with rigid configurations, whereas the third scenario allows small thermal fluctuations within the stable configurations. In the vast majority of mathematical muscle models in the literature, the sharpness is effectively infinite because of the relatively simple expression of the detailed balance condition. Although it is now 40 years since the theoretical formalism that relates the rate constant between states to the mechanical and chemical energy shape was introduced [5], to the best of our knowledge, there has been no quantitative analysis of the consequences of the above-mentioned hypothesis. With no further differences in the actomyosin energy or the whole cross-bridge cycle, we showed that the shape of the minima in the energy landscape plays a fundamental role in the predicted dynamical behavior of each myosin molecule attached to an actin filament. In particular, widening the minima induces several rapid state transitions in single attachment–detachment events, whereas sharp minima generate few or single power stroke events. Moreover, allowing thermal fluctuations improves the predicted efficiency of the muscle contraction at all levels (single molecule, fiber, and whole organ). In the consensus hypothesis, the myosin heads always attach in the pre-power stroke state. Consequently, they reach thermal equilibrium faster when rapid state transitions occur between the minima in the energy landscape. This dynamical behavior increases the populations of the force-generating states. The lower the lifetime of the actomyosin complex and the transition rates, the higher the percentage of time spent in the pre-power stroke (the low-force-generating state). In other words, because the cross-bridges always attach in the low-force state first, a faster state transition generates a higher probability of being in the high-force-generating state during the same actomyosin lifetime. As the values of the three energies in the minima are the same, this probability is the same in the steady state distribution, but the difference becomes negligible only at sufficiently long actomyosin lifetimes. Importantly, the shape of the biochemical component only has a non-trivial effect on the cross-bridge dynamics in relation to the mechanical components: wider minima reduce the effect of the mechanical energy, such that the minima in S1 and S2 have a lower energy barrier than in the infinitely sharp minima case, and shift toward the minimum in S0. The importance of this interaction is also revealed in the dependence of the rate constants on the stretching of the elastic component (S7 Fig). Comparing the forward rates k01 in SRII and the KS approximation of SL, we found that they substantially differ when the myosin is under high tension, but are similar when the tension is low or oppositely directed, as occurs in the step length protocols. As already mentioned in [24], single-molecule experiments may allow the testing of different energy shapes, through the exploration of observables under different setups. As we have said, relaxing the infinitely steep minima hypothesis results in some fundamental differences from previous models. The attached myosin moves faster between stable states, the mean force generated per attached head is higher, and the heart simulator shows a faster drop in tension during relaxation. In principle, these effects can also be reproduced by other models, imposing different rate constants that are largely unknown at present. Despite this, the detailed balance condition on the ratio of backward and forward rate constants between stable states must be respected to ensure the consistency of any model [3]. We have already shown that the behavior predicted by the open minima model cannot be exactly reproduced by the steep minima models by simply modifying a single rate constant (Fig 5). Thus, other parameters in the model must be modified, such as the energy minima values or the elastic stiffness. In turn, this will modify other behavior simulated by the model, and eventually force the introduction of new, ad hoc hypotheses to overcome the new discrepancies. Obviously, this is only true if the real shape of the energy in muscles is relatively wide. The influence of the wideness of the minima in the attached state has been presented in the framework of the lever arm hypothesis, currently the most widely accepted theory. However, to the best of our knowledge, the rotation of myosin molecules has not yet been observed. Indeed, the observed dynamical behavior of single myosin molecules may suggest a different mechanism [13], as has recently been supported by molecular dynamics simulations [37,38]. In the new model, the myosin head thermally fluctuates on the actin filament, generating force until it reaches a strong-binding site (“hopping” assumption). Our wide minima approach can actually be applied to both mechanisms, with a different interpretation of the multi-stable biochemical energy. In the lever-arm assumption, different minima are associated with different stable states of the lever arm, where the myosin globular portion is always attached to the same actin monomer. The wideness of the minima is then related to the internal rigidity of the actomyosin complex in each stable state. In the hopping assumption, the potential arises from electrostatic interaction [37] between myosin and different actin monomers, generating different minima. The two assumptions would not affect the mathematical treatment. Notably, the apparent stiffness used in our definition of Ec around the minimum is about 48pN/nm (see Material and Methods), comparable with the stiffness within the minima obtained in [37] (about 16pN/nm at 25mM ionic strength, which is expected to increase at physiological ionic strengths). We have imposed an energy drop in Ec that requires almost all ATP energy (16κb T0), whereas in this previous study, the energy drop seems to be much lower. A complete comparison of the two models is beyond the scope of this paper. All mathematical models of muscle mechanics are limited by the unknowability or wide variability of the real values of several parameters under physiological conditions. The present analysis is no exception. The influence of the energy landscape shape on myosin dynamics is probably affected by the choice of the attachment and detachment parameters and the common parameters of the actomyosin energy landscape. Moreover, although we have accounted for several components of the sarcomere and whole organs, other factors that have been neglected may influence the final behavior. For instance, we have not included the rigor states or the elasticity of the thin and thick actin filaments. Despite these omissions, the general behavior and time scales agree with several physiological and experimental data, demonstrating the robustness of the model with the chosen parameter set. Importantly, the heart simulator integrated with the SL model exhibits a quantitative fit with several experimental observations of muscle contraction, from the single-molecule level to the whole heart. Rather than averaging, we considered the dynamics of each single myosin head. The model quantitatively matched the contraction velocity and the T1 and T2 curves, supporting the choice of the common parameters used. Moreover, as shown in Fig 4, the model reproduced the experimental observations of the four phases when transiting from an isometric to an isotonic contraction. Using such a model, we might elucidate the physiological meaning of several cardio-myopathies at the single-molecule level and connect them to whole-heart dysfunctionalities.

Mathematical models are of fundamental importance in the quantitative verification of biological hypotheses. Muscle contraction models assume the existence of several stable states for the myosin head, whereas the transition rates between states are defined to fit experimental data. The ratio of the forward and backward rates is linked to the ratio of the probabilities of being in one or other stable state at equilibrium through a detailed balance condition. A commonly used assumption leads to a relatively simple expression for this balance condition that depends only on the values of the energy at the minima and not on the minima shape. Mathematically, this hypothesis corresponds to infinite sharpness at these minima; physically, it neglects the small thermal fluctuations within actomyosin stable states. In this work, we compare this classical approach with a model that includes thermal fluctuations within wide minima, and quantitatively assess how much this hypothesis affects the model outcomes at the single molecule, single fiber, and whole heart levels. It is shown that, using parameters compatible with known behavior in muscle mechanics, relaxing the infinitely sharp minima hypothesis improves the predicted force generation and efficiency at the macroscopic level.

Abstract Introduction Materials and Methods Results and Discussion

cell motility medicine and health sciences actin filaments myofibrils classical mechanics muscle tissue cardiovascular anatomy mathematical models muscle contraction mechanical energy molecular motors actin motors thermodynamics motor proteins research and analysis methods muscle physiology contractile proteins animal cells proteins mathematical and statistical techniques biological tissue muscle cells thermal stability physics biochemistry cytoskeletal proteins sarcomeres cell biology anatomy physiology myosins biology and life sciences cellular types physical sciences heart

2016

Including Thermal Fluctuations in Actomyosin Stable States Increases the Predicted Force per Motor and Macroscopic Efficiency in Muscle Modelling

Epilepsy is a common disabling disease with complex, multifactorial genetic and environmental etiology. The small fraction of epilepsies subject to Mendelian inheritance offers key insight into epilepsy disease mechanisms; and pathologies brought on by mutations in a single gene can point the way to generalizable therapeutic strategies. Mutations in the PRICKLE genes can cause seizures in humans, zebrafish, mice, and flies, suggesting the seizure-suppression pathway is evolutionarily conserved. This pathway has never been targeted for novel anti-seizure treatments. Here, the mammalian PRICKLE-interactome was defined, identifying prickle-interacting proteins that localize to synapses and a novel interacting partner, USP9X, a substrate-specific de-ubiquitinase. PRICKLE and USP9X interact through their carboxy-termini; and USP9X de-ubiquitinates PRICKLE, protecting it from proteasomal degradation. In forebrain neurons of mice, USP9X deficiency reduced levels of Prickle2 protein. Genetic analysis suggests the same pathway regulates Prickle-mediated seizures. The seizure phenotype was suppressed in prickle mutant flies by the small-molecule USP9X inhibitor, Degrasyn/WP1130, or by reducing the dose of fat facets a USP9X orthologue. USP9X mutations were identified by resequencing a cohort of patients with epileptic encephalopathy, one patient harbored a de novo missense mutation and another a novel coding mutation. Both USP9X variants were outside the PRICKLE-interacting domain. These findings demonstrate that USP9X inhibition can suppress prickle-mediated seizure activity, and that USP9X variants may predispose to seizures. These studies point to a new target for anti-seizure therapy and illustrate the translational power of studying diseases in species across the evolutionary spectrum. Mutations in the PRICKLE genes can cause seizures in humans, zebrafish, mice, and flies, suggesting the seizure-suppression pathway is evolutionarily conserved. [1–5]. Prickle binding partners have been studied extensively only in either non-neuronal vertebrate cell lines or non-neuronal tissues in the fly. (In both cases Prickles were shown to bind other WNT/PCP proteins. [6,7]) Such targeted approaches showed Prickles interact with REST, some kinases (including BCR), and post-synaptic density proteins, including TANC1 and TANC2. [4,6] To identify neuronal proteins that bind Prickles, recent work by our group and others showed PRICKLE1 also binds to Smurf1 (a ubiquitin ligase), [8] and Synapsin1, (a gene implicated in both epilepsy and autism) [1]; and PRICKLE2 also binds PSD-95 and p150Glued. [9] To identify other PRICKLE binding partners in neuronal-like cells, we used mass spectroscopy: a more global, unbiased approach. The interaction was monitored in a subclone of rat pheochromocytoma PC12 cells which, when treated with Nerve Growth Factor (NGF), assume a sympathetic neuron-like phenotype. [10] Clonal PC12 lines that overexpressed doxycycline-inducible GFP, GFP-PRICKLE1 (GFP-PK1), or GFP-PRICKLE2 (GFP-PK2) were produced (S1 Fig). Protein complexes immunoprecipitated with anti-GFP beads from whole-cell lysates (S1C Fig) were analyzed by mass spectrometry (IP-MS, S2 Fig). [1] Prickle interactors were considered candidates only if they recovered >10 peptide matches with both GFP-PRICKLE1 and GFP-PRICKLE2, but not with GFP alone (S1 Table). This dataset recovered both known Prickle-interactors (e. g. , Tanc2, and Bcr[6]) and novel Prickle interactors, including Usp9x, [11–15] a substrate-specific de-ubiquitinase. The putative Prickle-Usp9x interaction was of particular interest because Usp9x physically interacts with Smurf1 (to date, one of the few Prickle interactors identified in neural tissues[13] and both are implicated in neurite extension). [16] Moreover, since ubiquitination plays a role in cancer pathogenesis, a variety of reagents that modulate this system are already commercially available and in clinical trials. [17] The combination of previously identified Prickle-interacting partners with the present studies are depicted in the Prickle-interactome (Fig. 1) that we utilize to identify new seizure-modifying targets. To evaluate further USP9X as a PRICKLE binding partner, coimmunoprecipitation (co-IP) assays were carried out in the original PC12 GFP, GFP-PRICKLE1, and GFP-PRICKLE2 cells lines, utilizing Tanc2 and Bcr as positive controls. To detect the respective endogenous proteins, GFP-immunoprecipitates from the differentiated cell lines were immunoblotted with anti-BCR, anti-TANC2, and anti-USP9X. Fig. 2A demonstrates interaction by all three proteins in GFP-PRICKLE1 and GFP-PRICKLE2 immunoprecipitates. To validate this interaction, co-IPs were carried out in a different cell line. Flag-tagged PRICKLE1 and PRICKLE2 were immunoprecipitated from human embryonic kidney (HEK293T) cells and subjected to anti-USP9X Western blot analysis; Fig. 2B shows both PRICKLE1 and PRICKLE2 interact with endogenous USP9X in HEK293T cells. To map the PRICKLE1-USP9X interacting domain, recombinant vectors expressing full-length, N-terminal or C-terminal portions of PRICKLE1 (schematic diagrams in Fig. 2C, D) were transfected into HEK293T cells and immunoprecipitated via the Flag tag. Fig. 2C showed that endogenous USP9X interacted specifically with the PRICKLE1 C-terminus. On the other hand, BCR interacted with the N-terminus, a region that includes the PET/LIM domains, previously known to mediate Prickle1 protein-protein interactions. [20] The same approach defined the PRICKLE2-USP9X interacting domain. As with PRICKLE1, while the interaction with BCR mapped to the N-terminus, the PRICKLE2–USP9X interaction mapped to the C-terminus of PRICKLE2 (Fig. 2D). SMURF1 and USP9X are known to physically interact via the second WW domain of SMURF1 and the USP9X carboxyl terminus (a fragment named C2) [13] so, hypothetically, the PRICKLE interaction should also map to the same region. To test this, a PRICKLE1 or PRICKLE2 c-terminal fragment was expressed with one of two flag-tagged Usp9x deletion fragments: C1Usp9x (amino acids 1216–2107) and C2Usp9x (the carboxy terminal amino acids 2102–2560; Fig. 3A). Complexes were immunoprecipitated via the Flag tag. Fig. 3A shows that the PRICKLE1 and PRICKLE2 interactions also map to the carboxyl terminus of Usp9x (C2). The C-terminal end of both PRICKLEs and USP9X are therefore crucial for the PRICKLE-USP9X interaction. PRICKLE is ubiquitinated by the E3 ligase, SMURF2, which promotes PRICKLE degradation and turnover in HEK293T cells. [8] Ubiquitin tags from mono- and polyubiquitinated proteins can be removed by the C1-terminal catalytic motif of USP9X deubiquitinase. [12,13,21–23] We postulated the physical interaction between the PRICKLEs and USP9X might indicate that PRICKLE was a USP9X substrate, and that USP9X deubiquitinates and stabilizes both PRICKLE1 and PRICKLE2. To test this idea, PRICKLE substrates were monitored in USP9X deubiquitination assays. In previous studies, overexpression of the C1Usp9x catalytic fragment alone was sufficient to deubiquitinate USP9X substrates. [23] Accordingly, the robust ubiquitination of Flag-PRICKLE1 (Fig. 3B, C) was antagonized by overexpression of C1Usp9x; and PRICKLE was stabilized (Fig. 3). Similar results were obtained with full-length USP9X (S3 Fig). In the presence of the proteasomal inhibitor LLNL (N-acetyl-L-leucyl-L-leucyl-L-norleucinal), polyubiquitinated PRICKLE1 or PRICKLE2 accumulated, indicating that degradation of ubiquitinated PRICKLE is mediated by the proteasome as opposed to the lysosome (S3 Fig). To determine if a Prickle-Usp9x interaction has a functional effect in mice, we generated mice with deletion of Usp9x in the forebrain (Emx1-Cre/Usp9x loxtemp) [12] and examined expression of Prickle2. In the absence of Usp9x in 4-week old mice, Prickle2 was not detected in the cortex, CA1 region of the hippocampus, and dentate gyrus (Fig. 3D). Taken together, these data suggest that Prickles are novel Usp9x substrates. Genetic analysis was used to determine if Prickle and Usp9x interact genetically in fruit flies. We reduced the dosage of the USP9X orthologue (fat facets, or faf) in the context of a prickle mutation (pksple/+) which on its own promotes both behavioral and electrophysiologic seizure activity. [3,5] Three separate loss-of-function faf alleles were used to create transheterozygotes, which were then assayed with the bang sensitivity behavioral assay to assess recovery time from seizure activity. For every faf allele used, reducing its level suppressed the seizures in the context of the pksple/+ flies (Fig. 4A-C). Adding the USP9X small-molecule inhibitor Degrasyn/WP1130[24] to the fly food inhibited seizure activity in pksple homozygous flies (Fig. 4D) using a “fly flip” assay (see Materials and Methods). The demonstration that pharmacologic treatment with a USP9X inhibitor suppresses seizures in conjunction with genetic seizure suppression utilizing fly USP9X orthologue mutations suggests that USP9X mutations could protect humans from seizures. In contrast to this suggestion, in humans, USP9X has been postulated to be an epilepsy candidate gene. USP9X is X-linked; and in males has been associated with both autism spectrum disorder (ASD) and intellectual disability (which are frequently co-morbid with epilepsy). Moreover, mice with Usp9x-deficient neurons develop abnormal neuronal connectivity. [12,14,25,26] A similar apparent discrepancy has been observed in the case of the SCN1A gene (which codes for the NAV1. 1 channel) and its fly homologue para. Here, the great majority of 68 different mutations in para suppress seizures in the fly, [27–31] yet human SCN1A mutations are associated with severe epileptic encephalopathies, febrile seizure syndromes, and Dravet syndrome. Recent studies demonstrated that although loss-of-function para alleles suppress seizures, a few recently identified gain-of-function para alleles (e. g. , parabss1) actually cause seizures in the fly. [27] Accordingly, many of the most severe human SCN1A mutations are missense or truncation, gain-of-function alleles (see web resources). [32] To assess if USP9X coding variants might be associated with human seizures, the USP9X gene was resequenced in 284 male patients with epileptic encephalopathy. One male patient, T17133, presented with epileptic encephalopathy and was found to carry a de novo USP9X mutation (c. 3034T>C, p. Ser1012Pro) at a highly conserved residue (GERP 5. 62) that was predicted to be possibly damaging by PolyPhen (0. 898). Another patient with infantile spasms (parental genotypes unavailable) harbored variant (T2587c. 5669G>A, p. Gly1890Glu) a rare missense variant. Both mutations lie outside the identified Prickle-binding domain. The USP9X pSer1012 is very close to the Ubiquitin-like (UBL) domain; and the USP9X Gly1890 is in UCH (Ubiquitincarboxyl-terminal hydrolase) domain, just eleven amino acids from the proton acceptor in the active site, and in between the catalytic Cys and His motifs that form the catalytic domain, suggesting that both mutations likely alter normal USP9X function. Rare coding variants were sought in the cohort of 284 males with epilepsy and the male exomes in the NHLBI exome variant server EVS (see web resources) in a one-tailed Chi-squared analysis. For the combined African American and European American cohorts, we found a statistical association (p = 0. 0350) between USP9X rare coding variants and epilepsy (S2 Table). This further supports the assertion that USP9X may be a novel epilepsy gene. Our studies suggest USP9X-mediated de-ubiquitination stabilizes PRICKLE proteins in the nervous system. These results are consistent with results in other tissues, where faf RNAi reduced Prickle levels in Drosophila apicolateral junctions, [33] as well as studies showing USP9X stabilizes several proteins involved in various aspects of neuronal development and cancer. [12] Notably, USP9X was shown to stabilize oncoproteins (e. g. , MCL-1 and p53), sparking interest in modulating USP9X in tumors, with promising results. [24] Here, we showed that human USP9X mutations are associated with epilepsy. Our data corroborate other evidence showing USP9X mutations in males with co-morbid epilepsy conditions (e. g. , intellectual disability and ASD). In addition to USP9X, several other members of our identified Prickle-interactome were already implicated in seizures, intellectual disability, and ASD. (For example, a TANC2 variant in a patient with intellectual disability and febrile seizures[34] and in a patient with ASD, [35] and a SMURF1 variant in a patient with epileptic encephalopathy). [36] PRICKLE mutant humans, mice, zebrafish, and Drosophila all exhibit seizures. [3,5] In flies, the prickle seizure phenotype can be genetically suppressed by reducing Faf (creating transheterozygotes) or pharmacologically rescued by treating pksple mutants with Degrasyn/WP1130. Although USP9X is genetically associated with seizures in both flies and humans, mutations are seizure-protective in the flies but seizure-inducing in humans. This opposing effect is likely because the identified human USP9X mutations are all protein-changing mutations that reside outside of the Prickle-interaction domain whereas the fly faf mutations that suppress seizures are all amorphic alleles. This suggests that, for USP9X, complete loss of function may have a different phenotype than protein coding alleles. Before our study, neither USP9X inhibition nor stimulation was identified as a potential anti-seizure pathway. Yet, given the evolutionarily conservation of the Prickle-pathway, our results suggest Degrasyn/WP1130 and other USP9X-modulating molecules should be pursued as novel therapeutic agents for patients suffering from seizures. The antibodies employed were rabbit polyclonal anti-BCR (Santa Cruz), rabbit polyclonal anti-TANC2 (Bethyl), polyclonal rabbit-anti USP9X (Abcam), mouse monoclonal anti-β-actin (Sigma), mouse monoclonal anti-Flag (Sigma), rat monoclonal anti-HA-peroxidase (Roche), rabbit polyclonal anti-dsRed from Clontech (detects mCherry) mouse monoclonal anti-Myc (Santa Cruz), rabbit polyclonal anti-GFP (Santa Cruz). Anti-rabbit and anti-mouse horseradish peroxidase (HRP) -conjugated secondary antibodies (Thermo Scientific). The agarose antibody-conjugated beads were protein A/G beads (Pierce), anti-Flag (Santa Cruz), and anti-GFP Flag-PRICKLE1 plasmid was as previously described. [4] The N-terminus of PRICKLE1 (aa 1 to 313) and the C-terminus of PRICKLE1 (aa 314 to 831) were each Flag-tagged on the C-terminus and cloned into EcoRI (5’) and KpnI (3’) restriction sites of the pcDNA3. 1 vector. The N-terminus (aa 1 to 317) and C-terminus of PRICKLE2 (aa 318 to 844) were also Flag-tagged on the C-termini and cloned into EcoRI (5’) and KpnI (3’) restriction sites of the same vector. Using previous studies as a guide, we cloned deletion constructs of murine Usp9x. [23] C1Usp9x aa 1216–2107 (with the catalytic domain) and C2Usp9x aa 2102–2560 (the carboxyl terminus) were cloned into the BamH1 and BSpe1 restriction sites of the pmCherryC1 plasmid (Clontech). The previously described HA-UbiquitinC plasmids were kindly gifted by Pedro Gonzalez-Alegre (University of Iowa, Iowa City). HEK293T[37] cells were cultured in DMEM (Dulbecco’s modified eagle medium; Gibco) supplemented with 10% FBS (fetal bovine serum) and 1% penicillin-streptomycin (Gibco). PC12 cells were cultured in collagen-coated plates in RPMI (Roswell Park Memorial Institute medium) supplemented with 5% FBS, 10% HS (horse serum), 1% penicillin-streptomycin, hygromycin (100μg/ml), and blasticidin (5μg/ml). Cells were maintained in a humidified 37°C, 5% CO2 incubator. Transgene expression and differentiation were induced by 1. 5μg/ml dox and 100 ng/ml NGF treatment in low serum RPMI medium (2% HS, 1% FBS) respectively. PC12 neuronal-like cell lines are used as a model system for investigating neuronal differentiation in culture. [10] As reported before, a PC6-3 (sub-clone of PC12 cells) clonal line, stably expressing the tet-repressor protein pcDNA6/TR (PC6-3/TR), was generously provided by Pedro Gonzalez-Alegre (University of Iowa, Iowa City). [1] Clonal cell lines inducibly expressing GFP, GFP-PRICKLE1 or GFP-PRICKLE2 were generated as previously described. [1,38] GFP, GFP-PRICKLE1, and GFP-PRICKLE2 cDNAs were cloned into EcoRI and XhoI restriction sites of pcDNA5 TO (Invitrogen). Plasmids were transfected into the PC6-3/TR cells, with Lipofectamine 2000 (Qiagen), according to the manufacturer’s instructions. Clonal cells stably expressing the transgenes were selected (medium supplemented with hygromycin at 100μg/ml) and blasticidin (5μg/ml) and screened for doxycycline inducibility by fluorescence microscopy and anti-GFP Western blots. Dox-treated and untreated PC12 cells expressing GFP, GFP-PRICKLE1, or GFP-PRICKLE2 were lysed in ice-cold, NET-100 buffer (Tris 50mM, NaCl 100mM, EDTA 5mM supplemented with protease inhibitor (1X EDTA-free Complete Mini-tabs protease inhibitor cocktail from Roche). Equal amounts of proteins were resolved by sodium dodecyl sulfate-acrylamide gel electrophoresis (SDS-PAGE) in 4–20% acrylamide gels (pre-cast gels, Biorad) and transferred onto a Polyvinylidene fluoride (PVDF) membrane for 3 hours at 0. 30 Amps. The membrane was blocked in 5% non-fat milk for 2 hours at room temperature followed by incubation in anti-GFP antibody at 1: 1000 dilution overnight at 4°C. The membrane was then washed in TBST (Tris-buffered saline with Tween-20) at room temperature, followed by incubation in (horseradish) HRP-conjugated goat-anti-rabbit antibody (1: 10 000) at room temperature for 2 hours. The blots were developed using ECL detection kit (Thermo Scientific) after washing, as per the manufacturer’s instructions. Signals were captured on x-ray films. Immunoprecipitations were carried out as previously described. [1,38] Differentiated, dox-treated GFP, GFP-PRICKLE1, and GFP-PRICKLE2 PC12 cell lines were lysed in ice-cold NET-100 buffer supplemented with a protease inhibitor. Lysates were immunoprecipitated overnight with GFP-conjugated agarose beads after 1 hour of pre-clearing in A/G agarose beads. After 5 X 5 minute washes in NET-100 buffer, immunoprecipitates were eluted in 2X Laemmli buffer at 100°C for 5 minutes. Equal volumes of immunoprecipitates were resolved by SDG-PAGE in 4–20% acrylamide gels and then silver-stained. Following electrophoresis, a gel with resolved GFP immunoprecipitates was incubated in fixing solution (50% methanol, 10% acetic acid) for 30 minutes at room temperature and washed overnight in water. The gel was then incubated in 100mL of sodium thiosulphate solution (0. 33g sodium thiosulphate/ 1L water) for 120 seconds, followed by 3 X 30-second washes in water. This was followed by incubation in silver nitrate solution (0. 2g silver nitrate/100mL water) for 30 minutes. The gel was then washed in distilled water for 3 X 60 seconds, and then incubated in developing solution (3g sodium carbonate, 50μl formaldehyde, 2mL sodium thiosulphate solution, 93mL water) until proteins bands became visible. The stop solution (7g EDTA/500ML water) was added to the gel and shaken for 10 minutes to stop further development. This was carried out as previously described. [1] SDS-PAGE. 20 μL of clarified, soluble GFP, GFP-PRICKLE1, and GFP-PRICKLE2 immunoprecipitates were added to denaturing, SDS-PAGE, loading buffer (containing glycerin, beta-mercaptoethanol, and SDS in Tris buffer) and boiled for five minutes in preparation for electrophoresis. Bio-Rad precast 4–20% Tris-HCl gradient SDS-PAGE gels were run at 150 V for 45 minutes. Gels were then stained with Bio-Rad Flamingo fluorescent stain and imaged using a UVP PhotoDoc-It UV Imaging System (Upland, CA) followed by LC-MS/MS as described. HEK293T[37] cells were transfected with Flag-PRICKLE1 or Flag-PRICKLE2 with Polyfect (Qiagen), according to the manufacturer’s protocol. Cells were lysed in ice-cold NET-100 buffer after 48 hrs of incubation and immunoprecipitated overnight with anti-Flag beads at 4°C. Beads were washed for 5 minutes x 5 times in ice-cold NET-100 buffer. Immunoprecipitates were resolved by SDS-PAGE on a 4–20% gel and then subjected to anti-USP9X (1: 500) immunoblot analysis. HRP-conjugated anti-rabbit secondary antibody was used at a 1: 2000 dilution. For mapping studies, HEK293T cells transfected with Flag-PRICKLE1, Flag-NPRICKLE1 or Flag-CPRICKLE1; Flag-PRICKLE2, Flag-NPRICKLE2 or Flag-CPRICKLE2 with Polyfect (Qiagen). Transfected cells were treated as described above. Lysates were also immunoprecipitated overnight with anti-Flag beads, eluates resolved by SDS-PAGE and then subjected to anti-USP9X Western blot analysis. Membranes were stripped and reprobed with anti-BCR antibodies (1: 1000) overnight. HRP-conjugated anti-rabbit secondary antibody was used at a 1: 10000 dilution. Blots were developed; and signals captured on X-ray films. Flag-PRICKLE1 only, Flag-PRICKLE1 + HA-UbiquitinC or Flag-PRICKLE1 + HA-UbiquitinC + mCherryC1Usp9x were transfected into HEK293T cells and incubated for 48 hours. Cells were lysed and PRICKLE1 immunoprecipitated overnight via the Flag tag and resolved by SDS-PAGE. Ubiquitinated and de-ubiquitinated PRICKLE1 was detected by anti-HA (1: 1000) Western blot analysis. The membrane was incubated in anti-HA-peroxidase at room temperature for 2 hours and washed for 4 X 10 minutes in TBST. Blots were developed; and signals captured on X-ray films. The process was repeated for PRICKLE2 where Flag-PRICKLE2 only, Flag-PRICKLE2 + HA-UbiquitinC or Flag-PRICKLE2 + HA-UbiquitinC + mCherryC1Usp9x were transfected. The process was repeated with full-length USP9X with and without the proteasome inhibitor LLNL (N-acetyl-L-leucyl-L-leucyl-L-norleucinal) at a concentration of 50μm overnight.

Epilepsy is a common disabling disorder characterized by seizures with complex genetic and environmental components. The absence of a definitive pathophysiology for epilepsy stymies the development of effective treatment strategies. In a small fraction of epilepsy cases however, single gene mutations may illuminate seizure-causing mechanisms, which may open the door to the discovery of broader, more effective therapeutic strategies. We have previously shown that disruption of Prickle genes in multiple species including humans, results in a predisposition to seizures. Those findings support Prickle in a seizure-preventing role and presents a possible anti-seizure therapeutic target. We identified the deubiquitinase Usp9x (ubiquitin-specific peptidase 9 X-linked) as a new Prickle binding partner which stabilized Prickle by preventing its degradation. In mice lacking the Usp9x protein in their forebrains, Prickle2 was barely detectable. In seizure-prone prickle mutant Drosophila, reducing fat facets (Drosophila usp9x) genetically or by a small-molecule usp9x inhibitor (Degrasyn/WP1130) suppressed the seizures. We also found 2 epilepsy patients harboring mutations in USP9X. Our findings demonstrate that inhibition of Usp9x can arrest prickle-mediated seizures, and variations in USP9X may predispose to seizures. From these studies, we have elucidated a seizure-inducing mechanism, identified a potential anti-seizure target, and a potential anti-seizure drug.

Abstract Introduction Results Discussion Materials and Methods

2015

Seizures Are Regulated by Ubiquitin-specific Peptidase 9 X-linked (USP9X), a De-Ubiquitinase

The capacity of organisms to tune their development in response to environmental cues is pervasive in nature. This phenotypic plasticity is particularly striking in plants, enabled by their modular and continuous development. A good example is the activation of lateral shoot branches in Arabidopsis, which develop from axillary meristems at the base of leaves. The activity and elongation of lateral shoots depends on the integration of many signals both external (e. g. light, nutrient supply) and internal (e. g. the phytohormones auxin, strigolactone and cytokinin). Here, we characterise natural variation in plasticity of shoot branching in response to nitrate supply using two diverse panels of Arabidopsis lines. We find extensive variation in nitrate sensitivity across these lines, suggesting a genetic basis for variation in branching plasticity. High plasticity is associated with extreme branching phenotypes such that lines with the most branches on high nitrate have the fewest under nitrate deficient conditions. Conversely, low plasticity is associated with a constitutively moderate level of branching. Furthermore, variation in plasticity is associated with alternative life histories with the low plasticity lines flowering significantly earlier than high plasticity lines. In Arabidopsis, branching is highly correlated with fruit yield, and thus low plasticity lines produce more fruit than high plasticity lines under nitrate deficient conditions, whereas highly plastic lines produce more fruit under high nitrate conditions. Low and high plasticity, associated with early and late flowering respectively, can therefore be interpreted alternative escape vs mitigate strategies to low N environments. The genetic architecture of these traits appears to be highly complex, with only a small proportion of the estimated genetic variance detected in association mapping. Most organisms experience environmental heterogeneity. For mobile organisms, adverse environments can be avoided by migration and habitat selection. Immobile organisms, such as higher plants, can avoid adverse environments by early reproduction and survival as seeds, or they may mitigate negative environmental impacts by physiological and/or developmental adjustments. It is therefore not surprising that plant development is so remarkably plastic, with a single plant genotype able to give rise to a wide range of phenotypes, depending on the prevailing environmental conditions [1]. There is a substantial body of theory, with some experimental support, concerning the circumstances under which plasticity is adaptive [2,3]. Key factors include the temporal and spatial scales of heterogeneity in the environment, which affect how well future conditions can be predicted from current environmental cues. For example, environments that are highly stable and therefore highly predictable, or environments that change too rapidly and stochastically relative to developmental responses for robust future prediction, may favour phenotypic canalisation. In contrast slower and more predictable environmental variation may favour plasticity [4]. Therefore, the question of whether or not plasticity is adaptive in nature is a complex one, the answer to which depends on the nature of the trait and its relationship with fitness, the costs associated with being plastic, the frequency and predictability of changes in the environment, and the amount of genetic variation for plasticity in populations [3,5, 6]. In this context, it is interesting that closely related species in the plant kingdom can show widely differing degrees of phenotypic plasticity [4,7–11]. Quantifying the genetic variation in the plastic responses of a species and how it relates to fitness traits may help in understanding the ecological and adaptive significance of phenotypic plasticity more broadly. Here, we use the annual model species Arabidopsis thaliana (Brassicaceae) as a system to dissect the genetics of shoot branching plasticity in response to nitrate. Arabidopsis is an ideal system for the study of plasticity. Firstly, the availability of many inbred lines from a wide geographic range provides ample genetic material for quantitative genetic studies [12]. Secondly, because these lines are highly inbred, they facilitate the study of plasticity at the genotype-level (rather than at the population-level), since the same genotypes can be grown in different environments. Because Arabidopsis is a natural selfer with high levels of inbreeding in wild populations, trait variance and covariance estimates from empirical studies should be less susceptible to changes due to artificial inbreeding as is seen in other model systems [13,14]. Finally, several studies in this species revealed substantial genetic variation in plasticity for traits such as flowering time, height, shoot branching and silique number in different growth conditions [15–23]. We focus on the plasticity of shoot branching and its relationship with other morphological and life-history traits. The major determinant of branch number, and particularly its plasticity, is the degree of activity of axillary shoot apical meristems, laid down in the axil of each leaf as they form on the primary shoot apical meristem. Shoot apical meristems can remain dormant, restricting the shoot system to a single axis of growth, or they can activate to produce a branch, reiterating the development of the primary axis, and allowing the possibility of higher order branches. Axillary meristem activity is regulated by diverse inputs, including environmental factors such as nutrient availability, shading, and damage to the primary shoot apical meristem [24]. In addition, developmental inputs such as the position of an axillary meristem along the primary axis and the phase of growth of the plant (e. g. vegetative vs floral) have a profound effect on their activity. These inputs must be integrated to deliver an overall branching habit tuned according to the plant’s local environment, and there is compelling evidence that plant hormones are central to this integration. In Arabidopsis, a network of at least three interacting hormones—auxin, cytokinin and strigolactone—is required for the shoot branching response to nitrate supply [25,26]. Wild-type (Col) Arabidopsis plants grown under nitrate sufficient conditions produce more branches than those grown under nitrate deficient conditions. This is associated with a higher root biomass fraction in nitrate deficient conditions, a trait presumably associated with nitrate foraging [27]. Strikingly, plants deficient in cytokinin synthesis or signaling constitutively adopt branching levels similar to those of wild-type plants on low nitrate, whereas plants deficient in strigolactone synthesis or signaling constitutively adopt a high branching phenotype [25,26]. Thus, plasticity in response to nitrate supply depends on the hormone network, with constitutively extreme phenotypes associated with either low cytokinin or low strigolactone. This would suggest a mechanism for variation in branching plasticity in nature in which plants with low plasticity adopt these extreme phenotypes due to compromised strigolactone or cytokinin biology. However, the roles of these hormones in shoot branching have largely been elucidated using null alleles, which simultaneously affect both the activities of the genes in question and any ability to modulate these activities dynamically in response to environmental cues, making it difficult to assess their likely roles in variation in plasticity in nature. To address whether and how plasticity varies in natural genotypes, we have analysed shoot branching responses to nitrate supply in two populations of Arabidopsis: a set of recombinant inbred lines from a mapping population derived from 19 accessions (the MAGIC lines [28]) and a set of natural accessions for which genome-wide genotype data are available [29–31]. Our results show significant natural variation in shoot branching plasticity in response to nitrate in both populations. We show that this plasticity correlates strongly with flowering time and has contrasting effects on fruit set depending on the available nitrate. This is consistent with a continuum of responses to N limitation, with escape through early flowering and mitigation through nitrogen foraging at the extremes. These traits are genetically complex, likely due partially to allelic heterogeneity at the relevant loci. To investigate whether there is natural genetic variation for shoot branching plasticity in response to nitrate supply, we analysed a set of 374 Multi-parent Advanced Generation Inter-Cross (MAGIC) Arabidopsis lines grown under nitrate sufficient (high N—HN) and nitrate deficient (low N—LN) conditions [25,28]. Plants were monitored daily and the flowering time of each plant recorded. When the first two siliques were full, we scored the height of the main inflorescence and total number of primary branches (shoots > = 1cm). Four to eight plants from each genotype were scored in each condition (median n = 8), allowing us to assess how much of the variation in each trait was due to genetic and/or non-genetic effects. We partitioned the variance of each trait using linear mixed models that included terms accounting for differences between genotypes (genetic effects), differences between N treatments (environmental effects), and genetic differences in the degree of response to N supply (genotype-by-environment, GxE, interaction). The effect of N treatment on the different traits was variable (Fig 1A). Most of the variation in flowering time was due to differences between genotypes (~83% together on LN and HN), with virtually no response to the N treatment (no plasticity). For height and shoot branching, the total genetic component of variance was lower, respectively ~52% and ~32%. However, for both of these traits, ~15% and ~39% of the trait variation respectively was related to the added effects of the nitrate environment and its interaction with the genotype (Fig 1A). Because of this, the response to N was variable across the MAGIC lines, with ~22% of the variation in shoot branching being attributable to GxE interaction. In fact, for this trait, the GxE component of variance was as large as the genetic component alone, with a significant contribution when compared with a reduced model that excludes it (ΔAIC = -419. 32; likelihood ratio test p-value < 10−6); S2 Table), suggesting extensive genetic variation for shoot branching plasticity in these populations. Using these variance estimates from our models, we calculated the coefficients of variation for each trait’s component, to allow comparison among them (Fig 1B). This revealed that the largest relative variation was in shoot branching, whereas flowering time had comparatively little relative variation in our dataset. This is likely due to the fact that we worked primarily with rapid cycling lines that do not require vernalisation. GxE interactions affect the overall variance of a trait in such a way that the mean trait value for a genotype in one environment is a bad predictor of the mean for that trait in another environment. For shoot branching, this can be readily seen as a low correlation between the number of branches for each line under the two nitrate treatments (Fig 2A), with an estimated genetic correlation of only ~0. 33 (Fig 1A). By contrast, the flowering time for each line on HN and LN are very strongly correlated, as expected from the lack of plasticity we observed in this trait (Fig 2B), and the very high genetic correlation >0. 99 (Fig 1A). To assess whether natural accessions of Arabidopsis showed the same trends, 278 accessions were grown as described for the MAGIC lines. The same phenotypic traits were scored, but in addition the total number of secondary shoots was re-counted when the first siliques started to senesce, and the total number of siliques was scored as a proxy for reproductive fitness [17,32]. As expected, given the longer times available for branch growth, branch numbers on HN and LN were typically higher at senescence stage than at the 2-silique stage, with a strong positive correlation between the two developmental stages (S1 Fig). On both HN and LN, branch number correlated with silique number (S2 Fig), suggesting that branch number contributes to reproductive fitness, as observed in previous studies [15,17,33,34]. We note however that the number of seeds per silique can vary, associated with variation in silique length [35–38], which likely weakens this correlation between fruit number and seed number. Similarly to the MAGIC lines, there was virtually no plasticity for flowering time, with the variation in this trait being primarily due to genotype (Fig 1). This lack of plasticity is reflected in the strong positive correlation between flowering time for each line on HN vs LN (Fig 2D). Also, as for the MAGIC lines, a substantial proportion of the variation in branch number and height is due to N supply, and ~17% and ~8% of the variation respectively is estimated to be due to GxE interaction (Fig 1). This GxE interaction results in weak correlation between branch numbers on the two N treatments (Fig 2C), with a significant GxE effect assessed by comparison with a reduced model without this component (ΔAIC = -237. 95; likelihood ratio test p-value < 10−6); S2 Table). At the senescence stage, the variance was similarly partitioned for the number of branches and the number of siliques, consistent with the correlation between these two traits (S2C Fig). To analyse shoot branching plasticity directly, we calculated the shoot branching response to N supply in each line as the difference between the mean number of secondary shoots formed on HN vs LN. Under HN conditions, for both the MAGIC lines (Fig 3A) and the natural accessions (Fig 4A), there was a strong positive correlation between total number of secondary shoots and shoot branching plasticity, whilst on LN there was a negative correlation between these traits. This can be clearly seen when plotting the mean branch numbers of the 25 most and least plastic lines from the two populations on HN and LN (Figs 3B and 4B). Genotypes that are highly branched on HN respond strongly to N deprivation, resulting in a very low branch number on LN; whilst less responsive lines typically have a moderate number of branches both on HN and LN. Thus, for shoot branching response to N, high plasticity is associated with phenotypic extremes, while low plasticity is associated with a constitutively intermediate phenotype. This contrasts with the phenotypes of shoot branching mutants where low plasticity locks plants into a constitutively a extreme phenotype [25,26]. As expected given the correlation between branch number and fruit number, when the natural accessions were grown on HN, there was a positive correlation between branching plasticity and silique number, whereas on LN the correlation was negative (Fig 4C). Thus the low plasticity lines have more fruits on LN, but the ability to protect branch numbers and hence fruit numbers on LN appears to come at the expense of the ability to exploit HN conditions by increasing branching and thus fruit set. Conversely, the high plasticity lines produce more siliques on HN. We investigated the relationship between shoot branching plasticity and flowering time. Overall, there was a low or no significant linear correlation between shoot branching plasticity and flowering time in the two populations, although there is a clear non-linear trend in the data (Figs 3C and 4D). The trend seems strikingly linear for earlier flowering lines and to quantify this relationship, we excluded lines flowering after 25 days on LN, leaving 258 MAGIC lines, and 266 natural accessions. This was further justified by the fact that later flowering individuals showed some growth defects, especially on LN (e. g. stunted or aborted growth of the main stem and high levels of anthocyanin accumulation in the leaves). In both populations there is a strong positive correlation between the two traits (Figs 3C and 4D) with non-plastic lines flowering earlier than plastic lines. These data reveal a continuum, from lines that flower very early and produce a moderate number of branches regardless of N supply, to those that flower later and modulate their branch numbers according to N availability. The correlation is less strong in the natural accessions (compare Figs 3C and 4D), which in this experiment in general formed fewer branches, especially on LN. These results suggest alternative strategies for growth under N limitation. At one extreme there is a rapid exit, escape strategy where plants flower early and branch regardless of N supply. This strategy results in higher fruit numbers on constitutively LN. At the other extreme are the later flowering lines that adjust their branching according to N supply. This strategy results in higher fruit numbers on constitutively HN. The lines we assessed form a continuum between these extremes. The lack of response to N by lines at the low plasticity end of the spectrum could be due to an inability to sense nitrate. To test this hypothesis, we selected three lower plasticity lines (Sha and Hi-0 accessions and MAGIC. 11) and three higher plasticity lines (Rsch-4 and Tsu-0 accessions and MAGIC. 345) and assessed their primary nitrate response using a panel of seven nitrate responsive genes. Of these, six are up-regulated and one is down-regulated in response to nitrate resupply after a period of starvation (S3 Table; [39]). The standard lab accession, Col-0 (which has relatively low plasticity), was included for comparison. All the genes showed the expected nitrate response in all the lines, demonstrating that low plasticity lines are able to sense nitrate. For the genes tested, there was no clear correlation between the level of shoot branching plasticity and gene expression response to N supply (Fig 5). Rapid local transcriptional responses to nitrate are known to be distinct from plant N status responses [40–42]. We have previously shown that the shoot branching response to nitrate supply in Col is dependent on N status rather than nitrate per se [25]. Also, split root experiments have provided compelling evidence that N status assessment in Arabidopsis involves the shoot [41,42]. We therefore used reciprocal grafting experiments to assess whether the low plasticity and high plasticity syndromes were dependent on the shoot or root genotype. For the MAGIC lines, we used MAGIC. 11 as a representative low plasticity line and MAGIC. 345 as a high plasticity line. For the natural accessions, Sha (low plasticity) and Rsch-4 (high plasticity) were used. In both cases, the self-grafted and ungrafted controls reproduced the expected flowering time and plasticity phenotypes typical of low and high plasticity lines. In the grafts between genotypes, both the branching and flowering phenotypes were similar to those of the shoot parent (Fig 6). Thus, the shoot genotype determined the phenotype, with little evidence for any effect of root genotype on either flowering time or branching plasticity. We next investigated the genetic architecture of shoot branching and its plasticity. Compared with the other traits we measured, shoot branching has generally lower broad-sense heritability, ranging from ~25–35% depending on the nitrate treatment and population analysed (S3 Fig). The heritability for flowering time was generally the highest (~70%), except for the subset of early-flowering MAGIC lines (<25 days on LN), where it was considerably lower (~25% on LN and ~35% on HN). This might be due to the fact that the underlying genetic variability in the MAGIC line population is much lower (with only 19 ancestors) than that in the accessions, and restricting the analysis to only the earliest flowering lines likely removed some of the large effect loci from the remaining population. To identify QTL associated with each trait in the MAGIC lines, we focused on the early-flowering lines and used methods suitable for the analysis of multi-parent populations [43]. These methods differ from traditional bi-allelic SNP-based mapping methods, where the phenotype is associated with two genotypic classes, corresponding to the homozygous state of each allele. In the QTL mapping method used for MAGIC lines, the individuals’ genotypes at each locus are defined as probabilities that the allele derives by descent from each of the 19 founder accessions used to produce the mapping population [28]. This, therefore, is akin to a haplotype-based method of mapping, whereby the phenotype is associated with 19 possible genotypic states. In these MAGIC lines, variation in height is known to be largely due to the mutation in the ERECTA gene carried by one of the founding accessions, Ler-0 [28,44,45]. We could readily detect this QTL in our dataset, suggesting that the 258 lines used have sufficient power to detect associations with a simple genetic basis (S4 Fig). We also detected significant QTL for flowering time (Fig 7A), despite the lower heritability in this subset of lines when compared to the full set. There were two QTL coincident on HN and LN, both located in regions with genes previously implicated in flowering time regulation, such as VIP5 and FT on chromosome 1 and, among others, FLC, FY and CO on chromosome 5 [46–48]. Together, these QTL explain ~10% of the phenotypic variance in this trait (S1 Table). Therefore, even among these early flowering lines, there is variation in flowering time that can be explained by these QTL. We next mapped QTL for branch number under each N condition, as well as branching plasticity. Given the correlation between branch number and flowering time (Fig 3C), we expected some QTL to co-locate with flowering time QTL. Indeed, significant QTL were found on chromosome 5 for shoot branching on low N and shoot branching plasticity, which coincided with a QTL for flowering time (SB. LN. 5 and SB. Pl. 5 in Fig 7B). There was also a near-significant shoot branching plasticity QTL on the left arm of Chr4 that lies in the region of the flowering-related gene FRI (Fig 7B), which coincides with a non-significant peak for flowering time on LN (Fig 7A). These QTL are not significant when using flowering time as a covariate in the association model (Fig 7B, dotted line), suggesting that they are related to both traits, either epistatically or pleiotropically. There was a QTL for shoot branching on high nitrate at the end of Chr5 (SB. HN. 5 in Fig 7B), which remained even when using flowering time as a covariate in the model. This suggests a branching-specific association at this locus. We caution that this region of the chromosome suffers from poorer genotype imputations, with around half of the MAGIC line individuals having probability lower than 50% of assignment to a unique founder accession. We also found a QTL specific for shoot branching plasticity on chromosome 2 (SB. Pl. 2 in Fig 7B), which alone explains ~3% of the variance for shoot branching plasticity (S1 Table). To assess fully the independence of this GxE QTL from a common genetic effect across both nitrate treatments, we fitted the whole dataset simultaneously using a multi-trait model [20,49,50] (Fig 7C). Comparing the likelihood of models with and without an interaction term between nitrate and the genotype at each marker suggests that the QTL on chromosome 2 specifically controls GxE variation for shoot branching, but not common genetic effects across both nitrate treatments. Similarly, consistent with the covariate analysis, this model identifies the QTL on Chr5 that is related to the correlation with flowering time. Interestingly, when using flowering time as a covariate in the QTL mapping, two additional cryptic QTL for branch number on low N were identified (SB~FT. LN. 1 and SB~FT. LN. 3 in Fig 7B). We note that even though the QTL on Chr1 seems to coincide with the one for flowering time (FT. HN. 1 and FT. LN. 1 in Fig 7A), the peak SNP for this QTL is located ~2. 4Mb away from it, suggesting separate loci in this region are associated with each trait. In fact SB~FT. LN. 1 coincides with SB. HN. 1 and indeed this QTL is captured in the joint model (Fig 7C), suggesting a common effect in both nitrate treatments. Together, all the QTL identified for this trait explain ~10% of the trait’s added genetic and GxE variance (S1 Table). To explore whether similar loci could be identified in the natural accessions, we performed association mapping in 240 of the natural accessions used in the phenotyping experiments for which genotypes are available for 192 863 bi-allelic SNPs with >5% frequency [51]. Across all the traits, we found only one significant QTL for flowering time on HN in a region of Chr4, which has also been reported in other GWA studies for flowering time (Fig 8; [48,52]). Similarly to the MAGIC lines, we fitted shoot branching data from both nitrate treatments simultaneously using a multi-trait model to test for effects common across both nitrate treatments as well as GxE effects (S5 Fig) [20]. The result revealed a single SNP passing the 5% genome-wide threshold for a common genetic effect on the trait. The SNP was located on Chr5, between genes AT5G20680 and AT5G20670, both of unknown function. Generally, other SNPs neighbouring this SNP had high likelihood for no association, making it unclear whether this is a true or spurious association, as a correlated signal would be expected in neighbouring SNPs due to linkage (for example, the two closest SNPs within 1Kb had p ~ 0. 5 and p ~ 0. 8). A spurious association could be due to the fact that this SNP had relatively low minor allele frequency of 6. 5%. Furthermore, no QTL were identified using a multi-SNP approach, which takes advantage of increased power to detect associations based on local additive association signals around a focal window of the genome (S6 Fig) [53]. Overall, this suggests a low power to detect associations in our dataset. To investigate whether the failure to identify any QTL for the remaining traits was due to a lack of markers in linkage disequilibrium with causal loci, we estimated the heritability of each trait using a SNP-based relatedness matrix (h2GWAS; [54]), thus assessing the variance jointly explained by all markers used in the association test (S7 Fig). For flowering time h2GWAS ~ 0. 9, suggesting that our SNP panel captures most of the phenotypic variance for this trait, but this variance cannot be pinpointed to individual SNP loci (a case of “hidden” heritability). Shoot branching h2GWAS estimates were lower than our broad-sense heritability estimates, with the exception of branches at the senescence stage on HN. The apparent increase in h2GWAS for this trait between the 2-silique and senescence stage could be due to higher variability within genotypes at the 2-silique stage than at the senescence stage. The generally low h2GWAS values suggest that the markers do not capture the genetic component of the variance in this trait (a case of “missing” heritability). One explanation for this is that the markers used here are not in linkage disequilibrium with the causal loci. However, we also calculated h2GWAS using a panel of ~1. 7M imputed SNPs, which did not improve the result, suggesting this “missing” heritability might be due to other complexities of the genetic architecture of this trait that cannot be captured by h2GWAS. If shoot branching is a polygenic trait, our GWAS might be severely under-powered to detect low-effect loci due to the relatively low number of accessions used (240), when compared with other studies (nowadays on the order of one thousand [55]). Furthermore, Bonferroni-corrected thresholds are often over-conservative resulting in a high number of false negative results [56–58]. Therefore, in order to identify suggestive QTL, we defined a new genome-wide threshold by taking advantage of prior knowledge of flowering QTL on Chr4 (near the CCT gene) and Chr5 (near the FLC gene) [20,55,59]. This new threshold, at p < 10−5, allowed inclusion of these two flowering QTL on both LN and HN (orange points in Fig 8A). With this relaxed threshold we found 5 suggestive QTL for shoot branching (orange points in Fig 8B). Two of these were in the vicinity of QTL found in the MAGIC lines: the SNP on Chr1 for LN is 2Mb away from QTL “SB~FT. LN. 1”; the SNP on Chr3 for Plasticity is ~180Kb away from QTL “SB~FT. LN. 3”. This suggests that the new threshold might be picking biologically significant loci, and these tentative associations further confirm the hypothesis that shoot branching and its plasticity are complex polygenic traits. Another hypothesis to explain the failure of GWAS to find significant associations is that multiple alleles at a single locus might be associated with the trait of interest [20,51,60,61]. Because this allelic heterogeneity is not captured by bi-allelic SNP markers, it could lead to a failure to detect QTL. The MAGIC lines provide a good system to explore this allelic heterogeneity, because it is possible to infer the phenotypic effect of each parental genotype at a significant QTL from the phenotypic mean of MAGIC lines inferred to carry a particular parental haplotype at that locus [28]. For example, as mentioned above, the height QTL on chromosome 2 is known to be related to the large-effect null allele carried by the Landsberg erecta (Ler) accession, which is apparent in the predicted allelic effects on height at this locus in the MAGIC population (Fig 9). Besides the large effect Ler allele, there is further allelic heterogeneity captured in these MAGIC lines, which might be due to subtle effects of other alleles. Similarly, the Zu allele at FT. HN. 5 has a major effect on delaying flowering time compared to the other parents, again with substantial variation in the effect of the remaining alleles of around 1 standard deviation both above and below the respective mean. By contrast, the predicted effects of the parental alleles at the plasticity QTL detected on Chr2 (SB. Pl. 2) do not show a clear large effect allele, but rather a range of lower size effects within 1 standard deviation of the overall mean. This is consistent with a more subtle effect of multiple alleles at the locus affecting shoot branching plasticity, which might not be captured with simpler bi-allelic genotype associations. Theoretical analyses of the circumstances under which developmental plasticity is adaptive identify a number of important factors that influence the balance between the costs and benefits of plastic responses [2,5–7,62]. These include the spatiotemporal scales of environmental heterogeneity compared to the spatiotemporal scales over which plastic responses occur and the costs and benefits of those responses. Here we use shoot branching plasticity in response to N supply as a model system to investigate natural genetic variation in plasticity. This system is of interest because N availability in nature is known to vary extensively over short spatiotemporal scales [63,64] relative to branching responses; while the high costs of producing additional branches are balanced by the potential for high benefits through additional fruits and seeds. Thus costly shoot branching investment decisions must be made ahead of reliable information about the future availability of N. Optimal foraging strategy for N has been studied extensively in roots (reviewed in [65]). There is evidence to support sophisticated risk-benefit calculations underpinning root plasticity. For example, pea plants in which the root systems have been divided between two chambers will proliferate lateral roots into a chamber where the N supply is highly variable only if the supply to the other half of the root system, which was kept constant, was low [66]. While these studies demonstrate correlations between root behaviour and shoot biomass, plastic responses to nutrient supply in the shoot are in general less well characterised. Our results demonstrate that in natural accessions of Arabidopsis, and in a collection of MAGIC lines derived from 19 natural accessions, shoot branching plasticity correlates positively with flowering time and fruit set on high nitrate, but negatively with fruit set on low nitrate. In contrast, flowering time was relatively insensitive to N supply. Our results are comparable to those observed in a series of studies in Arabidopsis by Pigliucci and Schlichting, who found extensive variation in plasticity for shoot branching and other traits in response to nutrient levels [22,23,67]. In particular, a study of 37 families derived from three populations grown under high and low nutrient conditions showed similar trends to our findings, namely low plasticity for flowering time and variable plasticity for branching and height between families [67]. The relatively high branching of the low plasticity lines on low N, coupled with their earlier flowering can be interpreted as a rapid exit strategy in response to nutrient limitation. In contrast, the late flowering and low branching phenotype of the high plasticity lines on low N may represent an N foraging strategy [68]. Given the potentially high spatiotemporal heterogeneity of nitrate in soil [63,64], late flowering extends the time over which this nutrient could be encountered and captured by the plant. Indeed, late flowering has also been implicated as a phosphate foraging strategy [69]. One framework to interpret these data is to consider the relationship between phenotypic plasticity and the evolution of specialist and generalist lifestyles. Often, generalist species, which occupy a wide range of habitats, are associated with high plasticity (e. g. invasive species [6,70], although this may not be a general feature [9,10]). On the other hand, specialist species may be associated with extreme, stable environments, where low plasticity and extreme phenotypes might evolve. A well studied case is that of the shade avoidance response in plants. In many species, shading by neighbouring plants triggers stem elongation and a suite of other responses supporting share avoidance in a highly competitive environment [71]. While in some species shading by neighbouring plants triggers stem elongation and a suite of other plastic responses supporting shade avoidance, in other species that are adapted to shaded environments there is a lack of such response. For the shoot branching syndromes we identify, non-plastic lines could be considered as specialised for stably N-deficient soils, whereas plastic lines would be able to exploit a range of environments with variable N availability. However, rather than having an extreme branching phenotype, the low plasticity lines maintain a moderate branching phenotype on both high and low N, while the highly plastic genotypes make very few branches on low N and many branches on high N. Thus, in contrast to the shade avoidance example, for natural genetic variation in the shoot branching response to N supply, high and low branching extremes are associated with high plasticity. Across the two populations, there is a continuum of responses between the low and high plasticity extremes, correlating with flowering time. This could reflect a tradeoff between investing existing resources in the next generation and gathering more [3,8, 72]. In this context it is interesting that the ability to protect branching under low N appears to come at the expense of the ability to exploit high N, despite the developmental potential to do so. This suggests general N insensitivity. However, all the lines we tested are able to sense nitrate, as indicated by changes in transcript abundance for six nitrate responsive genes following N supply. This is similar to what is reported in other studies, in which variation in ability to response to N limitation was associated with only small changes in nitrate-responsive gene expression [73,74]. Consistent with this idea, we have previously shown that shoot branching responds to N sources other than nitrate and could therefore be a response to plant N status, in which shoot N plays an important role [25]. In this context it is interesting that reciprocal grafting experiments demonstrated that both the “low plasticity; early flowering” and “high plasticity; late flowering” phenotypes are determined by the shoot genotype. This contrasts to the branching phenotype of both strigolactone and cytokinin biosynthetic mutants, where wild-type roots can restore wild-type shoot branching to hormone deficient shoots [26,75,76]. This argues against the phenotypes of low or high plasticity lines being caused by constitutive changes in strigolactone or cytokinin biosynthetic capacity. Consistent with this interpretation, we have previously shown that low branching on low N is dependent on the plant hormones auxin and strigolactone [25], while high branching on high N requires a third hormone, cytokinin [26]. Double mutants defective in strigolactone and auxin synthesis/signalling are constitutively highly branched, whereas mutants defective in cytokinin synthesis/signalling show constitutively low levels of branching regardless of N supply. This contrasts with the moderate levels of branching associated with low plasticity in the populations we examine here. Nonetheless, since both strigolactone and cytokinin are synthesised throughout the plant, and their levels can be modulated by nutrient supply [77,78], a shoot specific effect on hormone synthesis linked to N status is possible. In this context it is interesting that the MAX3 strigolactone biosynthetic gene is located near the peak of the QTL for shoot branching plasticity detected in the MAGIC lines (SB. PL. 2 in Fig 7). In general, despite the fact that the shoot branching and flowering time traits we measured had substantial levels of heritability, we recovered relatively little of the variation in our mapping experiments. In the GWA analysis of the natural accessions, no SNPs significantly associated with any of the branching traits were identified and for flowering time only one SNP was identified in the region of the well-known FRIGIDA locus [48,52]. Even considering our stringent threshold for association significance, these results suggest that either there are many loci of small effect involved (e. g. [56,79]), and/or there are multiple alleles at each locus (e. g. [20,51,60]). Both of these genetic complexities affect the statistical power to detect associations, in particular in complex populations such as the ones used here [56,80]. Although we found some tentative associations when relaxing our significance threshold (based on prior knowledge of flowering time QTL), we caution that these may incur high false discovery rates and would require confirmation in studies with larger populations, or other advanced methods of association (e. g. genomic prediction and/or multi-marker methods [56,81]). Although our approach has the advantage of including a wide range of the natural genetic diversity in Arabidopsis, providing a broad picture of shoot branching GxE in this species, it comes at the expense of a lower statistical power compared to traditional bi-parental populations. Studies using bi-parental RIL populations are often able to identify many QTL for growth and physiological traits under different environmental conditions including N supply [82–84]. For example, a recent study using 4 Arabidopsis RIL populations under different water availability treatments was able to reveal a complex genetic architecture for several growth-related traits, which are, by their very nature, integrative of several developmental events in a plant’s life [85]. The results revealed several associations exclusive to only one of the RIL populations, suggesting that these QTL may not have been detected in traditional GWAS. Therefore, one possible way to dissect further the genetics of shoot branching GxE is to produce a RIL population specifically between a plastic and a non-plastic accessions. Consistent with this idea, mapping in the accessions was less effective than in the MAGIC lines, where there are fewer parental haplotypes involved [28,80,86]. For many of the loci where significant associations were detected in the MAGIC population, there was evidence for variable effects of the parental haplotypes on the traits, suggestive of allelic heterogeneity for some of these loci. Despite the relatively few loci identified, the mapping results from the MAGIC lines reveal some interesting features. First, there are significant peaks for shoot branching that are not significant or even clearly detectable for shoot branching plasticity. For example, two QTL for branch number on low N, SB~FT. LN. 3 and SB~FT. LN. 1 (Fig 7B), were detected when using flowering time as a covariate in the QTL model. There is no evidence of an effect of these loci on branching plasticity despite branch number on low N being inversely correlated with plasticity. The BRC1 and BRC2 genes, which have been implicated in branching and its plasticity lie within these two regions, respectively [87–89]. Conversely, there are peaks for branching plasticity that are not apparent when branch number is mapped, such as SB. Pl. 2 (Fig 7B and 7C). Together these data suggest that branch number and its plasticity can be tuned at least partially independently. Several significant peaks for flowering time were detected in regions of the genome known to include major flowering time regulators. In some but not all cases, these correlated with significant peaks for shoot branching traits, for example FT. HN. 5 and SB. Pl. 5 (Fig 7). The flowering time regions detected all include genes involved in season detection, such as FT, FLC and FRI [47,90]. This suggests the interesting possibility that branching plasticity may be seasonally controlled, with variation at these loci independently underlying variation in flowering time and plasticity. Indeed, a previous study using the outbred population from which the Arabidopsis MAGICs were derived revealed a pleiotropic role for the FRI gene, which besides controlling flowering time also affected the number of inflorescence nodes and associated branches of plants carrying recessive non-functional alleles [33]. This effect was dependent on FLC genotype, which is expected from the epistatic interaction between these two loci [91]. It is important to note that the lines we studied all flower rapidly, so the effects we see are primarily of relevance to a rapid-cycling lifestyle. FRI and FLC are typically studied in the context of vernalization requirement, so it is interesting that they may also contribute to flowering time in these early flowering lines. Although under our conditions flowering was not plastic, this trait is sensitive to seasonal and temperature changes [20,90,92]. It will therefore be interesting to understand how environmentally-induced changes in flowering interact with the branching architecture of plants and their response to nitrate. Our findings might also be of agronomic relevance, where breeding for increased nitrogen use efficiency is of importance [93]. There are several QTL studies looking at this issue in a range of crop species (e. g. in rice [94,95], wheat [96,97], barley [98,99], sorghum [100], maize [101]). Although the specific traits analysed vary across studies, the broad picture that emerges is the polygenic nature of yield-related traits, with dozens of candidate loci found across studies, often with GxE effects related to N availability. In this context, our study emphasises the importance of GxE in understanding the genetic architecture of such traits. Further, it illustrates that breeding efforts under non-limiting nitrate conditions might result in worse performing genotypes when fertilization is reduced. In fact, studies in maize have shown that yield improvements under low N conditions are lower if the cultivars were selected on high N, rather than directly on LN [93,102]. Dissecting the mechanisms behind these differences remains a challenge [93], and our study in a model organism may help to address some of these questions in the future. Overall, our work identifies intriguing associations between branching, its plasticity and flowering time, which may have adaptive significance. Our analysis suggests the hypothesis that a rapid escape strategy combining early flowering with uncoupling of shoot N status from branching suppression allows seed yield to be maintained in N-poor environments. This may provide tools to understand better shoot N status sensing, which is currently enigmatic. However, the genetic complexity of the natural variation we have identified suggests that selected bi-parental mapping populations may be more powerful in determining the underlying genetic basis for these traits and their association than the multi-accession approaches we used here. We used two Arabidopsis thaliana populations for our experiments: 297 natural accessions and 374 MAGIC (Multiparent Advanced Generation Inter-Cross) lines. The MAGIC lines are derived from 19 natural accessions that were randomly inter-crossed for 4 generations, followed by 6 generations of self-fertilisation to generate inbred lines, typically used for QTL mapping [28]. The natural accessions were obtained from the Nottingham Arabidopsis Stock Centre (NASC, www. arabidopsis. info accessed Dec 2018) and were selected from several collections [29–31]. Only accessions described to have a flowering time of less than 55 days were included in the experiments (data from [31,48]; Arabidopsis Biological Resource Center, http: //abrc. osu. edu accessed Dec 2018; own experimental data). For each line, seeds were sown on wet filter paper and stratified for five days at 5°C in the dark, and then transferred to 5. 5 cm diameter pots filled with low nitrate substrates consisting of 50% sand (Leighton Buzzard sand from WBB Minerals) and 50% Terragreen (Oil-Dri). The substrates were wetted with Arabidopsis thaliana salts (ATS) solution [103], containing either 9mM (high N treatment, HN) or 1. 8 mM NO3- (low N treatment, LN), which we have previously shown represent N-sufficient and N-deficient conditions for Col-0 [25]. After two weeks, plants were fed once a week with 10ml of nutrient solution per pot, and in-between watered with regular tap water as needed. In all experiments plants were grown under glasshouse conditions. For the QTL experiments plants were grown in the summers of 2008 (MAGIC lines) and 2012 (accessions). For each line, eight replicates were grown on each nitrate treatment, which were randomly allocated to trays around the glasshouse. For our main natural variation experiments we measured flowering time, total branches and height for each plant. Flowering time was measured as the number of days from germination to the day at which the first flower buds were visible at the rosette centre. Total branches were counted as the number of secondary shoots (from the axils of rosette + cauline leaves) that were more than ~1cm in length. Height was measured as the length of main inflorescence stem. Total branches and height measurements were made when plants had formed two full siliques (2-silique stage). For 278 of the 297 accessions, we also obtained measurements at a later stage when plants had at least two senescing siliques (senescence stage). Traits are reported as averages for each line, with each line represented by 4–8 replicate plants (median n = 8). Branching plasticity was calculated for each line as the difference between the mean number of secondary shoots formed on HN vs LN. This measure was chosen taking into consideration the biology of shoot branching. It may seem attractive to normalise the number of branches to the number of nodes, and thus the total number of possible branches. Similarly, branching plasticity could be expressed as a proportional change in the number of branches. These measures were rejected based on the fact that branching occurs in a strict basipetal sequence and nitrate supply modulates the stopping point of that sequence [25,104]. Since branch activation at any one node is therefore highly dependent on its position along the primary axis and on the behaviour of the buds at more apical nodes, proportional measures of branch activity are inappropriate. Node number does provide the upper bound for primary branch number, but this is seldom achieved except in extreme branching mutants. To allow plasticity comparisons across traits, we’ve also calculated a relative metric, the “relative distance plasticity index” adapted from [105]. For each line, we calculated all pairwise differences (between replicates) of the trait value on HN and LN and then divided it by the respective pairwise sum of those values. This division ensures the plasticity measure is unitless, allowing comparisons across traits on different scales. These pairwise scaled plasticity differences were then summed and divided by the number of pairwise comparisons, to get an average scaled plasticity. This results in a metric that varies between -1 and 1, with zero indicating no plasticity. The expression of primary nitrate responsive genes was analyzed by RT-qPCR in four natural accessions (Shahdara, Hi-0, Rsch-4 and Tsu-0), two MAGIC lines (MAGIC. 11 and MAGIC. 345) and the standard laboratory line Col-0. For each sample, 10mg of seeds were surface sterilized and stratified for 4 days at 5°C and then transferred to 25ml of liquid ATS in which nitrate was replaced by 0. 5mM ammonium succinate. The seeds were left to germinate and grow on a shaker (100-120rpm) in a controlled environment room (16h light/8h dark, 17–21°C). After 10 days, the seedlings were treated with 5mM KNO3 or KCl for 2 hours. After this treatment, they were quickly dried and flash frozen in liquid nitrogen, and stored at -80°C until RNA extraction. Total RNA was extracted using the RNeasy Plant mini kit including DNase I treatment (Qiagen), following the manufacturer’s instructions. RNA was quantified using a NanoDrop 1000 and 1 μg was used to produce cDNA using Superscript II (Invitrogen), following the manufacturer’s instructions. qPCR reactions were prepared using LightCycler 480 SYBR Green I Master (Roche), with 5ng of cDNA in 20μl reactions, following the manufacturer’s instructions. Reactions were performed in a LightCycler 480 II (Roche) machine and Cp values were determined based on the “second-derivative maximum” method implemented in the manufacturer’s software. There were two biological replicates for each line, with three technical replicates each. The Cp values of the technical replicates were averaged for each biological replicate and used in subsequent calculations. For each sample, the transcript levels of the primary nitrate response genes were normalised relative to the mean Cp value of two reference genes: APX3 (AT4G35000) and UBC9 (AT4G27960). Finally, we estimated the relative expression of those genes in the treatment (KNO3) relative to the control (KCl) conditions. These estimates were made using the ΔΔCp method, assuming equal primer efficiency [106]. All primers are listed in S3 Table. Two pairs of lines were used for grafting: two MAGIC lines (MAGIC. 11 and MAGIC. 345) and two accessions (Shahdara and Rsch-4). For each of these pairs we made four pairwise grafting combinations: two autografts and two allografts (one in each direction). Plants were germinated on ATS medium containing 0. 8% bacto-agar and either 9mM or 1. 8mM of NO3-. From thereon, the grafting experiment was performed as described in [26]. Flowering time and total branches were measured at silique stage as described above. 7 to 19 replicates (median n = 13) were sown for each graft combination, along with the ungrafted parents and the whole experiment was replicated twice. The lines used in these experiments represented the two ideotypes of focus: low plasticity and early flowering (MAGIC. 11 and Shahdara) and high plasticity and later flowering (MAGIC. 345 and Rsch-4). To assess whether each ideotype’s characteristic phenotype was mainly shoot or root driven, we tested the hypothesis of no effect of root and shoot ideotypes on each trait using a mixed model ANOVA. This model included fixed terms for nitrate treatment, shoot ideotype, root ideotype and experiment (this was included as a fixed rather than random term since there were only two levels for this factor). We further included interaction terms between nitrate and each of the root and shoot ideotypes to account for their contribution to the trait’s plasticity (ideotype-by-nitrate interaction). Finally, we included a random term for each graft’s ID (to account for variation in the base level, or intercept, of the trait for each particular graft combination) and a random slopes term for nitrate (to account for the specific plasticity of each graft combination). The mixed model was fitted with the statistical program R [107], using the lmer function of the lme4 package [108]. The hypothesis of no effect of root and/or shoot ideotypes along with the interaction of these terms and nitrate were tested using Wald F tests as implemented in the Anova function in the car R package [109]. We estimated broad-sense heritabilities based on replicate measurements of each genetic line. Strictly speaking, this is a measure of “clonal repeatability”, but in a randomized experiment like ours it should give a good estimate of the degree of genetic determination of the trait, i. e. its broad-sense heritability (p. 123 in [110]). Briefly, we used linear mixed models to partition the phenotypic variance into between-line (genetic) and within-line (residual) components. Broad-sense heritability was calculated by dividing the genetic variance by the total variance estimated from the model. Heritability estimates were obtained separately for each population and nitrate treatment. Confidence intervals for these estimates were obtained using a parametric bootstrap approach [111]. In summary, we simulated 1000 sets of phenotype data based on the fitted model and estimated broad-sense heritability for each. A 95% confidence interval was obtained by taking the 0. 025 and 0. 975 quantiles of the heritability distribution of simulated phenotypes. To partition the variance into genotype, environment and genotype-by-environment (interaction) components, we fitted a more complex random slopes linear mixed model [112]. We included nitrate treatment as a fixed effect, genotype ID as a random effect and a random slopes term for the nitrate-by-genotype interaction. In more detail, the GxE mixed model fit to each trait was Yij=β0+β1NITRATE+u0j+u1j+ϵij where: Yij is the trait value for the i-th individual from the j-th genotype (MAGIC line or accession ID); β0 is the intercept of the model, which in our specification is the mean on LN; β1 is the response when on HN; NITRATE is a dummy variable indicating whether the individual was grown on HN; u0j is the random term for varying intercepts of the j-th genotype (genotype-specific average on LN); u1j is the random term for varying slopes of the j-th genotype (the genotype-specific response to nitrate, or GxE component); ϵij is the residual term with ϵ∼Normal (0, σe2). The random part of the model has the following variance-covariance specification: [u0ju1j]∼Normal (0, Ωu) Ωu=[σu02σu01σu12] Where: σ2u0 is the variance of the trait on low nitrate; σ2u1 is the variance of the trait responses on HN (the GxE component); σu01 is the covariance between the two. The covariance parameter was used to calculate the correlation of the trait values between LN and HN. The estimates from the model are presented in S2 Table for each trait. We also compared this full model with a reduced model that excluded the GxE component: Yij = β0 + β1NITRATE + u0j + ϵij. We assessed differences between the models using a likelihood ratio test (to obtain a p-value) and difference in the Akaike Information Criterion (where negative values indicate a loss of information in favour of the more complex model). In all cases, linear mixed models were fitted with R [107], using the lmer function of the lme4 package [108] and variance components were extracted using custom scripts. Flowering time data were log-transformed and total siliques data were square-root-transformed to reduce distributional skews and heteroskedascitity, thus improving the model’s diagnostics. Despite the fact that shoot branching is measured on a discrete scale (count data), we did not observe a strong relationship between the mean and variance across samples as is expected with count data, typically modelled using Poisson likelihood models. For this reason, we modelled our branching data using a normal likelihood function assuming homogeneous variance. Association (QTL) mapping in the MAGIC lines was performed using the R/qtl2 package [43] and a custom R data package containing the genotype data in a suitable format for analysis (available at https: //github. com/tavareshugo/atMAGIC). In summary, for each of the 1254 available markers, the probability of ancestry of an individual’s genotype at that marker was inferred using the function qtl2: : calc_genoprob (), assuming a 1% genotyping probability error. The qtl2: : scan1 () function was then used to fit the QTL model to each marker for each trait analysed. For shoot branching, we also fitted a model that included flowering time as a covariate (to account for the correlation between these traits). We used a 5% genome-wide significance threshold obtained by permutation using the qtl2: : scan1perm () function. The founder accession’s effect at each candidate QTL was estimated using the qtl2: : scan1blup () function following [43]. These analysis used the average trait value for each line in each nitrate treatment (n = 4–8 replicates each). For shoot branching, we also fitted a more complex “multi-trait” model, to assess the independence of GxE QTL from common effect QTL [20,49,50]. This model was similar to the variance partitioning model detailed above (a random slopes mixed model), but with an added term to account for the genotype of each MAGIC line. Due to the model complexity, we converted the founder genotype probabilities obtained from R/qtl2 to a single genotype value corresponding to the founder allele with maximum probability at each marker for each individual (i. e. the genotype variable was a factor with 19 levels, corresponding to each founder accession). LOD scores were obtained for two model contrasts: the full model compared to a genetic model with no genotype-by-nitrate interaction (GxE) term; the genetic model compared to a null model (no genotype term). We obtained a 5% genome-wide threshold by permutation of the genotype data. This analysis was done using custom R scripts (see data availability section). Finally, we estimated the variance explained by each marker by comparing the genetic variance using the mixed models just described with a model excluding the QTL marker as a predictor variable, similarly to [113]. Association (GWAS) mapping in the accessions was performed using the—mlma-loco function in GCTA 1. 26. 0 [114]. This performs an association test for each SNP using a linear mixed model that includes a random term to account for population structure. This is achieved by using a SNP-based relatedness matrix to model the variance-covariance structure between genotypes in the population. We excluded the marker being tested from the relatedness matrix using the “leaving-one-chromosome-out” (LOCO) method implemented in GCTA, which should increase the power to detect associations. SNP genotypes for accessions were obtained from the 250K dataset of [51] (available at http: //github. com/Gregor-Mendel-Institute/atpolydb, last accessed Jul 2019), which were converted to plink format (www. cog-genomics. org/plink/1. 9/formats, last accessed Jul 2019) using a custom python script. SNPs with minor allele frequency below 5% were discarded, leaving 192863 biallelic SNPs. A 5% genome-wide significance threshold was obtained by Bonferroni correction. The “—reml” function in GCTA 1. 26. 0 was used to estimate the proportion of phenotypic variance explained by the SNPs used in the GWAS, referred to as GWAS heritability, h2GWAS. We also estimated h2GWAS using a panel of 1 763 004 imputed SNP genotypes provided by Ümit Seren in the group of Magnus Nordborg. This is the set of imputed SNP genotypes used in the web application “GWA-Portal” (http: //gwas. gmi. oeaw. ac. at, last accessed Jul 2019). For shoot branching, we also fitted a “multi-trait” model using the limix 0. 7. 12 Python package [115]. Similarly to what was done with the MAGIC lines, this was used to test for genetic effects common to both nitrate treatments as well as GxE effects (following [20]). Finally, we also obtained “multi-SNP” associations using sets of SNPs within 10Kbp windows centered on each gene’s annotation, using the “—fastBAT” method in GCTA 1. 26. 0 [53]. In all cases genome-wide thresholds were obtained by bonferroni correction. Data analysis and visualisation were carried out using the statistical software R version 3. 4. 1 [107]. The meta-package tidyverse [116] was used for data manipulation and visualisation. Where relevant, approximate 95% confidence intervals for mean estimates are presented as 2x standard error of the mean (assuming data follow a normal distribution). Other specific analysis or statistical tests are described in the relevant sections above or in figure legends. All analysis scripts are provided with the supplementary data (see Data availability statement), but are also available with detailed information at: https: //github. com/tavareshugo/publication_deJong2019_Nplasticity

Many organisms adjust their development depending on environmental conditions. This is particularly striking in plants, with development constantly tuned throughout their lives. A good example is the modulation of shoot branching in response to external environmental cues. We characterised hundreds of genetically distinct Arabidopsis thaliana lines for their branching response to the availability of a key nutrient, nitrogen. We found that some lines adjusted their degree of branching according to the level of nitrogen available, whereas others did not. These latter low plasticity genotypes constitutively produce an intermediate number of branches and also flower earlier than the more plastic genotypes that responded to the nutrient treatment. One interpretation of these results is that flowering time and branch number are traded off in some way, reflecting at the extremes alternative strategies to cope with low nitrogen availability: an escape strategy of early flowering, and a mitigation strategy involving additional nutrient foraging. Using quantitative genetics methods, we found that the genetic basis for this response is likely to be complex, although not intractable when analysis methods that take into account genetic variability at the relevant candidate loci are used.

Abstract Introduction Results Discussion Materials and methods

genome-wide association studies chemical compounds quantitative trait loci population genetics brassica nitrates model organisms experimental organism systems genome analysis population biology plants flowering plants arabidopsis thaliana research and analysis methods genomics genetic polymorphism animal studies chemistry genetic loci eukaryota plant and algal models heredity genetics biology and life sciences physical sciences computational biology evolutionary biology organisms human genetics

2019

Natural variation in Arabidopsis shoot branching plasticity in response to nitrate supply affects fitness

Hypoxic microenvironments are generated during fungal infection. It has been described that to survive in the human host, fungi must also tolerate and overcome in vivo microenvironmental stress conditions including low oxygen tension; however nothing is known how Paracoccidioides species respond to hypoxia. The genus Paracoccidioides comprises human thermal dimorphic fungi and are causative agents of paracoccidioidomycosis (PCM), an important mycosis in Latin America. In this work, a detailed hypoxia characterization was performed in Paracoccidioides. Using NanoUPLC-MSE proteomic approach, we obtained a total of 288 proteins differentially regulated in 12 and 24 h of hypoxia, providing a global view of metabolic changes during this stress. In addition, a functional characterization of the homologue to the most important molecule involved in hypoxia responses in other fungi, the SREBP (sterol regulatory element binding protein) was performed. We observed that Paracoccidioides species have a functional homologue of SREBP, named here as SrbA, detected by using a heterologous genetic approach in the srbA null mutant in Aspergillus fumigatus. Paracoccidioides srbA (PbsrbA), in addition to involvement in hypoxia, is probable involved in iron adaptation and azole drug resistance responses. In this study, the hypoxia was characterized in Paracoccidioides. The first results can be important for a better understanding of the fungal adaptation to the host and improve the arsenal of molecules for the development of alternative treatment options in future, since molecules related to fungal adaptation to low oxygen levels are important to virulence and pathogenesis in human pathogenic fungi. The genus Paracoccidioides is a complex of thermodimorphic fungi, and are causative agents of paracoccidioidomycosis (PCM) a deep systemic granulomatous mycosis, endemic in Latin America [1,2]. Paracoccidioides spp. grows as yeast in host tissue and in vitro at 36°C, and as mycelium under saprobiotic and laboratory conditions (18–23°C). As the dimorphism is dependent on temperature, when the mycelia or conidia are inhaled into the host respiratory tract, the transition to the pathogenic yeast phase occurs [3]. Once in the lungs, epithelial cells and resident macrophages are the first line of defence against Paracoccidioides cells. Inside macrophages, the parasitic yeast form subverts the normally harsh intraphagosomal environment and proliferates [4]. Adhesion to and invasion of epithelial cells and basal lamina proteins may be required for the extra pulmonary haematogenous fungal dissemination to organs and tissues [1,3, 5]. To survive in the human host, fungi must also tolerate and overcome in vivo micro environmental stress conditions. Conditions such as high temperature, distinct ambient pHs, carbon and metal ions deprivation, and gas tension (high levels of carbon dioxide and low oxygen levels) induce several stress responses in the invading fungus [6–10]. In Paracoccidioides spp. , previous analyses have demonstrated differential responses to iron and zinc deprivation, oxidative and nitrosative stress and carbon starvation faced by the fungus during infection [11–15]. In addition, Paracoccidioides spp. yeast cells recovered from liver of infected mice and from infected macrophages alter their metabolism in order to adapt to the host using available nutrition sources [16,17]. It is well established that oxygen levels vary throughout the mammalian body depending on numerous factors including tissue type and presence or absence of an inflammatory response [18]. Oxygen levels in most mammalian tissues are found to be considerably below atmospheric levels (21%) [19,20]. Also, oxygen availability at the sites of inflammation is significantly reduced compared to surrounding tissues [21,22] since, in inflamed tissues, the blood supply is often interrupted because the vessels are congested with phagocytes or the pathogen itself [23,24]. Thus, it seems highly probable that hypoxic microenvironments are generated during fungal infection [25,26]. Mechanisms used by fungi to sense oxygen levels have been characterized [27]. An SREBP (sterol regulatory element binding protein) ortholog, previously characterized in higher eukaryotes [28–32], was first identified and characterized in the fission yeast Schizosaccharomyces pombe as an oxygen sensor [33,34]. Later, it was characterized in the human pathogenic fungi Cryptococcus neoformans and Aspergillus fumigatus [35–37]. In A. fumigatus, the SREBP homologue, SrbA, controls the expression of genes involved in biosynthesis of lipids, ergosterol, and heme [37,38]. Recently, a new transcriptional regulator of the fungal hypoxia response and virulence that genetically interacts with SrbA, named SrbB, was also characterized in A. fumigatus [39]. In S. pombe and C. neoformans the SREBP homologues also regulate enzymes in the ergosterol biosynthetic pathway under hypoxic conditions [34,35,38]. Oxygen levels are low in subsurface layers of organic matter in natural environments that are habitats of environmental pathogens such Paracoccidioides and Aspergillus [40–43]. In this context, studies regarding the responses of Paracoccidioides to hypoxia are of relevance and in this study are described for the first time. Up to now, hypoxia has not been described in the Paracoccidioides genus, representatives of thermally dimorphic fungi, in which responses to hypoxia remain to be investigated. We observed that Paracoccidioides yeast cells respond to hypoxia regulating the expression of proteins from diverse metabolic pathways. We also observe that species of the Paracoccidioides genus have homologues of the key regulator of hypoxia adaptation in fungi, SrbA. Paracoccidioides srbA was characterized using a heterologous genetics approach that confirmed the functional conservation of this protein in the hypoxia response. Paracoccidioides srbA (PbsrbA) is likely involved in hypoxia, iron adaptation and azole drug resistance responses, as observed by functional complementation of the srbA null mutant in A. fumigatus by PbsrbA. The obtained data, may improve the arsenal of molecules for the development of alternative treatment options since molecules related to fungal adaptation to low oxygen levels are important to virulence and pathogenesis in human pathogenic fungi. Paracoccidioides, Pb01 (ATCC MYA-826), was used in the experiments. The yeast phase was cultivated for 7 days, at 36°C in BHI semisolid medium added of 4% (w/v) glucose. When required, the cells were grown for 72 h at 36°C in liquid BHI, washed with PBS 1X, and incubated at 36°C in McVeigh/Morton (MMcM) medium as previously described [44]. Pb01 yeast cells were subjected to normoxia and hypoxia as previously described [37,45]. Normoxia was considered general atmospheric levels within the lab (~21% O2). For hypoxia, an incubator (Multi-Gas Incubator MCO-19M-UV, Panasonic Biomedical) was used. The chamber was maintained at 36°C, and kept at 1% oxygen level, utilizing a gas mixture containing 1% O2,5% CO2 and 94% N2. Paracoccidioides yeast cells viability was determined as previously described: the number of viable cells was determined at times of 0,6, 12,18 and 24 h by staining with 0. 01% (w/v) trypan blue in PBS1X [12,46,47]. All A. fumigatus strains were routinely grown in glucose minimal medium (GMM) with appropriate supplements at 37°C as previously described [45,48]. To prepare solid media, 1. 5% (w/v) agar was added, before autoclaving. For protein extraction and associated mRNA abundance experiments, 0. 5% (w/v) yeast extract was added to liquid GMM to increase hypha mass [45]. For hypoxia cultivations, an incubation chamber (Invivo2 400; Ruskinn) was used. The chamber was maintained at 37°C and kept at 1% O2,5% CO2, and 94% N2, controlled through a gas mixer (Gas Mixer Q; Ruskinn/Baker Company). Normoxia was also considered general atmospheric levels within the lab (~21% O2). Following Paracoccidioides yeast cells incubation under normoxia and hypoxia, in biological triplicates, cells were centrifuged at 1,500 x g, resuspended in 50 mM ammonium bicarbonate pH 8. 5 and disrupted using glass beads and bead beater apparatus (BioSpec, Oklahoma, USA) in 5 cycles of 30 sec, while on ice. The cell lysate was centrifuged at 10,000 x g for 15 min at 4°C and the supernatants for each condition were polled in equimolar amounts and subjected to the nanoscale liquid chromatography coupled with tandem mass spectrometry in 3 technical replicates. The proteins were quantified using the Bradford reagent (Sigma-Aldrich) [49]. Sample aliquots (70 μg) were prepared for NanoUPLC-MSE as previously described [11,15,50,51], with some modifications. Briefly, 50 mM ammonium bicarbonate was added and was followed by addition of 35 μL of RapiGEST (0. 2%v/v) (Waters Corp, Milford, MA). The solution was vortexed and then incubated at 80°C for 15 min. The disulphide bonds were reduced by treating proteins with 10mM D-L-dithiothreitol for 30 min at 60°C. The sample was cooled at room temperature and the proteins were alkylated with 200 mM iodoacetamide in a dark room for 30 min. Proteins were digested with trypsin (Promega, Madison, WI, USA, 1: 25 w/v) prepared in 50 mM ammonium bicarbonate, at 37°C overnight. Following the digestion, 10 μL of 5% (v/v) trifluoroacetic acid was added to hydrolyse the RapiGEST, followed by incubation at 37°C for 90 min. The sample was centrifuged at 18,000 x g at 6°C for 30 min, and the supernatant was transferred to a Waters Total Recovery vial (Waters Corp). A solution of one pmol. ul-1 MassPREP Digestion Standard [rabbit phosphorylase B (PHB) ] (Waters Corp) was used to prepare the final concentration of 150 fmol. ul-1 of the PHB. The buffer solution of 20 mM ammonium formate (AF) was used to increase the pH. The digested peptides were separated further via NanoUPLC-MSE and analysed using a nanoACQUITY system (Waters Corporation, Manchester, UK). Mass spectrometry data obtained from NanoUPLC-MSE were processed and searched against the Paracoccidioides database (http: //www. broadinstitute. org/annotation/genome/paracoccidioides_brasiliensis/MultiHome. html) using ProteinLynx Global Server (PLGS) version 2. 4 (Waters Corp). Protein identifications and quantitative data packaging were performed using dedicated algorithms [52,53]. The ion detection, clustering, and log-scale parametric normalizations were performed in PLGS with an ExpressionE license installed (Waters, Manchester, UK). The false positive rate (FPR) of the algorithm for protein identification was set to 4% in at least two out of three technical replicate injections. Using protein identification replication as a filter, the false positive rate was minimized because false positive protein identifications, i. e. , chemical noise, have a random nature and do not tend to replicate across injections. For the analysis of the protein identification and quantification level, the observed intensity measurements were normalized to the intensity measurement of the identified peptides of the digested internal standard. Normalization was performed with a protein that showed no significant difference in abundance in all injections [54] to accurately compare the expression protein level to normoxia and hypoxia samples. For 12 and 24 h, the proteins oxidoreductase 2-nitropropane dioxygenase and 40S ribosomal protein S5 were used as normalizing proteins, respectively (PAAG_01321 and PAAG_05484 from Paracoccidioides genome database http: //www. broadinstitute. org/annotation/genome/paracoccidioides_brasiliensis/MultiHome. html). Furthermore, only those proteins with a fold change higher than 50% difference were considered to be expressed at significantly induced/ repressed levels. Paracoccidioides, Pb01 yeast cells, were grown under normoxia and hypoxia for 12 and 24 h, in biological triplicates. Following that, cells were harvested by centrifugation at 2,000 x g for 5 min at 4°C and diluted in PBS buffer at 106 cells/ml. Cells were stained with Rhodamine 123 (1. 2 mM) (Sigma Aldrich) according to the manufacturer' s protocol and then washed twice with 1X PBS. Stained cells were observed under a fluorescence microscope (AxioScope A1, Carl Zeiss) and analysed with the 546–512 nm filter. Rhodamine fluorescence intensity was measured using the AxioVision Software (Carl Zeiss). The minimum of 100 cells for each microscope slides, in triplicates, for cells submitted to hypoxia and normoxia for 12 and 24 h were used to measure the rhodamine fluorescence intensity. The software provided the fluorescence intensity (in pixels) and the standard deviation of each analysis. Statistical comparisons were performed using the student’s t test and p-values ≤ 0. 05 were considered statistically significant. The amino acid predicted sequences were obtained from GenBank (http: //www. ncbi. nlm. nih. gov/) to Paracoccidioides Pb01 (XP_002794199); Pb03 (KGY15961); Pb18 (EEH47197); Aspergillus fumigatus (XP_749262); Schizosaccharomyces pombe (NP_595694); Cryptococcus neoformans (XP_567526) and Homo sapiens (P36956). The SMART tool (http: //smart. embl-heidelberg. de) [55,56] was used to search for conserved domain bHLH (basic helix-loop- helix leucine zipper DNA-binding domain) and Phobius (http: //phobius. sbc. su. se/) [57] and SACS MEMSAT2 Prediction software (http: //www. sacs. ucsf. edu/cgi-bin/memsat. py) [58] were used to depict transmembrane segments. The amino acid sequences from all proteins were aligned using CLUSTALX2 [59] to show a conserved tyrosine residue (indicated by asterisk) specific to the SREBP family of bHLH transcription factors. Following Paracoccidioides incubation under hypoxia and normoxia, cells were harvested, and total RNA was extracted using TRIzol (TRI Reagent, Sigma-Aldrich, St. Louis, MO, USA) and mechanical cell rupture (Mini-Beadbeater-Biospec Products Inc. , Bartlesville, OK). Total RNA was extracted and treated with DNase (RQ1 RNase-free DNase, Promega). After in vitro reverse transcription (SuperScript III First-Strand Synthesis SuperMix; Invitrogen, Life Technologies), the cDNAs were submitted to a qRT-PCR reaction, which was performed using SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) in a StepOnePlus Real-Time PCR System (Applied Biosystems Inc.). The expression values were calculated using the transcript that encoded alpha tubulin (GenBank accession number XP_002796639) as the endogenous control, as previously reported [11] and, when required, the data were presented as relative expression in comparison to the experimental control cells value set at 1. Relative expression levels of genes of interest were calculated using the standard curve method for relative quantification [60]. Briefly, for each of the three replicates of a sample, the average quantity (avg) was calculated of target cDNA interpolated from the standard curve, the standard deviation of the average (stdev), and the coefficient of variation (CV) according to the formula CV = stdev/ avg. Any outlier points (>17% CV) was removed and avg, stdev and CV were recalculated. For each sample, the gene of interest (GOI) was normalized to the reference gene (RG) for the sample according to the following equation: normalized value = avg GOI quantity/ avg RG quantity. The standard deviation (SD) of the normalized value was calculated according to the equation: SD = (normalized value) x square root (CV reference gene + CV gene of interest) 2. The resulting values were plotted as a bar graph of normalized value versus sample name or experimental treatment group, with the error bars equal to the SD, of the biological triplicates of independent experiments [60]. Standard curves were generated by diluting the cDNA solution 1: 5. Statistical comparisons were performed using the student’s t test and p-values ≤ 0. 01 were considered statistically significant. Regarding to A. fumigatus, wild type, ΔsrbA and reconstituted strains were cultured in liquid GMM under normoxia or hypoxia. Germlings and mycelia were collected with vacuum filtration and lyophilized, prior to homogenization with 0. 1-mm glass beads. Total RNA was extracted, treated with DNase, reversed transcripted to cDNA and submitted to a qRT-PCR reaction, identically to which was performed to Paracoccidioides. Oligonucleotides to amplify the srbA gene from A. fumigatus and Paracoccidioides were used in the experiments. The data were normalized using the A. fumigatus tefA reference gene [61]. Primers are depicted in S3 Table. The A. fumigatus strains CEA10 (wild type) and a srbA null mutant of A. fumigatus were used in the genetic complementation assays. This srbA null mutant was previously generated by replacement of the srbA coding sequence in A. fumigatus strain CEA17 with the auxotrophic marker pyrG from A. parasiticus as previously described [45,62,63]. To perform genetic complementation of the respective ΔsrbA, the Paracoccidioides Pb01 srbA sequence was amplified from Paracoccidioides genomic DNA as template and linked together with a fragment of the gpdA (glyceraldehyde phosphate dehydrogenase) gene from Aspergillus nidulans, used as promoter and a functional pyrG gene from A. parasiticus, used to select the transformed strains (gpdA+PbsrbA+pyrG). The fused product was used to perform fungal transformations. Generation of fungal protoplasts and polyethylene glycol-mediated transformation of A. fumigatus were performed as previously described [45,64]. Reconstituted strains were confirmed by screening using hypoxia chamber, conventional PCRs, Southern blots, qRT-PCRs and immunoblot analyses. All primers used are shown in S3 Table. In order to eliminate the chance of heterokaryons, each transformant was streaked with sterile toothpicks a minimum of twice, to obtain colonies from single conidia. All strains were stored as frozen stocks with 50% (v/v) glycerol at -80°C. Ten-well 10% Mini-Protean precast gel (Bio-Rad) was used for SDS-PAGE. Denatured protein was loaded (40 μg per well). After gel electrophoresis, protein was transferred to a nitrocellulose membrane (Hybond-C Extra; Amersham Biosciences). PbSrbA was detected on blots using the A. fumigatus SrbA 1–275 recombinant primary N-terminus antibody at a 1/27,000 dilution and an anti-rabbit alkaline phosphatase (AP) -conjugated secondary antibody raised in goat (Abcam) at a 1/5,000 dilution, as previously described [45]. Chemiluminescence was measured following incubation of blots with Tropix CPD Star substrate (Applied Biosystems) with Immun-star enhancer (Bio-Rad) using a FluorChem FC2 imager (Alpha Innotech). DNA was isolated from overnight liquid cultures of A. fumigatus. The mycelium was separated from the medium by filtration and glass beads were used to disruption. Additional purification steps were used to isolate the genomic DNA and Southern blot was performed using the digoxigenin labelling system (Roche Molecular Biochemicals, Mannheim, Germany) as previously described [45,65]. Briefly, 30 μg aliquots of genomic DNA were digested with HindIII and EcoRI to detect gpdA and pyrG, respectively. Restriction digests were separated on a 1% agarose gel and blotted onto nylon membranes. The concentration of the probes in hybridization solution was 50 ng/ml, and hybridization was carried out at 50°C. Membranes were washed in a final solution of 0. 1 SSC and 0. 1% (w/v) sodium dodecyl sulphate, at 68°C. Production of biomass was performed to wild type, ΔsrbA and reconstituted strain 1 (Rec 1) of A. fumigatus. A total of 108 cells of each strain were grown under iron starvation (−Fe) and iron sufficiency (0. 03 mM, +Fe) in liquid medium for 24 h, at 37°C. The cells were harvested by vacuum filtration and then lyophilized. The data represent the mean ± SD of biological triplicates and the values were normalized to the reconstituted strain. Statistical comparisons were performed using the student’s t test and p-values ≤ 0. 01 were considered statistically significant. P. brasiliensis yeast cells were grown in McVeigh/Morton medium (MMcM) [44] and the yeast cells were incubated at 36°C with shaking at 150 rpm. In order to analyse the kinetic of expression of PbsrbA, the cells were cultivated under iron deprivation or supplementation, using the iron chelator bathophenanthroline disulfonate (BPS; 50 μM; Sigma-Aldrich, Germany) or 3. 5 μM Fe (NH4) 2 (SO4) 2, respectively. Total RNA was extracted at 30 min, 1,3 and 24 h and the quantitative real time PCR was performed as cited above. In order to start the characterization of Paracoccidioides hypoxia response, we utilized a proteomics approach. The NanoUPLC-MSE [50,51] was previously used to map metabolic changes in Paracoccidioides at a protein level [11,13,15,17] and was also used in this study. After exposing the cells to normoxia (21% pO2) and hypoxia (1% pO2) and using a proteomics approach at time points 12 and 24 h, we observed significant differences in protein expression indicating that the fungus responds to hypoxia. As described in Lima and co-workers [15], a 1. 5-fold change was used as a threshold to determine positively and negatively differentially proteins. In total, 134 and 154 proteins presented different abundances in 12 and 24 h under hypoxia, respectively, compared to normoxia. In 12 h, the same number of proteins (67) were increased and decreased, upon hypoxia, compared to control (normoxia). At 24 h, 102 proteins were increased and 52 were decreased (S1 and S2 Tables). The adaptation mechanism of Paracoccidioides to hypoxia, as represented by biological processes, as deduced from increased and decreased proteins is shown in Fig 1. Proteins associated with several subcategories of metabolism were represented in both analyses as increased and decreased proteins. Some of them were less represented in 24 h of hypoxia such as nitrogen, purine nucleotide/ nucleoside/ nucleobase and phosphate metabolism. Proteins associated with energy depicted an interesting profile of abundance. Those involved with electron transport/ membrane associated energy conservation were enriched for reduced levels in 12 h of hypoxia, and levels were subsequently restored at 24 h of hypoxia (Fig 1, S1 and S2 Tables). To further assess this observation, we evaluated mitochondrial activity using rhodamine, a permeable lipophilic cationic fluorescent probe that accumulates in mitochondria and is distributed electrophoretically into the mitochondrial matrix in response to mitochondrial electric potential [66,67]. The rhodamine probe has been used to stain yeast cells [67], including Paracoccidioides [13,68]. Consistent with the proteomics data that suggested reduced mitochondrial activity, a low level of staining of rhodamine was observed in yeast cells during 12 h of hypoxia. At 24 h, the intensity of detection was restored, which is consistent with proteomics data (Fig 2). Additionally, proteins such as catalase, thioredoxin, chaperones and gamma-glutamyltranspeptidase were up-regulated in Paracoccidioides in hypoxia for 12 h (S1 Table) and could be associated with the altered mitochondrial activity. Our suggestion is that the fungus possibly induces ROS scavenging enzymes to protect the fungus against low oxygen effects that induces a strong reduction in electron-transfer reactions. In C. neoformans, several genes associated with the mitochondrial activity were identified as essential for hypoxic growth [69]. Proteins in the energy subcategories glycolysis/ gluconeogenesis, TCA cycle and GABA shunt were also differentially abundant under hypoxia. The glycolysis/ gluconeogenesis and GABA shunt were increased at 24 h of hypoxia. On the other hand, proteins of the TCA cycle were reduced at both time points (Fig 1, S1 and S2 Tables). The mechanisms of hypoxia adaptation are variable among fungi [18,70]. At transcript level, for example, genes involved with glycolysis were induced, while those involved with aerobic respiration were repressed in Candida albicans, a facultative anaerobe, submitted to hypoxia [71–73]. However, in the obligate aerobic yeast C. neoformans, a general lack of changes in glycolytic mRNA abundance was observed in response to hypoxia, and genes involved in mitochondrial function have been observed to be critical for the hypoxia response [36,74]. In the obligate aerobic mold A. nidulans, exposure to hypoxia results in an increase in glycolytic gene transcripts and the GABA shunt, which bypasses two steps of the tricarboxylic acid (TCA) cycle [75]. Transcriptome data from A. nidulans largely correlated with the proteomic profile, in which proteins in core metabolism and utilization of the GABA shunt was identified [76]. Similar results were found in A. fumigatus, upon short-term hypoxia as the GABA shunt was also induced [77]. On the other hand, cultures exposed to long-term hypoxia revealed increased abundance of proteins involved in glycolysis, respiration, pentose phosphate pathway, and amino acid and pyruvate metabolism [78]. Fig 3 depicts probable mechanisms used by Paracoccidioides to overcome hypoxic environments. It does not represent an integral model of how Paracoccidioides adapts to hypoxia, but from our point of view is an important source to start the understanding of how this fungus adapts to low oxygen levels. The abundance of some enzymes involved in acetyl-CoA production are up-regulated in 12 h of hypoxia compared to normoxia. The induction, for example, of the aldehyde dehydrogenase and long-chain specific acyl-CoA dehydrogenase enzymes suggest that the acetyl-CoA is produced via acetaldehyde and beta-oxidation pathway, respectively. Consistent with these data, proteins involved in glycolysis were decreased in abundance (S2 Table). Acetyl-CoA can be used as an alternative carbon source under these conditions (Fig 3). In fact, the expression of proteins related to glycolysis, acetyl-CoA production from pyruvate and citrate, TCA cycle and oxidative phosphorylation were reduced (Fig 3, S2 Table). At 24 h, the detected up- and down-regulated proteins could show additional changes in Paracoccidioides strategies to adapt to hypoxia. For example, proteins involved in glycolysis are now increased supporting pyruvate production. In addition, the GABA shunt is increased at 24 h of hypoxia (Fig 3, S1 Table). The increased abundance of two enzymes involved with the GABA shunt pathway, NADP specific glutamate dehydrogenase and succinate-semialdehyde dehydrogenase, support this hypothesis in Paracoccidioides. Reports have shown that GABA is generated from 2-oxoglutarate via glutamate through the actions of glutamate dehydrogenase and glutamate decarboxylase, and that GABA transaminase irreversibly transaminates GABA to succinic semialdehyde, which is then oxidized to succinate by succinic semialdehyde dehydrogenase [76,79,80]. Transcripts for this pathway are also up-regulated in A. nidulans and A. fumigatus, under hypoxia [75,77]. The GABA shunt is hypothesized to help organisms to avoid accumulation of high NADH levels in the absence of a terminal electron acceptor such as oxygen, and also contributes to glutamate formation [77]. This pathway is also described as an alternative route to the TCA cycle [75]. Interesting, the TCA pathway was down-regulated, based on protein levels of key enzymes (Fig 3, S2 Table), although the role of the GABA shunt in the fungal hypoxia response remains to be conclusively determined. Moreover, enzymes involved in beta-oxidation and in production of ergosterol precursor molecules were also up-regulated according to proteomic data, at 24 h (Fig 3, S1 Table). During Paracoccidioides hypoxia adaptation, the detection of the long-chain specific acyl-CoA dehydrogenase, for example, shows that the fungus activates the beta-oxidation resulting in acetyl-CoA, that could be involved in fatty acid and ergosterol production. The enzyme 3-hydroxybutyryl-CoA dehydrogenase yields 3-acetoacetyl-CoA that together to acetyl-CoA supports ergosterol synthesis. Our suggestion makes sense since acetyl-CoA is probably not produced by pyruvate, neither acetate nor citrate, since enzymes related to their metabolism are down regulated in our data (Fig 3, S2 Table). The relative expression level of the transcript encoding Pberg3 was determined by quantitative real time PCR (Fig 4). The gene Pberg3 encodes C-5 sterol desaturase, an enzyme involved in the late steps in sterol biosynthesis [74,81]. The data provide additional evidence that Paracoccidioides faces hypoxia and regulates ergosterol production, to compensate the effects of low oxygen levels. Several enzymatic steps in ergosterol biosynthesis are catalysed by iron and oxygen-requiring enzymes including that performed by Erg3 [74]. Also, the metabolism of fatty acids and ergosterol are increased in C. albicans, C. neoformans, A. fumigatus and A. nidulans in response to hypoxia and these molecules are required for the stability, fluidity and structure of the fungus plasma membrane [36,72–74,76,77]. On this way, the fungus might be remodelling the fatty acid content of membrane lipids to keep the membrane fluidity in hypoxia. Along with ergosterol’s role as a target to antifungal drugs, the understanding of the mechanisms that regulate ergosterol biosynthesis is of interest to biomedical research [82,83]. In S. pombe, A. fumigatus and C. neoformans, the SREBP proteins are effectors which sense changes in oxygen levels indirectly through alterations in ergosterol levels [33,35,37]. Therefore, we addressed the question whether Paracoccidioides also relied on an SREBP like protein to adapt to hypoxia. We hypothesized that Paracoccidioides hypoxia response could be, in part, regulated by a homologue of the SREBPs, an ancient family of regulators, associated with the hypoxic response in fungi [27,33–35,37,84]. In silico analysis using Genbank (http: //www. ncbi. nlm. nih. gov/) and Paracoccidioides genome databases (http: //www. broadinstitute. org/annotation/genome/paracoccidioides_brasiliensis/MultiHome. html) showed that members of the genus Paracoccidioides, including the isolate 01, contain homologues of SREBPs. We named the gene srbA (PbsrbA), and the accession numbers in the Paracoccidioides genome database are PAAG_03792, PADG_03295 and PABG_11212 for Pb01, Pb18 and Pb03 strains, respectively. The SREBP proteins are basic helix-loop-helix leucine zipper transcription factors with a conserved tyrosine residue, specific to this family. In addition, the SREBP present transmembrane domains, responsible for associating the protein with endoplasmic reticulum (ER). The Paracoccidioides spp. srbA genes contain those domains (Fig 5 and S1 Text) suggesting that they are an integral membrane protein which requires to be processed to release the N-terminus containing the bHLH DNA binding domain. In mammals, SREBPs are synthesized as inactive precursors on the endoplasmic reticulum (ER) membrane where they bind to the SREBP cleavage activating protein (SCAP) which mediates sterol-dependent regulation of SREBP activity. The SCAP protein interacts with another ER-resident protein, named INSIG, and other proteases that cleave into the first transmembrane segment, to release the N-terminal transcription factor SREBP, which translocates to the nucleus and regulates expression of genes required when cholesterol levels are low [27,28,38,85]. In fungi, some differences are detected in the SREBP processing illustrating that, while many aspects of SREBP regulation are conserved across organisms, others are not [45]. In general, the differences are involved with the SREBPs processing for their activation. In S. pombe and C. neoformans, SREBPs are regulated in part by proteolysis, although in S. pombe, this processing is dependent on a Golgi E3 ligase complex, encoded by dsc (defective for SREBP cleavage) genes and not homologues of human proteases, as found in C. neoformans [86–88]. In A. fumigatus, the processing is similar to that found in S. pombe involving the Dsc complex, required for cleavage of SrbA. The hypoxic adaptation and virulence of A. fumigatus require both, SREBP and its processing mechanism, demonstrating an important mechanism to fungal pathogenesis [37,45]. Paracoccidioides spp. , in contrast to S. pombe and in accordance with A. fumigatus, does not depict in the genome database homologues for SCAP protein. On the other hand, there is an apparent homolog to the INSIG protein (Table 1). Moreover, the Site-1 and Site-2 proteases homologues were not identified in Paracoccidioides spp. genomes, as found in S. pombe and A. fumigatus (Table 1). These findings reinforce the relevance of studying activation of SrbA in Paracoccidioides spp. To determine if PbsrbA responds to hypoxia, we first examined mRNA levels of the transcript in different oxygen conditions. The fungus significantly increases the levels of PbsrbA after 1 h upon hypoxia exposure in comparison to normoxia (Fig 6). These results suggest that PbsrbA may be involved in the hypoxia response in Paracoccidioides spp. and further analyses were performed to test this hypothesis. There are a reduced number of works relating the functional analysis of genes in Paracoccidioides in the last six years because the achievement of viable and stable mutants of Paracoccidioides spp. is a hard task [11,13,89–94]. This parsimony in functional analysis surely reflects the complexity of those studies in the genus Paracoccidioides, as well as in other pathogenic fungi. Due these limitations in molecular genetic analyses available in Paracoccidioides, we utilized a heterologous genetics approach to test our hypothesis. In order to test whether Paracoccidioides srbA was able to replace the A. fumigatus SrbA function, we introduced Paracoccidioides srbA (PbsrbA) under control of the gpdA (glyceraldehyde-3-phosphate dehydrogenase) promoter from A. nidulans into a previously characterized srbA null mutant strain A. fumigatus (ΔsrbA) [45]. Ectopic introduction of the Paracoccidioides srbA gene (PbsrbA) into ΔsrbA allowed us to attribute all resulting phenotypes specifically to the absence of srbA in A. fumigatus [37,45]. Colonies were exposed to low oxygen growth condition (1% pO2) to randomized screening (S1 Fig) and confirmation of the strain genotype was done with Southern blot and PCR analyses (S2 Fig). A total of one and two copies of the PbsrbA and pyrG gene was observed in Rec1 (reconstituted strain 1) and Rec2 (reconstituted strain 2), respectively. The detected high band on pyrG Southern blot results (around 5 kb) is an unspecific cross-reactive detection because the probe is able to recognize the non-functional pyrG used to knockout the srbA gene in A. fumigatus genome [45] (S2A Fig). We next confirmed the PbsrbA genome integration using conventional PCR, using primers that amplify the PbsrbA sequence including the AngpdA promoter (S2B Fig). In addition, the PbsrbA transcript and protein expression were assessed (S3 Fig). As expected, the PbsrbA transcript was expressed only in the reconstituted strains (Rec1 and Rec2), increasing when the fungus was submitted to hypoxia (S3A Fig). In agreement, the AfsrbA transcript was not detected in the reconstituted strains. The transcript to AfsrbA was also analyzed and the results are consistent with previously published data and reinforce the obtained data with PbsrbA. Using quantitative real time PCR, we observed that the transcript to AfsrbA was expressed only in the wild type strain, increasing when the fungus faced hypoxia (S3A Fig). In addition, at the protein level, the western blotting analysis, using a polyclonal antibody against A. fumigatus SrbA amino acids 1–275, indicates that PbsrbA is expressed in the reconstituted strain (Rec1) (S3B Fig). The A. fumigatus SrbA protein was also detected in the wild-type strain showing the SrbA precursor and N-terminal cleavage protein [45]. In order to analyse the growth of the reconstituted strains exposed to hypoxia, we measured the colony diameter of each strain every 24 h (Fig 7). As previously described, the A. fumigatus srbA null mutant strain does not growth under hypoxia [37]. However, the PbsrbA reconstituted strains were able to restore the null mutant hyphal growth under hypoxia (Fig 7). This result indicates that Paracoccidioides has a functional SrbA protein that can rapidly promote adaptation to hypoxic microenvironments. Previous studies showed that the A. fumigatus SrbA protein coordinates iron and ergosterol homeostasis to mediate triazole drug and hypoxia responses [37,95]. The A. fumigatus SREBP is a key positive regulator of iron homeostasis, particularly related to iron acquisition, which is essential for adaptation to hypoxia and low iron microenvironments [95]. Iron homeostasis has been characterized in Paracoccidioides [11,12,96–98] and the elucidation of additional molecules involved in this process can be relevant in the understanding of fungus pathogenesis. In this sense, our purpose was firstly attempted to screen PbsrbA reconstituted strain susceptibility to antifungals drugs using ranges of azoles concentrations [37] (Fig 8). The results showed that PbsrbA restores the failed growth of the mutant and suggest its participation in mechanisms of resistance to azoles. Previous studies in S. pombe, C. neoformans, and A. fumigatus confirmed that fungal SREBPs are key regulators of ergosterol biosynthesis [33,36,39,77]. In A. fumigatus, the SrbA protein is involved, even in part, in regulation of the expression of several ergosterol biosynthesis genes [37,95]. Taken together, the results suggest that PbsrbA can also be involved in these mechanisms because transcript to Pberg3 involved in ergosterol biosynthesis production, is also regulated in Paracoccidioides Pb01 submitted to hypoxia (Fig 4). Possibly, the fungus increases the expression of genes related to ergosterol biosynthesis, in order to compensate the reduction in ergosterol production in low oxygen, as discussed before in Fig 3. Regarding iron homeostasis, previous studies showed that the initial responses to hypoxia in A. fumigatus involve transcriptional induction of genes involved in iron acquisition. The null mutant strain to srbA (ΔsrbA) has reduced growth under iron starvation in liquid medium because it coordinates responses to iron and oxygen depletion [39,95]. Here the PbsrbA reconstituted strain 1 (Rec1) was significantly able to restore the defective growth phenotype of the mutant (Fig 9A). In fact, the transcript to PbsrbA is up-regulated in Paracoccidioides sp. grown upon iron deprivation, mainly after 24 h of incubation (Fig 9B). Even in part, this gene could be important in the mechanisms to compensate the effect of iron depletion in Paracoccidioides sp. yeast cells. Altogether, the results show that the roles of srbA are also conserved in Paracoccidioides especially those related to hypoxia, susceptibility to the azoles and iron deprivation responses. Even partially, the Pb01 SREBP was able to restore the mutant phenotypes similarly to wild type strain. In this way, SREBP is a relevant molecule to compensate the effects of hypoxia in A. fumigatus and in Paracoccidioides. In conclusion, the hypoxia response of Paracoccidioides spp. was largely unknown. In this study, we used a large-scale proteomic approach and a detailed functional characterization of the homologue to the most important molecule involved in hypoxia responses in other fungi, the SREBP protein. Our results show that Paracoccidioides modulates several metabolic pathways in order to compensate for hypoxia stress and importantly it has a functional SREBP homologue, the SrbA protein, which could be involved in regulation of the majority of the hypoxia responses in this pathogen. Taken into account that hypoxia is an important condition faced by pathogens during infection, this characterization becomes relevant in the context of Paracoccidioides spp. pathogenesis and warrants further investigation.

The genus Paracoccidioides is composed of species that are causative agents of paracoccidioidomycosis (PCM), a neglected human granulomatous mycosis, endemic in Latin America. To survive in the human host, fungi must tolerate and overcome in vivo micro environmental stress conditions, including low oxygen levels. Paracoccidioides spp. depicts differential responses to several stresses such as iron/zinc deprivation, oxidative and nitrosative stresses and carbon starvation. In addition, Paracoccidioides yeast cells recovered from liver of infected mice demonstrated adaptability to the host conditions. Mechanisms by which fungi sense oxygen levels have been characterized, although this is the first description in Paracoccidioides spp. Little is known about hypoxia in thermally dimorphic fungi and nothing has been studied in Paracoccidioides genus, one of the representatives of this group of pathogens. A detailed characterization of the hypoxia responses was performed using proteomic and heterologous genetics approaches. Paracoccidioides genus have a functional homologue of the key regulator of hypoxia adaptation in fungi, SrbA, a SREBP (sterol regulatory element binding protein) orthologue. The proteome during hypoxia provided a global view of metabolic changes during this stress and species of the Paracoccidioides genus have a functional SrbA. Our study provides a better understanding of the fungal adaptation to the host and it can improve the arsenal of molecules for the development of alternative treatment options to paracoccidioidomycosis, since molecules related to fungal adaptation to low oxygen levels are important to virulence and pathogenesis in human pathogenic fungi.

Abstract Introduction Methods Results and Discussion

2015

Characterization of the Paracoccidioides Hypoxia Response Reveals New Insights into Pathogenesis Mechanisms of This Important Human Pathogenic Fungus

Multistability and scale-invariant fluctuations occur in a wide variety of biological organisms from bacteria to humans as well as financial, chemical and complex physical systems. Multistability refers to noise driven switches between multiple weakly stable states. Scale-invariant fluctuations arise when there is an approximately constant ratio between the mean and standard deviation of a system' s fluctuations. Both are an important property of human perception, movement, decision making and computation and they occur together in the human alpha rhythm, imparting it with complex dynamical behavior. Here, we elucidate their fundamental dynamical mechanisms in a canonical model of nonlinear bifurcations under stochastic fluctuations. We find that the co-occurrence of multistability and scale-invariant fluctuations mandates two important dynamical properties: Multistability arises in the presence of a subcritical Hopf bifurcation, which generates co-existing attractors, whilst the introduction of multiplicative (state-dependent) noise ensures that as the system jumps between these attractors, fluctuations remain in constant proportion to their mean and their temporal statistics become long-tailed. The simple algebraic construction of this model affords a systematic analysis of the contribution of stochastic and nonlinear processes to cortical rhythms, complementing a recently proposed biophysical model. Similar dynamics also occur in a kinetic model of gene regulation, suggesting universality across a broad class of biological phenomena. Biological systems are optimized to survive in environments whose properties may vary greatly, such as changes in the biochemical environment of bacteria across several orders of magnitude, or even qualitatively, such as seasonal variations that banish food sources and prohibit foraging behavior in some mammalian species. Multistable dynamics and scale-invariant fluctuations are two complex dynamical processes whose presence in a wide variety of biological organisms suggests an adaptive role where they occur. The former enables switching amongst a wide variety of dynamical scenarios, whereas the latter ensures sensitivity to environmental fluctuations even if their background ambient intensity scales across several orders of magnitude. Their co-existence would allow a system to express two (or more) fundamentally distinct dynamical behaviors whilst maintaining scale-invariant fluctuations within and between each of these. The objective of this paper is to elucidate the basic dynamical mechanisms of these two phenomena and show how they can be studied within a unifying framework. We take the human alpha rhythm, which exhibits both multistability and scale invariant fluctuations [1], as a paradigmatic example and show how a recently proposed biophysical mechanism [2] is a specific example of the present, more general dynamical framework. We also investigate multistable dynamics in a kinetic model of gene regulation [3]. Mathematically, multistability corresponds to the presence of multiple concurrent state-space attractors, each with their own basin of attraction. System noise is required to erratically knock the system' s state vector from attractor to attractor (for review, see [4], [5]. A classic example in the human perceptual system is binocular rivalry, the abrupt alternations between two discrete percepts that occur when different images are presented to each eye [6], [7]. Multistability is also found in the human motor system, for example when paced finger tapping switches between anti-syncopation and syncopation [8], [9]. In the setting of perceptual decision-making, multistability between both possible choices is thought to arise just before the outcome of a two alternative choice task [10], [11], [12]. Multistability has been reported in a wide variety of other biological contexts - including the cellular mechanisms of working memory [13], and gene expression, where it underlies cellular differentiation [14] and epigenetic variability in genetically identical cell lines [15]. These observations motivate the search for generic mechanisms not limited to a specific model or context. In the framework of dynamical systems, multistability corresponds to the exploration of a multi-attractor landscape under the influence of system noise. Although the nature of the dynamical landscape has been extensively mapped, system noise is almost invariably introduced in biological contexts as an additive stochastic term. This contrasts with the treatment of stochastic effects in econometrics, where a complex relationship between trade volume (system activity) and volatility (system stochasticity) is a well-known property of financial systems [16]. This more complex, state-dependent relationship is also observed in a wide variety of biological organisms including bacterial chemotaxis [17] as well as complex physical systems. State-dependent fluctuations are arguably a defining feature of human cognition, being present in perception (the “Weber-Fechner law”; [18]), movement (“Fitt' s law”; [19]) and even computation (“Hick' s law”) where such operations appear to underlie nonverbal numerical processing in humans and monkeys [20], [21] and infants [22]. The almost ubiquitous observation of “state-dependent computations” [23] challenges the approach of simply adding system noise to dynamical states in computational models of the brain. Hence, state-dependent fluctuations and multistability are both present in the perceptual, cognitive and motor systems of the brain, ostensibly allowing the brain to adopt distinct functional modes, whilst ensuring uncertainty can be represented adaptively within and between these modes. For example, perceptual switching during binocular rivalry between visual stimuli differing by an order of magnitude along at least one physical dimension (such as contrast) would necessitate tight coupling of bistability and scale-invariant fluctuations in visual cortex in order to ensure equivalent perceptual representations across transitions. More generally, consider a distributed perceptual system that encodes its beliefs about hidden sensory causes in its mean state, whilst the precision of those beliefs is encoded in the dispersion of those states (see for example [24]). Accordingly, in the presence of perceptual ambiguity, the co-existence of state-dependent fluctuations and multistability would allow such a system to switch between several competing perceptual representations whilst keeping the precision of those beliefs relatively constant. Without this coupling, beliefs regarding the more intense aspects of the external environment would inevitably be held with greater precision regardless of their veracity. Spontaneous activity of the human cortex is dominated by high amplitude 10 Hz oscillations, strongest over the posterior cortex - the so-called alpha rhythm. Knowledge of the human alpha rhythm dates back to the earliest recordings of electro-cortical activity by Hans Berger in 1924, yet its mechanisms remain poorly understood. In contrast to the widely held belief that the human alpha rhythm continuously “waxes and wanes”, it rather bursts erratically between two distinct ‘modes of activity’ [1]. Temporal fluctuations of power in each of these modes are not constant, but rather scale in proportion to the mean power of the modes. Spontaneous activity of the human cortex hence exhibits clear evidence of both multistability and scale-free invariance. A biophysical mechanism for these key features of the human alpha rhythm was recently established in a model of large-scale brain activity [2]. This neural field model describes the large-scale dynamics of corticothalmic activity, constrained by key neurophysiological properties [25]. When endowed with appropriate biophysical properties, this model showed a remarkable concordance with the multistable properties of the human alpha rhythm. A crucial process underlying this convergence between theory and experiment was the state-dependent gating of stochastic inputs to the specific thalamic nucleus (the key relay centre of the brain) by oscillatory feedback from the cortex. Proximity of voltage-dependent NMDA channels to ligand-gated ion channels was proposed to underlie this key “state dependent” innovation [2], [26]. The biophysical model employed by Freyer et al. [2] has been validated across a wide range of states of arousal hence positioning this finding within a broad and unifying account of cortical activity [27]. However, whilst Freyer et al. [2] add an explanation of the alpha rhythm to this framework, they do not elucidate the deeper dynamical mechanisms at play, or whether they could be achieved by other model innovations. The objective of the present study is to address this in a simple algebraic (“normal form”) model of multistable oscillations and discuss the broader implications for other complex biological systems by demonstrating the same phenomena in a modulatory genetic network. To provide an orientation for the present purposes, the quantitative properties of resting state, eyes closed EEG data are briefly re-iterated (see Materials and Methods for data acquisition and analysis). In addition to exemplifying the properties of multistability and scale-invariant fluctuations, analyses of all subsequent modeled systems follow the same principles as those employed here. Fluctuations in power of 10 Hz oscillations burst erratically between a low amplitude mode and highly variable large amplitude excursions, which range across several orders of magnitude (Figure 1c, d). When viewed in double logarithmic coordinates, the PDFs derived from these time series exhibit clear bimodality (Figure 1a). Each of these modes can be well described with a simple exponential PDF of the form where x is the power and γ is the corresponding shape parameter. The overall time series is closely described by their sum. To formally compare the bimodal to the unimodal fit, we compare their Bayesian information criterion (BIC) values. The better model will yield lower BIC values, reflecting small residual variances after penalization for the number of free parameters. Crucially, although the peaks of these modes differ by several orders of magnitude, when viewed in these double logarithmic coordinates their relative widths are almost identical: This shows that their standard deviation (SD) scales in proportion to their mean. The “dwell time” statistics further characterize the properties of these fluctuations [28]. These are defined as the successive durations that the system resides in each of the two modes (the boundary between the modes is defined by the crossing of the exponentials) and can be understood by inspecting their cumulative distribution functions (CDFs). For switching driven by a classic stochastic process - a simple Poisson process - the dwell time distributions conform to a simple exponential form evident as a straight line with slope -a in semi-logarithmic coordinates (Figure 1b). In contrast, the empirical dwell time distributions in alpha fluctuations show evidence of long tailed statistics, which are well described by stretched exponentials of the form. The parameter b captures the stretch, which imbues the dwell times with “fat right hand tails” and reflects the tendency of the system to be increasingly trapped in one state or another for 0<b<1. That is, the longer the system stays in one mode, the lower the likelihood of a switch to the other mode, in contrast to a Poisson process which has a constant failure rate. As reviewed above, multistability occurs in dynamic systems when system noise causes the states to jump between two or more co-occurring phase-space attractors, each with their own basin of attraction [29]. While a number of dynamical scenarios with different sets of attractors can give rise to multistability [30], the multistable behavior of the alpha rhythm suggests a particular setting. Given that the human alpha rhythm jumps between a low amplitude mode and a high amplitude oscillatory waveform [31], bistability in this setting likely reflects switching between a fixed point and a limit cycle attractor [2]. In particular, whereas high amplitude alpha oscillations are strongly nonlinear, the low power fluctuations lack nonlinear structure [31], [32], arguing that the low amplitude mode more likely corresponds to fluctuations around a fixed point rather than a second limit cycle. Although the co-existence of these attractors can arise in a wide class of models describing complex systems, all are mathematically equivalent to the normal form of a Hopf bifurcation. A normal or canonical form is a set of reduced, approximating equations which are considered to preserve the essential features of the original system [33], [34]. Normal forms are usually analytically solvable and therefore allow a deeper insight into the geometric structure of the approximating equations [35]. Here we study the general normal form of a Hopf bifurcation, namely (1) (2) (1) describes the amplitude dynamics with shape parameter λ and bifurcation parameter β (see below). (2) describes the dynamics of the system' s phase. The function f (r) defines the relationship between amplitude and phase, i. e. the phase dynamics is amplitude dependent if. In the present paper we solely focus on the amplitude dynamics, and for reasons of simplicity we will only consider the case of a constant phase velocity, θc = 10 Hz with f (r) = 0. Note that (1) is an odd function so that whatever holds for r>0 also holds for r<0 (because the amplitude and its derivative are both inverted). Hence, without loss of generality we need only consider the case r≥0. Because (1) is always zero for r = 0, the system always has a fixed point at the origin. This fixed point is an attractor if the RHS of (1) is negative in the immediate neighborhood of zero. Otherwise it is a repellor. If it is negative for all r>0 then the origin is the system' s global attractor. However this fifth degree polynomial can have at least two positive roots and, due to the rotation introduced by (2), the system can hence have up to two limit cycles of finite positive amplitude R. Any such limit cycle will be an attractor if the sign of (1) is positive for r<R and negative for r>R. Otherwise it will be a repellor. In Figure 2a, we consider a variety of scenarios with the shape parameter fixed at λ = 4 and different values of the bifurcation parameter β. In the first case (red; the RHS of (1) is strictly negative (leftward moving) for r>0 and hence the origin is the system' s global attractor. However as β increases, the local maximum at r∼1. 5 crosses zero at, ensuring two limit cycles. The preceding geometric considerations imply that there is hence an unstable repellor (at) and a stable attractor (at). An example (blue) is given for β = −2. With a further increase in β, the limit cycle repellor migrates inward and eventually (at β = 0) collides with the fixed point. Following this bifurcation (i. e. for β>0) geometry reveals that the fixed point is a repellor and the outer (and now only) limit cycle is the system' s global attractor. Panel (b) shows the three possible roots of (1) plotted across a continuous range of β with locations of the three preceding scenarios indicated accordingly. This plot reveals the continuous unfolding of the three scenarios just outlined, with the abrupt discontinuities - heralding bifurcations - at β = −4 and β = 0. This is the canonical sub-critical Hopf bifurcation, so-called because limit cycle solutions occur below the loss of stability of the fixed point at β = 0. Generally, bistability is confined to the region where the attracting fixed point and attracting limit cycles have basins of attraction separated by the unstable limit cycle (hence also called a seperatrix or basin boundary). Similar geometric considerations across a range of values of the shape parameter λ reveal all possible dynamic scenarios of this system (Figure 3). The limit cycle repellor - necessary for the subcritical Hopf bifurcation and hence bistability - occurs across a broad region of parameter space (one example in blue) and can hence be considered a structurally stable solution to the system. However, for λ<0, the quintic polynomial (1) has at most two zero crossings, i. e. two attractors. For β<0 the only attractor is at the origin r = 0, and as β increases through β = 0 (far left panels, red to green), this fixed point loses stability in favor of a single limit cycle attractor that grows continuously in amplitude from the origin. The absence of a point of inflexion for these values of λ precludes the possibility of the unstable limit cycle and hence bistability. Because limit cycle solutions occur strictly above the loss of stability of the fixed point at β = 0, this scenario corresponds to a supercritical Hopf bifurcation. This system hence captures the complete family of Hopf bifurcations with a continuous transition between the super- and subcritical cases. A third order (cubic) polynomial is sufficient to model either a super-critical or a subcritical Hopf bifurcation, but not both (i. e. there is no single cubic function that allows a smooth transition between the two). The fifth order term can be thought of as a higher order “correction term” that allows one to model a physical system that could, in principle, express either a subcritical or a supercritical bifurcation depending on the smooth tuning of a control parameter (see [33]). This model hence provides a single computational platform to explore our primary hypothesis as well as any putative alternative hypotheses. Whilst these basic processes are present in most human EEG recordings, there exists considerable inter-subject variability in the degree of bimodality, the relative height of the two modes and the stretch of the dwell time distributions ([1]; see supplementary Figure S1). For example, in a sample of 16 subjects the stretched exponential exponent b varied between 0. 4 and 0. 6 for the low power mode (mean 0. 49) and from 0. 5 to 0. 9 for the higher mode (mean 0. 68). Can these high-order statistics be used for subject-specific parameter estimation when inverting models of large-scale cortical activity? We exploit the simple parametric form of (4) by systematically varying the deterministic and stochastic terms and obtaining numerical time series, from which we estimate system statistics describing the extent of the bimodality and the shapes of the dwell time distributions for the two modes. This allows a proof of principle that the model parameters could be inferred from empirical time series data. The results for four summary statistics are shown in Figure 6, with attention restricted to regions in parameter space where bimodal activity arises. Results in Figure 6a–d show the impact of varying the two noise parameters ρ and η, while the model' s shape and bifurcation parameters λ and β are varied in Figure 6e–g. Note that, consistent with Figure 3, bistability only exists in a thin strip of parameter space where. Four types of summary system statistics are shown. The top row depicts the relative goodness of fit of a bimodal compared to a unimodal fit, after penalizing for relative model complexity with Bayesian Information Criteria (BIC). This captures the depth of the bimodality - that is, the overall distinctiveness and separation of the modes (see [1]). The second row shows the relative height of the two modes and hence the ratio of the time that the system is trapped in either mode. The lower two rows show the degree of stretching of the dwell time distributions for the low and high power modes (third and bottom rows, respectively). There are several noteworthy features of these plots: Firstly, the curves are well behaved (no smoothing has been applied) and are either monotonic in cross section, or unimodal. A specific experimentally derived estimate of any of the system statistics hence constrains the value of underlying system parameters to a relatively simple curve or loop (corresponding to contour lines) in each of these spaces. Secondly, although there do exist correlations between some of the parameters, they are sufficiently distinct that combining multiple system statistics generally further constrains the underlying parameter values. For example, the contours in Figure 6e for the BIC values are generally orthogonal to those in Figure 6d for the stretching of the low power mode. Hence specific values of each of these two system statistics will in general constrain the values of the model parameters λ and β to at most two possibilities (where the contour lines of each statistic cross). Fixing the parameters and simulating the system multiple times yields an ensemble of time series from which can be extracted an ensemble of data features. Ensemble simulations show that each of the summary statistics (BIC difference, etc.) follow a Normal distribution (Supplementary Figure S1) whose mean corresponds to the value shown in Figure 6. This is crucial for formal model inversion since this allows a (pre) metric to be defined over these spaces, using a suitable function such as Kullback–Leibler divergence and hence estimates of the distance between parameter values. We also note that although the contour lines in these particular planes may not cross, we actually seek the intersection of the surfaces which yield these lines in cross-section and which more generally will have mutual intersections in the full dimensional space. This can be achieved with a suitable variational scheme such as [37]. Although formal model inversion is beyond the scope of the present submission, two ‘proof-of-principle’ examples of model inversion from EEG data are shown in Figure 7. In the top row, system statistics of EEG data (Figure 7a) with frequent high amplitude excursions and short dwell-times (i. e. less stretched CDFs) are captured by the normal form model (Figure 7b) with noise parameters ρ = 0. 46, η = 30. 4, obtained from contours of Figure 6. In the bottom row, an empirical example with a smaller high power mode and longer dwelling (i. e. more stretched CDFs) is captured by increasing both the overall noise variance (to η = 57. 6) and the ratio of multiplicative noise (to ρ = 0. 71). The notion that resting state cortex operates in the vicinity of a nonlinear instability is certainly not unique to the present contribution, although the majority of previous accounts have assumed that the bifurcation was supercritical (e. g. [31], [38], [39], [40], but see also [41], [42]). In this case, switching between a low and high amplitude mode arises from a “wandering” of the system across either side of the bifurcation threshold, corresponding to a random walk on the system' s bifurcation parameter. To investigate the dynamics arising in this scenario, we set the shape parameter lambda to -4 so that the system possesses a supercritical bifurcation, and endow the bifurcation parameter beta with mean reverting stochastic (Ornstein Uhlenbeck) fluctuations, (5) where is an uncorrelated Wiener process. Beta will hence undertake stochastic excursions either side the bifurcation point, β = 0. The typical excursion duration and depth into either regime is determined by the correlation time τ and the variance σ. In particular, if τ diverges (i. e. 1/τ approaches zero), equation 5 becomes a Brownian walk on β. In this setting, the first return time for β (the dwell times for fluctuations on either side of the 0 axis) would show a power law distribution, which is contrary to the presence of stretched exponential dwell times in our EEG data. Decreasing τ (and hence introducing correlations) leads to an exponentially truncated first return time distribution (again not a stretched exponential). Should the system even yield a bimodal PDF, here we already encounter a major difference from the subcritical case, namely that the temporal statistics in the supercritical setting are determined by the (arbitrarily-imposed) characteristics of the noise through the correlation time parameter τ and do not emerge from the system' s inherent dynamics. Two examples of this dynamic scenario are presented in Figure 8. For small to moderate parameter noise, power fluctuations express a unimodal exponential PDF, despite the emergence of occasional oscillatory activity in the time series. For moderate to large parameter noise, the system exhibits periods of high amplitude oscillatory behavior corresponding to parameter excursions well into supra-threshold territory. The probability distributions in these settings do not, however, show the clear bimodal distributions. Rather they generate broad, approximately unimodal forms that do not converge, even over long integration times towards any obvious simple probability distribution (multiple local minima are occasionally present in the PDFs). This is not particularly surprising since mean reverting stochastic (Ornstein Uhlenbeck) fluctuations are maximally (and unimodally) distributed around zero. In brief, the parameters of (5) play a crucial role in determining the data features, but impart an expected high penalty cost and generate a poor model fit. It hence seems unlikely that this scenario, with the increased complexity, can parsimoniously account for bimodal statistics in human EEG. Having established the dynamical principles of bistability in a normal form model, we now revisit bistability in a detailed biophysical model of electrocortical activity (previously reported in [2]). This model describes local mean field dynamics of populations of excitatory and inhibitory neurons in the cortical gray matter as they interact with neurons in the specific and reticular nuclei of the thalamus [25]. These dynamics are governed by physiologically derived nonlinear evolution equations that incorporate synaptic and dendritic filtering, nonlinear firing responses, corticothalamic axonal delays and synaptic gains between presynaptic impulses and postsynaptic potentials. Resting state cortical activity has extensively been modeled by studying the noise-driven endogenous fluctuations. This yields a set of eight first-order stochastic delay differential equations. For a full model description including equations, please refer to [2], [25], [38]. Whereas in [2] examples of bistability were illustrated, we presently seek to more deeply explore the underlying bifurcation space and hence the core dynamical processes. We employ a numerical continuation scheme [43] to study the family of bifurcations occurring at the 10 Hz (alpha) instability within the subspace of this system' s parameter space spanned by the excitatory connection strengths for the reciprocal feedback between cortex and thalamus, νes and νse. Although this delay differential system is substantially more complex than the preceding normal form, it nonetheless exhibits a comparable family of Hopf bifurcations with a continuous transition from sub- to supercriticality along a branch of 10 Hz instabilities (Figure 9a). We hence identify a candidate set of parameter values for a subcritical Hopf bifurcation (Figure 9b) and drive the system with a combination of additive and multiplicative stochastic inputs. In particular, following [44] we introduce a state-dependent feedback term from descending excitatory cortical input to the specific thalamic nuclei which modulates the strength of ascending stochastic input from subcortical sources (as motivated in [37]). The model exhibits bimodal distributions of neural activity (Figure 9c), with long tailed dwell time distributions (Figure 9d), bearing strong resemblance to both the normal form model and empirical EEG. However, for other parameter combinations, the corticothalamic model also captures a broad range of other statistics of EEG, including epileptic seizures [27], [38], sleep spindles [27] and evoked potential waveforms [45] that do not occur in the simple normal form model. Intriguingly, the processes we presently study in the human brain have been observed across a range of distinct different biological systems. For example, the crucial role of bistability has been recently reported in gene expression [14], [15], [46], [47] as has the existence of long-tailed distributions in gene expression [48]. State-dependent noise has also been reported to play a key role in bacterial chemotaxis [17]. Our analyses may thus speak broadly to other biological contexts. We analyse a kinetic model of genetic regulation [3]. This model captures DNA-protein interaction in a small feedback network of genetic regulation in the temperate bacteriophage λ - a protoypical example of efficient phenotypic switching in response to an environmental signal [49], [50]. Please note that the bacteriophage λ is unrelated to shape parameter λ in equation (1). Briefly, the concentration of a single genetic repressor, x, initiates a sequence of DNA binding and dimerization at three sequential binding sites, each quickly reaching equilibrium due to its fast kinetic rates. The resulting DNA-protein complex then initiates transcription and production of x - hence enacting a positive feedback loop - which is offset by its degradation. Transcription also requires the presence of an RNA polymerase. Both transcription and degradation occur at slower time scales than binding and dimerization. By assuming the fast kinetic equations are kept at (or close to) equilibrium in comparison to transcription and degradation, the evolution of x can be described by the differential equation, (6) where the dimensionless parameter α is a measure of the degree in which the transcription rate is increased above its basal rate by repressor binding, and γ is proportional to the relative strengths of the degradation and basal rates [3]. Despite the complex algebraic form of (6), it can be seen that for physiologically realistic values of α = 10 and γ = 5. 5, this functional form has only one local maximum for x>0 and closely matches the qualitative shape of the best (least squares) fitting quintic polynomial (Figure 10a). Indeed, fixing the transcription rate parameter α = 10 and treating γ as a bifurcation parameter reveals a pair ofbifurcations yielding bistability between a high and low transcription rate occuring in the range 3. 79<γ<5. 73 (Figure 10b). Stochastic influences in genetic networks are widely modelled, representing the action of thermal fluctuations, parameter variability, and the impact of more complex chemical processes also not explicitly modelled. Similarly, in large-scale neuronal systems stochastic inputs typically represent non-specific inputs from neuronal regions (such as sensory organs) not otherwise explicitly modelled. In both settings, in systems where noise interacts with a feedback loop it can be introduced via a state-dependent stochastic term. Figure 10c–e shows an example time series with a state-dependent noise term added to the RHS of (6). It hence exhibits a bursty time series for the amplitude distribution, with a bimodal PDF showing scale-free fluctuations, and long-tailed dwell time distributions, all in excellent agreement with the properties of the normal form model. Intriguingly, these bimodal, scale invariant distributions appear to replicate those observed in other genetic regulatory systems, such as Figure 7 of [41]. Detailed biophysical models - derived and parameterized using prior biological knowledge - are useful for elucidating specific biological mechanisms underlying particular phenomena, and unifying different behaviours of the same system within a single framework (e. g. [27]). In contrast, simple algebraic (so-called normal forms) models provide deeper insights into the underlying dynamical processes at play, and whilst they lack the ability to prescribe specific biophysical mechanisms, they provide a window into unifying principles that exceed the confines of any one particular domain. A classic example is the logistic equation, which was derived as an abstraction of population dynamics [51] but became a mathematical paradigm for chaotic dynamics in complex systems. We presently show that a normal form of the Hopf bifurcation - a fifth order polynomial - is able to recapitulate the multistable and scale-invariant dynamics of human cortical activity when endowed with a subcritical instability and multiplicative noise. We also show how this model captures the dynamical principles at play in a detailed neural field model of cortical dynamics as well as a kinetic model of genetic regulation in a bacteriophage, hence suggesting universality in the biological world. Regardless of their niche, biological systems share a number of competing constraints such as flexibility, robustness, cost efficiency, fidelity, and reproduction. Is it hence surprising that apparently diverse biological systems exploit similar dynamical mechanisms in addressing these? Given that universality underpins many dynamical systems observed in nature [52], and indeed normal form models are formally equivalent to broad classes of more complex systems under fairly general conditions, it is perhaps to be expected that simple models should capture fundamental dynamical processes that arise throughout biology. We have argued that the combination of a subcritical bifurcation with multiplicative noise satisfies two key constraints, namely spontaneous switching between multistable modes of activity (flexibility) and scale-invariant fluctuations between these modes over several orders of magnitude (fidelity). The presence of bistability in both the detailed neural field model [2] and the genetic regulatory system [3] was already known: here we suggest that they share the same dynamical mechanism. Although Hasty et al. [3] illustrated exemplar time series with both additive and parameter noise, we add specific predictions arising from a state-dependent stochastic term, namely that the fluctuations may be scale invariant and the dwell times in each mode will be long-tailed. This is consistent with effects documented in other genetic systems [48], possibly due to similar mechanisms. By suggesting the functional form of the deterministic and stochastic terms in large-scale cortical dynamics, the present findings have direct and pragmatic implications for the analysis of neuroimaging data. Where stochastic processes have been previously introduced to dynamic models of cortical systems, they have almost invariably taken the form of a purely additive term. Our analysis mandates a state-dependent noise term for inversion of neural field (e. g. , Jirsa and Haken [53], Robinson et al. [54]) and neural mass (e. g. , Wendling et al. [55], David and Friston [56]) models from electrocortical data. We have given a proof-of-principle here by inverting the normal form model from two examples of bistable human EEG data (Figure 7). Naturally, formal inversion should leverage other (lower order) properties of the data such as spectral peaks and scaling regimes [25], [57] However, the bistable statistics are indispensable for disentangling the additive and multiplicative noise parameters, as well as the global properties of the system' s phase space. Moving beyond the exemplar model inversion method demonstrated here, a more formal framework, such as the variational framework implemented in Dynamic Causal Modeling (DCM) [58], would enable robust parameter estimation and model comparison. We provide a preliminary exploration of model space in duplets of parameters and show that (1) The space appears sufficiently well behaved (surfaces are smooth and unimodal), and (2) For fixed parameters, system statistics are Normally distributed. These observations suggest that the parameters will be identifiable and penalties for system complexity that rest upon the divergence between prior and posterior parameter distributions will be obtainable: Both of these are required for the variational model inversion scheme implemented in DCM [37]. This method has been successful in neuroscience for inverting both EEG and functional magnetic resonance imaging (fMRI) data. In fact the model typically used in DCM for fMRI is also a truncated power series (linear or 2nd-order), similar to our algebraic model, to which fluctuations [59] have been added. Our findings suggest that a higher order term (at least cubic) may be required for many of the rich behaviours observed near the edge of dynamical instabilities, and not necessarily only in EEG data. While EEG oscillations are much faster than fMRI responses, dwell time switching in cortical systems exhibits long tails and hence occurs well into the scales observable in fMRI data (∼10 sec). It is hence possible that a DCM with a higher order state term may be a better generative model for fMRI data than the current quadratic term. This is an empirical question that can be addressed using a suitable model inversion scheme [60], [61]. As noted before, a cubic polynomial is sufficient to describe the family of either sub- or supercritical Hopf bifurcations, but not both: A fifth order term is required to enable a smooth transition between the two [33]. Inclusion of higher order terms opens the possibility of modeling even more subtle features of the dynamics, such as sub-harmonics and nonlinear amplitude-frequency effects. These may be crucial to highly nonlinear neuronal computations such as pitch perception in the auditory system [62] and vibrotactile perception [63]. It is also important to note that the present normal form, which captures the basic dynamics of the biological systems we study, cannot be expected to capture all forms of multistability. In particular, the limit cycle is extinguished at the “lower end” of bistability (Figure 2A) through a saddle-node bifurcation. That is, the attractor loses structural stability and ceases to exist. An alternative dynamical scenario would be that such an attractor only loses asymptotic (Lyapunov) stability. That is, it remains an invariant set but loses attraction in one or more subspaces, hence becoming a saddle [64]. In this setting, system noise could still allow the system to itinerantly shadow the “ghost attractor” [4], although (in the absence of an attracting basin) the associated switching behavior may follow a simple Poisson process and not show any trapping (i. e. stretched, long-tailed dwell times). This dynamical scenario could only arise in a different normal form model. Whilst we have argued that modeling state-dependent noise and higher order state terms in neurophysiological data has pragmatic implications, they also speak to fundamental computational processes in the brain. As recalled in the Introduction, state-dependent fluctuations are arguably a defining feature of perception (the “Weber-Fechner law”), movement (“Fitt' s law”), and computation (“Hick' s law”). Intriguingly, multistability co-exists with scale invariant fluctuations in these basic cognitive systems. For example, multistability is a well known property of the function (as exemplified by binocular rivalry) and physiology (see for example [65]) of the visual system. Likewise, Weber' s law has well-known neurophysiological correlates in the early visual system [66]. Whilst the fixed ratio of the standard deviation with the mean intensity of a percept or motor action permits relative uncertainty to remain constant, the presence of multistability in these same systems additionally allows switching to distinct dynamical regimes which may confer adaptive advantage. If generative models that contain higher order terms are to provide stronger model evidence than simpler models (without those terms), it will arguably be most likely during experimental manipulations where such scale-free (or multistable) computations provide a performance advantage. That is, when the optimal generative model for the data embodies the same functional form as that required to optimize task performance. Resting state EEG data were acquired from 16 healthy subjects using BrainAmp amplifiers (hardware bandpass filter, 0. 1–250 Hz; BrainAmp; Brain Products) and EEG caps (Easy-Cap; FMS). Written informed consent was obtained from each subject prior to their participation. For detailed description of EEG data acquisition, preprocessing and analysis, please refer to [1], [2]. Time series were obtained from numerical integration of (4) in the presence of stochastic fluctuations. We used Heun' s integration scheme (an extension of the Euler integration into a two-stage second-order Runge–Kutta integration scheme [67]) with. In order to ensure stable and reproducible results, we obtained 10 different time series for each set of parameter values, starting from different initial conditions. Integration length was 900 s, matching the length of the EEG data sets. The first 10 seconds were always discarded to account for initial transients. For all following analyses the amplitude time series r (t) was transformed to its Analytic signal, where is the Hilbert transform of r, since the fitting of an exponential PDF, i. e. a chi-square distribution with two degrees of freedom (DoF), to a power time series mandates an underlying signal with two DoF. Substituting r by increases the DoF to two, thereby ensuring the valid application of the exponential PDF fitting procedure. Power PDFs were obtained by partitioning the time series of into 200 equally-sized bins and counting the number of observations in each bin. Using the maximum likelihood estimate (MLE), empirical PDFs were then fitted to two types of distribution functions: A simple exponential form [68], where x is the power and γ is the shape parameter; and a biexponential form, where γ1 and γ2 are the shape parameters of the two exponentials and δ is a weighting factor. We formally compared the bimodal to a unimodal fit using the Bayesian information criterion (BIC), which includes a penalty term for model complexity: , where L is the maximum likelihood estimate, i. e. the maximum value of the log likelihood function for the estimated model, n is the number of observations and k is the number of free parameters (k = 1 for the unimodal fit and k = 3 for the bimodal fit). Given two or more candidate models, the ‘best’ model will yield relatively low values of Λ – reflecting small residual variance after penalization for the number of free parameters. In order to gain a better insight into the functional form of the PDF, we formally evaluated the fitted PDFs in log-log coordinates. Following estimation of bimodal exponential distribution functions, the respective dwell time distributions were characterized by estimating their cumulative distribution functions (CDFs) [28]. Using the same procedure as described in [1], [2] we fitted the stretched exponential form to the dwell time CDFs. In order to estimate the parameters a and b, the equation was rewritten as. The parameters a and b were estimated from both the empirical and model data by means of a least squares linear regression in log (x) -log (log (P) ) coordinates. To map out the possible bistability and dwell time statistics, we explored the parameter space of our model by systematically varying stability parameters β and λ as well as the noise parameters η and ρ from (4). For λ>0, and the system exhibits two attractors (a fixed point and a limit cycle) coexisting in parameter space, and hence expresses bistability, given an adequate noise term. We hence chose fixed values λ = 4, β = −3. 4 and systematically varied the noise parameters η and ρ between 80 linearly equally spaced values in the defined ranges, and which were determined through iterative adjustment. For each combination of values (80×80), we iterated (4) as described in section Normal form model - numerical integration. For each time series, we applied the same fitting procedures and parameter estimation methods as for the EEG data (as detailed above), yielding two key system statistics: the BIC difference, quantifying the relative goodness of fit of a bimodal versus a unimodal PDF, and the stretched exponential exponent b indicating the degree of stretching of the dwell time distributions for the low and high power modes. BIC difference and exponent b were taken as averages across the 10 calculated time series at each point. The procedure was repeated for fixed noise parameters (η = 44. 945 and ρ = 0. 61) and varying stability parameters (,) as well as for all remaining possible combinations of two fixed and to varying parameters. This computational model describes local mean field dynamics [36], [53], [69] of populations of excitatory and inhibitory neurons in the cortical gray matter as they interact with neurons in the specific and reticular nuclei of the thalamus. These dynamics are governed by physiologically derived nonlinear evolution equations that incorporate synaptic and dendritic dynamics, nonlinear firing responses, axonal delays, and synaptic gains between presynaptic impulses and postsynaptic potentials [25]. The activity within each neural population is described by three state variables - the mean soma membrane potentials Va (x, t) measured relative to resting, the mean firing rate at the cell soma Qa (x, t), and local presynaptic activity φa (x, t) where the subscript a refers to the neural population (e: excitatory cortical; i: inhibitory cortical; s. : specific thalamic nucleus; r: thalamic reticular nucleus; n: nonspecific subcortical input). The global spatial mode is described by the eight first-order delay differential equations, (m1) (m2) (m3) (m4) (m5) (m6) (m7) (m8) Note that equation (m6) contains the stochastic term (m9) comprising additive and multiplicative noise terms. The multiplicative term is modulated by the delayed corticothalamic feedback. Lengthy time series (4200 s) of the excitatory cortical presynaptic activity where used to represent the cortical sources of scalp EEG and obtained by numerical integration of the model. As for the normal form model, we used Heun' s integration scheme with.

Biological systems are able to adapt to rapidly and widely changing environments. Many biological organisms employ two distinct mechanisms that improve their survival in these circumstances: Firstly they exhibit rapid, qualitative changes in their internal dynamics; secondly they possess the ability to respond to change that is not absolute, but scales in proportion to the underlying intensity of the environment. In this paper, we study a simple class of noisy, dynamical systems that mathematically represent a very broad range of more complex models. We hence show how a combination of nonlinear instabilities and state-dependent noise in this model is able to unify these two apparently distinct biological phenomena. To illustrate its unifying potential, this simple model is applied to two very distinct biological processes – the spontaneous activity of the human cortex (i. e. when subjects are at rest), and genetic regulation in a bacteriophage. We also provide proof of principle that our model can be inverted from empirical data, allowing estimation of the parameters that express the nonlinear and stochastic influences at play in the underlying system.

Abstract Introduction Results Discussion Materials and Methods

complex systems mathematics biophysic al simulations computational neuroscience applied mathematics biology computational biology

2012

A Canonical Model of Multistability and Scale-Invariance in Biological Systems

Germline mutations in the adenomatous polyposis coli (APC) gene are responsible for familial adenomatous polyposis (FAP), an autosomal dominant hereditary predisposition to the development of multiple colorectal adenomas and of a broad spectrum of extra-intestinal tumors. Moreover, somatic APC mutations play a rate-limiting and initiating role in the majority of sporadic colorectal cancers. Notwithstanding its multifunctional nature, the main tumor suppressing activity of the APC gene resides in its ability to regulate Wnt/β-catenin signaling. Notably, genotype–phenotype correlations have been established at the APC gene between the length and stability of the truncated proteins encoded by different mutant alleles, the corresponding levels of Wnt/β-catenin signaling activity they encode for, and the incidence and distribution of intestinal and extra-intestinal tumors. Here, we report a novel mouse model, Apc1572T, obtained by targeting a truncated mutation at codon 1572 in the endogenous Apc gene. This hypomorphic mutant allele results in intermediate levels of Wnt/β-catenin signaling activation when compared with other Apc mutations associated with multifocal intestinal tumors. Notwithstanding the constitutive nature of the mutation, Apc+/1572T mice have no predisposition to intestinal cancer but develop multifocal mammary adenocarcinomas and subsequent pulmonary metastases in both genders. The histology of the Apc1572T primary mammary tumours is highly heterogeneous with luminal, myoepithelial, and squamous lineages and is reminiscent of metaplastic carcinoma of the breast in humans. The striking phenotype of Apc+/1572T mice suggests that specific dosages of Wnt/β-catenin signaling activity differentially affect tissue homeostasis and initiate tumorigenesis in an organ-specific fashion. Epithelial malignancies such as colorectal and breast cancer are thought to arise and progress towards malignancy due to alterations in signal transduction pathways that regulate the balance between self-renewal and differentiation in adult stem cell compartments [1]. The canonical Wnt/β-catenin signal transduction pathway plays a rate-limiting role in embryonic and adult stem cell renewal, and its aberrant activation is among the most common signaling defect in human cancers [2]. Activation of the canonical Wnt pathway leads to intracellular β-catenin stabilization and its translocation to the nucleus where it interacts with members of the Tcf/Lef family of transcription factors to modulate the expression of specific Wnt target genes (http: //www. stanford. edu/~rnusse/pathways/targets. html). In the gastro-intestinal tract, Wnt/β-catenin signaling regulates stemness and differentiation of epithelial cells along the crypt-villus axis [3], [4]. Accordingly, truncating mutations in the APC tumor suppressor gene, the main negative regulator of the Wnt/β-catenin pathway, result in the constitutive activation of canonical Wnt signaling thus affecting stem cell differentiation and trigger tumor formation in the GI-tract and in other extra-intestinal tissues in a dosage-dependent fashion in man and mouse [5]–[8]. The structure and distribution of β-catenin binding and downregulating motifs along the APC tumor suppressor gene is particularly suited to study the effects of specific dosages of canonical Wnt signaling on the multiplicity and tissue-specific distribution of the resulting tumors (Figure 1A). The vast majority of APC mutations found in hereditary and sporadic colorectal cancers are distributed in the 5′ half of the gene and are predicted to encode for stable truncated proteins encompassing up to 3 β-catenin downregulating (20 a. a.) domains. Stable truncation of the mouse Apc gene at codon 1638 as encoded by the Apc1638T allele, results in a protein retaining a sufficient number of functional domains to ensure wild type β-catenin regulation, namely 3 of the 7 β-catenin down-regulating domains and one Axin-binding SAMP repeat (Figure 1A) [9]. Apc+/1638T animals are tumor-free and even homozygous Apc1638T/1638T mice are viable with no apparent predisposition to tumorigenesis [9], in sharp contrast with the marked tumor predisposition and embryonic lethality characteristic of all Apc-mutant mouse models described to date in hetero- and homozygosity, respectively [6]. Notably, Apc1572T, a targeted allele designed to truncate Apc immediately upstream of the only SAMP (Ser-Ala-Met-Pro) repeat encompassed by Apc1638T (Figure 1A), is characterized by an intermediate level of Wnt/β-catenin signaling activation, higher than wild type Apc and Apc1638T though significantly lower than other Apc targeted alleles known to result in GI tract tumors [9]. Here, we show that Apc+/1572T mice are characterized by a striking predisposition to multifocal mammary adenocarcinomas with no susceptibility to intestinal adenomas. To allow the biochemical and functional characterization of the Wnt/β-catenin signaling defect encoded by the Apc1572T allele [9], we established Apc1572T/1572T embryonic stem (ES) cells from pre-implantation blastocysts and compared them by TopFLASH reporter assays [10] with Apc+/+, Apc1638T/1638T, and Apc1638N/1638N ES lines [5] (Figure 1B). The results show that Apc1572T/1572T ES cells encode for intermediate Wnt/β-catenin signaling levels, in between those characteristic of Apc1638N/1638N and Apc1638T/1638T. The latter are in fact very close to those of wild type (Apc+/+) ES cells, as previously reported [9]. These differences in Wnt/β-catenin signaling dosage are likely to result from diminished efficiency of β-catenin downregulation by the Apc1572T truncated protein due to the deletion of the only Axin-binding SAMP domain encompassed by Apc1638T. Immuno-precipitation (IP) analysis of the Apc-bound β-catenin fractions in the different Apc-mutant ES cell lines confirmed that, when compared with wild type (Apc+/+) cells, decreasing amounts of Apc-bound β-catenin are observed in Apc1638T/1638T, Apc1572T/1572T, and Apc1638N/1638N ES cells (Figure 1C). Previously, we showed that different levels of β-catenin signaling affect the ability of mouse embryonic stem (ES) cells to differentiate towards specific lineages in a dosage-dependent fashion [5]. To address the same question for the Apc1572T allele, we have subcutaneously injected undifferentiated Apc1572T/1572T ES cells into syngenic mice to induce formation of teratomas, as previously described [5]. The differentiation profiles of the Apc1572T/1572T teratomas were then investigated by histological and immuno-histochemical analysis, and compared with those obtained with wild type (Apc+/+) and other Apc-mutant ES cells. In line with their intermediate level of constitutive Wnt/β-catenin signaling activation, Apc1572T/1572T teratomas show a more heterogeneous spectrum of ecto-, meso-, and endodermal lineages than Apc1638N/1638N (characterized by a higher TopFLASH reporter activity; see Figure 1B), though still more limited in their differentiation capacity than Apc1638T/1638T (characterized by TopFLASH reporter activity comparable with wild type ES cells) (Figure 2E). In agreement with previous observations [5], several differentiation types, namely neural, bone, cartilage and ciliated epithelia were absent in Apc1638N/1638N teratomas. In particular, markers employed to identify neuroectodermal lineages did not stain Apc1638N/1638N sections, in contrast with Apc1572T/1572T teratomas where a limited but significant number of the cells were GFAP positive. Differentiation to striated muscle was also severely affected and detectable in only a minority of the Apc1638N/1638N sections [5], whereas all Apc1572T/1572T teratomas analyzed revealed positive myosin staining. Notably, among the cell types positively identified in Apc+/+ teratomas, mammary epithelia were relatively more abundant in Apc1572T sections, as shown by the combined staining with SMA and CK8 and the typical tissue architecture with luminal cells on top of a myoepithelial basal layer (Figure 2A–2D and Figure 2F). Hence, homozygous Apc1572T ES cells are characterized by an intermediate differentiation defect between Apc1638N/1638N and Apc1638T/1638T, with an unusual enrichment in mammary epithelial differentiation. Apc+/1572T mice were generated from two independent ES clones previously obtained by targeting a hygromycin cassette at codon 1572 of the endogenous mouse Apc gene [9]. To assess the post-natal viability of Apc1572T/1572T mice, heterozygous Apc+/1572T animals were interbred and four litters analyzed (n = 18 mice). None of the resulting animals was found to be homozygous for the targeted allele (p = 0. 0034, χ2 test) indicating that the Apc1572T allele results in embryonic lethality, as previously observed for the majority of Apc-mutant mouse models with the only exception of Apc1638T [6], [9]. Thus, the difference between the Apc1572T and Apc1638T truncated proteins, namely the Axin-binding SAMP motif pinpoints to a key role for this functional domain in Wnt signaling regulation during embryonic development. Phenotypic analysis of Apc+/1572T heterozygous animals was performed on a total of 69 mice and compared with wild type, Apc+/1638N and Apc+/Min on two different genetic backgrounds, namely inbred C57Bl6/J and F1 C57Bl6/J×129Ola (Table 1). GI tumor multiplicities and localization in Apc+/Min and Apc+/1638N did not differ from previously published data [11], [12]. Notably, Apc+/1572T mice do not have an increased susceptibility to intestinal tumors when compared with wild type animals. Epidermal cysts and desmoids, previously shown in the Apc1638N model [12], were also observed among Apc+/1572T mice with the same gender-specific distribution though with diminished multiplicity and penetrance (Table 1). Together with the absence of intestinal tumors, the most striking phenotypic feature of the Apc+/1572T mouse model is undoubtedly represented by the highly penetrant incidence of multifocal mammary tumors among virgin females (100%) and males (30%), in sharp contrast with Apc+/Min and Apc+/1638N animals (1/6 and 0/14, respectively) (Table 1, Figure 3A–3D). These tumors typically arise around 3 months of age in C57BL6/J animals, though age of onset fluctuates in the different genetic backgrounds (Table 1). Histological analysis of the Apc+/1572T mammary tumors revealed a lobular arrangement with both acinar and glandular growth patterns (Figure 4A). Varying degrees of squamous metaplasia were observed in all tumors analyzed. These structures resemble skin and hair follicle differentiation (Figure 4B), in some cases strikingly similar to that observed in trichoepithelioma originated from the hair follicle. This highly heterogeneous histology with diffuse lobular hyperplasia and different degrees of squamous metaplasia was also present in smaller lesions. Thus, trans-differentiation of mammary epithelial cells takes place at an early stage during Apc–driven tumorigenesis. Immunohistochemistry (IHC) analysis revealed that all Apc+/1572T mammary adenocarcinomas (n = 12) encompass luminal and myoepithelial cell types together with areas of squamous metaplasia (Figure 4D–4E, Figure 4G–4H, and Figure 4J–4K). Heterogeneous patterns of β-catenin subcellular localization were also observed upon IHC analysis of Apc+/1572T mammary tumors with the majority of parenchymal cells showing membrane-bound and cytoplasmatic staining along with smaller patches characterized by strong nuclear staining (Figure 4M–4N). As observed in the vast majority of the intestinal and extra-intestinal tumors caused by APC gene mutations in man and mouse, LOH analysis of DNA and protein samples from Apc+/1572T mammary tumors revealed allelic imbalance in more than 90% of cases (21/23) (Figure 5A). These observations were validated by western analysis of tumor-derived cell lysates (Figure 5B). During necropsy, Apc+/1572T mice were identified with gross pulmonary alterations subsequently identified as metastases of the primary mammary adenocarcinomas by histological and IHC analysis. Similar to the primary mammary carcinomas, these lesions encompassed both luminal and myoepithelial cell types (Figure 4C, 4F, and 4I). Areas of squamous differentiation were also present, though significantly less abundant than in the primary mammary tumors (Figure 4L). β-catenin IHC analysis of the Apc+/1572T lung metastases recapitulated the staining pattern of the primary tumors (Figure 4O). To provide additional experimental support for the “just right” signaling model for Apc-driven mammary tumorigenesis, we have taken advantage of a recent study according to which Tgf-β signaling antagonizes canonical Wnt signaling thus negatively regulating stem cell self-renewal [13]. Hence, Tgf-β alterations such as those resulting from Smad4 loss of function mutations, are expected to lead to a further increase of Wnt/β-catenin signaling in the Apc-mutant cellular background. Therefore, we have bred Apc+/1572T animals with Smad4+/Sad, a mouse model for juvenile polyposis previously developed in our laboratory [14]. Smad4+/Sad animals are characterized by a late-onset predisposition to hyperplastic intestinal polyps which develop in the absence of a 2nd hit at either the Smad4 or the Apc locus [15]. As both these tumor suppressor genes map to chromosome 18 in the mouse, we have generated Apc+/1572T/Smad4+/Sad compound heterozygous mice where both targeted alleles are in the in cis phase on chr. 18 as previously described for the Apc1638N model [15]. As shown in Figure 6A, Apc+/1572T/Smad4+/Sad mice show a similar incidence of mammary adenocarcinomas similarly to Apc+/1572T, but are characterized by multiple GI-tract tumors. These polyps are of the adenomatous type and become apparent at a much earlier age than the hyperplastic lesions with a pronounced stromal component characteristic of the Smad4+/Sad model (Figure 6B). Moreover, the vast majority of the Apc+/1572T/Smad4+/Sad intestinal polyps show loss of the entire chr. 18 carrying the wild type alleles of both tumor suppressor genes [15] (Figure 6C), whereas the intestinal lesions characteristic of the Smad4+/Sad mice retain the wild type Smad4 allele at first, and show Smad4 LOH (but not at the Apc locus) only at more advanced progression stages [15]. Although it cannot be excluded that loss of Smad4 function underlies intestinal tumour formation in these animals through Tgf-β/BMP downstream effectors independent of Wnt signaling, the histology and molecular features of the Apc+/1572T/Smad4+/Sad GI polyps strongly suggest that the further increase of Wnt/β-catenin signaling conferred by the Smad4 mutation in the Apc-mutant background results in intestinal tumours in the compound mice without apparently affecting the mammary cancer phenotype. Although the role of Wnt/β-catenin signalling has been established for a broad spectrum of cancers [2], it is yet unclear which factors determine tissue and organ specificity of the tumors arising upon its constitutive activation. In familial adenomatous polyposis (FAP) for example, different APC germline mutations lead to different spectra of extra-colonic manifestations depending on their localization along the gene and on the stability of the resulting truncated polypeptide [16], [17]. In general, it appears that more hypomorhic APC mutants localized at the 5′ and 3′ ends of the gene result in atypical FAP phenotypes characterized by reduced intestinal adenoma multiplicities and enhanced tumorigenesis outside the GI tract (mainly desmoids and cutaneous cysts) [18]. Here, we show that a hypomorphic mutation in the mouse Apc tumour suppressor gene results in a highly penetrant predisposition to mammary adenocarcinomas without the intestinal tumours characteristic of FAP patients carrying germline APC mutations and of most Apc-mutant mouse models reported to date [6]. This unique tumor phenotype is even more accentuated by the presence of pulmonary metastases arising from the primary mammary lesions, a feature rarely observed in genetically modified mouse models of epithelial malignancies. Notably, DU4475, a human breast cancer cell line derived from a recurrent thoracic wall tumor following mastectomy due to a poorly differentiated invasive ductal carcinoma [19], carries a nonsense mutation at codon 1577 of the APC gene, only 5 residues downstream of the targeted Apc1572T allele [20], [21]. Hence, it is plausible to think that only very specific alterations result in the “just right” level of Wnt/β-catenin signaling activation and trigger neoplastic transformation in the mammary gland, presumably by affecting self-renewal of the stem cell population as shown by the observed metaplastic changes. In this hypothetical model, the critical Wnt signalling threshold level to ensure stem cell homeostasis, i. e. the equilibrium between self-renewal and differentiation, is considerably lower in the mammary gland than in the intestinal epithelium. However, this would imply that most Apc-mutant mouse models, characterized by a pronounced predisposition to multiple intestinal tumors [6], should also be susceptible to mammary tumorigenesis. Indeed, most targeted Apc models show predisposition to mammary adenocarcinomas, though with considerably less penetrance than the GI tract tumors [22]–[24]. More importantly, transplantation of mammary glands from Apc+/Min mice into wild type recipient animals results in metaplastic adenocarcinomas [22], thus showing that the propensity to develop mammary tumors is intrinsic to the Apc-mutant cells. Further confirmation of the validity of the “just right” signaling model has been more recently provided by the conditional inactivation of both Apc alleles in the lactating mammary gland cells (by BLG-Cre) which resulted in multiple metaplastic growths which do not progress to neoplasia [25]. Even more support for the “just right” signaling hypothesis has been delivered by the study by Kuraguchi et al. where the conditional loss of a single Apc-LoxP allele is specifically driven in mammary progenitor cells (by K14-Cre) and in lactating luminal cells (by WAP-Cre) [26]. Only the K14-Cre-mediated Apc heterozygosity resulted into mammary adenocarcinomas with similar histological features to those observed in Apc+/1572T tumors, thus suggesting the early progenitor or stem cell of origin of these mixed lineage tumors. Notably, analysis of the wild type Apc allele in the K14-Cre; Apc+/CKO mammary tumors revealed the presence of specific somatic point mutations clustering in the codon 1521–1570 region, i. e. very close to residue 1572 where our own mutation was targeted [26]. Thus, the genetic mechanisms underlying Apc-mediated mammary tumor formation are strikingly similar between the tissue specific conditional knock-out K14-Cre; Apc+/CKO model and the constitutive Apc+/1572T mice: in both cases one allele is completely lost (the germline conditional KO allele in K14-Cre; Apc+/CKO and the somatic loss in Apc+/1572T) whereas the other retains residual β-catenin downregulating activity (the somatic point mutations found in K14-Cre; Apc+/CKO, and the targeted germline mutation in Apc+/1572T). The final outcome of these selection processes is the retention of the dosage of Wnt/β-catenin signaling that is “just right’ to allow clonal expansion of mammary stem cells or early progenitors and their neoplastic transformation. A second implication of the “just right” signalling model is that an increase of the Wnt/β-catenin signaling dosage conferred by the Apc1572T mutation is expected to trigger intestinal tumor formation. Recently, it has been reported that Tgf-β signaling antagonizes canonical Wnt thus negatively regulating stem cell self-renewal [13]. Hence, Tgf-β alterations such as those resulting from Smad4 loss of function mutations, are expected to lead to a further increase of Wnt/β-catenin signaling in the Apc-mutant cellular background. Accordingly, Apc+/1572T/Smad4+/Sad compound heterozygous mice revealed a similar incidence of mammary adenocarcinomas as in Apc+/1572T, but were also characterized by multiple GI-tract tumors never observed in the parental strain, even when kept for longer than 1 year. These polyps arise from loss of the entire mouse chr. 18 where both the Apc and Smad4 tumor suppressor genes map, and are clearly different from the hyperplastic lesions characteristic of Smad4+/Sad mice which retain the wild type Smad4 allele [15]. Although it cannot be excluded that loss of Smad4 function underlies intestinal tumour formation in these animals through Tgf-β/BMP downstream effectors independent of Wnt signaling, the histology and molecular features of the Apc+/1572T/Smad4+/Sad GI polyps strongly suggest that the further increase of Wnt/β-catenin signaling conferred by the Smad4 mutation in the Apc-mutant background results in the observed predisposition to intestinal tumours in the compound mice without apparently affecting the mammary cancer phenotype. In view of the well known multifunctional nature of the APC tumor suppressor protein [27], one could also envisage that functional motifs other that those binding and downregulating β-catenin and Axin, could underlie the striking tumor phenotype of the Apc1572T model. However, with the only exception of the SAMP motif, all the known functional domains located in the COOH third of the protein which are truncated by the targeted mutation at codon 1572 are also absent in the Apc1638T model, previously shown to be tumor free even when bred to homozygosity [9]. Several members of the Wnt signaling cascade including the Wnt1 ligand and β-catenin, have been shown to result in mammary hyperplasia and tumors when overexpressed in transgenic mice in a tissue-specific fashion [26], [28]–[31]. Notably, the MMTV-driven Wnt1 and β-catenin transgenic models develop mammary metaplasia highly reminiscent of the tumors observed in Apc+/1572T animals, especially as far as the heterogeneity of their histology is concerned. Among the broad spectrum of breast cancers observed in man, the histological subtype that most closely resembles the Apc1572T mammary adenocarcinomas is represented by metaplastic carcinoma of the squamous type, responsible for 1 to 5% of the total breast cancer burden [32]. Notably, genetic alterations in different members of the Wnt signaling pathway including CTNNB1 (β-catenin), APC, and WISP3 (Wnt1 Induced Secreted Protein 3) are relatively common among metaplastic carcinomas of the breast [33]. Hence, it is plausible to think that only very specific gene alterations result in the “just right” level of Wnt/β-catenin signaling activation to trigger neoplastic transformation in the human mammary gland. On the contrary to other, more common gene mutations leading to breast cancer (e. g. ErbB2) which are likely to affect more committed progenitor cells, Wnt/β-catenin signaling activation presumably affects self-renewal and differentiation capacity of the mammary primordial stem cell or in very early progenitors as shown by the metaplastic changes and the presence of both myoepithelial and luminal lineages in Wnt-driven mammary tumors. Apart from their resemblance with breast metaplastic carcinomas in man, the mammary adenocarcinomas observed in Apc+/1572T mice without the concomitant presence of multiple intestinal polyps raise the possibility that a fraction of the hereditary breast cancer cases could be caused by germline APC mutations located in the proximity of codon 1572. We are currently testing this by sequencing the APC gene around codons 1500–1700 in hereditary breast cancer patients which tested negative for BRCA1/BRCA2 germline mutations. Also, we are searching whether FAP patients with APC germline mutations in the proximity of codon 1600 show any eventual predisposition to breast cancer. However, it is also possible that mutations in other members of the Wnt pathway may result in the “just right” signaling level and in tumor predispositions similar to the Apc1572T model. In conclusion, we have shown that a targeted Apc mutation encoding for intermediate levels of Wnt/β-catenin signaling results in a highly penetrant predisposition to multifocal mammary adenocarcinomas without the intestinal tumors characteristic of most Apc-mutant mouse models and individuals carrying germline APC mutations. Our results, also supported by several mammary cancer studies in mouse and man suggests that only specific dosages of canonical Wnt signaling are “just right” to expand the mammary stem/progenitor cell and result into mixed lineage (metaplastic) tumorigenesis. Apc+/1572T mice were generated from two independent 129Ola ES clones targeted at codon 1572 as previously described [9]. Stable expression of the truncated protein encoded by the targeted allele was confirmed by western blot analysis (not shown). Chimeras were bred with 129Ola and C57Bl6/J animals to generate inbred 129Ola, and F1 (C57Bl6/J×129Ola) Apc+/1572T mice. To generate inbred C57Bl6/J Apc+/1572T mice, F1 (C57Bl6/J×129Ola) animals were backcrossed to inbred C57Bl6/J mice for at least 8 generations. Control Apc1638N and ApcMin animals were generated by crossing inbred C57Bl6/J mutants with 129Ola for comparative purposes. Heterozygous mice were employed for the phenotypic characterization, together with wild type littermates as controls. Compound Apc+/1572T/Smad4+/Sad in cis mice were generated as previously described [15]. All animals were fed ad-libitum and housed in SPF facilities. Animal experiments were performed according to institutional and national regulations. Apc1572T/1572T ES cells were derived from pre-implantation (3. 5 dpc) blastocysts as previously described [34]. Teratomas were obtained upon subcutaneous injection of 5×106 cells into isogenic mice. 5×105 ES cells were plated on dishes coated with MEFs (mouse embryonic fibroblasts) and subsequently transfected by Lipofectamine 2000 (Invitrogen) with either 500ng of the TOP-FLASH or FOP-FLASH reporter constructs [10] together with 5ng of the Renilla luciferase vector for normalization purposes. Luciferase activity was measured by Dual–Luciferase Reporter Assay System (Promega). Apc IP analysis was performed according to previously published protocol [9] using the AFPN polyclonal antibody. Detection of Apc and β-catenin in the destruction complex was carried out by using the following antibodies: Apc Ab1 (OP44, Oncogene), β-catenin (610154, BD Biosciences). Tissues were fixed in PFA (4%) and embedded in paraffin. Five μm sections were mounted on slides stained by HE for routine histology. Antibodies employed for IHC analysis include: β-catenin (1∶2000,1247-1, Epitomics), Troma1 which recognizes a Ck8 epitope (1∶400, Hybridoma Bank), Sma (1∶200, M0851, DakoCytmomation), Ck14 (1∶10000, PRB-155P, Covance), Ck6 (1∶5000, PRB-169P, Covance). The Ck6 and Ck14 antibodies are employed to detect hair follicle and skin differentiation, respectively. However, when employed at a lower dilution (1∶1000), Ck6 also detects mammary epithelial cells. All IHC images presented here were obtained with higher dilution (1∶5000) aiming at the identification of squamous differentiation lineages. The same primary antibody dilutions were employed for IF analysis, rabbit anti-rat-FITC (Sigma) and goat anti-mouse-A594 (Invitrogen) were used for signal detection. LOH analysis for the Apc locus was performed as previously described [35]. In brief, tumour sections were obtained from Apc+/1572T mammary adenonarcinomas and stained by HE. Tumour areas were localized and microdissected by LCM (Laser Capture Microdissection; Leica Microssystems), followed by DNA isolation. PCR-amplified fragments were resolved in a denaturing 6% polyacrylamide gel, dried on paper and the number of counts per allele was determined on a phosphor imager. Subsequently, the number of counts of the larger allele was divided by the counts of the smaller allele to obtain an allelic ratio. A mean allelic ratio was calculated for at least five normal controls. This value was used to generate a comparative ratio (CR) >1. 0 by dividing the tumor allelic ratio by the mean normal allelic ratio. In this way, normalization of imbalances already observed in wild type controls due to preferential amplification of a specific allele, is not necessary. A CR ≥1. 5 was interpreted as significant, i. e. indicative of loss of the wild type allele. These PCR-based observations were further confirmed by western blot analysis. Mammary tumors and control tail tissue samples were digested with Blendzyme3 (Roche Diagnostics) in DMEM medium supplemented with gentamycin. Protein lysates, separation and blotting were performed using the NuPage Gel System (gel Tris-Acetate 3–8%) according to manufacturer' s protocol (Invitrogen). Apc detection was accomplished using the antibody Apc Ab1 (OP44, Oncogene). LOH analysis for the Smad4 gene was carried out exclusively at the protein level by IHC on tissue sections (1∶100, sc-7966, Santa Cruz). In this case, antigen retrieval was performed with Tris-EDTA pH 8. 0 and the signal was detected using Envision HRP-ChemMate Kit (DAKO).

Although signal transduction pathways are often described as “on–off” systems, the more quantitative aspects of signalling are likely to represent a very important means of regulation of the downstream biological outcomes. Mutations in members of the canonical Wnt signaling pathway represent among the most common defects in human cancers. However, it is yet unclear which factors determine tissue and organ specificity of the tumours arising upon Wnt constitutive activation. Previously, we have generated a series of hypomorphic alleles at the Apc tumor suppressor gene and showed that the differentiation potential of embryonic stem cells is dependent on the dosage of Wnt/β-catenin signalling these mutants encode for. Likewise, analysis of the two mutational hits occurring at the Apc gene in human and mouse intestinal tumors showed that these are selected to retain specific residual dosages of β-catenin downregulation. Here, we provide further support for this “just-right” signalling model by targeting a germline Apc mutation encoding for very low levels of Wnt signalling activation when compared with other mutants known to trigger intestinal tumorigenesis. Notwithstanding the constitutive nature of this mutation, heterozygous mice show a remarkable and highly penetrant predisposition to multifocal and metastatic mammary cancers without the GI tract tumor phenotype characteristic of the majority of Apc mouse models.

Abstract Introduction Results Discussion Materials and Methods

oncology/breast cancer developmental biology/stem cells cell biology molecular biology genetics and genomics/cancer genetics pathology/molecular pathology

2009

A Targeted Constitutive Mutation in the Apc Tumor Suppressor Gene Underlies Mammary But Not Intestinal Tumorigenesis

Variations on the statement “the variant surface glycoprotein (VSG) coat that covers the external face of the mammalian bloodstream form of Trypanosoma brucei acts a physical barrier” appear regularly in research articles and reviews. The concept of the impenetrable VSG coat is an attractive one, as it provides a clear model for understanding how a trypanosome population persists; each successive VSG protects the plasma membrane and is immunologically distinct from previous VSGs. What is the evidence that the VSG coat is an impenetrable barrier, and how do antibodies and other extracellular proteins interact with it? In this review, the nature of the extracellular surface of the bloodstream form trypanosome is described, and past experiments that investigated binding of antibodies and lectins to trypanosomes are analysed using knowledge of VSG sequence and structure that was unavailable when the experiments were performed. Epitopes for some VSG monoclonal antibodies are mapped as far as possible from previous experimental data, onto models of VSG structures. The binding of lectins to some, but not to other, VSGs is revisited with more recent knowledge of the location and nature of N-linked oligosaccharides. The conclusions are: (i) Much of the variation observed in earlier experiments can be explained by the identity of the individual VSGs. (ii) Much of an individual VSG is accessible to antibodies, and the barrier that prevents access to the cell surface is probably at the base of the VSG N-terminal domain, approximately 5 nm from the plasma membrane. This second conclusion highlights a gap in our understanding of how the VSG coat works, as several plasma membrane proteins with large extracellular domains are very unlikely to be hidden from host antibodies by VSG. VSGs are homodimers of two 50–60 kDa subunits held on the extracellular face of the plasma membrane by a glycosylphosphatidylinositol (GPI) anchor. VSGs have a large N-terminal domain of 350–400 residues and one or two small C-terminal domains of 20–40 residues each. The domains are connected to each other by flexible linkers [1–3]. The conformation of the linkers is unknown, as is their effect on the structure of the whole VSG. VSGs vary in sequence (for example, [4]), but have a conserved tertiary structure [5]. VSG molecules are free to diffuse in the plane of the membrane, and similar diffusion coefficients were obtained using the endogenous VSG coat on trypanosomes and VSG placed in the plasma membrane of mammalian cells in culture [6]. The rate of diffusion is high, similar to the rates measured for a range of other plasma membrane proteins, and equivalent to complete randomization of the VSG coat in 40 minutes [6]. The rate of diffusion provides strong evidence that there is minimal intermolecular affinity between VSG dimers, even at the high concentration present in the VSG coat. Estimates of the packing density of the VSG on the extracellular face of the plasma membrane have been derived from (i) measurements of the VSG copy number and estimates of the surface area (5. 7 x 106 VSG dimers and 180 μm2 [7]), and (ii) direct measurements of the cell surface area and percentage of VSG on the extracellular face of the plasma membrane (145 μm2 and 89% [8]). Thus, the estimated area available to each VSG dimer on the cell surface is between approximately 28 nm2 (cell surface 145 μm2) and 35 nm2 (cell surface 180 μm2), using the estimated VSG copy number above. It is worth noting that the first of the values for cell surface area was measured on cells grown in rodents, whereas the second was derived from trypanosomes grown in culture, and the discrepancy between the two values may represent a real difference due to growth conditions. The size of a VSG dimer can be derived from the structure of the N-terminal domain [5,9], and it is assumed that the long axis is perpendicular to the plasma membrane surface (Fig 1). The area taken up by each dimer can be approximated to a circle with an area of 28 nm2 [8], but note that this size estimate does not take into account any coordinated water molecules. This value is remarkably close to the estimates of the area available per VSG dimer as discussed above, strongly supporting the model that the vast majority of the plasma membrane is physically occluded by VSG. The VSG N-terminal domain has a long axis of approximately 10 nm, measured from the structure; allowing for some increase due to the C-terminal domain and GPI anchor, the thickness of the VSG coat is probably 12–15 nm. This value is in agreement with measurements from electron microscopy [10]. One conclusion that can be drawn from these estimates is that a significantly increased level of cell surface VSG can only occur if linked to an increase in cell surface area (for example, [11]). However, the estimates are not sufficiently accurate to distinguish between a model in which VSGs are always closely contacted by surrounding VSGs, forming a coat resembling a bubble raft [12], or whether there is a restricted amount of unoccupied space due to small variations in VSG-to-VSG distance. VSG N-linked oligosaccharides have several potential functions: (i) to provide a substrate for the unfolded glycoprotein glucosyltransferase (UGGT) that catalyses the addition of glucose to a terminal mannose on an incompletely folded protein, and where export from the endoplasmic reticulum (ER) does not occur until after the glucose has been removed, reflecting a folded state for the VSG; (ii) to act as a structural element of an individual VSG; and (iii) to act as a structural element of the VSG coat. In Trypanosoma brucei, there are two oligosaccharyltransferases (OSTs) that function in the bloodstream form: OST1 and OST2. OST1 recognises an N-linked glycosylation site in a low isoelectric point (pI) context (five residues on either side of the N-X-S/T signal) and adds a paucimannose oligosaccharide that can subsequently act as a substrate for UGGT and can eventually be further modified by trimming down to three mannose residues and, sometimes further, through the addition of an N-acetyl glucosamine and galactose decorations. OST2 adds an oligomannose structure in response to an N-linked glycosylation site in a high pI context, which can be processed by trimming [13–15]. The specificities of the OSTs are overlapping, so an N-linked site in a neutral pI context could receive either oligosaccharide. There is probably not a single fixed role for N-linked oligosaccharides in VSG function. Table 1 is an analysis of 33 distinctly expressed VSGs with A-type N-terminal domains, and shows length in residues and number, location, and nature of N-linked glycosylation sites, which are all features that will contribute to the overall dimensions of the VSG. There is an inverse correlation (R = -0. 63) between the number of residues in the N-terminal domain and in the C-terminal domain, indicating that there may be an upper and lower limit on the number of residues for a functional VSG (S1A Fig). In contrast, there is only a very weak correlation (R = -0. 11) between the molecular weight of the VSG and the number of N-linked sites (S1B Fig), suggesting that N-linked oligosaccharides do not normally have a role in increasing the size of VSGs, but do have a role other than structural in many cases, probably as substrates for UGGT. This view is supported by two other observations: (i) A small number of VSGs have no N-linked glycosylation sites, and so N-linked oligosaccharides can have no role in forming an effective coat. (ii) The majority of N-linked sites are in a low pI context (S2 Fig), and so will tend to have paucimannose glycans available for UGGT rather than the larger oligomannose glycans that might be more suitable for a space-filling role. If the role of the N-linked sites in most VSGs is to allow monitoring of folding, then it would follow that VSGs that fold efficiently no longer require such a site, whereas others that require reiterative folding cycles have retained one (or possibly more) sites. As such, the presence of an N-linked site could be more indicative of folding efficiency, rather than an element in forming a barrier. However, VSGs use every trick, and in some VSGs the oligosaccharide probably functions as a structural element in the barrier. The first example in which this might occur is in the minority of VSGs with multiple N-linked glycosylation sites in the N-terminal domain (Table 1), such as modelled for VSG118 [3], where the N-linked oligosaccharrides can occupy space between VSGs. However, it should be noted that in the one VSG with a known structure containing N-linked sugars, the oligosaccharide is held close to the VSG core and acts as a structural element in the molecules (Fig 1) [5,9]. A second example may be one N-linked glycosylation site location in VSG N-terminal domains, between residues 100 and 125, located at the plasma membrane distal tip of the VSG, where there appears to be selection for very high pI addition sites that would be almost exclusive addition of oligomannose (S2 Fig). A large oligosaccharide in this location could well affect access of external proteins. The VSG coat cannot be absolutely uniform, as there are other proteins present on the extracellular face of the plasma membrane, which raises the question of how the VSG coat acts as a physical barrier to prevent access of immunoglobulins to these non-VSG proteins. The plasma membrane can be divided into three discrete areas with different non-VSG protein compositions, each separated by some form of diffusion barrier. The first is the flagellar pocket, an invagination at the base of the flagellum, where all exo- and endocytosis occurs. The second is the flagellum membrane, and the third is the cell body membrane. Various combinations of distributions have been observed for different proteins, but the mechanism of segregating a protein to one compartment but not another is not understood. Two cell surface receptors for nutrient uptake have been identified: one for transferrin [16–18] and one for haptoglobin-haemoglobin [19]. These two receptors are concentrated in the flagellar pocket, with approximately 3,000 and 300–400 copies, respectively. The density of the VSG coat in the flagellar pocket is similar to that on the rest of the plasma membrane [8], but how the VSG density is maintained in the presence of a set of receptors and many other proteins is not understood [20]. It is also not known whether the concentration of the receptors in the flagellar pocket is advantageous for nutrient uptake and/or avoidance of immunoglobulin recognition, and/or for some other unknown reason. Many plasma membrane proteins—for example, hexose transporters—have only very small extracellular domains of less than 10 kDa, and it is likely that the VSG coat prevents access of antibodies to these domains. However, there are other proteins present on the cell body and/or flagellum plasma membrane that have extracellular domains similar in size to or even larger than the VSG. Two examples illustrate this point. First, ESAG4 and related genes encode a heterogeneous family of type 1 transmembrane proteins, some of which are localized to the plasma membrane of the flagellum. The ESAG4 family of proteins has an extracellular domain of 70–80 kDa and a cytoplasmic domain encoding an adenylate cyclase [21]. The extracellular domain is significantly larger than the VSG and can be modeled with very high confidence [22] onto a tandem di-domain of bacterial small-molecule transport proteins or substrate binding proteins (as reviewed in [23,24]). Second, the invariant surface glycoprotein (ISG) gene family also encodes type 1 transmembrane proteins localized over the whole cell surface [25–28]. ISGs have a small cytoplasmic domain and an extracellular domain that is similar size in size and structurally related to VSGs and the haptoglobin haemoglobin receptor through the use of an elaborated three-helical bundle [29,30]. ISG65 can be modeled on the haptoglobin haemoglobin receptor with a high degree of confidence, and the elongated structure has a long axis of approximately 10 nm, similar to a VSG (Fig 2) [31]. If ISGs are perpendicular to the cell surface, they would reach most or all of the way through the VSG monolayer. The copy number for individual ISGs has been estimated to between 5 x 104 and 7 x 104 [25,27]; if this level of expression is extended to the entire ISG family, it is likely that there are approximately 2 x 105 ISGs in total, roughly equivalent to one ISG for every 50 VSG molecules. These large, non-VSG proteins pose a potential threat through immunoglobulin recognition; how these proteins avoid recognition by immune effectors remains unknown. From the description above, there is an obvious dichotomy between two possible situations. In the most simplistic explanation, the VSG coat is structured to function by simply preventing access of host immunoglobulins to molecules such as ISGs. Alternatively, the VSG coat could function by combining a limitation on access with an active system that negates any antibody binding to proteins such as ISGs. In the context of the second model, the VSG coat is not a static entity that simply expands as the cell grows through the addition of new membrane and VSG. There is rapid endocytosis and recycling of the plasma membrane and VSG [34] that processes the equivalent to the entire cell surface every 12 minutes [35]. In addition, any VSG antibody complex that forms and protrudes above the surface of the VSG layer is subject to hydrodynamic flow resulting from movement of the trypanosome that both increases the rate of diffusion relative to uncomplexed VSG and gives the diffusion directionality [36]. The effect is to selectively force the complex toward the posterior pole, effectively concentrating it near the flagellar pocket and increasing its chances of endocytosis. It is thought that these two processes allow trypanosomes to persist as the antibody titre rises in the host until a threshold concentration is reached. The hydrodynamic flow-induced increased rate of endocytosis of surface-bound immunoglobulin does not appear to have evolved in African trypanosomes as a specialized adaptation, as it also occurs in the distantly related fish pathogen Trypanoplasma borrelli [37]; both might represent a specialization of an older mechanism to harvest material from the environment. Antigenic variation is a requirement for establishing persistent infection, as the mammalian immune system can kill trypanosomes once the immunoglobulin (Ig) titre is high enough to overwhelm the endocytosis and degradation pathway. Killing can occur through both opsonization [38] and complement-mediated mechanisms [39]. In rodent infections, near field isolates cause chronic infections lasting weeks, whereas monomorphic laboratory strains adapted for rodent growth proliferate until the rodent host dies after a few days. The difference in growth results from a loss of autoregulation of population density, leading to uncontrolled growth [40]. IgMs are important in controlling the acute infections caused by laboratory strains [41]. However, IgMs do not influence an infection when nearer field isolates are used to infect mice; the parasitaemia profile is the same in wild type and IgM-null mice [42]. This infers that the major interaction in adaptive immune system killing of trypanosomes in natural infections is probably mediated by interactions between the VSG coat and IgG. Specific VSG immunoglobulins are the mediators of clearance through the adaptive immune system, evidenced by VSG identity being the only known change in the trypanosome surface over the course of an infection. Antibodies against invariant cell surface proteins are produced during an infection but are not sufficient to produce immunity [43,44]. Binding of complement system components has also been detected. Binding of C3b and Factor B, two components of the alternative pathway C3 convertase (C3bBb), was detected after incubation in human serum [45]. Activation of the complement pathway downstream of C3 convertase did not occur, and so the trypanosomes remained viable. It is not known whether this binding is receptor-dependent and how further activation beyond C3 convertase is prevented by the trypanosome. Binding of complement C4 binding protein (C4BP), a regulatory component of both the classical and lectin pathways, has been detected, but, again, the molecular basis for the interaction has not been determined [46]. One way to determine how far extracellular proteins can penetrate toward the plasma membrane is to determine which structural features of the VSG are accessible on living trypanosomes. The proteins used have included: (i) VSG monoclonal antibodies (MoAbs), (ii) VSG monoclonal single domain antibodies (nanobodies, NAbs), (iii) polyclonal antisera recognising ISGs, (iv) lectins (in particular, Concanavalin A [Con A]), and (v) trypsin. Below, the results using each of these approaches are discussed, and some are re-evaluated in the light of more recent understanding of VSG structure to see how they illuminate the interaction between the trypanosome cell surface and molecules of the adaptive immune response. There are several reports on the production of anti-VSG MoAbs and analysis of their binding to live trypanosomes [47–54]; some details and the results are summarized in Table 2. The fraction of the MoAbs that bind live trypanosomes in different reports ranges from two out of 20 to nine out of nine. This variation is probably a result of the different methods used in the initial screen to identify VSG MoAbs, as the majority of laboratories did not use binding to live cells. A second difficulty in interpretation is the requirement to take great care to perform live cell binding experiments at <4°C throughout to prevent localisation of the VSG antibody complex to the flagellar pocket and subsequent endocytosis [35]. This requirement may explain why one report found seven out of 30 VSG MoAbs localised to the flagellar pocket ([48] and Table 2). Pooling the experiments shown in Table 1,43 out of 92 VSG MoAbs bind live cells. There are two observations that arise from these data: First, there are epitopes that are not accessible to antibodies—an observation consistent with dense VSG packing causing restricted access. Second, externally accessible epitopes are not a small percentage of the total number of epitopes. It is commonly believed that MoAbs observed to bind fixed trypanosomes but not live cells result from a disruption of the surface coat during the fixation process and a concomitant exposure of epitopes inaccessible in live cells. There is a further consideration that must be made to explain the increased accessibility of some MoAbs to VSG in fixed cells or in vitro (ELISA/western blot) but not in vivo. Denaturation of the VSG will expose epitopes normally hidden by being in the huge dimerisation interface [5] and/or internal within the structure VSG. Many of the screening procedures used to select MoAbs would have resulted in complete or partial VSG denaturation, including coating plates for solid phase radioimmunoassay (RIA) or ELISA, solvent fixation, air-drying, and, possibly, formaldehyde fixation. Such MoAbs would give the appearance of recognising epitopes that were inaccessible in live cells, and no analysis of the MoAbs above was performed to determine whether they recognised epitopes only after denaturation. A set of studies mapped the epitopes recognised by MoAbs that bound live cells onto the molecular structure of the VSG [49,50,53–55]. The first analysed nine monoclonal antibodies that were screened for VSG121 binding using solid phase RIA [49,50]. All nine MoAbs bound VSG in air-dried blood smears (no other fixation), and two bound live trypanosomes in suspension. The two MoAbs that bound live cells did not bind VSG in Western blots, whereas the other seven did. The epitopes were mapped using competition RIAs between the MoAbs and sera raised against purified cyanogen bromide peptides. The two MoAbs that recognised live cells were difficult to map but were weakly competed by anti-p19, which contains residues 1 to 111 of the mature VSG. The remaining MoAbs either recognised epitopes in p16 (residues 112 to 332) or conformational epitopes containing components from both p19 and p16. Subsequent to this report, it emerged that VSG structures are conserved [5], that p19 corresponds to the coil running the entire perpendicular length of the VSG, and that p16 contains most of the N-terminal domain (Fig 3A). Thus, it is not possible to estimate a value for the penetration of the two live cell binding MoAbs into the VSG monolayer, and the remaining seven probably recognised epitopes exposed by denaturation on drying or SDS treatment. The second study [54] went to great lengths to isolate a VSG117 MoAb that recognised VSG on both live cells and after western blotting. Unlike the analyses above, this work was performed after the structure of a VSG had been determined and exploited the conservation of structure to map the epitope using recombinant chimaeric VSGs. The epitope was mapped to a region more than halfway down the VSG N-terminal domain (Fig 3B). This set of experiments provided very strong evidence that immunoglobulin G molecules can penetrate a minimum of 6 to 8 nm into the VSG coat, and most of the surface of the VSG N-terminal domain is accessible. In a third study [55], a panel of seven MoAbs raised against VSG WaTat1. 1 were tested for binding to a second VSG, WaTat 1. 12, known to cross-react with WaTat1. 1 polyclonal antisera. One of the seven MoAbs did not cross-react, and, since there are only 24 point differences in the sequences of the two VSGs, it can be assumed that one of these differences occurs in the epitope. The differences are located throughout the structure of the N-terminal domain (Fig 3C). From this comparison, it is not possible to identify the epitope recognised by the single selective MoAb. It is attractive to speculate that the difference lies between residues 73 and 82, which contain seven out of 24 differences; however, these residues are largely buried in the dimerization interface and would not be accessible in the VSG dimer. The fourth study [53] used live trypanosomes and neutralising MoAbs that recognised VSG 78 to select mutants that escaped but still expressed a VSG recognised by a polyclonal anti-VSG 78. Several monoclonal antibodies were used to recognise different conformational epitopes. Two independent clones that escaped neutralisation with the first monoclonal antibody had changed serine 192 to arginine. The sensitivity to other monoclonal antibodies remained, showing that the overall structure was not affected by the mutation. Another mutant selected with the second monoclonal antibody expressed a VSG78 where glutamine 172 was changed to glutamic acid. The last isolated mutant selected with the third antibody had several changes in the VSG gene. There was a gene conversion in the 5′ region of the ORF and, in addition, a mutation in the codon 220 that was probably responsible for the resistance phenotype. All mutations identified are located in the loops at the membrane distal end of the VSG that would be readily accessible to antibodies on live cells. Single domain antibodies (nanobodies, or NAbs) are derived from classes of immunogloblins that contain only two heavy chains and are unique to camelids. The variable domain is approximately 15 kDa, containing the antigen binding variable loops, and can be made as a recombinant protein. When these were produced against VSG AnTat1. 1, a range of NAbs recognising different epitopes were isolated, including one that recognised the carbohydrate moiety on three different VSGs, all having N-linked oligosaccharides in the N-terminal domain [56]. The oligosaccharides on these three VSGs are located just over halfway down the N-terminal domain, and so the NAbs penetrate some distance into the VSG layer, as observed for one of the MoAbs described above [54]. While MoAb and NAb binding to the surface of the VSG N-terminal domain has been observed in multiple studies, the C-terminal domain does appear to be protected. Two polyclonal antisera to recombinant C-terminal domains both bound strongly after fixation but showed no binding to live trypanosomes [57]. This observation provides strong evidence that the VSG coat greatly reduces or eliminates penetration of immunoglobulins to the VSG C-terminal domain and, thus, the plasma membrane. ISG65 and ISG75 are the two best-characterised invariant proteins present over the extracellular face of the plasma membrane of the entire body [25–27,43]. As detailed above, the extracellular domains of approximately 350 and approximately 440 residues, respectively, are comparable sizes to a VSG. Modeling of the structure of the domain suggests that the ISGs have a long coil of approximately 10 nm (Fig 2) [29]. Are the ISGs accessible to antibodies on live cells? The interactions with immunoglobulins were tested in two ways: first, the binding of anti-ISG immunoglobulin to fixed and live cells was compared; second, mice were immunized with recombinant protein and challenged with a trypanosome infection. There was a discrete binding of anti-ISG75, but not anti-ISG65, to live cells in one study [43]; subsequently, however, binding of anti-ISG65 has been reported with an independent antiserum [58]. The binding of anti-ISG75 was low compared to binding after fixation and was dependent on the VSG expressed, suggesting some, but not complete, limitation on accessibility. These experiments are difficult to interpret; at a simple level they could be taken to show that ISGs are accessible, but the epitopes recognised by the antisera were not characterized, and a confirmation through the use of defined MoAbs is required. However, any ISG accessibility does not necessarily lead to immunity. Prior immunization with ISGs provided no protection against infection in mice [43]. It is also worth noting that infected people and animals produce anti-ISGs, but these do not provide protection [44]. Together, these observations allow a tentative conclusion that ISGs can be accessed by immunoglobulins, but binding is limited and tolerated by the trypanosome. The mechanism of this tolerance is probably related to the recycling of the cell surface [35]. One model for the tolerance might be that the combination of low ISG copy number and rapid recycling time does not allow the bound immunoglobulin to trigger a response. If this is the case, the VSG would shield part, but not all, of the ISG protein, ruling out a simplistic model of complete inaccessibility to non-variant surface proteins. The Con A monomer is 29 kDa and, at pH 7. 5 in physiological salt concentrations, is in approximately 1: 1 dimer-to-tetramer equilibrium [59]. The dissociation constant for monomeric interaction is around 50 μM, with a dissociation rate of 4/s; consequently, any binding detected to live or fixed cells after washing must be multivalent. Con A is subject to very complex post-translational modification [60], and the properties of different batches of Con A are affected by different amounts of proteolysis of the monomeric units. The main effect of this variability is not on binding but on valency, with proteolysed subunits remaining as dimers [61]. Succinylated Con A is locked in the dimeric form and has been used to reduce variability in the reagent in some experiments [62]. Con A preferentially binds a branched mannose trisaccharide [63] and, in VSGs, will bind N-linked sites modified with oligomannose rather than paucimannose. The response to the pI context of the N-linked glycosylation site is gradual, and away from the extremes of pI values, many sites are modified with either paucimannose or oligomannose side chains. This means that predictions of whether a live trypanosome expressing a particular VSG will bind Con A (Table 1) have to be taken with a pinch of salt. The ability of Con A to bind to live trypanosomes was determined in many labs (for example, [62,64–66]). To summarize these results, Con A bound trypanosome clones expressing some VSGs but did not bind other clones expressing different VSGs. The majority fell into the second category, consistent with the predictions in Table 1. Nearly all these studies were performed before detailed sequence and structural data were available for VSGs. In light of what is known now, some of these data can be explained. The locations of the sites vary, as some VSGs have a single site in the C-terminal domain and others a single site in the N-terminal domain (Table 1). Obviously, a VSG with no N-linked sites will not bind Con A; for other VSGs, binding will depend on accessibility, which itself will be related to the location of the N-linked oligosaccharide on the tertiary structure of the VSG and on the context pI of the N-linked site. Most of the experiments to determine the nature of the binding of Con A to live trypanosomes were performed using VSGs of unknown sequence (no sequences were available at the time), but a number of the VSGs have been subsequently characterized. For example, trypanosomes expressing VSG MITat 1. 6 (VSG 048) are not bound by Con A unless fixed or treated with trypsin [65], and this VSG was later shown to have a single N-linked glycosylation site in the linker between the two C-terminal domains [2]. The importance of the identity of the VSG in determining whether the N-linked oligosaccharide is accessible to Con A was clearly demonstrated in a study that used ten T. equiperdum clones expressing different VSGs; three were agglutinated, seven were not [64]. The sequences of one Con A binding VSG (BoTat 78) and one non-binding VSG (BoTat 1) have subsequently become available, and the location of N-linked glycosylation sites in these two VSGs provides an explanation for the difference: one site in VSG BoTat 1 is in the C-terminal domain; the other is at the base (plasma membrane proximal) of the N-terminal domain. In contrast, the sites in VSG BoTat 78 are located in the N-terminal domain, where they could present the oligosaccharides pointing toward the top of the VSG coat. Trypsin is a 23 kDa protease with specificity for lysine and arginine residues. When trypsin is added to live trypanosomes, it is able to digest VSG and release fragments from the cell ([67], for example). Different VSGs are released at different rates [67]. In terms of VSG release, the most trypsin-sensitive point is the hinge between the N- and C-terminal domains. One way to explain the variations in sensitivity to trypsin is that the enzyme is near the size limit able to gain access to the hinge part of the VSG, and some VSGs block access, whereas other do not. However, this model ignores the availability of substrate, and some VSGs may simply be better substrates than others. It would be interesting to compare the trypsin sensitivity of different VSGs comparing purified protein and live cells. Other species of African trypanosomes are much less well understood both in terms of the repertoire of functional VSGs and in non-VSG surface proteins. The genomes of T. congolense and T. vivax have been analysed for putative surface proteins [28], but there has been limited biochemical analysis to support this. In T. congolense, VSGs do not appear to have a structured C-terminal domain (s) but do have C-terminal extension of approximately 30 residues beyond the end of the structured N-terminal domain. The effect or role of this extension on the structure of the VSG coat is not known, and it may play an equivalent role to the C-terminal domain in T. brucei VSGs. T. vivax VSGs do not have any significant polypeptide extension on the C-terminus after the end of the structured VSG N-terminal domain, and there is little knowledge of experimentally identified non-VSG surface proteins. The molecular detail of how the VSG coat negates the adaptive immune response is interesting in itself, but is also relevant to identifying therapeutic strategies. One important question to be answered is how far extracellular macromolecules can penetrate into the VSG coat. The answer to this question will provide information on the effectiveness of the VSG coat as a physical barrier and whether the cell has evolved systems to overcome immunoglobulin binding to lower copy number invariant antigens. It is hard to draw many firm conclusions from the existing data, primarily due the absence of defined structures, ligands, and ligand binding sites. As examples: (i) The structure of ISG65 is only a model, and the epitopes recognised by the polyclonal ISG antisera have not been characterized. (ii) The structures of the N-linked oligosaccharides for some VSGs have been solved, but their location within the VSG coat is unknown (although it has been modeled [3]), and the relative affinity of Con A for the different N-linked oligosaccharides has not been determined. Another problem is that most of the data were collected before sequencing of VSGs became routine. Even now, only a subset of the experiments can be looked at with the sequence in one hand and structure-based hindsight in the other. The experiments with unambiguous data on the penetration of a macromolecule into the VSG coat provide very strong evidence that an intact immunoglobulin G could reach the lower part of the VSG N-terminal domain [54]. It is probable that the base of the VSG N-terminal domain, the region with the largest cross-sectional area perpendicular to the cell surface, represents the real physical barrier guarding the plasma membrane (Fig 4). It is possible that the C-terminal domain reinforces this barrier, as shown in Fig 4, but this location remains a model.

African trypanosomes have evolved two key strategies to prevent killing by the host immune response and, thus, maintain a long-term infection in a mammal. Both are based on a densely packed coat of a single protein, the variant surface glycoprotein (VSG), which covers the entire extracellular surface of the cell. The first strategy is antigenic variation, through which individual cells switch the identity of the expressed VSG at a low frequency and are selected by the host immune response. If the VSG is novel, the trypanosome proliferates, maintaining the infection; if it doesn' t switch, or if the new VSG is not novel, it will be killed. In the second strategy, the VSG acts as a protective barrier, shielding the cell from innate and adaptive immune factors until there is an overwhelming titre of antibodies recognising the expressed VSG. In this review, the VSG coat is modelled, and past experiments that investigated how it protected the trypanosome are revisited using current knowledge of VSG sequence and structure. The conclusions are: (i) the identity of the individual VSGs explains early experimental variation; (ii) most of the VSG molecule is accessible to antibodies. This second conclusion highlights a gap in our understanding of how the VSG coat works, as several plasma membrane proteins with large extracellular domains are very unlikely to be hidden from host antibodies by VSG.

Abstract The VSG Coat The Roles of VSG N-linked Oligosaccharides Non-VSG Proteins Present in the VSG Coat Interactions with the Adaptive and Innate Immune System Investigation of How the VSG Coat Functions VSG Monoclonal Antibodies VSG Single Domain Monoclonal Antibodies Antibodies Recognising Invariant Surface Glycoproteins Concanavalin A Trypsin Other African Trypanosomes Conclusions

2015

How Does the VSG Coat of Bloodstream Form African Trypanosomes Interact with External Proteins?

The Kato-Katz is the most common diagnostic method for Schistosoma mansoni infection. However, the day-to-day variability in host egg-excretion and its low detection sensitivity are major limits for its use in low transmission zones and after widespread chemotherapy. We evaluated the accuracy of circulating cathodic antigen (CCA) urine-assay as a diagnostic tool of S. mansoni. In comparison, a low sensitive CCA test (CCA-L) was assessed. The study was conducted in three settings: two foci with single S. mansoni infections (settings A and B), and one mixed S. mansoni – S. haematobium focus (setting C). Stool and urine samples were collected from school-children on three consecutive days. Triplicate Kato-Katz readings were performed per stool sample. Each urine sample was tested with one CCA and only the first urine sample was subjected to CCA-L. Urine samples were also examined for S. haematobium eggs using the filtration method and for microhaematuria using urine reagent strips. Overall, 625 children provided three stool and three urine samples. Considering nine Kato-Katz thick smears as ‘reference’ diagnostic test, the prevalence of S. mansoni was 36. 2%, 71. 8% and 64. 0% in settings A, B and C, respectively. The prevalence of S. haematobium in setting C was 12. 0%. The sensitivities of single Kato-Katz, CCA and CCA-L from the first stool or urine samples were 58%, 82% and 46% in setting A, 56. 8%, 82. 4% and 68. 8% in setting B, and 49. 0%, 87. 7% and 55. 5% in setting C. The respective specificities were 100%, 64. 7% and 100%; 100%, 62. 3% and 91. 3%; and 100%, 42. 5% and 92. 0%. Mixed infection with S. haematobium did not influence the CCA test results for S. mansoni diagnosis. Urine CCA revealed higher sensitivity than CCA-L and triplicate Kato-Katz, and produced similar prevalence as nine Kato-Katz. It seems an attractive method for S. mansoni diagnosis. Schistosomiasis remains of significant public health importance worldwide, with an estimated 207 million people infected. Over 90% of all schistosomiasis cases are found in sub-Saharan Africa [1], [2]. The disease is caused by six species of blood flukes: Schistosoma haematobium, S. mansoni, S. japonicum, S. intercalatum, S. guineensis and S. mekongi. S. haematobium causes urogenital schistosomiasis while the other species cause intestinal disease. The disease affects the poorest of the poor and compromises their development [3]. The development of preventive chemotherapy strategy, the greater access to praziquantel and the increased resources for control have amplified treatment possibilities to the majority of people who need it [4]. Significant progress has been made in the control of schistosomiasis within the past ten years, though this falls short of the target set by World Health Assembly Resolution 54. 19 adopted in 2001, which aimed to reach at least 75% of all school-aged children at risk of morbidity by 2010 [5]. In sub-Saharan Africa, S. haematobium and S. mansoni are the major causes of disease. S. haematobium infections can be diagnosed by several approaches, including detection of schistosome eggs in urines and rapid tests such as urine reagent strips for detection of microhaematuria [6]. On the contrary, there is currently no validated rapid test for S. mansoni. Because of its simplicity and relatively low-cost, the Kato-Katz technique [7] is widely used for epidemiological field surveys and is recommended by WHO for mapping to determine programme intervention zones, and for monitoring, evaluation and surveillance of intestinal schistosomiasis control programmes [8]. Though the specificity is very high, the sensitivity of Kato-Katz in single stool sample examination is limited by day-to-day variation in egg excretions leading to measurement error in estimating the presence of infection. This is particularly accentuated in areas with high proportions of light intensity infections [9]–[11]. The increase of large-scale interventions and repeated mass treatment with praziquantel will significantly reduce the prevalence and intensities of schistosomiasis. As consequence of the increase of low-intensity schistosome infections, more light infections will be often missed if single stool samples are examined by the Kato-Katz method, resulting in high underestimation of infection. Therefore, there is a need to develop and validate more sensitive diagnostic tools for S. mansoni infections. Within the Schistosomiasis Consortium for Operational Research and Evaluation (SCORE) framework, a multi-country evaluation of the commercially available point-of-contact (POC) circulating cathodic antigen (CCA) was conducted. The study aimed to evaluate the utility of the POC urine-based CCA assay as a survey tool to determine the prevalence of S. mansoni. The sensitivity and specificity of the POC-CCA for diagnosis of S. mansoni were assessed, using the cumulative results of the 9 Kato-Katz thick smear readings as ‘reference’ diagnostic test. In Cameroon, our study was conducted in three distinct epidemiological settings at different endemic levels; two areas of low and moderate endemicity of S. mansoni, and one mixed infection focus of S. mansoni and S. haematobium. As previous data suggested that POC-CCA testing may result in more children believed to be positive than by Kato-Katz testing, with a portion of these being false positives, the manufacturer developed a less sensitive version of POC-CCA. Therefore, in order to further assess the performance of CCA test, this experimental low sensitive CCA dipstick (designated CCA-L) was also tested in comparison to the commercially available CCA assay (designated CCA). The study was approved by the National Ethics Committee of Cameroon (Nr 084/CNE/DNM/09), and was a public health exercise through the Ministry of Health and the Ministry of Education. Stool and urine samples were collected from children in schools with the approval of the administrative authorities, school inspectors, directors and teachers. The objectives of the study were explained to the schoolchildren and to their parents or guardians from whom written informed consent was obtained. Children willing to participate were registered. Each child was assigned an identification number and results were entered in a database and treated confidentially. No identification of any children can be revealed upon publication. All children who participated in the study were treated with praziquantel. Other children were treated during the MDA campaign implemented by the national programme for the control of schistosomiasis and intestinal helminthiasis. Based on previous parasitological data, three epidemiological settings were selected for the study: (i) one setting of low endemicity of single S. mansoni transmission, i. e. Yaoundé (setting A) in the Centre region of Cameroon; (ii) one setting of moderate endemicity of single S. mansoni transmission, i. e. Makenene (setting B), Centre region; and (iii) one setting where mixed infections of S. mansoni and S. haematobium occur, i. e. Njombe (setting C) in the Littoral region. Yaoundé is the capital city of Cameroon. Makenene and Njombe are located at approximately 200 and 325 km from Yaoundé, respectively. Investigations were conducted in public primary schools of Obobogo (3. 82489 N, 11. 50071 E) in Yaoundé, Baloua (4. 88287 N, 10. 78953 E) in Makenene, and Kompita (4. 57898 N, 9. 64848 E) in Njombe. According to the literature, a single Kato-Katz thick smear for diagnosis of S. mansoni in low endemicity settings has a sensitivity of only 20–30% [10], [12]. However, since our study was to be carried out in both low and moderate endemicity settings, we assumed that a single Kato-Katz thick smear has a maximum sensitivity of 60%. The sensitivity of the CCA test is reported to be 80% or higher [13], [14]. Using these sensitivity estimates, a significance level of 5%, and a power of 80%, our sample size of complying children was calculated at 90. Assuming a compliance of 70% for the submission of each of three requested stool samples, the number of children to be included in each study setting was at least 199. To achieve this sample size, we selected a sample size of 250 children per study site. The study was conducted between December 2010 and January 2011. In each of the three settings, about 250 schoolchildren from the upper classes (age 8–12 years) were enrolled, approximately half boys and half girls. Urine and stool samples were collected from these children over three consecutive days. The samples were collected between 11. 00 and 14. 00, in 60 mL plastic screw-cap vials, transported to the laboratory and processed the same day. In the laboratory, three Kato-Katz thick smear slides per stool sample, using 41. 7 mg templates, were prepared and examined for S. mansoni and STH. All urine samples were tested using the CCA assays for diagnosis of S. mansoni. In addition, the first urine sample of each child was tested with CCA-L. Both CCA and CCA-L assays were obtained from Rapid Medical Diagnostics (Pretoria, South Africa) and performed at ambient temperature according to the manufacturer' s instructions. Briefly, one drop of urine was added to the circular well of the test cassette and allowed to be absorbed entirely into the specimen pad within the well. Then, one drop of buffer (provided with the kit) was added. The test result was read 20 minutes after adding the buffer. Results were determined in a blinded fashion by at least two individuals. In case the control band did not develop, the test was considered as invalid. Valid tests were scored as negative, trace (weak band) or positive (strong band). Due to the lack of standards designed for this test, the trace and positive results were classified as positive. In addition to the CCA tests, each urine sample was subjected to a filtration method for detection of S. haematobium eggs, and with reagent strips (Teco Diagnostics, USA) for microhaematuria assessment. Each urine sample was agitated to ensure adequate dispersal of eggs, 10 mL of urine were filtered through Nucleopore® filter, and filters were examined by microscopy for the presence of schistosome eggs. Schistosome infections were recorded; number of eggs was counted and intensity of infection was calculated and expressed as eggs per 10 mL of urine (eggs/10 mL) for S. haematobium or eggs per gram of feces (epg) for S. mansoni. For urinalysis, reagent strips were immersed in urine and removed immediately. The strip-results were read between 1–2 minutes by direct comparison of the colour blocks printed on the outside of the bottle label. The different data were analyzed by the epidemiological unit of the Centre for Schistosomiasis & Parasitology using appropriate statistical tests and methods. Data were entered in a Microsoft Excel spreadsheet, checked and validated. Statistical analyses were carried out using R software version 2. 10. 0. Only children with complete data records (i. e. 3 POC-CCA assays, 9 Kato-Katz thick smear readings, 3 reagent strip test results and 3 urine filtrations) were included in the final analysis. Sensitivity and specificity of CCA assays were estimated using 9 Kato-Katz thick smear readings as the reference test. The sensitivity was determined as the percentage of subjects with a positive CCA test among those positive for Kato-Katz; and the specificity was defined as the proportion of true negatives, i. e. the percentage of subjects negative for CCA among people negative for Kato-Katz. Positive predictive value (PPV) and negative predictive value (NPV) were also calculated for the different tests. Analyses were performed for the 3 study settings separately to determine differences in the sensitivity of the CCA assays potentially resulting from different S. mansoni endemicity levels (infection prevalence and intensity) and as a function of S. haematobium co-infection. To obtain a standardized measure of infection, the geometric mean infection intensity of S. mansoni, expressed as the number of eggs per gram of stool (EPG), was estimated for the three study cohorts. For each individual, the classification into light (1–99 EPG), moderate (100–399 EPG) and heavy (≥400 EPG) infection intensity was calculated based on the arithmetic mean of EPGs derived from the 9 Kato-Katz thick smear readings. The thresholds are set by the World Health Organization [3]. The strength of agreement between the CCA and the 9 Kato–Katz thick smears for each endemic setting was assessed by kappa statistics (κ), as follows: k<0 indicating no agreement, k = 0–0. 2 indicating poor agreement, k = 0. 2–0. 4 indicating fair agreement, k = 0. 4–0. 6 indicating moderate agreement, k = 0. 6–0. 8 indicating substantial agreement, and k = 0. 8–1 indicating almost perfect agreement [15], [16]. Day-to-day variation of Kato-Katz results was assessed using McNemar test [17]. An ordinal logistic regression approach was performed to assess the correlation between CCA and CCA-L categories and S. mansoni egg counts. The geometric mean egg counts of nine Kato-Katz thick smears per stool sample per day served as continuous explanatory variable, whereas the color reaction of the CCA test was considered as categorical outcome. This statistical procedure was also used to assess the association between CCA and CCA-L test results, expressed as binary outcome variable (negative/positive), with S. haematobium egg count as continuous explanatory variable and micro-haematuria as categorical explanatory variable. Non-overlapping 95% confidence intervals (CI) or p-values<0. 05 were considered as statistical significance. To further measure the discriminating power of the diagnostic CCA tests, receiver operating characteristic (ROC) curves were used to assess the association between sensitivity and specificity of the assays [18]. This allowed representing the variation of proportion of true positive individuals for Kato-Katz and CCA test in function of proportion of individuals negative for Kato-Katz but positive for CCA. The area under the curve (AUC) indicated the probability to identify accurately a true positive case when the result was simultaneously positive and negative for the CCA test, using single, triplicate and 9-Kato-Katz as the reference tests. AUC>0. 7 indicates high discriminating power. A total of 765 pupils were registered and included in the study: 258 in setting A, 251 in setting B and 256 in setting C. Of these children registered, 625 (81. 7%) aged 7–15 years old (mean 10. 7 years) provided all three requested urine and stool samples: 138 (53. 5%) in setting A, 245 (97. 6%) in setting B and 242 (94. 5%) in setting C. Table 1 summarizes the results of schistosomiasis prevalence obtained in the different study settings, as assessed by the different diagnostics approaches. The 95% confidence intervals (95% CI) for prevalence are shown. For the Kato-Katz method, there was a significant increase of infection prevalence with the increase of the number of Kato-Katz thick smear readings. Indeed, the prevalence figures for S. mansoni were 21. 0%, 23. 9% and 36. 2% in setting A; 41. 0%, 49. 0% and 71. 8% in setting B; and 31. 4%, 44. 6% and 64. 0% in setting C; with one, three and nine Kato-Katz thick smears, respectively. For the overall three study settings A, B and C, the prevalence of S. mansoni increased from 32. 8% with one Kato-Katz to 41. 8% with three Kato-Katz, and 61. 0% with nine Kato-Katz thick smear readings. Similarly, there was an increase of infection prevalence with the increase of the number of CCA tests. The prevalence of S. mansoni infections as determined by urine CCA were 52. 2% and 62. 3% in setting A, 69. 8% and 83. 7% in setting B, and 71. 5% and 86. 4% in setting C for one CCA and three CCA tests, respectively. For the overall three settings A, B and C the S. mansoni infection prevalence increased from 66. 6% with one CCA to 81. 6% with three CCA tests. For CCA-L, the S. mansoni infection prevalence after a single test was 16. 7% in setting A, 51. 8% in setting B, 38. 4% in setting C, and 38. 9% in the overall three settings. Table 2 shows the agreement between the different diagnostics approaches and the reference test, i. e. nine Kato-Katz thick smears, for the diagnosis of S. mansoni, stratified by study settings. There were a substantial agreement between the nine Kato-Katz thick smears and one Kato-Katz in setting A (k = 0. 6), and three Kato-Katz in settings A (k = 0. 7) and C (k = 0. 6). The agreement was moderate with one Kato-Katz in setting B (k = 0. 4) and three Kato-Katz in setting B (k = 0. 5); and with one CCA and CCA-L in all three settings (k = 0. 4–0. 5). Finally, the agreement was fair with three CCA in all three settings (k = 0. 3–0. 4) and with one Kato-Katz in setting C (k = 0. 4). The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the different diagnostic tests were determined and the results are summarised in Table 3. Using the nine Kato-Katz thick smears as ‘reference’ diagnostic test, the sensitivity varied from 46% to 96. 1%. In all three settings, there was an increase of sensitivity with the increase from single to triplicate Kato-Katz or CCA tests. The lowest sensitivity was obtained with CCA-L in setting A (46%). However, when considering the overall three settings, the lowest sensitivity was obtained with single Kato-Katz. The triplicate CCA produced the highest sensitivity in setting C (96. 1%) and for the overall three settings (93. 2%), followed by the single CCA (84. 5%) and the triplicate Kato-Katz (68. 5%). The specificity of CCA test was higher with CCA-L, reaching 100% in setting A. Apart from the setting C, the specificity of CCA assays increased from single to triplicate tests. For the overall three settings, the specificity increased from 40. 6% with triplicate CCA to 61. 5% with single CCA and 94. 7% with CCA-L. There was a significant variation of PPV and NPV of CCA and CCA-L between study sites, with a higher PPV value of 100% for CCA-L in setting B and a lower value of 53. 5% for triplicate CCA in setting A. The NPV varied from 53. 4% for CCA-L in setting B to 92. 3% for triplicate CCA in setting A (Table 3). In addition to the reference nine Kato-Katz, the sensitivity of the different assays was assessed towards the combined results of all tests (nine Kato-Katz, triplicate CCA and single CCA-L). Similar trends of increased sensitivities with the increase of number of tests were obtained (Table 3). To determine if the CCA and CCA-L results were affected by the intensities of S. mansoni infections, the data were analyzed considering the different intensity classes estimated from the 9 Kato-Katz thick smear readings. The results are summarized in Table 4, and the impact of infection intensities on CCA and CCA-L results is illustrated in Figure 1. For all three assays (triplicate CCA, single CCA and single CCA-L) there was an increase of prevalence and sensitivity with the increase of S. mansoni infection intensities in all three settings, apart from single CCA in setting A. In moderate and heavily infected children, the prevalence of single and triplicate CCA was above 90% in all three settings. High proportions of individuals negative for nine Kato-Katz thick smears were positive for all CCA tests. Indeed, for the overall three settings, the prevalence of positive CCA among Kato-Katz negative individuals varied from 5. 3% for CCA-L to 59. 4% for triplicate CCA. In setting C, the prevalence of S. haematobium increased from 2. 5% with one urine filtration to 12. 0% with three filtrations. Similarly, the prevalence of micro-haematuria increased from 9. 9% after one test to 14. 9% after three tests on three consecutive days (Table 1). Interestingly, micro-haematuria was also found in settings A (3. 6%) and B (8. 2%), non endemic for S. haematobium and where all urine filtrations were negative. These 25 children with micro-haematuria were aged between 7–14 years. They were mainly girls, with only 4 boys. A logistic regression analysis and adjustments were used to assess the correlation between CCA test categories and S. haematobium egg counts or microhaematuria. The results showed the absence of correlation between the CCA and CCA-L positivity and the concurrent infection with S. haematobium. OR = 0. 9 (p = 0. 10) for triple CCA, 0. 9 (p = 0. 16) for single CCA and 0. 87 (p = 0. 22) for CCA-L. Similarly, there was no significant association between the CCA or CCA-L positivity and the presence of micro-haematuria (p>0. 05). In order to assess the potential effects of infections with Ascaris lumbricoides, Trichuris trichiura or hookworms, the CCA and CCA-L responses was compared between individuals non infected and infected with any of these three species of soil-transmitted helminths (STH). The results show no significant difference of percentages of positive CCA and CCA-L based on STH infections status of individuals; 54. 2% vs 62. 1% for triple CCA, 36. 1% vs 39. 7% for single CCA and 6% vs 5% for CCA-L, for STH positive and negative individuals, respectively. This suggests that STH infections do not influence the urine CCA assay results. Detailed analysis of the data showed a day-to-day variation of the results of CCA in 21 children: 5 in setting A, 8 in setting B and 8 in setting C. It appears that in these few 21 individuals (16 positive for Kato-Katz and 5 negative) the CCA test varied from positive to negative from one day another, and vice versa, even in children positive for Kato-Katz, i. e. infected with S. mansoni. To assess the discriminating power of CCA tests, data of all three settings were merged together. The ROC curves and AUC of the different CCA and CCA-L assays are presented in Figure 2. It appears that the discriminating powers of single CCA and CCA-L were high in all cases, as the values of AUC were >0. 7. However, triplicate CCA was less discriminating, with values of AUC (0. 6–0. 7) <0. 7. The urine-CCA cassette test is a rapid antigen detecting test of active schistosome infection, more specifically S. mansoni. As the gastrointestinal tract of schistosome is a cul-de-sac, the parasite has to regurgitate at regular intervals the undigested particulate material as well as parasitic gut associated glycoproteins. CCA is one of the major antigens regurgitated by adult schistosomes and the major portion is later secreted in the host urine. The results of our study showed that urine CCA tests gave higher prevalences of S. mansoni than single and multiple Kato-Katz thick smears in all three investigated settings. This higher sensitivity of CCA is consistent with previous studies [19], [20]. However, detailed analysis of data revealed the complexity of results. In the three epidemiological settings, the S. mansoni prevalence increased by about twofold when moving from single to nine Kato-Katz thick smears. This result confirms the low sensitivity of single Kato-Katz and its improvement with the increase of the number of thick smears [21]–[23]. Taking into account the fact that repeating Kato-Katz allows a better estimation of the ‘true’ prevalence of schistosome infection, we adopted a 3×3 Kato-Katz design, i. e. 3 stool samples collected per child on three consecutive days and 3 Kato-Katz readings per sample; and the cumulative results of the 9 Kato-Katz thick smear readings was used as our ‘reference’ diagnostic test. The sensitivity and specificity of the POC-CCA and CCA-L assays were determined by comparison to this reference test. The sensitivity of the normal CCA was high (>82%) in all three settings, whereas the specificity was low, ranging from 42. 5% to 64. 7%. On the contrary, the sensitivity of CCA-L was low (overall 60%) and the specificity was quite high, above 91% in settings B and C, and up to 100% in setting A. This showed the poor accuracy of CCA-L in detecting S. mansoni infections. Therefore, this assay, in its current formulation, cannot be recommended for S. mansoni diagnosis. Similar results were obtained by Coulibaly et al. [23]. In all three settings, and for the overall settings, a single CCA showed higher or similar high prevalence than nine Kato-Katz thick smears. This suggests that the current commercially available CCA urine cassette assay maybe an appropriate test for the diagnosis of S. mansoni in moderate transmission zones, as reported by previous studies in different settings [20], [24]. Detailed analysis of our data showed a strong correlation between the S. mansoni infection intensities and the positivity rate of CCA, as well as the intensity of CCA test bands. In all settings, CCA prevalences were above 95% and up to 100% in children with high burden of infections (infection intensity ≥400 epg), whereas they were only about 65% in light infected children (Figure 1). This high variation of the sensitivity of CCA tests with the intensity of infections is in line with previous studies [19], [20], [25], [26]. Interestingly, our studies showed no influence of STH infections on the CCA results. Furthermore, there was no cross reactivity of S. haematobium infections nor microhaematuria (determined by urinalysis reagent strips) on the CCA test results. Similar results were obtained in recent studies in Kenya [20] and Côte d' Ivoire [23]. Although far less than for Kato-Katz, day-to-day fluctuations of CCA results were observed in few children all three settings. This daily variation of CCA results from positive to negative and vice versa were observed in 3. 4% (21/625) of children. From these 21 children with day-to-day fluctuation in urine CCA, 4 had positive Kato-Katz for all stools collected on the three consecutive days, 6 were negative for all nine Kato-Katz, and 11 exhibited day-to-day variation in Kato-Katz results. Apart from two children with moderate infection intensities, the daily fluctuation of CCA was observed mainly in children with very light infection, with an overall mean S. mansoni egg counts of 1. 7 epg. This variation of CCA test results might be linked to the day-to-day fluctuation of schistosome CCA levels in urine of humans infected with S. mansoni, as previously demonstrated in several countries [27]–[29]. Nevertheless, the fact that CCA is negative in those infected children with egg excretion detected by Kato-Katz raised concerns about the accuracy of CCA for diagnosing S. mansoni infections. According to the manufacturer' s leaflet, false CCA negative during the parasitic developing phase, usually in the first 4–8 weeks after infection. However, once adult worms start laying eggs, as detected by Kato-Katz, it becomes difficult to explain false CCA negative results without recourse to more detailed physiological examinations of the worms themselves. Further studies in the laboratory, perhaps, maybe required to address this issue. The comparison of single CCA and triplicate CCA results showed that the sensitivity of CCA increases with repetition of tests, whereas the specificity decreases. This is confirmed by the value the value of AUC (<70%). This indicates that repeating CCA tests significantly increases the number of false positive individuals to the detriment of true positive. In conclusion, considering the fact that urine CCA cassette produced similar prevalence as nine Kato-Katz thick smears, this assay seems an attractive tool for the diagnosis of S. mansoni infections. Moreover, its rapidity, easy to use, less time consuming than Kato-Katz, and the relative easiness to collect urine than stool samples are significant advantages for this test. Its field applicability for large scale screening for S. mansoni infections would be very useful in control programmes, especially as we are moving toward schistosomiasis elimination where feasible. However, further improvement of the assay and efforts to reduce the cost of CCA cassettes are required to enable its large scale. On the contrary, the results of the CCA-L dipstick, in its current formulation, were not satisfactory.

Recent momentum for the control of neglected tropical diseases has generated a renewed interest in the control of schistosomiasis, resulting in the increase of large scale mass drug administration of praziquantel in several countries. As a consequence, there is a need to develop more sensitive diagnostic tools, especially for Schistosoma mansoni, in order to help targeting control or develop better screening strategies for elimination. We evaluated the commercially available point-of-contact circulating cathodic antigen (CCA) as a potential tool for the screening of S. mansoni infections. Our results showed that urine CCA assay had a better sensitivity for detecting S. mansoni infections than single and triplicate Kato-Katz thick smears. The prevalence obtained with CCA was similar or higher than that obtained with nine Kato-Katz. Neither STH nor S. haematobium infections had any impact on the CCA test results. Although further studies are required for its refinement and validation, the overall results revealed CCA as a promising and easy to use tool for the diagnosis of S. mansoni.

Abstract Introduction Methods Results Discussion

medicine infectious diseases test evaluation schistosomiasis public health and epidemiology diagnostic medicine epidemiology neglected tropical diseases epidemiological methods parasitic diseases

2012

Evaluation of Circulating Cathodic Antigen (CCA) Urine-Tests for Diagnosis of Schistosoma mansoni Infection in Cameroon

Severe fever with thrombocytopenia syndrome virus (SFTSV), an emerging human pathogen naturally transmitted by ticks, has spread widely since it was first detected in 2010. Although SFTSV-specific antibodies have been detected in wild birds, these natural reservoir and amplifying hosts for the virus have not been well studied. Here we report an experimental infection of spotted doves (Streptopelia chinensis) with two strains of SFTSV, JS2010-14, a Chinese lineage strain, and JS2014-16, a Japanese lineage strain, which represent the main viral genotypes currently circulating in East Asia. In these studies, we have determined that spotted doves are susceptible to SFTSV and the severity of the viremia is dose-dependent. When challenged with 107 and 105 PFU, all doves developed viremia which peaked 3–5 days post infection (dpi). Only a subset (25–62. 5%) of the birds developed viremia when challenged at 103 PFU. Virulence of SFTSV in spotted doves was strain dependent. Infection with 107 PFU of strain JS2014-16 resulted in 12. 5% mortality over 6. 8 days and mean peak viremia titers of 106. 9 PFU/mL in experimentally inoculated birds. All doves inoculated with 107 PFU of the JS2010-14 strain survived infection with relatively lower mean viremia titers (105. 6 PFU/mL at peak) over 6. 1 days. Our results suggest that spotted doves, one of the most abundant bird species in China, could be a competent amplifying host for SFTSV and play an important role in its ecology. Between the two SFTSV strains, the strain of the Japanese lineage caused mortality, higher viremia titers in infected birds over a longer time period than did the Chinese strain. Our observations shed light on the ecology of SFTSV, which could benefit the implementation of surveillance and control programs. Severe fever with thrombocytopenia syndrome virus (SFTSV) is a phlebovirus in the family Phenuiviridae and causes severe fever with thrombocytopenia syndrome (SFTS), a severe hemorrhagic fever disease in East Asia [1,2]. The disease is characterized by high fever and a drastic reduction of platelets and leukocytes leading multi-organ failure with mortality up to 10% in patients. SFTSV was first isolated from a patient in eastern China in 2010. By the end of 2017, more than 12,000 cases were reported in 23 provinces of China making the disease an important public health concern [3,4, 5,6]. The SFTSV is a tick-borne zoonotic virus that has been detected in or isolated from several species of ticks, especially Haemaphysalis longicornis, a widely-distributed tick species in East Asia [7,8, 9]. SFTSV has a broad spectrum of animal hosts but none of the animals thus far have been confirmed as reservoir hosts. Previous studies conducted in East Asia including China, South Korea, and Japan showed that many domesticated and wild animals were susceptible to SFTSV infection but had no or inconspicuous clinical signs [10,11,12,13]. Additionally, in our previous study we demonstrated that some species of migratory birds, such as swan geese (Anser cygnoides) and spotted doves, could be both parasitized by H. longicornis and infected by SFTSV. These two characteristics demonstrate the potential for these species to contribute to the long-distance spread of SFTSV via migratory flyways [7]. This theory could explain why SFTSV has spread rapidly in China and genetically related viral strains were identified in China, Japan and Korea within a relatively short time span. Experimental infection with SFTSV causes mild clinical disease with moderate viremia levels in some vertebrate animals, which might serve as amplifying hosts in the natural transmission cycle of SFTSV [14,15,16]. However, susceptible avian species and their responses to SFTSV infection has not been established. Spotted doves are common birds in China. This species is found in most parts of China in summer months, but in winter, most migrate to warmer areas of southern China [17]. Spotted doves are also common birds in Japan and Korea where SFTSV also circulates [18,19]. In this study, we challenged naive spotted doves with two genotypes of SFTSV to establish an avian model of infection. Our objective was to determine the susceptibility of spotted doves to SFTSV infection, examine virulence and duration of viremia to assess the potential role of doves in SFTSV ecology. All bird transport, handling, daily husbandry, and study protocols were conducted in strict accordance with the Animal Ethics Procedures and Guidelines of the People’s Republic of China (Regulations for Administration of Affairs Concerning Experimental Animals, China, 1988). Protocols were pre-reviewed and approved by the Ethics Committee of the Jiangsu Provincial Center for Disease Control and Prevention (Certificate No. JSCDCLL [2016]032). Moribund birds and all birds remaining at the end of the study were anesthetized with isoflurane gas and then euthanized with cervical dislocation. Two SFTSV strains, JS2010-14 of Chinese lineage (hereafter JS2010) and JS2014-16 (hereafter JS2014) of Japanese lineage, were used in the study. Both viral strains were isolated from SFTS cases in Jiangsu province of China in 2010 and 2014, respectively. The spotted doves were purchased from a commercial breeder of the species in China and held for two weeks to acclimate prior to SFTSV challenge. The birds used for the study were determined to be clinically healthy by a qualified veterinarian. Upon arrival, each bird was given a numbered leg band and caged in a biosafety level 3 animal facility. The spotted doves were provided 12 hr light/12 hr darkness, housed in groups of four or eight in wire cages measured at approximately 80 cm (long) x 60 cm (wide) x 60 cm (height) and were provided a commercial seed mix and water ad libitum. All birds enrolled in the study were males, 2 months of age and approximately 400 grams in weight. Two independent challenge studies were conducted. To investigate the susceptibility of spotted doves to SFTSV infection, doves in the first study were randomly assigned to one of four treatment groups: procedural controls (n = 4), and three SFTSV challenge groups, each given a different SFTSV dose: 103 (n = 8), 105 (n = 8), and 107 (n = 8) PFU. This study was replicated for both SFTSV strains, JS2010 and JS2014. The birds in the control groups from each replicate were housed together and separately from the virus challenged groups. On day 0 of the study, control birds were sham inoculated, specifically they were injected subcutaneously (s. c.) with 100 μL of serum-free Dulbecco’s Modified Eagle Medium (DMEM) as previously described [20]. The individual birds in the three SFTSV challenge groups were each inoculated s. c. with 107,105, or 103 PFU of a low passage (<3) human origin isolate of SFTSV suspended in 100 μL of DMEM according to their group assignment. A bird was considered infected with SFTSV if live virus was isolated from a serum sample at any sampling time point or if the bird developed anti-SFTSV antibodies. Each dove in the virus-inoculated groups was sampled on day 1 through 14 post-inoculation (pi). On each sampling day, 100 μL of blood was collected. Whole blood was allowed to clot for 30 min at room temperature in blood collection tubes and held at 4°C until centrifugation at 2000 x g for 10 min. Serum was collected and diluted in DMEM for a final serum: media dilution of 1: 5. The resulting diluted serum samples were stored at -80°C until testing. Following SFTSV challenge, birds were observed for clinical signs daily over 14 days. Birds that were moribund, as characterized by difficulty perching or other neurological signs, were humanely euthanized. At 14 dpi all birds were anesthetized with isoflurane gas and then euthanized with cervical dislocation. In the second study, for each SFTSV strain, 12 spotted doves were inoculated with a dose of 105 PFU with an additional four birds inoculated with serum-free DMEM to serve as negative controls. Three birds were selected randomly from the inoculated ones at 2,4, 7, and 14 dpi and euthanized after blood collection for necropsy to collect heart, liver, lung, spleen, kidney, and brain tissues. A small portion of each organ was collected, weighed, and homogenized in 1 mL of lysis buffer (Qiagen, Germany) using a mini-bead beater instrument (TissueLyser LT, Qiagen, Germany). Real-time RT-PCR was used to quantify SFTSV in the homogenized organs. The negative control birds were euthanized and necropsied at 14 dpi and their organs processed in the same way. The sera from the challenged birds were also tested for the presence of SFTSV and anti-SFTSV antibodies at 2,4, 7, and 14 dpi by virus isolation and ELISA. Virus titration was performed as described [20] using a 24-well plate for a mini-plaque assay technique to accommodate small sample volumes. A half milliliter of Vero cells (2 x 105 cells/ml) in DMEM was added to each well. The plates were incubated for 4 days at 37°C in an incubator with 5% CO2. Sera were individually centrifuged at low speed for clarification. Supernatants were diluted 1: 10 in DMEM containing 10% fetal bovine serum. Individually diluted viral inoculum was added to three wells with confluent Vero cell monolayers in the 24-well plate and incubated for 45 min after which the inoculum was removed. One milliliter of complete agarose overlay was added to each well. The 1: 10 serum: DMEM samples were screened for the presence of virus by plaque formation and the positives were further titrated on 12-well plates with ten-fold serial dilution for endpoint titration. Cell cultures were examined for plaque formation at 96,120, and 144 hrs pi and the number of plaques was recorded. Infectious virus titers were calculated as PFU/ml. RNA was extracted from 140 μL of the 1: 10 diluted serum samples using the QIAamp Viral RNA Mini kit (Qiagen). RNA was extracted from brain tissuewith the RNeasy Lipid Tissue Mini extraction kits (Qiagen) and from other organs using the RNeasy Mini extraction kit (Qiagen). Real-time RT-PCR was performed using the QuantiTech RT-PCR kit (Qiagen). The primers were designed as previously described and used in a one-step real-time RT-PCR [21]. The forward (S-for) /reverse (S-rev) primers and MGB probe (S-pro) used in the real-time RT-PCR were targeted to the S segment of the viral genome. Conditions for the reaction were as follows: 50°C for 30 min, 95°C for 15 min, 40 cycles at 95°C for 15 sec, and 60°C for 1 min. Amplification and detection were performed with an Applied Biosystems 7500 Real-time PCR system (Applied Biosystems, Foster City, CA). Data were analyzed using the software supplied by the manufacturer. Prior to challenge, we sampled the blood of all birds. All were to be seronegative for specific antibodies to SFTSV by plaque reduction neutralization assay (PRNT). Serum samples were also collected at 14 dpi, prior to euthanasia or at the time of death in the first study. All serum samples were heat-inactivated at 56°C for 30 min and tested for anti-SFTSV antibodies by PRNT on 12-well plates as described [15]. Samples exhibiting a neutralization of ≧90% were considered positive for antibodies to SFTSV (PRNT90). Additional sera were tested for both IgG and IgM SFTSV antibodies with a commercial double antigen sandwich ELISA kit from Xinlianxin Biotech (Wuxi, China). The assay was developed for detecting total antibodies specific to SFTSV in various animal species including birds [13]. Positive sera were 2-fold diluted starting at 1: 10 for the assay to obtain endpoint titers determined by the cutoff values set by the positive and negative ELISA controls. All statistical analyses were performed with SPSS 19. 0 (SPSS, Chicago, IL) and statistical significance level was set at 0. 05. For categorical data, the proportion and 95% confidence interval (CI) were calculated and differences in proportions were compared with the Fisher' s exact test. Unless indicated, all tests of proportions or means were two-sided. In the first study, our results demonstrate that spotted doves were infected and developed viremia after inoculation with either viral strain (Table 1, Fig 1). Viremia appeared in all birds challenged with 107 and 105 PFU of either viral strain. When challenged with the dose of 103 PFU, viremia was detected in fewer birds than in the groups challenged at higher doses. When challenged at 103 PFU, more birds were viremic after challenge with strain JS2014 of the Japanese lineage (5/8) than with strain JS2010, of the Chinese lineage (2/8) (Table 1). We were able to detect SFTSV-specific antibodies in doves challenged with each of the SFTSV strains. In the JS2010 and JS2014 103 PFU challenged groups, two and one additional birds developed anti-SFTSV antibodies without being detected as viremic, respectively. All control birds (n = 4) were negative as determined by viral isolation, viral specific antibody detection, and real-time RT-PCR and none died. Mortality was observed only in the group of the birds challenged with 107 PFU of the strain JS2014 (1/8,12. 5%) on day 7 pi. No birds died in the groups inoculated with 105 and 103 PFU of either virus strain and in the group given the JS2010 strain at 107 PFU. Our data suggest that the infection of SFTSV in spotted doves was primarily self-limiting and infected birds recovered after a defined period of viremia. Mean SFTSV viremia levels were highest at 3 dpi in the JS2014 107 PFU challenge group (mean = 106. 9 PFU/mL, SD 100. 3), followed by the 105 PFU group at 4 dpi (mean = 105. 5 PFU/mL, SD 100. 2), and by the 103 PFU group on day 5 (mean = 105. 3 PFU/mL, SD 100. 2). The mean SFTSV viremia levels for the JS2010 107,105, and 103 PFU challenge groups peaked at 4 dpi (mean = 105. 6 PFU/mL, SD 100. 3), 5 dpi (mean = 104. 5 PFU/mL, SD 100. 4), and 6 dpi (mean = 104. 3 PFU/mL, SD 100. 2), respectively. The mean peak day of viremia for individual birds of the JS2014 107 PFU challenge group (mean 2. 6 d) occurred significantly earlier compared to the other challenge groups (Table 1) (overall F = 9. 2, p<0. 01, Tukey' s multiple comparisons of 107 mean to 105 and 103 PFU, q = 5. 2 and q = 5. 1, respectively). With the strain JS2010, the mean peak day of viremia for individual birds in the 107 PFU challenge group was 3. 7 d, significantly earlier compared to the other challenge groups (Table 1) (overall F = 10. 2, p<0. 01, Tukey' s multiple comparisons of 107 mean to 105 and 103 PFU, q = 6. 2 and q = 5. 5, respectively). Viremia detected in birds from both 107 PFU challenge groups fell to the threshold of detection (5–6 days) more rapidly than either the 105 or 103 PFU challenge groups (7–8 days). In the second study, the virus was successfully isolated from sera in 9 of the 12 spotted doves challenged with JS2014 at 105 PFU. SFTSV specific antibodies were detected in all three inoculated birds at 14 dpi while they were negative by viral isolation. There was no mortality. In the JS2010 105 PFU challenge group, 6 of the 12 spotted doves were positive by viral isolation and three birds, negative for viral isolation at 14 dpi, developed anti-SFTSV antibodies. As for temporal viral distribution in organs of the birds challenged with 105 PFU of JS2014, SFTSV was detected by viral isolation or RT-PCR in multiple organs, including kidney, liver, heart, lung, and spleen obtained from each of the three sacrificed birds at 2,4, and 7 dpi (Fig 2). At 14 dpi, however, SFTSV was no longer detectable by either viral isolation or RT-PCR in any tissues from the three sacrificed birds. For the birds inoculated with 105 PFU of JS2010, SFTSV was detected by RT-PCR only, but not by viral isolation, in the spleen of the three sacrificed birds at 2 dpi. At 4 and 7 dpi, kidney, liver, heart, lung, and spleen were all positive with either viral isolation or RT-PCR in all three sacrificed birds. At 14 dpi, SFTSV was not detected by either method in any tissues (Fig 2). All control birds were negative in either viral isolation or RNA detection. Prior to the study we sampled the blood of all birds that were confirmed to be seronegative for specific antibodies to SFTSV by the PRNT assay. Final serum samples were collected at 14 dpi or at the time of death in the first study, and the antibodies to SFTSV were detected by the PRNT90 in 8/8 (100%), 8/8 (100%), and 6/8 (75%) of the infected birds in the groups challenged with 107,105, and 103 PFU of the strain JS2014, respectively. In the groups challenged with 107,105, and 103 PFU of the strain JS2010, specific antibodies to SFTSV were detected in 8/8 (100%), 8/8 (100%), and 4/8 (50%) of the infected birds, respectively (Table 1). Neutralizing antibody titers for SFTSV by PRNT90 were up to 690–860 in mean (presented as reciprocal of serum dilution) at 14 dpi (Fig 3). Neutralizing antibodies for SFTSV were detected earlier in the spotted doves challenged with 107 and 105 PFU than in birds given 103 PFU (Fig 3). In the first study, all infected birds underwent a period of anorexia that coincided with detectable viremia in the groups challenged with 107 and 105 PFU of both SFTSV strains. Moreover, one dove challenged with 107 PFU of JS2014 became lethargic, had ruffled feathers and thus was euthanized on day 7. This bird had the highest viremia in the group, (up to 107. 3 PFU). In the groups challenged with 103 PFU of either viral strain, only the birds with detectable viremia were anorexic. No other clinical signs were observed. A dose-related loss of body weight was detected in birds following challenges with either of the SFTSV strains in the first study (Fig 4). The birds challenged with 107 PFU JS2014 had the largest drop in body mass, with an average of 4. 7% by 4 dpi. Challenged with 105 and 103 PFU of the strain JS2014, the birds lost 3. 5% and 2. 3% of average body mass at 5 and 6 dpi, respectively. The birds challenged with 107 PFU of the strain JS2010 showed the greatest body mass loss of 3. 6% average at 5 dpi. The doves lost 2. 4% and 1. 3% of average body mass at 6 and 7 dpi, respectively, when challenged with 105 and 103 PFU of the strain JS2010. The JS2014 strain appeared to cause the earlier and more severe loss of body weight than did the JS2010 strain. Around 10 dpi the mean body weights had either returned to or exceeded their starting levels in all challenge groups. SFTS is an emerging zoonotic disease which is traced back to reports of an unknown infectious disease in rural areas of Hubei and Henan provinces in central China in 2009 [1]. The causative agent was not initially identified due to similar clinical manifestations caused by Anaplasma phagocytophilum, Hantaan virus, and Rickettsia tsutsugamushi infections [1,2]. A phlebovirus was finally isolated from a farmer in Henan, China and confirmed as the cause of SFTS [1]. Surveillance data showed that SFTSV spread to 23 provinces in China from 2010 to 2017 [3,4, 5,6, 22]. Furthermore, SFTS cases have been reported in other Asian countries including South Korea and Japan [23,24]. Several studies on the geographic distribution, genetic diversity, and prevalence of SFTSV genotypes have led to the proposal that there are two major SFTSV lineages, the Chinese and Japanese lineages [25]. Phylogenetically the SFTSV strains are grouped in 5 clades (A, B, C, D and E) based on the sequences of their genome segments (S1 Fig). Clades A, B, C and D are classified as the Chinese lineage, while clade E is classified as the Japanese lineage. SFTSV strains isolated from China include all 5 clades, the strains from South Korea fall into 3 clades (A, D, and E), and all strains from Japan are from only clade E [4,25,26]. At present, the SFTSV strains of clade E are the most widely disseminated in East Asia. Recent analyses indicate that SFTSV might have originated in the Dabie Mountain area in central China. According to the theory, several decades ago the virus was introduced to Shandong Province from Henan Province, and then to Liaoning Province in Northeastern China, the Zhoushan Archipelago of China, Jeju Island of South Korea, and to Japan from Jiangsu province [26]. Transmission across land and sea must have happened to explain the distribution of SFTSV and particularly the viruses of clade E [26]. Previous studies have found that some domestic animals and wildlife may be infected with SFTSV and might serve as amplifying or reservoir hosts in the natural transmission cycle of SFTSV [11,12,13,14]. Our recent study showed that some species of migratory birds, such as swan geese and spotted doves, can be parasitized by H. longicornis and infected by SFTSV in nature [7]. Other studies have reported that migratory bird routes and the distribution of H. longicornis in East Asia overlap with the geographic distribution of SFTSV [27,28]. Migratory birds are known to be carriers and transmitters of infectious agents, like the causative agents of influenza, West Nile encephalitis, Crimean-Congo hemorrhagic fever, and Lyme disease [29,30,31]. Wild birds often travel long distances carrying parasites, including ticks, which may be infected with viruses and bacteria. While birds play a key role in spreading the above-mentioned virus or bacterium, mosquitoes and ticks are involved in the transmission of the agents for West Nile Virus, Crimean-Congo hemorrhagic fever virus, and Lyme spirochetes, respectively. It is reasonable to hypothesize that migratory birds may have an important role in spreading SFTSV in two scenarios, i. e. , either the birds are infected directly with the virus or the birds are carriers of parasitic ticks that bear the virus. Spotted doves are a common migratory species in China. In this study, we challenged spotted doves with Chinese (clade A) and Japanese lineage (clade E) SFTSV strains to establish a bird infection model of SFTSV. We were also interested in examining host susceptibility to infection, viral pathogenicity and duration of viremia, in order to assess the potential roles of spotted doves as a reservoir and/or amplifying hosts of SFTSV. The results showed that the spotted doves were susceptible to both clades of SFTSV. Viremia appeared in all birds challenged with the doses of 107 and 105 PFU of both viral strains. Most of the spotted doves challenged with the dose of 103 PFU were infected and had detectable viremia or SFTSV-specific antibodies. Mortality was observed (1/8,12. 5%) only in the group of the birds challenged at 107 PFU with the clade E virus, or the Japanese strain JS2014, on day 7 pi. No birds died in either the JS2014 105 or 103 PFU challenged groups and in all challenge levels of the Chinese strain JS2010. This suggests that the infection of spotted doves with SFTSV was primarily self-limiting and infected birds mostly recovered after a period of viremia. Our data in this report indicate that differential susceptibilities to the two clades of viruses may occur in spotted doves. Of the two viral strains tested, JS2014 led to one death and higher viremia titers among infected birds, while JS2010 caused no fatalities and had relatively lower virus titers in the blood of inoculated birds (Fig 1). Although necropsy was not performed on the bird that died, it did have the highest viremia levels of the group prior to death. Therefore, the mortality event was most likely caused by viral infection. We speculate that with higher viremia titers, the birds could transmit the virus to feeding ticks more efficiently. Thus, as a potential amplifying host, the spotted doves may be more efficient in transmitting the Japanese lineage SFTSV. This is consistent with studies on the geographic distribution of SFTSV genotypes, i. e. , the SFTSV strains of the Japanese lineage are more widely disseminated geographically, possibly due to its higher replication efficacy in migratory birds such as doves. To date, only a few mammals have been used as models for the study of SFTSV infection, including mice, goats, hamsters and macaques [10,14,15,16]. To our knowledge, this is the first study to examine spotted doves as a host of SFTSV. The results showed that the spotted doves are susceptible to SFTSV infection, particularly by clade E or the Japanese origin strains, and could be a competent amplifying host species. Thus, these birds may play an important role in the long-distance transmission of the virus. Heartland virus is a phlebovirus closely related to SFTSV. Infection of Heartland virus in field collected avian hosts was not observed [32,33]. However, infection of SFTSV was observed in chickens and some species of field collected birds as recently reported [14,15,16]. Experimental inoculation of chickens with Heartland virus resulted in no viremia and scanty immune responsiveness [32,33]. In contrast, this study demonstrates that the spotted doves can be experimentally infected by SFTSV resulting in detectable viremia and immune responsiveness. A differential susceptibility to closely related phleboviruses appears to occur in birds. The underlying mechanism for the difference, however, remains unknown and deserves further studies.

Severe fever with thrombocytopenia syndrome virus (SFTSV) is an emerging human pathogen naturally transmitted by ticks. Our recent study has showed that some species of migratory birds, such as swan geese and spotted doves, can be parasitized by Haemaphysalis longicornis, an SFTSV vector, and antibodies against the SFTSV detected in these species. These data demonstrate that migratory birds are infected with SFTSV and may also play a role in the infection of ticks with SFTSV. Other studies have reported that migratory bird routes and the distribution of H. longicornis in East Asia overlap with the geographic distribution of SFTSV. Migratory birds are known to be carriers and transmitters of infectious agents, like the causative agents of influenza, West Nile encephalitis, and Lyme disease. Wild birds often travel long distances carrying various parasites, including ticks, which may be infected with viruses and bacteria. It is therefore reasonable to hypothesize that migratory birds may have played an important role in spreading SFTSV in two potential transmission scenarios: 1) birds are infected with the virus and transmit it back to ticks endemically or in a distal region, or 2) they are carriers of parasitic ticks that are infected with the virus. Here we report an experimental infection of spotted doves (Streptopelia chinensis) with two strains of SFTSV, JS2010-14 from a Chinese lineage and JS2014-16 from a Japanese lineage, which represent the main viral genotypes currently circulating in East Asia. We determined that spotted doves are susceptible to SFTSV and that the severity of the viremia was dose and strain dependent. Infection with the strain of JS2014-16 led to a death rate of 12. 5% and higher viremia titers in experimentally inoculated birds while doves inoculated with the JS2010-14 strain survived infection with relatively lower virus titers in the blood. These findings provide novel insights for understanding the rapid spread of the virus in a short time span, in particular the SFTSV strains from the Japanese lineage (genotype E).

Abstract Introduction Materials and methods Results Discussion

reverse transcriptase-polymerase chain reaction invertebrates medicine and health sciences immune physiology ixodes immunology social sciences vertebrates parasitic diseases animals euthanasia animal behavior molecular biology techniques ticks antibodies zoology research and analysis methods immune system proteins infectious diseases birds artificial gene amplification and extension proteins behavior molecular biology disease vectors arthropoda biochemistry psychology animal migration eukaryota arachnida viremia polymerase chain reaction physiology biology and life sciences viral diseases species interactions amniotes organisms

2019

Susceptibility of spotted doves (Streptopelia chinensis) to experimental infection with the severe fever with thrombocytopenia syndrome phlebovirus

Telomeres distinguish chromosome ends from double-strand breaks (DSBs) and prevent chromosome fusion. However, telomeres can also interfere with DNA repair, as shown by a deficiency in nonhomologous end joining (NHEJ) and an increase in large deletions at telomeric DSBs. The sensitivity of telomeric regions to DSBs is important in the cellular response to ionizing radiation and oncogene-induced replication stress, either by preventing cell division in normal cells, or by promoting chromosome instability in cancer cells. We have previously proposed that the telomeric protein TRF2 causes the sensitivity of telomeric regions to DSBs, either through its inhibition of ATM, or by promoting the processing of DSBs as though they are telomeres, which is independent of ATM. Our current study addresses the mechanism responsible for the deficiency in repair of DSBs near telomeres by combining assays for large deletions, NHEJ, small deletions, and gross chromosome rearrangements (GCRs) to compare the types of events resulting from DSBs at interstitial and telomeric DSBs. Our results confirm the sensitivity of telomeric regions to DSBs by demonstrating that the frequency of GCRs is greatly increased at DSBs near telomeres and that the role of ATM in DSB repair is very different at interstitial and telomeric DSBs. Unlike at interstitial DSBs, a deficiency in ATM decreases NHEJ and small deletions at telomeric DSBs, while it increases large deletions. These results strongly suggest that ATM is functional near telomeres and is involved in end protection at telomeric DSBs, but is not required for the extensive resection at telomeric DSBs. The results support our model in which the deficiency in DSB repair near telomeres is a result of ATM-independent processing of DSBs as though they are telomeres, leading to extensive resection, telomere loss, and GCRs involving alternative NHEJ. The repair of DNA double-strand breaks (DSBs) is vital for preventing gross chromosome rearrangements (GCRs) leading to cell death or cancer [1]. There are multiple mechanisms for DSB repair, including classical nonhomologous end joining (C-NHEJ) [1], homologous recombination repair (HRR) [2], and alternative nonhomologous end joining (A-NHEJ) [3]–[5]. The initial steps in DSB repair are similar for all three pathways, involving the binding of the MRE11/RAD50/NBS1 (MRN) complex to the DSB, followed by activation of ATM [6]. Phosphorylation of proteins by ATM is then instrumental in assembling a repair complex at the DSB, modifying chromatin surrounding the DSB to allow access to repair proteins, and activating cell cycle checkpoints to delay traversal through the cell cycle until repair is complete. The primary repair mechanism for DSBs in mammalian cells is C-NHEJ, which involves the direct joining of two DNA ends, utilizing the proteins KU70, KU86, DNA-PKcs, LIG4, XRCC4, XLF, and Artemis [1]. The preference for C-NHEJ in DSB repair is insured by the ATM-mediated activation of proteins that protect of the ends of the DSB. This protection involves a variety of proteins associated with the DSB repair complex, including 53BP1 [7]–[10], histone γH2AX [11], and the MRN complex [12], [13]. When DSBs are not repaired in a timely manner, the ends of the DSB are eventually processed and resected to form single-stranded 3′ overhangs [5], [14], allowing the repair of DSBs by either HRR or A-NHEJ [2], [4]. The processing of DSBs is regulated by ATM through the activation of MRE11 [15] and CtIP [14], [16]–[18]. Following the processing of the DSB by MRE11/CtIP, resection of the 5′ end of the DSB is then mediated by EXO1 exonuclease in both yeast [19], [20] and mammalian cells [13], [21]. However, the extent of resection required, the timing in the cell cycle, and the consequences of HRR and A-NHEJ are very different. HRR requires large single-stranded 3′ overhangs to initiate repair using the complementary sequence on the sister chromatid [2], which involves activation of BRCA1 by ATM for removal of 53BP1 in late S phase and G2 [7]–[10]. Like HRR, A-NHEJ also requires the processing of DSBs by MRE11 [22]–[25] and CtIP [18], [26], [27]. MRE11 is also required for A-NHEJ in Xenopus [28] and S. cerevisiae, where the nuclease activity of MRE11 is necessary to release KU proteins and the MRN complex from DNA ends [12]. However, unlike HRR, DSB repair by A-NHEJ involves end joining at sites within the single stranded regions, which is often facilitated by the presence of microhomology [22], [23], [27], [29], [30], and is commonly associated with deletions [23], [31] and GCRs [26], [30]–[33]. Because of its unique characteristics, A-NHEJ is also referred to as microhomology-mediated end joining [4], [34], deletional NHEJ [24], or backup-NHEJ [5]. Although it is clear that a deficiency in C-NHEJ can promote repair of DSBs by A-NHEJ [3]–[5], it is less clear how DSBs are routed into the A-NHEJ pathway in cells that are proficient in C-NHEJ. Like HRR, ATM prevents A-NHEJ in G1 through the activation of 53BP1 and γH2AX, which work together to protect DNA ends [7]–[11]. However, unlike HRR, BRCA1 is not required for A-NHEJ, demonstrating that A-NHEJ can occur without extensive resection [18], [22]. Therefore, in addition to the inhibition of C-NHEJ, it has been pointed out that A-NHEJ can also be promoted by the stimulation of short-range MRE11/CtIP-mediated resection, the inhibition of HRR, or the inhibition of long range EXO1-dependent resection [34]. Some DSBs are more difficult to repair than others. This difference in repair efficiency is obvious from the fact that although most DSBs generated by ionizing radiation are repaired within a few hours, approximately 10 to 20% are repaired much more slowly [35]. Importantly, the DSBs that are slowly repaired are more likely to result in GCRs [36], [37]. Several factors can influence the efficiency of DSB repair. First, as originally proposed by John Ward [38], DSBs located adjacent to other radiation-induced DNA lesions, termed localized multiply damaged sites, are refractory to repair [36]. Second, DSB repair can also be influenced by chromatin structure, as shown by the fact that DSBs occurring within heterochromatin are repaired slowly and require ATM-mediated chromatin modifications that are not required for repair of DSBs that occur in euchromatin [14], [39]–[41]. Finally, as discussed below, the efficiency of DSB repair can also be influenced by the proximity of telomeres. Telomeres are cap structures found on the ends of chromosomes that protect chromosome ends and keep them from appearing as DSBs [42], [43]. Telomeres are therefore essential for preventing chromosome fusion and genomic instability [44]–[46]. Telomeres in mammalian cells are composed of a 6 base pair repeat sequence, TTAGGG, which is added on by the enzyme telomerase [42]. TRF1 and TRF2 specifically bind to these telomeric repeat sequences and recruit RAP1, TIN2, TPP1, and POT1, which combine to generate the shelterin complex that regulates telomerase activity and protects chromosome ends [43]. Apollo exonuclease is also recruited to telomeres through the interaction with TRF2 and generates single-stranded 3′ overhangs by resection of the 5′ end of the leading strand, which is initially blunt-ended following DNA replication [47], [48]. The nuclease activity of MRE11 also contributes to the maintenance of the single-stranded 3′ overhang in a TRF2-dependent manner [49]–[51], and EXO1 nuclease functions to elongate the single-stranded 3′ overhang [52]. The single-stranded 3′ overhang is required for the association of POT1 and its partner TPP1, which promote the formation of the t-loop that is necessary for telomere end protection. The extent of processing of the end of the chromosome is limited by the binding of POT1, so that deficiencies in POT1 or TPP1 result in long single-stranded 3′ overhangs on the end of the chromosome [53]–[55]. We have previously demonstrated that telomeric regions are deficient in NHEJ [56]. A similar deficiency in NHEJ at interstitial sites containing telomeric repeat sequences led us to propose that the telomere-specific binding protein TRF2 actively suppresses C-NHEJ as part of its role in protecting the end of the chromosome [44], [56]. One model for the inhibition of NHEJ near telomeres involves the inhibition of ATM by TRF2 [57]. As at interstitial DSBs, ATM may be involved in the protection of DSBs by activation of 53BP1 and H2AX [7]–[11]. Alternatively, ATM could be required for repair of DSBs near telomeres, because subtelomeric regions are heterochromatin [58], [59], and ATM is required for DSB repair in heterochromatin [14], [39], [41]. However, we found no difference in HRR near telomeres [56], which in view of the requirement for ATM in HRR [2], suggests that ATM is functional near telomeres. We therefore proposed a second model in which the sensitivity of telomeric regions to DSBs is due to the inappropriate processing of DSBs, which would generate large single-stranded 3′ overhangs that are poor substrates for C-NHEJ [11], [51], [60], [61], but good substrates for A-NHEJ [18], [22]–[27]. In this model, DSBs within subtelomeric regions are processed to generate a single-stranded 3′ overhang in the same fashion as telomeres, which occurs through an ATM-independent process involving the regulation of the Apollo and/or MRE11 nucleases by TRF2. Two recent studies have reported that persistent DSBs near telomeres in normal human cells in culture and in vivo contribute to ageing and ionizing radiation-induced senescence [62], [63]. Importantly, one of these studies showed that the ectopic localization of TRF2 caused a delay in repair of interstitial DSBs in mammalian cells, and the presence of telomeric repeat sequences inhibited NHEJ and the recruitment of the NHEJ protein LIG4 in yeast [63]. Cell senescence caused by oncogene expression was also shown recently to result from telomere dysfunction in normal human fibroblasts in culture and in preneoplastic cells in vivo [64]. The fact that the dysfunctional telomeres in most of the senescent cells still contained telomeric repeat sequences led the authors to conclude that irreparable DSBs near telomeres rather than telomere loss were responsible. Importantly, senescence due to telomere dysfunction was not observed in malignant tumors, consistent with our model that telomere loss resulting from oncogene-induced replication stress in tumor cells that lack of cell cycle checkpoints serves as a mechanism for GCRs in human cancer [46]. In the current study, we have investigated the effect of ATM deficiency on the consequences of DSBs near telomeres to determine how the mechanism of repair of DSBs differs at telomeric and interstitial sites. Importantly, investigating the role of ATM allowed us to determine whether the sensitivity of telomeric regions to DSBs is due to the inhibition of ATM, which could result in the loss of end protection and failure to repair DSBs, or whether it is due to the inappropriate processing of DSBs as though they are telomeres, which would be independent of ATM. Both of these mechanisms would involve the known functions of TRF2 mentioned above, and are consistent with the ability of TRF2 to inhibit DSB repair [63]. The approach we used involved generating DSBs at specific telomeric or interstitial locations with I-SceI endonuclease. We employed three different assay systems used in our earlier studies, one that uses the activation of the gene for green fluorescent protein (GFP) to monitor the frequency of NHEJ, one that uses the inactivation of the GFP gene to monitor large deletions, and a PCR-based assay to monitor the frequency of small deletions [56]. In addition, we included a fourth assay system that monitors the frequency of GCRs using the activation of the DsRed gene. While this assay does not detect GCRs involving large deletions, it does provide a method for determining the relative frequency of GCRs at different locations. Using these four assay systems, we compared how the inhibition of ATM kinase activity by KU55933 or the knockdown of ATM expression by shRNA affects the types of events resulting from DSBs generated at telomeric and interstitial sites. Importantly, comparing the relative proportion of the four different types of events rules out the possibility that the results can be explained solely by a difference in the frequency of DSBs generated by I-SceI at interstitial and telomeric sites. The results demonstrate that ATM is functional near telomeres and is required for the protection of DSBs, despite the fact that ATM can be inhibited by TRF2. The results also show that the large deletions resulting from DSBs near telomeres are independent of ATM, and therefore do not occur through the mechanism involved in the processing and resection of DSBs at interstitial sites. The results are therefore consistent with our model in which the sensitivity of telomeres to DSBs is due to the inappropriate processing of DSBs as though they are telomeres, which leads to extensive resection and GCRs involving A-NHEJ. The studies presented here rely on four assays specifically designed to compare the types of events occurring as a result of DSBs at interstitial and telomeric sites. The DSBs at specific locations in these assay systems are generated with the I-SceI endonuclease, which introduces DSBs at an 18 bp recognition sequence found in integrated plasmid DNA. The first assay system used in our studies determines the frequency of large deletions by monitoring the loss of GFP expression following the expression of I-SceI endonuclease in cell clones containing the pGFP-ISceI plasmid (Figure 1A). The I-SceI site in the pGFP-ISceI plasmid is located between the GFP coding sequence and its chicken β-actin promoter. The loss of expression of the GFP gene in cell clones containing the pGFP-ISceI plasmid requires deletions larger than 20 bps at the I-SceI site, which is necessary to delete the start codon for the GFP gene or truncate the chicken β-actin promoter. This assay system is therefore capable of distinguishing larger deletions from small deletions of a few bps, which are the most frequent events at interstitial I-SceI-induced DSBs [65]–[67]. We previously used this assay system to demonstrate a high frequency of large deletions at DSBs near telomeres [56]. Although not apparent from this GFP-based assay system, our previous studies involving Southern blot analysis of genomic DNA from individual subclones demonstrated that the I-SceI-induced deletions near telomeres are much larger than those observed at interstitial sites, typically resulting in the loss of the entire 7 kb plasmid and the telomere [68]. In addition to large deletions, GCRs or direct telomere addition occurring at or near the I-SceI site (chromosome healing) would also result in loss of GFP expression in this assay. However, the frequency of these events is very low [68]–[71]. The loss of GFP expression can also occur through changes in chromatin structure, which is increased near telomeres due to telomere position effect [59], [72], [73]. However, this does not affect the results of this assay, because although the expression of the telomeric GFP gene is gradually reduced during passage in culture, cells with complete silencing of the GFP gene are rare [56]. The second assay system used in our studies determines the frequency of NHEJ by monitoring the appearance of GFP+ cells following the expression of I-SceI endonuclease in cell clones containing the pEJ5-GFP plasmid (Figure 1A). The activation of the GFP gene in the pEJ5-GFP plasmid results from NHEJ between the distal ends of two different I-SceI sites located at either end of the puro gene, which is inserted between the GFP gene and its promoter [27]. The I-SceI site may or may not be retained in the process, depending on whether the ends are directly rejoined or joined after the loss or addition of nucleotides at the DSB. We previously used this assay system to demonstrate a deficiency in NHEJ near telomeres [56]. The third assay system used in our studies determines the frequency of GCRs using the same clones containing the pEJ5-GFP plasmid that were used for the analysis of NHEJ. However, this assay system monitors the frequency of activation of the DsRed gene as a result of rearrangements in which one of the I-SceI sites in the pEJ5-GFP plasmid is joined with the I-SceI site in a pDsRed-ISceI plasmid that is integrated at a different location in the genome (Figure 1B). The DsRed gene in the integrated pDsRed-ISceI plasmid is initially inactive due to the absence of a transcriptional promoter, but is activated when the I-SceI-induced DSB at the 3′ end of the chicken β-actin promoter in the pEJ5-GFP plasmid is joined with the I-SceI-induced DSB at the 5′ end of the DsRed gene. The cell clones containing both the pEJ5-GFP and pDsRed-ISceI plasmids can therefore be used to simultaneously monitor the frequency of NHEJ (joining two I-SceI sites in close proximity - green cells) and GCRs (joining two I-SceI sites on different chromosomes - red cells) as a result of interstitial or telomeric I-SceI-induced DSBs. A similar assay system involving the activation of a selectable neo gene by I-SceI-induced DSBs has previously been used to investigate the mechanisms involved in the formation of chromosome translocations [26], [33], [74], [75]. The fourth assay system used in our studies determines the frequency of small deletions occurring during rejoining of the ends of one of the I-SceI sites located in the pEJ5-GFP plasmid. For this assay, genomic DNA from cells expressing I-SceI endonuclease is first amplified by PCR using primers that span one of the I-SceI sites (Figure 1A), and the PCR product is then digested with I-SceI endonuclease to determine the fraction of the PCR product that has lost the I-SceI site, i. e. is not cut. The percentage of cells in the population that contain small deletions is then determined after correcting for the frequency of NHEJ and large deletions (see Materials and Methods). Using this assay system, we previously reported that there is little difference in the frequency of small deletions at interstitial and telomeric I-SceI-induced DSBs [68]. The cell clones described above that contain the pDsRed-ISceI plasmid integrated at an interstitial site and the pEJ5-GFP plasmid integrated at either an interstitial (EDS-7F) or telomeric (EDS-6J) site were used to determine the frequency of NHEJ and GCRs at interstitial and telomeric DSBs. Following infection with the pQCXIH-ISceI retrovirus and selection with hygromycin for 14 days, the percentage of cells expressing GFP or DsRed was determined by flow cytometry (Figure 2A). Consistent with our earlier studies [56], the frequency of NHEJ (GFP+ cells) was lower in clone EDS-6J8 with a telomeric pEJ5-GFP plasmid than in clone EDS-7F2 with an interstitial pEJ5-GFP plasmid (data not shown). In contrast, the frequency of GCRs (DsRed+ cells) was much greater in clone EDS-6J8 than in clone EDS-7F2 (Figure 2B). This difference in NHEJ and GCRs at interstitial and telomeric DSBs is evident from the much lower ratio of GFP+ to DsRed+ cells in EDS-6J clones containing the telomeric pEJ5-GFP gene compared to EDS-7F clones containing the pEJ5-GFP gene at an interstitial site (Figure 2C). The large standard deviation observed in the GFP+ to DsRed+ ratio in clone EDS-7F2 is a result of the extremely low level of DsRed+ cells in clones with an interstitial pEJ5-GFP plasmid. This low frequency of GCRs in the EDS-7F clones is consistent with the low frequency of translocations (3–5×10−5) previously reported to result from rearrangements between two I-SceI-induced DSBs on different chromosomes [26], [33], [74], [75]. Importantly, the frequency of GCRs at telomeric DSBs in the EDS-6J clones is underestimated in our system, because it does not detect GCRs that occur in combination with large deletions, which as we have previously shown, represent the majority of rearrangements at telomeric DSBs [68], [71]. In addition, the efficient repair of the DSB in the interstitial pDsRed-ISceI plasmid will limit the frequency of GCRs detected by this assay. Proof that repair of DSBs within the interstitial pDsRed-ISceI plasmid are rate limiting is demonstrated by the fact that the EDS-6J7 and EDS-6J10 clones that contain three tandem copies of the pDsRed-ISceI plasmid have approximately a 3-fold higher frequency of DsRed+ cells then the EDS-6J8 clone containing a single copy of the pDsRed-ISceI plasmid (Figure 2C). Regardless of these limitations, this assay provides an analysis of the relative differences in the frequency of GCRs at different locations, and clearly shows that the deficiency in NHEJ near telomeres is associated with an increase in GCRs. To confirm that the expression of the DsRed gene results from recombination between the I-SceI sites in the pEJ5-GFP and pDsRed-ISceI plasmids, PCR was performed using one primer specific for the chicken β-actin promoter in the pEJ5-GFP plasmid, and one primer specific for the DsRed gene in the pDsRed-ISceI plasmid (see Figure 1B). As expected, genomic DNA from the parental EDS-6J7 and EDS-6J8 cell clones produced no PCR product (data not shown). However, following infection with the pQCXIH-ISceI retroviral vector and selection with hygromycin, both clones showed a PCR product of the size expected for NHEJ between the I-SceI sites in the two plasmids (data not shown). To analyze individual recombination junctions in DsRed+ cells, we performed flow sorting to isolate pooled populations of DsRed+ cells from clones EDS-6J7 and EDS-6J8 expressing I-SceI endonuclease. The pooled populations of DsRed+ cells were then plated out at low density and individual colonies were selected to isolate individual subclones expressing the DsRed gene. DNA sequence analysis of the PCR products demonstrated deletions of 1 to 148 bps (average 21 bps) at the I-SceI site in 16 of 17 (94%) DsRed+ subclones (Figure 3). Insertions of 3 to 6 bps (3 of 17 subclones, 19%) and microhomology of 1 to 4 bps (10 of 17 subclones, 59%) were also observed. The high frequency of deletions and microhomology found at the recombination junctions suggests that A-NHEJ is commonly involved in the formation of GCRs in our assay, consistent with previous studies in which A-NHEJ was found to be involved in translocations involving two different I-SceI sites [26], [33], [74], [75]. Importantly, the frequency of loss of the I-SceI site during the formation of GCRs in our system is greater than was previously observed during NHEJ at interstitial DSBs (40%), but is similar to the frequency of loss of the I-SceI site during NHEJ at telomeric DSBs (60%) [56], suggesting that A-NHEJ is involved in DSB repair near telomeres. Consistent with our earlier studies [56], [68], the expression of I-SceI endonuclease in clone GFP-7F1 resulted in large deletions (loss of GFP expression) at interstitial DSBs in 6. 0% of the cells, while I-SceI-induced DSBs near a telomere in clone GFP-6D1 resulted in large deletions in 47. 3% of the cells (Figure 4). Therefore, compared to DSBs at interstitial sites, which usually result in small deletions [65]–[67], DSBs near telomeres are much more likely to result in large deletions. In clone GFP-7F1, the inhibition of ATM kinase activity with KU55933 or knockdown of ATM expression with shRNA resulted in a small increase in large deletions at interstitial DSBs beyond that caused by I-SceI endonuclease alone (6. 9% and 4. 0%, respectively). Combining KU55933 and shRNA knockdown resulted in a small additional increase in the frequency of large deletions beyond that seen with KU55933 alone (9. 0%). A deficiency in ATM therefore has little effect on the frequency of large deletions at interstitial DSBs. Although ATM plays an important role in protecting DSBs from resection [7]–[11], [13], the frequency of large deletions at interstitial DSBs in ATM-deficient cells may be minimized by a corresponding deficiency in the processing of DSBs, because ATM is also involved in the activation of the MRE11/CtIP nuclease activity that is required for the processing DSBs [14], [16]–[18], [76]. The role of MRE11 in C-NHEJ and A-NHEJ is also partially independent of ATM [22], however, this is unlikely to affect the frequency of large deletions, because without ATM, any processing by MRE11 would result in limited resection due to a failure to activate BRCA1 [18], [77] and EXO1 [13], [21] in ATM-deficient cells. A deficiency in ATM had a much greater effect on the frequency of large deletions at DSBs near telomeres. The treatment of clone GFP-6D1 with KU55933 increased the frequency of large deletions at DSBs near telomeres by an additional 17. 9% beyond the already high frequency caused by I-SceI endonuclease alone, so that 65. 2% of the cells contain large deletions (Figure 4). Knockdown of ATM with shRNA in clone GFP-6D1 also increased the frequency of large deletions, although to a lesser extent than with KU55933, increasing the frequency of large deletions by an additional 7. 9% beyond that seen with I-SceI endonuclease alone. Combining KU55933 and shRNA knockdown resulted in a slight increase in the frequency of large deletions beyond that seen with KU55933 alone, so that 67. 7% of the cells contain large deletions. The fact that KU55933 or knockdown of ATM does not prevent large deletions at DSBs near telomeres suggests that, unlike interstitial DSBs, the processing and resection of DSBs near telomeres is not dependent on ATM. In contrast, the increase in large deletions caused by a deficiency in ATM at telomeric DSBs suggests that ATM is involved in protecting DSBs near telomeres, as it is at interstitial DSBs. The mechanism of repair of DSBs near telomeres was also investigated by comparing the effect of KU55933 and/or knockdown of ATM expression on the frequency of NHEJ (GFP+ cells) in clones EDS-7F2 and EDS-6J8 that contain the pEJ5-GFP plasmid integrated at interstitial or telomeric sites, respectively. Our results demonstrated that treatment with KU55933 caused a 57% increase in the frequency of NHEJ in cell clone EDS-7F2, while the knockdown of ATM by shRNA caused no change in the frequency of NHEJ (Figure 5A). Previous studies have found variable effects of KU55933 on NHEJ using similar assays. One study found no effect of KU55933 on NHEJ [23], one study found a decrease in NHEJ [22], and one study found an increase in NHEJ [78]. The latter study proposed that the increase in joining the distal ends of two I-SceI-induced DSBs in ATM-deficient cells, as detected by this NHEJ assay, was a result of the loss of tethering of proximal ends of I-SceI-induced DSBs. However, this conclusion is not consistent with our results showing that knockdown of ATM by shRNA had no effect on NHEJ at interstitial DSBs. Although we cannot rule out that the knockdown of ATM was insufficient to affect NHEJ, other studies have shown that kinase dead ATM can have very different effects from a deficiency in ATM. One study reported that inhibition of ATM kinase activity by KU55933 prevented HRR, while ATM deficient cells showed no change in HRR [79]. Mouse ES cells expressing kinase-deficient ATM were also found to have more chromosome instability than ATM knockdown cells, and the expression of kinase-deficient ATM results in embryonic lethality, while ATM knockout mice are viable [80], [81]. Therefore, because the kinase activity of ATM is dispensable for recruitment of ATM to damaged sites [80], [82], it was proposed that the displacement of ATM or NHEJ proteins from damaged sites requires ATM kinase activity, and that without this kinase activity the chromatin-associated ATM physically blocks the resection that is required for HRR [83]. Importantly, the inhibition of HRR can promote NHEJ [14], which could explain the increase in NHEJ in cells treated with KU55933 in our assay. The effect of inhibition of ATM on NHEJ is very different at telomeric DSBs than it is at interstitial DSBs. Unlike clone EDS-7F2, KU55933 and/or knockdown of ATM expression in clone EDS-6J8 caused a large decrease in the already low frequency of I-SceI-induced NHEJ (Figure 5A). The frequency of NHEJ was further decreased by combining KU55933 and knockdown of ATM, which together resulted in a nearly a 10-fold reduction in NHEJ. Importantly, the decrease in NHEJ near telomeres in ATM-deficient cells corresponded to the increase in large deletions (see Figure 4), consistent with a failure to protect DSBs near telomeres in ATM-deficient cells. Unprotected DSBs would result in increased resection, which would reduce the frequency of NHEJ in our assay, both because of degradation of the GFP gene, and because single stranded overhangs are poor substrates for C-NHEJ. We next investigated the role of ATM in the formation of GCRs by analyzing the frequency of DsRed+ cells using the same EDS-7F2 and EDS-6J8 clones that were used for analysis of NHEJ. In clone EDS-7F2, the inhibition of ATM by KU55933 and/or shRNA-mediated knockdown of ATM caused no apparent change in I-SceI-induced GCRs (DsRed+ cells, Figure 5B). These results suggest that ATM is not required for GCR formation at interstitial DSBs, although small changes in the frequency of GCRs may not be detected due to the very low frequency of DsRed+ cells at interstitial DSBs. It is not immediately clear why ATM would not be required for GCRs, because CtIP, which is activated by ATM, is required for GCRs [26]. One possibility is that despite the requirement for ATM in the activation of CtIP [14], [16]–[18], a deficiency in ATM would also eliminate the requirement for MRE11/CtIP for processing of DSBs due to the lack of end protection, which also requires activation by ATM [7]–[11]. Therefore, ATM-deficient cells may have sufficient processing of DSBs to provide a substrate for A-NHEJ without resulting in the extensive resection leading to large deletions. This possibility is consistent with the fact that ATM is required for HRR, which requires extensive resection, but is not required for the limited processing required by A-NHEJ [18], [77]. In contrast to clone EDS-7F2, the inhibition of ATM by KU55933 and/or shRNA-mediated knockdown of ATM in clone EDS-6J8 with a telomeric pEJ5-GFP plasmid caused a large decrease in I-SceI-induced GCRs (Figure 5B). Moreover, similar to NHEJ, the decrease in the frequency of GCRs was additive when KU55933 and ATM knockdown were combined. Therefore, the decrease in GCRs near telomeres in ATM deficient cells, as is detected by this assay (those with relatively small deletions), is most likely a result of the increased frequency of large deletions due to excessive resection at the DSB. It is important to point out that these results do not mean that the inhibition of ATM prevents GCRs at DSBs near telomeres, because this assay only detects GCRs that occur with minimal degradation at the DSB, and large deletions at DSBs near telomeres commonly result in GCRs [68], [71]. We next determined the effects of ATM deficiency on the frequency of small deletions at one of the I-SceI sites in the pEJ5-GFP plasmid following expression of I-SceI endonuclease in clones EDS-7F2 and EDS-6J8 (Figure 6). Similar to the NHEJ assay, these small deletions also involve NHEJ, but unlike the NHEJ assay, the assay for small deletions only detects NHEJ events in which the I-SceI site is lost. The actual number of DSBs at an I-SceI site is much greater, since I-SceI sites are commonly restored by NHEJ [64], [84]. As we previously reported [68], the frequency of small deletions at interstitial and telomeric DSBs is very similar (Figure 6). As a result, small deletions outnumber large deletions at interstitial DSBs, while they equal less than half the number of large deletions at telomeric DSBs (compare Figure 4 and Figure 6) [68]. Similar to NHEJ, the inhibition of ATM kinase activity with KU55933 in clone EDS-7F2 with an interstitial pEJ5-GFP plasmid caused a significant increase in the frequency of small deletions, while the knockdown of ATM expression by shRNA had no effect (Figure 6). Unlike clone EDS-7F2, a deficiency in ATM in clone EDS-6J8 with a telomeric pEJ5-GFP plasmid dramatically reduced the frequency of small deletions at the I-SceI site, both with KU55933 and with knockdown by shRNA (Figure 6). The combined treatment with KU55933 and shRNA knockdown had an even larger effect, decreasing the percentage of cells with small deletions to only 2. 6%. The effect of ATM deficiency on small deletions therefore mimics the effect of ATM deficiency on the NHEJ assay (Figure 5A), suggesting that small deletions at a single I-SceI site and NHEJ between distal ends of two different I-SceI sites occur through the same pathway. As with NHEJ, the decrease in small deletions corresponds to an increase in large deletions, strongly suggesting that small deletions near telomeres become large deletions in ATM-deficient cells. These results therefore add additional support for our conclusion that DSBs near telomeres require ATM for protection, but that the extensive processing and resection at DSBs near telomeres is not dependent on ATM. The data presented here confirm and extend our earlier results that telomeric regions are highly sensitive to DSBs. In addition to the decrease in NHEJ and the increase in large deletions we have previously reported [56], [68], we now show that DSBs near telomeres have a much greater likelihood of generating GCRs. The frequency of GCRs is 20–50 fold higher at DSBs near telomeres compared to DSBs at interstitial sites (Figure 2). The effect of ATM deficiency on the types of events resulting from I-SceI-induced DSBs also demonstrates important differences in the mechanism of DSB repair at interstitial and telomeric sites. Treatment with KU55933 or knockdown of ATM expression by shRNA caused a much larger increase in the frequency of large deletions at telomeric DSBs than at interstitial DSBs. This increase in large deletions would explain the corresponding decrease in small deletions and NHEJ at telomeric DSBs, both indirectly because of increased degradation, and directly because resected DNA is a poor substrate for C-NHEJ [11], [51], [60], [61]. Although most cells in the population show no rearrangements at the I-SceI site in clones with interstitial DSBs (Figure 7), this does not mean that DSBs were not generated at the I-SceI site in these cells, because the I-SceI site is commonly restored during DSB repair [84]–[86]. Therefore, the predominance of large deletions in clones with telomeric I-SceI sites (Figure 7) strongly suggests that I-SceI sites that are restored by NHEJ at interstitial DSBs are much more likely to become large deletions at telomeric DSBs. This conclusion is consistent with our model in which DSBs near telomeres are prone to excessive resection that inhibits C-NHEJ [44], [56]. This model proposes that excessive resection at DSBs near telomeres results in GCRs involving A-NHEJ, because A-NHEJ requires single-stranded overhangs [18], [22]–[27]. A role for A-NHEJ at DSBs near telomeres is supported by the fact that A-NHEJ is associated with large deletions [23], [31] and GCRs [26], [30]–[33]. Excessive resection at telomeric DSBs would also explain why HRR is not deficient near telomeres [56], because HRR also requires extensive single-stranded 3′ overhangs [2]. The combined effect of ATM deficiency in our assays does not support the hypothesis that the sensitivity of telomeric regions to DSBs is due solely to the inhibition of ATM by TRF2, which would prevent the ATM-mediated changes that are required for repair of DSBs in heterochromatin [14], [87]. The fact that the inhibition of ATM increases the frequency of large deletions at telomeric DSBs demonstrates that ATM is functional near telomeres and would still be involved in the protection of DSBs, similar to its role at interstitial DSBs [7]–[11], although we cannot rule out the possibility that a partial inhibition of ATM near telomeres contributes to the sensitivity of telomeric regions to DSBs by decreasing end protection. Moreover, a localized deficiency in ATM would not in itself result in the high frequency of large deletions near telomeres, because the inhibition of ATM prevents resection of unprotected DSBs in G1 [11] and in heterochromatin in G2 [14]. Regardless, the observation that the inhibition of ATM did not prevent the formation of large deletions is consistent with our model in which large deletions at DSBs near telomeres occur by a mechanism that is different from the ATM-dependent processing and resection at interstitial DSBs. As an alternative, we have proposed that DSBs near telomeres are mistakenly processed as though they are telomeres, which involves Apollo and/or MRE11 and is mediated by TRF2, and is independent of ATM [44], [56]. This model is consistent with the observation that tethering TRF2 near interstitial DSBs inhibits their repair [63]. Unlike the processing of telomeres, the inability of POT1 to bind to the non-telomeric single-stranded 3′ overhangs at subtelomeric DSBs would result in extensive resection and GCRs involving A-NHEJ, as is the case with telomeres in POT1-deficient cells [53], [54], [88]. A role for A-NHEJ in repair of telomeric DSBs is supported by the presence of microhomology at the recombination junctions involved in GCRs resulting from DSBs near telomeres (Figure 3), since A-NHEJ commonly utilizes microhomology during end joining [22], [23], [27], [29], [30]. Microhomology has also been observed at recombination junctions involved in chromosome fusions resulting from DSBs near telomeres [68], [89] or telomere shortening [90]. However, one study found that although significant microhomology was found at chromosome fusions in fibroblasts in crisis and in the invasive ductal carcinoma stage of breast cancer, no significant microhomology was found in the ductal carcinoma in situ stage of breast cancer, suggesting that chromosome fusions commonly occur by C-NHEJ in early stages of breast cancer [91]. It is important to note, however, that the PCR-based assay used for analysis of chromosome fusion in this study does not detect the chromosome fusions described in our studies, which commonly involve large deletions that would result in the loss of the regions near telomeres required for PCR analysis. Although a deficiency in ATM reduced the high frequency of DsRed+ cells that result from DSBs near telomeres in our assay system, it is important to point out that this does not mean ATM is required for A-NHEJ, which is involved in the formation of most GCRs [26], [30]–[33]. As mentioned earlier, this assay does not detect GCRs that involve large deletions, because GCRs with large deletions would not activate the DsRed gene. A deficiency in ATM causes a substantial increase in the frequency of large deletions at DSBs near telomeres (Figure 4), which commonly result in GCRs [68], [71]. Therefore, a deficiency in ATM would not result in an overall decrease in GCRs at DSBs near telomeres, only a decrease in GCRs with little or no degradation at the recombination junction. The use of I-SceI endonuclease to generate DSBs involves certain caveats and limitations that should be considered when interpreting the results. First, the direct ligation of I-SceI-induced DSBs, which regenerates the I-SceI site, is a common event [84]–[86] that is not detected by our assay systems. As a result, each I-SceI site can be cut multiple times, which can alter the consequences of DSBs [85] and amplify the frequency of rearrangements at the I-SceI site. Second, due to differences in accessibility of the I-SceI endonuclease, I-SceI-induced DSBs may be more likely to be generated at specific times in the cell cycle. Because repair of DSBs can differ during the cell cycle [92], this may affect the types of rearrangements that would occur. Lastly, the integration of plasmid DNA containing active promoters can affect chromatin conformation at the integration site, which could affect repair of DSBs, since chromatin conformation can influence DSB repair [14], [39], [41]. As a result, the use of integrated plasmids containing selectable marker genes cannot address differences in repair due to naturally occurring differences in chromatin conformation or transcription. Additional studies using DSBs generated by nonenzymatic means will therefore be required to confirm our results. One potential artifact that can be ruled out in our studies is that the difference in the type of events observed at interstitial and telomeric sites is due to a difference in the frequency of DSBs generated by I-SceI endonuclease. Although subtelomeric regions are typically composed of heterochromatin [59], [72], [73], which could inhibit I-SceI-induced DSB formation near telomeres, this difference in chromatin conformation has been minimized in our cell clones by prior selection for expression of the puro or GFP genes. In addition, although the lower frequency of NHEJ near telomeres could be explained by fewer I-SceI-induced DSBs near telomeres, the frequency of GCRs and large deletions is increased near telomeres, and the frequency of small deletions and HRR is similar at telomeric and interstitial DSBs [56]. Moreover, a fundamental difference in repair of DSBs near telomeres is shown by the large size of the deletions at telomeric DSBs [68], [71], and by the different effects that ATM deficiency has on DSB repair at interstitial and telomeric sites. Despite the caveats and limitations of assays based on I-SceI-induced DSBs, the deficiency in repair of I-SceI-induced DSBs near telomeres in the EJ-30 human tumor cell line appears to be representative of the response of mammalian cells to telomeric DSBs. Similar rearrangements were also observed in response to I-SceI-induced DSBs near telomeres in mouse ES cells [89]. Moreover, the types of rearrangements observed as a result of I-SceI-induced DSBs near telomeres are very similar to the types of rearrangements resulting from spontaneous telomere loss in the EJ-30 human tumor cell line [69], [70], [93]. The observation that a high rate of spontaneous telomere loss is common among human tumor cell lines [94] previously led us to propose that the sensitivity of telomeric regions to DSBs plays an important role in chromosome instability in human cancer [46]. Consistent with this proposal, telomere dysfunction resulting from oncogene-induced replication stress was subsequently found to cause senescence in normal cells, leading to the proposal that oncogene-induced telomere dysfunction can also serve as a protection against cancer in cells with intact cell cycle checkpoints [64]. Because most dysfunctional telomeres in the senescent cells in this study retained telomeric repeat sequences, the investigators concluded that oncogene-induced senescence results from persistent DSBs that occur near telomeres during replication stress. The deficiency in DSB repair near telomeres is also important in senescence resulting from ionizing radiation, as shown by the fact that persistent DSBs resulting from ionizing radiation co-localize with telomeres and correlate with radiation-induced cell senescence, both in cultured human primary fibroblasts and in vivo in mice [62], [63]. Although the precise location of the DSBs near telomeres was not determined in these studies, as we have previously pointed out [44], the frequency of the persistent DSBs suggests a target size that included subtelomeric DNA, consistent with our earlier studies demonstrating that the region that is sensitive to DSBs extends at least 100 kb from the telomere [71]. A similar sensitivity to DSBs near telomeres in human germ line cells would also explain the high degree of variability in subtelomeric regions in humans, which have been attributed to a high frequency of translocations [95]. In addition, the sensitivity of telomeric regions to DSBs could explain the prevalence of human genetic diseases resulting from terminal deletions and inversions at the ends of chromosomes that are associated with translocations [96]–[100]. Further studies in the mechanism of sensitivity of telomeric regions to DSBs should therefore provide valuable insights into the mechanisms of human disease. The pEJ5-GFP plasmid (Figure 1A) has previously been used to monitor the frequency of NHEJ at telomeric and interstitial DSBs [27], [56]. The pGFP-ISceI plasmid (Figure 1A) has previously been used to monitor the frequency of large deletions at telomeric and interstitial DSBs [56]. The pGFP-ISceI plasmid was generated from the pEJ5-GFP plasmid by deletion of the puromycin-resistance (puro) gene following NHEJ between the two I-SceI sites. The pDsRed-ISceI plasmid (Figure 1B) was created by inserting an 18 bp recognition site for I-SceI endonuclease between the BglII and EcoRI restriction sites at the 5′ end of the promoterless DsRed gene in the pDsRed-Express-1 plasmid (Clontech). All of the cell lines used in this study were derived from clone B3 of the EJ-30 human bladder cell carcinoma cell line. EJ-30 is a subclone of the EJ human colon cancer cell line, which is also called MGH-U1 [101]. The cells were grown in MEM alpha media (UCSF Cell Culture Facility) supplemented with 5% fetal calf serum (Invitrogen-Gibco), 5% newborn calf serum with iron (Invitrogen-Gibco), 1 mM l-glutamine (Invitrogen-Gibco), and were propagated at 37°C in humidified incubators. The GFP-7F1 and GFP-6D1 clones containing the pGFP-ISceI plasmid integrated at interstitial and telomeric sites, respectively, were previously used to investigate the frequency of large deletions [56]. The EJ5-7F and EJ5-6J clones containing the pEJ5-GFP plasmid integrated at interstitial and telomeric sites, respectively, were previously used to investigate the frequency of NHEJ [56]. Clones EJ5-7F and EJ5-6J were transfected with the pDsRed-ISceI plasmid linearized with ApaLI, and colonies containing the stably integrated pDsRed-ISceI were selected with G418. The number of integrated copies of the pDsRed-ISceI plasmid was then determined by Southern blot analysis using a variety of restriction enzymes (see below). We identified four EJ5-7F clones that contain a single copy of the pDsRed-ISceI plasmid (EDS-7F2, EDS-7F4, EDS-7F5, and EDS-7F6), six EJ5-6J clones that contain a single copy of the pDsRed-ISceI plasmid (EDS-6J8, EDS-6J29, EDS-6J31, EDS-6J35, EDS-6J41, EDS-6J49), and two EJ5-6J clones containing three copies of the plasmid (EDS-6J7, EDS-6J10). Packaging of the pQCXIH and pQCXIH-ISceI retroviral vectors and infection of cell cultures was performed as previously described [68]. The selection for cells infected with pQCXIH-ISceI was achieved by growth in medium containing 50 µg/ml hygromycin (Sigma) for 14 days with medium changes every 2 days to allow for expression of I-SceI endonuclease and the generation of DSBs. After 12 days, the cells were trypsinized and replated. After an additional 2 days, the cells were trypsinized again, pooled, and either analyzed for the frequency of GFP-positive (GFP+) and DsRed-positive (DsRed+) cells, or replated for preparation of genomic DNA. KU55933 is an effective inhibitor of ATM kinase activity [102]. Treatment of cells with 10 µM KU55933 began the day after infection with pQCXIH or pQCXIH-ISceI and continued during the 14-day period prior to cell analysis using our assays. The shRNAs for knockdown of gene expression were introduced into cells using the pSiren RetroQ-Blasticidin retrovirus vector (kindly provided by Denise Chan, UCSF). The pSiren RetroQ-Blasticidin retrovirus vector was generated from the pSiren RetroQ retrovirus vector (Clontech) by replacing the puro gene with the blasticidin-resistance gene. The packaging the retrovirus vector was performed as previously described [68]. The shRNA sequence for knockdown of ATM was 5′-GCAACATACTACTCAAAGA-3′, which has previously been shown to effectively knockdown expression of ATM [103]. The efficiency of knockdown of ATM gene expression was determined by quantitative real-time PCR (see below). The efficiency of knockdown was 88. 5% for clone EDS-7F2,70. 7% for clone EDS-6J8,80. 5% for clone GFP-7F1, and 75. 9% for clone GFP-6D1. RNA isolation for analysis of gene expression was performed using an RNeasy kit (Qiagen) following the manufacturer' s instructions. cDNA was generated from 1. 5 µg of total RNA, using M-MLV-RT (Invitrogen) following the manufacturer' s instructions. Quantitative real-time PCR was performed on cDNA samples using a StepOnePlus Real-Time PCR machine (Applied Biosystems). PCR was performed using 2. 5 µl of the cDNA sample, 0. 2 µl of 10 µM forward primer, 0. 2 µl of 10 µM reverse primer, and 5 µl of Power Sybr Green PCR Master Mix (Applied Biosystems) following the manufacturer' s instructions. A mixture of cDNA from cell clones EDS-6J8, EDS-7F2, GFP-6D1 and GFP-7F1 that was undiluted, diluted 4X, 16X, and 64X, was used as a standard. The level of expression of the housekeeping gene GAPDH was also analyzed in each sample to control for the efficiency of PCR in each sample. The knockdown efficiency of ATM was calculated by comparing the expression level of the ATM and GAPDH genes in cell cultures with and without the shRNA for ATM. The expression level of the ATM and GAPDH genes were calculated by absolute quantification relative to the standard curve using the Standard Curve Method with the SDS software provided by the manufacturer (Applied Biosystems). The primers used for analysis of ATM expression by quantitative real-time PCR were ATM-F, 5′-TCCAGGGGAAGATGATGAAGA-3′, and ATM-R, 5′-TCTACAATGAGCTGCGTGTGG-3′. The primers used for analysis of GAPDH expression used as an endogenous control were GAPDH-F, 5′-GTTGCCATCAATGACCCCTT-3′, and GAPDH-R, 5′-ACTCCACGACGTACTCAGCG-3′. The analysis of the frequency of GFP+ and DsRed positive (DsRed+) cells was performed using an Accuri C6 Flow Cytometer (BD Biosciences). The cells were trypsinized, an equal volume of growth medium was added, and the cells were counted and pelleted. To prevent aggregation, the cells were then resuspended in 10 ml of ice-cold Dulbecco' s PBS (w/o Ca or Mg) containing 100 µg/ml Proteinase K (Sigma) by vigorous pipeting with a fine bore plastic pipet. The cells were then incubated 10 min on ice, pipeting twice more during the incubation. This treatment with Proteinase K is necessary with EJ-30 to keep the cells from aggregating. Following the incubation 2 ml of Dulbecco' s PBS (w/o Ca or Mg) containing 1% BSA (Sigma) was added to block further digestion with Proteinase K. The cells were then pelleted and resuspended in Dulbecco' s PBS (w/o Ca or Mg) at approximately 1×106 cells/ml for analysis by flow cytometry. Approximately 1×106 cells were counted for each sample. All samples were analyzed in triplicate. Error bars represent standard deviation of experiments that were conducted three times. DNA sequence analysis of recombination junctions involved in activation of the DsRed gene was accomplished by first isolating pooled populations of DsRed+ cells by flow sorting (in conjunction with the UCSF Cell Analysis Core Facility) from clones EDS-6J7 and EDS-6J8 that were selected for 14 days with hygromycin following infection with the pQCXIH-ISceI retrovirus. The pooled populations of DsRed+ cells were then plated as single cells and allowed to grow into colonies, after which individual colonies were selected at random. The genomic DNA from the subclones generated from the various colonies was then isolated and amplified by PCR using the EJ5-1 primer, 5′-ATGGTAATCGTGCGAGAGGG-3′, located at the end of the promoter in the EJ5-GFP plasmid, and the DSR-1 primer, 5′-TGAAGCGCATGAACTCCTTG-3′, located at the 5′ end of the DsRed gene (see Figure 1B). The conditions for PCR involved 94°C for 2 min, then 40 cycles of 94°C for 30 sec, 62°C for 30 sec, and 72°C for 45 sec. DNA sequence analysis was then performed directly on the PCR products using the EJ5-1 primer (MCLAB). The frequency of GFP+ cells was determined using a Cellometer Vision (Nexelcom). The cells were first trypsinized and 20 µl of growth medium containing approximately 1×104 cells was aliquoted into a counting chamber slide (Nexelcom). Two counting chambers were used for each sample, with each chamber being counted two times. All samples were analyzed in triplicate. Error bars represent standard deviation of experiments that were conducted three times. The results with the Cellometer Vision were verified by visual analysis of the cells being counted, and were very similar to results obtained using flow cytometry (data not shown). The presence of small deletions at a single I-SceI-induced DSB were analyzed by first generating PCR products spanning an I-SceI site in the integrated pEJ5-GFP plasmid, and then digesting the PCR products with I-SceI endonuclease. PCR was performed on genomic DNA isolated from the pooled hygromycin-resistant cell cultures 14 days after infection with the pQCXIH-ISceI retroviral vector. PCR was performed using Taq 2X Master Mix (New England Biolabs) and primers GFP-1 (5′-GCGGGGTTCGGCTTCTGG-3′) and GFP-3 (5′-CGCTTCCATTGCTCAGCGG-3′) (see Figure 1A). PCR involved 94°C for 2 minutes, then 40 cycles of 94°C for 30 seconds, 62°C for 30 seconds, and 72°C for 30 seconds. 25 µl of the PCR product was then digested with 20 units of I-SceI endonuclease at 37°C overnight, and the products were run on 4% agarose gels. After staining with ethidium bromide, digital images were analyzed using Image J software (http: //download. cnet. com/ImageJ/3000-2192_4-37303. html? tag=vtredir) to calculate the intensity of the bands. The fraction of cells containing small deletions (SD) at the I-SceI site was determined by dividing the intensity of the uncut band (UC) by the combined intensity of the cut (C) and uncut bands. The values for small deletions were then corrected for the fraction of cells that had large deletions or NHEJ, because these cells would not produce a PCR product, and would therefore cause an overestimation of the fraction of cells containing small deletions. The fraction of cells with small deletions therefore involves multiplying the fraction of uncut PCR product by 1 minus the fraction of cells with large deletions (LD), as determined in our large deletion assay, and by 1 minus the fraction of cells with NHEJ, as determined by our NHEJ assay. The final equation for the fraction of cells with small deletions is therefore: SD = UC/UC+C× (1−LD) (1−NHEJ). Although inversions of the fragment between the two I-SceI sites would also prevent small deletions, this was not corrected for because the frequency of these events is too low to significantly affect our results [31]. The validity of this correction was previously demonstrated by the analysis of the frequency of small deletions in 100 individual subclones selected at random [68]. All samples were analyzed in triplicate. Error bars represent standard deviation of experiments that were conducted three times. Southern blot analysis was performed on genomic DNA isolated from the individual G418-resistant subclones transfected with the pDsRed-ISceI plasmid to determine the number of integrated plasmid copies. Southern blot analysis was performed as previously described [104] using the pDsRed-ISceI plasmid as a probe. Copy number was determined by digesting genomic DNA with three separate restriction enzymes, BamHI, BglII, and EcoRI, that cut once in the plasmid adjacent to the I-SceI site. Clones containing a single copy of pDsRed-ISceI will show two plasmid-specific bands with each of the three restriction enzyme digests.

The ends of chromosomes, called telomeres, prevent chromosome ends from appearing as DNA double-strand breaks (DSBs) and prevent chromosome fusion by forming a specialized nucleo-protein complex. The critical function of telomeres in end protection has a downside, in that it interferes with the repair of DSBs that occur near telomeres. DSBs are critical DNA lesions, because if they are not repaired correctly they can result in gross chromosome rearrangements (GCRs). As a result, the deficiency in DSB repair near telomeres has now been implicated in ageing, by promoting cell senescence, and cancer, by promoting telomere dysfunction due to oncogene-induced replication stress. The studies presented here demonstrate that DSBs near telomeres commonly result in GCRs in a human tumor cell line. Moreover, our results demonstrate that the mechanism of repair of telomeric DSBs is very different from the mechanism of repair of DSBs at other locations, supporting our hypothesis that the deficiency in repair of DSBs near telomeres is a result of the abnormal processing of DSBs due to the presence of telomeric proteins. Understanding the mechanism responsible for the deficiency in DSB repair near telomeres will provide important insights into critical human disease pathways.

Abstract Introduction Results Discussion Materials and Methods

medicine biology

2013

The Role of ATM in the Deficiency in Nonhomologous End-Joining near Telomeres in a Human Cancer Cell Line

Many meiotic systems in female animals include a lengthy arrest in G2 that separates the end of pachytene from nuclear envelope breakdown (NEB). However, the mechanisms by which a meiotic cell can arrest for long periods of time (decades in human females) have remained a mystery. The Drosophila Matrimony (Mtrm) protein is expressed from the end of pachytene until the completion of meiosis I. Loss-of-function mtrm mutants result in precocious NEB. Coimmunoprecipitation experiments reveal that Mtrm physically interacts with Polo kinase (Polo) in vivo, and multidimensional protein identification technology mass spectrometry analysis reveals that Mtrm binds to Polo with an approximate stoichiometry of 1: 1. Mutation of a Polo-Box Domain (PBD) binding site in Mtrm ablates the function of Mtrm and the physical interaction of Mtrm with Polo. The meiotic defects observed in mtrm/+ heterozygotes are fully suppressed by reducing the dose of polo+, demonstrating that Mtrm acts as an inhibitor of Polo. Mtrm acts as a negative regulator of Polo during the later stages of G2 arrest. Indeed, both the repression of Polo expression until stage 11 and the inactivation of newly synthesized Polo by Mtrm until stage 13 play critical roles in maintaining and properly terminating G2 arrest. Our data suggest a model in which the eventual activation of Cdc25 by an excess of Polo at stage 13 triggers NEB and entry into prometaphase. The mechanism of the lengthy arrest in G2 that separates the end of pachytene from nuclear envelope breakdown (NEB) —which is a characterization of many female meiotic systems—has remained a mystery. One can imagine that both the maintenance and the termination of this arrest might involve either or both of two mechanisms— the transcriptional or translational repression of a protein that induces NEB, and thus meiotic entry, or the presence of an inhibitory protein that precludes entry into the first meiotic division. Because Drosophila females exhibit a prolonged G2 arrest (see Figure 1) and are amenable to both genetic and cytological analyses, they provide an ideal system in which to study this problem. The ovaries of Drosophila females are composed of a bundle of ovarioles, each of which contains a number of oocytes arranged in order of their developmental stages [1–3]. For our purposes, the process of oogenesis may be said to consist of three separate sets of divisions: the initial stem cell divisions, which create primary cystoblasts; four incomplete cystoblast divisions, which create a 16-cell cyst that contains the oocyte; and the two meiotic divisions. Although a great deal is known regarding the mechanisms that control cystoblast divisions and oocyte differentiation, relatively little is known about the mechanisms by which the progression of meiosis is controlled. As is the case in many meiotic systems, female meiosis in Drosophila involves preprogrammed developmental pauses. The two most prominent pauses during Drosophila meiosis are an arrest that separates the end of pachytene at stages 5–6 from NEB at stage 13, and a second pause that begins with metaphase I arrest at stage 14 and continues until the egg passes through the oviduct. It is the release of this second preprogrammed arrest event that initiates anaphase I and allows the completion of meiosis I followed by meiosis II. As shown in Figure 1, the end of meiotic prophase by dissolution of the synaptonemal complex (SC) at stages 5–6 [4,5] is separated from the beginning of the meiotic divisions, which is defined by NEB at stage 13, by approximately 40 h to allow for oocyte growth. We are interested in elucidating the mechanisms that arrest meiotic progression at the end of prophase, but then allow onset of NEB and the initiation of meiotic spindle formation some 40 h later. One intriguing possibility is that during this period of meiotic arrest, the oocyte actively blocks the function of cell cycle regulatory proteins such as cyclin dependent kinase 1 (Cdk1), the phosphatase Cdc25, and Polo kinase (Polo), all of which promote meiotic progression just as they do during mitotic growth. Recently, Polo was shown to be expressed in the germarium and required for the proper entry of Drosophila oocytes into meiotic prophase, as defined by the assembly of the SC [6]. Decreased levels of Polo resulted in delayed entry into meiotic prophase, whereas overexpression of Polo caused a dramatic increase in the number of cystocyte cells entering meiotic prophase, indicating that Polo is involved both in the initiation of SC formation and in the restriction of meiosis to the oocyte. How then is Polo, which is known to play multiple roles in promoting meiotic and mitotic progression [7,8], prevented from compelling the differentiated oocyte to proceed further into meiosis? One component of this regulation may well lie in the fact that Polo is not expressed during much of oogenesis. As shown below, Polo is clearly visible in the germarium but is then absent until stage 11, when it begins to accumulate to high levels in the oocyte (Figure S1). We show here that a second component of Polo regulation is mediated by binding to the protein product of the matrimony (mtrm) gene, which occurs from stage 11 until the onset of NEB at stage 13. This binding serves to inhibit Polo in the early stages of its expression, and thus prevents precocious nuclear envelope breakdown. The mtrm gene was first identified in a deficiency screen for loci that were required in two doses for faithful meiotic chromosome segregation [9]. mtrm/+ heterozygotes display a significant defect in achiasmate segregation (the meiotic process that ensures the segregation of those homologs that, for various reasons, fail to undergo crossingover). As a result of this defect, mtrm/+ heterozygotes exhibit high levels of achiasmate nondisjunction. As homozygotes, mtrm mutants are fully viable but exhibit complete female sterility. We show here that the Mtrm protein prevents precocious NEB. Indeed, as discussed below, the effects of reducing the dose of mtrm on meiotic progression and on chromosome segregation are easily explained as the consequence of precocious NEB at stages 11 or 12, and can be suppressed by simultaneously reducing the copy number of polo+. In addition, the effects of heterozygosity for loss-of-function alleles of mtrm can be phenocopied by increasing the copy number of polo+. These genetic interactions suggest that Mtrm negatively regulates Polo in vivo. Interestingly, Mtrm was shown to interact physically with Polo by a global yeast two-hybrid study [10]. We demonstrate that this yeast two-hybrid finding reflects a true physical interaction in vivo by both coimmunoprecipitation studies and by multidimensional protein identification technology (MudPIT) mass spectrometry experiments, which indicate that Mtrm binds to Polo with an approximate stoichiometry of 1: 1. Moreover, ablating one of the two putative Polo binding sites on Mtrm by mutation prevents the physical interaction between Polo and Mtrm and renders the mutated Mtrm protein functionless. This experiment, along with genetic interaction studies, provides compelling evidence that the function of the binding of Mtrm to Polo is to inhibit Polo, and not vice versa. The analysis of mtrm mutants allows us to examine the effects of premature Polo function during oogenesis. Our evidence shows that in the absence of Mtrm, newly synthesized Polo is capable of inducing NEB from stage 11 onward. As a result of this precocious NEB, chromosomes are not properly compacted into a mature karyosome and they are released prematurely onto the meiotic spindle. In many cases, the centromeres of achiasmate bivalents subsequently fail to co-orient. The mtrm gene was first identified as a dosage-sensitive meiotic locus; heterozygosity for a loss-of-function allele of mtrm specifically induced the failed segregation of achiasmate homologs [9]. The mtrm gene encodes a 217–amino acid protein with two Polo-Box Domain (PBD) binding sites (STP and SSP) and a C-terminal SAM/Pointed domain. The studies reported in this paper rely primarily on a null allele of mtrm (mtrm126), which removes 80 bp of upstream sequence and the sequences encoding the first 41 amino acids of the Mtrm protein (Figure 2A). Western blot analysis using an antibody to Mtrm reveals that Mtrm can only be detected in ovaries (Figure 2B). This is consistent with a previous report by Arbeitman et al. [11], which showed that the expression profile of the mtrm gene product was strictly maternal and that its expression was reduced greater than 10-fold over 0–6. 5 h of embryonic development. The specificity of this antibody is demonstrated by the fact that no signal was detected by either Western blotting or by immunofluorescence of ovarioles homozygous for the mtrm126 mutant (Figure 2C). Immunofluorescence studies using the same antibody reveal that Mtrm is expressed as a diffuse nuclear protein in the oocytes and nurse cells beginning at stage 4–5 (Figure 2C and 2D). As shown in Figure 2C, the Mtrm signal was not restricted to the karyosome itself; but rather Mtrm seems to fill the space in the entire nucleus. Although Mtrm is restricted to the nucleus until approximately stage 10, it localizes throughout the oocyte in later stages. Mtrm brightly stains both the oocyte nucleus and cytoplasm between stage 11 and stage 12, but staining is greatly reduced at stage 13, the stage at which NEB occurs (Figure S1). mtrm/+ heterozygotes display a substantial defect in the processes that ensure the segregation of achiasmate homologs. We show here that these meiotic defects are strongly suppressed by simultaneous heterozygosity for strong loss-of-function alleles of polo. (Our impetus for searching for a genetic interaction between mtrm and polo came from the finding that the mutants in the mei-S332 gene were partially suppressed by polo mutants [12].) We measure meiotic mis-segregation by assaying X and 4th chromosomal nondisjunction in females of the genotype FM7/X where FM7 is a balancer chromosome that fully suppresses X chromosomal exchange. (The 4th chromosome is obligately achiasmate.) As shown in Figure 3B, FM7/X; mtrm/+ females typically show frequencies of X and 4th chromosome nondisjunction in the range of 35%–45%, which is more than 100-fold above control values. However, FM7/X; mtrm126/+ females that were simultaneously heterozygous for either a deficiency (Df (3L) rdgC-co2) that uncovers polo or for either of two strong alleles of polo, poloKG03033 and polo16–1 (Figure 3A) displayed greatly reduced levels of meiotic nondisjunction (Figure 3B). The fact that the poloKG03033 mutation is due to a P element insertion allowed us to demonstrate that the observed interaction with mtrm was indeed a direct consequence of a reduction in polo activity. Two precise excisions of this insertion were generated, and neither was able to suppress the nondisjunctional effects observed in mtrm/+ heterozygotes (unpublished data). We also demonstrated that the poloKG03033 allele was able to suppress the meiotic defects generated by heterozygosity for mtrmexc13, an independently isolated allele of mtrm (unpublished data). Heterozygosity for these same loss-of-function alleles of polo has no detectable effect on meiotic chromosome segregation in mtrm+/mtrm+ females. In females of the genotypes FM7/X; poloKG03033/+ or FM7/X; polo16–1/+, the observed levels of nondisjunction for the X chromosome were 0. 2% and 0. 4%, respectively. Similarly, the observed levels of nondisjunction for the 4th chromosome were 0. 6% and 0. 5% respectively (n = 1,109 for FM7/X; poloKG03033 /+ and n = 1,226 for FM7/X; polo16–1/+ females). These data alone are consistent with either a hypothesis in which Mtrm acts to inhibit Polo, excess Polo creates a meiotic defect or a scenario in which Polo inhibits Mtrm, and the absence of sufficient Mtrm creates the defect. However, as we will show below, our additional data support the model whereby Mtrm inhibits Polo. If reducing the quantity of Polo suppresses the meiotic defects observed in mtrm/+ females, then over-expression of Polo alone should mimic the effects of reducing the dosage of mtrm+ (i. e. , we should see a chromosome segregation defect solely in the presence of increased dosage of polo+, even in mtrm+/mtrm+ oocytes). To test this hypothesis, we analyzed FM7/X females carrying two doses of a UASP-polo+ transgene construct driven by the nanos-GAL4 driver. As shown in Figure 3C, expression of the UASP-polo+ transgene construct results in a dosage-dependent increase in the frequency of achiasmate nondisjunction for both the X and the 4th chromosomes. Similar observations were made using chromosomal duplications that carry two copies of polo+ (Adelaide Carpenter, personal communication). Moreover, increasing the dose of Polo in females heterozygous for mtrm126 resulted in severe meiotic defects. Females carrying a single copy of the UASP-polo+ transgene and which were also heterozygous for mtrm126 were virtually sterile (unpublished data). Thus, increasing the dosage of Polo enhances the defect observed in mtrm/+ heterozygotes by inducing sterility. The genetic interaction between Mtrm and Polo during oogenesis is paralleled by their patterns of expression. Mtrm reaches its maximum level of expression from the end of stage 10 onward, filling the oocyte during stages 11–12, and then diminishes at stage 13. Analysis of Polo expression using an antibody to Polo [13,14] and wild-type oocytes revealed that Polo is present in the oocyte at low levels (except in the germarium) until stages 11 or 12 and then rapidly fills the oocyte cytoplasm from stages 12–13 onward (Figure S1). Taken together, these data support a model in which the presence of Mtrm inhibits Polo in the early stages of expression, while permitting the function of Polo at stage 13, when Mtrm is degraded. Data directly demonstrating that assertion are provided below. A large scale yeast two-hybrid screen identified Mtrm as a candidate interactor with Polo [10] and showed that Mtrm carries two putative PBD binding sites: STP and SSP (Figure 4A). To confirm that Mtrm interacts with Polo physically in vivo, we performed coimmunoprecipitation experiments on wild-type ovary extracts using a polyclonal antibody to Mtrm. As shown in lane 1 of Figure 4B, the antibody to Mtrm also precipitated Polo. We used two separate approaches to confirm the interaction between Polo and Mtrm. In the first experiment, we used ovary extracts from females expressing a Green Fluorescent Protein (GFP) -polo transgene [13] and performed the coimmunoprecipitation using an antibody to GFP. In the second experiment, we used ovary extracts from wild-type females and performed the coimmunoprecipitation using a monoclonal antibody to Polo [14]. In both experiments, we were able to show that Mtrm coimmunoprecipitated with Polo (Figure S2). In addition, MudPIT mass spectrometry reveals that Mtrm and Polo interact in oocytes with a stoichiometry of approximately 1: 1. We analyzed three independent affinity purifications from ovarian extracts expressing a C-terminally 3× FLAG-tagged Mtrm, and we used MudPIT mass spectrometry [15] to identify interacting proteins. We then compared the identified proteins to those detected in five control FLAG immunoprecipitations from control (w1118) flies. Among the proteins that showed reproducible and significant p-values (p < 0. 001) identified in all three analyses, Polo was detected by multiple peptides and stands out as the only protein recovered at levels similar to those of Mtrm, as estimated by normalized spectral abundance factor (NSAF) counts [16,17]. Although the NSAF values for Mtrm and Polo vary across the three biological replicates analyzed (Figure 4C), the ratio between the two proteins remains constant with an average of 0. 96 ± 0. 11, suggesting one Mtrm molecule binds to one molecule of Polo. Thus, three lines of evidence demonstrate that Mtrm physically interacts with Polo: the yeast two-hybrid work [10], our coimmunoprecipitation studies, and our MudPIT mass spectrometry experiments presented in this section. The observation of strong genetic interactions between mutants in these two genes (Figure 3) demonstrates a functional significance to this interaction. Polo interacts with target proteins via the interaction of its PBD and the sequences STP or SSP on the target protein. In both of these PBD-binding sites, the center residues (threonine or serine) are phosphorylated to facilitate Polo binding [18–20]. Mtrm carries two potential PBD-binding sites: STP with the central threonine at residue 40 and SSP with the central serine at residue 124 (Figure 4A). To determine whether the interaction between Mtrm and Polo is mediated through the interaction of the Polo PBD with either or both of these two potential PBD-binding sites, we created UASP-driven transgenes that carried mutations in either or both of the STP or SSP motifs. In each case, we mutated the central residue of the PBD-binding sites on Mtrm to the nonphosphorylateable residue alanine. These mutants are denoted as mtrmT (40) A, which disrupts the STP motif, and mtrmS (124) A, which disrupts the SSP motif. Each of these mutant constructs was expressed under the control of the nanos-GAL4 driver in a mtrm null background to insure that they were the only source of Mtrm protein in the oocytes. Coimmunoprecipitation experiments using antibodies to Mtrm revealed that MtrmS (124) A protein still interacted with Polo (Figure 4B). However, MtrmT (40) A failed to bind to Polo (Figure 4B), indicating that the STP residues define a motif that is critical for the Mtrm–Polo interaction. Mutation of both PBD sites also resulted in a version of Mtrm that did not interact with Polo (unpublished data). Because the interaction of Polo with target proteins via its PBD requires the phosphorylation of the center residues (threonine or serine) of the STP or SSP motifs [18–20], we searched the MS/MS dataset for phosphorylated peptides derived from Mtrm or Polo. For each of the detected sites, we estimated the levels of modification by dividing the number of spectra matching a particular phosphopeptide by the total spectral count for this peptide (Figure 4D). We were able to detect phosphorylation on both T40 and S124; although, in agreement with the second PBD not being the primary binding site, S124 phosphorylation was found less reproducibly. In addition, Mtrm S48, S52, and S137 were found phosphorylated at reproducibly high levels in two out of three experiments. We also observed that Polo T182 was detected as phosphorylated at high levels (over 80%) in all three immunoprecipitations, indicating that those Polo proteins that are bound to Mtrm were fully activated [21]. Not only is the STP motif important for Polo binding, but it is also required for proper Mtrm function (Figure 4E). We assayed the frequency of nondisjunction in females expressing either the mtrmS (124) A or the mtrmT (40) A construct in the germlines of FM7/X; mtrm/+ heterozygotes. Although the mtrmS (124) A construct was able to rescue the meiotic defects seen in mtrm/+ heterozygotes, the mtrmT (40) A construct failed to rescue the mtrm defect and maintained the high nondisjunction frequency seen in FM7/X; mtrm/+ heterozygotes. A similar failure to rescue was observed using a double mutant construct that carried both the mtrmS (124) A and the mtrmT (40) A mutations (unpublished data). Based on these observations, we conclude that the STP site is critical for Mtrm function and the T (40) A mutation ablates Mtrm function as a direct consequence of a failure to interact with Polo. In the previous sections, we presented three separate lines of evidence that Mtrm acts to inhibit Polo function, and not vice versa. First, effects of heterozygosity for mtrm can be suppressed by a corresponding reduction in the dose of polo+. Second, we observed that the phenotype created by reducing the dose of mtrm+ can be mimicked by increasing the dose of Polo. Third, and most importantly, the observation that mutating the STP Polo binding site by a conservative amino acid replacement (STP → SAP) ablates Mtrm function argues strongly that Mtrm functions as an inhibitor of Polo. Were it the case that Polo inhibits Mtrm, one would expect loss of the Polo interacting site to produce a hyperfunctional Mtrm, not a nonfunctional protein. The early stages of meiosis appear normal in both mtrm/+ and mtrm/mtrm oocytes. The germarium and early stages appear morphologically normal and, at least in mtrm/+ oocytes, both recombination and SC assembly are indistinguishable from normal ([9] and unpublished data). However, following stage 11 (the period during which Mtrm is maximally expressed), we observed multiple defects in oocyte maturation in both mtrm/+ and mtrm/mtrm oocytes. Most critically, we show that a loss-of-function allele of mtrm induces precocious NEB in a dosage-dependent manner. In wild-type oocytes, NEB usually does not occur until stage 13; only a single case of NEB at stage 12 was observed among the 61 stage 11 and 12 wild-type oocytes examined (Figure 5). However, in mtrm126/+ heterozygotes, more than a third of stage 12 egg chambers exhibited NEB. To ensure that the precocious NEB defect is the consequence of reducing the copy number of mtrm+, we repeated these experiments using females that are heterozygous for an independently isolated allele of mtrm; mtrmexc13. These females also displayed precocious NEB at stage 12 (data not shown). As is the case for the chromosome segregation defects observed in mtrm/+ oocytes, the precocious NEB that is seen in mtrm126/+ heterozygotes is strongly suppressed by simultaneous heterozygosity for a loss-of-function allele of polo (Figure 5B), suggesting that the timing of NEB is determined by the relative abundances of Mtrm and Polo. This conclusion is further strengthened by the observation that overexpression of Polo (using a UASP-polo+ transgene driven by the nanos-GAL4 driver) increases the frequency of precocious NEB in mtrm126/+ heterozygotes by nearly 2-fold (from 42% to 77%). The extent of the precocious NEB defect is even more evident in mtrm126 homozygotes. As shown in Figure 5, NEB had already occurred in 32 out of 33 stage 12 oocytes examined and in six of ten stage 11 oocytes examined. Thus, the loss of Mtrm causes precocious NEB in a dosage-dependent fashion. Taken together, these data argue that the presence of Mtrm prevents Polo from inducing NEB until stage 13 (see Discussion), and that a reduction or absence of available Mtrm allows the Polo that is synthesized during stages 11 and 12 to initiate NEB. The precocious breakdown of the nuclear envelope at stages 11 to 12 is important because the karyosome undergoes dramatic changes in structure during this period [2]. As noted above, in stages 9–10, the karyosome expands to the point that individual chromosomes can be detected [22–24]. These chromosomes recondense into a compact karyosome during stages 11–12, the exact time at which a reduction in the level of Mtrm causes precocious NEB. Thus, the early NEB events promoted by heterozygosity for mtrm might be expected to result in the release of incompletely condensed or disordered karyosomes. To test this hypothesis, we examined karyosome morphology during the 20 min that preceded NEB in wild-type, mtrm126/mtrm+, and mtrm126 polo+/mtrm+ polo16–1 oocytes. As shown in Figure 6, only two out of 28 (7%) wild-type oocytes with incompletely compacted or disordered karyosomes were observed. However, 7 out of 27 (26%) mtrm126/mtrm+ oocytes displayed a disordered karyosome, an effect that was largely suppressed (to 8%) by simultaneous heterozygosity for polo16–1 (Figure 6). These data support the view that the precocious NEB that is induced by lowering the level of Mtrm results in the release of improperly formed karyosomes into the cytoplasm and are again consistent with the possibility that Mtrm inhibits meiotic progression through its effects on Polo. The karyosome plays a critical role in directing the formation of the acentriolar spindle in Drosophila oocytes. In 8 out of 9 (89%) wild-type oocytes, the karyosome remains associated even after NEB; it is then surrounded by microtubules and forms a bipolar meiotic spindle (Figure 7 and Video S1). At metaphase I, chiasmate chromosomes are still condensed into a single mass at the metaphase plate in a tapered bipolar spindle [25–28]. However, in FM7/X; mtrm126/mtrm+ oocytes, the karyosome usually dissolved within 10–20 min after NEB, and the individual bivalents became clearly visible (Figure 7 and Video S2). In 15 out of 17 (88%) FM7/X; mtrm126/ mtrm+ oocytes examined, the chromosomes were individualized during spindle assembly. Indeed, in 14 of these examinations, all three pairs of major chromosomes were physically separated at some point during the time course of imaging (in the remaining case, the three bivalents could be distinguished but were still physically associated). As discussed in the legend to Figure 7, despite this dissociation into individual bivalents, in most oocytes the chromosomes are capable of reaggregating into a single mass and eventually forming a bipolar spindle. A striking example where all four chromosome pairs can be clearly distinguished is the image taken 26 min after NEB for FM7/X; mtrm126/mtrm+ oocytes (Figure 7). In those oocytes in which bivalent individualization was observed, the two major autosomes appeared to be held together by at least two chiasmata (one on each arm), suggesting that sister-chromatid cohesion along the euchromatic arms of these chromosomes still persists. The two X chromosomes remain physically associated, despite the lack of chiasmata, presumably as a consequence of the maintenance of heterochromatic pairing [29,30]. Because the nondisjunction of achiasmate chromosomes observed in mtrm126/mtrm+ heterozygotes was suppressed by heterozygosity for loss-of-function alleles of polo, we next tested whether a polo mutation could also suppress this karyosome maintenance defect. As shown in Figure 7 and Video S3, bivalent individualization was only observed in three out of 13 (23%) of FM7/X; mtrm126 polo+/mtrm+ polo16–1 oocytes, and thus 77% of the oocytes maintained the karyosome as a single mass throughout the process of spindle assembly. These data are consistent with the genetic data presented above: reducing the dose of polo+ strongly suppresses the deleterious effects of heterozygosity for mtrm. Because the karyosomes of mtrm/+ females were poorly formed before NEB and are usually transiently dissolved to individual bivalents shortly after NEB (see above), we also examined centromere co-orientation on bipolar prometaphase spindles using FISH probes (see Materials and Methods) directed against the X and 4th chromosomes (Figure 8) in both wild-type and mtrm/+ oocytes. In wild-type oocytes, the vast majority of most X and 4th chromosome centromeres co-oriented properly (Figure 8). The frequencies of abnormal centromere co-orientation in oocytes with chiasmate X chromosomes (XX) were only 2% for the X chromosome and 4% for the 4th chromosome. In FM7/X females, where X chromosomal crossingover is blocked, the frequencies of abnormal co-orientation were still quite low (4% for the X chromosome and 2% for the 4th). However, co-orientation of achiasmate centromeres was often aberrant in mtrm/+ heterozygotes, such that the centromeres of both homologs were often oriented toward the same pole (Figure 8A). In these cases, the two homologs also occupied different arcs of the meiotic spindle, a feature that is rarely, if ever, observed in wild-type oocytes. In chiasmate X females, 43% of observed oocyte nuclei displayed an aberrant co-orientation of 4th chromosome centromeres, and 6% of these oocytes displayed aberrant X centromere co-orientations (Figure 8B); these oocytes likely reflect the 8%–10% of oocytes that fail to undergo crossingover even in females bearing structurally normal X chromosomes. The defect in 4th chromosome centromere co-orientation was fully suppressed by simultaneous heterozygosity for polo16–1 (Figure 8A and 8B). As expected, due to the suppression of X chromosomal crossingover in FM7/X females, mtrm/+ heterozygotes displayed frequent abnormal centromere co-orientation for both X and 4th chromosomes, i. e. , 43% for X chromosomes and 37% for 4th chromosomes (Figure 8B). These results indicate that the mtrm heterozygotes display an obvious defect in centromere co-orientation. However, once again, both the defect in X and 4th chromosome centromere co-orientation was fully suppressed by simultaneous heterozygosity for polo16–1. Thus, as was the case with the previously considered defects, the deleterious effects of reducing the amount of available Mtrm can be suppressed by a simultaneous reduction in the amount of Polo. The trigger for the G2/M transition is activation of Cdk1 by Cdc25 (reviewed in [31]), and multiple lines of evidence suggest that Polo can activate Cdc25 [32]. First, in Caenorhabditis elegans, RNAi experiments demonstrate that ablation of Polo prevents NEB [33]. Second, the Xenopus Polo homolog Plx1 is activated in vivo during oocyte maturation with the same kinetics as Cdc25. Additionally, microinjection of Plx1 accelerates the activation of both Cdc25 and cyclinB-Cdk1 [34]. Moreover, microinjection of either an antibody to Plx1 or kinase-dead mutant of Plx1 inhibited the activation of Cdc25 and its target cyclinB-Cdk1. A later study by Qian et al. demonstrated that injection of a constitutively active form of Plx1 accelerated Cdc25 activation [35]. As pointed out by these authors, these studies support “the concept that Plx1 is the ‘trigger' kinase for the activation of Cdc25 during the G2/M transition. ” Finally, a small molecule inhibitor of Polo kinase (BI 2536) also results in extension of prophase [36]. These data are consistent with the view that the presence of functional (unbound) Polo plays a critical role in ending the extended G2 that is characteristic of oogenesis in most animals. We should note by Riparbelli et al. [37] that the careful study of female meiosis in polo1 homozygotes failed to observe a defect in the timing of NEB. However, as discussed in the legend to Figure 3, polo1, a missense mutant that is viable even over some deficiencies and does not suppress mtrm, is the weakest of the known polo mutants, and it is thus reasonable that no defect was observed. In light of these data, it is tempting to suggest that in wild-type Drosophila oocytes, the large quantity of Mtrm deposited into the oocyte from stage 10 onward inhibits the Polo that is either newly synthesized or transported into the oocyte during stages 11–12. However, at stage 13, an excess of functional Polo is created when the number of Polo proteins exceeds the available amount of inhibitory Mtrm proteins. This unencumbered and thus functional Polo then serves to activate Cdc25, initiating the chain of events that leads to NEB and the initiation of prometaphase. In the absence of a sufficient amount of Mtrm, an excess of Polo causes the precocious activation of Cdc25, and thus an early G2/M transition. A model describing this hypothesis is presented in Figure 9. Based on this model, one can visualize that decreasing the dose of Mtrm or increasing the dose of Polo will hasten NEB, whereas simultaneous reduction in the dose of both proteins should allow for proper timing of NEB. Two lines of evidence directly support a model in which Mtrm exerts its effect on Polo, with respect to preventing precocious NEB, by blocking the ability of Polo to activate Cdc25. First, as shown in Figure 10, mutants in the Drosophila cdc25 homolog twine fail to undergo NEB in stage 13. In addition, heterozygosity for twine also decreases the frequency of precocious NEB in mtrm126/+ heterozygotes from 42% (Figure 5) to less than 10% (7/72). Mtrm' s first PBD binding site (T40) is required for its interaction with Polo. Mtrm T40 has to be first phosphorylated by a priming kinase, such as one of the Cdks or MAPKs, and was indeed detected as phosphorylated in the mass spectrometry dataset. The NetPhosK algorithm [38] predicts T40 to be a Cdk5 site, and the serines immediately distal to T40—S48 and S52—which were also detected as phosphorylated (Figure 4D), are sites for proline-directed kinases such as Cdk or MAPK sites as well. The other prominent phosphorylation event occurs at S137, which could be a Polo phosphorylation site since it falls within a Polo consensus (D/E-X-S/T-Ø-X-D/E). Although the combined sequence coverage for Mtrm was 44%, indicating that some phosphorylated sites might have been missed, Mtrm S137 is a suitable binding site for activated Polo, in agreement with the processive phosphorylation model [18]. At this point of our studies, Mtrm T40 priming kinase or the kinase responsible for Polo activating phosphorylation on T182 has not been identified. The finding that Polo not only is able to bind to Mtrm in vivo in a 1: 1 ratio, but also is fully phosphorylated on T182 in its activation loop [21] suggests a method by which Mtrm serves to inhibit Polo. In general, enzymes are usually not recovered from affinity purifications at levels similar to their targets. They do not form stable complexes, but rather form transient interactions with their substrates, which is how efficient catalysis is achieved. Here, Mtrm is able to sequester activated Polo away in a stable binary complex over a long period of time. It is only when this equilibrium is disturbed at the onset of stage 13 by the production of an excess of Polo (as suggested in Figure 9) or by degradation of Mtrm that Polo can be released. The molecular determinants of the Mtrm: : Polo sequestration event are not clear, but it would be interesting to test whether the serines found phosphorylated in the vicinity of Mtrm PBD binding sites play a role in locking the binary complex into place. Our data demonstrate that a reduction in the levels of Mtrm results in the release of an incompletely compacted karyosome that rapidly dissolves into individual bivalents during the early stages of spindle formation. For chiasmate bivalents, this is apparently not a problem, because they still co-orient correctly (for example, the chiasmate X chromosomes shown in Figure 8 still achieve proper co-orientation in the vast majority of oocytes). However, the nonexchange bivalents frequently fail to co-orient properly, such that both homologs are oriented toward the same pole (but often occupy two different arcs of the spindle). This initial failure of proper co-orientation leads to high frequencies of nondisjunction as demonstrated by the genetic studies and analysis of metaphase I images presented in [9]. Although achiasmate homologs are properly co-oriented in wild-type oocytes [29,30], we have noted previously such homologs can often vacillate between the poles such that two achiasmate homologs are often found on the same arc of the same half-spindle during mid to late prometaphase ([25] and unpublished data). These chromosomes are often observed to be physically associated. This situation is quite different from the defect observed in mtrm/+ heterozygotes, where the homologs are neither physically associated nor on the same arc of the spindle. It is tempting to suggest that the chromosome segregation defects we observe in mtrm/+ heterozygotes are simply the result of precocious release of an incompletely re-compacted karyosome. According to this explanation, the defects observed in meiotic chromosome segregation are solely the consequence of premature NEB. (Implicit in this model is the assumption that it is the events that occur during karyosome re-compaction, at stages 11 and 12, that serve to initially bi-orient achiasmate chromosomes, and we do not have direct evidence to support such a hypothesis.) Alternatively, Polo plays multiple roles in the meiotic process [7,8], and it is possible that the chromosome segregation defects we see represent effects of excess Polo that are un-related to the precocious breakdown of the nuclear envelope. Such a view is supported by two observations. First, as shown in Figure 7, the bivalent individualization observed after NEB in mtrm/+ oocytes does not disrupt FM7-X pairings. Second, although heterozygosity for twine in mtrm126/+ heterozygotes suppresses the frequency of precocious NEB from 42% (Figure 5) to less than 10% (7/72), two alleles of twine tested (twe1 and twek08310) failed to suppress the levels of meiotic nondisjunction observed in FM7/X; mtrm126/+ heterozygotes. These data suggest that the effects of excess Polo on nondisjunction may not be regulated via Cdc25/Twine, but rather by the effects of excess Polo on some other as-yet-unidentified Polo target. This suggests that the effects of Mtrm on the level of Polo might affect multiple Polo-related processes. Support for such an idea that Mtrm can inhibit Polo-regulated proteins that are unrelated to NEB comes from the observation that the ectopic expression of Drosophila Mtrm in Schizosaccharomyces pombe blocks karyokinesis, producing long multi-septate cells with only one or two large nuclei ([39] and Bruce Edgar, personal communication). This phenotype is similar, if not identical to, that exhibited by mutants in the S. pombe Polo homolog plo1 (Plo1), which fail in later stages of mitosis due to the role of Plo1 in activating the septation initiation network to trigger cytokinesis and cell division. However, Plo1 also plays a role in bipolar spindle assembly that might also be inhibited in the Mtrm expressing cells, but this function of Plo1 is less well understood. Thus the possibility exists that the effect of mtrm mutants on meiotic chromosome segregation may well not be the direct consequence of early NEB, but rather may be due to the role of Polo in other meiotic activities, such as spindle formation or the combined effects of these defects with precocious NEB. Efforts to identify such processes and their components are underway in the lab. Finally, we should note that while Mtrm is the first known protein that is able to inactivate Polo by physical interaction with Polo itself; there is certainly additional mechanisms of Polo regulation. For example, Archambault et al. [40] have described mutants in the gene that encodes Greatwall/Scant kinase, which have both late meiotic and mitotic defects. Although there is no evidence for a physical interaction between these two kinases, the authors speculate that the function of the Greatwall kinase serves to antagonize that of Polo. The Scant mutations create a hyperactive form of Greatwall, which might be expected to lower the dosage Polo, and thus perhaps partially suppress the defects observed in mtrm/+ heterozygotes. Indeed, exactly such a suppressive effect has been observed in Scant homozygotes (however, this suppression is much weaker than that obtained by heterozygosity for loss of function alleles of polo). The data presented above demonstrate that Mtrm acts as a negative regulator of Polo during the later stages of G2 arrest during meiosis. Indeed, both the repression of Polo expression until stage 11 and the inactivation of newly synthesized Polo by Mtrm until stage 13 play critical roles in maintaining and properly terminating G2 arrest. Our data suggest a model in which the eventual activation of Cdc25 by an excess of Polo at stage 13 triggers NEB and entry into prometaphase. Although our data do shed some light on the mechanism by which Mtrm inhibits Polo, it is not entirely clear whether Polo' s ability to phosphorylate targets other than Cdc25 might be blocked by Mtrm: : Polo binding. These issues will need to be addressed in the future studies. Finally, we note that although small molecule inhibitors of Polo have been identified [36], Mtrm represents the first case of a protein inhibitor of Polo. It would be most exciting to identify functional orthologs of Mtrm outside of the genus Drosophila. Perhaps that might best be accomplished through a screen for oocyte-specific Polo-interacting proteins. Throughout this study, a w1118 stock served as our normal sequence X wild-type control, and for achiasmate X chromosome studies, FM7/yw was used as wild-type control. The GFP-polo stock was kindly provided by Adelaide Carpenter. The nanos-GAL4 driver was used to express UASP-driven transgenes (see below) in the ovary. All polo mutants, the P element insertion mutant, and deficiencies related to mtrm were acquired from the Bloomington Drosophila Stock Center. A P-element insertion mutant, KG08051, causing a mutation in the mtrm gene and exhibiting high levels of nondisjunction for achiasmate chromosomes [9], was obtained from the Bloomington Drosophila Stock Center. Although Harris et al. [9] positioned the insertion site for this transposon 90 bp upstream of the first ATG in the mtrm coding sequence, resequencing indicates that the true insertion site is in fact 80 bp upstream of the first ATG in the mtrm coding sequence. mtrm126 was generated by imprecise excision from the insertion of a null allele of mtrm. It is a deletion that removes 80 bp of 5′-UTR and 123 bp of coding sequence, deleting the first 41 amino acids (Figure 2A). Reverse-transcriptase (RT) -PCR, and Western blotting confirmed that mtrm126 homozygotes had no transcripts and no protein expression (unpublished data). Like the original P element insertion mutant, mtrm126 showed a dosage-sensitive effect on meiotic nondisjunction that was specific to achiasmate chromosomes and homozygous sterile females (homozygous males are fully fertile and meiotic segregation is normal in both mtrm heterozygotes and homozygotes). To construct the UASP-polo+ transgene, we amplified a 1. 74-kb XhoI-XbaI polo fragment from reverse transcribed cDNA by PCR using the primers 5′-CTCGAGGATGGCCGCGAAGCCCGAGGATAAG-3′ and 5′-TCTAGATTATGTGAACATCTTCTCCAGCATTTTCC-3′. The polo fragment was cloned into the pBluescript to generate pBlue-polo-cDNA. Then, a polo fragment was obtained by digestion with KpnI and XbaI from pBlue-polo-cDNA and cloned into the pUASP vector [41] to produce pUASP-polo+. The UASP-polo+ cassette in this plasmid was sequenced for confirmation. The transformation of the pUASp-polo+ and other plasmids (see below), to generate transgenic flies, was conducted by Genetic Services in Boston, Massachusetts, United States. To place the 3× FLAG downstream of mtrm, the PCR amplified 687-bp mtrm+1. 5×-FLAG fragment was created using primer pKpnI-mtrm-5,5′-GGGGTACCAAATGGAGAATTCTCGCACGCCCACGAACAAG-3′, and primer mtrm-3-FLAG (1. 5×), 5′-GTCCTTGTAGTCCTTGTCATCGTCGTCCTTGTAGTCAAGAGTGTGGAGCACATCCATGATACGG-3′. Then the 687-bp mtrm+1. 5× FLAG was amplified with the FLAG (3×) stop-Xbal primer, 5′-GCTCTAGATTACTTGTCATCGTCGTCCTTGTAGTCCTTGTCATCGTCGTCCTTGTAGTCCTTGTCATCGTCGTCCTTG- 3′, to produce the KpnI-XbaI mtrm-Flag (3×) fragment. The fragment was then cloned into the pUASP vector [41] to produce pUASP-mtrm-flag (3x). The Mtrm protein has two potential PBD binding sites: STP with the central threonine at residue 40 and SSP with the central serine at residue 124 (Figure 4A). To mutate the central residues to alanine in each motif, PCR assembly was used to make two separate codon changes in the mtrm gene, one at +118 from ACT to GCT to produce mtrmT (40) A and the other at +370 from CAG to CGC to produce mtrmS (124) A. To mutate the STP motif, primer pmtrm-mut-ATG: 5′-CGGGGTACCAAAAGATGGAGAATTCTCGCACGCCCACGAACAAGAC-3′ and primer pmtrm-STPre: 5′-GAGATTGGGCGAACGGAAGTTGCCAAAGATCGGAGCAGAGCATCGCACGTTGGAGGTGTTCACCTTCAG-3′ were used to amplify a 150-bp fragment for 5′-terminus of mtrm. The rest of mtrm was amplified with primers pmtrm-STP: 5′-CTGAAGGTGAACACCTCCAACGTGCGATGCTCTGCTCCGATCTTTGGCAACTTCCGTTCGCCCAATCTC-3′, and pmtrm-mut-TAA: 5′-GCTCTAGATTAAAGAGTG TGGAGCACATCCATGATACGCTTGC-3′ to produce a 520-bp fragment. The 150-bp and 520-bp fragments were combined in equal amounts and amplified by PCR to assemble the full-length KpnI-XbaI mtrmT (40) A gene introducing a point mutation. The KpnI-XbaI mtrmT (40) A was cloned in to pUASP to generate pUASP- mtrmT (40) A. After confirmation by sequencing, the plasmid was used for genetic transformation. To construct the mtrmS (124) A transgene, primer pmtrm-mut-ATG and primer pmtrm-SSPre: 5′-GGTCTCCATATTCGAGTCATCCGAACAGGTATCCGGGGCGCTGCAGCTCT-3′ were used to amplify a 420-bp fragment of the 5′ terminus of mtrm. The 3′ terminus of mtrm was amplified by using primer pmtrm-SSP: 5′-AGAGCTGCAGCGCCCCGGATACCTGTTCGGATGACTCGAATATGGAGACC-3′and primer pmtrm-mut-TAA to produce a 300-bp fragment. The two fragments in equal molar amounts were amplified by PCR to assemble a full-length KpnI-XbaI mtrmS (124) A gene with a point mutation introduced. The KpnI-XbaI mtrmS (124) A was cloned in pUASP to generate pUASP- mtrmS (124) A. The plasmid was used for genetic transformation after confirmation by sequencing. The mtrm gene was cloned into a pET-21a vector (Norvagen). 6×His-tagged Mtrm was expressed in the bacterial strain BL21 (DE3), isolated and purified using the Probed Purification System (Invitrogen), and used to raise rabbit and guinea pig polyclonal antisera by Cocalico Biologicals in Reamstown, Pennsylvania, United States. Affinity purification of the antiserum against Mtrm was performed by using a Sulfolink kit from the Pierce Company. Mouse monoclonal antibody to Polo was kindly provided by Moutinho-Santos [13]. Antibody to GFP from rabbits was purchased from Abcam. To prepare ovaries to fix for immunostaining, female fly preparation and ovary dissection were conducted as described in Xiang and Hawley [30]. Whole ovaries were collected and kept in 0. 75 ml of 1× Robb' s solution during the dissection. After egg chambers were manually teased apart, the ovaries were transferred to an Eppendorf tube. Then, 0. 25 ml of 16% formaldehyde was added and incubated for 15 min. The ovaries were washed three times in PBS + 0. 1% Triton X-100 (PBST) for 10 min each. After washing three times in PBST, they were incubated in PBST with 5% goat serum for at least 2 h at 4 °C with gentle shaking before being incubated overnight with primary antibodies. Egg chambers were washed four times in PBST and then incubated with proper fluorescently labeled secondary antibodies for 4 h at room temperature. Egg chambers were stained for 10 min in PBST with 0. 5 μg/ml DAPI and re-washed four times in the solution for a total of 40 min. The egg chambers were mounted on slides in Vectashield for analysis. Microscopy observation was conducted using a DeltaVision microscopy system (Applied Precision) as described in Xiang and Hawley [30]. To prepare the ovary extract for immunoprecipitation, ovaries from 100 yeast-fed female flies were dissected in 1× PBS. The ovaries were homogenized in an Eppendorf tube at 4 °C by a small pestle in 0. 5 ml of ovary extract buffer containing 25 mM Hepes (pH6. 8), 50 mM KCl, 1 mM MgCl2,1 mM DTT, and 125 mM sucrose with protease inhibitors cocktail (Calbiochem). The extract was centrifuged at 14,000g for 15 min at 4 °C, and the supernatant was collected. Protein A agarose beads were used for binding polyclonal antibodies from rabbit and guinea pig. Protein G agarose beads were used for binding monoclonal antibody from mouse. 50 μl of protein A- or G-coated agarose was washed three times with PBST (PBS + 0. 1% Triton X-100). 10 μl of antibody was added to the beads in a final volume of 500 μl of PBS and mixed on a shaker for 1 h at 4 °C. The beads then were washed twice with PBST. The ovary extract was immunoprecipitated with the beads for 1 h at 4 °C with continual shaking. After recovery by centrifugation at 1,000g for 3 min, the beads were washed four times with the cold ovary extract buffer with protease inhibitors, for 5 min each. For Western blotting, the beads were suspended in 30 μl of SDS loading buffer (50 mM Tris-HCl (pH6. 8), 100mM DTT, 2% SDS, 0. 1% bromophenol blue, 10% glycerol) and heated for 3 min at 95 °C before being loaded on a PAGE gel. Western blotting for Mtrm (Figure 2B) was conducted by using antibody to Mtrm from guinea pigs and an Alkaline Phosphatase chromogen kit (BCIP/NBT) (Roche). Fluorescent Western blotting techniques were used to display both Mtrm and Polo from coimmunoprecipitates on the same membrane. In order to prepare a C-terminally 3× FLAG-tagged Mtrm for the MudPIT mass spectrometry assay, the UASP-mtrm-Flag (3×) construct was expressed in ovaries under the control of the nanos-GAL4 driver in a wild-type background. The extraction of protein from the ovaries was the same as described above. 100 μl of anti-FLAG beads were washed two times with prechilled 1× PBS and then two times with prechilled ovary extract buffer. The anti-FLAG beads were mixed with the extract supernatant, incubated, and washed as described above. After washing, the beads bound with Mtrm-FLAG (3×) were finally transferred to a minicolumn and washed with 25 ml of TBS (50 mM Tris-HCl, pH7. 4,150 mM NaCl) at 4 °C. When washing was completed, 300 μl TBS with 100 μg/ml 3× FLAG peptide was added to elute proteins. TCA was added to the eluted protein solution at a final concentration of 20%. The solution was mixed and kept on ice for at least 30 min. The solution was centrifuged at 14,000 rpm at 4 °C for 15 min. The pellet was collected and 300 μl of prechilled acetone was gently added. After centrifuging again at 14,000 rpm at 4 °C for 15 min, the pellet was carefully collected. The pellet was air dried and ready for the MudPIT spectrometry analysis. TCA-precipitated proteins were urea-denatured, reduced, alkylated, and digested with endoproteinase Lys-C (Roche) followed by modified trypsin (Promega) as described in Washburn [15]. Peptide mixtures were loaded onto 100-μm fused silica microcapillary columns packed with 5-μm C18 reverse phase (Aqua, Phenomenex), strong cation exchange particles (Partisphere SCX, Whatman), and reverse phase [42]. Loaded microcapillary columns were placed in line with a Quaternary 1100 series HPLC pump (±Agilent) and a LTQ linear ion trap mass spectrometer equipped with a nano-LC electrospray ionization source (ThermoFinnigan). Fully automated 10-step MudPIT runs were carried out on the electrosprayed peptides, as described in [43]. Tandem mass (MS/MS) spectra were interpreted using SEQUEST [44] against a database consisting of 17,348 D. melanogaster proteins (nonredundant entries downloaded from the National Center for Biotechnology Information [NCBI] 28 November 2006 release), and 177 usual contaminants (such as human keratins, IgGs, and proteolytic enzymes). To estimate false discovery rates (FDR), each nonredundant protein entry was randomized, keeping the same amino acid composition and length, doubling the search space to a total of 35,050 amino acid sequences (17,525 forward + 17,525 shuffled sequences). Peptide/spectrum matches were selected and compared using DTASelect/CONTRAST [45] with the following criteria set: spectra/peptide matches were only retained if they had a DeltCn of at least 0. 08, and a minimum XCorr of 1. 8 for singly-, 2. 0 for doubly-, and 3. 0 for triply-charged spectra. In addition, peptides had to be fully tryptic and at least seven amino acids long. Combining all runs, proteins had to be detected by at least two such peptides or one peptide with two independent spectra. Under these criteria, the average FDR was 0. 34 ± 0. To estimate relative protein levels, NSAFs were calculated for each nonredundant protein, as described by Paoletti [16] and Zybailov [17]. Log-transformed NSAF values for proteins reproducibly detected in all three analyses were subjected to a two-tailed t-test to highlight proteins significantly enriched in the Mtrm purifications as opposed to negative controls as in Zybailov [17]. A differential modification search was set up to query a protein database containing only the sequences for Mtrm and Polo for peptides containing phosphorylated serines, threonines, tyrosines, and oxidized methionines, i. e. , SEQUEST “ASFP” (all spectra against few proteins). The maximum number of modified amino acids per differential modification in a peptide was limited to four. After this search, an in-house developed script—sqt-merge [46]—was used to combine the sets of SEQUEST output files (sqt files) generated from the normal “ASAP” search (all spectra all proteins; i. e. , without modifications) and the phosphorylation “ASFP” search described above into one set. This merging step allowed only the best matches to be ranked first. The peptide matches contained in the merged sqt files were compiled and sorted using DTASelect [45]. For the third round of searches, spectra matching modified peptides were selected if they passed the conservative filtering criteria: minimum XCorr of 1. 8 for +1,2. 0 for +2, and 3. 0 for +3 spectra, with a maximum Sp rank of ten, and fully tryptic peptides with a minimum length of seven amino acids. Xcorr scores for isopeptides, in which any of several adjacent residues could be modified, tend to close, resulting in low normalized differences in Xcorrs. The DeltaCn cut-off was hence set at 0. 01 to allow such peptides to be further examined (“–m 0 –t 0 – Smn 7 –y 2 –s 10 −2 2 −3 3 –d 0. 01” DTASelect parameters). The coordinates for these spectra were written out into smaller ms2 files using the “– copy” utility of DTASelect. Because these subsetted ms2 files contained, at best, a few hundred MS/MS spectra, they can be subjected to the same phosphorylation differential search against the complete Drosophila database (SEQUEST “MSAP, ” modified spectra against all proteins). This step allowed us to check that spectra matching modified peptides from Polo and Mtrm sequences did not find a better match against the larger protein database. Again, sqt-merge was used to bring together the results generated by these different searches. DTASelect was used to create reports listing all detected proteins and modified residues on Polo and Mtrm. All spectra matching modified peptides were visually assessed and given an evaluation flag (Y/M/N, for yes/maybe/no). The “no” matches were removed from the final data (-v 2 parameter in DTASelect). Results from different immunoprecipitations were compared using CONTRAST. NSAF5 (an in-house software by Tim Wen) was used to create the final report on all detected proteins across the different runs, calculate their respective NSAF values, and estimate false discovery rates (FDR). U_SPC6 software (also in-house by Tim Wen) was used to extract total and modified spectral counts for each amino acid within the proteins of interest and calculate modification levels based on local spectral counts. To investigate the timing of NEB, 3-d-old females were collected and fed on yeast for two days. Ovaries were dissected in halocarbon oil 700 (Sigma) on a slide, and egg chambers were separated by mixing using a metal rod. Then, a coverslip was gently put on without pressing and mounting. After waiting for 20–30 min, the egg chambers were observed by phase contrast microscopy in dark view. To facilitate live imaging of the karyosome before and during NEB, stage 11–12 oocytes from well-fed females were dissected in halocarbon oil and then co-injected with Oli-Green Dye (Molecular Probes) to visualize DNA and Rhodamine-conjugated tubulin (Cytoskeleton) to visualize the spindle and to determine timing of the NEB. Oocytes with germinal vesicles were imaged using a LSM 510 META microscope (Zeiss). Images were acquired using the AIM software v 4 by taking a 10-series Z-stack at 1-μm intervals. The 1. 686 satellite sequences (also known as the 359-bp repeats) on the X chromosome and AATAT repeats on the 4th chromosome were chosen as probes for in situ hybridization [29,30,47]. The 359-bp sequence of the 1. 686 satellite sequences and (AATAT) 6 repeats were used for probe preparation. Alexa Fluor 488 dye was used for probes of 359-bp sequence on the X chromosome. For probes (AATAT) 6 on the 4th chromosome, Alexa Fluor 647 dye was used. The details of probe generation and labeling, egg chamber dissection and fixation, fluorescent in situ hybridization, and microscopy observation were described previously [30]. In all oocytes examined for centromere co-orientation, 4th chromosomes were observed as red masses of hybridization, whereas the X chromosomes were observed as single bright green masses of hybridization. The FM7 balancer chromosome displays two green blocks of hybridization because of multiple inversions [30]. The AATAT probe is slightly hybridized with an X and FM7 balancer around the centromere region, and therefore both X and FM7 have a slight red signal at centromere location. The FlyBase (http: //flybase. bio. indiana. edu) accession numbers for genes and gene products discussed in this paper are: matrimony (mtrm) gene (FBgn0010431); polo (FBgn0003124); twine (FBgn0002673); and Greatwall/Scant kinase (FBgn0004461). The Yeast Resource Center (http: //www. yeastrc. org/pdr/pages/front. jsp) accession number for the S. pombe Polo homolog plo1 is CAB11167.

Many meiotic systems in females animals include a lengthy arrest period (spanning days in flies and to decades in humans) that separates the early and late stages of meiosis. Such an arrest raises the question: how can the quiescent meiotic cell cycle be precisely awakened or re-started? At least in principle, the answer to this phenomenon, which we refer to as “The Sleeping Beauty Kiss, ” might have two molecular solutions: the controlled expression of a protein that re-starts the cell cycle, or the inactivation of an inhibitory protein that prevents such a re-start. We show here that the re-start of the meiotic cycle in Drosophila depends on both mechanisms: the controlled expression of an “activator” known as Polo kinase, and the presence of a regulatory protein called Matrimony (Mtrm), which binds to and physically inactivates Polo. Indeed, Mtrm is the first known protein inhibitor of Polo kinase. The excess of Mtrm prior to the time of normal meiotic re-start, keeps Polo inactive. However, either the production of an excess quantity of Polo, or the destruction of Mtrm, at the appropriate time, releases active Polo, permitting a properly controlled re-start of meiotic progression.

Abstract Introduction Results Discussion Materials and Methods Supporting Information

cell biology genetics and genomics

2007

The Inhibition of Polo Kinase by Matrimony Maintains G2 Arrest in the Meiotic Cell Cycle

Insect cuticle is composed primarily of chitin and structural proteins. To study the function of structural cuticular proteins, we focused on the proteins present in elytra (modified forewings that become highly sclerotized and pigmented covers for the hindwings) of the red flour beetle, Tribolium castaneum. We identified two highly abundant proteins, TcCPR27 (10 kDa) and TcCPR18 (20 kDa), which are also present in pronotum and ventral abdominal cuticles. Both are members of the Rebers and Riddiford family of cuticular proteins and contain RR2 motifs. Transcripts for both genes dramatically increase in abundance at the pharate adult stage and then decline quickly thereafter. Injection of specific double-stranded RNAs for each gene into penultimate or last instar larvae had no effect on larval–larval, larval–pupal, or pupal–adult molting. The elytra of the resulting adults, however, were shorter, wrinkled, warped, fenestrated, and less rigid than those from control insects. TcCPR27-deficient insects could not fold their hindwings properly and died prematurely approximately one week after eclosion, probably because of dehydration. TcCPR18-deficient insects exhibited a similar but less dramatic phenotype. Immunolocalization studies confirmed the presence of TcCPR27 in the elytral cuticle. These results demonstrate that TcCPR27 and TcCPR18 are major structural proteins in the rigid elytral, dorsal thoracic, and ventral abdominal cuticles of the red flour beetle, and that both proteins are required for morphogenesis of the beetle' s elytra. How arthropods manufacture exoskeletons with a wide array of mechanical properties, ranging from hard and rigid to soft and flexible, is an important question in developmental biology. The insect exoskeleton, or cuticle, covers the entire body wall and attached appendages as well as the foregut, hindgut and tracheae. It is a complex extracellular biocomposite, secreted by the epidermis and consisting of several functional layers including the waterproofing envelope, the protein-rich epicuticle and the chitin-rich procuticle [1]. Cuticular proteins (CPs) and the polysaccharide chitin are the primary structural components of the exo- and endocuticular layers that comprise the procuticle. During cuticle maturation and tanning (sclerotization and pigmentation), some of the CPs are post-translationally modified and cross-linked by quinones or quinone methides produced by the laccase-mediated oxidation of N-acylcatechols [2], [3]. This process stabilizes and hardens the exoskeleton, protecting insects from microbial, physical and other environmental stresses. However, little is known about the functional importance of individual insect cuticular proteins in the morphogenesis and mechanical properties of the exoskeleton. More than 100 genes encoding CP-like proteins have been identified in the fruit fly, Drosophila melanogaster [4], with a similar number present in the red flour beetle, Tribolium castaneum [5]. Anopheles gambiae (malaria mosquito) and Bombyx mori (oriental silkworm) have an even larger number of genes encoding CP-like proteins, each species harboring more than 200 putative CP genes [6]–[10]. Expression of specific CPs may be required to produce cuticles with a range of morphological and mechanical properties in different regions of the insect body and at different developmental stages. Insect CPs are classified into several distinct families defined by the presence of specific sequence motifs [7], [10]. The largest of these is the CPR family, which includes proteins that have a conserved amino acid sequence known as the Rebers & Riddiford (R&R) motif [11]. The R&R motif contains a putative chitin-binding domain that may help to coordinate the interactions between chitin fibers and the proteinaceous matrix [12], [13]. A major event in the evolution and diversification of beetles was the transformation of the membranous forewings into thickened, hardened, non-flight covers (elytra) for protection of the delicate hindwings and dorsal abdomen [14]–[16]. The elytron is composed of ventral and dorsal layers of epidermal cells that secrete thin lower and thicker upper cuticular laminae [17], [18]. In the developing elytron, the space between these two layers is filled with hemolymph and supporting structures known as trabeculae that function as mechanical struts, connecting and fortifying the dorsal and ventral cuticular layers. As the elytron matures, the epidermal layers are reduced in size, possibly dying or fusing together, and the hemolymph is resorbed, leaving a cavernous interior. The dorsal layer of the elytron becomes highly tanned and rigid as a result of both pigmentation and sclerotization. The ventral layer also exhibits some pigmentation, but it remains thin and membranous in comparison to the dorsal layer. The surface of the ventral elytral cuticle is relatively smooth and makes close contact with the underlying and folded hindwings. The surface of the dorsal elytral cuticle, on the other hand, contains numerous sensory setae and rib-like structures (striae) that extend longitudinally, apparently adding rigidity to the structure [19]. Initially, the elytra are short, colorless and soft, but they expand in both length and width shortly after eclosion, and subsequently harden and darken. A similar cuticle tanning process occurs in most of the adult body wall. In this study we have identified two highly abundant proteins that are present in rigid cuticle of the elytron, pronotum and ventral abdomen but not in the flexible cuticle of the dorsal abdomen and hindwing of T. castaneum adults, characterized their genes and expression profiles, and analyzed their roles in adult cuticle formation and stabilization. We have also determined that these two CPR proteins are essential structural components in the sclerotized dorsal cuticle of the elytron and are required for normal morphological, functional and mechanical properties. Extracts of untanned elytra dissected from newly emerged adults contained two highly abundant proteins with apparent sizes of 10 and 20 kDa based on their electrophoretic mobilities (Figure 1). To characterize these major proteins further, each was digested with trypsin, and the resulting peptides were analyzed by MALDI-TOF/TOF mass spectrometry. Comparison of these results with conceptual trypsinization of the computed proteome of T. castaneum revealed two candidate genes, denoted as TcCPR27 (XP_971678) and TcCPR18 (XP_967633), which are members of the Rebers and Riddiford family of cuticular proteins (Figure 1, Table S1 and cuticle DB: http: //biophysics. biol. uoa. gr/cuticleDB). Peptide coverages for TcCPR27 and TcCPR18 were 68. 2 and 88. 7%, respectively (Figure S1). We cloned cDNAs corresponding to these cuticular protein genes (accession numbers HQ634478 and HQ634479). The cDNA sequence of TcCPR27 was identical to that of the NCBI RefSeq gene prediction, whereas the RefSeq prediction for TcCPR18 had one in-frame deletion of three consecutive nucleotides and a single nucleotide mismatch compared to the cDNA, resulting in a deletion of one amino acid (one of the twelve consecutive glycines at amino acid positions 65–76 in the RefSeq prediction) and a phenylalanine-to-leucine substitution at amino acid position 85 in the RefSeq prediction. TcCPR27 and TcCPR18 encode proteins containing putative secretion signal peptide sequences, with theoretical molecular masses for the mature proteins of 11. 4 and 16. 4 kDa, respectively. Each mature protein contains a single RR-2 cuticular protein motif. Nearly all RR-2 proteins have a consensus region as follows: G-X (8) -G-X (6) -Y-X (6) -GF [7]. Both of these Tribolium proteins, however, have a slightly different RR2 motif. TcCPR27 contains G-X (8) -G-X (6) -Y-X (5) -GA, whereas TcCPR18 has G-X (8) -H-X (7) -Y-X (6) -GF. The former has a GA rather than the almost universal GF or GY at the end of the consensus, and the conserved Y and the third G are interrupted by five amino acids instead of the typical six. In the case of TcCPR18, the second conserved G is replaced by an H with no G residue located nearby. TcCPR18 is rich in glycine (21. 6%), whereas TcCPR27 has a high content of both glycine (16. 5%) and histidine (15. 5%). TcCPR18 is an apparent ortholog of the ecdysteroid-regulated, adult-specific cuticle protein acp22 of Tenebrio molitor (yellow mealworm), with 67% sequence identity (Figure S2) [20]. Both TcCPR27 and TcCPR18 map to linkage group 3 of the T. castaneum genome, but they are not tightly linked (BeetleBase: http: //beetlebase. org) [21]. Elytra of three other Tribolium species, T. brevicornis, T. confusum and T. freemani, also contain predominant cuticular proteins with high amino acid sequence similarities to TcCPR27 and TcCPR18 (Figure S3 and Table S2). To investigate whether TcCPR27 and TcCPR18 are present in other regions of the adult cuticle, we extracted proteins from cuticular samples dissected from the pronotum and the ventral abdomen just after adult eclosion. As in the case of the elytra, TcCPR27 and TcCPR18 proteins were also the predominant protein constituents of the pronotum, although their yields were low relative to those obtained from the elytra (Figure 2). We hypothesized that because the extent of tanning of the pronotum just after eclosion is substantially greater than that of the elytron, which tans at a later time (Figure 2), pronotum cuticular proteins were already cross-linked and much less extractable at the time of adult eclosion. To delay pronotum cuticle tanning, dsRNA for laccase-2 (dsTcLac2) was injected into 0–1 d-old pupae [2], and the pronotum cuticular proteins were subsequently extracted from samples obtained soon after adult eclosion. The yields of TcCPR27 and TcCPR18 were much higher in those extracts, indicating that the two proteins had not undergone substantial cross-linking in the absence of laccase and therefore were more readily extractable (Figure 2). Like the elytron, the adult ventral abdominal cuticle undergoes tanning and becomes hardened 3–5 d post eclosion, whereas the dorsal abdominal cuticle in the adult remains relatively untanned, flexible and transparent like the hindwing. TcCPR27 and TcCPR18 were abundant in extracts recovered from ventral abdominal cuticle of newly emerged adults, but very little or no TcCPR27 or TcCPR18 was present in extracts of the dorsal abdominal cuticle (Figure S4A). Similarly, the levels of TcCPR27 and TcCPR18 in the flexible hindwing were very low or undetectable (Figure S4A). These results show that TcCPR27 and TcCPR18 are major proteins in cuticles that become highly sclerotized and rigid, but they are absent or only very minor components of cuticles that are more flexible and membranous. Few or no transcripts for TcCPR27 or TcCPR18 were detected during the egg, larval or early pupal stages of development. However, the transcript levels of these genes dramatically increased at the pharate adult stage 0–1 d before eclosion, declining soon thereafter (Figure 3A, 3B). Transcript levels of TcCPR27 and TcCPR18 in the elytron were approximately 1,700- and 55-fold higher, respectively, than those in the membranous hindwing (Figure S4B). Both genes were also expressed in the pronotum and ventral abdomen, whose cuticles become highly sclerotized and hardened in mature adults. Transcript levels of TcCPR27 and TcCPR18 in the ventral abdomen were approximately 3,000 and 770 times higher, respectively, than the levels in the transparent, flexible and membranous dorsal abdomen (Figure S4B). The high histidine content of TcCPR27 and TcCPR18 (15. 5% and 10. 1%, respectively) allowed us to purify these proteins from elytra dissected from newly molted adults by utilizing nickel-affinity chromatography (Figure 4A). A polyclonal antibody directed against purified TcCPR27 was then generated. The CPR27 antibody specifically detected CPR27 but not CPR18 (Figure 4A). In pharate adults, TcCRP27 was co-localized with chitin in the dorsal elytral cuticle as well as in the ventral abdominal cuticle, both of which become more rigid and darker as the adult matures (Figure 4B, panels 1,3). Little or no TcCRP27 immunoreactivity was detected in the pupal, hindwing, or ventral elytral cuticles. RNA interference (RNAi) was used to investigate the functions of TcCPR27 and TcCPR18. As a negative control we injected dsRNA for T. castaneum tryptophan oxygenase (the vermilion gene, abbreviated Ver), a gene required for normal eye pigmentation [22]. Following dsRNA injections into last instar larvae, mRNA and protein levels of TcCPR27 and TcCPR18 were analyzed by real-time PCR and SDS-PAGE. Injection of these dsRNAs led to substantial and specific down-regulation of each gene at the mRNA (Figure 5A) and protein (Figure 5B) levels. TcCPR27 immunostaining also was strongly reduced after injection of dsTcCPR27 (Figure 4B, panel 2). Chitin staining in TcCPR27-deficient insects, however, was detected at approximately the same level as in dsVer-treated control animals (Figure 4B, panel 4). These results were further supported by staining of elytral chitin with a fluorescent chitin-labeling reagent, FITC-CBD. There was no difference in chitin staining among elytra collected from dsTcCPR27-, dsTcCPR18- and dsVer-treated insects (Figure S5). Injection of dsTcCPR27 or dsTcCPR18 into larvae had no apparent effect on larval-pupal or pupal-adult molting or on the morphology of the pupal cuticle, as expected from the observed late pupa-specific expression of these genes. However, the elytra of the resulting adults were malformed (Figure 6 and Figure S6). The surface of the elytra of dsTcCPR18-treated adults was irregular and rough compared to those of control insects. Adults derived from dsTcCPR27-injected larvae exhibited even more severe morphological defects. Their elytra were very short, wrinkled, warped and fenestrated (Figure 6 and Figure S6) and their hindwings were unable to fold normally. Such insects died approximately one week after eclosion, apparently from dehydration that resulted from failure of the misshapen elytra to properly cover the membranous dorsal abdomen and to thereby seal it against trans-cuticular water loss. The shape of a normal elytron is “inverted boat-like” to fit snugly on top of the hindwings and abdomen in order to protect the latter structures (Figure S6). In contrast, elytra from dsTcCPR27-treated insects were flatter and/or warped and did not cover the entire abdomen. Manual excision of the distal half of the elytron from a mature adult also led to high mortality, whereas removing an entire hindwing did not cause significant mortality, as long as the elytra could adopt their normal juxtaposition (Figure S7), consistent with our observation that properly formed elytra are essential to prevent desiccation of the adult, in addition to other potentially protective functions. These results support the hypothesis that TcCPR27 and TcCPR18 are major structural proteins in rigid elytral cuticle, and are required for normal elytral morphogenesis, hindwing folding and body hydration. We also analyzed the effects of depletion of TcCPR27 and TcCPR18 on the mechanical properties of elytra. Dynamic mechanical experiments were carried out to determine the storage modulus E′ and the loss modulus E″ of elytra as a function of oscillation frequency and strain. E′ is a measure of the elastically recoverable deformation energy, whereas E″ is a measure of viscous energy dissipation (dampening) and hence is also known as the viscous modulus. The ratio E″/E′ is known as the “loss tangent” or simply tan δ, where δ is the phase angle between sinusoidally applied stress and strain. For materials such as the elytra where E′>>E″, E′ is approximately equal to the Young' s modulus obtained from the slope of simple stress-strain measurements at the same strain rate [23], [24]. Hence, E′ is a measure of the stiffness of the elytra. Elytra from animals injected with dsTcCPR27 were significantly less rigid (lower E′) than the dsVer-treated controls (Figure 7). Elytra from dsTcCPR18-injected beetles had intermediate values of strength, consistent with the less severe visible phenotype observed with TcCPR18 knockdown (Figure 6). In addition, the dsTcCPR27 elytra had reduced values for tan δ, an indication that they experienced a higher degree of cross-linking than the control. Lower E″ or tan δ in polymeric networks is typically associated with a reduction in the network of uncross-linked material, dangling chains (chains linked to the network at only one end), loops and other network imperfections [25]. It is a well-established principle in the synthesis of gels or networks by cross-linking polymers, that increasing the ratio of cross-linker molecules to polymer molecules will typically increase the overall cross-link density of a network and reduce the fraction that is not cross-linked [26]. Thus, for elytral cuticle, the deficiency of a major structural cuticular protein such as TcCPR27, while maintaining a constant concentration of quinone cross-linking molecules, would be expected to lead to a greater average number of cross-links per protein molecule. Thus, for elytral cuticle, reducing the expression level of a major structural cuticular protein such as TcCPR27 or TcCPR18, while maintaining a constant concentration of quinone cross-linking molecules, would be expected to lead to a greater average number of cross-links per protein molecule. The more severe phenotype for knockdown of TcCPR27, relative to TcCPR18 might be due to differences in the degree of knockdown of their expression in the RNAi experiments, or perhaps could be due to differences in their structural properties or cross-linking chemistry. A greater reduction in protein levels for TcCPR27 than TcCPR18 in the knockdown animals might have led to relatively more cross-linking, thus reducing tan δ, and a greater reduction in TcCPR27 protein expression would lead to a lower storage modulus. Similar observations were previously reported for elytra from insects subjected to Lac2 knockdown (reduced tan δ combined with reduced E′) [19]. The basic genetic patterning mechanism for dorsal appendages such as wings has been described for both D. melanogaster and T. castaneum [27]–[29]. Similar gene networks are used to pattern wings, elytra and halteres, despite the profound morphological and functional divergence of these appendages during insect evolution. Individual structural proteins are likely to substantially affect the physical properties of elytra. However, to date, there have been no detailed reports about the contributions of individual structural proteins to elytral morphogenesis. Like other beetle species, T. castaneum adults possess elytra, modified forewings with a highly sclerotized and pigmented dorsal cuticle. Immediately after eclosion, untanned elytra have a soft white cuticle. Elytra expand shortly thereafter and then become rigid and darker during cuticle maturation. The role of structural proteins in this developmental process is the focus of this study. We identified two proteins from the RR2 family of cuticular proteins, TcCPR27 and TcCPR18, which are highly abundant in protein extracts of elytra dissected from newly emerged adults. TcCPR27 and TcCPR18 transcripts were strongly up-regulated at the developmental stage when adult cuticular proteins are expected to be synthesized (pupal day 4), and were nearly absent at earlier immature stages. Highly abundant cuticular proteins related to TcCPR27 and TcCPR18 were also present in protein extracts of elytra dissected from three other Tribolium species including T. brevicornis, T. confusum and T. freemani (Figure S3 and Table S2). TcCPR27 and TcCPR18 were also identified in extracts of pronotum cuticle, in which tanning had been initiated before adult eclosion. The protein yields, however, were much lower than those obtained from elytral extracts unless cuticle tanning was suppressed by injection of dsRNA for the tanning enzyme TcLac2 (Figure 2). These results support the hypothesis that TcCPR27 and TcCPR18 are cross-linked by highly reactive quinones and quinone methides that are produced by the cuticle tanning phenoloxidase laccase-2, and that these proteins become inextractable after tanning has occurred. Previously, Missios et al. [30] extracted two major cuticular proteins of 10 and 20 kDa, consistent with the apparent masses of TcCPR27 and TcCPR18, respectively, from extracts of cuticle from whole bodies of newly eclosed T. castaneum adults. Neither of these proteins was extractable from 7 day-old adults, consistent with an interpretation that these proteins become cross-linked during maturation of the cuticle. To study the functions of TcCPR27 and TcCPR18, we performed RNAi and successfully down-regulated levels of TcCPR27 and TcCPR18 mRNAs and proteins (Figure 5). These deficiencies caused several elytral defects. Although TcCPR27 and TcCPR18 are also present in other body regions such as the cuticles of the pronotum and ventral abdomen, which are heavily tanned in the mature adult, we did not observe visible phenotypic changes in those cuticles after injection of dsTcCPR27 and dsTcCPR18. The size and shape of these body regions do not change much after adult eclosion, in contrast to the elytra that are greatly expanded shortly after eclosion. The elytra of both TcCPR27- and TcCPR18-deficient insects failed to fully expand, and their dorsal surfaces were not smooth (Figure S6). The elytra of TcCPR27-deficient insects, particularly, were very short, wrinkled and fragile. Elytra from dsTcCPR27- and dsTcCPR18- treated insects appear to contain more cross-linked proteins than the elytra from dsVer-treated control insects (Figure 6). Lacking these major cuticular proteins apparently increases the effective concentration of cross-linking agents (NADA and NBAD quinones), resulting in aberrant cross-linking among the remaining proteins and shortened warped elytra. The effect was seen most clearly in dsTcCPR27-treated insects, in which the modulus and tan δ were significantly reduced relative to control insects. The dsTcCPR18- treated insects had a smaller decrease in modulus and tan δ, perhaps because of a smaller degree of protein reduction. The difference could also arise from structural differences between TcCPR27 and TcCPR18, which could have different propensities for forming intermolecular vs. intramolecular cross-links. However, the present data cannot draw that level of distinction. All of these results suggest that TcCPR27 and TcCPR18 are critical for normal elytral morphogenesis and are required to prevent dehydration and death of the adult. In summary, we have identified in beetles two major structural proteins, TcCPR27 and TcCPR18, which account for approximately half of the extractable cuticular proteins in the elytra and also are major components of other hard cuticular structures. It is interesting to note that the proteins utilized for hard cuticles of other body regions of the beetle were apparently used to build the elytron' s hard cuticle. In some saturniid moth species, proteins from the same CPR family are also used to form rigid structures such as tubercles, head capsules and hard pupal cuticle [31]. We now have biomechanical evidence on just how important these kinds of proteins are. TcCPR27 and TcCPR18 are required not only for rigid cuticle development, but also for morphogenesis, elytral mechanical properties, and survival of the red flour beetle. In contrast, these proteins are essentially undetectable in soft cuticles. Expression of such cuticular proteins in the modified forewings appears to be a fundamental evolutionary step in transforming the flexible and thin membranous wing into a thickened and rigid elytron in the Coleoptera. In the case of TcCPR18, an orthologous gene is found in the only other beetle species examined, the lesser grain borer, Rhyzopertha dominca, in the family Bostrichidae (Schlipalius, D. and Beeman, R. W. , unpublished observations) but not in any of the other sequenced arthropod genomes, including representatives of the Diptera, Hymenoptera and Lepidoptera. These structural proteins are probably cross-linked during sclerotization, via formation of histidyl-catechol adducts [32], [33]. Rigidification of the beetle forewing has likely been achieved in part through both structural protein incorporation and multiple co-options of the sclerotization pathway acting downstream of conserved wing gene network components, with the final product being primarily a rigid interpenetrating network of chitin embedded in a cross-linked protein matrix [3], [29], [34], [35]. To gain a more comprehensive understanding of the roles of cuticular proteins in defining the morphology and properties of the beetle elytron and rigid body wall cuticle, future studies are required to determine, at the ultrastructural level, the precise localization of TcCPR27, TcCPR18 and other structural proteins, and to assess the nature and extent of their covalent cross-linking during sclerotization. The GA-1 strain of T. castaneum was used in this study. Beetles were reared at 30°C under standard conditions [36]. Elytra of newly emerged adults (n = 10) were homogenized in 100 µl of cold PBS containing protease inhibitor cocktail (Roche). The homogenate was centrifuged for 2 min at 4°C. The supernatant was collected as PBS soluble fraction. The pellet was homogenized in 100 µl of SDS-PAGE sample buffer, heated at 95°C for 10 min, centrifuged for 2 min. The supernatant was collected as PBS pellet fraction. Protein extracts were analyzed by 15% SDS-PAGE or 4–12% Bis-Tris gradient gel (Invitrogen). Proteins were digested with trypsin, and the resulting fragments were analyzed by MALDI-TOF mass spectrometry. After staining gels with Coomassie G-250, the selected gel band was excised as 1–2 mm diameter pieces and transferred to a 1. 5 mL Eppendorf tube. A protein-free region of the gel was also excised as background control. The control and test gel sections were destained using three 30 min washes of 60 µL 1∶1 acetonitrile: water at 30°C. Gel pieces were then dried for 10 min under vacuum. The gel sections were subjected to reduction and alkylation using 50 mM Tris (2-carboxyethyl) phosphine (TCEP) at 55°C for 10 min followed by 100 mM iodoacetamide in the dark at 30°C for 60 min. The carboxymethylated gels were thoroughly washed and re-dried in vacuo, then incubated with sequencing grade trypsin (Trypsin Gold, Promega, Madison, WI), 20 ng/µL in 40 mM ammonium bicarbonate, in 20 µL. Upon rehydration of the gels, an additional 15 µL of 40 mM ammonium bicarbonate and 10% acetonitrile was added, and gel sections were incubated at 30°C for 17 h in sealed Eppendorf tubes. The aqueous digestion solutions were transferred to clean 1. 5 mL Eppendorf tubes, and tryptic fragments remaining within the gel sections were recovered by a single extraction with 50 µl of 50% acetonitrile and 2% trifluoracetic acid (TFA) at 30°C for 1 h. The acetonitrile fractions were combined with previous aqueous fractions and the liquid was removed by speed vacuum concentration. The dried samples were resuspended in 10 µL of 30 mg/mL 2,5-dihydroxylbenzonic acid (DHB) (Sigma, St. Louis, MO) dissolved in 33% acetonitrile/0. 1% TFA and 2 µL of peptide/matrix solution was applied on a Bruker Massive Aluminum plate for MALDI-TOF and TOF/TOF analysis. Mass spectra and tandem mass spectra were obtained on a Bruker Ultraflex II TOF/TOF mass spectrometer. Positively charged ions were analyzed in the reflector mode. MS and MS/MS spectra were analyzed with Flex analysis 3. 0 and Bio Tools 3. 0 software (Bruker Daltonics). Measurements were externally calibrated with 8 different peptides ranging from 757. 39 to 3147. 47 (Peptide Calibration Standard I, Bruker Daltonics) and internally recalibrated with peptides from the autoproteolysis of trypsin. Peptide ion searches were performed with Beetlebase (http: //www. bioinformatics. ksu. edu/BeetleBase/) (as well as Metazoa domain_201000104 in NCBInr database) using MASCOT software (Matrix Science). The following parameters were used for the database search: MS and MS/MS accuracies were set to <0. 5 Da, trypsin/P as an enzyme, missed cleavages 1, carbamidomethylation of cysteine as fixed modification, and oxidation of methionine as a variable modification. Sequence motif analysis of the predicted protein sequence was searched in motifs database including PROSITE profiles and Pfam HMMs. The full-length coding sequences for TcCPR27 and TcCPR18 (351 bp and 504 bp, respectively) were amplified from total RNA extracted from pupae (mixture of 0 d- to 5 d-old pupae) by RT-PCR. The cDNAs for TcCPR27 and TcCPR18 were amplified using the following gene specific primers, which included predicted start and stop codons: 5′ ATG CAC GGT GGA GCA GTT C 3′ and 5′ TCA GTT GCC TCC AAT CCC G 3′ for TcCPR27, and 5′ ATG AGA TTA TTT ATT ACA TTG GCC 3′ and 5′ CTA GAT TAA TAA TGT GGT TTG TAA G 3′ for TcCRP18. PCR products were cloned into pGEMT (Promega) and sequenced. Total RNA isolation, cDNA synthesis and real-time PCR were done as described previously [37] using the following primer sets: 5′AGG TTA CGG CCA TCA TCA CTT GGA 3′ and 5′ATT GGT GGT GGA AGT CAT GGG TGT 3′ for TcCPR27,5′ GAA TAC CGC ATC CGT GAC CAC AAA 3′ and 5′CAG GTT CCA ACA AAC TGT AGG TTC CC 3′ for TcCPR18. Total RNA was isolated from whole insects (n = 5) to analyze developmental expression patterns and knock-down levels after RNAi of TcCPR27 and TcCPR18. Total RNA also was isolated from elytra, hindwings, ventral abdomens and dorsal abdomens of pharate adults (5 d-old pupae) (n = 10). The transcript levels of the T. castaneum ribosomal protein S6 (rpS6) were measured to normalize for differences between samples in the concentrations of cDNA templates. dsRNA for TcCPR27 and TcCPR18 was synthesized as described previously [38] using the primers 5′- (T7) -GAC CAC CAC ACC CAT G-3′ and 5′- (T7) -TCA GTT GCC TCC AAT C-3′ for TcCPR27, and 5′- (T7) -GGA AGA GTA CGG TCA TC -3′ and 5′- (T7) -GGT TCC CTT TAC TTT G-3′ for TcCPR18, where T7 indicates the T7 RNA polymerase recognition sequence. The sizes of dsRNAs for TcCPR27 and TcCPR18 were 204 bp and 325 bp, respectively. dsRNAs were injected into last instar larvae [39]. dsRNA for the T. castaneum vermilion gene (dsVer) was used as a negative control [40]. Proteins were extracted from 200 pairs of elytra of 5 d-old pupae as described in Materials and Methods. The homogenate was centrifuged for 2 min at 4°C. The supernatant was applied to a Ni-NTA column equilibrated with 50 mM Tris-HCl, pH 7. 5 containing 0. 2 M NaCl and 20 mM imidazole and washed with the same buffer. Bound proteins were eluted with a 20 to 200 mM imidazole gradient. The fractions were analyzed for protein content by SDS-PAGE. Purified TcCPR27 was used as antigen to generate rabbit antiserum by Cocalico Biologicals, Inc. , PA, USA. Mechanical analysis of elytra was carried out using a TA Instruments RSAIII dynamic mechanical analyzer, by methods described previously [34]. cDNA sequences are deposited at NCBI with accession numbers HQ634478 (TcCPR27) and HQ634479 (TcCPR18).

Primitive insects have two pairs of membranous flight wings, but during the evolution of the beetle lineage the forewings lost their flight function and became modified as hard, rigid covers called elytra for protection of soft body parts of the abdomen and also the delicate flexible hindwings, which retained their flight function. This transformation is manifested by a greatly thickened and rigid (sclerotized) exoskeletal cuticle secreted by the forewing epidermis. We demonstrate that this evolutionary modification is accompanied by the incorporation of two highly abundant structural proteins into the elytral cuticle, namely TcCPR18 and TcCPR27. Depletion of these proteins by RNA interference results in malformation and weakening of the elytra, culminating in insect death. These proteins are also abundant in hard cuticle from other regions such as the pronotum and ventral abdomen, but are absent in soft cuticles, and therefore may function as key determinants of rigid cuticle. Expression of such proteins at high levels in the modified forewing appears to have been a fundamental evolutionary step in the transformation of the membranous wing into a thickened and rigid elytron in the Coleoptera.

Abstract Introduction Results Discussion Materials and Methods

functional genomics rna interference gene function developmental biology molecular development proteins gene expression structural proteins biology extracellular matrix proteins morphogens biochemistry genetics genomics genetics and genomics evolutionary developmental biology

2012

Formation of Rigid, Non-Flight Forewings (Elytra) of a Beetle Requires Two Major Cuticular Proteins

Scrub typhus, murine typhus, and leptospirosis are widely neglected infectious diseases caused by Orientia tsutsugamushi, Rickettsia typhi, and pathogenic Leptospira spp. , respectively. Patients usually present with non-specific symptoms and therefore are commonly diagnosed with acute undifferentiated febrile illness. Consequently, patients face delayed treatment and increased mortality. Antibody-based serological test currently used as gold standard has limitations due to insufficient antibody titers, especially in the early phase of infection. In this study, we aimed to develop multiplex PCR to combine 3 primer pairs that target specific genes encoding 56-kDa TSA of O. tsutsugamushi, 17-kDa antigen of R. typhi, and LipL32 of L. Interrogans and evaluate its performance in comparison to the standard serological tests. Using EDTA blood samples of known patients, the sensitivity and specificity of our multiplex PCR was 100% and 70%, respectively. In addition, the assay was able to diagnose the co-infection of scrub typhus and leptospirosis. The assay may be useful in identifying causative agents during the early phase of these diseases, enabling prompt and appropriate treatment. Scrub typhus, murine typhus, and leptospirosis are widely neglected infectious diseases, especially in the tropical and temperate climate regions, caused by Orientia tsutsugamushi, Rickettsia typhi, and pathogenic Leptospira spp. , respectively. Orientia and Rickettsia are obligate intracellular bacteria. O. tsutsugamushi is transmitted to human through chigger bite, and R. typhi infection is transmitted by inoculation of rat flea’s feces on human skin. Pathogenic leptospires are spirochetal bacteria that are mainly transmitted through abraded skin or mucosa after contact with urine of infected animals, contaminated water, or contaminated soil [1–5]. Unfortunately, current diagnostic tools, together with awareness and experience of physicians, have been the limited. Early clinical manifestations of scrub typhus, murine typhus, and leptospirosis, such as high fever, headache, muscular pain, and anorexia, are non-specific and usually diagnosed as acute undifferentiated febrile illness. These clinical manifestations range from mild, severe, to possibly fatal [6–8]. Eschar caused by chigger bite is a clinical appearance of scrub typhus; however, it is not always present. Moreover, eschar-like lesion can occur in other diseases such as rickettsialpox and anthrax [7,9, 10]. Macular rash is generally present in murine typhus and scrub typhus as well [11]. The clinical manifestations of leptospirosis are biphasic fever and multi-organ failure in severe cases (Weil’s disease) [12]. Due to clinical manifestations of these diseases being recognized as acute undifferentiated febrile illness and subsequently underdiagnosed, a rapid and reliable laboratory investigation is necessary for confirmation and treatment efficiency. Standard diagnostic tests for scrub typhus and murine typhus, i. e. indirect immunofluorescence assay (IFA), and for leptospirosis, i. e. microscopic agglutination test (MAT), mainly depend on detection of antibodies which often return false negative results in the early phase of diseases. To identify pathogen in the early phase of disease, direct pathogen identification is of vital importance since these bacteria can be found in the bloodstream during the first week after onset [13,14]. Thus, a new diagnostic test development based on antigen or DNA detection is essential for rapid diagnosis and proper treatment. Multiplex PCR is a molecular laboratory test used for simultaneous amplification that utilizes different primers in a single tube [15]. It has been applied in diagnosis of several infectious diseases caused by bacteria, fungi, parasites, and viruses [16]. The multiplex PCR is fast and time-saving because it is capable of detecting multiple pathogens at the same time [17]. In this study, we developed multiplex PCR to detect O. tsutsugamushi, L. interrogans, and R. typhi and evaluated its efficiency compared to serological tests. This work selected three target genes encoding 56-kDa type-specific antigen (TSA) of O. tsutsugamushi, lipL32 of L. interrogans, and 17-kDa antigen of R. typhi to identify each bacterium in bacterial cell culture and blood samples of suspected patients. O. tsutsugamushi (Karp, Kato, and Gilliam strains) and R. typhi were obtained from Armed Forces Research Institute, Thailand. Pathogenic L. interrogans (serovar Pyrogenes, Pomona, and Bratislava) and non-pathogenic L. biflexa serovar Patoc were acquired from Department of Microbiology, Faculty of Medicine, Chulalongkorn University, Thailand. Other pathogens, consisting of Escherichia coli, Enterococcus faecalis, Staphylococcus aureus, Klebsiella pneumoniae, Salmonella spp. , Dengue virus (serotype1-4), Plasmodium falciparum, and Plasmodium vivax, were acquired from Department of Microbiology and Department of Parasitology, Faculty of Medicine, Chulalongkorn University, Thailand. This study used 83 EDTA blood samples taken from patients presented with acute undifferentiated febrile illness at Armed Forces Research Institute, King Chulalongkorn Memorial hospital, Loei hospital, Takuapa hospital and Chokchai hospital, Thailand. Patients included were over 18 years old, having acute fever (38°C or higher), exhibiting non-specific symptoms, (headache, muscular pain, anorexia, and rash for 3–5 days), and tested negative for both influenza and dengue antigens. Samples were tested in double-blind experiments and confirmed by IFA, for scrub typhus and murine typhus, and MAT, for leptospirosis. IFA and MAT are antibody detection techniques used as indicator of acute or current exposure; a fourfold increase of antibody titer in paired serum is determined as a positive result [2,13,14,18]. This study was approved by Institutional Review Board, Faculty of Medicine, Chulalongkorn University, Thailand (IRB No. 009/57,534/57, and 380/59). As we had used human EDTA blood obtained from hospital laboratories that perform serology and molecular analysis, an informed consent document was not required. We have not acquired any patient identification, and the data were analyzed anonymously. Bacterial DNA and blood samples were extracted using PureLink Genomic DNA Mini Kit (Invitrogen, USA). First, bacterial cells were lysed by PureLink Genomic Lysis/Binding Buffer. Then, DNA extraction was carried out according to PureLink manufacturer’s instruction. DNA was eluted in 50μl PureLink Genomic Elution Buffer. DNA concentration was measured by UV absorbance at 260/280. To isolate DNA from blood samples, each blood sample (200μl per sample) was added to 180μl PureLink Genomic Lysis/Binding Buffer. DNA isolation was carried out according to PureLink manufacturer’s instruction. Isolated DNA was kept at -20°C until use. The target genes consisted of 56-kDa TSA gene, 17-kDa antigen gene, and lipL32 gene were selected for this study. 56-kDa TSA (56K TSA_F and 56K TSA_R) primers were designed from accession no. M33004, using Primer 3 Software. 17-kDa (17K_F and 17K_R) primers followed the research conducted by Webb et al. [19]. LipL32 (LipL32_45F and LipL32_287R) primers were slightly modified from that of Bourhy et al. [20]. Primer sequences and their amplicon sizes are shown in Table 1. The specificity of primers was bioinformatically aligned with databases in National Center for Biotechnology Information (NCBI) using BioEdit software. Singleplex PCR condition was initially optimized according to GoTaq Flexi manual instruction (Promega, USA). Multiplex PCR was optimized to ensure appropriate amplification. Extension temperatures of 68°C and 72°C were tested in this experiment. Magnesium chloride concentration, ranging from 1. 5–4. 0 mM, was adjusted to give the most intensity for all PCR products. Annealing temperature, ranging from 55°C-61°C, was optimized using a gradient thermal cycler. PCR reaction was carried out in a total volume of 20μl per reaction. Multiplex PCR amplification was performed on Thermalcycler (Applied Biosystems, USA) under the following conditions; 94°C for 5 min, followed by 35 cycles of 94°C for 30 sec, 55–61°C for 1 min, 68/72°C for 1 min, and final extension at 68°C for 10 min. Limit of detection of multiplex PCR was determined using a tenfold serial dilution, ranging from 5x100 to 5x10-6 ng/μl per reaction, of Orientia, Rickettsia and leptospiral DNA mixture. Specificity of multiplex PCR was tested with unrelated pathogens such as Escherichia coli, Enterococcus faecalis, Staphylococcus aureus, Klebsiella pneumonia, Salmonella spp. , Dengue virus (serotype1-4), Plasmodium falciparum, and Plasmodium vivax. The efficiency of this assay was evaluated using masking conditions that mixes Orientia, Rickettsia or Leptospira with other bacteria. Multiplex PCR was validated in blood samples of acute undifferentiated febrile illness patients compared to serological methods (either IFA or MAT). Diagnostic accuracy of this assay was measured in term of sensitivity, specificity, positive predictive value, and negative predictive value compared to the gold standard methods. The equations for calculating sensitivity, specificity, positive predictive value, and negative predictive value are shown in Table 2. After PCR amplification, PCR products were loaded onto 2% agarose gel (Invitrogen, USA) in 0. 5X TBE (Tris-borate with EDTA buffer) and separated by horizontal gel electrophoresis for 30 min at 100 volts. Then, agarose gel was stained in 0. 5 μg/μl ethidium bromide (Sigma, USA) for 5 min and de-stained in deionized water for 10 min. Amplicons were visualized by UV transilluminator (Biorad, USA). In this experiment, optimal conditions were initially determined for singleplex PCR and, subsequently, for the multiplex PCR reaction. The results showed different sizes of PCR products containing a 166-bp fragment of Orientia DNA, 434-bp fragment of Rickettsia DNA, and 243-bp fragment of leptospiral DNA. In singleplex PCR, the 56-kDa TSA primer set was able to detect all strains of O. tsutsugamushi (Karp, Kato, and Gilliam strains). 17-kDa antigen primers could amplify DNA of R. typhi. LipL32 primers amplified pathogenic leptospiral DNA (serovar Bratislava, Pomona, and Pyrogenes), whereas this primer did not amplify non-pathogenic leptospiral DNA (L. biflexa serovar Patoc). (Fig 1). Multiplex PCR was adjusted to ensure appropriate amplification of each target. The optimal extension temperature at 68°C gave higher intensity of amplicons than at 72°C. After magnesium chloride concentrations ranging from 1. 5–4. 0 mM were tested, the optimal concentration of magnesium chloride was found to be 2. 5mM. For annealing temperature optimized using gradient PCR, the optimal annealing temperature of 61°C was determined (Fig 1). To determine the specificity of multiplex PCR, we tested three primer pairs with other pathogens, including E. faecalis, S. aureus, Salmonella spp. , K. pneumoniae, E. coli, Dengue viruses, P. falciparum, and P. vivax. No amplification was obtained from unrelated pathogens (Fig 2A). In addition, individual primer pair was able to amplify each target gene in the masking condition (Fig 2B). In this experiment, we evaluated limit of detection using tenfold serial dilution, ranging from 5x100 to 5x10-6ng/μl per reaction, of mixed DNA and mixed DNA-spiked blood samples. The results have shown that the lowest concentrations of mixed DNA were 0. 5pg/μl or 230 copies for O. tsutsugamushi, 0. 5pg/μl or 106 copies for L. interrogans, and 5pg/μl or 4,160 copies for R. typhi (Fig 3A). In contrast, the detection limit of mixed DNA-spiked blood samples was 0. 5ng/μl for all targets (Fig 3B). To validate efficiency of multiplex PCR assay, we performed multiplex PCR with 83 EDTA blood samples obtained from Thai patients presented with acute undifferentiated febrile illness. Multiplex PCR results were positive in 39 samples (47%), consisting of 22 samples of scrub typhus (26. 5%), 11 samples of leptospirosis (13. 25%), 5 samples of murine typhus (6. 02%), and 1 sample of co-infection between scrub typhus and leptospirosis (1. 2%). For comparison, standard serological methods results were positive in 20 samples (24. 09%). The results are shown in Table 3 and Fig 4. Twenty samples were detected by both serological methods and multiplex PCR. Nineteen samples were detected only by multiplex PCR but tested negative by standard serological methods. In addition, 1 in 39 samples exhibited co-infection between scrub typhus and leptospirosis. The diagnostic sensitivity and specificity of multiplex PCR were 100% and 70%, respectively, when serological methods were used as gold standard. Positive predictive value and negative predictive value of this assay were 51% and 100%, respectively. To our knowledge, this work is the first study to combine 3 primer pairs targeting 56-kDa TSA gene of O. tsutsugamushi, 17-kDa antigen gene of R. Typhi, and lipL32 gene of L. interrogans. Our results showed that this multiplex PCR assay was able to specifically detect each bacterial DNA target. The primer sets did not cross-react with other bacteria. Moreover, we validated this assay using EDTA blood samples from patients with acute undifferentiated febrile illness. Sensitivity and specificity of the multiplex PCR assay developed in this study were 100% and 70%, respectively. In the present, standard diagnostic tests for common causes of acute differentiated fever in Thailand are based on antibody detection, such as MAT for leptospirosis and IFA for both scrub typhus and murine typhus. However, these methods require paired sera, extensive labor, and specialized facilities only available at reference laboratories [18,21]. The delayed diagnosis and treatment may result in severity, complication, and mortality. Therefore, a rapid diagnostic test is necessary for early pathogen detection. Multiplex PCR can be completed within 5 hours, reducing turnaround time for pathogen identification [22,23]. Previous studies reported conventional PCR, nested PCR, real-time PCR, and loop-mediated isothermal amplification (LAMP) for diagnosis of scrub typhus, murine typhus, and leptospirosis [2,24]. However, multiplex PCR for simultaneous detection of these three neglected tropical diseases has never been reported. Multiplex PCR is faster and cheaper than conventional PCR for individual disease and serological methods [25]. In addition, this assay is easier and more convenient than nested PCR as well as LAMP. Multiplex PCR also reduces the risk of contamination between amplification [22]. Thus, multiplex PCR can be a practical and reliable diagnostic tool for multiple pathogens detection. Here, we developed multiplex PCR assay that could be used for diagnosis of three neglected diseases in a single tube. Our results have shown that the extension temperature of 68°C gave optimal result compared to 72°C. These bacterial DNA are highly AT-rich (65–70%), so the reduction in extension temperature can greatly increase PCR amplification for AT-rich templates [26–29]. On the other hand, extension temperature at 72°C prevents DNA synthesis at highly AT-rich region [30]. Because having multiple primer pairs in a single tube negatively affect amplification efficiency, further optimization is crucial. Optimal Mg2+ concentration is necessary for Taq DNA polymerase and specificity of primer-template binding. Inadequate Mg2+ concentration decreases PCR product yield and specificity due to incorrect primer-template binding [31]. The Mg2+ concentration for our assay was optimal at a final concentration of 2. 5 mM. Using a gradient thermal cycler, annealing temperature was optimized for sensitivity and specificity of primer-template binding [32]. Although AT-rich templates require lower annealing temperature [28], 61°C proved to be optimal for this multiplex amplification. The limit of detection of our multiplex PCR were 0. 5 pg/μl or 230 copies for O. tsutsugamushi, 0. 5 pg/μl or 106 copies for L. interrogans, and 5 pg/μl or 4,160 copies for R. typhi. Detection limits of this multiplex PCR were comparable to that of previous studies that detected O. Tsutsugamushi, R. Typhi, and L. Interrogans using singleplex PCR [5,33,34]. The limit of detection of multiplex PCR depends on the followings: 1) combination of multiple primers in a single tube that might affect effective amplification; 2) product size, because amplification is more effective in case of smaller product [15,35]. The higher limit of detection in mixed bacterial DNA-spiked blood samples, 0. 5ng/μl, might be a result of PCR inhibitors, such as hemoglobin and other components in blood [36,37]. Orientia and Rickettsia are obligate intracellular bacteria that can be found in buffy coat of blood specimens [2]. However, buffy coat is more difficult to collect than blood samples. Fortunately, leptospires could be found in bloodstream during the acute phase [18]. Previous studies had also reported detection of Orientia, Rickettsia, and Leptospira DNA in blood samples [2,38]. In this study, we used EDTA blood samples for multiplex PCR validation. Thirty-nine samples (47%) were found to be positive by multiplex PCR, whereas 20 samples (24. 09%) were positive by standard serological tests. Nineteen of the 39 samples, which tested positive in multiplex PCR analysis, exhibited negative serological results. These results may be explained by the period of blood collection and the phase of diseases. In the early phase of infection, bacteremia might not induce sufficient antibody level for serological detection [39–41]. Our study suggests that these 19 blood samples were bacteremic and, inferably, collected during the early phase of infection. Consequently, the multiplex PCR was able to detect bacterial DNA in the blood specimens, even though serological tests showed negative results. One sample exhibited dual positive of O. tsutsugamushi and L. interrogans in multiplex PCR assay, while tested negative in serological analyses. Previous studies had reported co-infection between O. tsutsugamushi and pathogenic Leptospira spp. in Thai and Taiwanese patients with acute undifferentiated fever [42,43]. This study suggests that multiplex PCR can be applied in clinical diagnostic test for identifying pathogenic agents that cause scrub typhus, leptospirosis, murine typhus, and co-infections. This multiplex PCR is a useful tool for early diagnosis of scrub typhus, murine typhus, and leptospirosis. Moreover, the combination of multiplex PCR and serological assays helps increase sensitivity for diagnosis and confirmation. However, sensitivity and specificity determined in this study are only preliminary measurements due to limited sample size. Further study should be performed using a large number of clinical samples. In conclusion, we developed a novel multiplex PCR assay for identifying causative agents of scrub typhus, murine typhus, and leptospirosis in blood samples. This method is a rapid, sensitive, and specific diagnostic test. The multiplex PCR assay will become useful for the development of better health care and treatment of patients presented with acute undifferentiated febrile illness, particularly in endemic areas of these diseases.

Scrub typhus, murine typhus, and leptospirosis are diagnosed as acute undifferentiated febrile illness. Diagnostic tests for these diseases depend on antibody detection. However, antibody detection is still limited by its tendency to return negative results during the early phase of aforementioned diseases. In this study, a novel multiplex PCR has been developed for detecting Orientia tsutsugamushi, Rickettsia typhi, and Leptospira interrogans that are simultaneously amplified in a single tube. The results have shown that multiplex PCR could be used as a diagnostic tool for detecting bacteria during the early phase of scrub typhus, murine typhus, and leptospirosis, allowing for administration of appropriate treatment.

Abstract Introduction Materials and methods Results Discussion

typhus medicine and health sciences leptospira pathology and laboratory medicine pathogens tropical diseases microbiology organisms bacterial diseases neglected tropical diseases molecular biology techniques murine typhus bacteria bacterial pathogens research and analysis methods orienta tsutsugamushi infectious diseases zoonoses serology artificial gene amplification and extension medical microbiology microbial pathogens scrub typhus molecular biology leptospirosis biology and life sciences leptospira interrogans polymerase chain reaction

2019

Development of multiplex PCR for neglected infectious diseases

Immune response against human cytomegalovirus (HCMV) includes a set of persistent cytotoxic NK and CD8 T cells devoted to eliminate infected cells and to prevent reactivation. CD8 T cells against HCMV antigens (pp65, IE1) presented by HLA class-I molecules are well characterized and they associate with efficient virus control. HLA-E-restricted CD8 T cells targeting HCMV UL40 signal peptides (HLA-EUL40) have recently emerged as a non-conventional T-cell response also observed in some hosts. The occurrence, specificity and features of HLA-EUL40 CD8 T-cell responses remain mostly unknown. Here, we detected and quantified these responses in blood samples from healthy blood donors (n = 25) and kidney transplant recipients (n = 121) and we investigated the biological determinants involved in their occurrence. Longitudinal and phenotype ex vivo analyses were performed in comparison to HLA-A*02/pp65-specific CD8 T cells. Using a set of 11 HLA-E/UL40 peptide tetramers we demonstrated the presence of HLA-EUL40 CD8 αβT cells in up to 32% of seropositive HCMV+ hosts that may represent up to 38% of total circulating CD8 T-cells at a time point suggesting a strong expansion post-infection. Host’s HLA-A*02 allele, HLA-E *01: 01/*01: 03 genotype and sequence of the UL40 peptide from the infecting strain are major factors affecting the incidence of HLA-EUL40 CD8 T cells. These cells are effector memory CD8 (CD45RAhighROlow, CCR7-, CD27-, CD28-) characterized by a low level of PD-1 expression. HLA-EUL40 responses appear early post-infection and display a broad, unbiased, Vβ repertoire. Although induced in HCMV strain-dependent, UL4015-23-specific manner, HLA-EUL40 CD8 T cells are reactive toward a broader set of nonapeptides varying in 1–3 residues including most HLA-I signal peptides. Thus, HCMV induces strong and life-long lasting HLA-EUL40 CD8 T cells with potential allogeneic or/and autologous reactivity that take place selectively in at least a third of infections according to virus strain and host HLA concordance. Human cytomegalovirus (HCMV; human herpesvirus 5) is the prototype member of β-herpesvirus and a widespread opportunistic pathogen. In healthy individuals, primary infection is asymptomatic and is followed by a life-long, persistent, infection that is controlled by host immune system [1]. Nevertheless, HCMV is a major cause of morbidity and mortality in immunocompromised individuals such as transplant recipients or HIV-infected patients. HCMV is the most common cause of congenital infection in the world that can result in neurodevelopmental delay and sensorineural hearing loss. HCMV disease can manifest in many forms, including infectious mononucleosis, hepatitis, post-transplant arteriosclerosis, pneumonia, colitis, immune senescence, and alteration to the immune repertoire [1]. The impact of HCMV on the outcome of solid organ transplantation (SOT) is substantial. HCMV not only causes a highly morbid and potentially fatal illness but also indirectly influences other relevant outcomes, such as allograft acute and chronic rejection, other opportunistic infections, post-transplant lymphoproliferative disorders, vascular disease, and overall patient and allograft survival [2,3]. Because of the magnitude of its direct and indirect impacts, there have been extraordinary efforts to define strategies for its prevention, monitoring and treatment [1,4]. Cellular immune response is the major mechanism by which HCMV replication is controlled [5,6]. Large human HCMV-specific T-cell responses have been described in numerous published reports, particularly in the transplantation setting and in ageing [7,8]. HCMV-specific T cells in healthy adults can constitute as much as 10% of the total memory CD4 and CD8 T cells that recognize multiple viral proteins, notably, pp65, IE1, IE2 and gB [9,10]. Suppression of the number and function of HCMV-specific CD4 and CD8 T cells allows reactivation of the virus from latency, leading to uncontrolled viral replication and clinical disease in immunocompromised hosts, including SOT recipients [5]. The CD8 T-cell response appears as the most important component of the anti-HCMV immune response [7], although CD4 T cells and natural killer (NK) cells also play a significant role [11,12]. Expanded HCMV-specific responses are often thought to be a requirement for protection and could result from the life-long latency of HCMV in specific cells, interspaced with episodic reactivations that gradually increase response size in a process called inflation [13,14]. HCMV-specific T cells were first described as those able to recognize the immunodominant antigen immediate early 1 (IE1) [15], but later studies emphasized the importance of T cells that target a tegument phosphoprotein of 65kDa (pp65/UL83) [16]. The original epitope identification studies focused on NLVPMVATV, a HLA-A*02 restricted epitope within pp65 that was defined as a “typical” epitope because of its common detection in HLA-A*02–positive individuals. The identification of other less common epitopes targeted by HCMV-specific T cells has extended the panel of HCMV-reactive T cells in humans [9,10]. In murine models, subdominant epitopes have been shown to be protective [17]. Despite technical advances in terms of HCMV-specific T-cell response monitoring [18], a correlation between T-cell responses and clinical protection has not been established to date. This underlines a need for a global analysis of anti-HCMV T cell responses at both qualitative and quantitative levels investigating response numbers, sizes, hierarchy levels, peptide specificities, time course and duration. HCMV-specific CD8 T cells directed against UL40 epitopes presented by HLA-E have more recently emerged as an additive piece in the complexity of anti-HCMV immune response [19]. HLA-E is a poorly polymorphic non-classical (MHC-Ib) HLA molecule. Although more than 20 HLA-E alleles have been registered, only two nonsynonymous HLA-E alleles: HLA-E*01: 01 (HLA-E107R) and HLA-E*01: 03 (HLA-E107G) that differ by a single amino acid (R107G) have been found in most populations [20]. Cell surface expression of HLA-E depends on binding of a conserved 9-mer peptide naturally provided by the N-terminal signal peptide of classical HLA-I or HLA-G molecules. HCMV UL40 signal peptide contains a 9-mer sequence with an exact sequence identity to endogenous HLA-E–binding peptides. The prototype of UL40 peptide loaded on HLA-E molecules is VMAPRTLIL provided by the AD169 HCMV strain [21]. As a consequence, HCMV UL40 promotes efficient cell surface expression and stabilization of HLA-E independently of TAP function [21,22]. HLA-E containing peptides engage two types of receptors. HLA-E binds the NK cell inhibitory receptor CD94/NKG2A [23,24] and, thereby, promotes efficient protection against lysis by CD94/NKG2A+ NK cells [22,23,25,26]. In addition to CD94/NKG2A, HLA-E interacts with CD94/NKG2C, albeit with lower affinity. CD94/NKG2C is an activating receptor predominantly expressed on a relatively small population of NK cells. Interestingly, the frequency of this CD94/NKG2C+ NK subset increases in HCMV-infected individuals [27] [6]. HLA-E/UL40 (HLA-EUL40) complexes also trigger TCR-dependent activation of a subset of CD8 αβ T cells [28–30]. UL40-specific/anti-HCMV HLA-E–restricted CD8 cytotoxic T-cell responses have been reported in healthy donors and in kidney and lung transplant recipients and associated with a possible harmful impact on graft endothelial cells [30] and allograft survival [31]. Characterization of these CD8 T-cell subsets in healthy and transplanted population remains sparse and no longitudinal study is available. In healthy hosts, the beginning and duration of HCMV infection are usually unknown, thus monitoring the development of T-cell responses starting at the time of infection is not possible except in the setting of organ transplantation where post-graft primary HCMV infections are frequent and require a specific follow-up. Our study investigated the presence of circulating HLA-E-restricted CD8 T cells in a cohort of kidney transplant recipients (KTR, n = 121) during either an active HCMV infection (at primary infection or at reactivation) or at latency and in HCMV seropositive (HCMV+) healthy volunteers (HV, n = 25). Using a set of HLA-E tetramers refolded with 11 different UL40 epitopes to cover the diversity of HCMV clinical strains, we provide here a quantitative analysis of HLA-EUL40-restricted anti-HCMV T-cell response in hosts. The frequency, the magnitude, the time course of HLA-EUL40-restricted anti-HCMV CD8 T-cell responses, as well as the phenotype and the specificity of peptide recognition of these subsets, were documented ex vivo in comparison to the conventional HLA-A*02pp65 CD8 T-cell responses. Altogether our findings reveal that HCMV induces early long-lasting HLA-E–restricted, UL40-specific unconventional CD8 T-cell responses that often parallels HLA-I-restricted CD8 T cells. Although their induction seems initially restricted by both the viral infecting strain and host’s HLA-I, extended peptide recognition may occur allowing these effector responses to potentially target self and allogeneic, donor-specific, HLA-I peptides. Although UL40-specific HLA-E-restricted CD8 T-cells have been described in a few HCMV seropositive (HCMV+) individuals [28–30,32], only sparse data are available concerning their characteristics and post-infection occurrence. To address this point, we performed a retrospective detection and quantification of circulating HLA-EUL40-restricted CD8 T-cell responses in a cohort of kidney transplant recipients (KTR, n = 121) and in HCMV+ healthy volunteers (HV, n = 25). Our study cohort included transplanted patients segregated into 4 groups according to recipient’s HCMV serology (HCMV- and HCMV+) and, for HCMV+ patients, the status of infection (primary, latent, active) at 12 months post-transplantation. Demographic and clinical characteristics of the cohort are presented in Table 1. UL40-specific HLA-E-restricted CD8 T cells were analysed ex vivo in blood samples after PBMC isolation using a multi-parameter (CD3+CD8α+TCRγδ-) flow cytometry assay subsequent to the blockade of the CD94 receptor using a specific blocking mAb. Our protocol was adapted from Pietra et al. [29] and allows a sensitive (threshold of detection: 0. 1% of total CD8 TCRαβ T cells) and peptide-specific analysis of HLA-EUL40 CD8 T-cell populations (S1 Fig). Detection of HLA-A*02: 01/pp65 CD8 T (HLA-A*02pp65) cells was carried out in parallel for a comprehensive analysis of unconventional (HLA-E-restricted) versus conventional (HLA-A*02: 01-restricted) anti-HCMV responses. Banked blood samples, harvested at M12 post-transplantation, were investigated using a set of HLA-E tetramers loaded with 8 different UL4015-23 peptides to encompass the usual UL4015-23 variability among common HCMV strains [33,34]. The 8 HLA-E tetramer/peptide complexes were tested individually. Fig 1 shows the distribution of HLA-EUL40 CD8 T-cell responses (detected for at least 1 tetramer HLA-E/UL40 peptide complex) versus HLA-A*02pp65 CD8 T-cell responses in the various groups. In HCMV- transplanted patients no HLA-EUL40 nor HLA-A*02pp65 T-cell response was detected, consistent with the concept that these responses are induced by and specific to HCMV infection. HLA-EUL40 CD8 T cells were detectable in all HCMV+ subgroups (primary, latency, reactivation) and were present in an overall of 28. 7% of HCMV+ transplanted recipients and 32. 0% in HCMV+ blood donors. By comparison the overall incidence of HLA-A*02pp65 T-cell responses in HLA-A*02 patients and healthy hosts was 65. 0% and 46. 1%, respectively. HLA-A*02pp65 T-cell responses were roughly similar in frequency upon primary (58. 3%), latent (68. 4%) and active (66. 7%) infection while HLA-EUL40 CD8 T-cell responses were lower upon primary infection compared to other groups. Interestingly, HLA-EUL40 CD8 T-cell responses was more frequent in HLA-A*02 as compared to non HLA-A*02 hosts (37. 5% versus 20. 0% for transplant recipients and 46. 1% versus 16. 7% for HV; p = 0. 0318). In HLA-A*02 recipients, HLA-EUL40 CD8 T-cell responses were found either associated with (32. 2%) or independent (16. 1%) of a HLA-A*02pp65 T-cell response. Coexistence of HLA-EUL40 and HLA-A*02pp65 CD8 T-cell responses also occurs in 33. 3% of healthy hosts. Together these results reveal a very high incidence (up to 46%) of HLA-EUL40 CD8 T-cell responses in HCMV+ hosts with no significant difference between transplanted patients and healthy individuals suggesting that antiviral and immunosuppressive regimens have no impact of these cell subsets at M12. These cells are detected more frequently in hosts carrying an HLA-A*02 allele. Unconventional CD8 T cells can be detected independently of detectable conventional HLA-A*02pp65 T-cell response. Presence of HLA-EUL40 CD8 T cells early post-infection (primary or reactivation) as well as at latency suggests long lived cell subsets consistent with memory anti-HCMV response. Of interest, the lack of HLA-EUL40 CD8 T cells in HCMV- transplant recipients may also indicate that, although a full sequence homology between common UL4015-23 peptides and signal peptides from most HLA-A, -B and -C molecules, presentation of allogeneic (i. e. donor) HLA-I signal peptides (HLAsp) by HLA-E-expressing uninfected cells in the graft doesn’t drive the generation of consequent anti-donor HLA-EHLAsp CD8 T-cell response. Detection of HLA-EUL40 CD8 T cells suggested that these unconventional responses occur more frequently in HLA-A*02 carriers. Genotyping of HLA-A was then performed to decipher this association. Firstly, HLA-A*02 allele frequency was 28% in the HCMV+ hosts in our study, a value similar to those found in the HCMV- counterpart (36%, p = 0. 1982) (Fig 2A). Our findings indicate that HLA-A*02 allele frequency was significantly higher in HCMV+ hosts with HLA-EUL40 responses than in non-responders (44% versus 22%, p = 0. 0026, Fig 2A). Next, distribution of HLA-A*02 genotypes were compared between HCMV+ and HCMV- individuals. A similar distribution of HLA-A*02 genotypes was observed in both groups (Fig 2B). However, HCMV+ hosts that display HLA-EUL40 responses were more frequently hosts carrying two HLA-A*02 alleles than hosts without response (19% versus 0%, p = 0. 0002). Similar analysis was performed for HLA-E alleles and genotypes. HLA-E sequencing allowed us to discriminate the two major HLA-E*01: 01 and HLA-E*01: 03 alleles. These variants differ in a single amino acid at position 107 when an arginine (R) in HLA-E*01: 01 is replaced by a glycine (G) in HLA-E*01: 03 resulting in different thermal stabilities and lengths of interaction with cognate receptors [33]. HLA-E allele distribution was found equal for HCMV- and HCMV+ hosts (Fig 2C, left panel) and no difference in HLA-E allele frequency was observed among HCMV+ individuals with or without HLA-EUL40 responses (57. 0% versus 53. 0%, and 43. 0% versus 47. 0%, for HLA-E*01: 01 and *01: 03 respectively, p = 0. 7541, Fig 2C, right panel). An equal distribution of HLA-E genotypes was calculated for HCMV+ and HCMV- hosts (p = 0. 1661) (Fig 2D, left panel). However, a significant change occurs in HLA-E genotypes for hosts that display or not HLA-EUL40 responses (p = 0. 0323) with a higher prevalence of heterozygous HLA-E*01: 01/HLA-E*01: 03 in hosts with HLA-EUL40 responses (Fig 2D, right panel). No impact of donor HLA was found. These findings support a role for immunogenetic factors in the occurrence of HLA-EUL40 responses upon HCMV infection and associate HLA-A*02/A*02 and HLA-E*01: 01/HLA-E*01: 03 genotypes as independent (p = 0. 85) positive factors promoting HLA-EUL40 responses. Next, we sought to determine the specificity of HLA-EUL40 CD8 T-cell responses toward UL40 peptide provided by the host’s HCMV infecting strain. To this aim, DNAs isolated from whole blood samples from transplant recipients undergoing either a primary infection (n = 18) of a reactivation (n = 7) of HCMV during the 12 months post-transplantation were used for UL40 protein (AA 1–221) sequencing. Sequencing identified a total 32 UL40 sequences for the 25 infected patients, some patients carrying more than one infecting strain (Table 2). Overall variability of full UL40 protein among strains is reported in S2 Fig and was consistent with a previous report [34]. UL40 variability affects 38 AA along the sequence but mostly concentrates within the region encoding the signal peptide (UL401-37) and in particular inside UL4015-23, the HLA-E binding epitope (Fig 3A). Notably, AA22 and to a lesser extend AA20 that correspond to the peptide position P8 and P6, respectively, two critical residues for the interaction with the CD94/NKG2-A or -C or with the TCR of specific T cells [32] [36], were the most variable, with respectively 48. 2% and 19. 6% of AA variability and up to 5 and 3 different AA (Fig 3A and 3B). In contrast, residues 16 (P2), 21 (P7) and 23 (P9) that correspond to the 3 major anchor AA for the peptide binding pockets of HLA-E, were highly conserved [37]. Three major UL4015-23 sequences (VMAPRTLIL, VMAPRTLLL, VMAPRSLLL) accounted for 62. 5% of the HCMV strains detected in patients (15 out of 25) while 10 other UL40 sequences were found only in a single patient (Table 2). These data confirmed that consensus UL4015-23 sequences such as VMAPRTLIL and VMAPRTLLL are highly predominant in clinical strains. Interestingly, VMAPRTLIL and VMAPRTLLL UL40 sequences are fully homologous to signal peptide sequence for the majority of HLA-A and HLA-C alleles excluding the most common HLA-A*02 and HLA-C*07. Since banked blood samples were available for 23 of these patients, we next assessed the presence of HLA-EUL40 CD8 T-cell responses using dedicated HLA-E/UL40 tetramers. HLA-EUL40 CD8 T cells were detected in 6 out of the 23 patients (26. 1%) and illustrated for 4 out of the 6 in the Fig 3C. As shown in Fig 3C, when HLA-EUL40 CD8 T-cell responses were investigated using HLA-E tetramers loaded with the UL40 peptide that we identified in their own infecting strain, HCMV strain-specific HLA-E-restricted T cells were detected in patients. Importantly, percentages of HLA-EUL40 CD8 T cells vary from 2. 9% up to 38. 6% of total CD8 αβT cells in the blood sample at the time of detection. Fig 3C also illustrates the complexity of the patterns of HLA-EUL40 CD8 T responses. Indeed, while in a large majority of hosts, homogenous CD8α bright populations were observed exemplified in patients #108 and #109, in few hosts, such as #026, multiple populations that display various levels of CD8α expression (low and high) were observed. This may reflect either the detection of concomitant subsets of HLA-EUL40 CD8 T at a particular time point or different stage of activation for a single population or both. Thus, our data sustain previous report showing the UL4015-23 nonapeptide as a highly polymorphic region inside the viral UL40 protein [34]. Our data show that UL40 polymorphism also drives (strain-specific) antigen-specific HLA-E-restricted T cells. However, in our study only a limited set of canonical UL40 peptides were found in the majority of clinical infecting strains (such as VMAPRTLIL, VMAPRTLLL and VMAPRSLLL identified in 20 out of 32 strains) and allowed strain-specific HLA-EUL40 CD8 T cells. Interestingly, about a third of patients were infected by an HCMV strain carrying a non-canonical UL4015-23 sequence that display variant amino acid on the residues P1, P3, P4, P5, P6 and P8. Thus we speculate that such HCMV strains for which no HLA-EUL40 CD8 T-cell response was detected in our assays may hold UL40SP probably not able to bind HLA-E. Nevertheless, we cannot rule out the possibility that detection of HLA-EUL40 CD8 T cells was underestimated in our study due to the lack of tetramers loaded with the full set of UL40 sequences identified in clinical isolates. To further characterize the HLA-E-restricted anti-HCMV T-cell responses, time course of these responses during the acute phase of infection and beyond, and T-cell activation markers were monitored post-infection in patients (n = 16) with either a primate infection or a reactivation of the virus. Results from 3 patients are illustrated in the Fig 4A that summarizes the most frequent profiles that we observed. Upon primary infection (exemplified by patient #109), HLA-EUL40 CD8 T cells develop early and most often concomitantly to HLA-A*02pp65 T-cell response. HLA-EUL40 CD8 T cells are detected in blood 1 month post-infection (patient #107 and #109) and may even precede detection of HLA-A*02pp65 T-cell response (patient #109). HLA-EUL40 CD8 T-cell response can be either predominant (patient #109) or lower in frequency among total CD8 αβT cells compared to conventional HLA-A*02pp65 response (patients #107 and #108). Patient #108 illustrates a HCMV reactivation with a pre-existing HLA-A*02pp65 population leading to a clear increase in the percentage of HLA-A*02pp65 CD8 T cells at the time of reactivation and a de novo induction of HLA-EUL40 CD8 T cells. For the 3 patients, consistent long term (M9-12 post-infection) responses were observed ranging from 1. 2 to 15. 6% for HLA-EUL40 CD8 T cells and 0. 4 to 47. 7% HLA-A*02pp65 CD8 T cells. Activation markers (CD69 and PD-1) were analysed by flow cytometry for both anti-HCMV CD8 T-cell subsets at each time point. Fig 4B reports on the relative expression of CD69 and PD-1 investigated ex vivo at M6 post-transplantation for the 3 recipients. Overall, we found that both subsets display similar rate of CD69+ cells. In contrast, there were striking differences in the programmed death-1 (PD-1) expression between the 2 subsets with a lower percentage of expression for PD-1 on HLA-EUL40 CD8 T cells as compared to HLA-A*02pp65 CD8 T cells (Fig 4C). These discrepancies were found at all time points post-induction (S3 Fig). More than 45% of HLA-A*02pp65 CD8 T-cell subsets express sustained levels of PD-1+ after a primary infection (patients #107 and #109) and up to 100% upon reactivation (patient #108). These investigations that shape the temporal occurrence of HLA-EUL40 CD8 T cells post-infection reveal that both responses, conventional and unconventional, may be very close in kinetic, persistence and in percentage of total CD8 T cells in blood. However, although similarly activated early post-infection, low expression of PD-1 could be a feature of HLA-E restricted anti-HCMV T-cell responses. The functional and phenotype description of HLA-EUL40 CD8 T cells is rather limited. Our phenotypic analyses by flow cytometry, performed ex vivo for 3 patients (#107, #108 and #109) confirmed that HLA-EUL40 T cells belong to the CD3+CD4-CD8αβ+TCRαβ+ T cell subset. HLA-EUL40 T cells exhibited a phenotype (CR45RAhighCD45ROlowCD27-CD28-CD57+/-CCR7-, S4 Fig) consistent with effector memory T-cell response as previously reported [38]. Furthermore, in our study, to better characterize anti-HCMV HLA-E-restricted responses, HLA-EUL40 CD8 T-cell lines were generated by cell sorting using for each patient an HLA-E tetramer loaded with the UL40 peptide identified in their own HCMV circulating strain. (Fig 3C). PBMCs from 5 HCMV+ patients with a primary infection or a reactivation (KTR #104, #105, #107, #108, #109) were sorted and then amplified in vitro to reach a purity>95% (defined by tetramer staining using the HLA-E/UL40 complexes employed for sorting). Amplified HLA-EUL40 T cells were used for the analysis of T-cell receptor β chain variable region (TCR-Vβ) expression by flow cytometry using 24 antibodies reactive to 70% of the human TCR-Vβ repertoire. Given the fact that HLA-E is a poorly polymorphic gene and that HLA-EUL40 CD8 T cells recognize a restricted number of UL4015-23 peptides, the question of the existence of a public T-cell repertoire between individuals was raised. Consistently, only few analysis of TCR sequences from UL40-specific T-cell clones have been reported yet and display a limited number of TCR including Vβ3,5. 1,9, 16,22 [29]. HLA-EUL40 CD8 T-cell population expressing a dominant Vβ chain sub-family was obtained for 3 patients while another one gives rise to oligoclonal populations (from 2 to 6 subsets detected) with variable distribution (Fig 5A). This suggests the sorting of multiple, coexisting, HLA-EUL40 CD8 T-cell populations in this patient. Interestingly, a broad TCR-Vβ repertoire was found with 16 Vβ identified (Vβ1,2, 3,5. 1,7. 1,8, 9,12,13. 1,13. 2,13. 6,14,16,17,22 and 23) thus enlarging the Vβ repertoire previously described for these cells. For patient #104 that exhibits oligoclonal T-cell populations only 19% of Vβ repertoire was identified suggesting that this patient carry a dominant Vβ not detectable in our assay. No dominant Vβ was identified for patient #109 with oligoclonal HLA-E-restricted subsets covering 82% of its repertoire. Next, amplified HLA-EUL40 CD8 T-cell populations were investigated for their capacity to produce TNF in response to TCR engagement in a peptide-specific manner. To this aim, the 5 enriched populations were stimulated with 11 HLA-E/UL40 peptide tetramers, used individually, for 5h before intracellular TNF staining. The set-up of experimental conditions are depicted in S5 Fig. In most cases, T cells were highly stimulated (up to 50% of cells expressing TNF) by the HLA-E tetramers loaded with the peptide corresponding to UL4015-23 identified in their own infecting stain (Fig 5B). However, consistent stimulations (10 to 50%) were also obtained for HLA-E tetramers loaded with other peptides. Interestingly, T-cell activation can be induced by peptides that correspond to self and donor-specific allogeneic HLASP supporting the idea that these T cells may be auto- and/or alloreactive (Fig 5B). In most of cases, changing in P8 or P6 residues of UL4015-23 peptides diminished or abolished the reactivity of T cells, showing the relative importance of these two amino acids for the interaction of the HLA-E complexes with the TCR. Such cross-stimulation was observed similarly for T-cell populations containing a single dominant Vβ subset or oligoclonal subsets. Magnitude of the stimulation was peptide-dependent and differs for each T-cell subpopulation. In most cases the dominant peptide issued from the infecting strain and used for sorting, gives the highest score of T-cell activation. Together, these data may suggest that a single dominant Vβ subset of HLA-EUL40 CD8 T cells induced in a UL40 peptide-dependent manner could be activated by HLA-E molecules presenting UL40 peptides with a degree of homology including a panel of HLASP. The use of 8 different HLA-E/UL40 peptide tetramers allowed us to decipher the spectrum of HLA-EUL40 responses generated post HCMV infection. This assay provided a qualitative and quantitative analysis of HLA-E-restricted responses for the 31 HCMV+ transplanted patients and HV initially found to carry at least one HLA-EUL40 CD8 T-cell response. Responses were analyzed to define, for each individual, both peptide specificity and relative strength of the responses (percentage of subset among total circulating CD8 T cells). As a result, consistent responses were observed for the 8 tetramers tested. VMAPRTLLL, VMAPRTLIL, VMAPRTLVL, VMAPRTVLL, VMAPRSLLL and VMAPRSLIL are the most frequently recognized peptides by HLA-EUL40 responses in terms of both occurrence and magnitude. The number of circulating HLA-E-restricted CD8 T cells varies in the range of 0. 1% (detection threshold) up to around 40% of total TCRαβ CD8 T cells. These percentages were similar or even higher than those we obtained for HLA-A*02: 01-restricted responses (Fig 6A). An overview of the panel of HLA-EUL40 T-cell responses detected in patients and HV is provided in the Fig 6B. This analysis shows that HCMV+ subjects usually display HLA-EUL40 T-cell responses against more than a single HLA-E/peptide complex. The total number of responses (from 1 to 8) detected as well as the nature of UL40 peptide recognition (specificity and magnitude) is variable among the hosts. Similar variability is observed for HV and transplant recipient populations. These ex vivo findings sustained our results above obtained with cell lines and showing that a monoclonal HLA-EUL40 CD8 T-cell subset can be activated by a set of UL40 peptides. Nevertheless, we cannot exclude that a multiplicity of responses can also coexist in hosts resulting from coinfection. Considering the ability of HCMV to generate HLA-EUL40 T-cell responses that recognize multiple peptides we next sought to determine whether the detected HLA-EUL40 CD8 T cells may target autologous or allogeneic (i. e. provided by the transplant donor) HLA-I signal peptide in the KTR. To this aim, sequence of HLA-I (-A, -B, -C) signal peptide carried by the host (KTR or HV) or by the transplant donors were compared to the UL40 sequences targeted by HLA-EUL40 CD8 T-cell responses to identify self and allogeneic, donor-specific or non donor-specific, peptides, respectively. Potential self or allogeneic recognition mediated by anti-HCMV HLA-E-restricted T-cell subset are presented in the Fig 6C. Due to full sequence homology between UL40 viral peptides and HLA-I signal peptides, most of UL40-induced responses were found to recognize at least one autologous HLA peptide for all HCMV+ individuals. Moreover, in most cases (70% of responders) HLA-EUL40 responses may also potentially target transplant HLA-I signal peptide presented by HLA-E molecules on the graft. Ex vivo HLA-EUL40 tetramer staining allowed us to provide a qualitative and quantitative assessment of unconventional CD8 T cells directed against HCMV. This unconventional T-cell subset is restricted by the MHC-Ib, HLA-E molecule, and targets UL40 signal peptide (UL4015-23). A number of conclusions can be drawn from this study. First, a major finding was the high prevalence of this CD8 T-cell population investigated in HCMV+ transplant recipients and healthy volunteers. HLA-EUL40 CD8 T cells were detected in 31 out of 105 (29. 5%) HCMV+ hosts. About half (46. 1%) of HCMV+ healthy HLA-A*02 blood donors possesses detectable HLA-EUL40 CD8 T cells. An equal proportion of HLA-A*02 blood donors (46. 1%) possess HLA-A*02pp65 CD8 T cells and 1/3 of these individuals display both anti-HCMV CD8 T cells. Although, in our cohort of transplant recipients, HLA-A*02pp65 specific T cells were more frequently detected than HLA-EUL40 CD8 T cells, the latter were found in over 35% of kidney transplant recipients. Nevertheless, we cannot exclude that these values were underestimated since ideally, a broader panel of HLA-EUL40 complexes would be used for an exhaustive detection. Moreover, T-cell populations below 0. 1% (our threshold of detection) were not considered. Together these results support the idea that HLA-E-restricted T-cell response belongs to the usual T-cell response against HCMV UL40. Conventional T-cell responses to HCMV peptides, such as dominant responses to the pp65 and IE epitopes presented by HLA-A*02 and HLA-B*07, can regularly reach 5–10% of total CD8 T cells in the blood of healthy adults and even greater with up to 30% of total CD8 T cells are reported in some studies [9,10,39]. However, there is extensive variability in the size of T-cell responses between individuals. The reasons for this variability are not fully understood but may include the dose and timing of infection, as well as the HLA restriction element. Here we show that similar disparity also occurs for HLA-E-restricted anti-UL40 CD8 T cells with frequencies varying from 0. 1% to over 38% of total CD8 T cells according to the hosts (median value: 2. 2%). These values are the highest reported for this unconventional subset ex vivo. Previous studies established ex vivo percentages of HLA-E-restricted anti-UL40 CD8 T cells from 0. 05% [31] to 14% [30,32]. Thus, HLA-E-restricted responses mirror HLA-Ia-restricted responses in both frequency and magnitude. Our longitudinal analysis demonstrated that these T-cell populations develop early post-infection and expand quickly to reach maximal rate between 2 to 11 months post primary infection and within 1 month post reactivation. Tetramer staining of HLA-EUL40 CD8 T cells showed continued expansion post-infection and stabilization at high frequencies. In our cohort study, anti-HCMV HLA-E-restricted, and to a lesser extent HLA-A*02-restricted T-cell responses appear more frequent during latent and reactivations/secondary infections than during primary infections. Although this difference may be due to variations in the time interval between infection and the time point selected for detection assay (M12) among individuals or an effect of immunosuppressive regimen, this could also reflect a delay in HLA-EUL40 CD8 T-cell induction. A key point of this study is to provide evidence for a positive correlation between HLA-A*02 allele and the occurrence of HLA-E-restricted anti-HCMV CD8 T cells. Firstly, using HLA-EUL40 tetramer staining, anti-HCMV HLA-E-restricted were detected more often in HLA-A*02 hosts. Next, HLA sequencing further confirmed a significantly higher rate of hosts carrying at least one HLA-A*02 allele among HLA-EUL40 CD8 T-cell responders compared to non-responders. Moreover, all HLA-A*02+/+ HCMV+ individuals (n = 6) developed an HLA-EUL40 CD8 T-cell response. The positive correlation between HLA-A*02 allele and HLA-EUL40 CD8 T-cell response could be related to the sequence of HLA-A*02 signal peptide (VMAPRTLVL). Indeed, HLA-EUL40 CD8 T-cell responses that have been identified in HCMV infection typically involved epitopes that are structurally related to canonical HLA-I leader sequences but foreign to the hosts [19,40,41]. Consistent with the paucity of the VMAPRTLVL sequence among viral strains, UL40 sequencing of host’s circulating strains allowed us to identify the VMAPRTLVL sequence only in a single clinical strain out of 32. Thus, it could be suspected that the presence of HLA-A*02 decreases the chances that a host will present a signal peptide derived from a different HLA-I allele, one that could cause negative selection of HLA-EUL40 reactive TCR. In that respect, when HLA-A*02 is present, deletion of HLA-EUL40-responsive T cells is less likely. HLA-E*01: 01 (HLA-E107R) and HLA-E*01: 03 (HLA-E107G) alleles only differ in a single amino acid at position 107 and the frequencies of these two variants are equal in most populations [20]. It has been shown that the HLA-E*01: 03 variant is usually expressed at higher levels than HLA-E*01: 01 [33]. Although located outside the peptide-binding groove, the mutant AA at position 107 may also possibly affect the conformation of HLA-E or its association with β2-microglobulin resulting in the presentation of different peptide repertoires [42]. We found no HLA-E allele preference associated with the establishment of an HLA-EUL40 CD8 T-cell response. Instead, we report a higher prevalence of HLA-E*01: 01/*01: 03 heterozygous among individuals carrying an HLA-EUL40 CD8 T-cell response. Interestingly, it has been demonstrated for HLA-E and for the non-human primate HLA-E ortholog that a large panel of identified peptides can be presented by all allotypes [43]. Both alleles present a limited set of peptides derived from class I leader sequences physiologically [42]. In stress conditions (viral infections, tumors), HLA-E can present peptides from other sources than the signal sequences of classical HLA-I molecules [38,44]. Recent studies demonstrated that the HLA-E alternative peptide repertoire observed in pathophysiological conditions seems not to be shared equally by the two HLA-E alleles [42,45]. Comparing the HLA-E*01: 03-restricted peptides to those of HLA-E*01: 01, Celik et al. demonstrated that the peptide repertoire of both alleles greatly differs in the absence of HLA class I molecules leading to functional disparity between both alleles [45]. Consistent with these observations, it is likely that bearing both *01: 01/*01: 03 alleles may improve HLA-E stability and the diversity of peptide presentation and thus increase HLA-EUL40 T-cell responses as suggested by our findings. In transplant recipients, the impact of donor HLA was investigated in parallel to the impact of recipient of donor HLA. We found no significant impact neither for HLA-A, -B, -C or HLA-E alleles nor for a mismatch between donor and recipient for HLA-A, -B, -C or HLA-E. An elegant study from Wang et al. suggested that HCMV-specific CD8 TCR repertoire diversity is more important than CD8 T-cell response magnitude for the control of persistent HCMV infection [46]. Using a single-cell based approach for the clonotype analysis of human CD8 TCRαβ repertoires they demonstrate a high prevalence of both TCRα and TCRβ public motif usage. Our analysis of TCR Vβ usage by HLA-EUL40-specific T cells investigated after in vitro expansion showed no predominating TCR Vβ usage for HLA-EUL40-specific T cells, indicative of an unbiased T cell response. A donor-specific focus revealed diverse and unique TCR Vβ chain repertoire in each host. Analysis of TCR Vα repertoire remains to be performed to fully define T-cell repertoire diversity. Ex vivo phenotype analysis at distance from the infection revealed that HLA-EUL40 CD8 T cells belong to effector memory cells, most probably TEMRA, that display CD45RAhigh/CD45ROlow. Chronic viral infections result in decreased function of virus-specific cellular and humoral immunity that occurs via upregulation of specific inhibitory receptors expressed on the immune cells. Our data showed that HLA-EUL40-restricted CD8 T cells express lower level of PD-1 as compared to HLA-A*02pp65-restricted CD8 T cells. It has been reported that CD8 T cells expressing high levels of co-inhibitory molecule PD-1 during the chronic infection are characterized by lower proliferation, cytokine production, and cytotoxic abilities [47]. PD-1 plays a significant role in establishment of virus-specific CD8 T-cell exhaustion and has been identified as a major regulator of T-cell exhaustion during chronic HIV/SIV infection [47]. Markedly upregulated on the surface of exhausted virus-specific CD8 T cells, PD-1 expression correlates with impaired virus-specific CD8 T-cell function and with elevated plasma viral load in chronic viral infections [48]. In our study, low levels of PD-1 expression compared to conventional HLA-A*02-restricted CD8 T cells appear as a feature of HLA-E-restricted CD8 T cells. The functional significance of the low PD-1 expression still requires investigations. It could be speculated that low PD-1 level on HLA-EUL40 CD8 T cells may reflect low TCR affinity as recently reported for antigen-specific CD8 T cells targeting melanoma peptides [49]. This feature could be important for homeostatic survival and proliferation to ensure long-term T cell survival [50]. It is interesting to notice that elected tropism of HCMV for endothelial cells also coincides with elevated basal level of HLA-E on this cell type as well as on hematopoietic cells as we previously reported [51]. Basal HLA-E expression can increase upon cellular stress caused by viral infection or heat shock and in inflammatory and cancer cells [41]. It can be speculated that HLA-E-expressing infected ECs play a role as both a trigger and a target of HLA-E-restricted CD8 T cells. We previously demonstrated in vitro the capacity of HLA-EUL40 CD8 T cells to efficiently kill primary allogeneic endothelial cell cultures presenting a homologous HLA signal peptide though HLA-E [30]. Consequently, HLA-EUL40 CD8 T cells could be involved in vascular injury and transplant rejection. The presence of UL40-specific CD8 T cells in the blood of lung transplant recipients was significantly associated with allograft dysfunction, which manifested as Bronchiolitis Obliterans Syndrome [31]. Although deciphering the clinical impact of HLA-EUL40 CD8 T cells was not the focus of the present study, clinical data indicated no significant impact on graft function (serum creatinine and proteinuria) at M12 post-transplantation (Table 1 and S1 Table). This could suggest that although we detected (by tetramer staining or TNF production) a multiplicity of HLA-EUL40 CD8 T-cell responses induced by HCMV and potentially cross-reactive toward a broad set of peptides including self and allogeneic HLAsp, their activation may be either controlled by co-inhibitory receptors or functionally impaired. Another important finding in the setting of organ transplantation also emerges from our work. No HLA-E-restricted CD8 T cells were detected in HCMV- transplant recipients suggesting that allograft does not induce per se consistent HLA-E-restricted CD8 T-cell response against allogeneic (i. e. donor) HLA-EHLAsp complexes as speculated in earlier studies [19]. The function (s) of HLA-EUL40 CD8 T cells still remain to be established in regard to the control of HCMV infection. Efficient lysis of infected cells expressing high levels of HLA-E (i. e. endothelial cells, monocytes) could be a primary function expected for this effector CD8 T-cell subset. Regulatory functions for some HLA-E/Qa-1-restricted CD8 T-cell populations have been well established in mice [52] and more recently in humans [53]. Considering the high expression of HLA-E on CD4 T and B cells [54], a regulatory role for HLA-EUL40 CD8 T cells in the homoeostasis of anti-HCMV cellular immune response cannot be excluded beyond an action on the elimination of infected cells. Moreover, our findings provide evidence for self and allogeneic HLA peptides as potential targets and triggers (for their maintenance) of HLA-EUL40 CD8 T cells supporting effector and regulatory functions for these unconventional CD8 T cells beyond HCMV infection. To conclude, HCMV UL40 induces specific HLA-E-restricted CD8 T-cell response with similar occurrence, magnitude, time course and long term persistence that pp65 viral protein. HLA-A*02 allele and HLA-E genotype are key determinants positively associated with HLA-EUL40 CD8 T cell response. HLA-EUL40 CD8 T cells are effector cells induced by HCMV in a strain-dependent manner that may specifically target and eliminate infected cells. We demonstrated that HLA-EUL40 CD8 T cells also display a potential reactivity toward self and allogeneic HLA peptides that may also contribute to the pathogenicity of HCMV, especially in immunocompromised patients. Banked biological samples (PBMCs and DNAs) were issued from the DIVAT biocollection (CNIL agreement n°891735, French Health Minister Project n°02G55). This retrospective study was performed according to the guidelines of the local and national ethics committees (CCPRB, CHU de Nantes, France). Blood samples collected from anonymous healthy volunteers (n = 25) were obtained from the Etablissement Français du Sang (EFS Pays de la Loire, Nantes) and collected with donor’s specific and written informed consent for research use. A total of 121 patients who underwent kidney (105/121) or kidney-pancreas (16/121) transplantation in our center (Institut de Transplantation/Urologie/Nephrologie, ITUN, CHU de Nantes, France) between 2006 and 2013 were retrospectively enrolled in our cohort study. The cohort includes 4 groups of transplant recipients defined by their HCMV status (Table 1). The groups were defined according to the HCMV serology of the recipient (HCMV- or HCMV+) and for HCMV+ the status of infection (primary, latent, reactivation) at M12 post-transplantation: HCMV active infection (AI) was defined by having at least two consecutive PCR with a viral load (VL) > 3 log10, expressed as number of viral DNA copies (log10cop) per 106 cells. No statistical difference (p > 0. 05) between the groups was found related to the age of the recipients at the day of transplantation, gender ratio, frequency of HLA-A*02 genotype, and the post-transplant time for the blood samples. There is also no statistical difference between the groups concerning the gender ratio of transplant donors, the rank of the transplantation and the duration of cold ischemia. Mismatches of total HLA-I and/or HLA-II for each donor/recipient pairs were equal in the groups. Finally, expected statistical differences between the groups only appeared related to HCMV primary infection status at 12 months post-transplantation. Healthy HCMV+ seropositive blood donors (n = 25) were also recruited in this study. No statistical differences were founded between HV and KTR patients related to age or gender ratio. Frozen PBMCs isolated from blood samples issued from kidney transplant recipients were prospectively stored at the Centre de Ressources Biologiques (CRB, Nantes, France). Cells were thawed 24 hours before use in RPMI-1640 medium (Gibco, Saint Aubin, France) supplemented with 8% human serum, 2 mM L-glutamine (Gibco), 100 U/mL penicillin (Gibco), 0. 1 mg/mL streptomycin (Gibco) and 50 U/mL human recombinant IL-2 (Proleukin, Novartis Pharma, Rueil-Malmaison, France). Blood samples from HCMV+ HV’s were provided by the Clinical Development and Transfer Facility (DTC Facility, INSERM/SFR Federative Structure Research Francois Bonamy, Nantes, France). PBMC were isolated by Ficoll density gradient (Eurobio, Les Ulis, France) and used immediately. For HCMV monitoring, EDTA blood samples were collected for blood donation from healthy volunteers, patient’s follow up or during the acute phase of HCMV infection. HCMV serology was performed using the LIAISON CMV IgG; LIAISON CMV IgM and LIAISON CMV IgG Avidity tests (Diasorin, Saluggia, Italy). Additional evidence of active HCMV replication was examined using an in-house real time HCMV PCR in whole blood, adapted from [55]. The combination of positive CMV IgG, positive IgM, and positive PCR was used for confirmation of primary HCMV infection. For UL40 sequencing, DNA were extracted using QIAsymphony system (Qiagen, Courtaboeuf, France) from 200μL of whole blood samples with the QIAamp DSP DNA Mini Kit (Qiagen). The HCMV UL40 region (858bp) was amplified by PCR using a protocol adapted from [56]. Briefly, the following specific forward and reverse primers were used for a long PCR: forward 5’-TCCTCCCTGGTACCCGATAACAG-3’ and reverse 5’-CGGGCCAGGACTTTTTAATGGCC-3’. Standard reaction mixtures were realized using SYBRPremix Ex Taq kit (Takara Bio Europe, Saint-Germain-en-Laye, France), with the following PCR parameters: one cycle of 94°C for 12 min, then 50 cycles of 94°C 30 sec, 63°C 30 sec and 72°C for 1 min 30, and finally one cycle of 72°C 7 min. PCR products were analyzed by electrophoresis through a 9% non-denaturing acrylamide-bisacrylamide 37. 5–1 gel stained with ethidium bromide. PCR products were purified using the enzymatic method ExoSAP-IT USB (Affymetrix, Thermo Fisher Scientific, Villebon-sur-Yvette, France). Bidirectional sequence was performed using the fluorescent BigDye terminator method (Big Dye version 1. 1 Cycle Sequencing Kit, Applied Biosystems, Courtaboeuf, France) and sequencing reactions were run on Applied Biosystems ABI Prism 3130 XL. Nucleotide and amino acid sequences analyses were performed using Seqscape software (version 2. 5, Applied Biosystems). All sequences were imported and aligned in MEGA5 software using the UL40 sequence from Human Herpesvirus 5 (Merlin strain) as reference sequence (NCBI Reference Sequence: NC_006273. 2). Sequence LOGO were created using the Los Alamos HIV Database tool Analyse Align (http: //www. hiv. lanl. gov/content/sequence/ANALYZEALIGN/analyze_align. html), which was based on WebLOGO3. Banked genomic DNAs (gDNAs) from the transplant donor/recipient pairs (n = 121) analysed in this study and available in the DIVAT biocollection (62 donors and 106 recipients) were harvested. Genomic DNA was extracted from blood samples issued from the HCMV+ HV (n = 25) using usual proteinase K/phenol-chloroform method and subsequently used for genotyping. For HLA-E*01: 01 and HLA-E*01: 03 determination, a first PCR product was generated from gDNA encompassing exon1 to 3 coding for the alpha domains and using the following primers: forward 5' -TCCTGGATACTCATGACGCAGACTC-3’ and reverse 5' -CCTCTTACCCAGGTGAAGCAGCG-3’. Next, a second run of amplification was performed into two separated nested PCR targeting exons 1–2 and exon 3, respectively with the primer pairs: 5-' TCCTGGATACTCATGACGCAGACTC-3’ and 5' -ATCTGGGACCCGAAGATTCGA-3’, 5' -TCGAATCTTCGGGTCCCAGAT-3’ and 5' -CCTCTTACCCAGGTGAAGCAGCG-3’. DNA sequencing was performed with BigDye Terminator v3. 1 kit (Applied Biosystems) according to the manufacturer' s instructions on the DNA Sequencing Core Facility (INSERM/SFR François Bonamy, Nantes, France), using a 48-capillary Applied Biosystems 3730 automatic system (Applied Biosystems). Sequences were analyzed using Chromas 2. 33 software (Digital River GmbH, Shannon, Ireland) using a SNP at AA position 107 to discriminate between *01: 01 and *01: 03 alleles. HLA-A, -B, -C genotypes of transplant donors/recipients pairs and HV were performed by either the EFS (Nantes, Pays de la Loire) or Histogenetics (Ossining, NY, USA), by using PCR-SSO (and completed by PCR-SSP if necessary) and based on the IMGT/HLA database nomenclature (www. ebi. ac. uk/ipd/imgt/hla/). Nine-mers UL4015-23 peptides from 11 different HCMV strains (VMAPRTLIL, VMAPRTLLL, VMAPRTLVL, VMAPRTVLL, VMAPRSLIL, VMAPRSLLL, VMTPRTLVL, VMAPQSLLL, VTAPRTLLL, VTAPRTVLL, VMAPRALLL) and the UL83 pp65495-503 peptide (NLVPMVATV) were synthesized (purity>95%) and purchased from Proteogenix SAS (Schiltigheim, France). HLA-E*01: 01/UL4015-23 (HLA-EUL40) and HLA-A*02: 01/pp65495-503 (HLA-A*02pp65) complexes were generated as described previously [57]. Recombinant HLA proteins were produced in E. coli and refolded with 15μg/mL of each UL4015-23 peptide for HLA-E-monomers or pp65495-503 peptide for HLA-A*02-monomers. Next, HLA-monomers were biotinylated for 4h at 30°C with 6μg/mL BirA (Immunotech, Marseille, France), purified and tetramerized with BV421- or APC-labelled streptavidin (BD Biosciences, Le Pont de Claix, France). Tetramerization was confirmed by gel filtration chromatography (Superdex 200 column, Sigma-Aldrich, Saint-Quentin Fallavier, France). To investigate the frequency of the anti-HCMV CD8 T-cell responses in individuals, PBMC (3x105 per condition) were pre-incubated with a blocking anti-CD94 mAb (clone HP-3D9,5 μg/mL, BD Biosciences) for 20 min at 4°C to completely abrogate the non-specific staining of CD94/NKG2+ T cells by HLA-E-tetramers (S1 Fig). PBMCs were then incubated with one of the different BV421-labelled HLA-E- or HLA-A*02-tetramers (10 μg/mL, 30min, 4°C), before costaining (30min, 4°C) with the following antibodies: anti-CD3 (clone SK7/Leu4, BV786,2 μg/mL, BD Biosciences), anti-CD8α (clone RPA-T8, BV650,0. 1 μg/mL, BD Biosciences) and anti-TCR γδ (clone 11F2, APC-Vio770,3 μg/mL, Miltenyi Biotec, Paris, France). Dead cells were excluded using NucRed Dead 647 ready probes reagent (Life technologies). As a control of tetramer staining, a FMO condition (Fluorescence Minus One; all labelled-markers except one) without tetramers was performed for each sample. Acquisition was performed on a BD LSR II and analyses were performed using BD DIVA Software v6. 0 as described below. Compensations were performed by using anti-mouse κ chain Ab-coated beads (anti-mouse Ig, κ chain/negative control compensation particles set, BD Biosciences) incubated with corresponding Ab at the same concentration during 15 min at room temperature. Data acquisition for the 121 KTR and 25 HV was normalized with application settings based on the KTR#001 patient. Gating analysis strategy was identical for all samples (S1 Fig). To follow-up the development of HCMV-specific T-cell subpopulations in KTR, banked PBMCs from 16 KTR prospectively collected at 1,2, 3,4, 5,6, 7,9, 10,12 and 13 months post-transplantation were used. For each time point tested, UL40-specific HLA-E-restricted (HLA-EUL40) and pp65-specific HLA-A*02: 01-restricted (HLA-A*02pp65) T cells were concomitantly stained and quantified as described above with the complementary mAbs: anti-CD69 (clone FN50, BUV395,2 μg/mL, BD Biosciences) and anti-PD1 (clone EH12 (. 1), PE, 2 μg/mL, BD Biosciences). Acquisition and analysis was performed on a BD LSR Fortessa X-20 with BD DIVA Software v8. 0. Longitudinal samples for each patient were all stained and acquired in the same experiment. HLA-EUL40 T cells were sorted for 5 transplant recipients (#104, #105, #107, #108 and #109) from PBMCs harvested at 12 months post-transplantation as previously described [58]. Briefly, streptavidin-coated beads (Dynabeads M-280 Streptavidin, Invitrogen, Villebon sur Yvette, France) were saturated with HLA-E/UL4015-23 monomers before incubation with PBMCs (5x106) for 4h. The UL4015-23 peptide corresponding to the own HCMV infecting strain was selected for each patient. HLA-EUL40 T cells were isolated by immunomagnetic sorting and then expanded for 21–30 days as follows: cells were seeded in 96-well plates (3x103/well) and stimulated with phytohemagglutinin (1 μg/mL, PHA-L; Sigma-Aldrich) in the presence of irradiated EBV-transformed B-cell lines and allogeneic PBMC from healthy donors (EFS, Nantes) as feeder. Cells were grown in RMPI-1640 medium supplemented with 8% human serum, 2 mM L-glutamine, 100 U/mL penicillin and 0. 1 mg/mL streptomycin and human recombinant IL-2 (150 U/mL). Purity (>95%) of each T cell population was defined after 14 days of culture by tetramer staining. The use of tetramers to activate T cells has been extensively reviewed by Wooldridge and colleagues [59]. T-cell activation by soluble peptide–MHC-I tetramers is very sensitive for inducing a full range of effector functions. In addition to inducing a normal pattern of T-cell signaling [60] tetramer activation results in lytic granule release, a full profile of cytokine and chemokine release and the production of a wide range of cell surface activation markers [61]. In the present study, a series of preliminary experiments were performed to set up the assay measuring T-cell activation in response to HLA-E/UL40 peptide tetramers. Representative results from these preliminary assays are illustrated in the S4 Fig. To determine the peptide specificity of HLA-EUL40-restricted T cells, purified cell lines (1x105 cells /condition) were stimulated for 5h at 37°C in 96-wells plates with one of the 11 HLA-E/UL40-tetramers at 20 μg/mL in RPMI 1640 medium in the presence of Brefeldin A (10 μg/mL, Sigma). Next, cells were incubated with an anti-CD8α mAb (clone RPA-T8,1 μg/mL, BioLegend) for 30 min at 4°C before fixation with 4% paraformaldehyde. After permeabilization with 0. 1% (w/v) saponin (Sigma-Aldrich), cells were stained for 30 min at room temperature with an anti-TNFα mAb (clone cA2, Miltenyi). Cells were finally washed twice in PBS-0. 1% (v/v) BSA and 0. 1% (w/v) saponin before sample acquisition on BD FACS Canto II. Phenotypic analyses were performed on PBMCs from 3 patients. Analysis of T cells before activation was performed ex vivo using the following mAbs: anti-CD3 (clone UCHT1), anti-TCR αβ (clone T10B9. 1A-31/T10B9), anti-TCR γδ (clone B1), anti-CD45RA (clone HI100), anti-CD45RO (clone UCHL1), anti-CD28 (clone CD28. 2), anti-CD27 (clone M-T271), anti-CD57 (clone NK-1) from BD Biosciences; anti-CD8β (clone SIDI8BEE) from eBioscience (Thermo-Fisher); anti-CD4 (clone RPA-T4), anti-CD8α (clone RPA-T8) from Miltenyi and anti-CCR7 (clone 150503) from R&D Systems. For Vβ TCR repertoire analysis, purified HLA-EUL40 T cells (2x105) were incubated 30 min at 4°C in PBS-0. 1% (v/v) BSA with the TCR Vβ Repertoire Kit (IO Test Beta Mark–TCR Vβ Repertoire Kit, Beckman Coulter, Villepinte, France). This kit allows detection of the following Vβ TCR: 1,2, 3,4, 5. 1,5. 2,5. 3,7. 1,7. 2,8, 9,11,12,13. 1,13. 2,13. 6,14,16,17,18,20,21. 3,22 and 23. All Abs were used at saturating concentration conforming to the manufacturer’s recommendation. Data are expressed as medians + interquartile range between Q1 and Q3, or percentages. Appropriate non-parametric statistical analysis (Kruskall-Wallis test, Mann-Whitney, Fischer’s exact test or Pearson’s chi-squared test with adequate post-tests) was performed using GraphPad Prism (GraphPad, San Diego, CA) and R softwares. The type I error rate α (probability threshold of rejecting the null hypothesis given that it is true) was set to 0. 05. A p-value <0. 05 was considered to represent a statistically significant difference.

Understanding the mechanisms of immune control of viral infection is crucial to improve diagnosis and to design efficient immunotherapies. CD8 T lymphocytes are key components of the cellular immunity against human cytomegalovirus (HCMV), a widespread pathogen that cause severe illness and poor outcome in immunocompromised hosts such as transplant recipients and HIV-infected patients. In this study we characterized a population of non-conventional CD8 T lymphocytes directed against the viral protein UL40 and presented by the non-classical HLA-E molecules in blood samples from HCMV seropositive hosts. This immune response was detectable in around 30% of hosts, may represent up to 38% of total blood CD8 T lymphocytes, persists for life and thus seems to belong to the common immune arsenal against HCMV. Genetic factors related to the host and to the different strains of HCMV are critical parameters for the existence of this immune subset. Although specifically induced in response to HCMV infection, a key feature of these cells is their potential ability to be also responsive against multiple HLA molecules. In conclusion, HCMV infection frequently leads to the long-term persistence of a large subset of lymphocytes with potential side effect requiring attention in contexts such as autoimmunity and transplantation.

Abstract Introduction Results Discussion Materials and methods

blood cells urinary system procedures medicine and health sciences organ transplantation immune cells pathology and laboratory medicine pathogens immunology microbiology cytomegalovirus infection cloning surgical and invasive medical procedures viruses dna viruses renal transplantation cytotoxic t cells molecular biology techniques herpesviruses research and analysis methods human cytomegalovirus transplantation immune system proteins infectious diseases white blood cells animal cells proteins medical microbiology t cells microbial pathogens molecular biology biochemistry signal transduction t cell receptors cell biology post-translational modification viral pathogens biology and life sciences cellular types immune receptors viral diseases signal peptides organisms

2018

HCMV triggers frequent and persistent UL40-specific unconventional HLA-E-restricted CD8 T-cell responses with potential autologous and allogeneic peptide recognition

HIV-1 undergoes multiple rounds of error-prone replication between transmission events, resulting in diverse viral populations within and among individuals. In addition, the virus experiences different selective pressures at multiple levels: during the course of infection, at transmission, and among individuals. Disentangling how these evolutionary forces shape the evolution of the virus at the population scale is important for understanding pathogenesis, how drug- and immune-escape variants are likely to spread in populations, and the development of preventive vaccines. To address this, we deep-sequenced two regions of the HIV-1 genome (p24 and gp41) from 34 longitudinally-sampled untreated individuals from Rakai District in Uganda, infected with subtypes A, D, and inter-subtype recombinants. This dataset substantially increases the availability of HIV-1 sequence data that spans multiple years of untreated infection, in particular for different geographical regions and viral subtypes. In line with previous studies, we estimated an approximately five-fold faster rate of evolution at the within-host compared to the population scale for both synonymous and nonsynonymous substitutions, and for all subtypes. We determined the extent to which this mismatch in evolutionary rates can be explained by the evolution of the virus towards population-level consensus, or the transmission of viruses similar to those that establish infection within individuals. Our findings indicate that both processes are likely to be important. Infection by HIV-1 is lifelong, and characterized by ongoing viral replication, and consequently the virus can undergo hundreds of rounds of replication between transmission events. This, combined with error-prone viral replication during reverse transcription, means that the viruses an individual transmits to a recipient are unlikely to be identical to those that initially infected them. A firm understanding of how evolution proceeds during the course of infection within an individual, and how this corresponds to evolution of the virus at the population scale, is therefore required to understand how selection acts at the point of transmission. This is important for vaccine design, and understanding how the virus evolves at the epidemiological scale. To understand the natural history of within-host HIV-1 infection, historical samples from untreated individuals are needed. The few datasets that exist, where multiple (>2) within-host sequenced samples are available, including from early infection, and spanning years rather than months, include nine subtype B individuals from North America [1], ten individuals from Europe, where eight were infected with subtype B, and two were infected with subtypes C and AE, respectively [2,3], and four female individuals from North America infected with subtype B [4]. Here, we add considerably to this small but important body of data by presenting longitudinal deep-sequencing data from 34 untreated individuals living in Rakai, Uganda, representing infection with pure subtypes A and D, and a variety of inter-subtype recombinants, thus expanding both the geographical regions and the viral subtypes for which data are available. As well as undertaking an evolutionary analysis of the within-host sequence data, we also determined the rate of evolution of subtypes A, C, and D at the population scale within Uganda, during approximately the same time period. Interestingly, we found that the nonsynonymous substitution rate in the gp41 region of env is twice as fast for subtype C compared to other subtypes. An indication that evolution at the population scale does not merely represent a continuation of directional within-host evolution, with repeated bottlenecks at the point of transmission, is the observation that HIV-1 evolves about two to six times faster within hosts than at the population scale [5–8] for both synonymous and nonsynonymous mutations [9–11]. Although here we examined different subtypes and gene regions to previous studies, our analysis confirms a similar mismatch in evolutionary rates. Three alternative hypotheses for the mismatch in HIV-1 evolutionary rates have been proposed [11]: The first, called ‘stage-specific selection’, suggests that transmission tends to occur early in infection, when within-host evolution is also suggested to be slow, resulting in slower than expected evolution at the population level [12]. Stage-specific selection is supported by the observation that the rate of evolution at the population level was found to be slower in populations where transmission probably occurs earlier during infection, such as among men-who-have-sex-with-men (MSM) [12]. Although still subject to debate, stage-specific selection is unlikely since cellular and antibody driven escape mutations have been shown to develop rapidly during early infection [13–17], and because stage-specific selection is expected to lead to a greater mismatch for nonsynonymous rather than synonymous substitutions; a pattern that we do not see [11]. Moreover, a more recent study found a faster rate of evolution in MSM than in heterosexual transmission chains [18], suggesting other factors likely explain differing rates of evolution among different groups. The second hypothesis, called ‘adapt and revert’, suggests that within-host evolution is dominated by the ‘reversion’ of mutations that were advantageous in the source host but detrimental in the recipient host [2,11,19–22]. This hypothesis is supported by the observation that within-host evolution is biased towards the accumulation of mutations towards the consensus at the population level [2]. The use of the term of reversion here is contentious because it implies adaptive evolution is going backwards, undoing what was selected for in a previous individual. However, at an individual level, adaptive evolution always goes forwards, thereby selecting for fitter genotypes. It is only when we step back to consider forward within-host evolution in the context of the viruses circulating at the population level that reversion is observed. The final hypothesis suggests that during the course of infection viral lineages that resemble the virus that initiated the infection are maintained, and that these ‘founder-like’ viruses are preferentially transmitted. In this scenario, the cycling of viral lineages through the HIV-1 reservoir maintains these founder-like viruses within the population [6,11,23–27], and therefore this process is referred to as ‘store and retrieve’. This hypothesis is supported by the analysis of HIV transmission chains, in which slower evolving lineages are transmitted [7]; studies of discordant couples showing evidence for the transmission of founder-like viruses [28,29]; variation in the rate of evolution along different within-host lineages [5,30]; and the observation that viruses founding new infections are adapted to early infection, yet under-represented in donor populations [17,31]. Adapt and revert and store and retrieve have largely been presented as either/or scenarios, but these processes need not be mutually exclusive [11]. Moreover, recombination and differing selection pressures in different genomic regions, means their contributions might not be homogeneous across the genome [2,26]. Using longitudinal deep-sequenced data from 34 HIV-1 seroconverters from the Rakai District in Uganda, we tested specific predictions of the adapt and revert and store and retrieve hypotheses, in an effort to quantify their relative contributions. First, we tested whether synonymous and/or nonsynonymous mutations show a strong bias in substitutions towards the population consensus, as predicted by ‘adapt and revert’ [2]. Second, we tested whether a sufficient number of founder-like viruses persist as infection progresses for their preferential transmission to explain the mismatch in phylogenetic rates for synonymous and/or nonsynonymous mutations, as required for store and retrieve [11]. Rather than being a case of either/or, our data suggest that both processes are likely to be important in explaining the mismatch in evolutionary rates. Moreover, our analysis supports previous work that purifying selection can explain why rates of viral evolution decline as the timescales over which they are measured increase from decades to millennia [32–36]. Thirty four HIV-1 seroconverters from the Rakai District in Uganda, previously found not to be superinfected with HIV-1 [37] using deep-sequencing (Roche 454, Pleasanton CA) were selected for further analysis based on sample availability (Table 1). In the previous superinfection screen, early and late samples were sequenced at the p24 region of gag and the gp41 region of env [37]. We sequenced serum samples from three additional time points after HIV-1 seroconversion and prior to initiation of ART (anti-retroviral therapy) using identical methods. These data were combined with the previous sequence data for all subsequent analyses, resulting in longitudinal deep-sequencing data for a 390 base pair (bp) region of p24 and a 324 bp region of gp41. Fig 1 illustrates mean diversity (at 1st, 2nd and 3rd codon positions) and divergence (for synonymous and nonsynonymous changes) over time among the 34 individuals for gp41 and p24. We fitted a linear regression to the diversity and divergence estimates over time among the 34 individuals. Diversity at third codon positions accumulated at similar rates in p24 and gp41, but diversity at the first and second positions accumulated much more slowly for p24 compared to gp41 (Fig 1). This is consistent with stronger purifying selection acting upon p24 and stronger diversifying selection acting upon gp41. This is further supported by the divergence patterns. In p24, nonsynonymous and synonymous substitutions accumulated at approximately similar rates (Fig 1; bottom left panel). In contrast, for gp41 notably greater nonsynonymous divergence than synonymous divergence was observed (Fig 1; bottom right panel). We further examined the diversity and divergence patterns by excluding nine individuals, which were likely infected with multiple viral variants (defined as > = 0. 02 mean pairwise diversity across all sites in the p24 gene region at the first time point). This may not identify all individuals where multiple viral variants were transmitted, particularly if these variants are very similar or where one variant has been lost. Nevertheless, no discernible impact on mean diversity and divergence trends was observed when these individuals were excluded (S1 Fig). Although diversity and divergence tended to increase as infection progressed, we observed considerable variation among individuals, with diversity notably not always increasing between subsequent time points (S2 Fig). To some extent, this reflects stochastic error associated with short gene regions and the 454-sequencing method (S2 Fig). In particular, biases in genome amplification during PCR are likely to result in diversity being underestimated. However, within-host evolutionary dynamics probably also has an important role. One individual (i24) had very high diversity in the gp41 gene region at the third codon position at the first sampling time point, which subsequently declined over time (S2A Fig; bottom right panel). Inspection of the maximum clade credibility trees (summary of the posterior tree distribution from BEAST) indicates this individual was infected by at least two different viral variants, which were very similar in p24, but very different in gp41 (S3 Fig). This is reflected in the low posterior support for the two infecting lineages in the p24 tree, but high posterior support (0. 82–1. 0) for multiple infecting lineages in the gp41 tree, which were maintained throughout infection (S3 Fig). The fall in diversity at the third codon position (S2 Fig) between the penultimate and last time points for i24 corresponds to one of the lineages predominating among the sampled viruses (S3 Fig). This probably reflects ongoing within-host competition between the two lineages, with eventually one lineage prevailing. A similar pattern of falling diversity is seen in the p24 gene region of i17, where a fall in diversity at the last sampled time point (1919 days) coincides with the loss of one of the within-host lineages (S4 Fig). Given sequence similarity at the first time point, the multiple lineages observed in i17 probably emerged during infection rather than due to coinfection. In the gp41 region of this individual we also see a drop in diversity between days 828 and 923. This possibly reflects the complex within-host evolutionary dynamics inferred by the MCC tree of this individual (S4 Fig), although since this individual had a relatively low SPVL (Table 1), amplification errors during sequencing could affect the estimates of diversity. Next we looked at absolute nonsynonymous and synonymous substitution rates at the within-host level. Estimated rates of within-host viral evolution varied considerably among individuals (Fig 2), and were consistent with previous measures of within-host rates of viral evolution for subtype B infected individuals in gag [4] and the gp120 region of env [4–6]. The comparatively large uncertainty in the estimates of within-host evolutionary rates are most likely due to the short gene regions analyzed and lower rates of evolution compared to gp120. Regardless, a significant mismatch was observed between the within- and between-host evolutionary rates in both gene regions for subtypes A, C, and D (Fig 2). Note that for recombinant infections, individuals have different subtypes at each of the two regions (Table 1). The mean ratios of the within- and between-host rates ranged from 3. 0 to 8. 8, and were similar for nonsynonymous and synonymous substitutions, and did not differ by gene region or subtype (S1 Table), although some of these estimates were associated with large uncertainties (standard deviation ranged from 1. 1 to 5. 9). It has previously been argued that higher rates of nonsynonymous and synonymous substitution at the within-host level compared to the between-host level is compatible with store and retrieve, but not stage-specific selection or adapt and revert [11], since under the latter two hypotheses a mismatch is only expected for nonsynonymous substitutions. However, this assumes that synonymous substitutions experience lower levels of selection compared to nonsynonymous substitutions, which might not always be the case due to their effects on RNA secondary structure [38]. In line with the patterns observed for divergence over time, a greater nonsynonymous substitution rate was noted for gp41 than for p24 among the different subtypes, whereas synonymous substitution rates were comparable (S5 Fig). These observations are consistent with gp41 being subjected to stronger diversifying selection than p24, together with p24 undergoing comparatively greater purifying selection. We see similar patterns in the between-host evolutionary rates (Fig 2). Notably, subtype C was characterized by the highest overall substitution rate in gp41 (~0. 002 subs/site/year), which appears to be driven by a comparatively higher nonsynonymous substitution rate (Fig 2 and S1 Table). This elevated rate for subtype C is in line with a previous study, but where substitution rates were estimated from whole-genome sequences that were sampled from a broader geographic distribution [39]. To corroborate the within-host evolutionary estimates using the renaissance counting method and hierarchical phylogenetic model, we performed three sets of auxiliary analyses in BEAST for a subset of ten individuals consisting of five pure subtype A and five pure subtype D infections, i. e. for a given HIV-1 infection, p24 and gp41 gene regions corresponded to the same subtype. Furthermore, based on mean diversity at the first time point in the p24 gene region, these individuals were unlikely to have been infected by multiple, genetically distinct strains. The results from the sensitivity analyses are briefly discussed here, but a full description can be found in S1 Text. First, we found a strong correspondence in the evolutionary rates for the ten individuals from the original analysis (based on 34 individuals) and the new analysis, which was based on ten individuals with pure subtype infections (S6 Fig). This strongly suggests that individuals infected with multiple variants have not impacted the evolutionary rate estimates for individuals infected with single viral variants. Next, we estimated evolutionary rates for the subset of individuals using a full codon substitution model [40], which were in good agreement with estimates from the renaissance counting method (S7 Fig). Finally, we examined the robustness of the evolutionary rate estimates for the ten individuals to different hyperpriors in the clock hierarchical phylogenetic model (see S1 Text for more details). The evolutionary rate estimates were found to be very similar across different hyperpriors (S8 Fig), strongly suggesting that the posterior estimates of the within-host evolutionary rates in Fig 2 are mostly informed by the sequence data and are robust to the choice of hyperprior in the hierarchical phylogenetic model. In a previous analysis comparing within-host rates of evolution with set-point viral load (SPVL), Lemey et al. observed a significant correlation between the synonymous substitution rate of the C2V5 region along the backbone branches (ancestral internal branches in the phylogeny that have given rise to the most recently sampled sequences) and the rate of disease progression [10]. Specifically, individuals with slower disease progression had lower rates of synonymous substitution than individuals with faster disease progression. Since SPVL is a strong predictor of disease progression in HIV-1 infection, we examined the correlation between the within-host evolutionary rate and SPVL to determine whether a similar relationship exists between synonymous substitution rate and disease progression in the Rakai cohort. In contrast to Lemey et al. , we found no significant correlation between the mean absolute synonymous or non-synonymous substitution rate (on either external, internal, or backbone branches) and SPVL (S9 Fig; S3 Table). We also tested for a subtype-specific effect, but again, we did not find a significant correlation. The lack of a significant correlation observed here could be because there is no link between substitution rate and SPVL among the Rakai individuals, or because of uncertainty in the substitution rate and/or SPVL estimates. In particular, the p24 and gp41 gene regions have low rates of evolution compared to the C2V5 region, which is typically associated with an order of magnitude higher rate of evolution (see Fig 3 in [2]). A prediction of adapt and revert is that, of sites where evolutionary change is observed, the accumulation of mutations should be biased towards the population consensus as infection progresses [2]. Limiting our analysis to sites within individuals where the most common (founder) allele at the first time point was at or close to fixation (frequency >0. 99), we defined polymorphic sites as those where a different (mutant) allele had reached a frequency of >0. 1 at some point during the sampling period. Across all individuals, between 10 and 30 percent of the mutant alleles observed at polymorphic sites represented changes towards population consensus in p24 and gp41 (Fig 3, top row). Moreover, there was a strong bias, with changes towards population consensus at polymorphic sites occurring approximately 1. 5 to 2 times more often than expected in the absence of selection (Fig 3, bottom row). Here, we assumed a mutational transition to transversion ratio of two, representing a biochemical preference for transitions. The broad pattern remains the same if we do not assume a preference for transitions, although the calculated bias is higher (S10 Fig). Taken together, these observations can explain a substantial mismatch in evolutionary rates, in broad agreement with a similar analysis using deep-sequencing data from nine European individuals [2]. In addition, if viruses that are more similar to the population consensus are preferentially transmitted [41], an even higher mismatch in evolutionary rates could be explained. We next considered nonsynonymous and synonymous changes separately. A high proportion of nonsynonymous changes (mean proportion of 0. 45 and 0. 17, respectively, for p24 and gp41, at 0–2 years post seroconversion) were towards population consensus (Fig 3, top row). Furthermore, at polymorphic sites, these nonsynonymous changes were associated with a strong bias, such that changes towards population consensus occurred twice as often as expected (Fig 3, bottom row). Considering all sites close to fixation at the first sampling time point (not just polymorphic sites), and assuming mutations are equally likely at all sites, nonsynonymous changes continued to show a strong bias towards population consensus (S10 Fig). Together, these observations provide strong evidence for the role of adapt and revert for nonsynonymous substitutions, particularly in p24, which is consistent with stronger functional constraints being present in p24 compared to gp41. A reasonably high proportion of synonymous changes were also towards population consensus (mean proportion of 0. 10 and 0. 11, respectively, for p24 and gp41, at 0–2 years post seroconversion; Fig 3, top row). This was accompanied by a small bias at polymorphic sites in p24, but there was no detectable bias in gp41 (Fig 3, bottom row). The absence of a strong bias at polymorphic sites suggests adapt and revert contributes little to the mismatch in evolution rates for synonymous changes when comparing rates at the within-host and population scales. Considering all sites, however, a bias towards population consensus was observed for synonymous changes in both gene regions, although this bias was much smaller than for nonsynonymous changes (S10 and S11 Figs). In other words, of all the changes that could occur (the vast majority of which will be away from population consensus, since most sites are at population consensus), and assuming mutation is equally likely at all sites, synonymous changes towards population consensus are much more likely than would be expected by chance. The detection of this bias for synonymous changes, when all sites are considered, provides evidence that a high proportion of synonymous changes are non-neutral, because, for example, they affect RNA secondary structure [38]. This is consistent with the observation that across the whole subtype B HIV-1 genome, synonymous mutations that occur away from population consensus during infection are often weakly deleterious, but with a significant proportion (~10% outside of env) being highly deleterious [42]. Evolution towards population consensus for synonymous changes, when all sites are considered, therefore most likely represents weak purifying selection, possibly exacerbated by the transmission of viruses harboring slightly deleterious mutations due to tight bottlenecks at transmission [33]. Assuming virus is not transmitted directly from the reservoir, a key prediction of store and retrieve is that founder-like viruses should be circulating in the viral population at sufficient frequencies, with a mismatch in evolutionary rates occurring if these founder-like viruses are preferentially transmitted. We determined how ‘founder-like’ a virus is within an individual by the number of mutations, d, the virus differs from the consensus virus (es) circulating at the first sampled time point. Thus, we are essentially using a small portion of the genome (p24 or gp41) as a surrogate for the whole genome. We next assumed that each viral sequence has a transmission fitness wd = e−α d, where α determines how rapidly transmissibility declines as sequences evolve away from the first time point consensus sequence (s), under the assumption that d is representative of how founder-like the whole genome is. The predicted contribution to the mismatch in evolutionary rates of an individual at a given sampling time point is then given by the mean value of d in the viral population at that time point, divided by the expected mean value of d in the transmitted viral population (see Methods). For both gene regions, a large mismatch in evolutionary rates can be explained if founder-like viruses have a strong transmission advantage (α ~ 2), and if transmission tends to occur during the first few years of infection (Fig 4). This pattern remains if we remove individuals likely infected by multiple variants (S12 Fig). As infection progresses, transmitted viruses are predicted to contribute less to a mismatch in evolutionary rates due to the gradual loss of founder-like viruses. A mismatch is still predicted if the data are partitioned between synonymous and nonsynonymous mutations (assuming selection at transmission is determined by the total number of mutations), although the predicted mismatch is generally less for nonsynonymous mutations. This is presumably because nonsynonymous mutations are subject to stronger within-host selection, and therefore founder-like variants are less likely to be preserved as infection progresses. When interpreting these results, it is important to acknowledge the role of recombination, which for within-host viral populations has been shown to limit linkage disequilibrium to about 100–200 bps [2]. However, if founder-like viral lineages are maintained during infection (because they have spent a long time in the reservoir where neither error-prone replication nor recombination occur) linkage across much longer regions, and possibly the whole genome, is expected for these lineages [26]. Because we do not have linkage information between sequences in p24 and gp41, we were unable to determine whether viruses that harbor founder-like p24 sequences also harbor founder-like gp41 sequences, as would be expected if the preferential transmission of founder-like viruses explains the mismatch observed in both regions, and more generally the mismatch observed across the whole genome [8]. Thus, although our observations are consistent with store and retrieve, due to short read lengths (390 bp for p24 and 324 bp for gp41), and relatively low rates of evolution for these two gene regions, we have insufficient power to test whether these founder-like viral variants are maintained because of cycling of lineages through the viral reservoir, or simply due to the stochastic nature by which mutations are accumulated along lineages in these two short gene regions. To resolve this question, longitudinal, long-read deep-sequencing data is needed. Using cryopreserved samples from individuals longitudinally sampled before the availability of universal treatment in the Rakai District of Uganda, we have substantially increased the number of individuals for which deep-sequenced data is available for HIV-1 during the course of untreated infection. The main subtypes represented in this HIV-1 cohort are A and D, rather than subtype B, which predominates in the more frequently studied European and North American cohorts. As well as analyzing this within-host sequence data, we also utilized publicly available population consensus sequences from Uganda, enabling us to directly compare rates of viral evolution at both the within-individual and population scales in the same population and for the same regions of the genome. It is notable that the virus evolves approximately three to nine times faster at the within-host than at the population scale. This pattern was observed for all three subtypes in both gene regions and for nonsynonymous and synonymous substitutions. These estimates are consistent with previous estimates for nine subtype B infected individuals for the gp120 region of env [1,5, 6], and for a subtype B infected individual measured across the whole genome [8]. Together with the observation that within-host viral lineages leading to transmission events evolve approximately half as fast as other lineages [7], these findings build a consistent picture of different rates of evolution for HIV-1 within- and between-hosts, for all of the subtypes that have been analyzed, and across the whole genome. Intriguingly, similar mismatches in evolutionary rates are observed for HIV-2, Hepatitis B, and Hepatitis C viruses [27,43–47], leading us to speculate that such mismatches are a general feature of rapidly evolving chronic viral infections in humans. We tested specific predictions of two of the mechanisms that have been implicated as contributing to the mismatch in evolutionary rates in HIV-1: adapt and revert and store and retrieve, with the aim of quantifying their relative roles. For nonsynonymous changes, our results are consistent with both adapt and revert, and store and retrieve, contributing to the mismatch in evolutionary rates. For synonymous changes, on other hand, our results are consistent with store and retrieve contributing to the mismatch in evolutionary rates, but with adapt and revert contributing little (p24) if at all (gp41). We conclude that both mechanisms are likely to have important roles, but that these differ for synonymous and nonsynonymous substitutions, and, given the differences seen between p24 and gp41 (Figs 3 and 4), our data suggests their relative contributions also differ across genome. Here, we have focused on the mismatch in evolutionary rates when within individual and population level rates are compared, with both rates measured across short timescales of years to a few decades. It is now well recognized that rates of viral evolution at the population level also decline as the timescales over which they are measured increase from decades to millennia [32,33]. This is often attributed to a combination of purifying selection, with the appearance and persistence of slightly deleterious mutations (independent of host genotype) over short time scales, but their eventual purging over longer time scales; and saturation effects, which are expected to be pronounced in RNA viruses due to their short genomes and high mutation rates [32–36]. Here, we also detected patterns of evolution within individuals that are consistent with purifying selection, specifically the purging of transmitted slightly deleterious mutations, which may contribute to the slowing of measured evolutionary rates at the population level as progressively longer timescales are considered. It is also possible that adapt and revert, and store and retrieve, might provide additional mechanisms leading to this slowing of measured evolutionary rates [7], but further work is needed to assess their likely importance. A unique feature of our analysis is the number of individuals included. This makes our overall analysis more robust than those based on fewer individuals, and also highlights the heterogeneity in patterns observed among individuals as well as between the two gene regions we looked at. There is ongoing interest in trying to estimate the number of variants initiating HIV infections [15,48–53]. Our analysis highlights that focusing on a single gene region can potentially be misleading. For example, diversity measurements for individual i24 indicate infection by a single variant when looking at the p24 gene region, but multiple variants when looking at the gp41 region. The most likely scenario is that the donor individual (or i24 before the first sampling timepoint) was superinfected by a distinct variant from an unknown individual, followed by recombination in either the donor or i24, which led to the maintenance of two distinct lineages in gp41 but not p24. Similarly, there is interest in estimating time since infection from measures of within-host viral diversity [1,3, 54–57]. However, diversity is not always a good measure, as it can be elevated for substantial periods of time due to the persistence of multiple founder lineages, as seen in individual i24, and can drop dramatically as a consequence of within-host population dynamics, as likely seen in individuals i24 and i17, although amplification biases cannot be ruled out. The continual adaptation of HIV-1 to different host environments and selection at the point of transmission are both likely to contribute to the complex patterns of HIV-1 evolution observed at the within-individual and population levels. Moreover, different selection pressures acting across the genome coupled with high rates of recombination further complicate the picture [26]. In particular, recombination is likely to elevate within-host evolutionary rates as a result of generating more diverse viral lineages. Disentangling the evolutionary pressures faced by chronic viruses will not only help us to understand how selection acts across multiple ecological scales [27], but will also have direct clinical importance by shaping our understanding of pathogenesis [58], how drug- and immune-escape variants are likely to spread through populations [41,59,60], and in the development of preventive vaccines [61]. Deeply sequenced, whole-genome reads that avoid PCR generated recombination and amplification errors will be needed to fully delineate the relative roles of all of these pressures across the genome. HIV-1 seroconverters participating in the Rakai Community Cohort study who were co-enrolled in the Molecular Epidemiology Research (MER) seroconverter study were previously screened for HIV-1 superinfection [37,62]. HIV-1 seroconversion date was estimated as the midpoint between the last seronegative and the first seropositive sample as tested in the Rakai Community Cohort study. Apart from i11, the first sequenced sample corresponds to the first positive sample (for i11, the first positive sample was 46 days earlier than the first sequenced sample). Individuals who had both gp41 and p24 regions sequenced for both time points previously screened, were not superinfected, and who had at least three additional serum samples available as part of the MER study were included in the study (Table 1). Previously generated sequences were used for this analysis [37]. In addition, serum samples from the three additional study time points were analyzed using identical next generation sequencing (NGS) methods, as described previously [37,63]. Briefly, HIV-1 RNA was extracted from 140μL plasma, reverse-transcribed, and amplified using a nested-polymerase chain reaction (PCR) to produce amplicons corresponding to portions of the viral p24 (~390 bp) and gp41 (~324 bp) gene regions. The corresponding HXB2 reference genome positions for p24 and gp41 used in this study are 1429–1816 and 7941–8264, respectively. Successfully amplified samples for both study visits (baseline and follow-up) in at least one region were sequenced using the 454 DNA Sequencing platform as previously described, with adjustments to use a 2-region format (Roche, Branford, CT) [37,62,63]. Pools of samples were processed using emPCR Amplification Manual-Lib-L-LV–June 2013 (Roche Branford, CT) using 25% of the recommended amplification primer amount and a 0. 2 copy-per-bead ratio [63]. The resulting sequencing reads were analyzed and similar sequences were combined into a single consensus sequence. The number of reads and consensus sequences for each sample in the study, plus viral load and CD4 counts where available, are shown in S2 Table. Short sequence reads (>10 bp short of the individual consensus) were removed, and consensus sequences that encompassed a cluster of at least ten individual, near-identical sequence reads were determined and used for all subsequent analyses [37,63]. In order to remove any residual contaminating sequences a representative sequence from all distinct viral populations for each sample run in a given NGS sequencing plate were combined in a neighbor-joining tree, and any micro-contamination or spill-over sequences that localized with another unrelated sample were removed. A final manual alignment of the sequences was performed to ensure the sequences aligned within and across the different individuals for all time points, and gaps were inserted where necessary to keep the reads in-frame. One or two base-pair insertions associated with homopolymeric tracks were removed. This realignment typically reduced the number of distinct consensus sequences at each time point because the position of gaps in the sequences was standardized. Through this procedure, most errors associated with 454 sequencing of HIV were corrected for, particularly indels associated with homopolymeric regions [64]. To reduce the impact of substitution errors introduced through 454 sequencing, we excluded in our evolutionary analyses sites within individuals where the second-most frequent allele frequency was <0. 056% (the estimated error-frequency per nucleotide [64]), under the assumption that polymorphisms at these sites were due to sequencing error. We also checked the resulting alignments for recombination using RDP4 [65], which did not detect any recombination breakpoints. Estimating the founder strain (s) that initiated each infection is challenging because the first sampling time point for each of the individuals in our study is estimated to be between 150 and 425 days since seroconversion. A common approach is to use the consensus sequence at the first sampled time point as a proxy for the founder strain. However, if an individual was infected by multiple strains, this can give misleading estimates and add considerable noise to the data. To help remediate this effect, we identified genetically distinct subgroups at the first sampling time point using hierBAPS (Hierarchical Bayesian Analysis of Population Structure) [66]. The consensus sequence of each of these subpopulations was then determined, with these representing our proxies of the founder strain (s). We note that although sufficient for our analysis, this is not a good method to determine the actual number of founder strains in our data, with some consensus sequences from the same individual differing by only a single base in our analysis. For gp41, six individuals had more than one consensus sequence, and for p24, eight individuals had more than one consensus sequence. Diversity and divergence over time were calculated on the full sequence data using custom-made Python scripts, which are available on github (https: //github. com/katrinalythgoe/RakaiHIV). For divergence, we estimated the mean pairwise genetic difference at each time point between each viral gene sequence from that time point and the consensus sequence (s) from the first time point. In cases where multiple consensus sequences were estimated, we inferred the most closely related ancestral sequence by only considering the minimum pairwise genetic difference for each sequence against the available consensus sequences. Diversity corresponded to the mean pairwise genetic differences among the sequences sampled at a particular time point. For both diversity and divergence, we only considered sites with minor allele frequency greater than 0. 056% (the estimated error-frequency per nucleotide from [64]). We estimated the within- and between-host evolutionary rates using BEAST [67] by employing a hierarchical phylogenetic model (HPM) [68] and a renaissance counting approach [69]. For the within-host evolutionary analysis, this approach has been shown to yield more precise estimates (e. g. [70]), as it enables information about the evolutionary parameters (e. g. substitution model and molecular clock) to be explicitly shared among the different individual datasets while allowing independent evolutionary histories for each individual. Specifically, renaissance counting is a probabilistic counting method for estimating nonsynonymous and synonymous substitution rates and site-specific dN/dS ratios using codon-partitioned nucleotide substitution models. It is based on a stochastic mapping approach, which infers the changes (or counts) at each site in the alignment across the phylogeny using a continuous-time Markov chain (CTMC) model of nucleotide substitutions, and empirical Bayes modeling to avoid inflated standard errors of the number of substitution counts, namely by excluding counts that are either zero or infinity. Site-specific dN/dS ratios can be estimated by dividing the observed nonsynonymous (cN) and synonymous substitutions (cS) with expected nonsynonymous (uN) and synonymous changes (uS), e. g. dN/dS = (cN/cS) / (uN/uS). As this method is implemented in a Bayesian phylogenetic framework, estimates of nonsynonymous and synonymous substitution rates also take into account phylogenetic uncertainty. Furthermore, it compares well with methods that use codon substitution models, with the advantage that it is more computationally efficient. For our analysis, we first estimated posterior tree distributions for each individual (for both gene regions), using a codon-structured nucleotide substitution model [71], a strict molecular clock, and a constant tree prior, and applied noninformative hierarchical priors on the substitution, clock, and population parameters. These were subsequently used as empirical tree distributions for the renaissance counting analysis, where noninformative hierarchical priors were similarly employed for all evolutionary parameters. To reduce the computational burden of the within-host evolutionary analysis, we used a subsampled dataset for each individual, where 25 sequences per time-point were randomly selected for each gene region. The final dataset comprised of 8100 sequences where each gene-specific individual dataset ranged from 75 to 125 sequences. For the BEAST sensitivity analyses we used the CIPRES Science Gateway [72]. To estimate the between-host evolutionary rates, we collated independent datasets for subtypes A, C and D HIV-1 infections from Uganda using the HIV LANL database. The sequences were subsequently randomly sampled, resulting in approximately 200 sequences in each dataset. For the subtype C dataset, there were fewer sequences available from Uganda (specifically, 43 and 90 respectively for p24 and gp41 gene regions). However, these datasets were considered to have sufficient temporal structure (along with subtypes A and D), as evaluated by root-to-tip regression method in TempEst [73]. Furthermore, the viral gene sequences corresponded to the same gene regions used in the within-host sequencing study. For this analysis, we employed BEAST using a codon-structured substitution model [71], uncorrelated log-normal distributed molecular clock [74], and a Bayesian skygrid prior [75]. Nonsynonymous and synonymous substitution rates were estimated using a similar approach outlined for the within-host evolutionary analysis. SPVL was calculated using similar criteria to [76], by taking the mean log10 viral load from all visits where viral load measurements were available, which were more than 6 months after the estimated date of seroconversion, and before the initiation of antiretroviral therapy or the onset of AIDS. This included viral load measurements taken from additional visits to those for which sequence data is available. The first viral load measurement from three individuals (i6, i15, and i20) was excluded because they were more than ten times higher than all subsequent measurements, indicating these individuals were in acute infection at the time (i. e. seroconverted soon before the first seropositive sample, rather than at the mid-point between the last seronegative and first seropositive samples, as assumed). The mean within-host evolutionary rates among the external, internal, and backbone branches for each individual were estimated from a subset of 500 posterior trees using a custom-made script in Java (https: //github. com/katrinalythgoe/RakaiHIV), which depends on the Java Evolutionary Biology Library available from https: //sourceforge. net/projects/jebl/. These estimates have been summarized in S3 Table. In line with Lemey et al. [10], association between evolutionary rate and SPVL was examined with a Pearson correlation test at the 5% significance level. We first calculated the population consensus sequences for subtypes A, D and C using the same sequences used to calculate the between-host evolutionary rates. For each of the 34 individuals in our study, we limited our analysis to sites that were fixed or nearly fixed for a single base at the first time point (>99% frequency). Of these sites, we defined them to be polymorphic for a given sampling time period (between 0 and 2 years since seroconversion; between 2 and 4 years since seroconversion; or over 4 years since seroconversion) if a mutation had reached an appreciable frequency (>10%) at least once during that period. For each sampling time period, we pooled data across all individuals and calculated the proportion of the changes at polymorphic sites that were towards the subtype-specific consensus for all mutations, synonymous mutations and nonsynonymous mutations (Fig 4). Additionally, we calculated the expected proportion of mutations towards the subtype-specific population consensus at polymorphic sites, assuming no selection, and a transition to transversion ratio of 2, thus accounting for a higher number of transitions in the absence of selection [77]. The bias towards population consensus was then calculated as the proportion of mutations towards subtype-specific population consensus, divided by the expected proportion of mutations towards population consensus (Fig 4). In S6 Fig, we also show the bias towards consensus for polymorphic sites assuming different transition to transversion ratios (0. 5 and 4), and for all sites (not just polymorphic sites). In addition, we calculated the proportion of mutations towards population consensus at sites that were not at population consensus at the first time point (S6 Fig). For each sampled time point, we calculated the expected contribution to the mismatch in evolutionary rate, conditional on transmission occurring, for all, synonymous and nonsynonymous mutations. Using a similar reasoning to [11], we assumed that each sequence has a transmission fitness wd = e−α d, where α determines how rapidly transmissibility declines as the distance from the consensus sequence (s) from the first sampling time point, d, increases. Where multiple consensus sequences were inferred, we assumed the ancestor to a given sequence was the genetically most similar one, including both synonymous and nonsynonymous mutations. The contribution to the mismatch was then calculated as the ratio of the mean number of mutations from the founder population, μX, to the expected mean distance of transmitted virus from the founder population, μTX, giving mX = μX/μTX. Here, X, refers to all, synonymous, or nonsynonymous mutations. Letting nX, d, δ represent the number of sequences distance d from the appropriate consensus sequence and that also harbor δ all, synonymous or nonsynonymous mutations, we can calculate μX = (∑d, δ δ nX, d, δ) / (∑d, δ nX, d, δ), and μTX = (∑d, δ δ wd nX, d, δ) / (∑d, δ wd nX, d, δ). When calculating the mean mismatch for a given time interval, for each individual we chose the mean mismatch calculated for all the sampled time points within that interval, to avoid the frequency of sampling from biasing the results. This project used stored samples from the Rakai Community Cohort Study in Uganda. All subjects involved in the study were adult and provided written informed consent for their samples to be stored and used for future unspecified HIV-related research. The study was approved by the Science and Ethics Committee of the Uganda Virus Research Institute, the Western Institutional Review Board, and the Committee on Human Research at the Johns Hopkins Bloomberg School of Public Health. All samples were anonymised.

The speed at which HIV-1 evolves within individuals and across epidemics is substantially different. Identifying the mechanisms shaping this phenomenon has important implications for understanding disease severity and transmission of HIV-1, especially when considering the spread of viruses that are resistant to drugs or natural host defenses in the population. In this study, we analyze newly generated HIV-1 sequences sampled from 34 individuals living in Uganda over multiple years, together with publicly available HIV-1 sequences from Uganda at the population scale. Our findings indicate that HIV-1 evolves around five times faster within individuals compared to the population scale. We demonstrate that there are likely two key processes driving the difference in HIV-1 evolutionary rate at the within individual and population scales. Specifically, we find support for 1) selection against variants in the within-host viral population, which were acquired in the donor because they were beneficial in that individual, and 2) preferential transmission of variants that are similar to those that initiated the infection.

Abstract Introduction Results Discussion Methods

organismal evolution medicine and health sciences pathology and laboratory medicine viral transmission and infection pathogens geographical locations microbiology uganda retroviruses viruses immunodeficiency viruses rna viruses microbial evolution viral load africa research and analysis methods sequence analysis evolutionary rate lentivirus bioinformatics medical microbiology hiv microbial pathogens hiv-1 evolutionary genetics viral evolution people and places virology viral pathogens database and informatics methods biology and life sciences evolutionary biology evolutionary processes organisms

2018

Evolution of HIV-1 within untreated individuals and at the population scale in Uganda

The intricate shaping of the facial skeleton is essential for function of the vertebrate jaw and middle ear. While much has been learned about the signaling pathways and transcription factors that control facial patterning, the downstream cellular mechanisms dictating skeletal shapes have remained unclear. Here we present genetic evidence in zebrafish that three major signaling pathways − Jagged-Notch, Endothelin1 (Edn1), and Bmp − regulate the pattern of facial cartilage and bone formation by controlling the timing of cartilage differentiation along the dorsoventral axis of the pharyngeal arches. A genomic analysis of purified facial skeletal precursors in mutant and overexpression embryos revealed a core set of differentiation genes that were commonly repressed by Jagged-Notch and induced by Edn1. Further analysis of the pre-cartilage condensation gene barx1, as well as in vivo imaging of cartilage differentiation, revealed that cartilage forms first in regions of high Edn1 and low Jagged-Notch activity. Consistent with a role of Jagged-Notch signaling in restricting cartilage differentiation, loss of Notch pathway components resulted in expanded barx1 expression in the dorsal arches, with mutation of barx1 rescuing some aspects of dorsal skeletal patterning in jag1b mutants. We also identified prrx1a and prrx1b as negative Edn1 and positive Bmp targets that function in parallel to Jagged-Notch signaling to restrict the formation of dorsal barx1+ pre-cartilage condensations. Simultaneous loss of jag1b and prrx1a/b better rescued lower facial defects of edn1 mutants than loss of either pathway alone, showing that combined overactivation of Jagged-Notch and Bmp/Prrx1 pathways contribute to the absence of cartilage differentiation in the edn1 mutant lower face. These findings support a model in which Notch-mediated restriction of cartilage differentiation, particularly in the second pharyngeal arch, helps to establish a distinct skeletal pattern in the upper face. Morphogenesis of the facial skeleton in zebrafish is tightly linked with the early differentiation of pharyngeal arch neural crest-derived cells (NCCs) into cartilage. Shortly after migration into the pharyngeal arches, NCCs form a series of pre-cartilage condensations that prefigure the distinct shapes of the later cartilage-replacement bones. As near-isometric growth of these cartilages during the later larval period largely preserves these initial shapes [1], early patterning, not later growth, is the major determinant of facial skeletal shaping. Identifying the local signals that sculpt and arrange early condensations in specific regions of the developing arches is therefore critical to understanding how the facial skeletal bauplan is established. Genetic studies in a wide range of vertebrates has revealed that patterning of arch NCCs along the dorsoventral axis is an important early step in regionalization of the face, with ventral (distal) cells generating the lower jaw and hyoid bone, maxillary cells forming the upper jaw, and more posteriorly located dorsal (proximal) cells making the lateral upper face. These dorsoventral domains are established in large part by interactions between dorsal Jagged-Notch, ventral/intermediate Endothelin1 (Edn1), and ventral Bmp signaling. Mutation of Edn1 signaling components and key downstream targets (e. g. Dlx5/6) in mice and zebrafish results in homeotic transformations and/or losses of skeletal elements derived from the ventral and intermediate domains of the arches, such that the lower jaw adopts an ectopic upper jaw morphology [2–12]. Downregulation of Bmp signaling results in a similar loss of ventral arch-derived structures in mice and zebrafish [13–16], whereas loss of the Notch ligand jag1b in zebrafish conversely affects bones and cartilages of the upper/dorsal face, particularly those from the second arch and the dorsal-posterior region of the first arch [17]. These pathways are actively antagonistic: Edn1 and Bmp signaling prevent jag1b expression in ventral/intermediate cells, Notch signaling blocks the expression of Edn1 target genes dorsally (e. g. dlx3b/5a/6a, msxe, nkx3. 2) [17], and Jagged-Notch and Edn1 signaling limit Bmp activity to the most ventral arches in part through upregulation of the Bmp antagonist Gremlin2 in the intermediate domain [13,14]. The end result of these interactions is the establishment of a distinct dorsal domain (excluding the anterior/maxillary region of the first arch, which is not patterned by Notch [17]) and the subdivision of an initial ventral arch region into distinct ventral/lower and intermediate regions [14]. How this dorsoventral patterning is translated into region-specific cartilage shapes has, however, remained unresolved. Previous microarray studies of dissected arches in mice lacking the key Edn1 target genes Dlx5/6 [18] or overexpressing Bmp4 [16] revealed a number of misregulated ventral- and dorsal-specific genes. However, an overarching logic by which the Edn1 and Bmp pathways impart region-specific skeletal shapes remained elusive, with the role of Notch signaling in this process even less clear. In the present study, we perform genome-wide expression analyses of purified arch NCCs to correlate how gene expression patterns change over time in wild-type zebrafish with how gene expression is affected by reduction or elevation of Edn1 or Jagged-Notch signaling. In so doing, we find a prominent role for Jagged-Notch signaling in repressing, and Edn1 in activating, the expression of a set of genes that are strongly induced as arch progenitors mature and begin to acquire cartilage fates, implying that Notch and Edn1 signaling exert opposite effects on cartilage differentiation within the arches. Two such downstream effectors identified in our genomic analysis are the pre-cartilage condensation marker barx1 (inhibited by Notch, activated by Edn1) and the early progenitor markers prrx1a and prrx1b (inhibited by Edn1). In mouse and chick, the homologs of prrx1a and prrx1b (Prrx1/PRRX1, previously called Prx1 or mHox) are expressed in uncondensed preskeletogenic mesenchyme [19–22], whereas Barx1/BARX1 is found in cells of nascent pre-cartilage condensations that have not or are just beginning to upregulate Sox9 [23–25]. Studies using a Prrx1 proximal promoter to drive lacZ expression [22,26] or Cre recombinase [27] revealed that the cells that make up the limb skeleton and associated connective tissues all pass through a Prrx1+ state at some point during their differentiation program. Though no similarly definitive lineage-tracing studies exist for Barx1, corollary evidence suggests that most populations of Barx1+ cells mature into Sox9+ chondrocytes [25,28–31]. In mammals, PRRX1 is required to repress cartilage differentiation in certain parts of the face: Prrx1 mouse mutants develop a large ectopic cartilage in place of the dermal squamosal bone on the side of the skull (derived from the dorsal first arch) as well as an aberrant sigmoidal process off of a shortened Meckel’s cartilage; these mutants also show chondrification of the stylohyoid ligament between the styloid process and Reichert’s cartilage (second arch), among numerous other craniofacial and limb skeletal defects [21,22]. By contrast, impaired cartilage development is observed in barx1 mutant zebrafish, particularly in the ventral/lower face [32]. In edn1 mutant zebrafish and mice mutant for the Edn1 receptor (Ednra2), defects in ventral and intermediate facial structures are preceded by a loss and shift of Barx1/barx1 expression, particularly in the second arch [10,33]. These studies indicate that although cartilage differentiation does not strictly require Barx1, chondrogenesis in the ventral and intermediate arches is most sensitive to its loss. Here we demonstrate that early arch patterning pathways compete to drive (Edn1) or restrict (Jagged-Notch, Bmp) the commitment of NCCs to chondrogenic differentiation, in part through antagonistic regulation of barx1 and prrx1a/b. These region-specific differences in the timing and extent of cartilage formation thus establish the template for the later formation of uniquely shaped bones in the upper and lower face. In an unbiased approach towards understanding facial skeletal patterning, we first performed a global gene expression analysis of pharyngeal arch NCCs at three time-points in wild-type embryos. To purify arch NCCs, we conducted fluorescence-activated cell sorting (FACS) on dissociated cells doubly positive for sox10: DsRed and fli1a: EGFP transgenes (and single-positive and double-negative cells for comparison) (Fig 1A and 1B). sox10: DsRed labels all NCCs and the ear, and fli1a: EGFP labels arch NCCs, blood vessels, and macrophages. These transgenes uniquely intersect within the arch NCC population, allowing us to selectively enrich for these cells shortly after NCC migration into the arches (20 hours post-fertilization, hpf) and during the initiation of pre-cartilage condensation formation (28 and 36 hpf). cDNA libraries were then constructed for each cell population and subjected to next-generation sequencing. To remove genes with low expression in the arches, we excluded genes with RPKM values ≤ 3 in the wild-type 36 hpf sample. However, a number of genes with known expression in the erthyroid lineage (e. g. hemoglobin genes hbae1/3, hbbe1/3), macrophages (e. g. mfap4 [34]), and the ear (e. g. mvp [35] and oc90 [36]) were found in this filtered list, suggesting some degree of contamination of the GFP/DsRed double-positive population by single-positive fli1a: EGFP or sox10: DsRed cells. We therefore further filtered for genes with expression ratios of 1. 5-fold or higher in the double-positive cells relative to both single-positive populations. This left 536,668, and 741 arch-enriched genes in the 20,28, and 36 hpf samples, respectively, with the latter group comprising the “total” arch gene list in Fig 1 (also see S1 Table). In order to understand how Edn1 and Notch signaling control the expression of these arch NCC-enriched genes, we next performed FACS purification and next-generation cDNA sequencing of GFP/DsRed double-positive cells from 36 hpf embryos with gain or loss of each signaling pathway. Specifically, we compared fold-change differences between edn1 mutants and stage-matched controls, jag1b mutants and wild-type siblings, and hsp70I: Gal4; UAS: Edn1 or hsp70I: Gal4; UAS: NICD (Notch1 intracellular domain) versus hsp70I: Gal4 controls (subjected to a 20–24 hpf heat-shock to overactivate Edn1 or Notch signaling) (see Methods). The top 20 genes up- and down-regulated in the Edn1 and Notch mutant and overexpression datasets (prior to filtering for arch NCC-enriched genes) are presented in S2 Table. Known targets of Notch (e. g. jag1b, hey2 and her2/4/15 genes) and Edn1 (e. g. dlx3b/4a/4b/6a and Evf1/2) are highly represented in these lists. All subsequent analyses were performed using the filtered list of 741 genes with arch-enriched expression in the 36 hpf wild-type sample. To identify those genes most strongly regulated by the Edn1 or Notch pathway, we divided the fold-change of the overexpression (OE) sample by the fold-change of the corresponding mutant (mut) sample. Genes considered ‘activated’ by the Edn1 or Notch pathways had an OE-fold-change/mut-fold-change ratio ≥ 1. 5 as well as an OE-RPKM/control-RPKM ratio ≥ 1. Genes considered ‘inhibited’ by Edn1 or Notch had an OE-fold-change/mut-fold-change ratio ≤ 0. 667 and a mutant-RPKM/control-RPKM ratio ≥ 1. Lastly, we performed one further refinement for the Notch lists by analyzing genes affected by treatment of embryos with the γ-secretase inhibitor dibenzazepine (DBZ), which blocks processing of the Notch receptor into its active intracellular form [37], starting at 24 hpf. After FACS-purification and next-generation sequencing of double-positive cells from 36 hpf embryos, we calculated the fold-change between DBZ-treated and control samples. Eleven of the top 20 genes downregulated in DBZ-treated embryos belong to the Her/Hes/Hey family of Notch targets [38,39] (7 of which were shared with the jag1b mutant list) (S2 Table), showing that DBZ is primarily affecting Notch signaling in this experiment. However, as γ-secretase inhibitors such as DBZ are also known to affect other signaling pathways [40], we only used the DBZ dataset to further refine the lists generated from the jag1b and NICD analyses. Specifically, we excluded genes from the ‘Notch activated’ list that were not also elevated in NICD versus DBZ (fold-change ratio ≥ 1. 25) and from the ‘Notch inhibited’ list those not also decreased in NICD versus DBZ (fold-change ratio ≤ 0. 8). These filtered gene lists (Fig 1E, S3–S6 Tables) were then used for the global analyses described below. Consistent with our previous data that the Edn1 and Jagged-Notch signaling pathways antagonize one another during facial development [17], we observed a disproportionately high number of genes oppositely regulated by these pathways. Of the 67 ‘Notch-inhibited’ genes and 107 ‘Edn1-activated’ genes, 22 were in common (Fig 1E). Conversely, 29 genes were in common between the 93 ‘Notch-activated’ and 137 ‘Edn1-inhibited’ genes (Fig 1E). These commonly regulated genes include many known positive Edn1 targets (e. g. dlx3b, dlx4a, dlx4b, epha4b, Evf1/2, msxe, and notch2) [5,17,18,33,41,42] and negative Edn1 targets (e. g. jag1b and pou3f3a/b) [17,18]. Smaller groups of genes were co-regulated by Notch and Edn1 in the same direction (positive, n = 9; negative, n = 6; S3–S6 Tables), as we previously observed for the BMP antagonist grem2 [14]. We next examined whether genes activated or inhibited by Notch or Edn1 presented any common temporal signatures during arch development in wild-type embryos. To do so, we first determined the fold changes in wild-type RPKM values for the 741 total arch genes from 20 to 28 hpf and from 28 to 36 hpf (Fig 1C, 1D and 1F; S1 Table). Total arch genes increased by a median of 1. 46-fold between 20 and 28 hpf, and 1. 18-fold between 28 and 36 hpf. In contrast, we found that the subset of genes that we had annotated as ‘Notch inhibited’ increased 2. 62-fold from 20 to 28 hpf in wild types, with many of these upregulated more than 10-fold (p < 0. 001). These strongly upregulated genes presented a range of expression levels at 20 hpf, showing that the stronger upregulation of ‘Notch inhibited’ genes is likely not an artifact of these having very low initial expression levels. ‘Edn1 inhibited’ genes also displayed a modest but significantly higher upregulation than total arch genes (median 1. 95-fold increase; p < 0. 001), though no significant differences were observed for ‘Notch activated’ genes (median = 1. 51). Although the ‘Edn1 activated’ genes were not more highly upregulated than total arch genes (median = 1. 68), the subset of ‘Edn1 activated’ genes in common with ‘Notch inhibited’ genes were the most strongly induced (median 4. 01 fold increase; p < 0. 001). Of these 22 common genes, 8 were induced more than 10-fold between 20 and 28 hpf, out of only 48 total >10-fold-upregulated arch genes. At later stages (28–36 hpf; Fig 1D and 1F), only ‘Edn1 inhibited’ genes (median 1. 52-fold increase) showed a small but significant difference (p < 0. 001) relative to all arch genes (median 1. 18-fold increase). In summary, these data show that genes commonly inhibited by Notch and activated by Edn1 are some of the most highly induced during early arch differentiation, consistent with a global role for Notch repression in limiting the differentiation of NCCs in the dorsal arches. Given the genome-wide role of dorsal Jagged-Notch signaling in repressing strongly induced genes during arch maturation, we investigated whether this might reflect a delay in cartilage differentiation in the dorsal domain versus the rest of the arches. From our genomic analysis, we observed that barx1, which marks early pre-cartilage condensations [23,29], was negatively regulated by Notch signaling (~5-fold lower in NICD versus jag1b mutant; S3 Table) and 12. 5-fold upregulated between 20 and 28 hpf in wild-type NCCs (S1 Table). By examining a time-course of barx1 expression (also see [29,32]), we find barx1 to be confined to the intermediate/ventral portions of the first and second arches at 26–28 hpf, with maxillary first arch and dorsal second arch expression not initiating until 30–32 hpf (Fig 2B). To determine whether this delay reflects later cartilage differentiation in the dorsal second arch, we made time-lapse recordings of fish expressing sox10: DsRed (which shows biphasic expression–first in all NCCs and later in differentiating chondrocytes) along with the arch NCC transgene fli1a: EGFP or the chondrocyte transgene col2a1aBAC: GFP (Fig 2C and 2D and S1 and S2 Movies). In both cases, chondrocyte differentiation was first evident in the palatoquadrate cartilage (Pq, primarily an intermediate first arch element with a small amount of dorsal contribution at its posterior end), the symplectic cartilage (Sy, intermediate second arch), at either end of the ceratohyal cartilage (Ch, ventral-intermediate second arch), and the proximal portion of Meckel’s cartilage (M, ventral-intermediate first arch). Chondrocyte transgene expression then spread into the center of the Ch and more ventral portions of the M cartilage. The last elements to differentiate were the hyomandibular cartilage (Hm, dorsal second arch) and the pterygoid process cartilage (Ptp, maxillary) (schematized in Fig 2A and 2E). We also observed that sox9a, an early marker of pre-chondrocytes [43–46] that is positively regulated by Edn1 signaling (S6 Table), was expressed only in ventral-intermediate arch NCCs at 36 hpf, with expression spreading to dorsal arch NCCs by 48 hpf (Fig 3A and 3K). Our findings point to cartilage differentiation occurring first in discrete zones, primarily within the intermediate arches, then spreading to other ventral regions and lastly to dorsal regions, consistent with previous studies based on Alcian Blue staining of sulfated proteoglycans typical of cartilage [47]. The earlier chondrogenic differentiation in intermediate/ventral arch cells relative to dorsal cells led us to hypothesize that antagonism between dorsal Jagged-Notch and ventral Edn1 signaling may serve to establish barx1+ condensations earlier and/or more extensively in the lower face. As reported previously [10], we find that barx1 expression is lost from the ventral second but not first arch of edn1 mutants at 36 hpf (Fig 3E). Conversely, elevation of Edn1 signaling (via 20–24 hpf heat-shock induction of hsp70I: Gal4; UAS: Edn1 fish) resulted in an expansion of barx1 expression throughout the arches (Fig 3F). In contrast, we find Jagged-Notch signaling to be required to restrict dorsal barx1 expression, consistent with our RNAseq data (S3 Table) and the mutually exclusive expression of barx1 and jag1b at 36 and 48 hpf (Fig 3B and 3C). In jag1b mutants, barx1 expression expands into the dorsal-posterior regions of both the first and second arch (Fig 3G; similar to Edn1 overexpression (Fig 3F) ), domains that correlate precisely with jag1b expression at this stage (Fig 3B). Conversely, jag1b expression is unaltered in barx1 mutants (S1 Fig), indicating that Jagged-Notch signaling functions largely upstream of barx1 and not vice versa. This ectopic dorsal barx1 expression was also observed in notch2; notch3 double mutants (Fig 3H), which display similar facial cartilage defects to jag1b mutants (consistent with notch2 and notch3, but not notch1a or notch1b, being expressed in arch NCCs; S2 Fig). Reciprocally, forced activation of Notch signaling in heat-shock-treated hsp70I: Gal4; UAS: NICD fish eliminated nearly all barx1 expression in the arches (Fig 3I). Finally, we find that the positive effect of Edn1 on ventral second arch barx1 expression can be explained at least in part by the previously reported role of Edn1 in blocking jag1b expression [17], as mutation of jag1b partially restored ventral second arch barx1 expression in edn1 mutants (Fig 3J). We next examined whether the ectopic dorsal expression of barx1 persisted in jag1b mutants, as well as the consequences of this for cartilage differentiation. At 36 hpf, sox9a expression marks the nascent cartilages in the ventral-intermediate arches that are the first to differentiate, with barx1 expression in a partially overlapping set of cells nearer to the poles of each arch (Fig 3A). By 48 hpf, sox9a expression has spread into the nascent dorsal cartilages yet remains only minimally overlapping with barx1 (Fig 3K, also see S3 Fig). These results are consistent with previous literature showing that Barx1 is expressed in nascent pre-cartilage condensations that have not or are just beginning to upregulate Sox9 [23–25]. In jag1b mutants at 48 hpf, we observe an expansion of barx1 but not sox9a expression in the dorsal first and second arch (Fig 3L), suggesting that a subset of dorsal arch NCCs may be trapped in an early barx1+ condensation state in the absence of Jagged-Notch signaling. This failure of ectopic dorsal barx1+ cells to transition to a more mature sox9a+ state may help explain why the dorsal cartilages of jag1b mutants are truncated rather than expanded (Fig 4B). Because ectopic dorsal barx1 expression correlated with dorsal cartilage defects in jag1b and notch2; notch3 mutants, we next investigated whether this reflected a common early requirement for Jagged-Notch signaling for both processes. To temporally inhibit Notch signaling, we treated embryos at different stages with 10 μM DBZ, and evaluated the effects on barx1 expression and skeletal patterning. Although DBZ may also affect other signaling pathways [40], our RNAseq analysis showed that the majority of the most strongly downregulated genes were canonical Notch targets (S2 Table). This analysis focused on the first arch phenotypes, which have proved the most penetrant and consistent across all of our Notch loss-of-function models. Compared with DMSO-treated controls, treatment of embryos with DBZ starting at 8 hpf resulted in a highly penetrant expansion of barx1 expression into the posterior dorsal first arch (12/12), as well as dorsal cartilage defects similar to jag1b mutants (19/20 with Pq malformations; Fig 3M and 3N). DBZ treatment starting at 24 hpf resulted in a weaker and less penetrant barx1 expansion (10/13 embryos with ectopic first arch barx1) and milder dorsal cartilage defects (41/44 with moderate Pq malformations). In contrast, treatment at 28 hpf only mildly affected barx1 expression in 7/13 embryos, with only 16/39 embryos displaying weak dorsal cartilage malformations (Fig 3M and 3N). Treatments initiated at 32 hpf did not affect barx1 expression or skeletal patterning. Inhibition of Notch signaling at these stages also had other effects on embryo development, including cardiac edema, which likely contributed to the general reductions in cartilage size. In summary, we observe a tight correlation between barx1 expression changes and subsequent malformations of dorsal cartilages in Notch-deficient embryos, with the requirement for Notch inhibition by approximately 24 hpf being consistent with the predicted global effects of Notch in repressing arch gene induction between 20–28 hpf (Fig 1F). We next investigated the extent to which the ectopic dorsal expression of barx1 in Notch pathway mutants contributes to the dorsal cartilage malformations. In particular, jag1b mutants display several characteristic changes in cartilages of the upper face, including truncation of the posterior end of Pq (i. e. the portion deriving from dorsal first arch NCCs; Fig 2A) and a variable reduction of the anterior part of Hm (Fig 4A and 4B). jag1b mutants also display a highly penetrant posterior shift of Hm such that it sits closer to the ventral Ch cartilage [17]. In barx1 mutants, the dorsal cartilages are largely unaffected, with there instead being conspicuous reductions of the ventral M and Ch cartilages (Fig 4C) [32]. In jag1b; barx1 mutants, we observed an incompletely penetrant rescue of posterior Pq (truncation in 11/27 double mutants versus 16/16 jag1b mutants, p < 0. 0001) and the position of Hm (posterior shift in 13/27 double mutants versus 14/16 jag1b mutants, p < 0. 0001) (Fig 4D and 4E). However, ventral M and Ch defects were not restored, and the anterior Hm was more prominently diminished (loss in 18/27 double mutants versus 5/16 jag1b mutants, p < 0. 0001). Of note, the two regions of skeletal rescue (posterior Pq and posterior Hm) correlate precisely with the earlier ectopic expression of barx1 in dorsal-posterior first and second arch domains of jag1b mutants (Fig 3G), suggesting that the ectopic barx1 may account in part for these phenotypes. On the other hand, the incompletely penetrant rescue of these elements, in addition to exacerbated phenotypes in other regions (e. g. anterior Hm), indicates the presence of other causative changes in jag1b mutants beyond barx1 misexpression. The finding that posterior-dorsal cells ectopically express barx1 but fail to turn on sox9a in jag1b mutants, as well as the fact that cartilage defects were only modestly rescued in jag1b; barx1 double mutants, suggest that Notch signaling has additional roles in dorsal cartilage development. In order to better understand the reductions of dorsal cartilage in jag1b mutants, we used photoconversion of the kikGR protein to follow the fate of dorsal second arch NCCs in wild types versus mutants (Fig 5). When wild-type cells were converted at 36 hpf and then re-imaged at 6 days post fertilization (dpf), we found that anterior dorsal second arch NCCs contributed to the anterior portion of the Hm cartilage, central dorsal second arch NCCs to the posterior portion of Hm, and posterior dorsal second arch NCCs to a small amount of Hm and the opercle bone to which it attaches (Fig 5A–5C). In jag1b mutants, all three domains contributed to similar portions of the malformed Hm cartilage as in wild types (Fig 5D–5F), indicating no major shift in the fate map of skeletal precursors in mutants. However, whereas cells from all three regions spread along the dorsoventral axis in wild types, cells from comparable domains in jag1b mutants gave rise to much smaller domains of cartilage (Fig 5G). These results suggest that Jagged-Notch signaling is also required for the expansion of the dorsal second arch NCCs that generate cartilage in the upper face. While loss of Jagged-Notch signaling can rescue barx1 expression and ventral cartilage development in edn1 mutants (Fig 3J; [17]), the partial and largely second-arch nature of this rescue implies the presence of other important pathways downstream of Edn1. Our expression analysis of sorted arch NCCs identified two genes implicated in early skeletogenic mesenchyme identity, prrx1a and prrx1b, which, like jag1b, were upregulated in edn1 mutants and downregulated in Edn1-overexpressing embryos (S5 Table). Loss of the homologous Prrx1 gene in mice results in ectopic dorsal facial cartilage [21,48], implying that Prrx1 genes may also restrict cartilage formation in the upper face. We thus reasoned that Edn1-mediated repression of prrx1a and prrx1b could help to explain the observed acceleration of cartilage differentiation in the intermediate domain (Fig 2B–2E). Consistently, we observed that expression of prrx1a and prrx1b was largely excluded from NCCs in the intermediate domain, instead being confined to the dorsal-most and ventral-most poles of the first two arches in 36 hpf wild types (Fig 6A and 6B). However, as predicted by our RNAseq analysis, prrx1a and prrx1b expression was upregulated along the ventral border and expanded into the intermediate arches of edn1 mutants, and lost in Edn1-overexpressing embryos (Fig 6C–6F), in accord with the elevated ventral Prrx1 expression observed in Dlx5/6 mutant mice [6]. Conversely, overactivation of Bmp4 signaling (via 20–24 hpf heat-shock induction of hsp70I: Gal4; UAS: Bmp4 fish) upregulated prrx1a and prrx1b expression throughout the arches (Fig 6G–6J), in accord with previous findings that Bmp signaling promotes genes associated with progenitor status and self-renewal in arch NCCs [16]. Positive regulation by the ventral Bmp signal, combined with negative regulation by intermediate Edn1, could help explain the restriction of prrx1a/b expression to the ventral pole of the arches. We next examined whether the Bmp4 induction of prrx1a/b is mediated by Hand2, a strong Bmp target that is specifically expressed in NCCs at the ventral border of the arches in both mice and fish [5,49], domains that closely overlap with ventral prrx1a/b expression. While Hand2/hand2 expression requires positive input from both the Edn1/Dlx and Bmp pathways [5,6, 13,49,50], overexpression of Bmp4 –but not Edn1 –induces its widespread ectopic expression [13,14], similar to the patterns observed here for prrx1a/b (Fig 6G–6J). However, consistent with previous results in mice [51], prrx1a and prrx1b were expressed largely normally in hand2 mutants, with a limited expansion of prrx1a in the ventral domain (S4 Fig). Thus, Bmp signaling appears to positively regulate prrx1a/b expression largely independently of Hand2 function. In order to interrogate Prrx1 function in zebrafish, we used TALENs to generate prrx1ael558 and prrx1bel491 mutant alleles resulting in early truncation of the Prrx1a and Prrx1b proteins upstream of the conserved DNA-binding homeobox domains (S5 Fig). Whereas prrx1a and prrx1b single mutants did not show craniofacial defects during larval stages (consistent with their identical expression patterns), double homozygous mutants exhibited highly penetrant abnormalities affecting dorsal skeletal elements of the first two arches (Fig 7A–7F). Identical dorsal skeletal phenotypes were seen in double mutants carrying prrx1ab1246 and prrx1bb1247 alleles independently generated by CRISPR-mediated mutagenesis (S5 Fig). In the first arch of double mutant larvae, ectopic cartilage develops along the dorsomedial surface of the Pq cartilage in place of the dermal entopterygoid bone. This extra cartilage is occasionally fused with the trabecular cartilages of the neurocranium. In approximately 40% of double mutant embryos, Pq also extended dorsal-posteriorly to fuse with the otic capsule. In the second arch, the top of the Hm cartilage is malformed, with two highly penetrant cartilaginous fusions to the anterior and middle parts of the otic capsule. The foramen of the Hm, a channel for the VIIth cranial nerve and the anterior lateral line nerve [52], is absent, and the opercle bone is reduced. Despite the expression of prrx1a/b at both the dorsal and ventral poles of the arches, double mutants had no detectable defects in ventral cartilages. Consistent with the ectopic dorsal cartilage, we also found that double mutants displayed ectopic barx1 expression at earlier stages (36 hpf) in dorsal arch regions that generate the parts of Pq, Hm, and otic cartilages affected in mutants (Fig 7G and 7H). This upregulation of barx1 in double mutants is consistent with the near mutually exclusive expression of prrx1a/b and barx1 in 36 hpf wild-type embryos (Fig 6G, 6H and 6L). In contrast to jag1b mutants (Fig 3L), these ectopic barx1 expression domains largely disappeared by 48 hpf (Fig 7I and 7J), perhaps accounting for the ectopic formation of cartilage in Prrx1 but not Notch pathway mutants. Interestingly, despite hand2 being expressed in a similar domain to prrx1a/b in the ventral arches, barx1 has been reported to be lost in hand2 mutants [32], opposite to the barx1 expansion we observe in prrx1a/b mutants. However, hand2 expression was unaffected in prrx1a; prrx1b mutants (S4 Fig), similar to previous observations in Prrx1-/- mice [53], suggesting that Prrx1 and Hand2 act antagonistically and independently to regulate barx1 expression and chondrogenesis in the ventral second arch. Despite both prrx1a/b and jag1b expression being mutually exclusive to barx1, we found only limited overlap between these genes (Fig 6K and 6L). We therefore hypothesized that these pathways function independently to limit barx1 expression and cartilage formation in distinct domains of the arches. Consistently, we observed no defect in prrx1a or prrx1b expression in 36 hpf jag1b mutants, although forced activation of Notch signaling expanded prrx1a and prrx1b expression ventrally and decreased it dorsally (Fig 8A–8F). Loss of jag1b also partially rescued the ventral expansion of prrx1b observed in edn1 mutants, especially in the second arch (Fig 8G–8I). These findings indicate that, although jag1b is not required for prrx1a/b expression, high levels of Notch signaling (either artificially or by loss of Edn1) can induce prrx1b expression ventrally. Reciprocally, a subset of prrx1a; prrx1b mutants showed a modest reduction of jag1b expression limited to the dorsal posterior second arch (Fig 8J and 8K). To further clarify the genetic interaction between these genes, we analyzed jag1b; prrx1a; prrx1b triple mutants (Fig 8L–8O). In 9/13 triple mutant sides examined, we observed the ectopic posterior extension and fusion of the Pq cartilage to the ear (and not the Pq truncations seen in Notch pathway mutants), indicating that prrx1a/b are largely epistatic to jag1b with respect to the ectopic Pq phenotype. However, in addition to this extra cartilage, skeletons of the triple mutants (but not prrx1a; prrx1b double mutants) showed irregular gaps within the body of Pq, reminiscent of abnormalities seen in Notch pathway mutants. These observations support the triple mutant phenotype being largely additive, in line with Prrx1a/b and Jagged-Notch signaling having distinct roles in regulating condensation and cartilage formation in the upper face. Because ventral prrx1a/b expression increases in the ventral arches of edn1 mutants (Fig 6C and 6D), we speculated that increased repression of cartilage differentiation by Prrx1 proteins might contribute to the ventral skeletal losses seen in edn1 mutants. Indeed, we found that homozygous loss of both prrx1a and prrx1b resulted in a modest rescue of ventral cartilage formation in edn1 mutants, particularly in the second arch (Fig 9B and 9C), as well as rescue of ventral barx1 in the second arch and dlx5a expression in both the first and second arch (Fig 9G–9J). The partial rescue of ventral cartilage in prrx1a; prrx1b; edn1 mutants, as well as the earlier recovery of barx1 and dlx5a expression, are qualitatively similar to the phenotypes seen in jag1b; edn1 mutants (Fig 9D and [17]). By contrast, there was no rescue of ventral hand2 expression, consistent with our finding that Prrx1a/b do not regulate hand2 (S4 Fig). As Hand2 normally restricts Dlx expression into the ventral-most arches [54,55], the lack of hand2 recovery in the triple mutants may explain the ectopic ventral expansion of dlx5a in the prrx1a; prrx1b; edn1 mutants. In jag1b; edn1 mutants, the partial recovery of ventral barx1 expression correlated with zones where prrx1b expression was reduced to control levels (Fig 8I). We therefore asked whether the remaining areas of elevated Prrx1 expression in jag1b; edn1 mutants might account for the incomplete rescue. Consistently, we found that progressive reduction of prrx1a/b gene dosage in jag1b; edn1 mutants resulted in a progressively better rescue of ventral cartilages, with 5/6 quadruple homozygous prrx1a; prrx1b; jag1b; edn1 mutants showing a prominent rescue of the ventral second arch-derived Ch cartilage and improved elongation of the first arch-derived M cartilage (Fig 9E and 9F). However, even in these quadruple mutants, the ‘rescued’ ventral cartilages are still smaller than in wild types, and the dorsal skeletal phenotypes associated with jag1b and prrx1a; prrx1b mutants are still present. These findings reveal important parallel contributions of ectopic Prrx1 and Jagged-Notch activity to the ventral craniofacial defects of edn1 mutants, yet indicate that Edn1 has additional functions beyond inhibiting Prrx1 and Notch activity. RNAseq analyses of facial NCCs confirmed our previous findings that Notch acts oppositely to Edn1 during pharyngeal arch development [17]. At a mechanistic level, this global analysis revealed that a major function of Notch signaling is to repress the expression of some of the most strongly upregulated genes in early arch development. These include the homologs of a number of genes implicated in mesenchymal condensation, chondrogenesis, and general skeletogenesis in mammals: e. g. barx1 [23,31,32], ctgfb [56], col6a1 and col6a6 [57,58], and tbx22 [59,60]. As we only profiled global gene expression patterns in mutants and overexpression embryos at 36 hpf, we cannot conclude whether Notch represses these highly-induced genes only at this later stage, or whether it also restrains their initial upregulation. The concept of Notch limiting differentiation is becoming a common theme in many developmental and regeneration contexts. For example, sustained Notch signaling in preskeletogenic mesenchyme in vivo or mesenchymal progenitors in vitro severely abrogates cartilage formation, with cells inappropriately maintained in a precursor state [61–64]. Likewise, Notch signaling promotes regeneration of the caudal fin of zebrafish by maintaining the blastema in a proliferative, undifferentiated state [65,66]. Though Notch can also promote differentiation in other contexts (e. g. stimulating maturation and hypertrophy in committed chondrocytes [reviewed by [67]]), our findings are consistent with the large body of literature describing roles for Notch in resisting differentiation of progenitor cell populations, in this case specifically in the dorsal arches. Our genomic analysis also identified two Prrx1 homologs (prrx1a and prrx1b) as negative targets of Edn1 that function in parallel to Jagged-Notch signaling to restrain cartilage differentiation, yet these pathways appear to do so in different ways (Fig 10). jag1b and prrx1a/b are expressed in largely non-overlapping domains and are generally not required for the other’s expression. The skeletal phenotypes of Notch and Prrx1 mutants also differ in critical ways. Mutants in both pathways develop ectopic barx1+ condensations and malformed cartilages in the dorsal arches, but only prrx1a; prrx1b mutants form ectopic dorsal cartilage. One potential explanation is that ectopic barx1 expression persists in dorsal NCCs at later stages in jag1b but not prrx1a; prrx1b mutants. Perhaps, Jagged-Notch signaling is also required for ectopic barx1+ cells to progress to a sox9a+ chondrogenic state. Further, our fate-mapping studies show that dorsal arch NCCs expand less in jag1b mutants compared to wild types, which could be due to persistent barx1 expression restricting the proliferation of chondrogenic cells. However, loss of barx1 improved only a subset of skeletal defects in jag1b mutants, suggesting that skeletal changes in Notch-deficient embryos result from more than just ectopic barx1 expression. In prrx1a; prrx1b mutants, the release of dorsal cells from a transient barx1+ state may instead allow these cells to reach a critical threshold for making ectopic cartilage (as proposed for the Prrx1 mouse mutant [68]). The finding that prrx1a; prrx1b double mutants presented skeletal defects only in dorsal elements was somewhat unexpected, given the expression of prrx1a and prrx1b in both dorsal and ventral arch regions. While homozygous or dominant-negative mutations in PRRX1 have been associated with loss of the lower jaw in humans [69–73], Prrx1 mutant mice are similar to zebrafish mutants in displaying ectopic dorsal cartilage, but dissimilar in showing minor abnormalities of the lower jaw [21]; Prrx1; Prrx2 double mutants display much more pronounced jaw reductions [22,48,53]. As zebrafish lack a Prrx2 homolog [74,75], the lack of lower jaw defects in prrx1a; prrx1b double mutants could reflect redundancy with other pathways, or, alternatively, the evolution of different requirements for Prrx1 genes between fish and mammals. At a molecular level, the ectopic barx1 expression we observe in the dorsal arches of prrx1a; prrx1b mutant fish is reminiscent of the medial expansion of Barx1 seen in the ventral first arch of Prrx1; Prrx2 mutant mice [53]. While our data implicate Prrx1 genes and Jagged-Notch as two important negative targets of Edn1 in the ventral arches of zebrafish, the fact that edn1 mutant phenotypes are only partially rescued by the combined loss of Prrx1a/b and Jag1b suggests other yet to be identified key targets of Edn1. Indeed, our RNAseq analysis revealed two different classes of genes activated by Edn1: (1) those that are highly upregulated during early arch development and also inhibited by Notch (including many well-known Edn1 targets such as dlx3b/4a/4b/6a, hand2, epha4b, Evf1/2, and msxe) and (2) those that are Notch-independent and only modestly upregulated during early arch development. Given that jag1b itself is negatively regulated by Edn1 signaling [17], many of the genes on the first list may in fact be Notch targets that are only indirectly stimulated by Edn1. Functional interrogation of these two classes of target genes should help uncover additional functions of Edn1 in arch development. Our findings in zebrafish also support a greater role for the Jagged-Notch and Prrx1 pathways in patterning the second arch and posterior half of the first arch compared with the anterior portion of the first arch, which generates the bulk of the lower and upper jaw skeleton (Fig 10). For example, jag1b is expressed in only a limited posterior dorsal domain of the first arch and not in the maxillary or mandibular prominences [17], and first arch-derived skeletal structures are less affected than second arch-derived structures in jag1b and prrx1a; prrx1b mutants. barx1 expression is also primarily lost in ventral NCCs of the second but not first arch in Edn1 pathway mutants [10,33], consistent with the more pronounced upregulation of jag1b and prrx1a/b in this domain. This second arch bias is also reflected by greater rescue of second versus first arch ventral cartilages upon loss of Prrx1 and Notch signaling in edn1 mutants. Given the very different arrangements of cartilage and bone in the second versus first arch, it is not surprising that programs that restrict cartilage formation have distinct roles in each arch. In the future, it will be interesting to explore how Hoxa2 and Hoxb2, which confer second arch identity [76–79], impact the Notch- and Prrx1-based cartilage restriction programs we have identified. We also note that other major signaling pathways, such as Bmp, Fgf, Tgfβ, Shh, and Wnt, also influence the spatiotemporal patterns of differentiation within the arches–including control of prrx1a/b and barx1 expression [13,14,25,29,80–86]. For example, our work indicates that Bmp signaling likely helps to establish prrx1a/b expression at the ventral poles of the zebrafish arches. Whether Bmp signaling regulates Prrx1 genes in other vertebrates remains unclear, as previous studies did not detect changes in Prrx1 expression in conditional Bmp4 deletion mice [15] or chicken mandibular explants exposed to exogenous Bmp ligands and antagonists [80]. Future work will need to integrate these other key patterning programs into the model to more fully explain how the timing and extent of chondrogenesis is precisely controlled in the developing face. Heterochrony in skeletal differentiation is an important mechanism contributing to the evolution of morphological differences between species [reviewed by [87], also see [88,89] and references therein]. This concept of variation in developmental timing of homologous structures between species has been proposed to explain, for example, differences in beak length and morphology, as well as the shape of Meckel’s cartilage between quail and duck [90–92]. Our work supports the idea that differential developmental timing can also be a critical driving force for varying skeletal structure within an individual. In the arches of zebrafish, chondrocyte differentiation invariably occurs first in intermediate/ventral before dorsal regions [47]. We have found that these events are presaged by an earlier initiation of barx1 and sox9a expression in intermediate/ventral relative to dorsal arch NCCs, with Jagged-Notch and Prrx1a/b circumscribing the size of the later-forming dorsal condensations. In a previously proposed ‘hinge-and-caps’ model [93,94], arch polarity is established by differential signaling in the intermediate regions of the arches (i. e. , ‘hinges’) versus the dorsal and ventral poles of the arches (i. e. , ‘caps’). Our work provides potential cellular correlates to these hinges and caps in zebrafish, particularly in the second arch and posterior portion of the first arch [95]. We propose that the poles of the arches, or caps, represent progenitor domains, consistent with their expression of the mesenchyme progenitor marker Prrx1 in many species [22,80,96]. In contrast, the intermediate arches, or hinges, reflect the sites of initial chondrogenesis, as evidenced by their earlier expression of sox9a. Whereas this model predicts that Prrx1 expression at both the ventral and dorsal poles would restrict chondrogenesis relative to the intermediate hinges, dorsal-specific Jagged-Notch signaling would further restrict chondrogenesis in dorsal relative to ventral regions. This model would explain our observations that cartilages generally form first in the intermediate regions, then spread next to the ventral pole, and lastly to the dorsal pole due to combined repressive effects of Prrx1 and Jagged-Notch. However, the timing of cartilage differentiation is clearly more complex. For example, our time-lapse imaging revealed that the Ch cartilage first undergoes chondrogenesis at its tips and then later in its center, potentially correlating with expression of the Bmp target gene msxe in a subset of ventral second arch Ch precursors [14]. Hence, layering of additional signaling pathways, such as Bmp, may further refine the timing of cartilage differentiation within the arches. Given the expression of Prrx1 homologs at the ventral and dorsal poles of the arches from sharks through mammals [96], and conserved expression of Jag1 in the dorsal arches of mice [97,98], it appears likely that a similarly regulated intermediate—ventral—dorsal gradient of chondrogenesis may be conserved across vertebrates. For example, in human embryos, Meckel’s cartilage (ventral) differentiates before those elements that form in more proximal/dorsal positions (i. e. the malleus, sphenoid, and styloid process) [99]. On the other hand, differences in the timing and extent of cartilage differentiation might account for the striking differences in facial form between species. In larval zebrafish, the majority of the bony visceral skeleton arises through cartilage templates in the first two arches. In contrast, much of the mammalian facial skeleton forms through direct ossification, with exceptions including Meckel’s cartilage in the lower jaw and the ossicles of the middle ear. These differences might be reflected in the fact that loss of the pre-cartilage marker Barx1/barx1 has more profound effects on the facial skeleton of zebrafish than mice [32,100,101], and, reciprocally, that loss of Prrx1 genes impacts lower jaw development in mammals but not fish [21,22,48]. It will therefore be interesting to examine whether differences in the requirements and/or regulation of Prrx1 and Barx1 genes underlie differences in the extent and timing of chondrogenesis between species. An unanswered question is how heterochrony in cartilage differentiation might translate to the distinct shapes of skeletal elements along the dorsoventral axis. One possibility is that dorsoventral differences in the timing at which progenitors commit to a cartilage fate influences the duration and types of signals they encounter from the surrounding endoderm and ectoderm. For example, Jagged-Notch signaling in the dorsal posterior second arch would protect progenitors from early chondrogenesis, thus allowing these cells to receive later osteogenic cues that direct them to form the large, fan-shaped opercle bone. Such an interpretation is consistent with the reciprocal expansion of barx1+ pre-cartilage condensations and loss of opercle bone in jag1b mutants, and the formation of an ectopic opercle bone upon forced expression of JAG1 in ventral regions [17]. Another possibility is that the timing of condensation formation and subsequent chondrogenesis influences the degree of proliferative expansion of elements in different arch domains [1]. In conclusion, our study revisits heterochrony, a fundamental concept of evolutionary biology, from a developmental perspective, showing that the timing and extent of cartilage differentiation within specific arch regions contributes to the diversity of skeletal shapes within the skull. All zebrafish (Danio rerio) were maintained and handled in strict accordance with good animal practices as defined by the relevant national and local animal welfare bodies. Zebrafish embryos were anesthetized for time-lapse imaging or prior to fixation by adding tricaine to their water. All animal experiments performed in this study were approved by the Institutional Animal Care and Use Committee of the University of Southern California (No. 10885,20193). Zebrafish (Danio rerio) embryos were reared at 28. 5°C and staged as previously described [102]. The following transgenic lines were maintained as heterozygotes: Tg (fli1a: EGFP) y1 [103], Tg (sox10: DsRed-Express) el10 [104], Tg (col2a1aBAC: GFP) [105], Tg (hsp70I: Gal4) kca4/+ and Tg (UAS: myc-Notch1a-intra) kca3 (hereafter UAS: NICD) [106], Tg (UAS: Edn1; α-crystallin: Cerulean) el249 and Tg (UAS: Bmp4; cmlc2: GFP) el49 (hereafter UAS: Edn1 and UAS: Bmp4, respectively) [14]. The hsp70I: Gal4 and UAS: NICD lines do not contain selectable markers and were genotyped using primers for Gal4 (F: 5′-CTCCCAAAACCAAAAGGTCTCC-3′; R: 5′-TGAAGCCAATCTATCTGTGACGG-3′) and UAS: NICD (F: 5’-CATCGCGTCTCAGCCTCAC-3’; R: 5’-CGGAATCGTTTATTGGTGTCG-3’). For the UAS: Edn1 line, in cases where it was not possible to ascertain α-crystallin: Cerulean expression in living animals, individuals carrying the transgene were identified by genotyping for the lens marker (F: 5’-TGGTGCAGATGAACTTCAGG-3’ and R: 5’- GCATGCAGACAGCAGCAATA-3’). Gal4 expression was induced in hsp70I: Gal4; UAS: NICD, hsp70I: Gal4; UAS: Edn1, and hsp70I: Gal4; UAS: Bmp4 embryos by heat-shocking from 20–24 hpf in a 40°C incubator. The sucker/edn1tf216 [5], jag1bb1105 [17], barx1fh331 [32], notch3fh332 [107], and Df (Chr1) hand2s6 [108,109] mutant lines were described previously and genotyped by PCR using GoTaq (Promega, Madison, WI) with published primer sequences followed by digestion with the appropriate restriction enzymes. Three new mutant lines (notch2el515, prrx1ael558, prrx1bel491) were generated for this study via TALEN-mediated mutagenesis. The notch2el515 allele was generated with the same TALEN pair used for the previously reported notch2el517 allele [110]. Exon 2 (of 4) of prrx1a was targeted with TALENs that recognize the following sequences: Left: 5’-CGTTGAGCTGCTCGTCTGGA-3’; Right: 5’-TGTTTCGCCTCTGTTTACGC-3’, and exon 1 (of 5) of prrx1b was targeted with TALENs that recognize the following sequences: Left: 5’-TGGCGAAACGGGCAGGACTA-3’; Right: 5’-TGTATCACTGCCACTCGTTA-3’. TALEN constructs were produced using a PCR-based platform [111]. The TALEN plasmids were linearized by StuI digestion (New England Biolabs, Ipswich, MA), and RNAs were synthesized with the mMessage mMachine T7 Ultra kit (Ambion/Life Technologies, Carlsbad, CA, USA). TALEN RNAs (100 ng/μl) were injected into 1‐cell-stage embryos. Germline founders were identified among the injected individuals by screening outcrossed progeny by PCR followed by restriction digestion. The primers used to identify mutations in each gene are listed in S7 Table. Stable mutant alleles predicted to result in immediate stop codons or frameshifts followed by stop codons were identified by sequencing PCR products in the F1 generation. The notch2el515 allele consists of a 2-bp deletion and a single nucleotide polymorphism (SNP) that destroy a ClaI site in the target region and result in an immediate stop after aa 208 (of 2471), within the extracellular EGF-like domains. The prrx1ael558 allele is an 8-bp deletion that destroys a BseRI site and produces a frameshift after aa 90 (of 245; upstream of the homeodomain at aa 101–155), causing the addition of one incorrect amino acid followed by a stop codon. The prrx1bel491 allele is a 2-bp insertion that abolishes a HinfI site and causes a frameshift after aa 68 (of 245; upstream of the homeodomain at aa 87–165), resulting in the addition of four incorrect amino acids followed by a stop codon. Two additional alleles, prrx1ab1246 and prrx1bb1247, were independently generated via CRISPR-mediated mutagenesis. CRISPR gRNA templates were produced via PCR following a published protocol [112], and gRNAs were synthesized with the MEGAScript T7 transcription kit (Ambion) and column-purified with the mirVana miRNA isolation kit (Ambion). Cas9 RNA was transcribed from pT3TS-nCas9n with the T3 mMessage kit (Ambion) and purified with an RNeasy Mini Kit (Qiagen, Hilden, Germany) [112]. gRNAs (25 ng/μl) plus Cas9 RNA (50 ng/μl) were injected into 1‐cell-stage embryos, and stable lines were identified by sequencing as described above. The prrx1ab1246 allele is an 11-bp deletion that causes a frameshift after aa 62, which results in the incorporation of 28 additional amino acids followed by a stop codon. The prrx1bb1247 allele consists of an 8-bp deletion that causes a frameshift after aa 24 and the inclusion of 29 incorrect amino acids before termination. All animal experiments performed in this study were approved by the Institutional Animal Care and Use Committee of the University of Southern California. For RNA sequencing experiments, fli1a: EGFP fish were crossed to the sox10: DsRed line, and doubly transgenic fli1a: EGFP; sox10: DsRed fish were further crossed to the edn1, jag1b, hsp70I: Gal4, UAS: Edn1, and UAS: NICD lines. Each of these lines were then separately incrossed to generate embryos for FACS sorting. Wild-type fli1a: EGFP; sox10: DsRed (20,28, and 36 hpf) embryos were sorted for co-expression of GFP and DsRed expression under a fluorescent dissecting stereomicroscope (Leica M165 FC, Wetzlar, Germany) prior to dissociation. Single-positive and double-negative embryos were also saved as controls for FACS. Mutant edn1; fli1a: EGFP; sox10: DsRed embryos were selected under the fluorescent microscope at approximately 34 hpf based on the reduced distance between the bottom of the first pharyngeal pouch and the ventral border of the arches. To identify jag1b mutants and doubly-transgenic hsp70I: Gal4; UAS: Edn1 or hsp70I: Gal4; UAS: NICD individuals, we genotyped cell lysates of tail biopsies collected from anesthetized individual 24-hpf fli1a: EGFP; sox10: DsRed double-positive embryos. To induce Edn1 or NICD overexpression in the hsp70I: Gal4; UAS: Edn1 and hsp70I: Gal4; UAS: NICD lines, embryos were heat-shocked from 20–24 hpf in an incubator set at 40°C. As another means of inhibiting Notch signaling, fli1a: EGFP; sox10: DsRed embryos were treated with dibenzazepine (DBZ; Tocris, Bristol, UK; final concentration of 10 μM in embryo medium) from 24–36 hpf. The number of embryos used for each sort and the number of cells obtained are presented in S9 Table. To facilitate FACS analyses at the 36 hpf time point, embryos were moved at 27 hpf to an incubator set at 22°C to delay their development such that they reached an approximation of the 36 hpf stage the following morning. fli1a: EGFP; sox10: DsRed double-positive embryos were dissociated following [113], with minor modifications. Briefly, 30–40 dechorionated embryos were incubated in fresh Ringer’s solution for 5–10 minutes and agitated by pipetting to remove the yolk. The deyolked embryos were then mixed with a protease solution containing 0. 25% trypsin (Life Technologies), 1 mM EDTA, and 2 mg/ml Collagenase P (Roche Life Science, Indianapolis, IN) in PBS and incubated at 28. 5°C for 15 min, pipetting up and down every 5 min to aid the dissociation. The reaction was stopped by the addition of a 6x stop solution consisting of 6 mM CaCl2 and 30% fetal bovine serum (FBS) in PBS. The cells were pelleted via centrifugation at 2000 rpm for 5 min at 4°C, resuspended in suspension medium (1% FBS, 0. 8 mM CaCl2,50 U/ml penicillin, and 0. 05 mg/ml streptomycin (Sigma-Aldrich, St. Louis, MO) in phenol red-free Leibovitz’s L15 medium (Life Technologies) ), pelleted again as above, and then resuspended in 500 μl suspension medium and placed on ice. Cells were sorted by FACS for GFP and DsRed expression on a MoFlo Astrios instrument (Beckman-Coulter, Brea, CA, USA). GFP/DsRed double-positive, double-negative, and single-positive populations were collected directly into RLT lysis buffer (Qiagen). Total RNA was immediately extracted using the RNeasy Micro kit (Qiagen) following the manufacturer’s protocol and quantified on a NanoDrop 2000 spectrophotometer (NanoDrop Products, Wilmington, DE, USA). The quality and quantity of extracted RNA were assessed on a Bioanalyzer Pico RNA chip (Agilent, Santa Clara, CA). cDNA was then made from the extracted RNA using the SMARTer V3 kit (Clontech, Mountain View, CA), according to the manufacturer’s instructions. The number of amplification cycles for cDNA synthesis was estimated based on input amounts of RNA. The size and amount of the resulting cDNA were then confirmed by Bioanalyzer. Sonication was performed on a S2 ultrasonicator (Covaris, Woburn, MA) according to Clontech’s recommended conditions. DNA libraries were constructed using the Kapa Hyper prep kit (Kapa Biosystems, Wilmington, MA) and NextFlex adapters (Bioo Scientific, Austin, TX). Libraries were visualized by Bioanalyzer analysis and quantified by qPCR (Kapa library quantification kit). Sequencing was performed on Illumina HiSeq 2000 (50-bp paired end reads) and NextSeq 500 (75-bp paired end reads) machines (Illumina, San Diego, CA). DNA libraries were constructed and sequencing was performed at the Norris Cancer Center Molecular Genomics Next Gen Sequencing Core at USC. Raw sequencing data in Fastq format was imported into the Partek Flow interface for alignment and quantification. Pre-alignment QC showed that the reads from all samples had generally high quality, with the average Phred quality score for each sample being above 30. Reads were then trimmed from both ends based on Phred quality score with a minimum end quality level of 20 and a minimum acceptable read length of 25. The TopHat 2 algorithm was used to align the trimmed reads to the zebrafish GRCz10 genome assembly (Ensembl_v80). Aligned reads were then quantified using the Partek E/M algorithm with default parameters to yield the RPKM values. RNAseq files have been deposited in NCBI’s Gene Expression Omnibus and are accessible through the GEO Series accession number GSE72985. Filtered gene lists were derived in MS Excel as described in the Results section. Six genes on the list of arch NCC-enriched genes had passed the ≥ 3 RPKM threshold at 36 hpf but had RPKM values of 0 in the 20 hpf sample, leading to a division error that would have precluded their inclusion in the temporal expression analysis; we thus set the 20 hpf RPKM value for these genes to 0. 01 based on the lowest positive RPKM values in the dataset. Alcian Blue and Alizarin Red staining to detect cartilage and bone, respectively, was performed on 4–6 dpf larvae as previously described [114]. Two-color fluorescent in situ hybridizations were carried out as previously reported [17]. Published probes used in this study include dlx2a [115], dlx5a [10], notch2, jag1b [17], and sox9a [44]. Partial cDNAs for barx1, notch1a, notch1b, notch3, prrx1a, and prrx1b were cloned into the pCR-Blunt II-TOPO vector (Life Technologies) and sequence-verified prior to plasmid linearization and in vitro transcription with Sp6 or T7 polymerase (Roche) (S8 Table). To determine when Notch signaling affects skeletal patterning, we treated embryos with the γ-secretase inhibitor DBZ. DBZ dissolved in dimethyl sulfoxide (DMSO; 10 mM stock) was added to embryo medium to a final concentration of 10 μM. Embryos (n = 30–50 per treatment) were incubated in this solution starting at 8,24, or 28 hpf until fixation at 36 (8 hpf group) or 42 hpf (other groups) for in situs or at 4 dpf for Alcian and Alizarin staining. In the groups used for skeletal staining, the DBZ solution was refreshed at 48 hpf and thoroughly washed out at 56 hpf. Clutch-mate controls were exposed to the same concentration of DMSO. Embryos were dechorionated at 24 hpf to improve drug accessibility. Fate-mapping was performed with the green-to-red photoconvertible kikGR protein [116]. kikGR RNA was injected into embryos from a jag1b; fli1a: EGFP cross at the one-cell stage. At 36 hpf, embryos were anesthetized in Tricaine (Sigma-Aldrich) and mounted in 0. 2% agarose for confocal imaging. Small groups of cells in the dorsal arches were selected using the region of interest tool in the Zeiss LSM software and exposed to the UV 405 laser until the red photoconverted protein became apparent (typically ~10 seconds using 50% laser power). The same animals were reimaged at 6 dpf to determine the destination of the photoconverted cells and then genotyped for the jag1bb1105 mutation. Confocal images of in situ hybridizations (~30 μM z-stacks) were captured on a Zeiss LSM5 microscope using ZEN software. Time-lapse imaging of doubly transgenic fli1a: EGFP; sox10: DsRed and col2a1aBAC: GFP; sox10: DsRed larvae followed [117], with ~130 μM of z-stacks collected every 10 or 12 minutes starting at 48 hpf. Skeletal preparations were photographed using a Leica DM2500 microscope. Image levels were adjusted in Adobe Photoshop CS6, with care taken to apply identical adjustments to images from the same data set and to avoid removing information from the image. To analyze changes in gene expression between 20 and 28 hpf and 28 and 36 hpf for the total arch and Edn1- and Notch-regulated gene lists, we first calculated the median and quartile values for each list. The full lists were then collectively compared first by a Kruskal-Wallis test and then pairwise by Mann-Whitney U tests, with the Bonferroni correction applied to an α-value of 0. 05 to account for multiple comparisons. Chi-square was used to compare the proportions of embryos showing different skeletal phenotypes in jag1b, barx1, and jag1b; barx1 mutants, with p < 0. 05 considered significant. JMP 7. 0 software (SAS) was used for statistical analysis. Numbers for each experiment are presented in S10 Table.

The exquisite functions of the vertebrate face require the precise formation of its underlying bones. Remarkably, many of the genes required to shape the facial skeleton are the same from fish to man. In this study, we use the powerful zebrafish system to understand how the skeletal components of the face acquire different shapes during development. To do so, we analyze a series of mutants that disrupt patterning of the facial skeleton, and then assess how the genes affected in these mutants control cell fate in skeletal progenitor cells. From these genetic studies, we found that several pathways converge to control when and where progenitor cells commit to a cartilage fate, thus controlling the size and shape of cartilage templates for the later-arising bones. Our work thus reveals how regulating the timing of when progenitor cells make skeleton helps to shape the bones of the zebrafish face. As mutations in many of the genes studied are implicated in human craniofacial defects, differences in the timing of progenitor cell differentiation may also explain the wonderful diversity of human faces.

Abstract Introduction Results Discussion Materials and Methods

medicine and health sciences condensation face bmp signaling condensed matter physics vertebrates animals notch signaling animal models osteichthyes immune receptor signaling developmental biology model organisms membrane receptor signaling embryos cartilage research and analysis methods embryology fishes connective tissue biological tissue head phase transitions physics signal transduction zebrafish anatomy cell biology biology and life sciences physical sciences cell signaling organisms

2016

Competition between Jagged-Notch and Endothelin1 Signaling Selectively Restricts Cartilage Formation in the Zebrafish Upper Face

Sensory processing in the brain includes three key operations: multisensory integration—the task of combining cues into a single estimate of a common underlying stimulus; coordinate transformations—the change of reference frame for a stimulus (e. g. , retinotopic to body-centered) effected through knowledge about an intervening variable (e. g. , gaze position); and the incorporation of prior information. Statistically optimal sensory processing requires that each of these operations maintains the correct posterior distribution over the stimulus. Elements of this optimality have been demonstrated in many behavioral contexts in humans and other animals, suggesting that the neural computations are indeed optimal. That the relationships between sensory modalities are complex and plastic further suggests that these computations are learned—but how? We provide a principled answer, by treating the acquisition of these mappings as a case of density estimation, a well-studied problem in machine learning and statistics, in which the distribution of observed data is modeled in terms of a set of fixed parameters and a set of latent variables. In our case, the observed data are unisensory-population activities, the fixed parameters are synaptic connections, and the latent variables are multisensory-population activities. In particular, we train a restricted Boltzmann machine with the biologically plausible contrastive-divergence rule to learn a range of neural computations not previously demonstrated under a single approach: optimal integration; encoding of priors; hierarchical integration of cues; learning when not to integrate; and coordinate transformation. The model makes testable predictions about the nature of multisensory representations. The brain often receives information about the same feature of the same object from multiple sources; e. g. , in a visually guided reach, both vision and proprioception provide information about hand location. Were both signals infinitely precise, one could simply be ignored; but fidelity is limited by irrelevant inputs, intrinsic neural noise, and the spatial precisions of the transducers, so there are better and worse ways to use them. The best will not throw away any information—in Bayesian terms, the posterior probability over the stimulus given the activities of the integrating neurons will match the corresponding posterior given the input signals. Encoding in the integrating neurons the entire posterior for each stimulus, and not merely the best point estimate, is crucial because this distribution contains information about the confidence of the estimate, which is required for optimal computation with the stimulus estimate [1], [2]. A sensible code will also “compress” the information—for example, by representing it in fewer neurons—otherwise the brain could simply propagate forward independent copies of each sensory signal. Psychophysical evidence suggests that animals—and therefore their brains—are indeed integrating multisensory inputs in such an “optimal” manner. Human subjects appear to choose actions based on the peak of the optimal posterior over the stimulus, given a variety of multisensory inputs [1], [3]–[7]. Prism and virtual-feedback adapation experiments [8]–[12] have demonstrated the plasticity of these multisensory mappings, and it is not likely limited to recalibration: Deprivation studies [13]; afferent re-routing experiments [14], [15]; the ability to learn novel, cross-modal mappings; and genetic-information constraints together suggest that integration is learned, with the organization of association cortices driven by sensory data. A plausible neural model of multisensory integration, then, must learn without supervision how to combine optimally signals from two or more input populations as well as a priori information, encoding both the most likely estimate and certainty about it—even when the relationship between the signal spaces is nonlinear (like retinotopic and proprioceptive-encoded hand location), and when their relationship is mediated by another variable (like gaze angle). Existing computational models of multisensory integration or cross-modal transformation neglect one or more of these desiderata (see Discussion). Here we show that the task of integration can be reformulated as latent-variable density estimation, a problem from statistics that can be implemented by a neural network, and the foregoing requirements thereby satisfied. The goal is to learn a data distribution (here, the activities of populations of visual and somatosensory neurons while they report hand location in their respective spaces) in terms of a set of parameters (synaptic strengths) and a set of unobserved variables (downstream, integrating neurons). In particular, we model the cortical association area with a restricted Boltzmann machine (RBM), an undirected generative model trained with a fast, effective Hebbian learning rule, contrastive divergence [16], [17]. By making the machine a good model of the distribution of the training data, learning obliges the downstream units to represent their common underlying causes—here, hand location. The same formulation turns out to be equally suited to coordinate transformation as well. We begin by examining the ability of our model to perform optimal multisensory integration, in the sense just described. We use our “standard” network, with a visible layer of 1,800 Poisson units, comprising two 30×30 input populations, and a hidden layer of half that number of Bernoulli units. We trained and tested this network on separate datasets, with stimuli chosen uniformly in the 2D space of joint angles (see Methods and Fig. 1B). We now relate our model to some familiar results from psychophysical investigations of multisensory integration. In the foregoing simulations, the input populations were driven by the same stimulus. The most common experimental probe of integration, however, is to examine the effects of a small, fixed discrepancy between two modalities—with, e. g. , prism goggles or virtual feedback [1], [3], [4], [18]–[20]. Integrated estimates tend to fall between the means of the discrepant inputs, revealing the relative weighting of the two modalities. The mean location of the integrated estimate therefore allows experimenters to assess integration without having to obtain reliable estimates of the error covariance. Notice this point will not necessarily lie along the straight line connecting the input means, since the sensory covariances need not be aligned [1]. To replicate these experiments, the trained network from Fig. 2 was tested on sets of “shifted” data in which joint angles had been displaced from their corresponding visual locations by a fixed quantity, the “input discrepancy, ” before being encoded in the prop population. To determine how large to make this discrepancy, we returned to the original, unshifted data. Although the average discrepancy between the two inputs in this data set is zero (as seen in the locations of the magenta and orange ellipses in Fig. 2), the noisy encoding renders the discrepancy on single trials non-zero, with the probability of finding such a discrepancy determined by the sum of the input covariances, . This quantity providing, then, a natural measure of discrepancy, each set of shifted data was created with an input discrepancy of standard deviations of, with. Note that large enables a further investigation—into the generalization of the trained network: The extent to which the RBM' s optimality is maintained as the input discrepancy grows indicates, qualitatively, the generalization powers of the machine on these data. Fig. 5A shows the error statistics for these testing datasets for several discrepancy magnitudes along a single direction (discrepancies along other directions, not shown, were qualitatively similar). Psychophysicists examine conditional errors, but again for generality we have averaged across stimulus locations to produce marginal errors. The RBM-based estimator (green) becomes noticeably suboptimal by 7. 5 standard deviations. Furthermore, the distribution of errors becomes distinctly non-normal for large input discrepancies, spreading instead over the arc connecting the centers of the input error distributions. This arc corresponds to the location of the optimal estimate for varying relative sizes of the input error covariances [1]. Whether such a pattern of errors is exhibited by human or animal subjects is an interesting open question. Another way of measuring machine generalization is to test its performance under gain regimes outside its testing data. Since no discrepancy is enforced between the modalities, biases should be zero. Performance should be approximately optimal in the training regime, where gains ranged from 12 to 18 spikes. And indeed, Fig. 5B shows that neither the error covariance (the relative shapes of the green and black ellipses) nor the bias (the relative positions of the green and black ellipses) are noticeably worse than in the training regime until the gain ratios (PROP/VIS) reach the extreme values on the plot. Finally, we examine machine performance under both input discrepancy and gain modulation, with a constant input discrepancy of 2. 5 standard deviations and various gain ratios Fig. 5C. The black and green dotted lines, nearly identical, track the movement of the error means of the optimal and RBM-based estimators, respectively. This reproduces the familiar psychophysical finding that varying the relative reliability of two discrepant inputs will bias downstream activity (sc. , behavior) toward the more reliable modality' s estimate [1]. We now examine some of the properties of the hidden units, especially those that electrophysiologists have focused on in multisensory neurons in rhesus macaques. One example of a statistical feature that is constant across trials is the prior distribution of the stimulus, which the network therefore learns to encode in its weights. Whether prior distributions in the brain are encoded in synaptic weights [46], [47], as a separate neural population [2], or something else again, remains an area of active research (see also Text S1). An interesting consequence of the present formulation is that it renders the gains random variables (see e. g. Fig. 1A), no less than the stimulus location; that is, they represent information that is not constant across trials. This has testable implications for multisensory populations. For an -dimensional stimulus, the posterior precision (inverse covariance) of the multisensory neurons is an symmetric matrix and therefore has independent entries. But if the precisions of the two input populations are each functions only of a single parameter (their respective gains, reflecting the confidence in each modality), then the multisensory activities need only encode two, rather than, numbers on each trial. Conversely, in the case of a one-dimensional stimulus, a population of multisensory neurons ostensibly need only encode the single value of the posterior variance, , but the density-estimation approach predicts that the hidden-unit activities on a given trial will nevertheless encode both of that trial' s input-population gains—and indeed they do in our model, albeit imperfectly (Fig. 3A). Testing these predictions experimentally would be straightforward—try to decode unisensory covariances from a multisensory population—but it has never been done. The question of whether cortical circuits learn to encode any posterior covariance information at all, as opposed to merely the point estimate that psychophysical experiments elicit, is itself a crucial, open one. Of course, in theory one can always compute a posterior over the stimulus given some population activities [48]; but whether the posterior conditioned on activities deep in the hierarchy matches that conditioned on the activity in early sensory cortices, as in our model, is unknown. Our model also predicts that such constancy would emerge during learning—which could be tested, for instance, by training an animal on a novel multisensory pairing (e. g. , audition and touch). That fewer units are used to represent the same information (half as many in our simple integration model; see Multisensory integration in the RBM), and that the maximum spike count of each hidden neuron is bounded by the maximum mean spike count of the inputs, constrains the amount of information that can be transmitted. This forces the hidden units to represent the information more efficiently—i. e. , to “integrate” it. In fact, without that constraint, no learning would be required to satisfy the information-retention criterion: A random weight matrix has rank almost surely, and the neuron nonlinearities are likewise invertible, so any random set of synaptic connections would suffice (since any invertible transformation is information-preserving). We chose to constrain the multisensory representational capacity, so that the synaptic connections form an matrix, which will not in general preserve stimulus information. One promising theoretical strategy would be to take “passing on all the information” as a given, and then to seek the set of constraints—fewest spikes [49], topography [50], fewest neurons, least processing time, computational efficiency [51], etc. —that yields the most biologically realistic activity patterns in the multisensory units. Multisensory integration was first considered from the standpoint of information theory and unsupervised learning in [52], and in a related work [50], and our approach is similar in spirit, but with important differences. Crucially, a different objective function was minimized: integration was achieved by maximizing mutual information between the hidden/output units of two neural networks, each representing a modality, forcing these units to represent common information, the latter additionally constraining topography. In our model, contrariwise, integration is enforced indirectly, by requiring a reduced number of (hidden) units to represent the information in two populations. This allows for greater generality since it does not require foreknowledge of which populations should be forced to share information: if the information in the input populations is redundant, it will be “integrated” in the hidden units, and conversely. More recently, the idea of treating multisensory integration as a density estimation problem has been proposed independently by [53], a complementary report that explores both cognitive and neural implications of this view, without proposing an explicit neural implementation. As in [50], [52], then, no attempt is made to employ biological learning rules. Most significantly, none of these models invokes the criterion for optimal integration that we have argued to be central—the correct posterior distribution over the stimulus given hidden-unit activities (, in the notation of this paper). This approach renders the combination of three signals of two independent causes—coordinate transformation—a matter simply of allowing another population to feed the hidden units; whereas the other models would require something more sophisticated. More recent models of multisensory integration or cross-modal transformation neglect some combination of the desiderata listed in the introduction. Basis-function networks with attractor dynamics [27], [30], [54] ignore prior distributions but more significantly require hand-wiring (no learning). The models of [46] and [47] extend these attractor networks to include the learning of priors, but even these must be hand wired and so are practical only for simple representations. Other models of learning [24]–[26], [55] disregard variance information, so that what is learned is essentially a mapping of means; nor, correspondingly, do they account for the learning of priors. The probabilistic population coding model [2] makes explicit the notion of encoding a posterior, but includes no model of learning. Finally, many authors have either anticipated [51], [56], [57] or explicitly proposed [58]–[60] that learning to process early sensory information might be viewed as forms of density estimation. Our work shows that the range of computations that can be assimilated to this statistical problem extends to the acquisition of two key operations for motor planning and control: multisensory integration, even when the underlying stimulus is distributed non-uniformly, and coordinate transformations; and further that these computations can be combined hierarchically, as is observed in the the neural circuits underlying these operations. Throughout, we work with the example case of integrating two-dimensional (2D) proprioceptive and visual signals of hand location, but the model maps straightforwardly onto any pair of co-varying sensory signals. These two signals report elbow and shoulder joint angles (prop,), and fingertip position in Cartesian space (vis, X), respectively. Choosing the forward kinematics, , to be invertible renders the variables isomorphic, so that we can refer generically to them as a “stimulus” (S), independent of space. The kinematics model for most of the results has joint ranges of (shoulder) and (elbow) and limb lengths of 12 (upper arm) and 20 (forearm) cm; see inset of Fig. 1A. The exception is Fig. 7C, D, in which a one-degree-of-freedom (1D) arm was used for simplicity: , with link length cm and joint range, and the position of the eye (eye, gaze angle). Below, we describe data generation from the 2D kinematics; the modifications for 1D are straightforward. Each training vector consists of a set spike counts, , generated by choosing a random stimulus (s, i. e. and x) and a random global gain for each modality (), and encoding them in a populations of neurons with Gaussian tuning curves () and independent Poisson spike counts—a “probabilistic population code” [2]: (2) as illustrated in Fig. 1A. Each gain, , can be thought of as the confidence in its respective modality, since the posterior covariance of a single, sufficiently large population, , is inversely proportional to its gain [2]. The tuning curves of each population are two-dimensional, isotropic, unnormalized Gaussians, whose width (variance) is, and whose centers form a regular grid over their respective spaces. To avoid clipping effects at the edges, the space spanned by this grid of neurons is larger than the joint space (or, for VIS, than the reachable workspace). Thus the grid consists of a central “response area” whose neurons can be maximally stimulated, and a “margin” surrounding it whose neurons cannot. The margin width is four tuning-curve standard deviations (), making spiking of putative neurons outside the grid extremely unlikely even for stimuli at the edge of the response area. In accordance with the broad tuning curves found in higher sensory areas and with previous models of population coding in multisensory areas [2], [27], tuning-curve widths were themselves chosen so that their full width at half maximum embraced one-sixth of the response area. The prior over the stimulus is either uniform or Gaussian in the space of joint angles. (Implementation of the Gaussian prior is detailed in Learning non-flat priors.) Since both dimensions of prop space are allotted the same number of neurons () and the tuning curves are isotropic and evenly spaced, but the physical ranges of these dimensions differ (and for the shoulder and elbow, resp.), the induced covariance in the population code is anisotropic, being more precise in elbow than shoulder angle. The nonlinearity of the forward kinematics likewise ensures anisotropy of; see Fig. 1A. This makes the problem more interesting, anisotropic covariances entailing, for example, optimal estimates that are not on the straight-line path between cue means (see e. g. Fig. 1 of [1]). The priors over the gains, and, which set the maximum mean spike counts, are independent and uniform between 12 and 18 spikes. Unless otherwise noted, gains in the testing data were drawn from the same distribution as the training-data gains. To show that the model works, we must compare two posterior distributions over the stimulus: the posterior conditioned on the input data, —i. e. the “true” or “optimal” posterior—and the posterior conditioned on the downstream/integrating units, (see The RBM, below). That comparison is easiest to make, and to exhibit, when the optimal posterior is as simple as possible—ideally, a Gaussian, which has only two nonzero cumulants, mean and covariance. With a flat or Gaussian prior over the stimulus, the probabilistic population code that we are using does indeed have an approximately normal posterior for a unimodal population [2]; but to guarantee this for two populations that are encoding the stimulus in different (i. e. , nonlinearly related) spaces, the unimodal posterior covariances (and) also must be small enough that typical errors lie within the linear regime of the arm kinematics (see Text S1). Given the gain (G) regime and the tuning-curve widths (), choosing neurons in the grid yields variances between 2 and 9 mm2 for the two populations, satisfying the requirement. These values are also comparable to empirical values for visual and proprioceptive localization variances from human psychophysics, 5 mm2 and 50 mm2, resp. [1]. These latter are in fact an upper bound, since they are with respect to behavior, the furthest downstream assay of certainty. In any case, we stress that this and other compromises of the population code with biological realism (uniform tiling of the stimulus space, identical tuning curves, etc.) serve to simplify the analyses and interpretation rather than reflecting any limitation of the neural-network model. Now, whereas a Gaussian posterior requires a flat or Gaussian prior, such a prior in prop space will induce an irregular prior in vis space (and vice versa; see again Fig. 1A) —so there can be a Gaussian posterior only in one space. Results are therefore computed in the space of the flat or Gaussian prior. Observing these constraints, the posterior cumulants can be written: (3a) (3b) (See Text S1 for a derivation.) Intuitively, the posterior precision (inverse covariance, Eq. 3) is a sum of three precisions: the prior precision, ; the weighted prop () tuning-curve precision, ; and the weighted vis () tuning-curve precision, . (Since the posterior is expressed over rather than X, the latter' s precision must be warped into -space by the Jacobian, , of the forward kinematics, which is evaluated at the center of mass of the proprioceptive population.) The weights are the total spike counts for each population, , . The posterior mean (Eq. 3b) is a normalized, weighted sum of three estimates: the prior mean, ; the center of mass of the population, ; and the (transformed) center of mass of the population, . The weights are the three precisions. The center of mass, with the preferred stimulus, is likewise intuitive, being the maximum-likelihood estimate of the stimulus for a single population [61]. The nonlinearity (cosine) in the 1D “coordinate-transformation model” (Fig. 7C, D), , likewise allows the posterior to be normal in only one space. Since two of the variables live in Cartesian space—X (vis) and E (eye) —and only (prop) lives in joint-angle coordinates, we chose uniform priors over the former, sampling them between and, so that their sum never exceeded the bounds of the joint range (see above, Input-data generation). Zero in this space corresponds to hand position at the center of fixation for, and to central fixation for E. The addition of a non-flat prior (Fig. 6) will only have an appreciable effect on the posterior if the width of the prior distribution is comparable to that of the likelihoods, i. e. the single-modality localization covariances. The covariance of the prior was therefore constructed so that, along both dimensions, the extreme angles were 150 standard deviations apart—a reasonable prior distribution, perhaps, after extensive training on a reaching task to a single target location [47]. Using more realistic, broader priors would require relaxing the constraint that the optimal posterior distribution over the stimulus be Gaussian—which again we insist upon only for ease of analysis. The neural circuit for sensory integration was modeled as a restricted Boltzmann machine, a two-layer, undirected, generative model with no intralayer connections and full interlayer connections (Fig. 1A, bottom right) [17], [62]. The input layer (R) consists of Poisson random variables, whose observed values are the population codes just described. The hidden-layer units (V) are binary, indicating whether or not a unit spiked on a given trial, making them Bernoulli random variables. Unless otherwise noted in the results, the number of hidden units in the model is equal to half the number of input units, i. e. the number of units in a single input population—thus forcing the model to represent the same information in half the number of neurons. During RBM training [17], [62], input and hidden units reciprocally drive each other through the same weight matrix: (4a) (4b) which corresponds to Gibbs sampling from the joint distribution represented by the machine. Here is the entry of the vector z; and are, respectively, the vectors of biases for the hidden and observed units; is the matrix of synaptic strengths; and is the logistic (sigmoid) function. (The lack of intralayer connections is what allows the entire joint to be sampled in just two steps.) As in a standard stochastic neural network, each unit' s mean activity is a nonlinear transformation of a weighted sum of its inputs. To ensure that this mean is in the support of its associated exponential-family distribution, the nonlinearities are chosen to be the inverse “canonical links” [63]: the logistic function for the Bernoulli hidden units, and the exponential function for the Poisson input units. (Technically, the use of Poisson input units makes the model an “exponential family harmonium” [62] rather than a restricted Boltzmann machine, which would have all Bernoulli units.) The unit' s activity (presence of a spike, or spike count) is sampled from this mean. Weights and biases were initialized randomly, after which the networks were trained on batches of 40,000 vectors, with weight changes made after computing statistics on mini-batches of 40 vectors apiece. One cycle through all 1000 mini-batches constitutes an “epoch, ” and learning was repeated on a batch for 15 epochs, after which the learning rates were lowered by a factor of. This process was repeated a total of seven times, i. e. 90 epochs, after which learning was terminated. (The number of epochs and the learning-rate annealing schedule were determined empirically.) Weight and bias changes were made according to one-step contrastive divergence [16], [17]: (5) where the circumflexes differentiate the zeroth (no hat) and first (hat) steps of Gibbs sampling. That is, the input data (r) are propagated up into the multisensory (hidden) layer (v), back down into the input units (), then back up into the multisensory neurons (); see Fig. 1B. This is repeated for all the data (that is, for each drawn from Eq. 2, for each stimulus and set of gains drawn from and). The change in the weight connecting neuron to neuron is thus proportional to the difference between the first and second pair of correlations between them—a Hebbian and an anti-Hebbian term. This rule approximates gradient descent on an objective function for density estimation (Hinton' s “contrastive divergence” [17], or alternatively “probability flow” [64]). Although this specific learning rule has not been documented in vivo, it is constructed entirely of components that have been: change in firing rate based on (local) correlations between pre- and postsynaptic spike counts. Anti-Hebbian learning has been observed in a neural circuit [65], albeit not in mammalian cortex, and plausible cellular mechanisms for it have been described [66]. After training, learning was turned off, and the network was tested on a fresh batch of 40,000 data vectors (Fig. 1B): stimuli were again drawn uniformly from the grid of joint angles, and the corresponding spike counts simulated by drawing from the two populations of Gaussian-tuned, Poisson neurons. For each input vector, hidden-layer activities were computed by drawing 15 sample vectors (from) and averaging them. Since the input gains are between and, and assuming that hidden and input units integrate information over the same-sized time window from the past, this implies that hidden neurons fire no faster than input neurons—which would otherwise constitute a violation of the information bottleneck. This is essential for our task, since we require an efficient coding, not merely a different one. For each trial, decoding the hidden vector consists of estimating from it the mean and covariance of the optimal posterior —that is, all the information in the network about the stimulus. Generally, finding a good decoder can be hard; but because the network is a generative model, we can use its generative (hidden-to-input) weights to turn the hidden vector back into expected input spike counts () —which we know how to decode: Eq. 3. In practice, it often turns out that the weighted sum in Eq. 3b is unnecessary: the center of mass from a single (updated) population suffices. When showing results in joint angles, we take the center of mass of the prop population; likewise for Cartesian space and vis. Also, reconstruction of the total spike counts was mildly improved by first mapping them to the true (input) total spike counts via a standard neural network; in cases where this final step was applied (Fig. 3A), training and testing used different data. The posterior covariances used in Fig. 3B–D, however, did not use any such trained decoder; they were reconstructed just as the posterior means were, i. e. by using the generative weights and then applying equation Eq. 3a.

Over the first few years of their lives, humans (and other animals) appear to learn how to combine signals from multiple sense modalities: when to “integrate” them into a single percept, as with visual and proprioceptive information about one' s body; when not to integrate them (e. g. , when looking somewhere else); how they vary over longer time scales (e. g. , where in physical space my hand tends to be); as well as more complicated manipulations, like subtracting gaze angle from the visually-perceived position of an object to compute the position of that object with respect to the head—i. e. , “coordinate transformation. ” Learning which sensory signals to integrate, or which to manipulate in other ways, does not appear to require an additional supervisory signal; we learn to do so, rather, based on structure in the sensory signals themselves. We present a biologically plausible artificial neural network that learns all of the above in just this way, but by training it for a much more general statistical task: “density estimation”—essentially, learning to be able to reproduce the data on which it was trained. This also links coordinate transformation and multisensory integration to other cortical operations, especially in early sensory areas, that have have been modeled as density estimators.

Abstract Introduction Results Discussion Methods

circuit models neural networks computational neuroscience psychophysics sensory systems biology computational biology sensory perception neuroscience learning and memory coding mechanisms

2013

Learning Multisensory Integration and Coordinate Transformation via Density Estimation

Blocking Plasmodium falciparum transmission to mosquitoes has been designated a strategic objective in the global agenda of malaria elimination. Transmission is ensured by gametocyte-infected erythrocytes (GIE) that sequester in the bone marrow and at maturation are released into peripheral blood from where they are taken up during a mosquito blood meal. Release into the blood circulation is accompanied by an increase in GIE deformability that allows them to pass through the spleen. Here, we used a microsphere matrix to mimic splenic filtration and investigated the role of cAMP-signalling in regulating GIE deformability. We demonstrated that mature GIE deformability is dependent on reduced cAMP-signalling and on increased phosphodiesterase expression in stage V gametocytes, and that parasite cAMP-dependent kinase activity contributes to the stiffness of immature gametocytes. Importantly, pharmacological agents that raise cAMP levels in transmissible stage V gametocytes render them less deformable and hence less likely to circulate through the spleen. Therefore, phosphodiesterase inhibitors that raise cAMP levels in P. falciparum infected erythrocytes, such as sildenafil, represent new candidate drugs to block transmission of malaria parasites. Recent renewed emphasis on the eradication of malaria has highlighted the need for novel interventions to target the parasite during transmission from the human host to the mosquito. Drug treatments to clear asexual blood stage parasites (that cause pathology) do not kill mature gametocytes and therefore allow transmission to continue [1]. Transmission of malaria parasites relies on the sexual stages, termed gametocytes that circulate in the peripheral blood and are taken up by Anopheles mosquitos during a blood meal. For Plasmodium falciparum, the causative agent of the most severe form of human malaria, gametocyte maturation requires about 10 days and is divided in five morphological stages [2]. During this period, immature gametocyte-infected erythrocytes (GIE) sequester in internal organs such as bone marrow and spleen [3–6]. Sequestration mechanisms of GIE are still unknown, although failure of immature GIE to adhere to endothelial cell lines in vitro [7], and absence on their surface of parasite structures allowing cytoadhesion of asexual stages [8], suggest that GIE-host interactions are unlikely to be mediated by cytoadhesion. Recent evidence rather suggests that GIE biomechanical properties may play an important role in this process [9]. At maturation GIE are released into the blood circulation, where they can persist for several days [10], thus increasing the likelihood of parasites being taken up during a mosquito blood meal and ensuring transmission. This remarkable ability of mature GIE to circulate through the spleen is due to the important deformability that they acquire during the transition between stages IV to V [9,11,12]. By contrast, immature GIE are particularly stiff, which likely contributes to their sequestration by mechanical retention [9]. Therefore, modulation of GIE mechanical properties plays a key role in their microcirculatory behaviour and it has been proposed that interfering with mature GIE filterability through spleen capillaries may represent a novel way to block parasite transmission [4,9]. However, mechanisms mediating the switch in GIE deformability late in the maturation process are still elusive. The disassembly of the microtubule subpellicular network subtending the trilaminar membrane structure in the transition from stage IV to stage V gametocytes probably contributes to this process [12–15]. The switch in deformability is also linked to the de-association of the parasite-derived STEVOR proteins from the infected erythrocyte membrane [9]. These processes must be tightly controlled and signalling likely plays a regulatory role. In uninfected erythrocytes, changes in phosphorylation status, including phosphorylation by cAMP-dependent kinase A (PKA), are known to regulate mechanical properties of the erythrocyte membrane [16,17]. For instance, phosphorylation of band 4. 1 by PKA may be central to the regulation of erythrocyte cytoskeletal organization and membrane mechanical properties [18]. PKA phosphorylation of dematin has also been shown to modulate the association between actin and spectrin in the erythrocyte cytoskeleton [19,20]. In infected erythrocytes, PKA activity results from both the human and the parasite enzymes [21]. During the parasite’s life cycle, plasmodial PKA activity is implicated in a wide variety of processes including P. berghei sporozoite motility and liver cell invasion [22], P. falciparum erythrocyte invasion by merozoites [23,24], or modulation of infected erythrocyte membrane permeability [25]. So far, there is no evidence for a regulatory role for cAMP-signalling during sexual development. However, adenylate cyclase alpha (PfACα) is highly expressed in gametocytes [26], and PKA activity is reportedly higher in gametocyte-producing parasites compared to parasites defective in gametocyte production [27], suggesting a potential role for cAMP-signalling in sexual development. Here, we have investigated the role of cAMP-signalling in modulating GIE mechanical properties. Using both genetic and pharmacological manipulation of cAMP signalling in conjunction with the microsphiltration method to assess the ability of GIE to circulate through inter-endothelial splenic slits, we show that a decrease in cAMP levels increases mature GIE deformability, and conversely, increasing cAMP levels increases GIE stiffness. These findings provide the proof of principle that molecules targeting phosphodiesterases (PDE) represent a novel drug class capable of blocking malaria transmission. To investigate whether PKA activity modulates GIE mechanical properties, we assessed the filterability of GIE using the microsphiltration method, which mimics the physical constraints experienced by infected erythrocytes in the splenic microcirculation [28,29]. In this system, increased retention rates correspond to decreased erythrocyte deformability and impaired filterability. We treated stage III GIE with KT5720 and H89, two independent and widely used PKA inhibitors that also block a few other kinases [30,31] and that have already been shown to inhibit PKA activity in P. falciparum [21,24,32]. Before incubation with these inhibitors, approximately 94% of stage III GIE and 30% of stage V GIE were retained on the microspheres (Fig 1A and S1 Fig), confirming the retention rates previously observed [9]. Importantly, incubation with H89 and KT5720 significantly decreased the retention rates of stage III GIE (P = 3. 10e-6 and 6. 10e-6 for H89 and KT5720, respectively; Fig 1A) consistent with PKA activity contributing to immature GIE stiffness, whereas incubation of stage V GIE with H89 did not alter their retention rates (S1 Fig). By contrast, stage III GIE retention rates were not affected upon incubation with compound 2, a highly selective inhibitor of apicomplexan cGMP-dependent protein kinase (PKG) [33], or with GGTI-298, an inhibitor of the cAMP-effector exchange protein activated by cAMP (EPAC) [34]. Furthermore, GIE filterability was not affected upon incubation with PKI-m, a membrane permeable inhibitor of the human PKA (PKA) that is a poor inhibitor of parasite PKA (PfPKA) [32]. This suggests that immature GIE filterability is modulated by PfPKA, and not by human erythrocyte PKA. To confirm this notion, we measured retention rates of a transgenic parasite that exhibits a down-regulation in PfPKA activity due to episomal overexpression of the regulatory (PfPKA-R) subunit of PfPKA (pHL-pfpka-r) [25]. cAMP binding to PKA-R liberates the catalytic subunit (PKA-C) from inactive R/C complexes and over-expression of PKA-R acts as a cAMP sink dampening complex dissociation, so decreasing PfPKA activity. Immunoblotting and immunostaining of gametocytes with specific antibodies indicated that PfPKA-R expression was significantly increased in pHL-pfpka-r transgenic parasites (Fig 1C, 1D and 1E). We note that two PfPKA-R specific bands are identified in gametocytes compared to a single band in schizonts [35], indicating that the R subunit undergoes translational modification during gametocytogenesis. Microsphiltration analysis of pHL-pfpka-r immature GIE showed a significant decrease in retention rates, similar to that observed upon incubation of wild-type GIE with H89 (P = 0, Fig 1B). To confirm that decreased retention rates were due to dampened cAMP-signalling, we measured deformability of pHL-pfpka-r immature GIE following incubation with the cell permeable, phosphodiesterase resistant cAMP analogue, 8Br-cAMP; raising cAMP levels restored retention rates to wild-type phenotype (Fig 1B). Loss of episomally derived PfPKA-R over-expression in pHL-pfpka-r transgenic parasites should result in a regain in PfPKA activity and restoration in the levels of retention associated with endogenous PfPKA expression. To promote shedding of the episome-encoded R subunit, the pHL-pfpka-r transgenic line was cultured for several generations without pyrimethamine selection required to retain the episome, leading to a new line called pHL-pfpka-r-wt. Immunoblotting and immunostaining of stage III GIE with specific anti-PfPKA-R antibodies confirmed that PfPKA-R expression in pHL-pfpka-r-wt had reverted to that of wild type levels, consistent with loss of the episome harboring the pfpka-r expression cassette (Fig 1C, 1D and 1E). Retention rates of pHL-pfpka-r-wt immature GIE were similar to wild-type GIE, indicating that the retention phenotype mediated by over-expressed PfR-induced down-regulation of PfPKA activity had reverted (Fig 1B). These results indicate that PfPKA-mediated phosphorylation contributes to immature GIE stiffness. To further demonstrate the contribution of PfPKA to immature GIE stiffness we probed membrane extracts of stage III and stage V GIE using a monoclonal antibody specific for canonical phospho-PKA sites (RRXS*/T*) (Fig 2A and 2B). The intensity of phosphorylation was 2-fold lower in stage V GIE compared to stage III, indicating that membrane components were less phosphorylated by PKA in mature than in immature gametocytes. At least five proteins displayed a higher degree of PKA site phosphorylation in stage III compared to stage V, suggesting that these PKA substrates are potentially involved in mediating the membrane rigidity phenotype (Fig 2A and 2B). Reduced PKA phosphorylation at stage V indicates a drop in cAMP levels accompanies the switch in deformability that occurs at the transition between immature and mature stages. The degree of cAMP-mediated phosphorylation can be increased by treatment with calyculin A, a serine/threonine phosphatase inhibitor (known to inhibit P. falciparum phosphatase-1-like activities [36,37]) that diminishes dephosphorylation of PKA substrates. Incubation of stage V GIE with calyculin A increased phosphorylation intensity at 2-fold to levels observed in stage III (Fig 2A and 2B). Consistently, incubation of stage V GIE with calyculin A markedly impaired their filterability by increasing the retention rates in microspheres up to 90% (P = 0. 0038), whereas it did not significantly affect filterability of uninfected erythrocytes (P = 0. 7; Fig 2C). To validate that increased retention rates triggered upon incubation with calyculin A corresponded to a decrease in GIE deformability, we visualized the shape of GIE as they flowed through the matrix by adding a paraformaldehyde-fixation step to the microsphiltration experiment (Fig 2D and 2E). 30% of untreated GIE maintained their original shape, whereas 70% displayed a twisted and deformed shape, likely reflecting their ability to squeeze and slide between microspheres [9]. Upon incubation with calyculin A the proportion of deformed GIE decreased from 70% to 40% (Fig 2D and 2E). To rule out any cytotoxic effect of calyculin A on gametocytes that could un-specifically result in higher cell rigidity, we measured parasite lactate dehydrogenase activity immediately and 72 h after calyculin A treatment and validated that gametocyte viability was not affected (Table 1). Manipulating the phosphorylation status using calyculin A clearly affects GIE mechanical properties and given the documented effect of calyculin A on PKA activity in other systems [37], it is consistent with PfPKA activity in mediating gametocyte deformability. The above results suggest that changes in cellular cAMP levels influence GIE deformability via PKA activation. We first measured cAMP concentrations in MACS-purified GIE at immature and mature stages. Intracellular levels of cAMP decrease approximately five-fold in stage V GIE compared to stage III GIE (P = 0. 008; Fig 3A), while PfPKA-R levels are unaltered between immature and mature stages (Fig 3B). Thus, increased deformability of stage V GIE is accompanied by reduced PKA activity due to a decrease in cAMP levels. This notion was underscored by increasing cAMP levels via addition of 8Br-cAMP and measuring the filterability of stage III compared to stage V gametocytes. Upon incubation with 8Br-cAMP, retention rates of stage III GIE were not modified, whereas those of stage V GIE were proportionally augmented with increasing concentrations of 8Br-cAMP. High concentrations of 8Br-cAMP did not affect filterability of uninfected erythrocytes, indicating that 8Br-cAMP specifically affects P. falciparum mature GIE (Fig 3C). Analysis of Giemsa stained smears of stage V GIE upstream and downstream of the microsphere matrix showed unaltered male: female ratios indicating that both male and female gametocytes use cAMP to regulate infected cell deformability. The drop in cAMP concentration in stage V GIE might be a result of either reduced cAMP synthesis by adenylate cyclases, or increased degradation by phosphodiesterases (PDEs). In P. falciparum, four genes encode PDEs: PfPDEα (PF3D7_1209500. 1), PfPDEβ (PF3D7_1321500. 1), PfPDEγ (PF3D7_1321600) and PfPDEδ (PF3D7_1470500) [38,39]. We performed real-time RT-PCR to quantify mRNA levels for all PDEs in asexual blood-stages as well as in immature and mature gametocytes. PfPDEα and PfPDEβ are mainly expressed in asexual blood-stages, PfPDEγ is minimally expressed in all blood-stages, whereas PfPDEδ is highly expressed in stage V gametocytes (Fig 4A). Importantly, mRNA levels of PfPDEδ increase approximately two- to three-fold in stage V compared to stage III gametocytes. These results are consistent with expression data available at PlasmoDB (http: //www. plasmodb. org/), where PfPDEδ is annotated as almost exclusively expressed in stage V gametocytes [40,41]. To determine whether PfPDEδ is involved in triggering the switch in deformability observed in stage V GIE, we analysed retention rates for mature GIE from transgenic parasites in which the PfPDEδ gene had been deleted [42]. We first measured cAMP concentration in MACS-purified stage III and stage V GIE and found loss of PfPDEδ activity led to a two-fold increase in cAMP levels in both immature and mature GIE (Fig 4B). Microsphiltration analysis of mature GIE from the PfPDEδ-mutant line showed a significant increase in retention rates (Fig 4C), indicating that this phosphodiesterase participates in regulating cAMP levels in mature gametocytes and hence GIE deformability. As an alternative way to investigate the effects of raising cAMP levels in GIE, we used pharmacological agents such as forskolin, an activator of mammalian adenylate cyclase, and zaprinast, an inhibitor of PDEs. Although zaprinast is well known as an inhibitor of mammalian PDE5, a cGMP phosphodiesterase, it is known to increase cAMP levels in human erythrocytes [43], and can inhibit both cAMP- and cGMP-PDE activities in P. falciparum (Table 2) [39]. To validate the effect of these molecules on intracellular levels of cAMP, we measured its concentration in MACS-purified stage V GIE and found a two-fold increase in cAMP levels in cells treated with either compound (P = 0. 029 and 0. 024 for forskolin and zaprinast, respectively; Fig 5A). Importantly, incubation of stage V GIE with either forskolin or zaprinast markedly increased microsphere retention rates by up to 82% (P = 0) and 86% (P = 0. 0002), respectively. Both reagents showed no significant effect on filterability of uninfected erythrocytes (P = 0. 49 and 0. 20 for forskolin and zaprinast, respectively; Fig 5B). Upon incubation with forskolin or zaprinast, the proportion of paraformaldehyde-fixed mature GIE that exhibit a deformed shape as they flowed through the matrix was decreased to 45% and 31%, respectively, compared to 70% of untreated cells (Fig 5C). The ensemble indicates that pharmacological agents that raise levels of cAMP affect mature GIE deformability impairing their ability to pass through an in vitro model for splenic filtration. PDEs have been well studied as potential drug targets in relation to numerous diseases. For example, the PDE5 and PDE6 inhibitor sildenafil (Viagra) is a widely used to treat erectile dysfunction. As a first step to address the potential of sildenafil to impair the circulation of P. falciparum mature gametocytes in humans and thereby block malaria transmission to mosquitoes, we measured the cAMP concentration in MACS-purified stage V GIE following incubation with sildenafil. We found that 100 μM sildenafil triggered an approximately four-fold increase of intracellular levels of cAMP, leading to levels similar to those of immature GIE (P = 0. 0068; Fig 6A). An increase in cGMP levels was also measured following sildenafil treatment (S2 Fig), consistent with our observations that both zaprinast and sildenafil can inhibit both cAMP and cGMP hydrolytic activities (Table 1). Upon incubation with different concentrations of sildenafil from 10 nM to 100 μM, retention rates of stage V GIE increased in a dose-dependent manner and reached 92% retention at 100 μM (Fig 6D). At 100 μM the proportion of paraformaldehyde-fixed GIE that exhibit a deformed shape as they flowed through the matrix decreased to 29%, compared to 70% of untreated cells (Fig 6B and 6C). Importantly, mature GIE exhibited greatly reduced filterability with more than 75% retention with 1 μM sildenafil (666. 7 ng/ml), which approximately corresponds to the reported peak serum concentration reached in humans after 60 min following 100 mg oral dose (Cmax 440 ng/ml) [44,45]. The filterability of uninfected erythrocytes was not significantly affected at these sildenafil concentrations indicating that it specifically affects infected erythrocytes (P = 0. 7012; Fig 6D). A novel paradigm that recently emerged postulates that the dynamics of immature P. falciparum GIE sequestration in the extravascular spaces of bone marrow and release of mature forms into peripheral circulation through bone marrow endothelial slits depends on an increase in GIE deformability at the transition from immature to mature stages [6,9]. The ability of mature P. falciparum GIE to circulate and pass through narrow blood capillaries and splenic slits depends on the deformability of both the parasites and the infected erythrocytes. Here, we have provided new mechanistic insight into the regulation of GIE deformability, and importantly, by demonstrating that GIE deformability is altered by zaprinast and sildenafil treatments we provide the proof of concept that PDE inhibitors could open new avenues towards the design of malaria transmission-blocking drugs. Stiffer immature gametocyte stages have high levels of cAMP, but become more deformable upon inhibition of PfPKA either pharmacologically or by overexpression of the PfPKA regulatory subunit. Deformability was not affected upon incubation with GGTI-298, making unlikely that cAMP regulates GIE stiffness via activation of PfEPAC [35]. So one mechanism by which raising cAMP levels induce GIE stiffness is via activation of PfPKA and its possible substrates could be parasite proteins involved in mediating the deformability of the parasite itself, a parasite protein (s) that remodels the erythrocyte membrane, or a protein (s) of erythrocyte origin. For instance, glideosome associated protein (GAP) 45 and myosin A, which are components of the microtubule subpellicular network subtending the trilaminar membrane structure in immature gametocytes, have been identified as PKA substrates in asexual stages [12,46]. In uninfected erythrocytes PKA-mediated phosphorylation of dematin and protein 4. 1 decreases their ability to promote spectrin binding to F-actin, therefore modulating mechanical properties of the erythrocyte [18,20]. In GIE these proteins may be phosphorylated by PfPKA in immature stages and may become dephosphorylated by parasite phosphatases in deformable stage V gametocytes. The effect on deformability of the phosphatase inhibitor calyculin A is consistent with dephosphorylation playing a role. The detection of five bands showing a decrease in PKA site phosphorylation in mature stages suggests that more than one PKA substrate likely contributes to regulation of GIE deformability. Besides proteins from the erythrocyte cytoskeleton, parasite-encoded STEVOR proteins are also attractive candidates to be PfPKA substrates. They have been shown to impact on the deformability of erythrocytes infected with sexual and asexual P. falciparum parasites, and the switch in deformability during gametocyte maturation is linked to a de-association of STEVOR proteins from the erythrocyte membrane in mature stages [9,47]. Interestingly, three serine residues and one threonine residue conserved across the entire STEVOR family are predicted to be PKA sites (cbs. dtu. dk/services/NetPhosK). Thus, phosphorylation of STEVOR proteins and its impact on STEVOR subcellular localisation and/or their interaction with erythrocyte cytoskeleton could be dependent on PKA-mediated phosphorylation and future studies will address this point. In addition, the mature parasite-infected erythrocyte surface antigen (MESA) is also expressed in sexual stages and interacts with the erythrocyte cytoskeleton, where it binds to protein 4. 1 [48,49]. The presence of four classical PKA phosphorylation sites in MESA suggests that this protein might also be a target of PKA-mediated changes in GIE mechanical properties. PfPKA-mediated phosphorylation of proteins associated with the erythrocyte cytoskeleton, or parasite proteins located in the erythrocyte membrane implies that the parasite kinase might be exported into the erythrocyte cytosol. PfPKA-R and PfPKA-C sequences do not have a recognizable PEXEL/HT motif and whether they are secreted remains hypothetical; however, there are proteins that lack a clear secretion signature that make up the non-PEXEL exportome [50]. Furthermore, in the absence of compelling secretion data for PfPKA-R and PfPKA-C, we entertain the possibility that PfPKA effects on erythrocyte membrane deformability may be mediated indirectly, through PfPKA-dependent phosphorylation of other, as yet to be identified, secreted parasite effectors. We found that the overall GIE cAMP concentration drops at the transition between immature and mature gametocyte stages, concomitant with the switch in deformability necessary for transmission. We established that raising cAMP levels in stage V GIE with forskolin, zaprinast or sildenafil rendered them stiff, like immature GIE. Zaprinast and sildenafil treatment of stage V GIE clearly led to a rise in cAMP levels and sildenafil also induced a small increase in the amount of cGMP (S2 Fig). Consistent with these observations, retention rates of stage V GIE proportionally augmented with increasing concentrations of 8Br-cGMP (S2 Fig). However, cGMP levels do not change between stage III and V gametocytes [42] and moreover, inhibition of the cGMP-dependent protein kinase (PfPKG) with the specific inhibitor compound 2 did not alter stage III GIE retention rates at a concentration 20-fold higher than required to inhibit erythrocyte invasion by merozoites [33]. Nonetheless, cGMP could impact indirectly on GIE deformability via crosstalk with cAMP signalling, as already reported for uninfected erythrocytes [43]. These observations validate the use of sildenafil to increase mature GIE stiffness. Several PDE inhibitors have been developed and used as therapeutic agents, and PDE5 has received considerable attention over the last 10 years, with three selective inhibitors now on the market (sildenafil, vardenafil, and tadalafil) [45]. In humans sildenafil acts by inhibiting both PDE5 and PDE6, and we found that in P. falciparum-infected erythrocytes this inhibitor clearly led to a rise in cAMP levels. Consistently, it increased GIE retention in an in vitro model for splenic filtration, demonstrating that administration of PDE inhibitors could be a new way to block parasite transmission to mosquitoes. Interestingly, previous studies with PfPDEδ-mutant line pointed to an essential role for cGMP-signalling in P. falciparum gametogenesis and ookinete formation in the mosquito vector [42,51]. This suggests that inhibition of plasmodial PDEs with sildenafil or derived analogues has the potential to raise both cAMP and cGMP levels resulting in a block in both gametocyte transmission via changes in GIE deformability and on ookinete development in mosquitoes. Our observations provide an opportunistic approach towards the discovery of new malaria transmission-blocking drugs, by taking advantage of the wealth of clinical data available for sildenafil, which been approved by the Food and Drugs Administration and is widely used in humans with little side effects to treat erectile dysfunction. In opposition to the Ehrlich’s “magic bullet” which consists of targeting pathways that are essential for parasites, but absent in humans, this strategy, referred to as “inverted silver bullet” [52], open new avenues towards the design of novel interventions to halt the spread of malaria to humans. The P. falciparum clonal lines B10 and 3D7 (clones of NF54), and the transgenic lines pHL-pfpka-r, and PfPDEδ- clone 4 have been described elsewhere [25,42,53]. Parasites were cultivated in vitro under standard conditions using RPMI 1640 medium supplemented with 10% heat-inactivated human serum and human erythrocytes at a 5% haematocrit [54]. Synchronous production of highly specific gametocytes stages was achieved according to described protocol [55]. For the isolation of gametocytes, culture medium was supplemented with 50mM (final concentration) N-acetylglucosamine (GlcNAc) from day 0 onwards and medium replacement was continued for 5 days to eliminate the asexual stages. Gametocyte preparations were enriched in different experiments by magnetic isolation using a MACS depletion column (Miltenyi Biotec) in conjunction with a magnetic separator. To measure cAMP and cGMP levels, 6. 106 GIE were purified by magnetic isolation and the cell pellets were incubated with drugs at 37°C, centrifuged at 1,500 x g for 5 min and washed with PBS. Sample diluent containing detergents to lyse the cells, inactivate endogenous phosphodiesterases and stabilize the cyclic nucleotides was added to the pellet for 10 min at room temperature, as described in the kits protocol (FluoProbes, powered by Interchim). To avoid eventual interference with the assay stemming from haemoglobin contained in the red blood cell lysate, proteins were precipitated with 5% trichloroacetic acid (TCA) for 10 min on ice, and the precipitate was removed by centrifugation at 1,500 x g for 10 min. The supernatant was carefully removed and transferred to a clean tube. TCA was removed by four successive extractions with water-saturated Diethyl ether. The cyclic nucleotides content was measured after acetylation using a commercially available cAMP High Sensitivity Chemiluminescent Assay Kit or cGMP High Sensitivity Chemiluminescent Assay Kit (FluoProbes, powered by Interchim), cyclic nucleotides content was expressed as picomoles of cAMP or cGMP per 4. 107 cells. Synchronized cultures containing 1 to 5% GIE were incubated 15 min to 2 h at 37°C with 10 nM to 150 μM 8Br-cAMP (8-Bromide-cyclic adenosine-monophosphate), 100 to 150 μM forskolin, 65 μM zaprinast, 50 nM Calyculin A, 10 nM to 100 μM sildenafil citrate, 10 μM compound 2 (4-[7-[ (dimethylamino) methyl]-2- (4-fluorphenyl) imidazo[1,2-a]pyridin-3-yl]pyrimidin-2-amine), 10 μM H89,10 μM KT5720,10 μM PKI-m (Protein Kinase Inhibitor myristoylated), 10 μM GGTI (GeranylGeranylTransferase I) 298 trifluoroacetate salt hydrate. None of the compounds reported above, except Compound 2, PKI-m and KT5720, affected stage V GIE viability measured as parasite lactate dehydrogenase (pLDH) levels [56]. Stage V GIE were treated with compounds at the highest dose, for the indicated times, washed to remove the compounds and assayed for pLDH activity both immediately and after 72 h incubation at 37°C. All reagents were purchased from Sigma-Aldrich or Euromedex, except Compound 2 that was provided by DB. Calibrated metal microspheres (96. 50% tin, 3. 00% silver, and 0. 50% copper; Industrie des Poudres Sphériques) with 2 different size distributions (5- to 15-μm-diameter and 15- to 25-μm-diameter) composed a matrix used to assay infected erythrocyte deformability under flow, as described [28,29]. Suspensions of synchronized cultures containing 1% to 5% GIEs were perfused through the microsphere matrix at a flow rate of 60 mL/h using an electric pump (Syramed _sp6000, Arcomed_ Ag), followed by a wash with 5 mL of complete medium. The upstream and downstream samples were collected and smeared onto glass slides for staining with Giemsa reagent, and parasitaemia was assayed by counting 2000 erythrocytes to determine parasite retention versus flow-through. Retention rates of uninfected erythrocytes were monitored after labelling a subpopulation of erythrocytes with PKH67 (Sigma-Aldrich) according to manufacturer’s instructions. The proportion of labeled erythrocytes in upstream and downstream samples was determined by fluorescence microscopy using a Leica DM 5000 B at 100X magnification. To visualize GIE shape during their flowing through the matrix, 1 mL of PBS/4% paraformaldehyde was added after perfusion of the GIE-containing culture on the microsphere matrix. After 5 min of incubation, fixed GIEs were separated from the microspheres by a 3-step decantation procedure, and GIE morphology was observed on a glass slide by light microscopy using a Leica DM 5000 B at 100X magnification. Microsphiltration experiments were performed in triplicate and 100 cells were counted per experiment. PDE activity in native parasite fractions was measured using a modification of a previously published method [57]. Briefly, parasites were frozen in liquid nitrogen and stored at -80°C until use. Parasites were resuspended in 500 μl lysis buffer (20 mM hepes and 250 mM sucrose, pH 7. 0), subjected to 5 cycles of freeze-thaw in liquid nitrogen and pelleted at 100,000 g for 30 min. Particulate fractions were resuspended in lysis buffer containing EDTA-free protease inhibitors (Roche). PDE assays were carried out in triplicate wells of a 96-well plate in the presence of [3H]-labelled cGMP or cAMP (GE Healthcare) for 30 min at 37°C. Reactions were terminated by boiling the plate for 1 min, followed by a 3 min centrifugation at 900 g. 1 unit of alkaline phosphatase was added to each well and incubated for 30 min at 37°C. [3H]-labelled guanosine was separated from the radiolabelled cAMP/cGMP substrate using ion exchange (BioRad AG 1 x 8 resin). Supernatants containing the [3H]-labelled guanosine product were added to scintillation fluid (Optiphase Supermix, Wallac). Scintillation was measured using a Wallac 1450 Microbeta Liquid Scintillation Counter (Perkin Elmer) and PDE activity was expressed in pmol cAMP or cGMP/min/mg protein. Inhibition assays were carried out in the presence of compounds dissolved in DMSO. PDE assays for specific activity and IC50 determination were carried out at a native lysate dilution that gave 30% cGMP/cAMP hydrolysis. RNA was isolated from purified GIE or asexual stages using Trizol (Invitrogen) according to the manufacturer’s instructions and treated with DNAse I (Roche). RNA was reverse-transcribed using Superscript II that was primed with random hexanucleotides (Invitrogen). Real-time PCR was performed using an ABI Prism 7900HT sequence detector (Applied Biosystems). Relative quantification of cDNA was performed using 2ΔCt method (User Bulletin 2, ABI, http: //www. appliedbiosystems. com). Triplicate PCR reactions were analyzed for each sample. Transmission-blocking antigen precursor Pfs48/45 (PF13_0247) and ookinete surface antigen precursor Pfs25 (PF10_0303) were used as markers of stage III and stage V gametocytes, respectively. Transcript abundance was compared using mean of ΔCt values calculated using ubiquitin-conjugating enzyme (PF08_0085) transcript (HK gene) as endogenous normalizer. Gene-specific primers used to profile the expression of PDEα, PDEβ, PDEγ and PDEδ were published in Wentzinger et al (2008) [39], and gene-specific primers used to profile the expression of Pfs48/45, Pfs25 and HK were published in Joice et al (2013) [6]. GIE were purified by magnetic isolation and pelleted by centrifugation at 1800rpm. To prepare membrane extracts, 1. 107 GIE were resuspended in 100 μl of PBS1X/1%Triton X-100, freezed at -80°C overnight and centrifugated at 16000 g for 5 min at 4°C. To prepare total extracts, 5. 106 GIE were used. Pellets were denatured in protein loading buffer 5 min at 100°C and were separated by 4–12% SDS-PAGE, transferred to PVDF membrane and blocked for 1 h in 5% nonfat dry milk. Immunoblots were probed overnight with a purified rabbit antiserum against PfPKA-R at 1: 16, a rabbit mAb anti-phospho-PKA substrate (RRXS*/T*, 100G7E, Cell Signaling) at 1/1000, a mouse mAb anti-HSP70 antibody at 1/1000, or a mouse mAb anti-Band3 antibody (Sigma) at 1/5000 followed by 1 hour with horseradish peroxidase-conjugated anti-mouse or anti-rabbit IgG secondary antibodies (Promega) at 1: 10 000. Detection step was performed using the Pierce chemoluminescence system (Pierce) following the manufacturer’s instructions. The levels of PfPKA-R or phospho-PKA were quantified by densitometry using the Quantity One analysis software (BioRad). For each sample, we then calculated the ratio of protein levels relative to loading control HSP70 or Band 3. GIE were air-dried on glass blood smears and methanol-fixed for 10 min at -20°C. After 1h pre-incubation in PBS1X/2% BSA, slides were incubated overnight with a purified rabbit antiserum against PfPKA-R at 1/50 and with AlexaFluor 594-conjugated goat anti-rabbit affinity-purified IgG (Molecular Probes) for 1 hour. Parasite nuclei were stained with Hoechst 33342 (diluted 1: 20000, Life technologies). Samples were observed at 100X magnification using a Leica DM 5000 B. Statistical significance for differences in cAMP concentration and in protein levels was established using student test and Wilcoxon Mann-Whitney rank sum test. Statistical significance for differences in retention rates was established using Wilcoxon Mann-Whitney rank sum test. Statistical significance for differences in proportion of GIE showing different shape was established using a Chi-square test.

Malaria transmission is ensured by deformable mature gametocyte-infected erythrocytes being taken up when a mosquito bites. Non-deformable immature gametocyte stages are sequestered in the bone marrow, as their lack of deformability would lead to their splenic clearance. In the present study, we apply nano-filtration technology to mimic splenic retention and demonstrate that deformability of transmissible mature stage V gametocytes is regulated by parasite cyclic AMP-dependent kinase signalling. Importantly, when we used drugs to raise cAMP levels we render transmissible mature gametocytes as stiff as non-transmissible gametocytes. In contrast, when we inhibit the cAMP-dependent kinase we render immature gametocytes more deformable. Thus, by two different approaches we confirm that the drop in cAMP levels in mature gametocytes leads to an increase in their deformability and hence more likely to circulate through the spleen. Our molecular observations have the potential to be translated into therapies for blocking malaria transmission by demonstrating that raising cAMP levels with sildenafil also known as “Viagra” renders mature gametocytes rigid. These findings provide the proof of principle that deformability of circulating gametocytes is targetable by pharmacological agents and as such, it provides a novel approach to prevent the spread of parasites. PDE inhibitors therefore represent novel drug leads potentially capable of blocking transmission and improving the worldwide fight to eliminate malaria from the human population.

Abstract Introduction Results Discussion Materials and Methods

2015

cAMP-Signalling Regulates Gametocyte-Infected Erythrocyte Deformability Required for Malaria Parasite Transmission

Scabies is a parasitic skin infestation caused by the burrowing mite Sarcoptes scabiei. It is common worldwide and spreads rapidly under crowded conditions, such as those found in socially disadvantaged communities of Indigenous populations and in developing countries. Pruritic scabies lesions facilitate opportunistic bacterial infections, particularly Group A streptococci. Streptococcal infections cause significant sequelae and the increased community streptococcal burden has led to extreme levels of acute rheumatic fever and rheumatic heart disease in Australia' s Indigenous communities. In addition, emerging resistance to currently available therapeutics emphasizes the need to identify potential targets for novel chemotherapeutic and/or immunological intervention. Scabies research has been severely limited by the availability of parasites, and scabies remains a truly neglected infectious disease. We report development of a tractable model for scabies in the pig, Sus domestica. Over five years and involving ten independent cohorts, we have developed a protocol for continuous passage of Sarcoptes scabiei var. suis. To increase intensity and duration of infestation without generating animal welfare issues we have optimised an immunosuppression regimen utilising daily oral treatment with 0. 2mg/kg dexamethasone. Only mild, controlled side effects are observed, and mange infection can be maintained indefinitely providing large mite numbers (>6000 mites/g skin) for molecular-based research on scabies. In pilot experiments we explore whether any adaptation of the mite population is reflected in genetic changes. Phylogenetic analysis was performed comparing sets of genetic data obtained from pig mites collected from naturally infected pigs with data from pig mites collected from the most recent cohort. A reliable pig/scabies animal model will facilitate in vivo studies on host immune responses to scabies including the relations to the associated bacterial pathogenesis and more detailed studies of molecular evolution and host adaption. It is a most needed tool for the further investigation of this important and widespread parasitic disease. Scabies, or sarcoptic mange, is an infectious skin disease caused by the mite Sarcoptes scabiei. Human scabies is a widespread disease in developing regions of the world, and remains a significant problem amongst indigenous populations in developed countries [1]. Secondary bacterial infections of scabies lesions, most notably with Group A Streptococcus or Staphylococcus aureus [2], have been linked to serious complications, such as renal damage and rheumatic heart disease [3]. Although scabies is unusual in urban Australia, it is currently endemic in remote northern and central Australian Aboriginal communities and remains a major public health problem in these socially disadvantaged communities [4], [5]. Over 70% of two year old children in Australia' s remote Aboriginal communities have been at least once infected with scabies, most of them acquiring the first infection as infants. Importantly, these numbers are reflected in the rates of observed streptococcal skin infections in over 80% of these children [6]. Parasitic mites of the genus Sarcoptes infest up to 40 different mammalian hosts across 17 families [7]. Commonly described hosts include dogs, pigs, foxes and wombats. Sarcoptic mange causes significant losses to primary industries, especially in pig herds, where it leads to decreased growth rates and subsequent reduced feed conversion efficiency [8]. Although effective control has been achieved in many regions by using macrocyclic lactones such as ivermectin, administered to sows 4 weeks before farrowing, sarcoptic mange remains common in piggeries, with reported prevalence from 20–86% [9]. Despite the economic and health significance of S. scabiei infestation in both human and animal populations, the pathogenesis and immune responses to this disease is not well understood [10]. Scabies has historically been a difficult disease to study. Scabies mites cannot be maintained or propagated away from their animal host, and it is difficult to collect mites in large quantities, as a typical infestation of human scabies involves fewer than twenty mites. Access to hyper-infested hosts has enabled the construction of cDNA libraries from human [11], [12] and fox [13] mite populations. Sequencing these libraries has resulted in the identification of genes which reveal some unexpected features of scabies biology and immunopathology [reviewed in 14]. We have recently shown that scabies mite gut proteases play major roles in maintaining the mite infestation within the epidermis, either as digestive enzymes [15], or by providing host complement evasion mechanisms [16]. Another research focus has been the characterisation of the molecular basis of emerging acaricide resistance [17], [18]. Despite these recent advances, access to hyper-infested hosts remains opportunistic and sporadic. Moreover, there has been an almost complete failure in efforts to maintain viable mites in the laboratory for longer than 24–48 hours, and no established methods are available to propagate mites in vitro. To overcome these barriers, the availability of a tractable animal model for scabies would be of enormous benefit. Despite being morphologically very similar [19], S. scabiei variants appear to be predominantly host specific, and investigations whether they are genetically distinct are ongoing [20], [21]. The majority of cross-infectivity studies have been unsuccessful, with infestations transient and self limiting [22]. Over several years we have tried unsuccessfully to propogate mites from crusted scabies patients, pigs and dogs on immunosuppressed mice. There is currently only one other animal model for scabies worldwide, that of canine scabies mites maintained on rabbit hosts at a U. S laboratory [22]. This trans-species adaptation was successful only after many attempts, and has not been able to be repeated (personal communication, L. Arlian Ohio). While this host-mite system has been extremely valuable for immunological studies, it has been maintained under strong drug selection pressure for over two decades [18], and consequently its utility as a source of live, viable mites is limited due to logistical difficulties of access for international research and regulatory restrictions. When infested with mange, pigs show similar epidermal, morphological, and immunological changes to humans [23]. Of significant importance is the fact that pigs have been described to develop a manifestation of scabies closely resembling human crusted scabies [24], a poorly understood manifestation. A further potential advantage of this host-parasite system is the fact that the complement system in pigs is comparable to humans [25], and pigs are susceptible to Streptococci, including Group A streptococci (GAS) [26]. Experimental models of porcine mange have been used previously to study clinical manifestations [27], effects on production and transmission dynamics, as well as limited studies of immunology and histopathology [28]. The majority have been short-term, naturally resolving infections. Our objective was to achieve a consistent, reliable quantity of S. scabiei. Herein we report progress in developing a sustainable experimental model of chronic porcine mange, which now consistently provides large numbers of mites, facilitating the less restrictive conduct of research on human and animal scabies. All animals were handled in strict accordance with good animal practice as defined by the Australian code of practice for the care and use of animals for scientific purposes and the NHMRC' s Animal Code of Practice, and all animal work conducted with ethical approval from both DEEDI and QIMR Animal Ethics Committees (DEEDI-AEC SA2009/07/294, QIMR A0306-621M). Samples were initially sought from pigs for the purpose of obtaining sufficient mites to attempt transmission of infection to mice. Six ears from slaughtered pigs with mangy appearance were supplied to the laboratory in a weekly basis by a Southeast Queensland abattoir from June 2004 to November 2007. Skin pieces were dissected from the inner ear and incubated in glass petri dishes at 27°C, which encourages mites to crawl out towards the heat source. Dishes were examined and mites picked under a dissecting microscope using microscopic needles. Pigs were housed at the DEEDI Animal Research Institute, Yeerongpilly, QLD, and at the Centre for Advanced Animal Studies, Gatton, QLD. It was ensured that the care and the experimental practices conformed to the Australian animal ethics guidelines. Pigs with suspected natural mange infections (Table 1, groups 1–4) were sourced by approaching commercial piggeries, saleyards and private owners. At a later stage of the study (groups 5–10) naïve piglets of the “large white” breed - a common meat producing breed in Australia [29] - were obtained from the University of Queensland piggery, Gatton, QLD and experimentally infected with mites. To passage the infection, new piglets were placed into pens adjacent to infected pigs from the previous group. A heater placed on the fence line encouraged pigs to congregate and thus enhanced the potential for mite transfer. From groups 6 onwards, mite transmission was additionally ‘boosted’ by the direct transplant of mite infested skin crusts obtained directly from the previous group. Briefly, crusts were harvested from infected pigs and dissected into small pieces (approx 0. 5cm2). Several crusts were inserted into both ear canals of naïve piglets. Pigs were temporarily restrained to prevent dislodgement of the crust by agitation, thereby allowing successful infestation. Skin samples were collected by gently scraping and lifting off encrusted areas from the inner ear area of the pig with a sharpened teaspoon and subsequently examined for mites. S. scabiei var. suis DNA from 10 mites collected from one pig in group 4 in 2006 and from another pig in a later group 10 in 2010 was prepared as described previously [30]. The SMIPP-S-B2 gene was amplified in a proofreading PCR and products were purified and sequenced in both directions, with chromatograms manually inspected for quality. These sequences were then aligned with SMIPP-S-B2 (AY333073), with the related SMIPP-S-B1 (AY333076), and SMIPP-S-B3 from S. scabiei var. hominis cDNA used as outgroups (data not shown). Sequence alignments and subsequent phylogenetic analysis was performed using MEGA4 [31]. The quantity of mites obtained from abattoir samples was variable, exceeding 3,000 mites per ear on some occasions, whereas at other times no mites were observed for several months (Figure 1). Although seasonal trends for mange occurrence have been reported [32], this trend is likely dependent on other determinants including housing conditions, with pigs in confinement showing less seasonal differences in mite burden [33]. While there is no obvious seasonal trend apparent from our data (Figure 1), without having prior knowledge of the source environment from the samples obtained, we could not investigate for any association. The results however accord with the known prevalence of sarcoptic mange in the pig industry, despite the availability of effective control measures. The sporadic supply of mites obtained from this approach meant that it was unsuitable as a sustainable source for laboratory experiments. Clearly, an alternative approach was needed. Despite having clinical features of mange, such as scratching behaviour, hair loss, reddened and flaky skin, the untreated pigs in groups 1 and 2 either had no or very few mites and symptoms resolved within 6 weeks of acquisition (Table 1). These observations are typical for acute-hypersensitive mange in pigs, which is generally described as a short term infestation with low mite numbers [27], a clinical picture akin to ordinary human scabies. To obtain sufficient mites for our studies, pigs with chronic mange were needed. Chronic mange closely resembles human crusted scabies, with the formation of hyperkeratotic plaques and proliferation of mites. It is more commonly observed in older sows or when pigs are immunosuppressed, and is generally uncommon in growing pigs [27], [33]. To increase mite numbers and maintain a prolonged infestation by counteracting natural immunity, it was proposed to treat mange infected pigs with corticosteroids. This concept is supported by the observation that corticosteroid therapy often results in the development of crusted scabies in humans [34]. Dexamethasone is a synthetic glucocorticoid commonly used in animal models to promote infection and exacerbate infectious responses. It has pleiotropic effects on immune responses, depressing lymphocyte production, antibody production and inflammatory responses. Due to its previous use in other immunosuppression models, including pigs [35], [36], it was selected for study. Because dexamethasone had not been tested previously in scabies infected pigs, and to give appropriate consideration to animal welfare issues, a gradual, conservative treatment program was initiated. As naturally infected pigs self cured over time (Table 1, Groups 1 and 2) pigs with existing, naturally acquired mange were treated (Table 1, Group 3) with weekly injections of 0. 01mg/kg of dexamethasone (Dexafort, Provet, Brisbane). However this did not increase mite numbers or prolong infection. Mite numbers improved markedly when dexamethasone was given tri-weekly but the infestation was not sustained for longer than 12 weeks and only a single viable mite harvest was possible from each pig (Table 1, group 4). Because of the possibility that the immune system was overcoming infection prior to steroid treatment, dexamethasone treatment was commenced in naïve piglets prior to infection. In light of marginal effects observed at lower dosage, the dexamethasone dosage and frequency was increased to 0. 1mg/kg daily. For these groups (5 and 6) a successful experimental transfer was observed, with up to 500 mites obtained in a single ear scraping. Because of concerns regarding aversion behaviours to daily injections, the delivery method of dexamethasone delivery was changed from injection (groups 3–5) to oral (Dexamethasone tablets, 1 or 4mg, Provet, Brisbane), with pigs readily accepting tablets offered in marshmallows. A further alteration (applied to groups 6–10) was to boost the passage of mites by directly placing mite infested skin crusts deep into the ear of piglets. Combined, these protocol modifications greatly enhanced rate of infection, with new piglets developing crusted lesions within four weeks. Mite numbers continuously increased, and one pig maintained infection for over six months (Table 1, Group 7). Since 0. 1mg/kg appeared to be inadequate to achieve sufficient immunosuppression, moderate increments in dosage were commenced, whereby a sentinal pig was exposed to the higher dose for six weeks prior to the rest of the group, allowing for observation of detrimental effects. At 0. 3mg/kg pigs started to develop much larger crusted lesions which also spread to regions beyond the ears (Figure 2B). At this point, an excess of 10,000 mites were obtained per harvest, and crusting would re-develop after removal, enabling multiple harvests (Table 1, Group 8). Given that our objective of developing sustainable mite supply was achieved, further dose increases were not undertaken. At the maximal dose of 0. 3mg/kg dexamethasone mite infestation dramatically increased, but noticeable side effects were observed. The most obvious of these was a tendency of growth retardation and a change of body shape. Retardation in weight gain in dexamethasone treated pigs has been reported [36]. Decreased growth can be attributed to the interference of the corticosteroid on the hypothalamic-pituitary axis, thus depressing production of growth hormone. Stunting was not associated with decreased nutrition uptake, and all pigs had healthy appetites. At higher doses some pigs also developed hirsutism. Several pigs developed laxity in hind foot tendons, and in one pig evidence of excess adiposity and bone demineralisation was observed on post mortem. Frequent vomiting was observed in one pig. While dexamethasone is known to increase the risk of gastric ulcers, no evidence of ulcers or gastrointestinal irritation were observed post-mortem. Corticosteroids have been used to induce stress responses, and this was observed in several pigs having decreased ability to cope with bullying and natural dominance behaviours. This effect was controlled by housing pigs in pairs or singly, thereby minimising the requirement to compete for social positioning. Most these side effects parallel that of iatrogenic Cushing' s syndrome seen in humans, with glucocorticoid excess resulting in symptoms such as centralised adiposity, bone osteoporosis, hirsutism, depression and anxiety. It should be emphasised that the side effects observed in this study were mild, and aside from the physical effects, pigs appeared normal in spite of their mange infestation. Pigs were closely monitored by skilled veterinary staff for side-effects including increased susceptibility to disease, but other than mange, no additional infections were observed. Good husbandry and infection control measures were paramount to this process. Since the optimal treatment regimen has now been established, the dosage of dexamethasone can be adjusted accordingly to maintain required mite burden, while minimising side effects. Continued tailoring of dexamethasone is also important since it became evident through the trial that pigs showed great variation in inherent immunologic responses to both steroid treatment and scabies. For example, one pig on low dose dexamethasone still developed chronic mange and the side effects described above. Similarly, other pigs had relatively low mite burdens despite receiving higher doses of dexamethasone. Variation in pig mange responses are common [27], with small number of pigs in the population often harbouring the majority of the mite burden [8]. Our data (summarized in Table 1) suggest the development of a stable and transmissible mite population established from an originally natural infection over five passages. We observed a striking increase in mite numbers per harvest especially for the last three groups. We propose that the increased mite numbers are at least in part a direct consequence of the immunosuppression regimen and reflect the immune status of the pigs. It is however also possible that the selection towards adapted subpopulations of mites has occurred via passage through five generations of pigs. Any adaptation of the mite population should be reflected in genetic changes. Such changes over time can only be observed in a mite population within a closed cohort, where there had been no introduction of mites with potentially new genotypes. As a indispensable prerequisite for such a study the presented pig/mite model provides a continuous source of mites and may offer a controlled setting to monitor genetic changes over time in closed mite populations. To test the feasability of the model for a large experimental setup at a later stage we aimed here to compare limited sets of genetic data obtained from pig mites collected from a naturally infected pig in group 4 (2006) with data from pig mites collected from one pig in group 10 (2009). Thus, we were investigating genetic changes in a closed mite population maintained isolated for over 3 years. We focussed on the SMIPP-S-B2 gene, belonging to a multigene family of at least 32 closely related gut proteases homologous to the group 3 major allergens of astigmatid house dust mites [37] [38]. Due to mutations in the catalytic triad and considerable structural rearrangements, these have no proteolytic function, but have instead evolved into potent inhibitors of each of the three pathways of the host complement system [16]. The genes in the SMIPP-S family show considerable sequence diversity, and are hypothesised to mediate a novel host immune-evasion strategy. Synonymous versus non-synonymous changes within the coding region of the mature SMIPP-S protein indicate that SMIPP-Ss most likely have functions that would be affected by non-synonymous changes and therefore be subjected to selection pressure [37]. The pig mite derived SMIPP-S-B2 sequences form a distinct cluster from the corresponding human mite sequence. We also observed a considerable degree of intra-species heterogeneity. In the 2006 mite population, three distinct B2 isoforms (B2-1 to 3) were observed, with polymorphisms at 7 amino acid sites (Figure 3A) with isoform B2-3 observed most frequently. Analysis of the 2009 mite population revealed the emergence of five new isoforms (B2-4 to 8), with isoforms B2-1 and 2 not seen in the 2009 mites. Neighbour-joining analysis suggests these new isoforms are derived from B2-3 (Figure 3B). Some of the bootstrap values were low, which is most likely attributable to the proposed recent evolutionary divergence, also indicated by the short branch lengths at the respective nodes. Indeed, the primary source of sequence variation involved polymorphisms at amino acid residue 184. Whereas human mite sequences contain an alanine at this site, pig mites in 2006 contained threonine or aspartic acid, and 2009 pig mites had alanine, threonine, aspartic acid or glutamate. Sequence alignments of the complete SMIPP-S family show that this region is highly polymorphic [38]. Moreover, the predicted surface location of this residue suggests this rapidly evolving divergence may be of functional relevance. It is suggested the SMIPPs provide a complex network of immunologically cross-reactive sequences in response to host antibodies, and thus represent specific genetic adaptations to the intimate host-parasite relationship [37]. While heterogeneity within the SMIPP family most likely happens in vivo in human and pig populations, it is difficult to undertake such studies in human settings. In previous genetic analysis on mites from human crusted scabies, patients had been infected over many years and had undergone multiple treatments over several episodes. In microsatellite data reported by Walton et al. [20], [39] a large degree of genetic diversity in the mite population within individual hosts was possibly the result of both recrudescence and re-infection from multiple community sources. In contrast our animal model provides a controlled setting to monitor genetic changes over time in closed mite populations. Although this SMIPP-S analysis was limited in its scope, it nevertheless highlights the potential of this model for investigations of the molecular evolution of scabies mite genes. Representative isolates collected over the continuing passage of mites through successive generations may provide fascinating insights into genetic diversity, evolution and host-adaption of scabies mites, particularly within the novel SMIPP family. In conclusion, over a prolonged timeframe of five years, involving ten independent cohorts (32 animals in total), we have successfully developed a sustainable experimental infestation of scabies mite on immunosuppressed pigs. Encrustment on the ears occurs after 6–12 weeks and can be maintained for at least 12 months, depending on drug dosage and individual pig response. This animal model now consistently provides large mite numbers (>6000 mites/g skin) for molecular-based research on scabies. Projects that have benefitted from this to date include detailed studies of gene expression in scabies mites [40], and the development of new techniques to measure drug sensitivity (submitted), pilot studies in preparation to sequence the scabies mite genome, and histological localisation of various molecules involved in evasion of host defences [15], [16], [41]. Most importantly, further research may utilise the full features of this animal model which facilitates in vivo studies, including investigation of pig immune responses to scabies and dexamethasone pathways of immunosuppression, in addition to more detailed studies of molecular evolution and host adaption. Furthermore, our pig model is now at a stage to be optimised as pig/scabies/GAS model. Such research should result in innovative tools for the prevention, monitoring and further investigation of this important and widespread parasitic disease.

Scabies, a neglected parasitic disease caused by the microscopic mite Sarcoptes scabiei, is a major driving force behind bacterial skin infections in tropical settings. Aboriginal and Torres Strait Islander peoples are nearly twenty times more likely to die from acute rheumatic fever and rheumatic heart disease than individuals from the wider Australian community. These conditions are caused by bacterial pathogens such as Group A streptococci, which have been linked to underlying scabies infestations. Community based initiatives to reduce scabies and associated disease have expanded, but have been threatened in recent years by emerging drug resistance. Critical biological questions surrounding scabies remain unanswered due to a lack of biomedical research. This has been due in part to a lack of either a suitable animal model or an in vitro culture system for scabies mites. The pig/mite model reported here will be a much needed resource for parasite material and will facilitate in vivo studies on host immune responses to scabies, including relations to associated bacterial pathogenesis, and more detailed studies of molecular evolution and host adaptation. It represents the missing tool to extrapolate emerging molecular data into an in vivo setting and may well allow the development of clinical interventions.

Abstract Introduction Methods Results and Discussion

infectious diseases/bacterial infections infectious diseases/neglected tropical diseases infectious diseases/tropical and travel-associated diseases infectious diseases/skin infections

2010

A Tractable Experimental Model for Study of Human and Animal Scabies

Trypanosoma cruzi, causative agent of Chagas disease in humans and dogs, is a vector-borne zoonotic protozoan parasite that can cause fatal cardiac disease. While recognized as the most economically important parasitic infection in Latin America, the incidence of Chagas disease in the United States of America (US) may be underreported and even increasing. The extensive genetic diversity of T. cruzi in Latin America is well-documented and likely influences disease progression, severity and treatment efficacy; however, little is known regarding T. cruzi strains endemic to the US. It is therefore important to expand our knowledge on US T. cruzi strains, to improve upon the recognition of and response to locally acquired infections. We conducted a study of T. cruzi molecular diversity in California, augmenting sparse genetic data from southern California and for the first time investigating genetic sequences from northern California. The vector Triatoma protracta was collected from southern (Escondido and Los Angeles) and northern (Vallecito) California regions. Samples were initially screened via sensitive nuclear repetitive DNA and kinetoplast minicircle DNA PCR assays, yielding an overall prevalence of approximately 28% and 55% for southern and northern California regions, respectively. Positive samples were further processed to identify discrete typing units (DTUs), revealing both TcI and TcIV lineages in southern California, but only TcI in northern California. Phylogenetic analyses (targeting COII-ND1, TR and RB19 genes) were performed on a subset of positive samples to compare Californian T. cruzi samples to strains from other US regions and Latin America. Results indicated that within the TcI DTU, California sequences were similar to those from the southeastern US, as well as to several isolates from Latin America responsible for causing Chagas disease in humans. Triatoma protracta populations in California are frequently infected with T. cruzi. Our data extend the northern limits of the range of TcI and identify a novel genetic exchange event between TcI and TcIV. High similarity between sequences from California and specific Latin American strains indicates US strains may be equally capable of causing human disease. Additional genetic characterization of Californian and other US T. cruzi strains is recommended. Trypanosoma cruzi is a protozoan parasite that, in both humans and dogs, may cause an insidious onset of fatal cardiac disease[1]. Known as Chagas disease, T. cruzi is the most economically important parasitic infection in Latin America, where an estimated 8–9 million people are living with the chronic disease [2]. The parasite is most commonly transmitted vectorially, by numerous species of triatomine bugs (commonly called “kissing bugs”), distributed from Chile and Argentina in South America to approximately 42. 5 degrees northern latitude of the United States of America [3,4]. Only seven authochthonous clinical cases of Chagas disease in humans have been officially documented in the United States despite the fact that nine endemic Triatoma species are known to harbor T. cruzi [1]. The prevalence of infection varies among Triatoma species and across geographic regions [5,6] and has been reported to be as high as 61% in Louisiana [7]. In the US, T. cruzi has been found in wild canids; numerous rodent species; and mesomammals such as raccoons, opossums and skunks [1]. The prevalence of T. cruzi in various wildlife species has ranged upwards from 50% in Texas and some southeastern states [8,9]. Many of these mammals are peri-urban species that adapt well to human-modified landscapes and, if infected, can bring T. cruzi closer to humans and their canine companions. In turn, when triatomines are present in the local environment, there may be a subsequent increased risk of vectorial T. cruzi transmission to people, and an even greater transmission risk to dogs, who likely acquire T. cruzi via ingestion of infected vectors [10,11]. In 2006, the Texas Veterinary Medical Diagnostic Laboratory reported 18. 6% of 532 dogs presumably clinically ill with cardiac disease to be seropositive for T. cruzi [12]. In addition, canine serological surveys in states such as Tennessee, Louisiana and Texas indicate that T. cruzi infection is not an uncommon occurrence, even in apparently healthy domestic dogs [13–15]. Likewise, recent human serological surveys and Triatoma blood meal analyses suggest that human T. cruzi exposure may also occur more frequently than previously realized [16–18]. Physicians and veterinarians are not well-trained to recognize this disease in the US; treatment is not readily available [19]; and there are no drugs approved for veterinary use [1]. Understanding of the ecology of T. cruzi in the US, including vector and reservoir distribution, and of the molecular epidemiology of endemic strains will enable health and disease control professionals to better respond to the likely rising incidence of Chagas disease. Trypanosoma cruzi taxonomy has been revised, with the most recent consensus classifying the organism into six subtypes or ‘discrete typing units’ (DTUs), designated TcI to TcVI [20]. Within each DTU fall numerous strains whose unique identities are generally determined via typing of several independent genetic loci. Very little T. cruzi molecular epidemiology research has been done in the US as compared to that accomplished in Latin American countries [21], despite the concern that Chagas may become an emerging disease in the country [19,22,23]. Most research on US T. cruzi has been restricted to typing to the DTU level, and to date, only TcI and TcIV have been detected in local vectors and wildlife [9,24]. Researchers have recently begun to explore intra-DTU molecular diversity, focusing on isolates from the southeastern US [21]. However, data on genetic diversity in southwestern regions (e. g. California, Arizona, and New Mexico) are very limited [25]. California has the largest influx of migrants of any state in the US [26], with 53% of the immigrant population of Latin American origin [27]. Additionally, 2011 US census data indicates that more than 21% of the nearly 3 million South American migrants residing in the US live in California, with estimates of 75,000–399,000 living in Los Angeles alone [28]. It is therefore probable that many exogenous strains of T. cruzi enter California every year via human migration. It has been experimentally demonstrated that at least one virulent Honduran strain can be viable if introduced into Tr. protracta, the most common triatomine bug vector in California [29]. Thus, the pool of T. cruzi strains present in the US may potentially become more diverse. Additionally, with global climate change, it has been predicted that the human population at risk for T. cruzi transmission will increase in southern California due to increased triatomine activity associated with warmer temperatures [23]. Therefore, in addition to monitoring T. cruzi vector distribution, it is important to investigate the molecular genetics of endemic strains; how they compare to virulent strains in Latin America; and whether recently introduced strains may already exist in local vectors. To this end, the goals of this study were to: 1) compare the prevalence and DTUs of T. cruzi within triatomine bug populations from two regions of California and 2) further characterize the California T. cruzi samples via molecular genetics to assess whether there are regional differences and to determine how the California samples compare to those present in other regions of the US and Latin America. Triatoma protracta specimens were actively collected from private residences in two study regions. All landowners consented to the collection of bugs from their properties. The first study area was located in southern California, in the town of Escondido (33. 1247° N, 117. 0808° W). This study site was chosen because previous research had identified T. cruzi in the resident triatomine bug population [25]. Abundant woodrat (Neotoma macrotis) nests were found, and much of the terrain was covered with large granite-based boulders and smaller rocks that provided crevices for triatomines. The second study area encompassed several residences in the town of Vallecito, situated in northern California (38. 0903° N, 120. 4736° W). This location was selected based on knowledge that multiple triatomine bugs collected there in 2011 were positive for T. cruzi (M. Niemela, pers comm). Woodrat nests at these properties varied by site but were generally less abundant than the Escondido location. Black light traps were used in July and August 2012 to collect adult bugs from both study regions. Lights were turned on approximately 30 minutes before sunset and left on for at least two hours after sunset to coincide with the evening hours during which the adult bugs were flying (C. Conlan, pers comm) [30]. The bugs often did not fly the complete distance to the light trap; therefore, combing the surrounding area facilitated capture of bugs crawling on the ground nearby. This trapping method worked well in Escondido, where the trap was strategically placed at the top of a hill and the vegetation on the slope below consisted of small shrubs that did not obscure the emanating light. Light trapping was less successful in Vallecito. Hence, to augment the triatomine sample size from this region, we enlisted the help of property owners to collect bugs found in their homes. We also partially excavated several woodrat nests to obtain both adult and nymphal bugs. All bugs were placed in tubes and frozen at -20C° until laboratory processing. In addition, we opportunistically obtained specimens from public health employees in southern California, who often received bugs from concerned citizens, especially if the bug had bitten someone within the home. These bugs were shipped to the laboratory either frozen or in ethyl alcohol during the months of April-July 2012 and June-August 2013. The program Geneious was used for the assembly and alignment of maxicircle and nuclear sequences. Previously published COII-ND1, TR and RB19 partial sequences included in our alignments are listed in S1 Table. We prioritized the selection of the RB19 and TR isolates included in the alignment, based on the availability of sequences for both genes, overlap with isolates included in the COII-ND1 alignment when possible, as well as representation from a broad geographic range. Two different approaches have been used to amplify the contiguous COII-ND1 maxicircle genes, which span a region of approximately 1,594 bp (CL Brener strain, GenBank #DQ343645). The first approach results in separate partial sequences for each gene, yielding short fragments of 417 bp and 369 bp for COII and ND1, respectively [42,46]. The second approach, and the one applied in this study, generates a COII-ND1 combined partial sequence of approximately 1,272 bp in length [21,35,39]. The two shorter gene fragments obtained via the first method are completely nested within this longer combined sequence. The goal of our phylogenetic analyses was to maximize the number of unique T. cruzi sequences included, while ensuring that isolates represented a wide geographic range, especially within the TcI DTU. Therefore the alignment of our COII-ND1 sequences was not limited to published sequences of similar length, but also included the shorter separate gene sequences obtained in studies where the first approach was applied (S1 Table). Following alignment, all COII-ND1 sequences (n = 62) were trimmed and manually concatenated to a final length of 786 bp (369 + 417), representing the two separate gene fragments. Phylogenetic trees for the RB19, TR and COII-ND1 gene sequences were re-constructed in MEGA6 via Neighbor-Joining (NJ) and Maximum Likelihood (ML) methods. In the NJ approach, the evolutionary distances were computed using the maximum composite likelihood method [47] with 2,000 bootstrap replicates. For the ML trees, the best fit model (as determined via the Model Test option in MEGA6) was run with 500 bootstrap replicates. The bootstrap support of the resulting NJ and ML phylogenies were compared for each genetic marker, and the best supported tree was selected. For trees displaying similar topology, both NJ and ML bootstrap values were included at appropriate nodes. In all cases, the trees were outgroup rooted with T. cruzi marinkellei. The discovery of an apparent TcI/TcIV hybrid was further evaluated via the comparison of pairwise-distances between this sample and representative samples of TcI and TcIV for each genetic marker (i. e. T. cruzi sequences included in the reconstruction of the respective phylogenetic trees). The uncorrected p-distances were calculated in MEGA6 using pairwise deletion and transitions/transversions as the substitution types. The program Dna-SP version 5. 10 [48] was used to calculate diversity indices for the TR, RB19 and COII-ND1 TcI sequences obtained in this study. The TcIV sequences were not analyzed due to their limited number. Haplotype diversity (Hd), nucleotide diversity (Pi), G+C content and the number of segregating sites (singleton + parsimony informative polymorphic sites) were calculated for all genes. The number of synonymous and non-synonymous mutations, as well as the ratio of number of nonsynonymous substitutions per site to synonymous substitutions per site (dN/dS), were calculated for the TR and RB19 genes but were omitted from the COII-ND1 analysis due to the putative RNA editing that occurs within the maxicircle gene [49]. A total of 29 triatomine bugs were collected from the Vallecito study area, of which 24 (two adults and 22 nymphs) were found within woodrat houses. The five remaining bugs were adults obtained from either light traps or within a resident’s home. All identified bugs were Triatoma protracta. The two PCR-based screening assays targeting different T. cruzi genomic loci showed a high degree of concordance in the DNA extracted from all bugs. The 121/122 kinetoplast minicircle assay was marginally more sensitive, detecting 16 positive bugs, whereas the TcZ1/TcZ2 nuclear assay only identified 15 of these same bugs as positive. Kinetoplast DNA sequences obtained from the single discordant sample confirmed the presence of T. cruzi DNA. Thus 55. 2% of bugs at the Vallecito site were infected with T. cruzi. At the Escondido study site, 53 bugs were collected, all of which were adult Tr. protracta drawn to light traps. Thirteen bugs were positive for both T. cruzi PCR screening assays; however, positive amplification was detected for an additional six bugs using only the kDNA minicircle assay. Five of these discordant samples were successfully cloned and sequenced to confirm the presence of T. cruzi DNA, yielding the conclusion that 18 bugs (34%) were T. cruzi positive at the Escondido location. There were 15 Tr. protracta bugs submitted from public health employees in southern California, three of which were positive for parasite DNA on both screening assays (20%; there were no discordant results for screening assays among this set of bugs). With the exception of one specimen from San Diego, these bugs represented a range of locations within the Greater Los Angeles Area: Agoura Hills, Altadena, Los Angeles, Northridge, Oak Hills, Santa Clarita, Simi Valley, Tarzana, and Thousand Oaks (S1 Fig). Of the 11 bugs for which addresses were provided, area visualization via GoogleEarth revealed that the homes primarily abutted natural canyon areas designated as parks or were within housing tracts interspersed with parcels of undeveloped land. A summary of the T. cruzi positive bugs is shown in Table 2. Together these data confirm that wild populations of Tr. protracta at multiple sites in California are frequently infected with T. cruzi. We next aimed to investigate which T. cruzi subtype/DTUs were present using lineage-specific genotyping on a subset (n = 29) of the positive bugs, as described in Fig 1. Within this subset, DTU determination was successful only for those samples that were positive for both of the T. cruzi screening assays described above (n = 22). Samples that were parasite positive only for the more sensitive 121/122 assay (n = 7) likely had insufficient DNA to amplify the lower copy number DTU gene targets. We detected 13 and 7 TcI samples from the Vallecito and Southern California locations, respectively. We found only two TcIV samples, both of which were from the Escondido location. Thus, the T. cruzi TcI and TcIV DTUs are both endemic in California. The T. cruzi DTUs that we identified are known to contain substantial genetic diversity [35,50–53]. We therefore generated nucleotide sequence data to investigate our sample diversity at the intra-DTU level and to enable comparison with strains from other studies. Sequences from two nuclear genes (TR and RB19) consistently classified our Californian (CA) samples into TcI (n = 10) and TcIV (n = 2) DTUs, confirming our previous genotyping results (Figs 2 and 3). The TR gene demonstrated greater sequence diversity across CA samples than did the RB19 gene: thirteen vs. two unique haplotypes identified respectively. The RB19 gene sequences for the 10 CA TcI samples were identical and indistinguishable from a single TcI sequence from an opossum isolate [38] obtained from the US state of Georgia (Fig 2). Likewise, the TcIV RB19 sequences were identical for the two CA samples (Esc19 & Esc26), as well as for two other US samples in GenBank (from a dog of unknown origin and a raccoon from Georgia). In contrast, for the TR gene, 2 to 8 single nucleotide polymorphisms (SNPs) were detected among both the TcI and TcIV-positive CA samples (Fig 3). The CA samples were distinguished from two GenBank TcIV sequences from Guatemala and Brazil by 11 to 14 SNPs. Within the TcI group, all the sequences from this study were closely related. The most closely related database sequences were from the US and northern South America (Colombia and Venezuela). Three of the southern CA sequences (Esc2, SoCal1 allele 1 and SoCal3) were identical to each other, as well as to an isolate obtained from a bug collected in the state of Florida (GenBank #AF358970) [35]. For the TR gene, the phylogenetic reconstruction between the NJ and ML trees was very similar, and bootstrap values for both trees are presented at congruent nodes (Fig 3). In contrast, the NJ and ML topology for the RB19 gene varied within major clades, and only the NJ tree is represented (Fig 2). For the maxicircle COII-ND1 genes, the NJ and ML tree topologies were very similar within the TcI clade, but the NJ tree provided better support within the TcIV clade. We therefore present the NJ tree with both NJ/ML bootstrap values indicated at congruent nodes (Figs 4 and 5). Eleven of the twelve samples in this study were categorized as TcI based on the analysis of the maxicircle COII-ND1 genes (Table 2 and Fig 4). Phylogenetic analysis of the concatenated COII-ND1 sequences revealed that the CA TcI sequences obtained in this study were grouped with strong bootstrap support (96%) in a subclade (Fig 5, subclade 1) with other North American isolates (US and Mexico), as well as several isolates from Central America (i. e. Guatemala and Honduras). In addition, Colombian and Venezuelan isolates previously classified within either “sylvatic” or “domestic” genetic populations [42] were also included in this subclade. Thus, the composition of sequences within subclade 1 closely corresponded to that of the group described elsewhere as TcI-Dom, which contains a high proportion of TcI strains associated with human infection across the Americas [46,51,53]. Esc19 was the only sample classified as having a TcIV maxicircle sequence, varying by only 2–3 SNPs from the southeastern US isolates (Fig 4). Interestingly, the Esc26 COII-ND1 sequence, which was defined as TcIV via the RB19 and TR nuclear gene sequences, as well as the DTU assays, was classified as TcI and was identical to those sequences obtained for Esc2 and Esc46, both of which were typed as TcI by all other markers tested. These data are most consistent with the Esc26 sample being the product of a genetic exchange event between TcI and TcIV ancestors, leading to TcI mitochondrial introgression into a TcIV nuclear genomic background. The p-distances presented in Table 3 highlight the genetic exchange between TcI and TcIV observed in sample Esc26. With respect to the RB19 and TR markers, Esc 26 was more closely related to TcIV than TcI by an order of magnitude. In contrast, for COII-ND1, the reverse finding was apparent. Table 4 provides values for the diversity indices calculated in Dna-SP. As seen in the phylogenetic analyses, no diversity was observed within the TcI sequences for the RB19 gene, whereas the TR and COII-ND1 genes are more genetically diverse. Our research is one of only two molecular studies on T. cruzi in California and the first to investigate this parasite’s genetic diversity in the northern portion of the state. While this study’s prevalence of T. cruzi in Tr. protracta populations in southern California (~30%) was similar to earlier findings, an even higher prevalence was detected in our northern California study region. The genetic markers employed in this study allowed us to demonstrate the close similarities between T. cruzi strains in California and those present in other US states, as well as some Latin American countries. Thus, vectors across California present a clear transmission risk to humans and dogs. Additionally, experimental studies have already proven that Tr. protracta can sustain a Honduran T. cruzi isolate [29], and Triatoma infestans and Rhodnius prolixus, two vectors from Latin America, have been shown capable of harboring US T. cruzi isolates [4]. Although the divergence and migration of T. cruzi strains occurred over a period of millions of years [50,53,74], in this new era of global connectivity, vectors and the pathogens they carry may unknowingly be transported between countries [75,76]. Therefore if triatomine vectors efficient in T. cruzi transmission, and perhaps more readily able to colonize human homes, were to be unwittingly introduced to the US, potential mixing of T. cruzi strains could occur within and among vector species. Despite the fact that T. cruzi has been known to exist in the US for at least 80 years, only four states consider Chagas disease to be reportable. In late 2012, Texas became the fourth state to declare Chagas a reportable disease, a decision preceded by a series of in-state studies and clinical case reports on the disease in both canines and humans. Consistent with Texas, our research implies that some areas of California may have a similar risk for T. cruzi transmission and suggests that California physicians and veterinary practitioners should consider Chagas disease as a potential cause of cardiac illness in regions where Tr. protracta populations are evident.

Trypanosoma cruzi is a protozoan parasite that causes Chagas disease in humans and dogs and may eventually lead to mortalities related to cardiac failure. This parasite is most frequently transmitted by triatomine bug vectors, commonly called “kissing bugs. ” Although Chagas disease is predominately acquired in Latin American countries, T. cruzi exists in wildlife and vectors in some parts of the United States of America (US), including regions of California. Within the US, occasional cases of locally acquired Chagas disease have been reported, and recent serological surveys indicate that T. cruzi exposure may be occurring more commonly than previously realized. However, relatively little molecular research has been performed on the T. cruzi strains present in the US, especially within California. In this study, we collected nearly 100 kissing bugs from regions of northern and southern California to determine the T. cruzi prevalence and genetic diversity for each region’s kissing bug population. We compared DNA sequences obtained in this study to those of several T. cruzi strains found in Latin America and the southeastern US. Based on our data, we conclude that Californian T. cruzi samples are closely related to strains found in Latin America known to be associated with human infections.

Abstract Introduction Methods Results Discussion

2016

Molecular Diversity of Trypanosoma cruzi Detected in the Vector Triatoma protracta from California, USA

We define an interface-interaction network (IIN) to capture the specificity and competition between protein-protein interactions (PPI). This new type of network represents interactions between individual interfaces used in functional protein binding and thereby contains the detail necessary to describe the competition and cooperation between any pair of binding partners. Here we establish a general framework for the construction of IINs that merges computational structure-based interface assignment with careful curation of available literature. To complement limited structural data, the inclusion of biochemical data is critical for achieving the accuracy and completeness necessary to analyze the specificity and competition between the protein interactions. Firstly, this procedure provides a means to clarify the information content of existing data on purported protein interactions and to remove indirect and spurious interactions. Secondly, the IIN we have constructed here for proteins involved in clathrin-mediated endocytosis (CME) exhibits distinctive topological properties. In contrast to PPI networks with their global and relatively dense connectivity, the fragmentation of the IIN into distinctive network modules suggests that different functional pressures act on the evolution of its topology. Large modules in the IIN are formed by interfaces sharing specificity for certain domain types, such as SH3 domains distributed across different proteins. The shared and distinct specificity of an interface is necessary for effective negative and positive design of highly selective binding targets. Lastly, the organization of detailed structural data in a network format allows one to identify pathways of specific binding interactions and thereby predict effects of mutations at specific surfaces on a protein and of specific binding inhibitors, as we explore in several examples. Overall, the endocytosis IIN is remarkably complex and rich in features masked in the coarser PPI, and collects relevant detail of protein association in a readily interpretable format. Protein-protein interaction (PPI) networks aim to capture the interactions between proteins that mediate many of their molecular functions [1]–[3]. However, with one node per protein and one edge per binary interaction, PPIs provide only a coarse rendering of the nuanced molecular level interactions. With exposed surfaces ranging from tens to hundreds of residues, proteins may present multiple distinct binding interfaces. Each interface can mediate binding to a single partner, or to multiple partners. The cooperative or competitive character of these interactions tunes protein availability in the cell, the formation of higher order complexes, and ultimately many important biological functions. Proteins with multiple binding interfaces can bring together distinct partners to assemble transient or permanent complexes. In contrast, multiple distinct partners competing for a single shared interface may function to connect disparate functional modules in the cell [4], [5], with such competitive binding having arisen, for instance, as a result of gene duplication [6], [7]. Distinguishing the types of binding interfaces a protein uses for each interaction partner is a key step to resolving the cooperativity inherent in functional protein interactions. Moreover, a protein interaction network with resolved interfaces helps to connect gene mutations with disease [8], and to identify possible drug targets, with inhibitors of protein-protein binding receiving increasing attention [9]–[12]. In particular, by targeting interfaces shared in multiple binding interactions, one may be able to shut down entire pathways, whereas targeting more isolated interactions offers a route for a more measured intervention. Assigning interfaces to protein interactions thus has both fundamental and practical relevance. To refine the coarse protein-protein interaction network and to capture these important structural and chemical aspects of interactions [5] requires the identification of the binding domains or interfaces on each protein. Importantly, one needs to distinguish on the basis of clear rules between binding partners that target overlapping or distinct surface regions. By systematically cataloguing these details it is possible to create not only a map of shared versus distinct binding interactions [5], [13], [14], but an entirely new sub-network of the protein interaction network, as we do here. A PPI network with interfaces overlaid on the proteins highlights the number of interfaces each protein uses to mediate binding and the number of binding partners per interface (see Figure 1b). An interface-interaction network (IIN) is what one gets by visualizing the protein interface connectivity as separated from the underlying PPI network. Unlike in the PPI network representation, in the IIN representation distinct patterns of connectivity between interfaces emerge, and this network topology can be analyzed to yield insight into the specificity and possible cooperation and competition of protein interactions. Although the importance of structural details in protein interactions has led to increasing efforts to identify protein-binding interfaces in a systematic way [13], [15]–[18], PPI networks with interfaces overlaid on them and detailed IINs have not previously been created. Earlier studies have used protein structures combined with homology modeling [5], genomic data [19], and docking algorithms [20]–[22], to both assign and infer [23] binding interfaces. While it would be possible to construct an IIN from the residue level details collected from some structural data [21], [23], both the accuracy and coverage of the network would be limited by the errors inherent to homology modeling or docking methods, and by the fact that crystallized protein complexes cover only a small percentage of known protein-protein interactions [20]. In particular, using structural homology to infer binding partners provides important guidance but may overestimate the number of binding partners because structural homologs reflect evolution but not necessarily shared functions [24], and small differences in sequences can separate specific from non-specific binding [25]. Proteins with disordered regions and without structural or domain information would be absent from the network, thus sampling only subsets of interaction types within the proteome. The presence of false positives in the IIN would obscure the diverse patterns that emerge in the network and distinguish the network structure from that of the parent PPI. The topology of a network reflects functional pressures acting to connect nodes in specific ways, whether the nodes are proteins, interfaces, or airports. One force shaping the PPI network is the need to transmit information across diverse functional modules, resulting in a giant connected component with few unconnected proteins. Differences in the IIN topology imply different functional and physical forces acting on the component interfaces. In order to analyze the specificity, cooperativity, and topological properties of an IIN one requires both an accurate assessment of shared and distinct binding interfaces and a dense collection of protein-protein interactions. Therefore, we here combine both structure-based computational approaches with literature-curated biochemical data to build an IIN for the proteins involved in clathrin-mediated endocytosis (CME) in yeast [26]. CME is a central pathway for internalizing cargo such as nutrients and signaling molecules into the cell. Assigning the interfaces mediating each protein interaction would be severely limited if we relied on structural data alone, as several of these interactions are mediated by short peptide motifs on disordered regions. Furthermore, any uncertainty associated with using models of possible domain interactions is completely bypassed by exploiting the wealth of established domain and residue information known from biochemical experiments. Through this process, we are able to carefully evaluate whether a protein' s interaction partners bind to the same interface, or to distinct interfaces. Figure 1a illustrates the results of the interface assignment for the yeast actin protein, ACT1, and Figure 1b shows the PPI network with interfaces overlaid for the subset of actin binding proteins. This representation captures both the ability of actin to bind several proteins simultaneously through four distinct interfaces, and the competition between multiple proteins to bind each of these interfaces. The resolution of this network exceeds in two important ways that of networks obtained by only defining PPI network edges as competing. First, because all of actin' s binding partners compete with at least a few others to bind actin, all PPI edges from actin would be marked as competing. Therefore it would not be possible to distinguish which of actin' s binding partners can bind at the same time. Second, because an edge connects two proteins, a network with edges marked as competing does not clarify which protein surface (say, actin' s or its partner' s) is actually shared, as sharing can occur on one or both proteins. In addition to collecting detailed data on protein structures, a particular advantage of our curated approach is to eliminate false positives from the PPI by creating a coherent and consistent picture of the protein interactions. We identify the specific mechanism mediating an observed protein-protein interaction and determine whether the interaction is direct or indirect. Of particular concern are indirect interactions, mediated through intervening proteins, because they are not always distinguishable from direct interactions in high-throughput affinity purification/mass spectrometry (AP/MS) [3], protein-fragment complementation assay (PCA) [27] or, to a lesser extent, yeast two hybrid (Y2H) experiments [3]. Literature sources also document protein pairs tested and found to not bind. Therefore, by curating the literature we do not predict new interactions but we do remove spurious interactions. We also compile the number and types of experiments used to identify the interfaces in each protein-protein interaction, as the interfaces can vary from high-resolution selections of specific residues to low-resolution large regions of the protein. This compilation provides a starting point for improving the resolution of the structural interaction. The CME network constructed and characterized in this way reveals significant complexity with permanent and dynamic assemblies of few or many proteins, a mixture of binding modes with both shared and distinct binding, and both large and small binding interfaces. This detailed information is necessary for building models of protein-protein interactions where both competitive and cooperative binding reactions contribute to function. The accuracy and coverage of the protein IIN we have generated allows us to draw generalizable insights about the structure of the IIN, the overlap of binding interfaces, the identification of indirect interactions, and the implications towards the biological functions with the parent PPI. Compiling this information for more parts of PPI networks will help prune indirect and spurious interactions, highlight areas of poorly resolved structural and biochemical characterization, and facilitate investigation of the physical and evolutionary origins of the IIN topology and in turn of protein binding. The aims of the present work are (1) to develop a general framework for the construction of IINs from a combination of structural and biochemical data that measure the support of proposed protein interactions; (2) to characterize the general network properties of the resulting endocytosis IIN as compared to PPI networks and randomized networks, and (3) to demonstrate the applications of the IIN, including as a resource for predicting response to mutation and to specific binding inhibitors. At the network level, we examine whether the IIN retains the complex characteristics of the PPI network, including a high connectivity, hub structures, local clustering, and a scale-free character manifested in power-law distributions of the number of binding partners. We also quantify the fragmentation of the fully connected parent PPI network into separate interface modules at the IIN level. We provide several examples for the use of the IIN in selecting possible drug targets and in predicting the effects of mutations by identifying specific pathways of communication between proteins via their interfaces. For the CME proteins we discuss the central role of SH3 domains and multi-interface proteins. We emphasize that the process of assigning protein interfaces has generated not only a useful map of interactions among these highly-studied proteins but has highlighted the difficulties associated with trying to make automated assignments, including overlapping residues and inconsistencies between sources. Therefore, we discuss the insights derived from our interface procurement process that are relevant for high-throughput methods of interface determination. As a first step, we constructed a curated PPI network of 56 proteins involved in CME in yeast [26]. Following the approach described in Methods, we first combined 337 edges downloaded from BioGRID [28] via the Saccharomyces Genome Database (SGD) [29] and 49 additional distinct edges collected from IntAct, MINT, DIP, and BIND. 177 edges had interfaces assigned to both proteins in the interaction and nine additional edges were added from literature evidence. We note that for these 56 proteins, we observed significant overlap in the interactions reported in each protein-protein interaction database, as listed in Table 1. Of the assigned protein-protein interactions, sixteen had two binding modes, and two had three binding modes, resulting in a total of 206 assigned interface-interface interactions from 186 assigned protein-protein interactions. We removed 35 edges from the original network because they were suspected to be indirect, shown not to bind in further experiments, or they occurred only in a study of yeast prions, suggesting that the observed binding may not normally be functional. For 28 interactions identified in multiple high-throughput studies no evidence from the literature was found to assign interfaces, and 145 interactions were found only in one reference without sufficient information to assign an interface. Nearly all of the 145 unassigned interactions that were implicated in a single experiment came from high-throughput studies, and because the proteins in the CME subset form the connected clathrin coat and actin patch together, many of these observed interactions could be indirect. The differing support for the CME protein interactions is represented visually in Figure 2 and collected along with specific details of their assignment in tabulated form in Table S1. The tabulated list contains all the currently known interactions between these 56 CME proteins and the interface assignment status effectively ranks them in terms of their reliability. Interactions with interfaces assigned are further classified in Figure 3 according to the experimental data used to make the assignment. The blue edges in the Figure 2 network are unresolved interactions that have the most evidence (more than one study) supporting their potential functionality in the cell. The red edges are most likely to be artifacts of the experimental probes of their interaction, on the basis of evidence listed in Table S1. Of the red ‘false positive’ edges, most were indirectly interacting through a larger complex. A few had been shown in detailed biochemical characterizations not to bind to one another. In Figure 3 we also distinguish the amount and type of evidence used to support each interface assignment by coloring each edge. The number of interface interactions assigned directly from crystal structures is shown in black, and represents a minor fraction of the total assignments for this network. For blue edges, both interfaces were resolved using biochemical studies, typically by truncating or mutating the constituent proteins. We note that although these binding interactions have been tested in vitro and in some cases in vivo (thicker blue lines), some of the interfaces encompass large folded domains rather than specific surface binding residues. These domains could therefore be segregated further into more than one interface given additional resolution of the specific residues involved in each interaction. Green and cyan edges had one or both interface inferred. For these assignments we relied on homology to other proteins either through sequence, function or crystal structure. Alternatively, a lack of competition for binding to a surface or a lack of any structural or functional homology was sometimes used to infer distinct vs shared interfaces. Finally, the yellow edges are speculated to be distinct interfaces due to a lack of observed similarity to known partners or domains, and as such have the weakest support. We determined the degree distributions of both the original and the curated PPI networks, as a statistical measure of the number of interaction partners of each node. Upon going from the full combined-database PPI network (plus the added 9 edges; see Methods) to the curated PPI network, the decrease in total edges results in a less dense network in which the average number of partners per protein dropped significantly from 13. 5 to 6. 4. Although large PPI networks typically have degree distributions characterized as power law or truncated power laws [30], [31], neither the curated PPI nor the combined-database PPI have degree distributions statistically consistent with a power law density (Figure 4a). The deviations from a power law could be due to the small size of the two networks (only 56 proteins) and the fact that they are all part of a functionally related module. Such deviations would be exacerbated in the original un-curated network, where the distribution is more uniform for N<10, by spurious and likely indirect interactions within the set of proteins absent in the curated PPI network. Lastly, both of the original and curated endocytosis PPI networks had high clustering coefficients (0. 56 and 0. 46, respectively; see Methods) indicating that proteins that interacted had partners that were likely to interact with one another. Not surprisingly, the clustering coefficient of 0. 28 in a full yeast PPI collected from several large-scale studies in yeast [3], [27] is lower, since the CME proteins were specifically chosen to be part of the same functional module. The majority of proteins in this endocytosis network have multiple interfaces, with an average number of 3. 5 distinct binding interfaces per protein. The correlation between protein size and the number of interfaces is quite weak (R = 0. 26). One reason for this weak correlation is that several of these proteins have additional binding partners outside the CME network module considered here. Another reason is the size of the interfaces varies broadly, from just a few residues (for NPF motifs [32]) to hundreds of residues (e. g. , the clathrin-clathrin leg binding [33]). For example, LAS17 is a medium sized protein at 633 residues that has the most interfaces thanks in part to five short proline-rich domains (PRDs) we assigned as distinct interfaces due to their specificity towards different binding partners [34], [35]. The black edges in Figure 3 connect interfaces determined through crystal structures that therefore are defined by a subset of non-contiguous residues. Interfaces from biochemical studies tend to lack single residue resolution but instead span stretches of residues or complete domains. In the 3D protein structure, only a fragment of these residues would be expected to contact binding partners and as such some of these interfaces could be split or refined further. It is interesting to compare the change in network properties between the curated PPI network (shown in Figure 3a with interfaces overlaid) and its IIN (Figure 3b). In general, a PPI network and its IIN should have equal numbers of edges, but it is possible for an IIN to have more edges if a pair of interacting proteins has multiple modes of binding to one another. Proteins that act as alternating subunits in a symmetric complex, for example, will contact two copies of the same partner through distinct interfaces. The CME IIN contains several instances of multiple binding modes, resulting in an increase in edges from the PPI. Such distinct modes for the same two proteins to bind one another can act as a regulatory mechanism controlling the accessibility of surfaces on the protein, or as sources of extra stability to the protein-protein interaction. For example, the protein CRN1 contains two distinct actin-binding domains that bind separate regions on the actin surface and are modulated by the nucleotide bound state of actin. Through these multiple binding modes, CRN1 can have opposite roles in either inhibiting or activating the severing of actin filaments [36]. In another example, the SH3 domain of LSB3 binds three distinct PRDS on LAS17. The PRDS on LAS17 follow one after another and the flexibility to bind any one of them to the SH3 domain of LSB3 could help stabilize the binding interaction at different geometries as part of the higher-order actin patch assembly. The IIN contains more nodes than the PPI, with each node now representing a distinct interface rather than a protein. In general, one would expect such an increase because proteins are known to have evolved multiple domains or interfaces to bind specific partners. The increase in nodes is much greater than the increase in edges from the PPI to the IIN, and therefore the IIN is substantially less densely connected than the PPI, with the average degree dropping from 6. 4 to 2. 06. In this now sparsely connected network, the clustering coefficient has dropped from 0. 46 in the PPI to zero in the IIN. To quantify the significance of this result we generated randomized versions of the IIN that maintain the same number of nodes and the same degree distribution. We find that the randomized networks have distinctly higher clustering coefficients than the IIN (Table 2), suggesting that the structure of the IIN has evolved against having interfaces that bind to one another sharing the same partners. This is in contrast to the PPI network, where the relatively large clustering coefficient reflects the likelihood that two proteins that interact with one another share interaction partners. To further quantify the significance of the local structural elements in the IIN, we evaluate the relative abundance of all 6 different types of 4-node motifs in the network. As shown in Table 2, their abundance in the IIN differs significantly from randomized networks. In particular, 4-node hubs with one shared interface binding to four non-shared interfaces are abundant in the actual IIN, and 4-node chains with four shared interfaces forming a linear chain of interactions are suppressed. The low abundance of these motifs is expected from network specificity optimization [37]. Interestingly, 4-node squares formed by four shared interfaces binding as in A1-B1, B1-A2, A2-B2, B2-A1 are also enriched in the IIN, forming a motif that has high specificity and can arise from gene duplication. Collectively these abundance shifts suggest that the local structure of the IIN is not random but reflects distinct evolutionary mechanisms acting on its topology. Moving to the global properties of the IIN, we find that the distribution of binding partners per interface follows a power law quite well, with most interfaces having only a single binding partner (Figure 4b). The degree distribution of the IIN is constrained by the parent PPI degree distribution but not fully determined by it, as a PPI can theoretically give rise to many IINs with distinct numbers of nodes and connectivity (but each IIN uniquely defines its parent PPI). Hence a power-law distribution of the number of partners per interface is not a trivial outcome of having a power-law distribution of the number of partners per protein. We do expect that the number of nodes in the IIN will increase relative to the PPI and therefore the number of partners per node will be split between more nodes (assuming the number of edges stays about the same). How exactly the degree distribution changes from PPI to IIN then depends on whether it is mostly the highly connected hub proteins that are split about equally between multiple interfaces, or whether some interfaces retain large portions of binding partners and several single partner interfaces are created. In the CME network, the maximally connected node in the PPI (actin) is split between interfaces, but not evenly, such that one interface retains the majority (16) of the 23 binding partners. Overall, the IIN contains a significant number of highly connected nodes, just as in the PPI. The biggest change in the degree distribution from the PPI to the IIN was the formation of many single-partner interfaces in the IIN, whereas a protein in the PPI was more likely to have at least 3 partners. We discuss further below whether these trends might be conserved in other IINs. Another distinguishing feature of the IIN is its fragmentation into modules, unlike the densely connected PPI. Compared to randomized networks, the CME IIN has a diverse distribution of module sizes, with many small fragments, whereas randomized networks all have a single giant connected component alongside many small fragments (Figure 5). In fact, the number of interfaces in each CME fragment again appears to follow a power law distribution with an exponent of about −2 (Figure 5). As a result, isolated small modules dominate, but larger connected networks even at the interface level are not uncommon. One must keep in mind, though, that here we focus on only a limited, functionally defined module. In future studies, it will thus be interesting to examine other IINs resolved at the same level of detail. The modules in the IIN start to show clustering of interfaces with shared properties, although to varying degrees. In Figures 3a and 3b, we colored the interfaces according to specific domain types that are repeated in the network: PRDs and SH3 domains; EH domains and NPF motifs; phosphorylation sites and kinase domains; clathrin boxes; acidic domains; and subunit-subunit interfaces. As seen in Figure 3a, at the PPI level these interface types are mixed (i. e. , distributed across different proteins); by contrast, we find them to be clustered into separate IIN modules. In randomized networks such clustering is not observed. This clustering of interface types reflects the need for binding interfaces to maintain high specificity towards their complementary binding partners and against binding towards unrelated interface sequences [37]. We note that our choice of defining all phosphorylation sites as distinct interfaces places them all in the same module (see Methods), whereas an alternative definition (for example, treating any phosphorylated residues overlapping with other interfaces as forming shared interfaces) would distribute some of them throughout the network. By contrast, the actin ACT1. 2 interface is part of a large module with significant heterogeneity in domain types, as discussed further below. Because these binding interfaces do not all contact the same residues of the ACT1. 2 interface, they do not all classify according to a single domain type. The convergence of these distinct partners to bind a single protein surface seems more likely a result of functional selection rather than duplication and divergence [38]. The IIN shares some of the scale-free characteristics of PPI networks [30], yet differs markedly in a number of network topological properties, including a lower average degree of the IIN and a more fragmented structure. While strictly applying only to the CME IIN, we expect many of these results to be conserved in IINs derived from larger PPI networks. First, the comparison of the IIN structure with randomized networks suggests evolutionary pressure acting on the IIN to prevent both giant connected components and a high clustering coefficient (where two interacting interfaces have the same partners). Second, interfaces that have only a single partner should be robustly conserved even for larger networks because they frequently mediate inter-subunit contacts (see light green nodes in Figure 3), and can evolve to high specificity [37]. A noticeable increase in singly connected nodes when transitioning from PPI to IIN would contribute to a steep power law-type degree distribution as a general trend. If a few hub interfaces and many single interfaces were maintained in other IINs, their degree distributions would resemble power laws. The degree distribution of interface partners is noteworthy because power-law distributions indicate networks robust against attacks on specific nodes [39], as would occur from mutations to specific binding surfaces or targeting by binding inhibitors. In a separate study, we will pursue the hypothesis that the structure of the IIN evolved to minimize nonspecific binding, and that therefore the network features of the IIN encode important physical and biological functions of the proteins. Since minimization of nonspecific binding is a physical pressure common to all proteins [37], [40], we would predict that these topological features would then be conserved in all IINs, not just for CME proteins. A number of distinct patterns emerge in the CME IIN. From the degree distribution of the IIN, we can contrast the properties of single interfaces from hub interfaces. More than a quarter of the single partner interfaces come from interfaces between subunits of a multi-subunit complex like ARP2/3 [41]. Dimerization interfaces also tend to be single partner interfaces. The most highly connected interfaces, or hub interfaces, are a surface on the actin protein with 16 partners, and several SH3 domains. The actin surface is distinct from the SH3 domains in that its binding partners do not all conform to the same binding type. The binding interface ACT1. 2 is a relatively large and flat region spanning parts of subunits I, II and III of actin (Figure 1a), where not all binding partners use the same set of residues to stabilize their interactions, but the overlap is still significant. While it is certainly possible that with additional residue information this interface could be refined and split into more than one binding site, the extensive sharing of the ACT1. 2 interface is consistent with earlier studies that found flat interfaces to provide a better platform for binding a large variety of partners [42], as geometrical packing need not be as optimized. Furthermore, we note that the nucleotide binding state of actin strongly tunes the affinity for its distinct partners. The IIN overlaid on the PPI reinforces that many of these endocytic proteins are able to bind multiple partners simultaneously because of the number of distinct interfaces. This directly observable insight would be lost if one only categorized protein-protein interactions (i. e. , edges in the PPI) as either competing or not, since many proteins have multiple shared interfaces. The interface assignments also highlight redundancy in the network, where the recruitment of a particular protein during the endocytic pathway could happen via multiple mechanisms, as many of the proteins are chimeras of the most frequently represented domains [43] in this network. Of the endocytic proteins, 16% contain SH3 domains, whereas in the entire yeast proteome <1% of proteins contain SH3 domains. The designation of distinct domains on each protein allows one to contrast the specific structural elements present in these CME proteins versus CME proteins in other organisms. Much of the CME pathway between yeast and mammals is conserved. However, a major distinction is that CME in yeast requires the actin network to initiate the membrane invagination [44], whereas in mammals the actin network is engaged only in some cases in the later stages of vesicle budding [45]. It is interesting to note that of the 9 CME proteins in yeast without functional homologs in mammals [26], all but one (PAL1) engage in SH3 or PRD interactions (LSB3, LSB4, LSB5, BBC1, AIM21, BSP1, AIM3, APP1). This finding is statistically significant, having only a ∼0. 15% probability to occur by chance (as determined by the probability of choosing at random 8 or more proteins out of 9 that have SH3 or PRD interactions, with 22 candidates among the 56 proteins of the CME network). The SH3 domains in the CME network recruit proteins throughout the progression of the vesicle budding process after the initial clathrin coat assembly [35]. The abundance of SH3 domains in yeast CME proteins likely reflects the central role of their interactions in connecting the growing clathrin coated pit to the actin cytoskeletal network of yeast. Distinguishing interface domains in each protein also enables direct visual identification of multi-interface proteins that act to bring together multiple proteins with different functions, which again is not possible if one only marks edges in the PPI as competing. Both PAN1 and SLA1 have many interfaces that can connect simultaneously to both the scaffolding proteins of the clathrin pit formation (through PAN1' s EH domain [32] and SLA1' s clathrin box [46]), and to the actin polymerization proteins via PRDs, SH3 domains, and acidic domains. LAS17, on the other hand, does not connect directly to the scaffold proteins of clathrin pit formation but rather has distinct interfaces to bind both SH3 proteins and the ARP2/3 complex. While the role of LAS17 is not fully understood [26] and appears to involve both activation and inhibition of actin branching [47], the designated interface-interface pairs provide a basis for grouping the many functions of this protein along with distinct CME proteins according to domain types (including PRDs, acidic domains, the C-helix and WH2 domains). Lastly, some multi-interface proteins in the network, such as Arc15 and Arc19, contain only subunit-subunit interface domains, indicating that they function as structural components of a multi-subunit complex. Designing any ligand, and in particular a drug molecule, to bind exclusively to its intended target without cross-reactivity requires not only positive selection for the specific target but also negative selection against related targets [37]. The clustering of interfaces in modules in the IIN provides a tool for predicting which binding partners of an interface are the most selective for its surface and do not bind to related domains. For example, both RVS167. 2 and the SH3 domain of YSC84/LSB4 (YSC84. 1) bind several of the same PRDs. Obtaining target specificity for only one of those interface sites benefits from knowing which PRDs are specific to only one of these interfaces. The interfaces VRP1. 0, BSP1. 3, and ABP1. 1 that bind RVS167 but not YSC84, and ABP1. 5, LAS17. 6 and AIM21. 0 that bind YSC84 but not RVS167, could be used as templates for targeting only one of the two SH3 domains. Collectively, the information on interface connectivity and protein connectivity combined in a network format provides important guidance for the selective inhibition or activation of specific pathways, for drug targeting, and for predicting response to surface mutations. The PPI network is essential for identifying which proteins interact in a functional pathway, but the details of the IIN allow one to isolate specific binding sites while conserving the functionality of other sites. The IIN also allows one to predict how drugs designed as roadblocks along a certain pathway could be bypassed by alternate available interface interactions. For example, one might expect that inhibiting or mutating ‘hub’ interfaces, much like knocking out ‘hub’ proteins, would induce a more severe phenotype. RVS167. 2 is found to interact with 12 PRD interfaces as part of the PRD-SH3 IIN sub-network in the top right of Figure 3b. These interactions are not immediately apparent in the PPI network, lacking interface resolution. While early studies [34], [48], [49] already pointed to the prevalence of these interactions, their functional importance and temporal recruitment in endocytosis is emerging only now [35], [50], [51]. Mutations of the SH3 domain of RVS167 that leave its membrane shaping BAR domain intact still significantly alter the endocytic phenotype [50]. A substantial phenotypic response to such a localized mutation would be anticipated from the IIN because removing that particular node removes multiple edges. However, the fragmented and clustered structure of the IIN also provides a more detailed perspective on the response to deleting this node. Although targeting the SH3 domain of RVS167 would inhibit 12 RVS167 binding interactions, one can see in the IIN that most of those interface partners can also bind to alternate SH3 domain containing proteins and all of the interface partners are on proteins with a PRD that can bind an alternate SH3 domain. These alternate pathways may help explain why mutations of the SH3 domain of RVS167 do not eliminate endocytic function in yeast [50]. The inhibition of particular binding sites would have unexpected results if there were nodes or edges missing from the network. For example, truncation of the clathrin N-terminal domain (CHC1. 1 in Figure 3b) was accurately predicted to cause a severe endocytic phenotype by preventing recruitment of clathrin to the membrane. However, when the known binding sites on the N-terminal domain were mutated, the expected result was not observed, and this led to the identification of duplicate binding sites in the N-terminal domain [52]. We do note that the CME IIN overlaid on the PPI proposes another mechanism for recruitment of clathrin to the membrane via binding of the clathrin light chain (CLC1) to SLA2, which can then bind the membrane or other membrane bound proteins. This interaction may be too weak to recruit clathrin on its own, or SLA2 may be too small to bridge the large separation from the clathrin light chain to the membrane. The IIN also indicates sets of opposing or reciprocal mutations, or truncations that should result in the same phenotypic response. The prediction of the identical responses assumes that the binding partners of the targeted interface act independently of other binding partners. The extent to which these assumptions are violated could suggest allostery or cooperativity between the affected partners. Identifying interfaces for reciprocal mutations could then offer a tool for testing cooperativity or dependence between binding interactions or for identifying missing interfaces. In the clathrin CHC1. 1 interface example given above, the IIN would predict a reciprocal mutation to all five clathrin boxes to give the same phenotype as the removal of the CHC1. 1 interface. If, instead, clathrin were still recruited to the membrane, then one expects other clathrin N-terminal binding sites to be missing from the network. In another module, the EH domains are shown with their NPF motif binding partners. Based on the specificity of these interfaces for one another only in the IIN, one would expect that mutations to either the NPF motifs (including all copies) or to the EH domains (including all copies) would generate the same phenotype. The extent to which they do not match would first indicate possible missing nodes from the network. Alternatively, the result could indicate that one of these domains acts cooperatively with another domain to affect the global behavior of the protein, not just this specific interaction. In terms of the anticipated biological response to mutation of either the EH domain or the NPF motifs (assuming independence), this interaction helps stabilize a scaffold of proteins at the membrane that recruits the clathrin trimer. From the IIN combined with the PPI, cutting these edges out of the network would not prevent any of the proteins from connecting to the membrane or the early coat module, as PAN1 could still connect via SLA2 and EDE1 via SYP1. Clathrin and actin would still be recruited normally. What this mutation should affect is crosslinking between these proteins and therefore clustering of these proteins in one place on the membrane. If crosslinking and clustering of proteins is necessary for efficient coat formation then eliminating these interactions could decrease or slow down clathrin-pit formation. As one of the main challenges in IIN construction, there is more than one way to define whether a binding interface on a protein is shared between multiple binding partners or is completely distinct. The two main criteria we use to characterize shared and distinct interfaces are (1) if the same residues are present in both interfaces, (2) if the binding of one protein partner would interfere with the binding of another partner due to structural overlap or allosteric effects. Both criteria are important to the function of the proteins in the cell. Concerning the first criterion, the sequence makeup of the interface is central to achieving binding specificity, as even proteins with the same domain structures do not necessarily share the same partners [35]. Furthermore, the residues involved in a binding interaction are not only important for binding to their specific partner but also for avoiding the formation of nonfunctional interactions with the other proteins in the cell [37]. This negative selection on an interface can contribute to optimizing the specificity and strength of functional binding interactions [25]. Concerning the second criterion, determining whether two potential binding partners can both bind at the same time to form a trimer is important for modeling the dynamics of protein association, as competition for binding partners will affect concentrations of available protein. The same is true if a protein has repeated copies of the same domain and can therefore bind multiple copies of the same binding partner. However, it may not always be possible to assign distinct interfaces that meet both criteria of sequence specificity without any steric obstruction and therefore in this work we consistently emphasize residue detail where the information is available. Otherwise we did use competition for binding partners as grounds for defining shared versus distinct interfaces. In future work it would be valuable to annotate both the residues involved in each interface as well as whether each pair of distinct interfaces on a protein can bind their partners simultaneously. The procedure of manually assigning interfaces has also highlighted some important issues for consideration in computerized interface assignment. For one, residue overlap does not necessarily mean that proteins compete for binding to the protein, as demonstrated by multi-subunit complex formation (Table 3). None of the interface interactions within the complex would be considered shared because they are all bound together at the same time. The majority of interfaces do not overlap, but ∼30% of the bound partners share one or more residues in their interactions. Most commonly the overlap was only one or two residues, and the corresponding percentage of the interface varied substantially depending on the size of the protein. Thus it seems reasonable to allow 1–2 residues of overlap before defining interfaces as shared. This policy is also consistent with the assignment of different domains as distinct interfaces, even though the 3D structure of the protein might produce some residue overlap between two distinct domains. We note that here we did not use a strict cutoff in our assignments because through manual curation we treated each interaction on a case-by-case basis, merging residue level detail with experimental data on simultaneous binding partners. For attempts at homology modeling or docking, it would first be useful to assess how reliable a purported interaction might be. Particularly in the case of interactions involving subunits of multi-protein complexes, many of the interactions are actually indirect. Arp2, for instance, has relatively high homology to actin and shares several binding partners; however, Arp2 acts as part of a multi-subunit complex and binds to these shared partners (such as LAS17) in distinct ways. Also, higher thresholds for sequence similarity could be warranted in particular cases, such as SH3 domains, where small variations in sequence distinguish specific from nonspecific partners [25]. One of the major distinctions between the procedure used here and current automated methods is the inclusion of detailed information on binding interfaces between proteins from biochemical studies, not just from high-resolution protein structures. This information preempts or complements the use of homology or docking models of protein interactions. Unfortunately, the domain or interface details from these studies is not collected in a convenient database, whether it is the specific residues that comprise the interface or more general information on inhibition or competition between binding partners. There are also ambiguities and inconsistencies in existing data that are difficult to resolve without combining multiple literature resources in a coherent analysis. Nevertheless, mining these data would provide a valuable resource for generating more complete networks of interface-interface interactions. Our protein list is composed of 56 proteins that were selected because they all participate in the yeast clathrin-mediated endocytosis pathway and have been identified as central components [26]. We downloaded the physical interaction partners of the 56 proteins of the endocytic functional module in yeast via the Saccharomyces Genome Database (SGD) [29] interaction list compiled from BioGRID [28] and directly from the IntAct, MINT, DIP, and BIND protein interaction databases. We kept only the interactions between the subset of 56 proteins to define the initial set of experimentally determined protein-protein interactions. We disregarded genetic interactions, as they do not imply that the proteins directly interact with one another, but rather that their expression or phenotype is correlated. The overlap in databases was quite large for these proteins, with BioGRID containing the largest number of interactions and missing interactions coming not from missed references but from missed interactions within the same references. Given a PPI network, the first step in assigning the binding interfaces was collecting information on the particular proteins from the SGD [29]. The SGD combines information from various databases on each yeast gene. The major data sources we used were the list of referenced physical interactions loaded from the various PPI databases and the available PDB structures. The protein tab also provides a useful guide to the size, sequence, domain structure, and function of each protein. Crystal structures of complexes were available for a few of the protein interactions, including the ARP2/3 complex and several actin binding interactions (shown as black edges in Figure 3). We ensured that we matched the numbered PDB residues (which sometimes started at zero arbitrarily) to the correct sequence region on the protein of interest. For protein homologs from species other than yeast, the sequence alignment is also provided for positioning the interface on the yeast protein of interest. Compared across species, actin has high (87%) sequence homology and structures from other species were simply used as proxies for the expected interaction in yeast. To assign residues involved in the protein interfaces from a PDB complex we used a 4-Å cutoff between non-hydrogen atoms and required that at least 3 residues contacted one another in each interface. Cofactors such as metal ions and water molecules were not considered in assigning whether two proteins interacted or which residues formed the interfaces. Some of the protein structures had multiple missing residues for crystallization purposes, such that the assigned interface may be smaller than in the complete protein. By using the PDB structures we eliminate all indirect interactions that are often assigned to protein subunits of a large complex in high-throughput AP/MS and PCA. We did not use any predicted models of protein complexes [23] because direct information was generally available through literature studies and because protein homologs (e. g. , Arp2 and actin) do not always share the same set of binding interactions. In most cases, crystal structures were not available and instead the literature references from the PPI databases were used to assign interfaces. Binding to proteins outside the endocytic network, as listed in the SGD, was ignored. Nearly all of the edges to which we assigned interfaces were implicated as binding in more than one experiment. We have collected all the justifications for each assignment into a spreadsheet with references (see Table S1), categorized the support for each interface assignment with edge colors in Figure 3, and below we describe additional criteria we used to define the interfaces for the specific cases of kinase binding and SH3 domains binding to PRDs. Several of the endocytic proteins have SH3 domains (BZZ1, ABP1, LSB3, LSB4, RVS167, BBC1, MYO3, MYO5, and SLA1) and PRDs to which SH3 domains bind (VRP1, LAS17, MYO5, APP1, AIM21, AIM3, SCD5, BBC1, ABP1, ARK1, PRK1, INP52, SCP1, BSP1, SLA1, SYP1, GTS1). We took advantage of several large-scale studies [34], [35], [48] focused on identifying which PRDs bind to which SH3 domains by compiling all interactions noted for our 56 proteins (including those interactions missing from the PPI databases). Tonikian et al. [35] provide the most recent and comprehensive study to identify PRDs by combining data from three independent experiments. We assigned the PRD and SH3 interfaces if the interactions were observed by Tonikian et al. and at least one other experimental study. As one exception to this criterion, if there is only one supporting experiment, yet that experiment found a different PRD site, then the interface was left unassigned. Lastly, if more than two references reported binding and the PRDs were different, the two PRDs were combined into one binding site. Binding multiple PRDs on the same protein has been experimentally demonstrated [34], but Tonikian et al. only report the most likely PRD, so this does not rule out additional PRDs. We merged the two SH3 domains of BZZ1 to improve the consensus of their binding partner interfaces but kept the two SH3 domains of SLA1 separate. We separated the multiple PRDs of LAS17 into distinct binding sites as multiple lines of evidence implicated specific binding partners for specific regions. These details are collected in Table S1, under tabs 2 and 3. The endocytic protein subset contains three kinases (ARK1, PRK1, AKL1) and similar to the SH3 domains, the specificity of kinases for their phosphorylation targets has also been studied at large scale [53], [54]. We here again compiled the interactions from Breitkreutz [53], Mok [55], and Ptacek [54] and their collaborators (again including some interactions missing from the PPI databases), and assigned the interactions if the binding was reported in at least two references. Because most of the targets in Mok et. al. [55] were predicted but not verified, we included these sites as references only if they were also experimentally tested or observed in previous mass spectrometry experiments. These details are collected in Table S1, under tab 4. In some cases the data implicating two proteins as interacting only came from high-throughput studies and these interactions were generally unassigned. Others came from a literature source that did not isolate binding interfaces, with no additional evidence available from homologs or functionally related proteins. Edges that were identified between the ARP2/3 complex subunits and other proteins were considered indirect if PDB structures or biochemical evidence implicated a specific subunit in the direct interaction. For a few interactions, evidence from the literature suggested that such proteins did not bind directly to one another upon further investigation, and as a result these edges were removed. We note these in the interaction table. For example, we were unable to find any evidence for the protein RVS161 forming direct physical interactions with any proteins other than RVS167. Furthermore, there was some biochemical evidence suggesting that proposed edge interactions were mediated via RVS167 rather than directly through RVS161 [56], as they operate as an obligate dimer. We added 9 new edges to the network to account for literature studies providing evidence for the binding interactions. These were largely actin related interactions that lacked references in the PPI databases but have been well established as functionally important binding partners of actin. One was an SH3-PRD interactions defined in two separate publications that were missing from the database. Several of these proteins have domains known to bind at the membrane that are important to their function in endocytosis. Therefore we pointed these out on the protein-interface interaction network in Figure 3a to facilitate the prediction of functional responses to mutation. As the first criterion to assign an interface, we used the residues involved in the binding, if available. Specific residues were available from PDB structures and for several peptide motifs like PRDs [34], [35] and clathrin boxes [32], [46], [57]. If two partners of a protein bind to an interface using some overlapping residues we did not automatically classify the interface as shared. There are two reasons for this decision, the main one being that sharing one or a few residues does not mean those two proteins cannot bind simultaneously. To demonstrate this point we calculated the percent of distinct interface pairs within a multi-subunit protein complex that had overlapping residues. For each of the complexes we considered, there are some pairs of interfaces that have one or more residues in common (Table 3). Even if the interfaces are defined at the atomic rather than residue level, there is still a fraction of atoms within the cutoff distance of both binding partners. The second main reason is that even if the binding partners cannot bind simultaneously, the specificity and stability of their interactions may be mediated through chemically distinct binding sequences. For example, we chose to treat a kinase' s phosphorylation binding site as distinct from other protein binders that may interact with the phosphorylated residue because of their distinct binding modes. However, if the residue overlap is substantial, as is the case for many proteins that bind to actin in similar but not identical ways, then the interface is considered shared. When the specific residues of the folded protein interfaces were not available, the next description of the interface was the domain structure represented by sequential sequence residues (e. g. , the SH3 domain contained in residues 1–51). These domains were generally identified in biochemical studies and the sizes of the domains varied from a few residues (e. g. , clathrin boxes) to hundreds of residues (e. g. , coiled-coil domains). In some cases the assigned interfaces may not represent a known domain but they are designated as unique interfaces because they do not overlap with any of the protein' s other binding partners. Lastly, if residue level detail is not available, then the fact that two binding partners are competing with one another is used as justification for listing the interface as shared. To summarize, we did not use a strict cutoff of overlapping residue numbers for defining shared versus distinct interfaces. All subunits of a multi-subunit complex were assigned distinct interfaces for these inter-subunit contacts because they could clearly bind simultaneously. This is despite the fact that pairs of proteins could have as many as 10 overlapping residues if a long disordered region of a protein sat at the seam of an interaction between two other proteins. For most biochemical studies, stretches of residues were identified and shared interfaces were assigned when proteins bound to overlapping stretches of residues and there was no evidence that they could bind simultaneously. For the distinct surfaces in actin, there was in some cases overlap between residues, but there was also evidence that the proteins could bind simultaneously. For example, several actin binding proteins bind to the actin filaments, and therefore they can bind simultaneously with the actin-actin binding interactions, despite overlapping with some residues. In representations simpler than the IIN, edges in the PPI network have been marked as shared. To extend the representation to full interface assignments, one must keep track of possible overlap in all pairs of binding interactions for each protein. Given a protein that has k binding partners, there are k (k−1) /2 possible pairs of partners sharing an interface. To keep track of the interface assignments, each protein had its own file with a k-by-k matrix indicating the overlap between the k binding partners (Table 4). The diagonal entries are null and the off-diagonal entries of the symmetric matrix are 0 if the two partners use separate interfaces and 1 if the two partners use the same interface. Some protein-protein interactions are controlled by more than one set of interfaces and would require an additional entry into the matrix. The binding interfaces from each protein can then be consolidated into a network representing a connected set of interface interactions. We note that in a matrix representation it is possible to define a case where one interface overlaps with two others that do not overlap with each other, and this detail cannot be captured in a simple interface network picture. This would be the case, e. g. , if two proteins A and B bind to two distinct parts of a protein X and the third protein C binds across those two complete interfaces on protein X. However, this issue can easily be fixed by splitting protein C' s interface into two interfaces to bind the two parts of protein X. For example, this splitting was done for LAS17' s CA region that binds to ARP3 through both its C interface and its separate A interface [47]. We evaluated clustering coefficients of our networks using the expression [58]where Nclosed (i) counts how many distinct pairs of the k (i) partners of interface i have an edge between them to form closed triangles with node i. Self-loops were ignored in this calculation. We also use a global clustering coefficient Cglobal as the number of distinct closed triangles Ntriangle in the network divided by the total number of distinct triplets, with Nopen the number of open triplets. We computed degree distributions, p (k), where k counts the number of partners per node (either protein or interface), and p (k) is the probability for finding a node in the network with that degree. For the degree distribution, we note that we treated self-loops as a single partner, rather than the standard method of treating a self-loop as counting as degree of 2, so that the degree would reflect the physical number of binding partners per protein (or interface). We enumerated the 4-node motifs present in our networks by identifying all distinct sets of 4-node subgraphs that are connected by at least one path (each node in the subgraph can be reached by the others). There are six distinct 4-node subgraph architectures [59] and we note that they are all counted mutually exclusive to one another, i. e. , a set of 4 nodes uniquely classifies as one of the six subgraphs. A single node may belong to more than one 4-node subgraph. Hub and chain motifs have 4 nodes connected by 3 edges, flag and square motifs have 4 nodes connected by 4 edges, and the other two 4-node subgraphs contain 5 and 6 edges. To generate networks that shared the same number of interfaces, edges, and the same degree distribution as the IIN in Figure 3b, we used the Monte Carlo method of Maslov and Sneppen [60]. Specifically, in a trial move two interfaces were selected randomly and a partner from each of these interfaces was randomly selected. The partners were then swapped between interfaces, unless one of these new edges already existed, in which case the move was rejected. We fit our degree distributions to power laws using the maximum likelihood method, where the discrete data is fit to a power law distribution x−γ/ζ (γ) normalized over the range x≥xmin [31]. We measure the goodness-of-fit using the Kolmogorov-Smirnov metric and calculate the p-value for the data being drawn from a power law density using the method of ref. [31]. For the p-value calculation, our null hypothesis is that the data is drawn from a power-law density. Therefore, a small p-value of <0. 05 would reject this null hypothesis and demonstrate that our data is not described by a power law. A large p-value, on the other-hand, indicates that the data is consistent with the hypothesis that it was drawn from a power law distribution.

Much of the work inside the cell is carried out by proteins interacting with other proteins. Each edge in a protein-protein interaction network reflects these functional interactions and each node a separate protein, creating a complex structure that nevertheless follows well-established global and local patterns related to robust protein function. However, this network is not detailed enough to assess whether a particular protein can bind multiple interaction partners simultaneously through distinct interfaces, or whether the partners targeting a specific interface share similar structural or chemical properties. By breaking each protein node into its constituent interface nodes, we generate and assess such a detailed new network. To sample protein binding interactions broadly and accurately beyond those seen in crystal structures, our method combines computational interface assignment with data from biochemical studies. Using this approach we are able to assign interfaces to the majority of known interactions between proteins involved in the clathrin-mediated endocytosis pathway in yeast. Analysis of this interface-interaction network provides novel insights into the functional specificity of protein interactions, and highlights elements of cooperativity and competition among the proteins. By identifying diverse multi-protein complexes, interface-interaction networks also provide a map for targeted drug development.

Abstract Introduction Results Discussion Methods

macromolecular complex analysis systems biology biochemistry physics proteomic databases protein interactions proteins protein structure macromolecular assemblies biology computational biology proteomics biophysics macromolecular structure analysis

2013

Interface-Resolved Network of Protein-Protein Interactions

Zinc is an essential micronutrient for all living organisms and is involved in a plethora of processes including growth and development, and immunity. However, it is unknown if there is a common genetic and molecular basis underlying multiple facets of zinc function. Here we used natural variation in Arabidopsis thaliana to study the role of zinc in regulating growth. We identify allelic variation of the systemic immunity gene AZI1 as a key for determining root growth responses to low zinc conditions. We further demonstrate that this gene is important for modulating primary root length depending on the zinc and defence status. Finally, we show that the interaction of the immunity signal azelaic acid and zinc level to regulate root growth is conserved in rice. This work demonstrates that there is a common genetic and molecular basis for multiple zinc dependent processes and that nutrient cues can determine the balance of growth and immune responses in plants. Zinc (Zn) is an essential micronutrient for humans, animals, and plants [1]. It is of particular importance for the function of numerous metalloenzymes that are involved in a plethora of processes such as energy metabolism, nucleic acid and protein synthesis, and protein catabolism [2]. These key biological processes can be adversely altered in situations in which Zn availability is limited. Low Zn manifests itself at physiological and molecular levels, and can cause deleterious effects such as growth retardation and malfunction of immune responses. Recent studies in mammalian systems have shifted the focus on the role of Zn from simply a nutrient to a signalling molecule that fine-tunes intracellular signalling events (for an overview see [3]) and an important player in nutritional immunity [4]. In particular, in the role of Zn for host defence, a complex role is emerging. On one hand, Zn is actively depleted from infection sites restricting the ability of the pathogen to proliferate [5,6] and on the other hand high Zn levels are generated by the host that contribute to kill the pathogen [7]. However, despite its fundamental importance, it remains unclear whether there is a common molecular basis for these multiple functions involving Zn and whether the signalling and immune related functions of Zn are also relevant for plants. Plants are the first link in the chain of human Zn nutrition and therefore studying Zn related processes in plants is of particular relevance. Land plants acquire Zn at the root–soil interface and multiple processes are crucial for efficient Zn acquisition. While Zn transport is clearly very important for Zn acquisition and homeostasis [8], other processes also play pronounced roles for efficient growth under Zn limited conditions. For instance, when responses of two lines of rice that displayed a contrasting tolerance to -Zn were quantified, these lines didn’t differ in Zn-transporter activity but mostly in their maintenance of root growth and the exudation rates of organic acids [9]. A similar correlation of increased root growth and increased tolerance to –Zn conditions had also been described in wheat [10]. Overall, this indicated that Zn levels in the environment are perceived by the plant and lead to distinct changes in root growth that might be important for adaptive responses to low Zn conditions. Nevertheless, so far, neither a role for Zn as signal, nor the genetic and molecular bases of root growth changes upon -Zn conditions in plants has been clearly established yet. One important biotic stress response process that Zn and its availability has been shown to be involved in animals [11] as well as in plants is disease resistance [12]. One gene that is involved in the response to biotic and abiotic stresses in Arabidopsis thaliana is AZI1 (AZELAIC ACID INDUCED 1, At4g12470) [13,14]. It encodes for a lipid transfer protein (LTP) -like protein and belongs to the EARLY ARABIDOPSIS ALUMINIUM-INDUCED GENE1 (EARLI1) gene subfamily [15]. AZI1 was named after its unique response to the systemically active compound azelaic acid (AzA, a nine-carbon dicarboxylic acid) [16]. In the Arabidopsis genome, AZI1 clusters in a tandem array on chromosome 4 with three other EARLI1-type genes, namely At4g12480 (EARLI1), AT4G12490 (ELHYPRP2 (EARLI1-LIKE HYBRID PROLINE-RICH PROTEIN 2) and AT4G12500. Among these genes, the role of AZI1 for long-distance signals related to systemic acquired resistance (SAR) is the best documented so far. In addition to AZI1, SAR involves an important hormone, namely salicylic acid (SA). SA accumulates upon pathogen attack [17], and leads to the induction of PATHOGENESIS-RELATED GENE 1 (PR1, SA marker gene) [18]. In Arabidopsis thaliana, mutation of AZI1 causes a specific loss of systemic immunity triggered by pathogens [16]. Beyond its role in biotic stress responses, the involvement of AZI1 in response to abiotic stress, such as the regulation of seedling growth under salt stress, was demonstrated [14]. But, whether AZI1 is involved in the response to nutrient levels that potentially affect the plants capability to defend itself or the capability for pathogens for infection (e. g. low Zn), remains unexplored. Here, we study the genetic basis of low exogenous Zn levels on primary root length by exploring natural genetic variation. We find that there is heritable natural variation of root length responses to low Zn and that natural allelic variation of the immune gene AZI1 determines a significant proportion of this response. We further reveal an intriguing evolutionarily conserved interaction between exogenous Zn levels and AzA dependent defence pathways to regulate primary root elongation. To identify genetic components that regulate plant growth upon low Zn (-Zn) conditions, we determined the length of the primary root of 231 genetically diverse natural accessions of Arabidopsis thaliana (S1A Fig) from the RegMap population [19] grown on +Zn and -Zn medium over 7 days (S1 and S2 Tables). Importantly, while still being correlated, the primary root length of the vast majority of accessions clearly differed in +Zn and –Zn conditions (S1B and S1C Fig). To assess whether these root length responses were specifically due to the -Zn treatments, we determined mRNA levels of four Zn-deficiency responsive marker genes ZIP3, ZIP5, ZIP12 [20] and PHO1; H3 [21] in the Col-0 accession under our screening conditions. All of these four genes were significantly up-regulated in –Zn conditions (S2 Fig), demonstrating that the plants sensed and responded to the -Zn conditions. In the panel of screened accessions, we observed broad phenotypic variation for root length (S3 Fig) that was highly heritable (broad sense heritability (H2) ranging [22] from 0. 36 to 0. 44 for –Zn, and from 0. 42 to 0. 48 for +Zn) (S3 Table). Our analysis revealed that nearly 20% of the respective phenotypic variance is accounted for by genotype by environment (G X E) interactions for all days (S4 Table). We then conducted Genome Wide Association Studies (GWASs) using the AMM method that corrects for population structure confounding [23], to identify loci that were associated specifically with root length under -Zn (Fig 1A and 1B, S4 Fig). No association was observed under +Zn (S5 Fig). We then corrected the association P-values for all SNPs for multiple testing using the Benjamini-Hochberg-Yekutieli method [24]. Due to the limited power of our association study and the potentially high false negative rate due to population structure correction, and because we aimed to test any major emerging candidate experimentally, we selected a relatively non-conservative 10% false discovery rate (FDR) as our threshold for significant association (we thus expect that 10% of the significant SNPs are false positives). Using this criterion, we identified two chromosomal regions associated with root length in –Zn conditions. On chromosome 2, the significant peak (P-value = 3. 27*10−7; FDR ≈ 7%) was located in a region with a cluster of similar genes encoding a Cysteine/Histidine-rich C1 domain family (At2g21810). It was detected on the last day of the time course (day 7). These proteins require Zn ions for their function [25] (pfam, PF00130). On chromosome 4, the significant peak (P-value = 4. 40*10−7; FDR ≈ 6%) was located in in a region that contains the lipid transfer protein (LTP) -like AZI1 (At4g12470) gene and the 7 additional genes encoding for lipid transfer proteins as a cluster (Fig 1B and 1C). This peak was already detected early in the duration of our time course (day 2) suggesting that this locus was relevant for regulating early primary root length in response to limitations in external Zn rather than being a consequence of low internal Zn levels. As AZI1 itself was known for being involved in signalling: it mediates azelaic-acid-induced systemic immunity [16]; we hypothesized that AZI1 was involved in mediating crosstalk between nutrient and immunity signals. However, as the GWAS peak spanned multiple genes, we first tested whether the best candidate in this region was indeed AZI1. For this, we assessed all 8 genes in the genomic region surrounding the association peak. Of these 8 genes, only AZI1 showed significant transcript level alteration in response to -Zn in Col-0 (Fig 1D), suggesting that it was involved in –Zn dependent root growth regulation. To test this further, we determined the root lengths of Arabidopsis Col-0 (WT), two azi1 (T-DNA) mutant lines (azi1-1 and azi1-2) and an AZI1 overexpressing (OE AZI1) line (35S: : AZI1) grown in +Zn or -Zn conditions over 7 days (S6 and S7 Figs). In presence of Zn, no significant differences in root length could be observed between azi1 mutant lines and wild-type plants (day2, Fig 1E; 7 days, S6A Fig). Grown under -Zn, the root length of azi1 was significantly shorter than Col-0 and 35S: : AZI1 plants (day 2, Fig 1E; 7 days, S6B Fig). Starting from day 5 onwards, roots of azi1 mutant lines were still significantly shorter and roots of 35S: : AZI1 plants become significantly longer than Col-0 roots (S6 Fig, day 5; S7A and S7B Fig), which suggests that the expression level of AZI1 is involved in controlling this trait (root length) (S6C Fig). The effect of Zn limitation was only visible on primary roots length. Day 5 was therefore chosen as time point for further analysis. To assess whether this is a function of AZI1 that is common to other micronutrients, we grew the same set of lines under low iron (-Fe) conditions (S7C Fig). There, no significant root length difference could be detected for the four tested lines, which indicates that growth responses to comparable nutrient limitations are not dependent on AZI1 and supports the notion of a rather specific AZI1 dependent response to Zn. While it has been shown that AZI1 transcripts accumulate in the aboveground tissues [16], they can be detected in roots as well (S8 Fig, [26]). Therefore, in order to assess whether the effects observed on root length in –Zn are associated with a systemic role of AZI1 or its local expression in roots, we expressed AZI1 under the control of the promoter of the zinc transporter ZIP1, which is known to be predominantly expressed in roots specifically under Zn deficiency [27]. In line with these results [27], AZI1 transcript was only detected in the roots of Zn-deficient pZIP1: : AZI1 seedling (S9A Fig). Consistent with a local role of AZI1 in roots, 2 independent single-insertion pZIP1: : AZI1/azi1 lines displayed longer roots than Col-0 under –Zn conditions, while the roots of azi1 mutant lines were significantly shorter than Col-0 (S9B and S9C Fig). Taken together, these data show that AZI1, previously described as a key component of plant systemic immunity involved in priming defence [16,28], modulates primary root length in a Zn level dependent manner. While we had shown that AZI1 was involved in modulating primary root length in a Zn level dependent manner, this was not proof that the allelic variation of AZI1 is causal for the observed root length differences under –Zn. We therefore set out to test this. Sequence analysis of the AZI1 genomic region (promoter and coding region) showed multiple polymorphisms in the regulatory region (S5 Table) as well as synonymous changes in the coding region that were consistently different between contrasting groups of accessions with either long or short roots on –Zn (Fig 2A and 2B, S10 Fig, S1, S2 and S6 Tables). Consistent with causal regulatory polymorphisms, AZI1 expression was significantly higher upon –Zn in accessions with longer roots (Fig 2C, 2D and 2E). For further analysis, we focussed on two contrasting accessions, Col-0 and Sq-1, which were among the most contrasting accessions regarding their root length on –Zn (S1 and S2 Tables) and each displayed the variant of the marker SNP that was associated with long and short roots on –Zn respectively. To then experimentally test whether the difference in AZI1 expression level was due to the natural allelic variation and whether the allelic variation was also causal for the longer roots, we transformed the azi1 mutant (Col-0 background) with constructs containing 1. 6kbp of the promoter and the coding region from either Col-0 (long roots in -Zn) or Sq-1 (short roots in -Zn), and an empty vector (control). In five independent homozygous single insertion lines complemented with the Col-0 pAZI1: AZI1 the expression level of AZI1 under –Zn was significantly higher (P<0. 01) than that in plants transformed with the Sq-1 pAZI1: AZI1 construct (Fig 3A). Consequently, we tested these T3 lines for root length differences under –Zn and +Zn. Consistent with the hypothesis that our AZI1 variants determine root growth specifically under –Zn, no difference in term of root length was observed between the T3 lines grown on +Zn (S11 Fig), while under -Zn, we observed significantly longer roots (P< 0. 05) in the Col-0 pAZI1: AZI1 plants compared to Sq-1 pAZI1: AZI1 plants or azi1 plants transformed with the empty vector (Fig 3B). Taken together, these data demonstrate that allelic variation at the AZI1 locus can cause variation of AZI1 expression levels and at the same time leads to variation of primary root length under -Zn. Moreover, as there were multiple polymorphisms in the regulatory region of the AZI1 gene (Col-0 and Sq-1 accessions, S12A Fig), and only synonymous changes in the coding sequence in these constructs (S12B Fig), we can rule out that the observed effect is due to changes in the protein sequence of AZI1. Therefore, our data suggest that the differences caused by the two AZI1 alleles are due to regulatory elements or posttranscriptional regulation such as RNA stability. We note, that we cannot completely exclude the additional involvement of other genes in the associated region in contributing to this response. While AZI1 had not been implicated in any known process involving Zn, it is known to mediate signal mobilization for systemic defence priming that can be triggered by AzA [16,17]. We therefore hypothesized that AZI1 would modulate growth and immunity programs depending on Zn and AzA status. To test this hypothesis, we first established the effects of the exogenous application of AzA on root growth. AzA affected root growth in a dose-dependent manner starting with a relatively mild reduction of growth at 100 μM to complete inhibition of root growth at 200 μM AzA (Fig 4A). We then determined whether this response is dependent on AZI1, and assessed root growth in Col-0, the Sq-1 accession and azi1 mutant lines at 100 μM AzA and in presence or absence of Zn after 5 d of treatment. While, AzA severely inhibited root growth in Col-0 plants in presence of AzA and Zn (+AzA+Zn), the azi1 mutant lines and the Sq-1 plants were significantly more resistant to the inhibitory effect of AzA (Fig 4B). This demonstrated that AzA modulates root growth in an AZI1 dependent manner. To test whether Zn modulates this response, we conducted the same assays under –Zn conditions. Strikingly, low levels of Zn alleviated the growth inhibitory effect of AzA on Col-0 to a large extent, and led to a further increased growth of the azi1 mutant lines (Fig 4B). The primary root length of these mutant lines (azi1) grown in presence or absence of Zn and/or AzA was similar to those observed for Sq-1 genotype grown under same conditions (Fig 4B). Taken together, these data show that AzA induced reduction of root length is modulated by Zn levels, and that AZI1 is a key component for this modulation. To further test whether the interaction between Zn and AzA is specific to early developmental stages or if it is retained later in plant development, Arabidopsis Col-0 and azi1 mutant were grown for 10 days in +Zn condition, then transferred in +Zn, -Zn, +Zn+AzA, or -Zn+AzA conditions for 5 additional days. Also at this later stage, –Zn treatment leads to slightly increased primary root length of Col-0 while the root length of azi1 is decreased compared +Zn treatments (S13A and S13B Fig). However, AzA application has an inhibitory effect on root growth regardless of the presence or absence of Zn in the medium in both Col-0 and azi1 seedlings (S13A and S13B Fig). We therefore conclude that Arabidopsis plants could prioritize root growth over defence during early development in response to low Zn, which is in line with our initial GWAS analysis where the association of AZI1 to primary root length in -Zn was highly significant during early development (day2) of Arabidopsis. It is proposed that AZI1/AzA regulates one branch of SAR, and that the second branch is regulated by salicylic acid (SA) (for review, [29]). To determine whether the interaction of Zn limitation and AzA is specifically regulated by AZI1, or could involve SA-defence related genes, we assessed the effect of AzA treatment in presence or absence of Zn on primary root elongation in mutants for genes such as ISOCHORISMATE SYNTHASE 1 (ICS1) and CAM-BINDING PROTEIN 60-LIKE g (CBP60g) (for review, [30]). Our results showed that the primary root lengths of 5-day-old ics1 and cbp60g seedlings were similar to WT (Col-0) in presence or absence of Zn or AzA (S14 Fig). Therefore, we concluded that AZI1 plays a specific role in regulating root growth in response to Zn limitation and in combination with AzA. Our data had not only shown that Zn levels and AzA modulate root growth, but also that the root growth responses to these treatments strongly interact (Fig 4B). To test whether this interaction is due to the modulation of molecular responses to AzA by Zn levels, we measured the expression levels of 18 defence-related genes that had been shown a mild but significant expression change upon AzA treatment (P < 0. 05) in leaves of wild-type plants (Col-0) [16], as well as 2 additional genes regulating salicylic acid biosynthesis (WRKY28 and WRKY46) [31], [32,33], AZI1, and a marker gene frequently used as a reliable molecular marker for SA-dependent SAR (PR1) [18]. Our q-RTPCR based gene expression analysis showed that almost all (16) of these defence-related genes were upregulated in response to the application of AzA (Fig 5). Notably, the group of most strongly induced genes contained genes involved in salicylic acid (SA) biosynthesis, such as ISOCHORISMATE SYNTHASE 1, WRKY28, WRKY46, as well as the SA response marker PR1. AzA treatment of plants grown on –Zn medium (-Zn+AzA) resulted in the upregulation of only 9 of the 16 genes that were upregulated in +Zn/+AzA (Fig 5). Notably, the SA response marker gene PR1 was not among these. Furthermore, consistent with an effect of Zn levels on the expression of these genes, plants grown on low Zn showed a down-regulation of 6 of the 16 defence-related genes induced by AzA alone (Fig 5). Overall this suggests that interaction of –Zn and AzA is not due to a general lack of induction of SA biosynthesis genes, but rather acts more downstream during SA signalling. In roots of the Sq-1 accession (short root under Zn limitation) expression patterns provided a more complex picture (S7 Table). Here, the expression of AZI1 was downregulated in response to -Zn-AzA and no significant changes were recorded for its expression in response to -Zn+AzA or +Zn+AzA treatments compared to control (+Zn-AzA). Expression of the ICS1, WRKY28 and PR1 genes showed an increase in response to -Zn-AzA, but no significant changes in response to -Zn+AzA compared to control (+Zn-AzA). Prompted by this difference to the Col-0 genotype, we measured whether the expression of Zn deficiency responsive genes (ZIP3, ZIP5, ZIP12 and PHO1; 3) in response to low Zn is altered upon AzA treatment in the Sq-1 accession. While expression of these four genes was induced by low Zn treatment (consistent with Sq-1 sensing the -Zn conditions), -Zn + AzA treatment significantly reduced the expression level of the four genes compared to low Zn treatment alone, AzA supply in Zn sufficient conditions significantly induced two of them (ZIP3, ZIP5) (S7 Table). Overall, these results demonstrate the presence of complex low Zn and AzA signal interaction in plants, and that Zn status impacts the expression of defence-related genes and modulates the response to AzA in plants, and that this is subject to natural variation. The link of Zn status and AZI1 to both, growth and defence prompted us to test whether interaction of AZI1 and Zn status impacts growth responses to biotic stimuli. For this, we chose to measure primary root responses upon infection with the bacterium Azospirillum brasilense (Sp245) as it had been described that root growth as well as AZI1 expression is changed in A. thaliana in presence of A. brasilense [34]. Consistently with [34] root growth of 5 d old Col-0 or Sq-1 plants was reduced when grown on complete medium (+Zn) and inoculated with A. brasilense (Fig 6). When grown on –Zn, this A. brasilense induced root length reduction was slightly less pronounced (Fig 6). The bacterially induced growth inhibition under +Zn was significantly less pronounced in the azi1 mutant lines, showing that AZI1 is involved in regulating this growth response. Under –Zn, bacterial incubation resulted in further decrease of root growth of azi1 mutant lines compared to +Zn (Fig 6). Overall, these data clearly show that there is a complex interaction of Zn levels and that the AZI1 gene determines the balance of growth and defence. The gene family to which AZI1 belongs, is strongly conserved throughout the Viridiplantae (green plants) [35]. This led us to hypothesize that the AzA mediated root growth regulation and its modulation by Zn status is a conserved growth-immunity regulating pathway. We therefore investigated the effect of AzA on root growth in presence and absence of Zn in the monocot species rice (Oryza sativa) (Fig 7A and 7B). Root growth increases in rice grown in low Zn compared to +Zn condition (Fig 7A and 7B). Interestingly, while seedlings grown in medium that contained AzA (300 μM) and Zn didn’t develop any roots, this growth inhibition was not observed when germinating the grains on –Zn medium containing AzA (Fig 7, S15 Fig). This demonstrated that AzA mediated growth inhibition, as well as its regulation by Zn levels is an evolutionary conserved mechanism. Plants must sense changes in external and internal mineral nutrient concentrations and adjust growth to match resource availability [36,37]. Responses to nutrient limitations manifest very early in plants life cycle, and root related processes are a major target of responses to nutrient constraints in plants. Consequently, primary root growth responds early on and drastically to nutrient limitations [38] and genome wide association mapping approaches are now being used to understand the genetic and molecular factors that govern these early growth responses [39]. However, apart from these abiotic variables (e. g. nutrients and water levels), the root is continuously exposed to changing biotic factors. Our study to identify genes and their variants that determine early root growth responses to low Zn levels revealed a genetic and molecular link between root growth response to abiotic (Zn) and biotic (defence related signalling) factors. In particular, our study revealed a significant association between the AZI1 locus and primary root length in Arabidopsis accessions grown in –Zn condition, as well as an intriguing AZI1 dependent interaction between Zn levels and the AzA pathway. One interesting question is whether the AZI1 function that relates to abiotic factors is specific to Zn or not. While there has been one report in which it was shown that ectopic expression of AZI1 improved Col-0 seed germination under high salinity stress condition and azi1 mutants were overly sensitive [14], many leads point in the direction of a largely Zn specific function: Significant associations close to the AZI1 locus were neither found in the GWAS for accessions that were grown at the same time on +Zn (S5 Fig), nor in other published root GWAS datasets for growth on MS medium [40], under different nitrogen conditions [41], under NaCl stress [42], and under Fe deficiency [39]. Moreover, while loss of function of AZI1 caused decreased root growth of Arabidopsis plants grown under –Zn conditions, and AZI1 overexpression caused the roots to be longer than WT (Fig 1), this was not observed when azi1 or 35S: AZI1 were grown under -Fe conditions (S7C Fig). Nevertheless, the level of exclusiveness or the extent of this specificity of AZI1 function for Zn can only be elucidated by further studies. The responses that we have observed and studied occur early-on during plant development, when Zn levels in the seedlings will not be significantly depleted. Moreover, it is likely that traces of Zn will have remained in the washed agar that we used for these assays, generating an environment very low in Zn, rather than a fully Zn depleted growth environment. Overall, our findings thus exposed mechanisms that will relate more to sensing than to a Zn starvation/depletion response, consistent with our observation that these low Zn levels promote early root growth rather than to inhibit it. While we have identified a signalling mechanism that contributes to this –Zn dependent increase in root growth, the Zn sensor still remains elusive. Our study proposes a possible mechanism for the regulation of root growth depending on the environmental Zn level, in which AZI1 plays an important role, and probably in an interaction with SA. AZI1- and SA-related signals are known to interact [17], and possibly involved in a self-amplifying feedback loop (for review [43,44]). Our gene expression analysis revealed that Zn status impacts the expression of AZI1, as well as other immune- and SA-related genes in Arabidopsis (Fig 5). Variation in SA concentration in leaves and roots of Arabidopsis plants upon nutrient deficiency stresses has been reported [45], and appear to be nutrient specific [45]. For example, SA levels significantly increased in response to potassium deficiency while low Fe caused a significant decrease of SA in roots [45]. The involvement of phytohormones in controlling root growth under different nutritional status has been documented [46], and hormone accumulation “thresholds” appear to be critical for hormones action [47]. For example, in various plants species, treatment with low concentrations of SA led to root growth promotion while treatment with high concentration of SA caused an inhibition of root growth [43,44]. In our study, -Zn treatment is associated with longer primary roots of wild-type plants (Col-0), together with no significant accumulation of SA as revealed by the absence of induction of a marker gene frequently used for SA accumulation, namely PR1 [18]. Interestingly, in contrast to azi1, low SA-accumulating mutants, ics1 and cbp60g, did show longer roots in –Zn similar to wild-type plants (Col-0). This suggests that the elongation of primary root in –Zn is a result of an active AZI1 and likely low SA level. The increase of SA [17], AZI1 expression level and AzA [34] was reported in plants exposed to bacterial infection presumably to promote defence response, often at the cost of growth. In our gene expression analysis PR1 was induced by +Zn+AzA conditions, suggesting an accumulation of SA under these conditions. Consistently, root growth of wild-type plants (Col-0) treated with (+Zn+AzA) was severely reduced compared to control condition (+Zn-AzA). Similarly, plants (Col-0) treated with Sp245 showed shorter root than the ones grown on control condition. The negative effect of either AzA or Sp245 treatment on root growth was alleviated when these stresses were combined with low Zn treatment, (-Zn+AzA or -Zn+Sp245). This was accompanied by an absence of PR1 induction, indicating that the otherwise root growth inhibitory SA response is affected by low Zn levels at the molecular level. Our work exposed an interaction of Zn levels and immunity and a common genetic and molecular basis for this. Interestingly it seems that Arabidopsis can prioritize root growth over defence responses, if Zn levels are low during early development. Prioritizing root growth over defence in –Zn in early developmental stages could be a way to explore soil for available Zn. The data from rice point in the same direction and provide a hint that this interaction is evolutionary conserved. It will be interesting to elucidate how this relates to the molecular role of Zn in nutritional immunity in plants, perhaps somehow similar to the role of Zn levels in infection sites in animal systems [5–7]. AZI1 belongs to a large family of pathogenesis-related proteins; LTPs. While some LTPs were suggested to play a role in defence reaction in a root specific manner [48], the role for AZI1 in defence responses was mainly investigated in aboveground tissues, and similar data for role of AZI1 in roots are still absent. Our data demonstrate that the expression of AZI1 predominantly in roots is sufficient to control the primary root length in response to Zn availability. Much like AZI1 function, the role of AzA had been also analysed only in the aboveground tissues. Our data reveal an important role of this pathway in the root, which extends the current models for underground defence priming. Importantly, the conserved specific interaction of Zn and AzA that can be observed not only in the dicot species Arabidopsis but also in the monocot species rice, is not only interesting from the view of basic biology, but also harbours very interesting perspectives for an innovative biotechnological application. This is because AzA is thought to directly mediate crop plant responses to pathogens and herbivores or to mimic compounds that do [49] and is listed among the natural compounds that induce resistance by a priming mechanism [50]. To activate these plant responses, AzA like others organic and inorganic chemicals can be applied as a foliar spray, seed treatment, or soil drench [49]. However, our work revealed that the soil-based application of AzA might significantly impact root traits depending on Zn bioavailability (AzA severely inhibited the root growth in +Zn), which could have an enormous and direct impact on plant growth in the field. We used 231 different genotypes of A. thaliana from different geographic origins (S1 and S2 Tables, S1A Fig) and for each genotype grew 12 seedlings. The previously described [51] azi1 insertion mutant lines (SALK_017709 (N517709) and SALK_085727C (N657248) available in the Nottingham Arabidopsis Stock Centre. SALK_085727C was provided by Peter Urwin (University of Leeds, UK). The ics1 (SALK_133146C) and cbp60g (SALK_023199C) mutants were used in this work. Lines overexpressing AZI1 under the 35S promoter, under the ZIP1 promoter, expressing the Col-0 or Sq-1 AZI1 locus under their native promoters were generated in this azi1 mutant background. Plant phenotyping for GWAS was as described previously [40]. Briefly, for each growth condition all lines were grown side by side. For both Zn conditions, GWAS assays were performed in the same growth chambers under the same 22°C long-day conditions (16 h light, 8 h dark). Seeds were placed for 1 h in opened 1. 5 ml Eppendorf tubes in a sealed box containing chlorine gas generated from 130 ml of 10% sodium hypochlorite and 3. 5 ml of 37% hydrochloric acid. For stratification, seeds were imbibed in water and stratified in the dark at 4 °C for 3 days to promote uniform germination. On each plate, eight different accessions with three seeds per accession were then germinated and grown in a vertical position on agar-solidified medium contained 0. 5 mM KNO3,1 mM MgSO4,1 mM KH2PO4,0. 25 mM Ca (NO3) 2, l00 μM NaFeEDTA, 30 μM H3BO3, l0 μM MnCl2, l μM CuCl2,15 μM ZnSO4,0. 1 μM (NH4) 6Mo7O24, and 50 μM KCl, in presence of 1% (wt/vol) sucrose and 0. 8% (wt/vol) agar. -Zn or -Fe medium was made by not adding the only source of Zn (ZnSO4) or FeEDTA to the medium, respectively. For the assays involving azelaic acid treatments, AzA (246379 ALDRICH, Sigma) was added in different concentration ranging from 25 to 200 μM. For the assay with rice (Oryza sativa L.), Niponbare was used and seeds were soaked in deionized water over night in dark then transferred in a controlled-environment chamber (light/dark cycle of 14/10 h, 200 μmol photons m-2s-1, temperature of 28/25 °C and RH of 80%) to ¼ Yoshida media for 5 d [52,53]: 0. 36 mM NH4NO3; 0. 41 mM MgSO4; 0. 19 mM CaCl2; 0. 13 mM K2SO4; 0. 08 mM NaH2PO4; 4. 72 μM H3BO3; 2. 37 μM MnCl2; 8. 90 μM Fe-NaEDTA; 0. 62 μM ZnSO4; 0. 04 μM CuSO4; 0. 02 μM (NH4) 6Mo7O24, adjust to pH 5. 5. ZnSO4 was removed in -Zn medium. For rice treatments, azelaic acid (246379 ALDRICH, Sigma) was added in different concentration ranging from 75 to 300 μM. For the control condition, rice plants were kept in nutrient solution with the above-mentioned composition. Rice seedlings were grown in a growth chamber under the following environmental conditions: light/dark cycle of 14/10 h, temperature of 28/25 °C, and RH of 80%. GWAS root trait quantification was conducted using the BRAT software [40]. Root length in assays involving mutant and transgenic lines, as well as rice plants was measured using ImageJ software, version 2. 0. 0 (http: //rsb. info. nih. gov/ij/). Statistical differences between genotypes were calculated using t-test analyses and ANOVA with subsequent post hoc tests using Graphpad Prism (GraphPad Software Inc. , San Diego, CA, USA) or Microsoft Excel (Microsoft, USA). Total RNA was extracted from roots of Arabidopsis wild type plants (different accessions) grown in presence or absence of Zn, and with or without azelaic acid (246379 ALDRICH, Sigma), using Plant RNeasy extraction kit (Qiagen) and RQ1 RNAse-free DNAse (Promega). Two μg of total RNA were used to synthesize cDNA using poly-A oligos. Real-time quantitative reverse-transcription PCR (RT-qPCR) was performed with a Light Cycler 480 Real-Time PCR System (Roche; Roche Diagnostics, Basel, Switzerland) using LightCycler 480 SYBR Green I Master mix (Roche, IN, USA). Primer list is provided in S8 Table. Primers were designed in conserved regions between tested accessions. Gene transcript accumulation quantification were performed in a final volume of 20 μL containing optimal primer concentration 0,3 μmol, 10 μL of the SYBR Green I master mix, and 5 μL of a 1: 25 cDNA dilution. Real time-PCR conditions were as 95°C for 5 min, and followed by 40 cycles of 95°C for 10 s, 60°C for 10 s, 72 °C for 25 s, and finally one cycle 72 °C for 5 min. As a negative control, template cDNA was replaced by water. All PCR reactions were performed in triplicates. For each sample, a cycle threshold (Ct) value was calculated from the amplification curves. For each gene, the relative amount of calculated mRNA was normalized to the level of the control gene ubiquitin10 mRNA (UBQ10: At4g05320). For every sample, the relative gene expression of each genes was expressed following normalization against the CT values obtained for the gene used for standardization, for instance ΔCT, AZI1 = CT, AZI1 − (CT, UBQ10). Quantification of the relative transcript levels was performed as described previously [54–57]. Briefly, -Zn treatment was compared to +Zn, the relative expression of a each gene was expressed as a ΔΔCt value calculated as follows: ΔΔCt = ΔCT, AZI1 (-Zn) − ΔCT, AZI1 (+Zn). The fold change in relative gene expression was determined as 2−ΔΔCT. Wild-type strain of Azospirillum brasilense is used in this study [58]. These bacteria strains were cultivated and inoculated in plant culture medium as described previously [59]. Statistical differences between genotypes were calculated using t-test analyses and ANOVA with subsequent post hoc tests using Graphpad Prism (GraphPad Software Inc. , San Diego, CA, USA). The AZI1 locus from Col-0 and Sq-1 accessions were cloned with primers spanning the region ranging from 1614 bp upstream of the AZI1 transcription start site to the stop codon of AZI1 into the binary vector pCAMBIA1301 by restriction enzymes of BamHI and PstI using primers listed in S8 Table. AZI1 transcription start site to the stop codon of were cloned in fusion to 1129 bp ZIP1 promoter using primers listed in S8 Table. The constructs were transformed into Agrobacterium tumefaciens strain GV3101 and then used for Arabidopsis transformation by the floral dip method [60]. Transgenic plants were selected by antibiotic resistance, and only homozygous descendants of heterozygous T2 plants segregating 1: 3 for antibiotic sensitivity: resistance were used for analysis. For GWAS, mean total root length values of 231 natural accessions were used (S1 and S2 Tables). The GWA analysis was performed in the GWAPP web interface using the mixed model algorithm (AMM) that accounts for population structure [23] and using the SNP data from the RegMap panel [61,62] [19]. Only SNPs with minor allele counts greater or equal to 10 were taken into account. The significance of SNP associations was determined at 10% FDR threshold computed by the Benjamini-Hochberg-Yekutieli method to correct for multiple testing [24]. Using a Multi-trait mixed model ([63]) we decomposed the variance of the root length response under plus/minus Zn Conditions for each day into a genetic term, the G X E interaction and an environmental term. The respective variance parameters were estimated in a null model without single marker effects and a global Kinship estimated from all markers. The analysis has been performed in R (R Development Core Team 2008). The respective R scripts are available at https: //github. com/arthurkorte/MTMM.

Plants have evolved mechanisms to cope with complex environments in which resources as well as potential threats are fluctuating. Thereby, plants modulate their growth based on multiple cues from the environment. In this study, by exploring natural genetic variation in Arabidopsis to study the role of zinc in regulating primary root length, we find a major locus governing this is the AZELAIC ACID INDUCED (AZI1) locus, previously known to be involved in systemic acquired resistance. We then showed that regulatory variation at AZI1 contributes significantly to this natural variation. Importantly, the known AZI1 function led us to show that there is an interaction of zinc deficiency and the defence pathway. While the studies of the roles of the defence signal AzA and AZI1 had been restricted to the aboveground tissues, we clearly showed an important role of this pathway in the root, which is zinc-dependant. Our observations regarding the interaction of zinc and AzA-dependent defence pathways on root growth are not a fluke of evolution, but they are evolutionary conserved between dicots and monocots. Taken together, these results will serve as a basis to design new strategies for improvement agricultural crop species able to modulate growth and defence.

Abstract Introduction Results Discussion Materials and methods

genome-wide association studies plant growth and development brassica plant physiology developmental biology plant science model organisms experimental organism systems genome analysis seedlings plants arabidopsis thaliana research and analysis methods genomics zinc gene expression chemistry genetic loci chemical elements plant defenses eukaryota plant and algal models root growth genetics biology and life sciences physical sciences computational biology organisms human genetics

2018

Natural allelic variation of the AZI1 gene controls root growth under zinc-limiting condition

Streptococcal toxic shock syndrome (STSS) is a severe invasive infection characterized by the sudden onset of shock and multiorgan failure; it has a high mortality rate. Although a number of studies have attempted to determine the crucial factors behind the onset of STSS, the responsible genes in group A Streptococcus have not been clarified. We previously reported that mutations of csrS/csrR genes, a two-component negative regulator system for multiple virulence genes of Streptococcus pyogenes, are found among the isolates from STSS patients. In the present study, mutations of another negative regulator, rgg, were also found in clinical isolates of STSS patients. The rgg mutants from STSS clinical isolates enhanced lethality and impaired various organs in the mouse models, similar to the csrS mutants, and precluded their being killed by human neutrophils, mainly due to an overproduction of SLO. When we assessed the mutation frequency of csrS, csrR, and rgg genes among S. pyogenes isolates from STSS (164 isolates) and non-invasive infections (59 isolates), 57. 3% of the STSS isolates had mutations of one or more genes among three genes, while isolates from patients with non-invasive disease had significantly fewer mutations in these genes (1. 7%). The results of the present study suggest that mutations in the negative regulators csrS/csrR and rgg of S. pyogenes are crucial factors in the pathogenesis of STSS, as they lead to the overproduction of multiple virulence factors. Streptococcus pyogenes (group A Streptococcus; GAS) is one of the most common human pathogens. It causes a wide variety of infections, ranging from uncomplicated pharyngitis and skin infections to severe and even life-threatening manifestations, such as necrotizing fasciitis (NF) and bacteremia. Several streptococcal virulence factors, including pyrogenic exotoxins, streptokinase, and streptolysins, are reportedly involved in these diseases. Streptococcal toxic shock syndrome (STSS) is a severe invasive infection that has been recently characterized by the sudden onset of shock and multiorgan failure; it has a high mortality rate, ranging from 30% to 70% [1]. There is controversy as to whether the cause of STSS largely depends on host factors or bacterial factors. Although many studies have sought to determine the crucial factors behind the onset of STSS, the responsible GAS genes have not been clarified. Recently, we and others have reported that mutations in the csrS (covS) gene—a sensor gene of a two-component regulatory system—were detected in a panel of clinical isolates from severe invasive streptococcal infections, but not in non-STSS isolates [2]–[4]. Mutations in the gene caused an increased expression of various virulence genes; the upregulation of streptolysin O (SLO) induced necrosis of neutrophils and prompted the escape of csrS mutated strains from being killed by neutrophils, resulting in increased virulence in lethality in the mouse model [2]. Complementation of the wild csrS gene into csrS-mutated STSS isolates dramatically decreased their virulence in lethality [2]. Similarly, csrR (covR) mutations were found in the clinical isolates of STSS patients [5]. Such results suggest that csrS/csrR mutations are closely associated with the onset of STSS. However, several study groups that investigated the csrS/csrR gene sequence in each STSS isolate [3], [4], [6], [7] also report that there is no mutation in the csrS/csrR gene of STSS isolates [4]. These results raise questions as to how frequently STSS isolates have mutations in the csrS/csrR genes in a mass of isolates, and what mutations other than csrS/csrR genes may be responsible for the onset of STSS. In this study, we sequenced the csrS/csrR genes of 164 GAS strains that have been isolated from STSS patients in Japan since 1992. Almost one-half of the STSS isolates had a mutation in the csrS/csrR genes. In addition, we found a mutation in the rgg (ropB) gene, instead of the csrS/csrR genes, in the clinical isolates of patients with STSS. Since the rgg gene negatively regulates various virulence genes in a manner similar to that of the csrS gene, a mutation of the rgg gene in STSS clinical isolates increased the expression of several virulence genes and virulence in lethality in the mouse model. Such mutations were detected at a high frequency in more than 50% of STSS isolates. These findings suggest that mutations in the negative regulators such as csrS/csrR and rgg of S. pyogenes bring about overproduction of a number of virulence factors, such as SLO, and play a crucial role in the onset of STSS. In our previous study, we reported that there were various types of mutations in the csrS gene of emm49 clinical isolates from STSS patients [2] and in the csrR gene in emm3 clinical isolates from STSS patients [5]. These findings strongly suggest that csrS/csrR mutations play important roles in the pathogenesis of STSS. To evaluate the frequency of these csrS/csrR mutations in isolates from clinical cases of STSS [8], we sequenced the csrS and csrR genes in STSS clinical isolates from sterile sites (164 isolates) and non-STSS clinical isolates from non-sterile sites (59 isolates). The diagnoses, sites of bacteria isolation, and characteristics of S. pyogenes isolates are described in Table 1. Of the 164 STSS clinical S. pyogenes isolates, 55 isolates (csrS, 46 isolates; csrS + rgg, 9 isolates) (33. 5%) had mutations in the csrS gene, 19 isolates (csrR, 13 isolates; csrR + rgg, 6 isolates) (11. 6%) had mutations in the csrR gene, and 2 isolates (1. 2%) had mutations in both genes (Tables 1 and 2). The csrS and csrR genes of these isolates had deletions, point mutations, or insertions that created non-functional CsrS and CsrR products, as shown previously [2], [4], [5]. Therefore, 76 isolates (46. 3%) had mutations in the csrS and/or csrR genes, while the remaining 88 STSS isolates (53. 7%) had mutations in neither csrS nor csrR (Tables 1 and 2). On the other hand, non-STSS GAS isolates had a significantly lower number of mutations in the genes [csrS mutation, 1. 69% (1/59); csrR mutation, 0% (0/59); total, 1. 69% (1/59); p = 0. 00000000062 by χ2 analysis]. Although csrS/csrR mutations were more common among STSS isolates examined than among non-STSS isolates, they were not present in all STSS isolates. This may suggest that mutations in other regulatory genes may also be found among STSS isolates. To identify novel bacterial factors that may contribute to the pathogenesis of STSS, we next investigated the expression pattern of virulence factors in S. pyogenes isolates. We determined the sequence of the csrS/csrR genes from a panel of emm-matched GAS isolated from STSS patients; NIH1 (also called SSI-1), NIH3, NIH8, NIH34, NIH152-3, NIH249, NIH327, and NIH352 were clinical isolates of the emm3 genotype from STSS and C500, OT22, and K33 were emm3 non-STSS isolates (Tables 1 and S1). A mutation in the csrS gene was found in NIH152-3, NIH249, NIH327, and NIH352 of the STSS isolates; however, the other STSS and non-STSS GAS isolates had mutations in neither the csrS nor the csrR gene (data not shown). To determine whether other emm3 STSS strains have possible mutations in their genomes, we used comparative genome sequencing (CGS) [9], a microarray hybridization-based method developed to search for single-nucleotide polymorphisms (SNPs) and insertion–deletion sites within a genome between emm3 STSS and non-STSS isolates. We found several genes with SNPs in the NIH1 genome in comparison with that of non-invasive isolates K33. Three (codY, csrR and rgg) of them had non-synonymous amino acid change in NIH1 but not in K33 and C500 (Table S2). We further sequenced these 3 genes in other non-invasive isolate, OT22 and STSS isolates, NIH3, NIH8 and NIH34. A couple of genetic differences which affect amino acid sequence were detected between the STSS and non-STSS GAS isolates (Table 3). All four STSS isolates (NIH1, NIH3, NIH8, and NIH34) had some difference in SPs1742 (Rgg) but not in non-STSS isolates (C500, OT22, and K33) (Table 3). SPs1742 is identified as the rgg gene, a transcriptional regulator of multiple genes [10]–[13], although the role of the rgg gene is controversial [14]. We [2] and others [4] have previously reported that STSS emm49 and emm1 clinical isolates exhibit a higher expression of the SLO gene (slo) than non-STSS isolates, due to a mutation in the csrS gene. Therefore, we examined whether a panel of emm3-genotyped STSS isolates possessing mutations in the csrS or rgg gene and emm3 non-STSS isolates lacking mutations could produce SLO (i. e. , secretion of SLO in the culture supernatant). The comparison of the supernatants at the same growth condition (data not shown) showed that larger amounts of SLO were secreted by STSS isolates possessing mutations in the csrS gene (NIH152, NIH249, NIH327, and NIH352) or rgg gene (NIH1, NIH3, NIH8, and NIH34) than by non-STSS isolates (C500, OT22, and K33) (Figure 1). These data suggest that rgg mutations may be related to an increased expression of SLO, as observed in the case of csrS mutations. To clarify the role of rgg gene mutation in STSS isolates in terms of SLO production, we created the rgg mutants K33rgg and OT22rgg, non-STSS isolates into which an rgg mutation had been introduced. They exhibited increased SLO secretion, as observe with STSS isolates (Figure 1). In contrast, when an intact rgg gene was integrated into the genomic DNA of the STSS isolates NIH8 and NIH34 (NIH8rgg+ and NIH34rgg+), the SLO secretion was decreased to the level of that in non-STSS isolates (Figure 1). Taken together, it appears that the mutation of the rgg gene was responsible for increased SLO production in the culture supernatant as that of csrS gene was. It has been reported that Rgg influences the transcription of many virulence-associated genes in S. pyogenes [10]–[13]. To test the possibility that the transcriptional expression levels of virulence genes are regulated by the function of the rgg gene, we performed real-time polymerase chain reaction (RT-PCR) with specific primers for each virulence-associated gene. The amounts of mRNA of protein G-related alpha2-macroglobulin-binding protein (grab), nicotine adenine dinucleotide glycohydrolase (nga), streptodornase (phage-associated) (sdn), streptokinase (ska), and slo in the STSS isolate of NIH34 with the rgg mutation were larger than those of the pharyngitis isolate of K33 with the intact rgg gene (Figure 2). On the other hand, the amounts of mRNA of the cystein protease (speB) and streptolysin S (sagA) genes in the STSS isolate of NIH34 were less than a half of those in the non-STSS isolate of K33 (Figure 2). The amounts of mRNA of the IgG-degrading protease of GAS, Mac-1-like protein (mac), C5a peptidase (scpA), IL-8 protease (scpC), superantigen (speA), and DNA gyrase (gyrA) genes in NIH34 were almost the same as those in K33 (Figure 2 and data not shown). NIH34rgg+ suppressed the expression of virulence-associated genes to the levels found in non-STSS isolates; further, the expression of speB and sagA genes was increased to levels observed in non-STSS isolates (Figure 2). Additionally, the expression pattern of the virulence genes in K33rgg was similar to that in the STSS isolate NIH34 (Figure 2). These findings suggest that the transcriptional expression of multiple virulence genes, including the slo gene in GAS, was strongly influenced by the mutation in the rgg gene. To elucidate the role of rgg in infections in vivo, we used GAS intraperitoneal injections to compare the lethality and histopathology of NIH34 with that of the K33 strain in a mouse model. The NIH34 strain showed significantly higher lethality than the K33 strain (p = 0. 00027) (Figure 3A). Introduction of the rgg mutation in the K33 strain (K33rgg) resulted in higher lethality among infected mice than the K33 strain (p = 0. 00067) and exerted a level of lethality similar to NIH34. The NIH34 strain into which an intact rgg gene (NIH34rgg+) had been introduced exhibited less lethality than the NIH34 strain (p = 0. 0000097) and possessed the same level of lethality as the K33 strain. We confirmed that bacteria isolated from kidney or liver of infected mice at day 6 retained the mutation (data not shown). Therefore, the mutation of the rgg gene in the STSS isolates contributed to enhanced lethality in the mouse model. Histopathological examination of mice infected with NIH34 or K33rgg strains was carried out. Scattered multiple inflammatory foci containing bacterial colonies were observed in the kidney. The foci were accompanied with neutrophil infiltration, cell debris and hyalinization (Figure 3B). In contrast, no significant pathological change was observed in mice inoculated with the K33 or NIH34rgg+ strains (Figure 3B). In another mouse model of soft-tissue infections, subcutaneous infection with NIH34 or K33rgg resulted in significantly larger lesions as compared to the infection with NIH34rgg+ or K33 (p<0. 01) (Figure 3C). Bacteria were isolated from spleen and kidney after the subcutaneous infection of the rgg mutants but not the intact rgg strains. We confirmed that bacteria isolated from lesions retained the mutation (data not shown). This showed that subcutaneous inoculation of mice led to the systemic spreading in the rgg mutant. These results suggest that the rgg-mutated strains isolated from STSS patients are more virulent in vivo than strains from patients with non-invasive infections, and that the increase in virulence in vivo is canceled by introducing an intact rgg gene. In our previous study, using the Transwell system, we showed that SLO, which causes necrosis in neutrophils, and an IL-8 protease of ScpC are important for bacterial resistance to killing by neutrophils [2]. Here, we examined the effect of rgg mutation on resistance to killing by neutrophils. As shown in Figure 4A, the migration ability of human neutrophils in response to chemokine IL-8 did not significantly differ between K33 and K33rgg or between NIH34 and NIH34rgg+. Furthermore, the scpC mutation in the NIH34 strain did not have a significant influence on the migration of human neutrophils, compared to the csrS mutation, as previously reported (Figure 4A). This finding is in accordance with the less influence of ScpC expression in the rgg mutation (Figure 2). Collectively, the mutation of the rgg gene had little influence on the migration of human neutrophils in response to IL-8. As previously reported [2], migrated neutrophils may be killed by the STSS GAS isolates via enhanced SLO production, and therefore we examined this possibility. Human neutrophils were efficiently killed by the rgg-mutated strains (NIH34 and K33rgg), whereas strains with the intact rgg gene (K33 and NIH34rgg+) did not cause obvious impairment of neutrophils (Figure 4B). In the slo-deficient mutant, the ability to kill neutrophils was abolished. Nicotine adenine dinucleotide glycohydrolase (Nga) is a cytotoxic protein secreted through the SLO complex [15]. Based on the results that the nga expression was negatively regulated by the rgg gene (Figure 2), we examined the lethal activity of the nga mutant against neutrophils. The neutrophil-killing activity was significantly decreased in an nga-deficient mutant (NIH34nga), but to a lesser extent as compared to the activity of NIH34slo. Therefore, these findings strongly suggest that SLO is a factor essential for neutrophil-killing activity in rgg-mutated emm3 STSS isolates, and that Nga partially influences the neutrophil-killing activity. In our previous study, a csrS mutation in the emm49-genotyped strains was a key to the onset of severe invasive streptococcal infections [2]. The csrS mutant showed higher lethality in a mouse model and more efficiently killed human neutrophils than the non-mutated strain [2]. Therefore, we next compared the effect of the mutation in the csrS gene with that in the rgg gene, in terms of in vivo virulence in lethality and impairment of neutrophil function in vitro. Intraperitoneal infection of mice with the csrS mutant (K33csrS) caused earlier death and higher lethality than did infection with the rgg mutant (K33rgg) (p = 0. 017) (Figure 3A). Furthermore, K33csrS strains decreased the migration ability of neutrophils in response to IL-8, and they induced necrosis of migrated neutrophils to a greater degree than did the rgg mutants (Figures 4A, B). These and the aforementioned results suggest that the rgg mutant can escape being killed by neutrophils only because of the SLO function, and not because of ScpC, whereas both SLO and ScpC in the csrS mutant contribute to the escape. This suggests that the csrS mutant may be more virulent in systemic infections than the rgg mutant, owing to its ability to up-regulate more virulence factors such as ScpC (Figures 2 and 3A, B). In this study, we found that there are mutations in the rgg gene or the csrS/csrR genes in STSS clinical isolates. We sequenced the rgg gene in strains isolated from sterile sites of STSS patients and found that 42 of 164 (25. 6%) isolates carried some mutations (deletion, point mutation, or insertion) in the rgg gene. To determine whether these mutations contributed to a loss of Rgg function, we examined the level of SLO and SpeB [14] secretion and compared it with that in non-STSS isolates because overproduction of SLO [This study, 16] and less secretion of SpeB are also reported in the rgg mutation [14], [16]–[17]. We defined these phenotyped isolates as Rgg non-functional mutants. In 33 of 42 isolates, SLO production had increased and SpeB production had decreased (Tables 1 and 2 and data not shown). All of remaining nine rgg mutants (strains with mutation only in rgg) showing no increase of SLO expression were emm12-genotyped strains and had a mutation at the same position in comparison with other non-invasive strains. This mutation was synonymous in the level of amino acid, so we defined the mutants are functional as shown in Table 2. Collectively, 11. 0%, 28. 0%, 7. 9%, 1. 2%, 5. 5%, and 3. 7% of the 164 STSS clinical isolates carried non-functional mutations in the rgg, csrS, csrR, both csrS and csrR, both csrS and rgg, and both csrR and rgg genes, respectively, so that a total of 57. 3% of the STSS isolates carried mutations in one or more of these negative regulator genes (Tables 1 and 2). On the other hand, the frequency of mutations in these genes was very low (1. 7%) in non-invasive isolates (Tables 1 and 2). Therefore, the incidence of mutations in these genes is higher in STSS isolates than in non-invasive isolates (p<0. 01 by χ2 analysis). This finding suggests that mutations in the csrS/csrR genes or the rgg gene are crucial factors causing severe invasive infections, such as STSS. Since the late 1980s, STSS caused by S. pyogenes has become a serious health problem in both developed and developing countries. In this study, we found a high frequency of mutations of negative regulators in STSS clinical isolates. The rgg mutant killed human neutrophils, impaired multiple organs, and enhanced lethality in the mouse model, similar to the csrS mutant. These findings suggest that the impairment of negative regulators of S. pyogenes virulence genes induces neutrophil incompetence and subsequent STSS infection. This study is the first to show that clinical S. pyogenes isolates from STSS patients have mutations in one or more of genes––rgg, csrS, and csrR––which are involved in the negative regulation of multiple virulence genes. In our previous study, we found mutations in the csrS/csrR genes of 5 emm49 strains isolated from patients with severe invasive infections [2]. In the present study, we further examined whether STSS isolates other than the emm49 genotype possess mutations in the csrS and csrR genes: 46. 3% of the STSS isolates including various emm genotypes had non-functional mutations in one or more of the csrS/csrR genes. This finding suggests that mutations in the csrS/csrR genes are commonly recognized in STSS clinical isolates with various emm genotypes. We have shown that the amount of SLO protein produced in STSS isolates is greater than that in non-STSS isolates, and that this effect is due to mutations in both the rgg and csrS genes of the isolates. The loss of function incurred by the mutation in the rgg gene in emm3-genotyped S. pyogenes affected the regulatory network of the virulence-associated genes; hence, the mutated strains could resist killing by neutrophils and caused damage to various organs in the mouse models. Therefore, the mutated emm3-genotyped S. pyogenes strains may potentially cause severe infections such as STSS in humans. Hollands et al. [14] reported that a mutation of the rgg (ropB) gene reduces M1T1 group A streptococcal virulence. We examined the contribution of Rgg to the pathogenesis of systemic infections by using a clinical emm1-genotyped STSS isolate, NIH186, and an emm1-genotyped pharyngitis isolate, Se235. NIH186 and Se235rgg, both of which had a mutation in the rgg gene, showed higher lethality than NIH186rgg+ and Se235, in both of which the rgg gene is intact (data not shown). The rgg mutants impaired neutrophils to a greater extent than the rgg-intact strains did (Figure S1); this finding suggests that rgg mutants are more virulent than rgg-intact strains, in the emm1 genotype. Therefore, the discrepancy between the finding in this study and that of Hollands et al. [14] may be attributed to modified regulation of SLO expression in rgg-mutated isolate in the latter, but not downregulation of speB and sagA operons. Rgg is reported to regulate the transcription of many virulence-associated genes in S. pyogenes [10]–[13], and its regulatory profile varies among strains used [16]–[17]. Nevertheless, up-regulation of the slo, nga and ska genes and down-regulation of the speB gene are commonly found in the rgg mutation of emm3-genotyped isolates (Figure 2) and of M49 serotyped-strains, NZ131 and CS101 [16]–[17], suggesting they are the Rgg core regulon of GAS strains. In recent studies, it has been reported that expression of the rgg gene is positively regulated by CsrS [4], while it is negatively regulated by CsrR [16]. Expression of the slo gene is enhanced in the csrS mutant (Figure 2) [2], [4], but not in the csrR mutant [18]. In this study, the expression of the slo gene was enhanced in the rgg mutant (Figure 2), suggesting that the enhancement of the slo gene may serve as the same regulatory pathway as the effect of the csrS mutation. These findings suggest that CsrS affects the Rgg regulon as well as the CsrR regulon (Figure 5); in the csrS mutant, CsrR is not phosphorylated by CsrS, and Rgg expression is suppressed. It has been reported that the csrR null-mutation does not affect the expression of SLO [18]. However, Treviño et al. [19] reported that SLO production increases as a result of a csrR mutation in which histidine replaces arginine at position 119 of the CsrR protein; however the protein retained DNA-binding activity. The strains carrying such a kind of mutation are phenotypically identical to the csrS mutants [19]. Nine csrR mutants in this study showed increased SLO production (Tables 1 and 2), 2 (NIH136 and NIH300) of which had an amino acid replacement at position 119 of CsrR protein. Other 7 isolates showed mutation in the N-terminal amino acid of CsrR, but the exact mechanism of the CsrR mutant remains to be solved. The csrS/csrR and rgg genes negatively regulate various virulence genes; however, they regulate different virulence genes. The slo, nga, and ska genes are negatively regulated by both CsrS/R and Rgg. The grab gene is negatively regulated by Rgg, while the mac, scpA, and scpC genes are negatively regulated by CsrS [2] (Figure 2). Thus, in terms of impairing neutrophil function, the csrS mutant inhibits the migration of neutrophils due to the destruction of IL-8 by the increased expression of scpC (Figure 5) [2]; meanwhile, the rgg mutant does not significantly affect the expression of scpC. On the other hand, since both rgg and csrS genes negatively regulate the expression of slo, infections with these mutants result in damage of neutrophils due to the increased production of SLO in the foci. This may explain why neutrophils are observed histopathologically in some cases of severe invasive infection, but are not in others. Indeed, our mouse model shows that neutrophils clustered around the foci of bacteria in the kidney infected by the rgg mutant (Figure 3B) but not by the csrS mutant [2]. The slo, nga, and ska genes are negatively regulated by both CsrS and Rgg [2] (Figure 2). We previously reported that SLO is an important virulence factor for the necrosis of neutrophils, which leads to higher lethality of infected mice [2]. Nucleosidase (NADase), which is encoded by the nga gene, contributes to severe invasive infections by GAS in the murine model of infection [20]. Streptokinase, which is encoded by the ska gene, has an important role in GAS invasion and proliferation [21]. STSS isolates carrying mutations in the csrS gene and/or the rgg gene commonly increased the expression of these genes [2; this study]. Thus, overproduction of these factors in the mutants could cooperatively contribute to increased virulence, thus causing the onset of STSS. Notably, the mutation frequency of these genes in STSS isolates (57. 3%) was much higher than that in non-invasive isolates (1. 7%). These results suggest that mutations in the negative regulators of various virulence genes are important to the STSS onset. However, 42. 7% of the STSS isolates did not have mutations in the csrS/csrR or rgg genes. Such strains may have mutations in other various other two-component regulatory systems or regulators in the S. pyogenes genome [22], which would be the focus of our research. We could not exclude the possibility that clinical severity of infection by strains lacking any mutations in the three genes depends on host factors, and not on bacterial factors. Specific human leukocyte antigen class II haplotypes are associated with a risk of disease severity [23], and the importance of both host and environmental factors has been reported [24]. In the mouse model, the csrS mutant (K33csrS) showed higher lethality than the rgg mutant. However, in the present study, the mortality rate of STSS patients infected with the rgg mutant was 60. 9%, while that of patients infected with the csrS mutant was 47. 2% (data not shown). These findings suggest that the rgg mutant also causes high lethality in humans, which may indicate differences in disease severity between humans and mice. Streptokinase is highly specific for human plasminogen, exhibiting little or no activity to those of other animal species [25]. Human-specific pathogenic factor (s) may influence virulence in cases of infection with the rgg mutant. Collectively, we showed that mutations of negative regulators that result in the overproduction of multiple virulence factors are important to the onset of severe invasive infections such as STSS. Recently, it has been reported that community-associated methicillin-resistant Staphylococcus aureus (CA-MRSA) causes severe invasive infections, resulting in NF or even death [26], [27]. The enhanced virulence of CA-MRSA has been linked to an overproduction of leukolytic peptides, phenol-soluble modulins (PSMs) [28], [29]. The production of PSMs is regulated under the strict control of agr [29]. The change of expression of the agr regulator results in increased expression of virulence factors and increased virulence. Severe invasive infections are caused not only by S. pyogenes but also by other bacteria such as other Streptococcus, Staphylococcus aureus, Vibrio vulnificus, and Aeromonas spp. Such severe invasive infections may be caused by the coordinated overexpression of multiple virulence factors that are affected by the global regulatory network. This study complies with the guidelines of the declaration of Helsinki. This study protocol was approved by the institutional individual ethics committees for the use of human subjects (the National Institute of Infectious Diseases Ethic Review Board for Human Subjects) and the animal experiments (the National Institute of Infectious Diseases Animal Experiments Committee). Written informed consent was obtained from all study participants or their legal guardians for the patients who died. All clinical samples and healthy human neutrophils were stripped of personal identifiers not necessary for this study. All animal experiments were performed according to the Guide for animal experiments performed at National Institute of Infectious Diseases, Japan. The S. pyogenes strains and plasmids used in this study are described in Tables 1 and S1. The STSS criteria in this study are based on those proposed by the Working Group on Severe Streptococcal Infections [8]. The clinical isolates used were isolated from sterile sites of patients with STSS (164 isolates; age 0–99 years) and from non-sterile sites of patients with non-invasive infections (59 isolates; ages 1–67 years). The isolates from STSS and non-invasive infections were collected by the Working Group for Beta-hemolytic Streptococci in Japan, as previously reported [30]. Escherichia coli DH5α was used as a host for plasmid construction and was grown in a Luria-Bertani liquid medium with shaking or on agar plates at 37°C. S. pyogenes was cultured in Todd-Hewitt broth supplemented with 0. 5% yeast extract (THY medium) without agitation or on tryptic soy agar supplemented with 5% sheep blood. Cultures were grown at 37°C in a 5% CO2 atmosphere. When required, antibiotics were added to the medium at the following final concentrations: erythromycin, 300 µg/mL for E. coli and 1 µg/mL for S. pyogenes; and spectinomycin (Sp), 25 µg/mL for each of E. coli and S. pyogenes. The growth of S. pyogenes was turbidimetrically monitored at 600 nm, using a MiniPhoto 518R (Taitec, Tokyo, Japan). The nucleotide sequences of the csrS, csrR, and rgg genes were determined by automated sequencers, i. e. , an Applied Biosystems 3130xl Genetic Analyzer and an Applied Biosystems 3130 Genetic Analyzer (both Applied Biosystems, Tokyo, Japan). Sequencing data were deposited in the DNA Data Bank of Japan (DDBJ). Male five to six-week-old outbred ddY and hairless mice were purchased from SLC (Shizuoka, Japan) and maintained in specific pathogen-free (SPF) conditions. All animal experiments were performed according to the guidelines of the Ethics Review Committee of Animal Experiments of the National Institute of Infectious Diseases, Japan. The replacement of a mutated rgg gene by an intact rgg gene was performed by allelic recombination. Specifically, the chromosomal DNA derived from the GAS strains K33 (for emm3) and F482 (for emm1) was purified and used as a template for the PCR amplification of the intact rgg gene. The primers used were 5′-GGGGATCCTTATGGCTATATCATAGCTG-3′ (sense) and 5′-GGGAATTCTGTTGAGATAAACTACACC-3′ (antisense). The PCR fragment was ligated into the plasmid pSF152, and the resultant plasmids pSFrgg3+ (for emm3) and pSFrgg1+ (for emm1) were used for chromosomal integration into the mutated rgg gene of isolates from STSS patients, as described previously [31]. The integrated strains (Spr) were then selected by using spectinomycin (Sp) -containing agar plates. Integration of the intact rgg gene was confirmed by PCR. A total of 1 mL of the supernatant of an overnight bacterial culture (OD600 = 1. 0) was passed through a 0. 45-mm pore size membrane filter (Nippon Millipore, Tokyo, Japan), to remove the remaining cells. Proteins in the resulting cell-free supernatant were precipitated with 10% trichloroacetic acid and resuspended in a sample loading buffer containing a saturated Tris base. Samples were heated at 100°C for 3 min and separated on sodium dodecyl sulfate (SDS) –12. 5% polyacrylamide gels. To detect SLO, the proteins on the gels were electrophoretically transferred onto a PVDF membrane. The membrane was blocked with 5% nonfat milk +0. 2% Tween-20 and reacted with primary anti-SLO polyclonal antibody (American Research Products, Belmont, MA, USA), secondary antibody peroxidase-conjugated anti-rabbit Ig (GE Healthcare, Tokyo, Japan), and an ECL Plus Western blotting Detection System (GE Healthcare). Complete-genome comparisons were performed with an array-based service (CGS) provided by NimbleGen Systems Inc. (Madison, WI, USA) [9]. The reference genome sequence used in the microarray was that of S. pyogenes SSI-1 (GenBank accession No. BA000034). Total RNA was extracted from bacterial cells using the RNeasy Protect Bacteria Mini Kit (QIAGEN, Tokyo, Japan), according to the manufacturer' s instructions. Complementary DNA synthesis was performed with the PrimeScript RT reagent kit (Perfect Real Time) (Takara Bio, Otsu, Japan), also following the manufacturer' s instructions. Transcript levels were determined using the ABI PRISM Sequence Detection System 7000 (Applied Biosystems) and Premix Ex Taq (Perfect Real Time) (Takara). For real-time amplification, the template was equivalent to 5 ng of total RNA. Measurements were performed in triplicate; a reverse-transcription-negative blank of each sample and a no-template blank served as negative controls. The primers and probes used are listed in Table S4. GAS was grown to late-log phase (OD600 = 0. 6−0. 8) at 37°C in a 5% CO2 atmosphere, pelleted by centrifugation, washed twice with sterile phosphate-buffered saline (PBS), suspended in sterile PBS. A total of 1×107 CFU of GAS suspended in 0. 5 mL of PBS was injected intraperitoneally into five to six-week-old ddY outbred male mice (10–16 mice/GAS isolate). The number of surviving mice was compared statistically, using the Kaplan-Meier log-rank test. For the subcutaneous infection model, male hairless mice Hos: Hr-1 were injected with 1×107 CFU of GAS in a 100-µl suspension of GAS in PBS. The lesion area was measured daily and analyzed. Dissemination in kidney and spleen of GAS was evaluated by colony counting at day 7 post-infection. For histopathological analysis, the tissues from GAS-infected mice were directly fixed in 10% neutral-buffered formalin, embedded in paraffin, sectioned and stained with hematoxylin and eosin (H&E). Human neutrophils were isolated from the venous blood of five healthy volunteers, in accordance with a protocol approved by the Institutional Review Board for Human Subjects, National Institute of Infectious Diseases [2]. This study complies with the guidelines of the declaration of Helsinki. Chemotaxis assays were performed as previously described [2]. Briefly, 5×105 neutrophils in Roswell Park Memorial Institute (RPMI) medium containing 25 mM HEPES and 1% FCS in Transwell inserts (3-µm pore size; Coaster, Corning, NY, USA) were placed in 24-well plates containing 600 µl medium or 100 nM interleukin (IL) -8 solution (Pertec, London, UK); the plates were then incubated with or without 5×106 bacteria for 1 h at 37°C, in advance of the assay. After 1 h of incubation, cells in the lower wells were collected and 104 10-µm microsphere beads (Polysciences Inc. , Warrington, MA, USA) were added. Cells were stained with propidium iodine (Sigma, St Louis, MO, USA) for flow cytometry to quantify the viable neutrophils; analysis was performed, using the FACS Calibur (BD Biosciences, San Jose, CA, USA). The DNA Data Bank of Japan (DDBJ) (http: //www. ddbj. nig. ac. jp/index-e. html) accession numbers for the genes and gene products discussed in this paper are: TK283 csrR locus - AB517797; TK929 csrR locus - AB517804; NIH43 csrR locus - AB517807; NIH75 csrR locus - AB517814; NIH136 csrR locus - AB517819; NIH157 csrR locus - AB517822; NIH212 csrR locus - AB517826; NIH216 csrR locus -AB517827; NIH252-2 csrR locus - AB517838; NIH259 csrR locus - AB517839; NIH273 csrR locus - AB517842; NIH300 csrR locus - AB517850; NIH301 csrR locus - AB517851; NIH323-1 csrR locus - AB517853; NIH381-1 csrR locus - AB517863; NIH404 csrR locus - AB517867; NIH406 csrR locus - AB517868; NIH447 csrR locus - AB517877; NIH5 csrS locus - AB517796; TK76 csrS locus - AB517800; NIH18 csrS locus - AB517801; TK280 csrS locus - AB517803; NIH35 csrS locus - AB517805; NIH44 csrS locus - AB517809; NIH49 csrS locus - AB517810; NIH55 csrS locus - AB517812; NIH75 csrS locus - AB517815; NIH102 csrS locus - AB517817; NIH152-3 csrS locus - AB517820; NIH156-1 csrS locus - AB517821; NIH205 csrS locus - AB517823; NIH200-4 csrS locus - AB517825; NIH220-1 csrS locus - AB517828; NIH222 csrS locus – AB517829; NIH230 csrS locus - AB517830; NIH236 csrS locus - AB517831; NIH238 csrS locus - AB517833; NIH243 csrS locus - AB517834; NIH253-1 csrS locus - AB517835; NIH250-2 csrS locus - AB517836; NIH263-2 csrS locus - AB517840; NIH268 csrS locus - AB517841; NIH283-1 csrS locus - AB517844; NIH286 csrS locus - AB517845; NIH287-1 csrS locus - AB517846; NIH296 csrS locus - AB517847; NIH297 csrS locus - AB517849; NIH317 csrS locus - AB517852; NIH325-1 csrS locus - AB517854; NIH345 csrS locus - AB517855; NIH372 csrS locus - AB517859; NIH437 csrS locus - AB517862; NIH403 csrS locus - AB517866; NIH421 csrS locus - AB517871; NIH424-1 csrS locus - AB517873; NIH433 csrS locus - AB517874; NIH453 csrS locus - AB517875; Se202 csrS locus - AB517643; NIH3 rgg locus - AB517795; NIH8 rgg locus - AB517798; TK65 rgg locus - AB517799; NIH18 rgg locus - AB517802; TK1097 rgg locus - AB517806; NIH43 rgg locus - AB517808; NIH50 rgg locus - AB517811; NIH60 rgg locus - AB517813; NIH91 rgg locus - AB517816; NIH118 rgg locus - AB517818; NIH186 rgg locus - AB517824; NIH236 rgg locus - AB517832; NIH250. 2 rgg locus - AB517837; NIH273 rgg locus - AB517843; NIH293 rgg locus - AB517848; NIH357 rgg locus - AB517856; NIH366 rgg locus - AB517857; NIH371 rgg locus - AB517858; NIH372 rgg locus - AB517860; NIH374-2 rgg locus - AB517861; NIH381-1 rgg locus - AB517864; NIH390 rgg locus - AB517865; NIH406 rgg locus - AB517869; NIH409 rgg locus - AB517870; NIH422 rgg locus - AB517872; NIH445 rgg locus - AB517876.

Group A streptococcus (GAS) causes life-threatening severe invasive diseases, including necrotizing fasciitis and streptococcal toxic shock-like syndrome. Although many studies have attempted to determine factors that are crucial for the onset of streptococcal toxic shock syndrome (STSS), bacterial factors responsible for it have not been clarified. By comparing genome sequences of clinical GAS isolates from STSS with those of non-invasive infections, we showed that mutations of negative regulator genes (csrS, csrR, rgg) were detected at a high frequency of more than 50% in STSS isolates, but at a low frequency of less than 2% in non-invasive isolates. These mutations of negative regulators were found in various emm-genotyped STSS isolates but not in a particular emm genotype. These mutants enhanced virulence in mouse models. Such results indicated that mutations of bacterial negative regulators are crucial for the pathogenesis of STSS due to the overproduction of multiple virulence factors under the de-repressed conditions.

Abstract Introduction Results Discussion Methods

genetics and genomics/disease models genetics and genomics/genetics of disease microbiology/medical microbiology infectious diseases/bacterial infections

2010

Highly Frequent Mutations in Negative Regulators of Multiple Virulence Genes in Group A Streptococcal Toxic Shock Syndrome Isolates

Aberrant DNA methylation disrupts normal gene expression in cancer and broadly contributes to oncogenesis. We previously developed MethylMix, a model-based algorithmic approach to identify epigenetically regulated driver genes. MethylMix identifies genes where methylation likely executes a functional role by using transcriptomic data to select only methylation events that can be linked to changes in gene expression. However, given that proteins more closely link genotype to phenotype recent high-throughput proteomic data provides an opportunity to more accurately identify functionally relevant abnormal methylation events. Here we present a MethylMix analysis that refines nominations for epigenetic driver genes by leveraging quantitative high-throughput proteomic data to select only genes where DNA methylation is predictive of protein abundance. Applying our algorithm across three cancer cohorts we find that using protein abundance data narrows candidate nominations, where the effect of DNA methylation is often buffered at the protein level. Next, we find that MethylMix genes predictive of protein abundance are enriched for biological processes involved in cancer including functions involved in epithelial and mesenchymal transition. Moreover, our results are also enriched for tumor markers which are predictive of clinical features like tumor stage and we find clustering using MethylMix genes predictive of protein abundance captures cancer subtypes. Genomic characterization can elucidate underlying biology, disease etiology and reveal biomarkers of cancer development and progression; however, each molecular feature is susceptible to different sources of biological and technical measurement noise and provides only one view on the cell state. Therefore, comprehensive studies are needed to understand the molecular basis of disease. Toward this end a multi-institutional consortium, The Cancer Genome Atlas (TCGA), has extensively characterized numerous cancer sites producing genome wide data for mutations, copy number alterations (CNA), RNA expression, microRNA expression, and DNA methylation [1–5]. As part of this project, the proteome was probed using protein array Reverse Phase Protein Assay (RPPA) technology. However, antibody based analysis are inherently limited because of the reduced coverage and inability to easily compare across proteins due to differential binding effects [6,7]. Transcending these limitations, recent advancements in proteomics through high sensitivity mass-spectrometry (MS) are opening new opportunities in cancer research [8]. To accelerate the uptake of proteomics the Clinical Proteomic Tumor Analysis Consortium (CPTAC) is performing proteomic analyses of TCGA tumor bio-specimens for a growing number of tissue types and establishing standardized workflows using high-throughput liquid chromatography tandem mass-spectrometry (LC-MS/MS) to capture the proteome as a whole [6,9, 10]. To best leverage this new technology comparative analysis between protein abundance and RNA expression can highlight factors influencing concordance and inform how to best interpret proteomic data [11]. For example, multiple studies have proven that concordance between mRNA and protein is highly variable, such that one cannot be used to reliably predict the other. Correlation between mRNA and protein has been repeatedly shown to vary by tissue type and cancer status among other molecular features like biological function or molecular stability [7]. It was shown across multiple cancers that dynamic proteins involved in metabolism show strong agreement whereas housekeeping proteins and RNA processing proteins are weakly or negatively correlated [6,9, 10]. So, although many biological functions are regulated primarily through RNA expression—producing moderate correlation between proteomic and transcriptomic data, with mean spearman rho: 0. 23–0. 47 –post-transcriptional mechanisms also play a significant role that cannot be overlooked. The proteome represents the final link from genotype to molecular phenotype, so proteins are of special importance among molecular features and likely provide a more accurate depiction of cell state; this enhanced view on disease can be leveraged to assess functional effects of upstream aberrations, such as epigenetic modifications. Multi-level epigenetic features such as DNA methylation and histone modification work in concert to regulate gene expression. DNA-methylation, the covalent addition of methyl groups to CpG dinucleotides to form 5-methylcytosine (5mC), is catalyzed by DNA methyltransferases, and is influenced by both environmental and hereditary factors [12]. Previous studies have shown that DNA methylation plays a key role in health and is involved in processes of embryonic development and cellular differentiation, where changes can occur through imprinting, inheritance, or de novo events [13,14]. Furthermore, DNA methylation has been numerously cited as a potentially causative event in cancer [15,16]. Among potential DNA methylation drivers, silencing of tumor suppressors through promoter CpG island hypermethylation is best understood and linked to corresponding gene silencing [13,17,18]. Global hypo-methylation on the other hand can potentially result in genomic instability and reactivation of oncogenes [12,13,15]. To elucidate the role of DNA methylation in disease, our goal is to investigate whether linking proteomic data with DNA methylation data identifies key genes, describes molecular features and subtypes in cancer. Previously we presented MethylMix an algorithm that formalizes the identification of DNA methylation driver genes using a model-based approach [19–23]. Recognizing the complex role of the methylome in epigenetic regulation of cancer, MethylMix uses mRNA data to select only differentially methylated genes that show a downstream effect on gene expression (MethylMix-GE). This selects for likely functional aberrations with the aim of discriminating between true driver genes, and passenger events which are characteristic of genome wide dysfunction in cancer. Herein we present MethylMix-PA (Protein Abundance), a MethylMix analysis which refines candidate nominations for epigenetic driver genes by excluding aberrations that are buffered at the protein level; this likely selects for events which are functional over those which may accumulate during cancer but do not drive pathogenesis. Using proteomic data generated by MS technology from three cancer cohorts: breast invasive carcinoma, colorectal adenocarcinoma, and ovarian serous cystadenocarcinoma, we report MethylMix-PA genes, which include potential cancer progression markers and therapeutic targets. We describe MethylMix-PA’s ability to elucidate key molecular and higher level disease features and evaluate MethylMix-PA performance against MethylMix-GE. In summary, our study highlights the differences between integrated epigenomic-proteomics and epigenomic-transcriptomics analyses. For each cohort both models identify genes that are 1) differentially methylated when compared to normal adjacent tissue and 2) functionally predictive of downstream effects at the level of RNA expression in the case of MethylMix-GE or protein abundance in the case of MethylMix-PA (Fig 2). Among all three cancer cohorts we observe significant correlations between RNA expression and protein abundance (mean rho: 0. 23–0. 47), indicating that most genes are regulated at the transcript level (Fig 3, S2 Table). Therefore, it is unsurprising that MethylMix-PA shows high agreement with MethylMix, where more than 90% of MethylMix-PA genes are also identified by MethylMix. However, MethylMix-PA lists are more conservative identifying fewer candidate genes across all three cancers, where often the effect of methylation is present at the RNA level, but not detected at the protein level (Fig 2), likely because they are buffered at the protein due to post-transcriptional, translational, or degradation regulation. Therefore, MethylMix-PA better enriches for methylation-states that more likely execute functional roles in cancer development. For each cancer cohort using protein abundance data also identifies a few unique driver genes, the majority of which have documented roles in carcinogenesis. Explanative mechanisms by which the effect of DNA methylation may be undetected at the RNA level but functional at the protein level are further addressed below in the discussion. In breast cancer we discovered 19 novel differentially methylated genes of diverse biological functions. MethylMix-PA identified hyper-methylation of FSTL1, an autoantigen that promotes immune response. This candidate tumor suppressor, FSTL1, has also been shown to mediate tumor immune evasion in nasopharyngeal cancer through hyper-methylation silencing [24]. MethylMix-PA also found hyper-methylation of DHX40 which has an unclear link to cancer; although it is of note that RNA splicing proteins—like DHX40 –are highly stable, perhaps explaining the particularly stronger effect of DNA methylation on protein abundance than mRNA [25] (S1 Table). Next, MethylMix-PA identified hypo-methylation of CEACAM5 (also known as CEA), a cell surface glycoprotein that is used as a clinical biomarker for gastrointestinal cancers and may promote tumor development through its role as a cell adhesion molecule. High levels of CEACAM5 have been associated with operable early breast cancer [26,27]. Next, MethylMix-PA also identified hyper-methylation of FOXO1, a transcritionf factor where low expression has been associated with cancer [28]. In colorectal cancer the MethylMix-PA analysis uniquely recovers several genes associated with immune function and inflammation, which is known to play a key role in pathogenesis. We found that MethylMix-PA identifies a functional effect of UTR hypo-methylation of the PTPRC gene. PTPRC belongs to a family of protein tyrosine phosphatase which contains oncogenes regulating cell growth and differentiation. PTPRC is also related to tumor necrosis and disrupts normal T- and B-cell signaling through SRC kinase pathways—which are separately implicated in colorectal cancer through amplification [9,29]. Next, MethylMix-PA identified upregulation of S100A9 through promoter hypo-methylation. Of note, elevated S100A9 mRNA and protein levels are commonly observed in many conditions associated with inflammation [30]; additionally in hydropharangeal cancer where knockdown inhibited cell growth and invasion, S100A9 is also prognostic of worse outcome and indications like metastasis [31]. Of note MethylMix-PA filtered out functional effects of a UTR hypo-methylation in S100A9 previously detected by MethylMix-GE. Next, MethylMix-PA identified hyper-methylation across the promoter region of LTF, a likely tumor suppressor which is produced by neutrophils to regulate growth and differentiation. In the context of colorectal tissue LTF has been shown to restrict inflammation by regulating T cell interaction [32]. Additionally, gene expression of LTF has previously been shown to correlate with tumor size and survival in breast cancer [33]. MethylMix-PA picks up hypo-methylation states in 18 unique genes in ovarian cancer related to processes of invasion and proliferation. MethylMix-PA uniquely identifies hypo-methylation in the promoter region of EVL a key regulator of the actin cytoskeleton, associated with invasion and metastasis. Overexpression of EVL is also indicative of advanced stage in breast cancer [34] and has been implicated in malignancies due to inappropriate recombination [35]. Next, MethylMix-PA identified hypo-methylation of CTSZ, also known as cathepsin Z, a a lysosomal cysteine proteinase that has been shown to be involved in many primary tumors. For example, high levels of CTSZ promote epithelial to mesenchymal transition and are associated with the mesenchymal-like cell phenotype [36]. We also found hypermethylation of GSTM2, a gene that is normally high expressed in ovary, but has been shown to be a hypermethylated in lung cancers [37] and colorectal cancers [38], suggesting a tumor suppressor role for GSTM2 across tissues. Lastly, MethylMix-PA also identifies hypo-methylation in the mitochondrial genes SPG7, speculatively linked to cancer through metabolic function [39]. We conducted enrichment analysis to identify biological processes that are overrepresented in MethylMix-PA and MethylMix-GE genes (Table 2, S3 Table). Given the large proportion of common genes, across all three cancers both models capture many of the same annotations. However, comparing enrichments found for each cancer site, we find that broadly MethylMix-PA results include more significant enrichments for functions associated with cell adhesion and migration of epithelial and endothelial cells; these processes increase cell motility and invasiveness and are indicative of epithelial to mesenchymal transition (EMT) which is key to cancer development. Additionally, we observed that enrichment for immune functions are highly variable between each model’s results. Comparing unique annotations among breast cancer genes, MethylMix-PA includes enrichments for cell-cell adhesion, STAT signaling, response to interferon-gamma and immune cell functions, whereas MethylMix-GE similar pathways, but is also enriched for several other functions with less relevance to cancer such as homeostasis, muscle cell proliferation and skin development. In colorectal cancer, the MethylMix-PA gene list is shorter as fewer MethylMix-PA genes have been identified (Fig 2). These genes are only enriched in cell-cell adhesion (Table 2). The MethylMix-GE list for colorectal cancer is also enriched in cell-cell adhesion but also includes seemingly irrelevant enrichments for humoral immune response and detection of stimulus involved in sensory perception (S3 Table). For ovarian cancer, the MethylMix-PA enrichment mirrors the MethylMix-GE enrichment almost exactly with enrichments for metabolic processes, NF-kappa-beta signaling and interleukin-1 production. Taking an orthogonal approach, we identified putative biomarker of disease progression based on correlations between gene expression and clinical features (Table 3). Although MethylMix-PA gene lists contain much fewer identifications, we find that across all three cancers MethylMix-PA lists include a larger proportion of markers of tumor stage and size and show stronger odds of containing such genes (Table 3). The greatest difference in frequency of tumor stage marker is observed in breast cancer where 12% versus 7% of genes show correlation in MethylMix-PA and MethylMix-GE gene lists respectively. The most significant associations however are observed in colorectal cancer where 15% of MethylMix-PA genes show correlation between gene expression and tumor stage, this includes LTF which is mentioned among unique MethylMix-PA genes (Table 3A, S1 Table). The same trend applies when correlating gene expression with tumor size where the largest difference in enrichment can be seen in colorectal cancer where 7% versus 3% of genes correlate with size when comparing models. However, the enrichment is much stronger for breast cancer where 29% of genes correlate with tumor size compared to 21% of MethylMix-GE genes (Table 3B). Clustering on methylation has been shown to stratify patients into clinically relevant subgroups [2,20,21,23]. We performed consensus clustering using the DM values for MethylMix-PA and MethylMix-GE genes evaluating clusters sizes from two to six (Table 4); for clarity we discuss clusters at K = 2, examining the gross differences between MethylMix-GE and MethylMix-PA. We evaluated if these epigenetically defined subgroups correspond to previously published subtypes and clinical and genetic features and found that MethylMix-PA identifies subgroups of patients that enriched for specific cancer subtypes and other molecular features and performs similarly to MethylMix-GE (Fig 4, S4 Table). In breast cancer MethylMix-PA clusters significantly correlate with molecular subtypes and other molecular features such as Progesterone and Estrogen Receptor (PR, ER) status (Fig 4A). Similar to other studies our clusters differentiate between canonical breast cancer molecular subtypes: Cluster-1 and Cluster-2 containt the majority of patients with Luminal A/B type tumors while Cluster-3 contains the majority of patients with Basal-like tumors and as expected is enriched for samples negative for ER, PR, or HER2. HER2 and Normal subtypes are less clearly distinguished in MethylMix-PA clusters. Among colorectal samples we are able to confirm the CpG island methylator phenotype (CIMP) (Fig 4B). Cluster-2 contains all but one of the patients labeled CIMP-High using methylation signatures and 75% of patients labeled Microsatellite Instable/CIMP using gene-expression signatures. The CIMP subtype has known association with MLH1 silencing through hyper-methylation, which is reflected in our MethylMix-PA subtypes where we find cluster-1 to include the majority of samples with non-silenced MLH1. MethylMix-PA subtypes also significantly correlate with Microsatellite Instability where samples labeled as Microsatellite Instability-Low (MSI-L) or Microsatellite Stable (MSS) are found by majority in cluster-1. Examining subtypes in ovarian cancer our MethylMix-PA clusters agree well with molecular subtypes and are significantly correlated (Fig 4C). Cluster-1 contains 78% of Immunoreactive subtype and 78% of Differentiated subtype patients, while about half of cluster-2 is comprised of patients labeled as Proliferative. Lastly Mesenchymal subtype patients can be found with relatively equal frequencies in each cluster [40–42]. MethylMix-PA clusters also significantly correlate with tumor features, where cluster-2 and cluster-1 roughly correspond patients with lower-grade and higher-grade tumors. Epigenetic aberrations contribute to oncogenesis, where DNA hypermethylation inactivates tumor suppressor genes, while hypomethylation is known to promote genomic instability and activate oncogenes [12,20]. Therefore, DNA methylation has potential to inform patient treatment and improve patient outcomes through new diagnostics and therapeutics. When identifying epigenetically driven cancer genes, it is of note that most biological functions—subject to genomic and epigenomic dysregulation—are ultimately executed at the protein level, so we can expect neutralization of non-functional upstream effects at—or before—the proteome. Herein we confirm the potential of using proteomic data to elucidate functional DNA methylation events by conducting the first genome wide analysis of epigenome-proteome relationships across three large human cancer cohorts. We present MethylMix-PA, a method that formalizes the identification of abnormally methylated genes that are predictive of protein abundance, like MethylMix-GE, and uses a model-based approach, negating the use of arbitrary user-defined thresholds for abnormal DNA methylation, and identifies subpopulations of hypo or hypermethylated samples within a heterogeneous population. By integrating DNA methylation array and quantitative MS technologies, MethylMix-PA identifies candidate epigenetic driver genes with clinical value as potential therapeutic targets and protein biomarkers for assessing prognosis and treatment stratification. MethylMix-PA builds on our model MethylMix and addresses the potential limited predictive value of mRNA as proxy for phenotype due to the role of post-transcriptional mechanisms. MethylMix-PA identifies oncogenes and tumor suppressors and—with the exception of a few genes—returns a subset of MethylMix identifications, where often the effect of DNA Methylation does not propagate to the proteome (Fig 1, S1 Table). In other cancer studies similar buffering has been observed in both cis and trans CNA effects, suggesting that many detectable aberrations in cancer do not manifest in expression changes at the protein level [6,10]. Otherwise put, many abnormally methylated genes are likely only passengers and do not functionally contribute to cancer development. Identification of a reduced set of genes in our study has pragmatic benefits for cancer research, where narrowing nominations to fewer high-quality candidates increases the likelihood of finding true targets; strongest candidates include genes identified by both models that show negative correlation between DNA methylation and both gene expression and protein abundance, and therefore have clear biological interpretations amenable to validation in the laboratory. Similar methods to identify true targets have been described, where genes that show correlation between mRNA and protein are more likely to have tumor promoting effects [10]. Conversely, novel MethylMix-PA genes should be taken with due consideration given the lack of clear mechanisms explaining how changes in DNA methylation may alter protein levels, but be undetectable at the transcript level—plausible explanations that remain to be tested include erroneous or noisy gene expression data, low mRNA stability or alternative splicing confounding expression at the RNA level. Nevertheless, most new identifications are well supported to have tumor promoting effects and therefore warrant further investigation to uncover how DNA-methylation may influence regulation of genes like EHF, FSTL1, PTPRC, S100A9, LTF, EVL, and TSTA3. Importantly, in all these cases the type of DNA methylation is consistent with gene function, where known tumor-suppressors are hyper-methylated and oncogenes are hypo-methylated at regions where DNA methylation negatively regulates transcription. Taken together MethylMix-PA genes highlight important features in cancer related to tumor features and subtypes. MethylMix-PA genes also capture oncogenic biological processes based on enrichment analysis showing key aspects of cancer development such as processes related to EMT, immune function, and proliferative signaling (Table 2, S3 Table). MethylMix-PA also elucidates more shared annotations between cancer types, and thus a greater ability to identify genes of core cancer pathways that are shared across cancer sites. Next, using a completely orthogonal approach we also find that MethylMix-PA is more descriptive of tumor progression; although this new analysis produces a reduced number of identifications, MethylMix-PA genes are more likely to correlate in expression with disease features such as tumor stage and size (Table 3). Lastly, we find MethylMix-PA performs reasonably well for patient clustering recapitulating established molecular subtypes. Given the limitations of our study, we expect our clustering to have reduced discriminative power, since we limit our observations to genes for which we have both matched gene expression and protein abundance measurements in our analysis. This significantly diminishes the feature space we used for learning. Nevertheless, we find that MethylMix-PA performs similarly to MethylMix-GE in identifying cancer subtypes such as luminal and basal types of breast cancer, the CIMP type in colorectal cancer and all subtypes in ovarian cancer, with the exception of the mesenchymal subtype which is the least clearly defined subtype [40] (Fig 2, S4 Table). These findings suggest the reduced number MethylMix-PA genes capture the major sources of variation in each cancer cohort and facilitate translatability into feasible panels for testing. Overall MethylMix-PA shows practical utility for improving nominations of cancer driver genes and elucidating new mechanisms of cancer development missed by our previous model. More broadly our study supports using proteomic data to better understand how epigenetic deregulation promotes cancer. Similar approaches have been applied and found to potentially improve aspects of patient care. For example, a retrospective analysis of outcomes in an oncology trial for glioblastoma—which tested efficacy of different temozolomide regiments—found that updating the clustering model to incorporate MGMT protein expression and c-MET protein abundance provided better separation of overall survival prognostic groups than incorporating MGMT promoter methylation alone [43]. These findings and ours support the claim that protein data combined with DNA methylation is a better way to stratify patients and understand cancer features then using DNA methylation alone. Although milestone initiatives like TCGA and CPTAC provide valuable date for the acceleration of discovery and research in cancer, we acknowledge the limitations of this study and further work required. A barrier to translation, the number of specimens used here is insufficient to draw conclusive clinical correlations and require replication of these results by independent studies. Importantly molecular measurements used here are also subject to sources of technical and biological bias. For example, it is known that bulk measurements obscure the complex nature of tumor microenvironment which includes many cell types including vascular, lymphatic, and immune cells. This confounding effect is compounded considering that each molecular feature was measured using different tumor fragments, which may have very different cellular compositions due to intra-tumor heterogeneity. Additionally, we recognize further characterization of genome wide proteomic studies is required to fully understand possible biases, such as worse detection of highly hydrophobic and hydrophilic peptides, or low-abundance peptides co-eluting with very high-abundance peptide [9]. Moreover, early proteomic techniques such as those utilized in CPTAC’s Common Data Analysis Pipeline have not yet reached the genome level resolution of other omic measurements; these methods require refinement to address low coverage due to inherent limitations of proteolytic measurements such immeasurable peptides that are excessively large or small tryptic fragments and the inability to distinguish some amino acids [9]. This reduced coverage to a few thousand genes in our study excludes many genes with possible roles in cancer. The complex nature of disease development and interplay between interacting biological aberrations—genetic, epigenetic, somatic or germline—often makes it difficult to elucidate causal mechanisms of cancer development. Furthermore, there is still much work in multi-omics to elucidate causal flows of information influencing cellular physiology and pathology and to discriminate how separate phenomena are linked to create cancer [3,5, 42,44]. However, integrated multi-omic approaches like MethylMix-PA can provide additional insights into pathways and processes involved in oncogenesis and how they manifest as clinical phenotypes. As CPTAC moves into its second phase and characterizes more samples across more cancer types, models such as MethylMix-PA may leverage this valuable data to improve understanding of the molecular basis of cancer. All data used in this study is third party data, and is available from the following articles [6,9, 10,45–47]. All other data is available within the paper and Supporting Information files. Molecular data were produced from tissue bio-specimens from three cancer cohorts: breast invasive carcinoma (BRCA), colorectal adenocarcinoma (COADREAD), and ovarian serous cystadenocarcinoma (OV) (Table 1).

To elucidate the molecular basis of cancer we examine the variation and dynamics characterizing the flow of information from epigenome to the transcriptome and proteome. Conducting the first genome wide analysis of epigenome-proteome associations, we present a MethylMix analysis that leverages protein abundance data taking advantage of recent high-throughput proteomic data generated using mass-spectrometry technology to elucidate the role of DNA methylation in cancer. By integrating across molecular data types, we confirm the benefit of using protein abundance data to provide additional insights into pathways and processes involved in oncogenesis and how they manifest as clinical phenotypes. Applying our method across three large cancer cohorts including breast cancer, ovarian cancer and colorectal cancer, MethylMix identifies key genes and describes molecular features and subtypes in these cancers.

Abstract Introduction Results Discussion Methods

medicine and health sciences breast tumors cancer risk factors cancers and neoplasms protein abundance oncology epigenetics dna medical risk factors dna methylation chromatin research and analysis methods chromosome biology epidemiology gene expression breast cancer biological databases chromatin modification proteomics dna modification biochemistry genetic causes of cancer proteomic databases colorectal cancer cell biology nucleic acids database and informatics methods genetics biology and life sciences

2019

The impact of DNA methylation on the cancer proteome

The exosome complex functions in RNA metabolism and transcriptional gene silencing. Here, we report that mutations of two Arabidopsis genes encoding nuclear exosome components AtRRP6L1 and AtRRP6L2, cause de-repression of the main flowering repressor FLOWERING LOCUS C (FLC) and thus delay flowering in early-flowering Arabidopsis ecotypes. AtRRP6L mutations affect the expression of known FLC regulatory antisense (AS) RNAs AS I and II, and cause an increase in Histone3 K4 trimethylation (H3K4me3) at FLC. AtRRP6L1 and AtRRP6L2 function redundantly in regulation of FLC and also act independently of the exosome core complex. Moreover, we discovered a novel, long non-coding, non-polyadenylated antisense transcript (ASL, for Antisense Long) originating from the FLC locus in wild type plants. The AtRRP6L proteins function as the main regulators of ASL synthesis, as these mutants show little or no ASL transcript. Unlike ASI/II, ASL associates with H3K27me3 regions of FLC, suggesting that it could function in the maintenance of H3K27 trimethylation during vegetative growth. AtRRP6L mutations also affect H3K27me3 levels and nucleosome density at the FLC locus. Furthermore, AtRRP6L1 physically associates with the ASL transcript and directly interacts with the FLC locus. We propose that AtRRP6L proteins participate in the maintenance of H3K27me3 at FLC via regulating ASL. Furthermore, AtRRP6Ls might participate in multiple FLC silencing pathways by regulating diverse antisense RNAs derived from the FLC locus. The regulation of gene silencing occurs at multiple levels and non-coding RNAs (ncRNAs) have emerged as important regulators of genome silencing at the transcriptional and posttranscriptional levels [1]. At the chromatin structure level, chromatin remodeling factors and DNA- and histone-modifying enzymes alter chromatin to control its structure and compaction, thus affecting silencing [2]. ncRNAs significantly contribute to the regulation of chromatin structure and play important roles in eukaryotic genomes by affecting the epigenetic architecture, including both establishment and maintenance of epigenetic marks. ncRNAs can initiate heterochromatin formation through the RNA interference (RNAi) pathway, or independently of RNAi, and through the RNA processing machinery [3], [4]. The exosome is an evolutionarily conserved complex of RNase-like and RNA binding proteins involved in 3′ to 5′ decay and processing of various RNA substrates [5]–[10]. The exosome complex plays an important role in regulating both coding and ncRNAs [11]. The nuclear and cytoplasmic forms of the eukaryotic exosome complex share ten common subunits [12]. In most organisms, all nine subunits of the exosome core complex are inactive, and enzymatic activities are provided by the tenth catalytic RRP44 subunit, a 3′-5′ hydrolytic exoribinuclease, which also has endonucleolytic activity [13]. The nuclear form of eukaryotic exosome also associates with the second catalytic RRP6 subunit, a substoichiometric, nuclear-specific, 3′ to 5′ exoribonuclease [14]–[16]. The RRP6 subunit has a number of unique functions in the exosome [14], and also additional functions not associated with the exosome core [17]–[19]. Arabidopsis has three possible functional homologs of RRP6: the nuclear proteins AtRRP6L1 and AtRRP6L2, and the cytoplasmic protein AtRRP6L3 [20]. When we purified the exosome complex from Arabidopsis, AtRRP6L proteins were not detected in our experiments, which is likely due to the fact that, as a substoichiometric subunit restricted to a nuclear form, it was underrepresented in our preparations [8]. Therefore, whether these Arabidopsis RRP6Ls physically associate with the exosome core remains to be tested. The exosome complex broadly affects epigenetic silencing of heterochromatic and euchromatic loci by regulating a variety of ncRNAs [11], [21]–[24]. In fission yeast (Schizosaccharomyces pombe), the exosome acts in several different small RNA (smRNA) pathways to affect constitutive and facultative heterochromatin silencing, in either RNAi-dependent or RNAi-independent manners [4], [25]–[27]. The exosome also acts in gene silencing through RNA quantity and quality surveillance, and in collaboration with the 3′ termination machinery [21], [22], [28]–[30]. Our previous genome-wide survey revealed that many exosome targets in Arabidopsis correspond to ncRNAs, many originating from heterochromatic loci, suggesting that the exosome participates in various silencing pathways in Arabidopsis [8]. Our recent analysis of exosome functions in smRNA-mediated silencing of genes in Arabidopsis showed that the exosome has little effect on the smRNAs that function in the main silencing mechanisms, siRNA-dependent methylation of DNA (RdDM). Rather, we showed that the exosome associates physically with long, polyadenylated RNAs transcribed from the scaffold regions of several heterochromatic loci, and exosome defects affected the level of histone H3K9me2, an epigenetic mark that alters chromatin structure [31]. We also found that the Arabidopsis rrp6 homologues AtRRP6L1 and AtRRP6L2 participate in these epigenetic mechanisms and may function redundantly [31]. With the exception of AtCSL4, the genes encoding the Arabidopsis exosome core complex subunits are essential for viability [8]. By contrast, unlike the core exosome subunits, the exosome nuclear catalytic subunit RRP6 and the Arabidopsis RRP6-Like proteins are not essential for viability [14], [20]; thus, the rrp6 and rrp6l mutants provide tools to study the role of the exosome during development. Epigenetic regulation by long non-coding RNAs (lncRNAs) and histone modifications plays a key role in controlling the expression of Arabidopsis FLC (FLOWERING LOCUS C), which encodes a MADS-box transcription factor that suppresses flowering [32]–[35]. FLC regulation through chromatin modifications has been well studied [36], [37]. FLC expression requires H3K4 methylation, a permissive chromatin modification, on FLC chromatin, and H3K4 demethylation leads to FLC repression [37]–[39]. FLC silencing requires Polycomb complex 2 (PRC2) -deposited H3K27me3, a repressive chromatin modification [36], [40], [41]. Repressing the expression of FLC provides a central mechanism for both the vernalization pathway, which regulates flowering time in response to periods of prolonged cold, and the autonomous pathway, which regulates flowering independently of environmental signals [37]. The processing and metabolism of lncRNAs play a crucial role in FLC silencing, and different lncRNAs produced from the FLC locus may have distinct functions. For example, lncRNAs COOLAIR and COLDAIR participate in the epigenetic silencing of the FLC locus [40], [42]. The COOLAIR antisense transcript is produced from the FLC locus as two alternatively polyadenylated isoforms, AS I and AS II, and it was shown that the processing of AS I and II function in FLC epigenetic silencing by affecting H3K4 demethylation [37], [38], [42]. COOLAIR transcription does not appear to be required for vernalization, but it has been implicated in FLC repression early during cold treatment, possibly mediated by direct effects on the FLC promoter [37]. By contrast, the COLDAIR sense transcript is produced from within FLC intron 1, and plays a role in FLC silencing via recruitment of Polycomb repressing complex 2 (PRC2) to FLC during vernalization [40]. The establishment of H3K27 trimethylation to silence FLC in vernalization requires COLDAIR [40], but not AS I and AS II [41], [43]. Also, in the autonomous pathway, the proteins involved in 3′-end RNA processing act as the main factors in FLC silencing [38], [39], [44]–[46]. Here, we report that mutations of AtRRP6L1 and AtRRP6L2 result in delayed flowering in early-flowering Arabidopsis ecotypes that do not require vernalization for flowering. We found that AtRRP6L1 and AtRRP6L2 epigenetically regulate FLC silencing by regulating different antisense transcripts and modulating H3K4me3 and H3K27me3 histone modifications at the FLC locus. Moreover, we discovered a novel antisense transcript, termed Antisense Long (ASL), which originates from the FLC locus in wild type plants and is regulated by AtRRP6L1 and AtRRP6L2. Our study demonstrates that Arabidopsis RRP6L proteins play an important role in the regulation of genes expressed in specific developmental phases via participating in lncRNA-mediated epigenetic silencing. We previously examined T-DNA mutations in Arabidopsis RRP6L1 and RRP6L2 and found that these mutations lead to de-repression of known heterochromatic loci and that RRP6L1 and 2 likely function redundantly in this process [31]. The T-DNA insertion allele of AtRRP6L1 was isolated from the Wisconsin population of T-DNA mutants [31], in the Wasilevskaya (Ws) ecotype, and the T-DNA insertion allele of AtRRP6L2 comes from the SALK collection, in the Columbia (Col-0) ecotype [31]. We previously used RT-PCR analysis to demonstrate that rrp6l1-3 mutant is a null allele and rrp6l2-3 is nearly null [31]. To control for the ecotype, we examined the phenotypes of rrp6l1-3, rrp6l2-3, and rrp6l1-3 rrp6l2-3 double mutants (hereafter rrp6l1/2) and compared them to wild-type plants of Col-0 and Ws ecotypes. When we examined the phenotype of different rrp6l1 and rrp6l2 alleles, we found that these single mutants did not show significant phenotypic alterations, although they exhibited a mild delay in flowering, as measured by leaf number at bolting (Fig. 1A). By contrast, the late-flowering phenotype becomes pronounced in rrp6l1/2 double mutants grown under long day conditions (Fig. 1A) and becomes very severe in plants grown under short day conditions (Fig. 1B). These findings indicate that AtRRP6L1 and AtRRP6L2 likely have redundant functions in the pathways activating flowering. The flowering defects in plants grown under short and long day conditions suggest that AtRRP6L1 and AtRRP6L2 function in the regulation of flowering through a pathway different from the photoperiod pathway, which promotes flowering in response to day length. The vernalization and autonomous pathways promote flowering through repression of FLC expression. During vernalization, cold temperatures repress FLC expression, and FRIGIDA (FRI) activates FLC expression [35]. Ecotypes that lack functional alleles of FRI, such as the early-flowering accessions Col-0 and Ws ecotypes used in this study, do not require vernalization for flowering. Thus, our data suggest that AtRRP6L1and AtRRP6L2 could be involved in regulation of FLC expression through the autonomous flowering pathway. To examine the roles of AtRRP6L1and AtRRP6L2 in regulation of FLC expression, we examined whether mutations of AtRRP6Ls affect the expression of FLC. We observed no change in FLC expression in the rrp6l1-3 and rrp6l2-3 single mutants, consistent with the degree of the observed phenotypic alterations, and we confirmed this observation with additional AtRRP6L1 and AtRRP6L2 T-DNA alleles (Fig. S1B). By contrast, we observed an increase in the levels of FLC transcript in the rrp6l1/2 double mutant relative to wild-type plants (Fig. 1C). These results suggest that AtRRP6L1 and AtRRP6L2 affect FLC expression and function redundantly in this process. To make sure that this phenotype does not result from transgressive segregation, we constructed AtRRP6Ls mutants using different AtRRP6L alleles isolated from the same Col-0 background and confirmed that this mutant combination causes a similar delay in flowering and derepression of FLC (Fig. S1A). FLC acts as a dosage-dependent floral repressor [47]. The exosome complex and RRP6 subunits affect the processing and turnover of various RNAs, and thus regulate RNA quality and quantity [5], [8], [11]. To find out whether the FLC transcript, which increased in abundance in rrp6l1/2 mutants, corresponds to a functional transcript rather than nonfunctional byproduct, we examined the expression of the flowering genes SUPPRESSOR OF CONSTANS OVEREXPRESSION 1 (SOC1) and FLOWERING LOCUS T (FT), which act downstream of FLC. We found that the rrp6l1/2 mutants showed lower levels of both SOC1 and FT transcripts (Fig. 1C), indicating that the increased level of FLC transcript in rrp6l1/2 mutants corresponds to a functional FLC transcript, and the increase in FLC expression enhances the repression of the downstream genes. In addition to increased levels of FLC transcript, we also detected increased amounts of the unspliced FLC RNA in rrp6l1/2 mutants (Fig. 2B). The autonomous pathway constitutively represses FLC, independent of environmental inputs [35]. To find out whether AtRRP6L1 and AtRRP6L2 affect FLC expression directly or by regulating the expression of the upstream genes that silence FLC in the autonomous pathway, we examined the expression of upstream genes in the AtRRP6L1 and AtRRP6L2 mutants. We found that the rrp6l1/2 plants showed no changes in expression of the genes that act upstream of FLC (Fig. 1D). These data imply that the de-repression of FLC observed in AtRRP6L mutants results from a direct effect of AtRRP6Ls on the expression of FLC, not from regulation of the genes acting upstream of FLC. To test whether the effect of rrp6l1/2 on FLC expression requires the core exosome complex, we next examined AtCSL4-2 and AtRRP41, which encode core exosome complex subunits. We did not observe de-repression of FLC transcription in an AtRRP41 inducible RNAi line or in AtCSL4-2 T-DNA insertion mutants (Fig. 1E), suggesting that AtRRP6L1 and AtRRP6L2 likely function independently of the exosome in the regulation of FLC expression. Antisense transcripts regulate FLC silencing and antisense expression appears to independently intersect with both the vernalization and autonomous pathways to repress FLC expression [37]. During the vegetative phase, the FLC locus produces two alternatively spliced, polyadenylated regulatory antisense (AS) transcripts, AS I and AS II [37] (Fig. 2A). Targeted 3′ end processing of these antisense transcripts affects the recruitment of histone chromatin remodelers to the locus, which results in reduced FLC transcription [38], [39], [46]. Therefore, we investigated whether the rrp6l1/2 mutants showed changes in the ratio of 3′ end processing and polyadenylation of these antisense transcripts. We found that, compared to wild type plants, the rrp6l1 or rrp6l2 single mutants, and the rrp6l1/2 double mutants had lower levels of processed AS I and II transcripts (Fig. 2C and Fig. S1C); also, the rrp6l1/2 plants showed reduced levels of AS I and II, consistent with the stronger phenotype of the rrp6l1/2 mutants (Fig. 1A and S1C). Interestingly, the pattern of down-regulation of AS I and II transcripts in rrp6l1/2 mutants was similar to the pattern observed in the mutants of cleavage stimulation factors CstF64 and CstF77, components of the cleavage polyadenylation machinery required for the 3′-end processing of AS I and II transcripts [38]. RRP6 plays an important role in formation of the 3′ ends of many RNAs in yeast and humans [14], [48], [49], and in budding yeast, also participates in the regulation of antisense transcripts derived from the PHO84 locus [21], [50]. Therefore, it is possible that AtRRP6L1 and AtRRP6L2 proteins, along with CstF, could participate in the 3′ end processing of both antisense transcripts. To further investigate the relationship between CstF64 and AtRRP6Ls, we attempted to construct a triple rrp6l1-3 rrp6l2-3 cstf64-2 mutant; however, since cstf64-2 mutants are sterile, we could not obtain the triple homozygous mutant. Taken together, our data suggest that AtRRP6L proteins could negatively regulate FLC expression by affecting the expression of the regulatory antisense transcripts. Methylated H3K4 marks active chromatin states and the antisense transcripts synthesized from FLC may function in FLC silencing by recruiting chromatin remodeling factors that drive H3K4 demethylation [38], [39]. To investigate whether the decrease in the antisense transcripts in the rrp6l1/2 mutants leads to changes in histone modifications at the FLC locus, we used chromatin immunoprecipitation (ChIP) to analyze the levels of H3K4me3 at various regions of the FLC locus (Fig. 2A). We found that the rrp6l1/2 mutant had significantly increased levels of H3K4me3 along the entire length of FLC, compared with wild type (Fig. 2D). These data suggest that the decrease in the level of antisense transcripts in rrp6l1/2 mutants might lead to the decreased recruitment of chromatin remodeling factors required for H3K4 demethylation, thereby regulating the accessibility of the transcription machinery to the locus, and in turn leading to FLC de-silencing, similar to previously reported observations [38]. The expression of FLC sense and antisense transcripts together with the increased level of H3K4 trimethylation in rrp6l1/2 mutants suggested that AtRRP6s could participate in FLC transcriptional silencing by affecting the chromatin structure at the FLC locus. Therefore, we asked whether AtRRP6L proteins can directly interact with the FLC locus to participate in the silencing pathway. To address this question, we constructed transgenic rrp6l1-3 lines that were complemented by a wild type copy of AtRRP6L1 fused with the TAP-tag for affinity purification, AtRRP6L1-TAP (see Methods). We then used ChIP on these lines to examine the association of AtRRP6L1 protein with several regions of FLC (Fig. 2A). ChIP showed a modest enrichment of AtRRP6L1 protein in the regions corresponding to the 3′-UTR and intron 1 of FLC (Fig. 2E). The AtRRP6L1 binding region within intron 1 appears to be further downstream of the 3′-end of the AS I transcript (with respect to the direction of AS I and II transcription), implying that AtRRP6L1 may bind in this region to process a longer antisense precursor transcript. The AtRRP6L1 binding regions do not overlap with the regions reported to be bound by FPA and FCA proteins, RNA-binding 3′-end processing factors required for the processing of the AS I transcript [39], [46]. However, FPA binds to the region between exon 4 and 5 (Fig. 2A), which is also downstream of the 3′-end region of AS I [46]; thus, AtRRP6L1 and FPA have somewhat similar, but not identical, binding patterns (Fig. 2A and E). The association of AtRRP6L1 protein with a larger region of the FLC locus (downstream of the 3′end of AS I) also suggests that AtRRP6L proteins might participate in the processing of different types of antisense RNAs, in addition to the known antisense transcripts derived from the FLC locus. Alternatively, AtRRP6L could also participate in co-transcriptional regulation of antisense transcription by binding to nascent antisense transcripts. Also, the level of AtRRP6L1 enrichment at the FLC locus was relatively modest (Fig. 2E). Thus, we cannot rule out the possibility that AtRRP6L proteins could associate with the locus by binding other protein and RNA complexes that physically interact with the locus. When we were examining the pattern of known antisense transcripts produced from the FLC locus, we observed the presence of a different antisense transcript in wild type plants, but not in rrp6l1/2 plants. Therefore, we set out to investigate the features of this antisense transcript by tiling RT-PCR using a set of primers that cover the entire FLC locus (Fig. S2A). We found that the transcript is a novel antisense RNA of over 2000 nucleotides in length. The sequence of this antisense RNA, which we termed ASL (Antisense Long), corresponds to intron 1 and the 3′-UTR region of the sense FLC transcript (Fig. 3A). Interestingly, 5′ region of the ASL transcript overlaps with the 5′ region of the AS I and II transcripts. Sequencing of ASL revealed that it has two different isoforms, 2,236 nt and 2,536 nt (ASLa and b, respectively), produced by alternative splicing (Fig. 3A). Moreover, ASL spans intron 1, an important region for maintenance of FLC silencing [51]. Notably, ASL also overlaps with the COLDAIR lncRNA, which is transcribed within intron 1 in the sense direction during vernalization [40] (Fig. 3A). To determine whether the ASL RNA has a 5′ cap, we used Terminator 5′-Phosphate-Dependent Exonuclease (TPE), which degrades uncapped RNA. We found that TPE treatment did not affect ASL levels, indicating that ASL has a 5′ cap (Fig. 3B). Next, we examined the 3′ end of ASL by performing cDNA synthesis primed by either sequence specific primer or oligo-dT primers. To our surprise, we found that ASL is not polyadenylated, as we detected ASL only from the cDNA primed by specific primers, not by oligo-dT (Fig. 3C). In plants, the RNA polymerases RNA Pol IV and Pol V [52] participate in gene silencing through smRNA-mediated mechanisms [53]–[55]. To investigate which RNA polymerase synthesizes ASL, we treated plants with α-amanitin, an inhibitor of RNA Pol II, and then used RT-PCR to examine ASL levels. We did not detect ASL in plants treated with α-amanitin (Fig. 3D), implying that RNA Pol II synthesizes ASL. To confirm this, we also examined the presence of ASL in nrpd1 (Pol IV) and nrpe1 (Pol V) mutants; we detected ASL in these mutants, indicating that Pol IV and V do not affect ASL, although nrpd1 mutants showed a minor change in ASL levels (Fig. S2B). Taken together, our data indicate that ASL is capped, synthesized by RNA Pol II and non-polyadenylated, and also suggest that it is distinct from the known antisense transcripts originating from the FLC locus. The tiling RT-PCR analysis indicates that the same promoter produces ASL, AS I, and AS II. Thus, it is possible that the ASL transcript could function differently from AS I and II in the silencing of FLC. Indeed, different antisense RNAs transcribed from same promoter of the human pseudogene PTENpg1 have different functions in transcriptional and post-transcriptional silencing of the tumor suppressor gene PTEN [56]. Next, we investigated how the AtRRP6Ls participate in the regulation of ASL expression. We found that the level of ASL transcript decreased in rrp6l1-3 and rrp6l2-3 single mutants (Fig. 4A and Fig. S2C). Moreover, we detected little or no ASL transcript in rrp6l1/2 double mutants (Fig. 4A), indicating that AtRRP6L proteins function as the main factors regulating the levels of the ASL transcript. Consistent with AtRRP6L functions in regulation of ASL expression, we observed that the level of the ASL transcript recovered to wild type levels in rrp6l1-3 mutant complemented by the wild type copy of AtRRP6L1-TAP (Fig. 4B). These data suggest that both AtRRP6L1 and AtRRP6L2 directly regulate the expression of ASL and are the main factors in this process. RRP6 is a 3′ - 5′ exoribonuclease and RRP6 defects usually result in abnormal accumulation of various RNAs due to failure to degrade or process them [21], [49], [50], [57]. We next asked how AtRRP6Ls could regulate ASL levels. The rrp6l1/2 mutants showed nearly undetectably low levels of ASL. We reasoned that, if AtRRP6Ls regulate the stability of the ASL transcript, then we would observe a difference in ASL decay rate in AtRRP6L single mutants. Therefore, we used the rrp6l1-1 single mutant and conducted an α-amanitin chase to compare the rates of ASL transcript decay in rrp6l1-1 and wild type plants. We found that the rrp6l1-1 mutant and wild type had similar rates of ASL transcript decay (Fig. 4C), suggesting that AtRRP6L proteins do not directly participate in the degradation of the ASL transcript but rather affect its production or biogenesis. To find out whether AtRRP6L1 could play a direct role in the expression of ASL, we examined if AtRRP6L1 protein physically associates with the ASL transcript. To this end, we conducted RNA immunoprecipitation (RNA-IP) in wild type plants using antibodies against AtRRP6L1 protein (Fig. 5A). The RNA-IP showed that AtRRP6L1 protein physically associates with the ASL transcript (Fig. 5A). We also obtained identical results from RNA-IP in the rrp6l1-3 mutant complemented with AtRRP6L1-TAP (Fig. 5B). Together these results indicate that AtRRP6L1 protein likely participates directly in the regulation of ASL transcript levels. We then hypothesized that the ASL RNA may play a role distinct from that of AS I and II. Histone remodeling factors affect FLC silencing and several lncRNAs affect H3K4 demethylation and H3K27 trimethylation [41]. The sense lncRNA COLDAIR participates in recruiting PRC2 and is necessary for the establishment of H3K27 trimethylation during vernalization [40], but this does not require AS I and II [41], [43]. However, H3K4 demethylation in the autonomous pathway does require AS I and II [37]–[39]. COLDAIR may also contribute to the maintenance of H3K27me3 during vegetative growth. In addition, the H3K27me3-binding protein LHP1 functions in the maintenance of H3K27me3 during the vegetative phase in actively dividing cells, suggesting that H3K27me3 maintenance could be important for FLC silencing during the vegetative phase after H3K27me3 has been established [58]. To examine whether ASL has a role distinct from that of AS I and II, we examined the level of the repressive histone mark, H3K27me3 at the FLC locus in rrp6l1/2 mutants. To our surprise, we observed that the rrp6l1/2 mutants showed significantly decreased levels of H3K27me3 along the entire FLC locus (Fig. 5C). The rrp6l1-3 single mutant also showed a mild decrease in the level of H3K27me3 (Fig. S1D). This observation is consistent with the very mild phenotype and the decreased level of the ASL transcript observed in rrp6l1 single mutants (Fig. 1A, 4A and S2C). Together, our data indicate that the knock-out of both AtRRP6L1 and AtRRP6L2 affected the levels of both H3K4me3 and H3K27me3 in the FLC locus (Fig. 3B). This is in contrast to AS I and II, which affect only the levels of H3K4me3 at FLC, at least in vernalization pathway [37], [41], [43]. The level of H3K27me3 correlates with the nucleosomal density [59], [60]. To examine whether the reduction in the level of H3K27me3 in rrp6l/2 mutants affects nucleosome positioning, we performed Micrococcal Nuclease (MNase) -ChIP assays using anti-H3 antibodies. MNase degrades nucleosome-free regions, allowing the estimation of nucleosome density. We found a lower nucleosomal density at the FLC locus in rrp6l1/2 mutants (Fig. 5D). These data indicate that the reduction of H3K27me3 levels observed in the rrp6l1/2 mutants could also result in relaxation of the chromatin state, which then allows factors involved in FLC transcription to gain easier access to the locus. Taken together, our results suggest that the regulation of the ASL transcript by AtRRP6L proteins may contribute to the maintenance of H3K27me3 at the FLC locus, which in turn contributes to the compact chromatin structure of the locus. The rrp6l1/2 mutations lead to a decrease of H3K27me3 levels and also affect the nucleosome density at the FLC locus (Fig. 5C and D). Therefore, we asked whether the ASL transcript could be directly involved in H3K27 trimethylation, a role similar to that played by the COLDAIR lncRNA. To answer this question, we performed RNA-IP using antibodies against H3K27me3. We found that the ASL transcript physically associates with H3K27me3 regions (Fig. 5A). Taken together, our data suggest that the ASL transcript could function in the maintenance of H3K27 trimethylation during the vegetative phase. For most exosome complex subunits, mutations cause a lethal phenotype, which indicates that, for most organisms, development requires exosome-mediated regulation of diverse RNAs [8], [61]. Unlike exosome core subunits, RRP6 and AtRRP6-Like proteins are not essential for viability [14], [20]. Most studies of exosome-dependent and exosome-independent RRP6 functions in developmental processes have been performed in fission and budding yeast [18], [21], [24], [49], [57], [62]. In these systems, RRP6 participates in facultative gene silencing and also regulates the transition from mitosis to meiosis through RNA-mediated epigenetic mechanisms [24], [57]. Similar to these findings, we observed that AtRRP6L1 and AtRRP6L2 function in regulation of flowering time by repressing FLC expression. The defect in the AtRRP6L proteins results in mis-regulation of antisense RNA production from the FLC locus and affects the level of histone modification of the locus. AS I and II transcripts function in FLC silencing in both vernalization and autonomous flowering pathways [37], possibly by affecting the level of H3K4 demethylation [38], [39], although the exact mechanism remains unknown. A decrease in the levels of processed AS I and II in mutants of CstF64 and CstF77, proteins involved in 3′-end processing, leads to increased levels of H3K4 trimethylation and subsequent de-repression of FLC [38]. Thus, AtRRP6Ls may contribute to 3′-end processing of the antisense RNAs similarly to the CstFs; indeed, 3′-end processing of various RNAs is one of the well-known functions of the exosome complex. The exosome processes a number of structural RNAs including rRNA, snRNA and snoRNA via trimming the 3′-ends of their precursors [63], [64]. Furthermore, RRP6 acts together with the Nrd1-Nab3 termination complex in budding yeast in non-canonical 3′-end processing and termination of the antisense RNA derived from PHO84, as well as processing of several other mRNAs [49], [50]. Disruption of the Nrd1-exosome pathway leads to de-repression of reporter genes integrated into heterochromatic regions and results in alteration of chromatin structure at specific loci and heterochromatic regions [11], [21], [22]. The exosome function in processing of mRNA and antisense lncRNA is likely to be conserved in plants, suggesting that AtRRP6L proteins may participate in regulating synthesis of the antisense RNAs derived from the FLC locus and this regulation could be important to maintain a repressive chromatin state for silencing of FLC. Surprisingly, we found that the defect of AtRRP6Ls caused a reduction of the level of H3K27 trimethylation, which has not been reported in studies of AS I and II in the vernalization and autonomous flowering pathways [38], [43]. A different intronic sense lncRNA, COLDAIR, derived from intron 1 of FLC, physically associates with the PHD-PRC2 complex to establish H3K27 trimethylation during vernalization [40]. Moreover, the AS I and II transcripts are not required for PcG-mediated silencing via regulation of H3K27me3 trimethylation in the vernalization pathway [41], [43]. Together, these data suggest that other lncRNAs, not AS I and II, function in regulating H3K27 trimethylation. AtRRP6Ls might affect the level of H3K27me3 indirectly, by regulating H3K4 demethylation through regulating either AS I and II, or ASL transcripts, via a mechanism similar to the interplay between Trithorax and Polycomb groups, which antagonistically regulate the levels of H3K4me3 and H3K27me3 at the FLC locus during vernalization [40], [41]. Alternatively, AtRRP6Ls may regulate the level of H3K27me3 directly via an unknown mechanism that functions independently of the silencing pathway involving AS I and II. Indeed, ARABIDOPSIS TRITHORAX 1 (ATX1), an Arabidopsis homolog of Trithorax 1, dynamically regulates activation of FLC through trimethylation but not dimethylation of H3K4 and atx1 mutations led to the loss of H3K4me3 and gain of H3K27me2 during the vegetative phase, but did not affect H3K4me2 and H3K27me3 [65]. This indicates that the regulation of the levels of H3K4me3 and H3K27me3 at the FLC locus could be independent of each other, at least during the vegetative phase. Together, the previous reports and our data suggest that the decrease of H3K27me3 in rrp6l1/2 mutants could be caused by the decrease in ASL RNA expression, and AtRRP6Ls may participate in the respective pathways for FLC silencing through regulating the expression of AS I, AS II, and ASL RNAs. We identified ASL, a novel, long antisense RNA that is distinct from the previously-described antisense RNAs. We observed that the ASL transcript physically associates with H3K27me3, suggesting that it could play a role in H3K27 trimethylation and function differently than AS I and II RNAs. The PRC2 complex binds ncRNAs with high affinity but does not recognize specific sequences, while its binding affinity correlates with the length of the RNA [66], [67]. Thus, it is possible that ASL transcript could also participate in recruitment of the PRC2 complex to the FLC locus, which leads to maintenance of H3K27 trimethylation during vegetative growth. Taken together, the previous reports and our data suggest that the regulation of chromatin structure via various lncRNAs is a central mechanism in FLC silencing, and different lncRNAs may function in different chromatin modification pathways. The AtRRP6L proteins may play a role in silencing pathways by regulating antisense transcription. ASL has several features in common with AS I and AS II. First, ASL is transcribed by RNA Pol II; second, it is alternatively spliced, existing in 2 isoforms; third, its transcription is driven by the same promoter that drives AS I and AS II; fourth, the 5′ part of ASL overlaps with the 5′ region of AS I and II. However, in contrast to AS I and II, the ASL transcript is long (over 2,000 nucleotides long), is non-polyadenylated, and extends into intron 1 of FLC. These differences suggest that ASL may have functions distinct from the functions of AS I and II in FLC silencing, and the mechanism of FLC silencing could be more complicated than previously thought. RNA Pol II transcribes ASL. Various non-polyadenylated Pol II RNAs, such as snRNA, snoRNAs, and some mRNAs, are processed by the exosome, which is recruited by the Nrd1-Nab3-Sen1 termination complex in the noncanonical 3′ end-processing pathway [22], [49], [50], [63], [68]. Thus, the noncanonical 3′ end-processing pathway may also participate in processing of ASL, if this pathway is conserved in plants. However, Arabidopsis homologs of Nrd1, Nab3 and Sen1 have not yet been characterized. The rrp6l1-3 and rrp6l2-3 single mutants showed decreased levels of ASL, and ASL was almost undetectable in the rrp6l1/2 double mutant. Based on the results of the α-amanitin chase experiments, the rrp6l1-1 mutation does not affect the stability of ASL, suggesting that AtRRP6L1 and AtRRP6L2 could be the main regulators of ASL synthesis. This finding is very intriguing, since defects in the exosome and RRP6 usually lead to abnormal accumulation of various RNAs due to failures of RNA degradation or processing [21], [22], [49], [50], [57]. This result may be caused by an unknown function of AtRRP6L proteins, which participate in either the synthesis or biogenesis of ASL, rather than in its degradation. Indeed, we previously reported that the expression of a number of loci decreased in AtRRP4 and AtRRP41 inducible RNAi plants and the AtCSL4-2 mutant [8], and many of these loci are located within euchromatic regions as well as in regions harboring H3K27me3 (unpublished data). Therefore, the exosome and AtRRP6Ls may function in regulation of RNA synthesis, different from their conventional functions in RNA degradation. Similarly, inactivation of the human homologue of RRP6 leads to dramatically reduced levels of Xist ncRNA involved in X-chromosome inactivation, although it remains to be seen whether this effect is direct [69]. In our study, we found that AtRRP6L1 protein physically associates with the ASL transcript, suggesting that AtRRP6L1 plays a direct role in regulation of ASL levels, likely through ASL synthesis rather than degradation. In addition, recent work demonstrated that another 5′-3′ exoribonuclease, Xrn1, also directly contributes to RNA synthesis of several mRNAs in budding yeast, by physically associating with chromatin and contributing to transcription elongation [70]. Alternatively, the decrease in ASL levels in the AtRRP6L mutants may indicate that different RNA decay proteins participate in degradation of these RNAs. We found that the level of FLC transcript was unaffected in the exosome core subunit mutants, AtRRP4 and AtRRP41 inducible RNAi lines and AtCSL4-2 T-DNA mutants, suggesting that the function of AtRRP6L1 and AtRRP6L2 in regulation of FLC expression could be independent of the exosome complex. This could also indicate that the exosome core complex is not necessary for metabolism of RNAs produced from the FLC locus and different RNA decay factors could participate in their degradation. For example, in yeast, the XRN family of 5′ to 3′ exoribonucleases works in both the nucleus and cytoplasm, and has diverse functions in RNA metabolism [71], including in the degradation of XRN1-sensitive unstable antisense RNAs [72]. Silencing of FLC is regulated mainly through histone modifications rather than DNA methylation [73]. We previously reported that the exosome complex and AtRRP6L proteins function in DNA methylation-independent silencing and affect the histone modification pathway in some heterochromatic loci in Arabidopsis [31]. Along with our previous findings, regulation of FLC silencing mainly by histone modifications suggested that the FLC locus could be one of the targets of the AtRRP6L proteins. In accord, we observed that defects in AtRRP6L proteins caused a decrease in antisense RNAs, resulting in the alteration of histone modifications and de-repression of FLC. It is intriguing to speculate that AtRRP6L proteins may have dual functions in FLC silencing via regulation of antisense transcription, which means that AtRRP6L proteins could participate in 2 different pathways, one involved in H3K4 demethylation and the other involved in H3K27 trimethylation. First, AtRRP6Ls could participate in the H3K4 demethylation pathway via regulating synthesis of AS I and AS II. Second, AtRRP6L proteins could function in the H3K27 trimethylation pathway via regulating the synthesis of ASL. However, more work will be needed to untangle the interrelationships of the different lncRNAs and the roles they play in the epigenetic architecture at FLC. How AtRRP6L proteins and the ncRNAs controlled by them help recruit chromatin modifiers to modulate silencing by affecting histone modifications that repress transcription remains an intriguing topic for future work. The atrrp6l1-3 allele was isolated from the BASTA population from the University of Wisconsin [31]; the atrrp6l1-1, atrrp6l2-2, atrrp6l2-3, atrrp6l2-4 and atrrp6l3-1 alleles correspond to SALK_004432, SALK_113786, SALK_011429, and SALK_149898, and SALK_122492, respectively. iRNAi lines of RRP41, csl4-2, RNA Pol IV (SALK_128428, nrpd1a-3, nrpd1-3), RNA Pol V (SALK_029919, nrpd1b-11, nrpe1-11) mutants were described previously [8], [74], [75]. All Salk alleles are in the Col-0 ecotype and the University of Wisconsin alleles are in the Ws ecotype. The RNAi-mediated knockdown of RRP41 was induced by germinating and growing seedlings on ½× MS plates containing 8 mM 17β-estradiol, following a previously published method [8]. Long day and short day conditions for plant growth were 16 hours light/8 hours dark and 8 hours light/16 hours dark, respectively. Flowering time was measured by counting rosette leaf number at the time of flowering [76]. For chromatin immunoprecipitation (see below), we used the Tandem Affinity Purification (TAP) affinity tag to selectively precipitate RRP6L1 by expressing a RRP6L1- TAP fusion protein. For construction of plant lines with affinity-tagged RRP6L1 for RNA-IP and ChIP, we complemented atrrp6l1-3 with the RRP6L1-TAP transgene. For RRP6L1-TAP complementation, the entire genomic region of RRP6L1 including 1. 5 kb upstream from the ATG codon was amplified by PCR using LA taq polymerase (Takara) and cloned into TAP-tag carrying destination vector pDB1008 [8]. For complementation with the TAP tagged transgene, the homozygous atrrp6l1-3 mutant was transformed using Agrobacterium-mediated transformation and the progeny plants containing both the T-DNA insertion allele and the transgene were identified by PCR. Trizol reagent (Invitrogen) was used to isolate total RNA from seedlings. 10-, 14-, 18-, and 21-day-old seedlings were used for examining the expression of ASL. 11-day-old seedlings were used for investigating expression of genes and antisense RNAs examined in our study. For RT-qPCR, 2–4 µg of total RNA was digested with DNase I (Fermentas) and reverse transcribed for one hour at 42°C (oligo-dT primers) or at 50°C (gene-specific primers), with 100 units of PrimeScript reverse transcriptase (Takara). RT-qPCR (MyiQ-iCycler; Bio-Rad) was used to quantify transcripts using the comparative threshold cycle method (ΔΔCt, Table S1 shows primer sequences), with ACTIN 7 (At5g09810) as an internal reference. For tiling RT-PCR, sets of serial primers were designed in intervals of 100–200 nt. After obtaining the full-length ASL transcript, another set of overlapping primers was designed to make sure the 3′ and 5′-ends of the RNA have been isolated. The PCR products amplified by tiling RT-PCR were cloned into the pBluescript KS vector and sequenced using T7 and T3 primers. ChIP was conducted following a previously-described method [77]. Each experiment used 1. 5 grams of tissue from 11-day-old seedlings. All ChIP experiments used at least two biological replicates and at least two technical replicates. Anti-H3K4me3 (07-473) and anti-H3K27me3 (ab6002) were purchased from Millipore and Abcam, respectively. IgG Sepharose 6 Fast Flow (GE Healthcare) was used for ChIP using RRP6L1-TAP tagged line. The mock antibody control used an equal amount of chromatin that was not treated with antibody. The ChIPed DNA was purified using PCR purification kit (Fermentas) and qPCR was performed. Supplemental Table S1 lists the primers used for PCR. MNase-ChIP was performed following a previously-described method [78]. Two grams of tissue from 11-day-old seedlings was fixed using 1% formaldehyde solution for 10 min and washed with distilled water several times. The fixed samples were homogenized with HONDA buffer (20 mM HEPES-KOH pH 7. 4,0. 44 M sucrose, 1. 25% Ficoll, 2. 5% Dextran T40,10 mM MgCl2,0. 5% Triton X-100,5 mM DTT, 1 mM PMSF, 1% plant protease inhibitors) and then filtered through miracloth. After isolation of the nucleus-containing fraction by centrifugation, the fraction was treated with MNase (NEB) at 37°C for 10 min. Anti-histone H3 (ab1791) was used for the ChIP. The purification of ChIPed DNA and qPCR was performed as described in ChIP assay. RNA-IP assays were performed as described previously [31], [79]. Two grams of tissue from seedlings at 11-days-old was fixed with 1% formaldehyde. For purification of RRP6L1-TAP or RRP6L1 RNA complexes, the chromatin was incubated with prewashed IgG Sepharose 6 Fast Flow (GE Healthcare) or with polyclonal anti-AtRRP6L1 antibodies, respectively, at 4°C overnight. H3K27me3-RNA complex purification was performed using anti-H3K27me3 (ab6002) overnight following by incubation with protein A agarose beads. Immunoprecipitated RNA purification used phenol∶chloroform and PrimeScript reverse transcriptase (Takara) and sequence specific primers were used for cDNA synthesis. Supplemental Table S1 lists the primers used for PCR. Eleven-day-old seedlings were treated with 5 µM α-amanitin (Sigma) for 0,6 and 9 h or 17 h. After RNA extraction, cDNA was synthesized using ASL-specific and ACTIN 7 primers, followed by qRT-PCR. The level of ASL transcript was normalized relative to the level of ACTIN 7 transcript. Total RNA extracted from 11-day-old seedlings was treated with TPE (Epicentre) at 42°C, and purified with phenol∶chloroform. Complementary cDNA was synthesized using RNA sequence specific primers followed by RT-qPCR.

Arabidopsis FLOWERING LOCUS C (FLC) delays flowering; therefore, repressing expression of FLC provides a critical mechanism to regulate flowering. This mechanism involves multiple levels of regulation, including genetic regulation by transcription factors, and epigenetic regulation by modifications of genomic DNA and histones at the FLC locus. This work examines the role of non-coding RNAs in the epigenetic regulation of FLC, finding that the different RNAs produced from the FLC locus may have different functions in altering the epigenetic landscape at the FLC locus, and revealing that AtRRP6L1 and AtRRP6L2, two components of the exosome, an RNA-processing complex, play roles in regulating these non-coding RNAs. Therefore, this work explores the complex ties between RNA processing, non-coding RNAs, and epigenetic regulation of FLC, a key repressor of flowering.

Abstract Introduction Results Discussion Materials and Methods

biology and life sciences

2014

Arabidopsis RRP6L1 and RRP6L2 Function in FLOWERING LOCUS C Silencing via Regulation of Antisense RNA Synthesis

The effectiveness of a mass vaccination program can engender its own undoing if individuals choose to not get vaccinated believing that they are already protected by herd immunity. This would appear to be the optimal decision for an individual, based on a strategic appraisal of her costs and benefits, even though she would be vulnerable during subsequent outbreaks if the majority of the population argues in this manner. We investigate how voluntary vaccination can nevertheless emerge in a social network of rational agents, who make informed decisions whether to be vaccinated, integrated with a model of epidemic dynamics. The information available to each agent includes the prevalence of the disease in their local network neighborhood and/or globally in the population, as well as the fraction of their neighbors that are protected against the disease. Crucially, the payoffs governing the decision of agents vary with disease prevalence, resulting in the vaccine uptake behavior changing in response to contagion spreading. The collective behavior of the agents responding to local prevalence can lead to a significant reduction in the final epidemic size, particularly for less contagious diseases having low basic reproduction number R 0. Near the epidemic threshold (R 0 ≈ 1) the use of local prevalence information can result in divergent responses in the final vaccine coverage. Our results suggest that heterogeneity in the risk perception resulting from the spatio-temporal evolution of an epidemic differentially affects agents’ payoffs, which is a critical determinant of the success of voluntary vaccination schemes. Immunization through the vaccination of populations has been estimated to annually prevent 2-3 million deaths from infectious diseases such as measles, diphtheria, pertussis and tetanus [1]. This number may rise substantially with the development of strategies to further increase global vaccine coverage [2]. Apart from conferring a long-term protection against the disease to the vaccinated individual, vaccination has an even more important community-level benefit. A sufficiently high vaccine coverage makes it difficult for the pathogen to find susceptible hosts, thereby conferring herd immunity to the whole population [3,4]. Consequently, even those members of the community who are unable to get vaccinated, such as newborns and immune-suppressed individuals, are protected against the disease. In principle, any disease caused by a pathogen that only has human hosts can be eradicated by mass immunization, provided there is a sufficiently efficacious vaccine that is readily available. Such an outcome has been realized for smallpox [5,6] and is expected to be achieved for polio [7,8]. Conversely, the presence of a significant fraction of non-immunized individuals, which disrupts the population’s herd immunity, can result in the recurrent outbreaks of vaccine-preventable diseases such as measles, mumps and pertussis [9]. Elucidating the mechanisms that promote wider acceptance of vaccination in the population can therefore help explicate the reasons behind the failure of immunization programs. One of the most important challenges in implementing an effective immunization program is to ensure that enough individuals agree to get vaccinated. This decision could be based on many factors such as an individual’s knowledge about the costs, including perceived side-effects, and benefits of vaccination, as well as the social, economic and cultural environment to which they belong [10,11]. The lack of public confidence in the efficacy and/or safety of vaccines can give rise to vaccine hesitancy (i. e. , delay or refusal to get vaccinated despite the availability of vaccine services) [12], and in extreme cases generate vaccine scares [13,14]. Even in the absence of any bias against a vaccine as such, vaccine uptake in the population may vary over time with changing prevalence of the disease. Indeed, it is expected that individuals will be more likely to get themselves vaccinated when there is a higher risk of getting infected [15]. Conversely, low disease incidence may often lead to a significant drop in vaccine uptake, presumably because of the lower perceived risk of contracting the disease [16]. This suggests that when the threat of infection is high the individual has a strong incentive to get vaccinated, while at times of lower risk she may be tempted to avoid vaccination and free-ride on the herd immunity provided by immunized members of a population without bearing any cost herself. However, if everyone argues in this manner and avoids vaccination, it would leave the population completely exposed to invasion by the pathogen. This is essentially an instance of a social dilemma [17] that often arises in strategic interactions between rational individuals, who are trying to maximize the benefits accruing to them from their actions and those of others [18]. That is, while free-riding appears to be optimal from an individual’s perspective, it leads to a clearly undesirable collective outcome. This is one of the problems central to game theory, which therefore provides a natural framework for understanding the conditions under which a population of rational individuals will voluntarily decide to get vaccinated. Most earlier studies of interaction between disease spreading and vaccine uptake behavior that incorporated a game theoretic framework have assumed homogeneous, well-mixed populations [19–22]. Thus, the risk of infection for every individual, as well as the protection offered to them by immunized individuals in their neighborhood, is identical. However, in reality, individuals interact primarily with neighboring members of their social networks and can have widely different contact structures [23]. Considering the network microstructure governing contacts between individuals can explain aspects of the collective outcomes of spreading contagion processes [24–31] and strategic interactions [32] that do not manifest in well-mixed models of populations. However, models that investigate vaccine uptake behavior by individuals in social networks typically do not incorporate strategic considerations in terms of explicit payoffs, i. e. , the net benefit associated with specific collective actions. Instead, agents are assumed to imitate the behavior of their more “successful” neighbor [33,34]. Additional model-based studies of voluntary vaccination by agents located on a social network have offered new perspectives on the impact of network contact structure [35], presence of local sub-groups [36], and the role of beliefs [37] and learning [38] on decision-making. As models incorporating strategic decision-making and those utilizing social network approaches each describe different aspects of vaccine uptake behavior (see [39] for a review), a framework combining both may come closer to capturing the complexity associated with such behavior in reality. To understand the interaction between human behavior and epidemic dynamics [40–42], in this paper we present a model in which rational agents take strategic decisions to vaccinate themselves on the basis of information about the disease prevalence and the immune status of their neighbors on a social network. Each agent decides their action by playing a game against a hypothetical opponent who shares the same neighborhood as it. Unlike previous studies that use a similar framework of strategic interactions, in our model the payoffs defining the structure of the game incorporate real-time information on the specific situation prevailing in the network neighborhood and consequently vary dynamically amongst individuals. Thus, the games played by the different agents change over time with the spread of the disease across the network, resulting in an emergent spatio-temporal heterogeneity in the nature of the games. We find that this heterogeneity at the level of individual agents, in terms of both information available to them as well as their response, can have significant implications for population-level outcomes such as the final epidemic size and the extent of vaccine coverage. We also examine how the source of the information, viz. , global (fraction of the population that is infected) or local (fraction of infected neighbors), that agents may use in assessing the risk of getting infected can lead to very different collective outcomes. The implications of our results reported here suggest that access to real-time information about the state of an evolving epidemic can change the risk perception and affect the vaccine uptake decisions taken by individuals. These in turn result in emergent patterns of collective choice behavior that may provide useful insights into the mechanisms driving vaccine acceptance, which could be relevant for public health planning. In our model, we study the dynamics of two coupled processes, namely epidemic spreading and the evolution of vaccine uptake behavior, on a social network of N agents. The connection topology of the network is specified by the contact structure among individuals in a given population. The time-scale of an epidemic considered here is much shorter than durations over which the network structure may change significantly as a result of births, deaths and migrations of individuals. This makes it relevant in contexts where a population is suddenly confronted with a situation that warrants vaccination within a short time-frame, such as a scenario involving the accidental release of a pathogen for which a vaccine is available, or perhaps a bioterrorism incident [43,44]. The spread of the disease over the network changes the status of an agent which, at any instant, can be in one of three possible states, namely, susceptible (S), infected (I), and recovered (R). We assume that recovery from the disease confers immunity from further infection to an agent. The disease is assumed to spread through direct contact between agents with a transmission rate β, while infected agents recover from the disease after an average time period of τI. Thus, the disease dynamics follows the well-known SIR model [45]. We have explicitly verified that qualitatively similar results are obtained upon varying either β or τI, keeping all other parameters fixed (see S1 Fig). Introducing vaccination in this framework allows a susceptible individual to avoid the possibility of getting infected by immediately achieving an immune status (which effectively corresponds to the R state). As an epidemic propagates through the population, each agent can have access to local information about the number of infected cases among her network neighbors (i. e. , with whom she has direct contact), as well as global information about the disease prevalence in the entire network. In reality, such information is obtained through different channels, e. g. , via mass-media in the case of global information and through word of mouth for local information. The agents also have information about the extent to which their neighborhood offers them protection from the disease. This is provided by their knowledge of how many of their neighbors are immune as a result of either having recovered from the disease earlier, or through vaccination. Each agent utilizes the above information to determine their likelihood of getting infected. Based on this threat perception, the agents subsequently make a strategic decision on whether to get vaccinated by taking into account the “cost” associated with vaccination. This cost arises from the threat of side-effects, either real or perceived, as well as the effort involved in getting vaccinated, and tempts the agent to free-ride on the protection that may be offered by the immunity of their neighbors, particularly when the prevalence is low. By engaging in such behavior agents can enjoy the benefits of immunization without bearing the cost of getting vaccinated themselves. However, if every agent argues along the same lines, it will lead to extremely low vaccine uptake, causing the loss of herd immunity and exposing the population to the risk of an epidemic outbreak of a vaccine-preventable disease. This results in a dilemma for a population of well-informed rational agents, who decide their actions entirely on the basis of maximizing their individual payoffs. As a game-theoretic framework provides a natural setting for investigating such social dilemmas, we model the vaccine uptake decision process of individual agents in terms of games. In order to make a strategic decision each agent plays a symmetric 2-person game against a virtual opponent who shares the same neighborhood and hence has identical information. Note that in the heterogeneous setting that we consider where the network neighborhood of each agent is distinct, the information on the basis of which she takes a decision also differs from agent to agent. Thus, each agent asks whether by changing her action she could have increased her payoff given her unique situation. In order to achieve this, we allow the focal agent to consider a virtual opponent to which she attributes information identical to that which she possesses, and follows the same decision process as herself. In other words, the agent plays against her virtual self in order to see if she could have done better had she chosen a different action with the same information and in the same setting. At each round of the game, an agent has a choice of two possible actions, i. e. , to get vaccinated (v) or not (n). The cost and benefit associated with the choices is represented in terms of a payoff matrix. An important feature of our approach is that the payoffs evolve with the progress of the epidemic and the ensuing change in vaccine coverage in the population. The payoff received by the focal player j, where j ∈ [1, N], is represented by a function of the form Uxy (fi, fp), where x, y ∈ {n, v} are the actions of the focal player and the virtual opponent, respectively (see table in Fig 1). Here, fp is the fraction of neighbors that are immune and fi is a linear combination of local and global information about the disease prevalence: f i (j) = α (I / N) ︸ global + (1 − α) (k i (j) / k (j) ) ︸ local. Note that I is the number of infected agents in a population of size N, while k (j) is the total number of neighbors of the focal agent j, of which ki (j) individuals are infected. By tuning the parameter α ∈ [0,1], we can consider any information scenario between the two extreme cases wherein an agent uses exclusively local (α = 0) or global (α = 1) information. As mentioned earlier, a high disease prevalence (i. e. , a large value of fi) ensures that the benefits of vaccination outweigh its cost, thereby acting as an incentive for the focal agent to get vaccinated. It is reasonable to assume that the values of the payoffs associated with the decision to vaccinate increase with prevalence as, all other things remaining same, susceptible agents will be more likely to get infected when fi is high. Specifically, it is beneficial to get vaccinated if even one of the neighbors of the focal agent remains susceptible to the disease. Thus the value of fp will have less relevance in such a situation. Therefore, a reasonable simplification is to assume that Uvv and Uvn are increasing functions of fi and independent of fp. Another important consideration is when all the neighbors of the focal agent are immune to the disease. In this situation, there is a high probability that the agent will successfully escape infection even if she opts not to get vaccinated. Hence, analogous to the arguments used above, it is reasonable to assume that the payoffs associated with the decision to “not vaccinate” increase with the fraction of protected neighbors. In view of the fact that the utility of getting vaccinated depends primarily on the number of neighbors not protected against the disease, which is directly related to the probability of the focal agent to get infected, we consider Unv and Unn as increasing functions of fp and independent of fi, as a simplification. For concreteness, we choose the simplest possible linear functional form for Unv, Unn, Uvv and Uvn as follows: U nv = a f p + b, U nn = c f p + d, U vn = e f i + f, U vv = g f i + h. This linear form in fi and fp has the added advantage of not having multiple solutions (i. e. , Nash equilibria, explained later) for any particular choice of fi and fp, which would have required invoking additional selection criteria for choosing among them. As the payoff functions are time-varying, the nature of the game can change depending on the hierarchical relation between the payoffs that prevails at any instant. To characterize the hierarchy of payoff functions in the (fi, fp) space, we note that when fi is high and fp → 1, it is possible to escape infection as long as most of the neighbors are immune but in the absence of protection from the neighborhood, vaccination is vital to an individual. This suggests the following relation between payoffs: Unv > Uvv > Uvn > Unn, i. e. , the game is Hawk-Dove [46]. When fi is low and fp → 1, the non-vaccinators prevail as there a very low risk of infection and most of the population is immune to the disease. This would result in Unv > Unn > Uvv > Uvn, i. e. , the game is Deadlock [47]. When fi is high and fp → 0, the benefits of vaccination outweigh the perceived cost of vaccination because of the high risk of contracting disease. This results in the hierarchal relation Uvv > Uvn > Unv > Unn, i. e. , the game is Harmony [48]. When fi is low and fp → 0, it is extremely tempting to not get vaccinated because of low prevalence. However the possibility of being infected is non-zero, which makes vaccination a viable choice. This results in Unv > Uvv > Unn > Uvn, i. e. , the game is Prisoner’s Dilemma [49] (see Fig 1, inset). These four games govern the preference that an agent has for each action (viz. , to vaccinate or to not vaccinate) at the four extremities of the (fi, fp) parameter space. In the interior of this space, the hierarchies among the payoffs gradually change, thereby giving rise to different games. To ascertain that the system behaves in the same way as explained above at these four extremities, we choose the parameters a − h such that Unv, Unn, Uvn and Uvv satisfies the inequalities mentioned above. The payoff associated with not getting vaccinated when the opponent chooses to vaccinate (Unv) is always greater than the corresponding payoff for the case where both do not get vaccinated (Unn), as the latter situation exposes both to the risk of being infected. We hence set a = c without loss of generality. Similarly, the payoff received when both the focal player and her virtual opponent get vaccinated (Uvv) is greater than that obtained when only the focal player is vaccinated (Uvn). This is because, in principle, the latter scenario implies that she, instead of her virtual opponent, could have avoided the cost associated with vaccination. We hence set e = g without loss of generality. If the parameters a − h satisfy the following relations: a + b > e + h > e + f > b, a + d > h > d > f, (1) then the situations discussed above (Hawk-Dove, Deadlock, Harmony and Prisoners’ Dilemma) will prevail at the four extremities of the (fi, fp) space. As the information on the basis of which the agent decides whether to vaccinate changes over time, the nature of the game played by her also varies. Thus, while under certain conditions, agents can exhibit a propensity to free-ride (e. g. , when prevalence is low), our model is also consistent with recent observations about the prevalence-elasticity of the demand for vaccines [50–52]. As the epidemic spreads in the population each susceptible agent j will, at any time t, choose an action such that a unilateral change of action will not yield a higher payoff. In game theory, such an action profile is known as a Nash equilibrium [53]. If player j (and her opponent) decides to vaccinate with probability pj (po) and not vaccinate with probability 1 − pj (1 − po), the expected utility for agent j can then be calculated as ϵ j = p j (p o (U vv + U nn − U nv −U vn) + U vn − U nn) + p o (U vn − U nn) + U nn. Given that the game is symmetric, the Nash equilibrium would be either pj = 0 or pj = 1 if it is pure, or if it is mixed then the agent j would vaccinate with the probability p j = U nn − U vn U vv + U nn − U nv − U vn. Note that the expression for the vaccination probability for a mixed strategy Nash equilibrium is similar to the strategy referred as mixed ESS in the Bishop-Cannings theorem [54]. As this probability will be different for each susceptible agent, it introduces heterogeneity in the individuals’ decision across the network due to differences in the risk-perception of each agent. Also, as this probability can change with time, an agent can change her decision as the disease spreads over the network. Incorporating such spatio-temporally varying strategies for the vaccine uptake of agents on a network presents a more realistic way of examining the coupled dynamics of vaccination and disease. In order to study the consequences of the interplay between the strategic decision-making process for vaccine uptake and epidemic spreading, we simulate the stochastic spread of a directly transmitted disease on empirical social networks of villages in southern India [55], as well as model networks (the simulation algorithm is outlined in S1 Text). All agents in our model are initially susceptible and 0. 5% of the nodes in a network are randomly chosen to become infected to simulate the onset of an epidemic. Note that no node is initially in a vaccinated state. We employ the Gillespie stochastic evolution algorithm [56] to determine the time at which the next event will happen and which node would take part in that event. The event could be one of the three different types of transitions that can change the state of a node: (i) disease transmission (S → I), (ii) recovery (I → R), and (iii) vaccination (S → R). Disease transmission is a contact-dependent transition and can take place only when node j in state S is in contact (i. e. , has a connecting link) with nodes in state I. Recovery is a time-dependent transition and depends on the time interval spent by a node j in infected state (for more details see [57]). Vaccination is an information-driven transition, which involves strategic decision making (as shown in Fig 1). The simulation is stopped when there are no infected nodes remaining in the network. The payoff parameter values used for all simulations reported here are a = 0. 45, b = 0. 3, d = 0. 002, e = 0. 5, f = 0 and h = 0. 2, which satisfy the relations (1). As shown in S2 Fig, the results reported here are robust with respect to different choices of these parameter values (which are in any case highly constrained by the above-mentioned inequalities). The goal of our study is to see if voluntary vaccination can emerge as a result of spatially heterogeneous strategic decision making in response to individual-based assessment of an epidemic threat and if so, what role the source of information (local or global) may play in shaping this collective response. Fig 2 shows the results obtained for a simulated epidemic on the social network of one of the 75 villages in southern India from the data set of [55]. We stress, however that our results are qualitatively similar for other choices of social network (as shown in the subsequent figure). Fig 2 (a) and 2 (b) illustrates the final outcome of a simulated epidemic with transmission rate β = 0. 025 and average infectious period τI = 10 on the empirical social network of a specific village (village 55 in the data set), for the two extreme values of α (results for intermediate values of α are shown in S3 Fig). The blue color represents the nodes that escaped infection without getting vaccinated. Note that as all nodes were initially susceptible, the vaccine uptake behavior is entirely epidemic-driven. It is evident from the figure that more agents experience the disease (as indicated by red colored nodes) when the information available about prevalence is global (α = 1) as compared to when it is local (α = 0), although the vaccine coverage (as indicated by yellow colored nodes) is almost same. To understand the reason behind this disparity in the final outcome of epidemic simulated when considering different sources of information, we consider the time evolution of the fraction of nodes in different states, as shown in Fig 2 (c). For α = 0, the final fraction of agents that were infected during the epidemic, inf∞, is 0. 17 and the final fraction of agents vaccinated during the epidemic, vac∞, is 0. 22. In contrast, for the case α = 1, inf∞ is 0. 42 and vac∞ is 0. 19. Hence, even though vac∞ is similar for the two cases, there is a significant difference in the value of inf∞. It is clear from the figure that voluntary vaccination behavior emerges much later in the case α = 1 (at t = 20) as compared to α = 0, where it emerges almost immediately after initiating the simulated epidemic. As highlighted in the inset of the right panel of Fig 2 (c), in the case α = 1 the agents start getting vaccinated when the epidemic prevalence becomes significantly high. This emergent behavior is a reasonable description of how responses to epidemics typically unfold. For instance, in the absence of an efficient mechanism for the dissemination of incidence data, the media usually reports an outbreak only when the reported cases of the disease becomes sufficiently high. Once a disease has affected a significant proportion of population, even a subsequent high vaccine coverage would be unable to reduce the final fraction of infected agents. To test the robustness of these results with regard to the contagiousness of the epidemic, we simulated epidemics with different values of the basic reproduction number R 0, the average number of secondary cases resulting from a single primary infection in a completely susceptible population. For each value of R 0 we conduct 1000 trials to average over the effect of noise on the final size of the epidemic and vaccine coverage. On comparing the final outcome of these simulations, it is apparent that the value of inf∞ for α = 1 is always greater than the corresponding value for α = 0, independent of any choice of R 0 (Fig 2 (d) ). This underpins the previous observation that the epidemic infects a larger proportion of agents in the network when agents decide to get vaccinated based on the information about the global disease prevalence, as compared to local. However, a comparison of vac∞ for α = 0 and α = 1 reveals a more complex situation (see Fig 2e). For both low and high values of R 0, vac∞ is higher for α = 0 than for α = 1, but there is an intermediate range of R 0 in which the values of vac∞ for α = 1 are higher than for α = 0. Thus, there is a crossover of both the curves of vac∞ for α = 0 and α = 1. This shows that an epidemic simulated with these intermediate values of R 0 results in higher vaccine coverage when agents base their vaccination coverage on the global information as compared to local. An important point to note here is that the effect of high vaccine coverage in this regime of R 0 for α = 1 is not reflected in the final size of the epidemic (Fig 2 (d) ). This shows that even if the vaccine coverage in this regime of R 0 is high, the simulated epidemic affects more agents for α = 1 than for the case α = 0. A possible explanation of this is that in the case of global information the threat perception does not appear significant unless a large proportion of agents are affected by the epidemic and hence fails to overcome the perceived cost of vaccination. This results in limited vaccine uptake which does not provide any significant check on the spread of the epidemic. Furthermore, use of local prevalence information leads to localized elevated vaccine uptake in the neighborhoods of infectious agents which allow for efficient intervention. By contrast, vaccine usage is dispersed throughout the network in the case where global prevalence information is used, resulting in sub-optimal outcomes. To gain more insight into the dynamics of the model, we simulated epidemics on Erdős-Rényi networks with N = 1024 and average degree 〈k〉 = 10, for both α = 0 and α = 1 (see Fig 3 (a) –3 (c) ). The results are consistent with those obtained for the empirical social network. We note that similar results are obtained by increasing R 0 by changing τI instead of β as is done here (see S1 Fig). To see how the crossover behavior near epidemic threshold depends on the average degree 〈k〉 of the network, we simulated the epidemic on Erdős-Rényi networks having different average degree and on empirical social networks from the dataset of Ref. [55] whose largest connected component (LCC) size is greater than 1000. We calculated the area A enclosed between the two vac∞ vs R 0 curves for α = 0 and α = 1. In Fig 3 (d), we have shown how this area decreases with an increase in the value of 〈k〉. This indicates that this intriguing behavior is dependent on the average degree of network. In order to examine how the results are affected by the size of the population being considered, we display the dependence of vac∞ on N for α = 0 (top) and α = 1 (bottom) in Fig 4 (a). We observe that the change in the values of inf∞ and vac∞ with respect to R 0 show similar behavior on increasing the size N of the network. The vac∞ versus R 0 curves for these two different values of α show two different kinds of behavior, on increasing the system size. To investigate this change in behavior, we looked into the probability distribution of the final number of vaccinated agents V∞ calculated over 2000 trials. We found that for α = 1 this distribution is unimodal for all values of R 0, whereas for α = 0 a bimodal distribution is observed for some values of R 0, i. e. the probability distribution has peaks at two different locations. To identify where this behavior changes in the (R 0, α) parameter space, we characterize the bimodal nature of the probability distribution of V∞ by calculating the bimodality coefficient [58]: BC = m 3 2 + 1 m 4 + 3 (n − 1) 2 (n − 2) (n − 3) where, n represents the sample size, m3 and m4 refer to the skewness and kurtosis of the distribution, respectively. A value of BC greater than 5/9 suggests that the distribution is bimodal. Our computational study indicates that the probability distribution of V∞ is bimodal for values of α < 0. 5 [shown in Fig 4 (b) ]. This can be observed from Fig 4 (c), which shows how the probability distribution of V∞ changes on increasing the value of α from 0 to 1. This can be a potential signature of a subcritical (discontinuous) transition for local information and a supercritical (continuous) transition for global information. Vaccine hesitancy typically rises with decreasing disease incidence as a consequence of reduced risk perception among individuals of contracting the disease. Understanding the mechanisms driving such behavior is important as it can reverse the success of any immunization program close to achieving the eradication of a disease [59]. We utilize the framework of game theory to investigate vaccine uptake behavior, as it provides an intuitive description for the action of rational agents, i. e. in absence of any social or religious bias against decision to get vaccinated. In contrast to previous approaches, we simulate the spread of an infectious disease on a social network, where each agent can, at every time step, decide whether to get vaccinated. The decision-process of each agent is modelled by a game, in which the payoffs for different actions vary over time as the epidemic progress and the immunization status of the neighboring agents change. Each agent plays against a hypothetical opponent who shares the same neighborhood and thus has identical information, imposing symmetry on the payoff matrix. We examined whether information about an epidemic outbreak at the local or global level can lead to the emergence of voluntary vaccine uptake behavior in a population of agents that are aware of the benefits of free-riding on the immunity of their peers. In particular, we focused on how spatio-temporal heterogeneity in individuals’ vaccine uptake decisions can affect the overall vaccine coverage at the population level, and consequently determine the fate of an epidemic outbreak. We would like to stress that this heterogeneity is both in terms of the information an individual receives from the network neighborhood, as well as, the response based on her individual risk perception [60]. We observe that a defining factor for efficient disease control through voluntary vaccination is the source of information. Faster and more efficient vaccine coverage is observed for the case when individuals assess their risk of catching infection based on the prevalence in the local social network neighborhood, as opposed to that in the whole population of their social network. Compared to the size of the entire population, the number of cases that are reported in the initial phase of an epidemic are fairly low, and therefore an individual who only has access to the global prevalence information may not perceive the disease to be a serious threat. Consequently, the perception of risk in contracting the disease takes some time to become significant enough to incite vaccine uptake among individuals. However, by the time global prevalence becomes high enough so that the perceived risk of infection outweighs the cost of vaccination, the epidemic will have already affected a large fraction of the population. We find that this delay in the emergence of vaccination behavior can sometimes manifest as a large final size of the epidemic despite high vaccine coverage. On the other hand, the presence of disease in an agent’s neighborhood increases the risk of infection even at the early stage of an epidemic, and thus leads to an immediate increase in vaccine uptake. This not only increases the total vaccine coverage but also reduces the burden of disease. An intriguing observation in the case of agents using local information is that the emergence of voluntary vaccination results in bimodal distributions of the final epidemic size and vaccine coverage for diseases with R 0 ≈ 1. This behavior, observed close to the epidemic threshold, can be attributed to competition between the two possible final outcomes for the state of an initially susceptible individual, namely to get vaccinated or to get infected. Previous game theory based models of vaccination during epidemic outbreaks have considered the effect of strategic decision-making in well-mixed populations where all individuals have the same risk assessment [19,20]. In contrast, our model captures the impact of inhomogeneous risk and benefit perception at the individual level, which gives rise to spatio-temporally diverse games and hence different Nash equilibria across the population. Consequently, the whole population would never converge to a state in which every agent has the same strategy, unless the disease is completely eradicated. This also rules out the possibility that the strategic decision to vaccinate will disappear from the population with time, unlike in models that utilize imitation game dynamics to describe vaccination behavior. Indeed such models suggest that the persistence of high vaccine coverage can only be ensured by incentivizing vaccine distribution [61]. Our findings show that the model presented here provides a complementary mechanism for the emergence of voluntary vaccination. This arises as a response to the potential threat of an epidemic outbreak if each agent utilizes the information available to them and makes a rational decision whether getting vaccinated might be beneficial to her or not. One of the key assumptions that underpins our approach is that agents are well-informed and make rational decisions based on the information available to them. In reality, the conditions under which individuals make vaccination decisions may, of course, deviate from this assumption. However, the rational agent framework, where individuals take decisions purely based on self-interest, provides a benchmark for investigating voluntary vaccination behavior. This can be then extended to include, for example, the effect of personal beliefs and peer influence [52], which can result in anti-vaccine sentiments [62] or vaccine scares [63]. While we have investigated how the final size and vaccine coverage varies for diseases with different contagiousness (i. e. R 0), it is also possible to augment our model with additional parameters that capture other features such as case fatality ratio. For instance, two diseases with comparable R 0, such as Ebola and Influenza, and thus similar transmission rate and vaccination costs, could result in different coverages, based on the subjective perception of how harmful (or severe) a disease is. The dynamics of disease progression may also be modified by including additional stages, for instance to account for appreciably long infection periods [64]. Additionally, one could also explore the consequence of differential vaccine efficacy among individuals and finite durations for the protection afforded by the vaccine. The social network on which the disease spreads has, for simplicity, been assumed to be static through the course of an epidemic. However, over time the network can indeed change by vital dynamics, i. e. through individuals dying and new ones being born. An additional source of temporal variation in the connection structure arises from the changing behavior of the agents [65] including actions taken by them in response to the epidemic, such as social distancing [66,67]. We would like to stress that our results are independent of population size and meso-level structural details, such as the existence of modularity, but depend strongly on the degree (average number of contacts a person has) of the network. This could partly be because we are primarily considering the final outcome of the simulated epidemics, such as final epidemic size and total vaccine coverage. Another potential reason is that the strategic decision making in our model depends crucially on the neighborhood which is a micro-level detail of the social network. From a policy-making viewpoint, it is easier to estimate how many social contacts a person has on average rather than meso- and macro-level details, which widens the scope of our model and its results. We also stress on the importance of taking into account the heterogeneity in the disease status of neighbors in a social network for risk assessment when deciding whether to vaccinate. The prevalence aggregated over the whole population may sometimes result in a false perception of risk, especially if the disease is in one’s vicinity. The key outcome for public health planning is that accurate and localized reporting of disease outbreak is crucial for changing individuals’ risk perception and thereby their attitude towards vaccination, especially during the initial phase of an epidemic.

A major factor underlying the success of voluntary vaccination schemes is the public perception about the costs and benefits associated with vaccines. Individuals may avoid vaccination if they perceive the risk of infection to be low compared to the potential hazards and inconveniences associated with getting vaccinated. However, in the course of an epidemic outbreak individuals may opt to vaccinate because of the associated higher risk perception. Modeling individual decision-making in the presence of an evolving epidemic using games, we show that spatial heterogeneity in the vaccine-uptake behavior emerges with the spread of disease on social networks. Our results highlight the crucial importance of the information source shaping an individual’s risk perception for achieving high vaccine coverage.

Abstract Introduction Model Results Discussion

medicine and health sciences infectious disease epidemiology applied mathematics immunology sociology social sciences vaccines preventive medicine social epidemiology probability distribution mathematics network analysis social networks infectious disease control vaccination and immunization public and occupational health infectious diseases computer and information sciences epidemiology probability theory game theory biology and life sciences physical sciences

2019

Epidemic prevalence information on social networks can mediate emergent collective outcomes in voluntary vaccine schemes

Clinical strains of HCMV encode 20 putative ORFs within a region of the genome termed ULb′ that are postulated to encode functions related to persistence or immune evasion. We have previously identified ULb′-encoded pUL138 as necessary, but not sufficient, for HCMV latency in CD34+ hematopoietic progenitor cells (HPCs) infected in vitro. pUL138 is encoded on polycistronic transcripts that also encode 3 additional proteins, pUL133, pUL135, and pUL136, collectively comprising the UL133-UL138 locus. This work represents the first characterization of these proteins and identifies a role for this locus in infection. Similar to pUL138, pUL133, pUL135, and pUL136 are integral membrane proteins that partially co-localized with pUL138 in the Golgi during productive infection in fibroblasts. As expected of ULb′ sequences, the UL133-UL138 locus was dispensable for replication in cultured fibroblasts. In CD34+ HPCs, this locus suppressed viral replication in HPCs, an activity attributable to both pUL133 and pUL138. Strikingly, the UL133-UL138 locus was required for efficient replication in endothelial cells. The association of this locus with three context-dependent phenotypes suggests an exciting role for the UL133-UL138 locus in modulating the outcome of viral infection in different contexts of infection. Differential profiles of protein expression from the UL133-UL138 locus correlated with the cell-type dependent phenotypes associated with this locus. We extended our in vitro findings to analyze viral replication and dissemination in a NOD-scid IL2Rγcnull-humanized mouse model. The UL133-UL138NULL virus exhibited an increased capacity for replication and/or dissemination following stem cell mobilization relative to the wild-type virus, suggesting an important role in viral persistence and spread in the host. As pUL133, pUL135, pUL136, and pUL138 are conserved in virus strains infecting higher order primates, but not lower order mammals, the functions encoded likely represent host-specific viral adaptations. Human cytomegalovirus (HCMV) is a member of the β-herpesvirus subfamily that, like all herpesviruses, persists indefinitely in infected individuals through a latent infection. HCMV persistence is associated with an increased risk of age-related pathologies including atherosclerosis [1], [2], immune senescence [3], [4], [5] and frailty [6], [7], [8] in otherwise healthy individuals. Reactivation of HCMV from latency in individuals with compromised T cell immunity, including transplant and AIDS patients, is a significant cause of morbidity and mortality [9], [10], [11], [12]. Further, HCMV is the leading cause of infectious disease-related birth defects [12], [13], [14]. The mechanisms controlling the outcome of infection, and in particular the latent infection, in the diverse cell types infected by HCMV in the human host are poorly understood. Understanding these mechanisms is essential to ultimately controlling the overt viral pathologies in individuals with weakened or insufficient T cell-mediated immunity as well as non-overt pathologies associated with viral persistence. Clinical isolates of HCMV uniformly contain a unique region of the genome, termed ULb′, that encodes 20 predicted open reading frames (ORFs) [15], [16], [17]. While the actual coding potential is known for only a few ULb′ ORFs, these ORFs are considered dispensable for viral replication in laboratory models such as fibroblasts since laboratory-adapted strains of the virus lacking the entire ULb′ region replicate with increased kinetics and to increased viral yields relative to clinical strains. As such, it is postulated that ULb′ ORFs function in latency, immune evasion, virus dissemination in the host, or other aspects of pathogenesis. We have previously identified sequences in the ULb′ region of the HCMV genome encoding the UL138 protein (pUL138) that are required for a latent infection in CD34+ hematopoietic progenitor cells (HPCs) infected in vitro [18], [19]. Disruption of the UL138 coding sequence (cds) results in a virus that replicates with increased efficiency relative to the wild-type virus in HPCs in the absence of a reactivation stimulus. While disruption of UL138 ablates the latent phenotype, a more robust loss of latency phenotype results from the disruption of additional ULb′ sequences around and including the UL138 locus, indicating that other viral sequences in addition to UL138 contribute to the outcome of infection in HPCs. The mechanism by which pUL138 functions in viral latency is unknown; however, it has recently been reported that the pUL138 enhances levels of tumor necrosis factor receptor (TNFR) on the cell surface [20], [21]. We have recently reported that UL138 is part of a larger 3. 6-kb polycistronic locus [22]. pUL138 is expressed from the 3′ end of three overlapping transcripts (3. 6-, 2. 7-, and 1. 4-kb) by both canonical and stress-inducible alternative mechanisms of translation initiation [19], [22]. These transcripts encode three additional putative ORFs, UL133, UL135, and UL136 upstream of UL138. We detected proteins derived from these ORFs during transient expression of UL138 cDNAs, as well as during HCMV infection [22]. This locus may serve to coordinate the expression of pUL133, pUL135, pUL136 and pUL138 for a common function in dictating the outcome of infection in the cell. The present study represents an initial characterization of the unique HCMV genetic locus encoding UL133, UL135, UL136, and UL138, the proteins expressed from this locus, and the role of the locus in infection. We collectively refer to this locus as the UL133-UL138 locus. pUL133, pUL135, and pUL136 are previously uncharacterized proteins. Like pUL138, pUL133, pUL135, and pUL136 were expressed early during productive infection and ultimately localized to the Golgi apparatus. These proteins were each associated with the Golgi as integral membrane proteins with large C-terminal cytosolic domains. Despite localization to the Golgi, pUL133, pUL135, pUL136, and pUL138 were only partially co-localized. We hypothesized that the UL133-UL138 locus functions in mediating context-dependent outcomes of infection. As would be expected for ULb′ sequences, the UL133-UL138 locus was dispensable for viral replication in primary fibroblasts. We demonstrate that like UL138, the UL133-UL138 locus or UL133 alone impeded replication in CD34+ HPCs, consistent with a role for the encoded proteins in latency. Surprisingly, the locus augmented replication in endothelial cells. The disparate cell-type dependent phenotypes associated with the UL133-UL138 locus correlated with differential profiles of expression from the locus in endothelial and CD34+ HPCs. While all four proteins were expressed in fibroblasts, we fail to detect pUL136 in endothelial cells and do not detect pUL135 or pUL136 in CD34+ HPCs. Further, the UL133-UL138NULL virus exhibited an increased capacity for replication and/or dissemination in a NOD-scid IL2Rγcnull-humanized mouse model following stem cell mobilization relative to the wild-type virus, further suggesting an important role for the UL133-UL138 locus in latency and reactivation. The role of individual proteins encoded by this locus in infection and latency awaits further investigation. These proteins likely represent virus adaptations to higher order primates acquired through co-speciation as the protein sequences are conserved in chimpanzee CMV (ChCMV) and to some extent in rhesus CMV (RhCMV), but are not present in CMV strains infecting lower vertebrates. Our work defines a novel locus that underscores the complexity of the virus-host interactions governing HCMV replication. We have previously demonstrated the coding potential of UL133, UL135 and UL136 within the ULb′ region of the HCMV genome [22]. pUL133, pUL135, and pUL136 are encoded by three polycistronic transcripts of 3. 6-kb, 2. 7-kb and 1. 4-kb, respectively, which also encode pUL138, an established determinant of HCMV latency [19], [22]. The expression of the transcripts is sensitive to inhibition of protein synthesis, but not to inhibition of viral DNA synthesis, indicating early kinetics of expression [19]. In the present study, we have characterized the expression and localization of these novel proteins as well as identified a role for this novel locus in infection. To aid in the analyses of pUL133, pUL135, pUL136, and pUL138, we constructed a series of recombinant viruses in the BAC clones of the FIX strain of CMV. We inserted the myc epitope tag in-frame at the 3′ terminus of each ORF (Figure 1A). The resulting viruses are termed FIX-UL133myc, FIX-UL135myc, FIX-UL136myc [22], and FIX-UL138myc [19]. The kinetics of productive viral replication in primary human embryonic fibroblasts (MRC5) infected with each of these viruses or the parental strain, FIX-WT, were measured by TCID50 over a time course. Despite variation in the eclipse phase, the recombinant viruses containing epitope tags replicated with kinetics and to yields reflecting that of the wild-type virus in MRC5 cells (Figure 1B and S1). The differences between the viral yields are not significant. The analogous recombinant viruses were also made in the TB40E strain of HCMV. TB40E viruses also replicated with wild-type kinetics similar to the FIX viruses. To analyze the kinetics of pUL133, pUL135, pUL136, and pUL138 expression, MRC5 cells were infected with FIX-UL136myc (Figure 1C) or TB40E-UL136myc (Figure 1D) at a multiplicity of infection (MOI) of 2. FIX-UL136myc was used in these experiments to aid in the detection of pUL136 because this protein is expressed at low levels in infected cells and we have not been able to generate an adequate UL136-specific antibody. Proteins were detected by immunoblotting with a monoclonal antibody specific to the myc epitope tag to evaluate pUL136 or rabbit polyclonal antibodies raised against pUL133, pUL135 or pUL138 [19], [22]. pUL133 and pUL135 have an apparent molecular mass of 39- and 43-kDa, respectively. Similar to pUL138, pUL133 and pUL135 were expressed by 12 hours post infection (hpi). pUL133 and pUL138 were expressed throughout the time course of 120 hpi, while pUL135 expression tapered off dramatically at 84 hpi. Multiple isoforms of pUL136 were detected at 37-kDa, 27-kDa, and 20-kDa, termed pUL136-37K, pUL136-27K, and pUL136-20K, respectively. An additional minor isoform was detected at 24-kDa. The 27K and 20K isoforms of pUL136 are predominantly expressed by 12 hpi and persist throughout the time course of infection. pUL136-37k exhibited slightly delayed kinetics of expression and was detected robustly at 36 hpi. Further studies are required to determine the origins of the pUL136 isoforms. Presumably, full-length pUL136 is derived from the 3. 6- and 2. 7-kb transcripts encoding UL138. We previously detected at least one smaller pUL136 isoform expressed from the 1. 4-kb UL138 transcript [22]. The relative ratios of the pUL136 isoforms differ in the TB40E infection relative to the FIX expression. This difference was consistently observed in multiple experiments; however, the reason for this difference is not known. The dynamic expression of pUL133, pUL135, and pUL136 from the UL138 transcripts and their expression patterns are intriguing and may have important implications for their function during infection. We previously determined that pUL138 localized to the Golgi apparatus during infection or transient expression [19]. To determine the subcellular localization of pUL133, pUL135, and pUL136, we infected MRC5 fibroblasts with FIX-UL133myc, FIX-UL135myc, or FIX-UL136myc at an MOI of 2 and analyzed the subcellular distribution of each protein at 24 and 48 hpi by indirect immunofluorescence using a monoclonal antibody specific to the myc epitope tag. Cells were co-stained with an antibody against the Golgi marker GM130 and with DAPI to identify the nucleus. Cells infected with FIX-UL138myc were used as a reference. By 48 hours post infection, each protein accumulated in the Golgi similarly to pUL138myc (Figure 2). pUL136myc resembled pUL138 in that pUL136myc was Golgi associated at both 24 and 48 hpi. FIX-or TB40E-UL136myc express myc-tagged versions of all pUL136 isoforms (Figure 1C and 1D); however, these forms cannot be differentiated in these experiments. pUL133myc and pUL135myc exhibited more diffused cellular staining at 24 hpi with predominant localization to the Golgi by 48 hpi. Similar to pUL138, pUL133, pUL135, and the full-length pUL136 isoform (s) are predicted to have amino terminal transmembrane (TM) domains (Figure 3A). The TM domains predicted for pUL135 and pUL136 span the membrane once, whereas pUL133 has two predicted membrane spanning domains. We investigated the membrane association of these proteins by analyzing crude membrane preparations from MRC5 cells either uninfected or infected with FIX-UL136myc at an MOI of 1. Cells were treated with cycloheximide 4 hours prior to harvest to allow newly synthesized proteins to traffic to their resident compartments. 3K and 12K fractions contain cytoplasmic and nuclear membranes, respectively, whereas the 25K and 100K pellet fractions contain lighter vesicles and microsomal membranes. The 100K supernatant contains soluble proteins. Proteins were detected in each fraction using the myc antibody to detect pUL136 or polyclonal antibodies to each HCMV protein. The major histocompatability complex I (MHC I) protein was analyzed as a control using a monoclonal antibody. Similar to pUL138, pUL133 and pUL135 were concentrated in the 25K and 100K pellets indicating their association with lighter microsomal membranes (Figure 3B). This data is consistent with the localization of these proteins to the Golgi (Figure 2). pUL133, pUL135 and pUL138 also accumulated in the 3K pellet representing heavier membranes. This broad distribution may reflect trafficking of these proteins through the secretory pathway and is similar to the distribution of MHC I. pUL136-20K was predominantly associated with the 100K supernatant, suggesting that this is a cytosolic protein. Consistent with this finding, we have previously demonstrated that smaller truncated forms of pUL136 are expressed from the 1. 4-kb UL138 transcript, which begins 300 nucleotides downstream of the full-length UL136 start codon and lacks a predicted TM domain [22]. The pUL136-27K protein was associated with the 25- and 100-K pellets as observed for pUL133, pUL135, and pUL138, suggesting that this protein contains the N-terminal portion of the protein, including the transmembrane domain. The predicted molecular mass of the full-length pUL136 protein based on amino acid sequence is 27-kDa. Intriguingly, we did not detect pUL136-37k in experiments with a cycloheximide chase. This may be due to the low abundance and/or the narrow window of expression of this protein (Figure 1C). Alternatively, this result may be due to the rapid turnover of the protein or the sensitivity of a modification resulting in the 37-kDa mass to cycloheximide treatment. The predicted TM domains for pUL133, pUL135, and full length pUL136 are in the N-terminus of each protein, as is the case for pUL138 (Figure 3A). In order to determine if pUL133, pUL135, and pUL136-27k are integral membrane proteins, we treated 25k microsomal membrane fractions with buffer containing 100mM sodium carbonate (Na2CO3), which is typically used to disrupt protein-protein interactions without affecting protein-lipid interactions. Following salt extraction, the pellet and the supernatant were analyzed by immunoblotting using antibodies as described for Figure 3B. pUL133, pUL135 and pUL136-27k were resistant to Na2CO3 extraction and recovered exclusively in the pellet (Figure 3C), indicating that these proteins are integral membrane proteins similar to pUL138 and MHC I [19]. To determine the orientation of pUL133, pUL135, and pUL136 in membranes, 25k microsomal membrane fractions from cells infected with FIX-UL136myc were treated with proteinase K in the presence or absence of 1% Triton-X100. Lysates were analyzed by immunoblotting as described previously for Figure 3B (Figure 3D). Approximately 50% of MHC I was digested by proteinase K under native conditions to yield a lower molecular mass band as previously reported [19], [23], [24]. MHC I was completely converted to the lower molecular mass form in the presence of detergent. As previously reported, pUL138 associated with microsomal membranes was efficiently digested (60–90%) in the absence of detergent [19]. Similarly, proteinase K treatment of microsomal membranes resulted in efficient digestion of pUL133 (60–90%), pUL135 (80–90%), and pUL136 (95–100%) in the absence of detergent as observed in four independent experiments. The predicted N-terminal position of the TM domains for each of these proteins and the near complete digestion of these proteins in the absence of detergent suggest that, like pUL138, the large C-terminal domains of each protein is exposed on the cytosolic face of Golgi membranes. The less efficient digestion of pUL133 may be related to the possibility that pUL133 may span the membrane twice. Given that the majority of pUL133 is digested, we interpret these results to indicate that both the N- and C-termini of pUL133 are on the cytosolic face of 25K fraction membranes. The membrane association and the topology for each of these proteins were unchanged when expressed individually by lentivirus transduction (Figure S2). The observed topology of these proteins in Golgi membranes likely has important implications for their function during infection. Taken together, these data demonstrate that the three proteins, pUL133, pUL135, and pUL136 expressed with pUL138 from polycistronic transcripts have similar properties to pUL138 with regard to localization and membrane association. To address the significance of the UL133-UL138 locus in viral infection, we constructed a recombinant virus in the TB40E strain that lacks the entire UL133-UL138 locus termed TB40E-UL133-UL138NULL (Figure 4A). The TB40E virus strain was used for these studies to analyze the role of the UL133-UL138 locus in infection as TB40E exhibits broader tropism than the FIX strain. MRC5 cells were infected with the parental TB40E strain (TB40E-WT) or TB40E-UL133-UL138NULL at an MOI of 0. 2. Viral replication was measured over a time course by TCID50 (Figure 4B). As would be expected for viruses lacking ULb′ sequences, TB40E-UL133-UL138NULL replicated with kinetics and to yields similar to TB40E-WT. Not unexpectedly, TB40E-UL133-UL138NULL also replicated to similar yields as the wild-type virus at high MOI (Figure S3A). The analogous virus was also made in the FIX strain of HCMV and this virus also replicated to similar yields as the wild-type virus (Figure S3B). These results indicate that the UL133-UL138 locus is dispensable for viral replication in fibroblasts. Since each protein from the UL133-UL138 locus localized to the Golgi (Figure 2), we next wanted to determine the extent to which the proteins co-localize in the context of productive infection. MRC5 cells were infected with TB40E-UL133-UL138NULL at an MOI of 2 to provide the context of viral infection. At 6 hpi, cells were then co-transduced for 48 h with four lentivirus contructs expressing pUL133, pUL135, pUL136, and pUL138, each with a different carboxy terminal epitope tag. Antibodies specific to the Flag (FLAG), hemagglutinin (HA), myc, or glu-glu (EE) epitope tags were conjugated with Quantum dots and used to label pUL133FLAG, pUL135HA, pUL136myc, and pUL138EE, respectively. When expressed independently, pUL133FLAG, pUL135HA, pUL136myc, and pUL138EE localized in part to a perinucler region resembling the Golgi, implying that the localization of these proteins to the Golgi did not require other viral proteins or the context of infection (Figure S4). Similar to infection (Figure 2), both pUL133FLAG and pUL135HA exhibited a more diffuse localization showing both cell surface and perinuclear staining resembling the Golgi (Figure S4). Importantly, Figure S4 demonstrates that there was no appreciable background or bleed through of these signals in each of the five channels. When expressed together, there was substantial overlap in the signals for all four proteins (Figure 5). Due to the high level of pUL133FLAG expression, merged images are shown of all viral proteins with and without pUL133FLAG. pUL136myc and pUL138EE exhibited the greatest overlap in their localization pattern. There were some notable differences in staining where pUL135HA appears to be excluded from perinuclear regions strongly stained with pUL133FLAG and pUL138EE. It is interesting that regions of pUL135HA staining are typically juxtaposed to regions of intense pUL133FLAG and pUL138EE staining. Importantly, no signal was detected in infected cells transduced with an empty vector, demonstrating the specificity of the antibodies and that these antibodies were not bound by viral Fc receptors. The significance of these patterns and regions of overlapping and non-overlapping protein localization awaits further investigation. Given the role of pUL138 in latency, the other proteins expressed from the UL133-UL138 locus may function to cooperate with pUL138 in establishing a latent infection in CD34+ cells. We first analyzed protein expression from the UL133-UL138 locus in infected CD34+ HPCs. CD34+ HPCs, freshly isolated from umbilical cord blood were infected with TB40E-WT or TB40E-UL136myc at an MOI of 2. Pure populations of infected (GFP positive) CD34+ cells were isolated by fluorescent activated cell sorting (FACS) and seeded into long-term bone marrow cultures (LTBMC) over a stromal support. Whole cell lysates were prepared at 2 and 5 days post infection (dpi) and analyzed for protein expression by immunoblotting using polyclonal antisera to HCMV proteins, the monoclonal antibody specific to IE1 and IE2, or a monoclonal antibody recognizing the myc epitope tag. Representative blots are shown from five independent experiments. We detected transient expression of IE1, which did not persist past 2 dpi (Figure 6A). IE1 transcript expression in HPCs has been detected previously [19], [25], [26]. Similarly, pUL133 was detected at 2 dpi but not at 5 dpi. As previously observed, pUL138 was expressed in CD34+ HPCs at 2 and 5 dpi [19]. By contrast, we detected very low to undetectable levels pUL135 or pUL136 in these cells under conditions where the expression of pUL138 was readily detected. While we cannot exclude the possibility that pUL135 and pUL136 are expressed in CD34+ HPCs, their relative abundance is substantially diminished compared to infected fibroblasts and we consistently fail to detect these proteins in CD34+ HPCs with our current tools. These results indicate that expression of individual proteins from the UL133-UL138 locus may be differentially modulated based on the context of infection. Given the low levels of expression of pUL135 and pUL136 detected in CD34+ HPCs, it is uncertain as to what role these proteins may play in the establishment and maintenance of latency. The differential expression of UL133-UL138 locus proteins in CD34+ cells may be due to the relative abundance of transcripts encoding pUL138. As we have shown previously, pUL138 can be expressed from 3. 6-, 2. 7-, and 1. 4-kb transcripts, although it appears to be most efficiently expressed from the 1. 4-kb transcript [22]. By contrast, pUL133 and pUL135 are expressed only from the 3. 6- and 2. 7-kb transcripts, respectively. We reasoned that if the 1. 4-kb transcript was expressed more abundantly in CD34+ HPCs than the 3. 6- and 2. 7-kb transcripts, this could explain the enhanced levels of pUL138 expression. RNA was isolated from infected CD34+ HPCs at 2 dpi and the 3. 6-, 2. 7-, and 1. 4-kb transcripts were quantitated by real time reverse-transcriptase PCR (qRT-PCR) using primers specific to UL133, UL135, UL136 or UL138. As a control, we used TB40E-UL133-UL138NULL where none of these transcripts are expressed. We detected significant levels of 3. 6-, 2. 7- and 1. 4-kb transcripts compared to control infection. Transcript levels of IE1 and IE2 were nearly identical in cells infected with both TB40E-WT and TB40E-UL133-UL138NULL (fold change of 1. 3 and 1. 4, respectively; data not shown). The 3. 6-, 2. 7-, and 1. 4-kb transcripts were present in 1∶3∶3 ratio, respectively (Figure 6B). While these results could explain the higher level of pUL138 expression, they do not explain the absence of pUL135 and pUL136 since the ratio of 2. 7- to 1. 4-kb transcripts is 1. Therefore, translational regulation of these transcripts may also play a role in protein abundance. We next sought to determine the role of the UL133-UL138 locus in latency. We have previously demonstrated a role of pUL138 in the FIX strain for promoting the latent infection [18], [19]. For a relative comparison, we generated a virus using the TB40E strain, TB40E-UL138Stop, where UL138 was disrupted by the substitution of the initiator codon with a stop codon. Further, as pUL133 expression was reliably detected in infected CD34+ HPCs (Figure 6A), we generated a virus containing a similar disruption in UL133. We analyzed the replication of these recombinants relative to TB40E-WT in MRC5 cells using multi-step growth curves (Figure 7A). Similar to FIX strains containing disruptions in UL138, TB40E-UL133Stop and TB40E-UL138Stop replicated with slightly enhanced yields relative to TB40E-WT [19]. These results demonstrate that UL133, like UL138, is dispensable for viral replication in fibroblasts. To analyze latency, CD34+ HPCs were infected with TB40E-WT, TB40E-UL133-UL138NULL, TB40E-UL133Stop, or TB40E-UL138Stop at an MOI of 2. Pure populations of infected CD34+ cells were incubated in LTBMC for 10 dpi. Cell lysates were analyzed by an infectious centers assay to determine the number of cells required to form an infectious center [18], [26]. This assay is a measure of virus replication, but distinct from a plaque forming or TCID50 assays. The infectious centers assay is appropriate for these measures because each infected CD34+ HPC does not go on to produce virus upon reactivation. It is thought that the differences between viruses with regards to the establishment of or reactivation from latency is the number of infected cells producing virus as opposed to the yield of virus per cell [18], [26]. TB40E-UL133-UL138NULL (p = 0. 0043), TB40E-UL133Stop (p = 0. 0005), and TB40E-UL138Stop (p = 0. 0143) replicated with increased efficiency in HPCs relative to the wild type virus (Figure 7B), producing 5-fold greater infectious centers compared to cells infected with the wild-type virus. Similar to previous findings for UL138, these data suggest a role for UL133, UL138, and the entire UL133-UL138 locus in suppressing viral replication, presumably for the latent infection in HPCs [18], [19]. Further work is required to determine the individual contributions of these and other UL133-UL138 locus proteins to viral replication or latency in this model. HCMV infects a wide array of cell types in the human host. ULb′ genes are predicted to encode functions that mediate viral replication, dissemination, and persistence in the host. To determine if the UL133-UL138 locus is important for viral replication in other cell types, we compared viral yields in endothelial and epithelial cells infected with TB40E-WT or TB40E-UL133-UL138NULL (Figure 8A). We chose 3 different primary human endothelial cell types (microvascular lung, HMVEC; umbilical vein, HUVEC; aortic, HUAEC), one endothelial cell line (HAEC), and one primary human renal epithelial cell type HRCE. In each case, cells were infected with an MOI of 0. 1 and cell lystates were harvested 10 dpi. Surprisingly, TB40E-UL133-UL138NULL exhibited a modest to severe replication defect (5–200 fold) in all endothelial cell types analyzed. By comparison, TB40E-UL133-UL138NULL replicated similarly to TB40E-WT in the epithelial cell type tested. To further explore the role of the UL133-UL138 locus in endothelial cells, we analyzed multi-step replication of UL133-UL138NULL in HMVECs. HMVECs were infected with TB40E-WT, TB40E-UL133-UL138NULL or TB40E-UL138Stop at an MOI of 0. 2 and whole cell lysates analyzed for virus production over a time course following infection by TCID50 (Figure 8B). TB40E-UL133-UL138NULL exhibited a 2-log defect in replication relative to TB40E-WT. This defect in replication was not due to a failure of the mutant virus to enter or spread in HMVEC cells based on the initial number of infected (GFP+) cells and the formation of plaques, respectively (data not shown). Further, this defect cannot be overcome by infecting cells at higher multiplicities (MOI of 2; data not shown). Viruses lacking only UL138 exhibited no defect relative to TB40E-WT, suggesting pUL138 is not required for replication in these cells. We next analyzed protein expression from the UL133-UL138 locus in endothelial cells. HMVECs were infected at an MOI of 2 and whole cell lysates were analyzed for pUL133, pUL135, pUL136 and pUL138 expression at 2 and 5 dpi by immunoblotting. As a control for infection, we analyzed the expression of the IE1 and IE2 proteins. We detect IE1 in HMVECs, but IE2 is consistently expressed at low to undetectable levels in these cells (Figure 8C). With regards to the UL133-UL138 locus, we readily detected expression of pUL135 and pUL138 at 2 dpi. pUL133 was detected, but only at the 5 dpi time point. Expression of pUL136 was undetectable in each of three independent infections. The failure to detect pUL136 may be due to the variability in expression of the pUL136 isoforms or in the inherent instability of pUL136 (Cicchini and Goodrum, unpublished results). Given the three cell type-dependent replication phenotypes associated with the UL133-UL138 locus, we analyzed viral replication and dissemination in a NOD-scid IL2Rγcnull-humanized mouse model. This model represents the only animal model to effectively study HCMV infection parameters including of latency and reactivation [27]. NOD-scid IL2Rγcnull mice were engrafted with human CD34+ HPCs. The huCD34+-engrafted mice were transfused with human fibroblasts infected with TB40E-WT or TB40E-UL133-UL138NULL or uninfected fibroblasts as a negative control (8 mice per experimental group). At 4 weeks post infection, four mice in each group were treated with granulocyte-colony stimulating factor (G-CSF) and AMD-3100 to induce stem cell mobilization and viral reactivation. At two weeks post mobilization, we measured viral genome loads in bone marrow and spleen tissues by quantitative TaqMan PCR with probes and primers specific for HCMV US28. HCMV genomic DNA was detected in the bone marrow of both wild type and UL133-UL138NULL infected non-mobilized mice (351 copies/µg DNA for TB40E-WT vs. 291 copies/µg for TB40E-UL133-UL138NULL; p = 0. 17; not significant) and did not increase significantly upon mobilization (Figure 9A). Both viruses showed an increase in splenic viral DNA loads following mobilization suggesting that cells infected with both viruses were disseminated to the spleen (Figure 9B). However, mobilization of the mice infected with TB40E-UL133-UL138NULL resulted in a 2- to 3-fold higher levels of viral DNA load in the spleen compared to wild-type-infected animals (p = 0. 08). Mice infected with TB40E-WT had an overall 1. 4-fold increase in spleen viral DNA load following mobilization versus a 43-fold increase in TB40E-UL133-UL138NULL infected mice. Low levels of viral DNA were detected in the spleens of unmobilized mice infected with either TB40E-WT or UL133-UL138NULL because infected cells do not efficiently traffic out of the bone marrow in the absence of mobilization. These data indicate that the UL133-UL138 locus is important for modulating viral replication, reactivation or dissemination in this model. In an effort to understand the possible function of the proteins encoded from the UL133-UL138 locus, we searched the known protein sequence databases for protein sequence similarity using BLASTpsi (http: //blast. ncbi. nlm. nih. gov/Blast. cgi). Further, we used PHOG (http: //phylofacts. berkeley. edu/orthologs/) to predict super-orthologs based on phylogenetic analysis [28]. Finally, we used Phyre (http: //www. sbg. bio. ic. ac. uk/~phyre/) to predict three dimensional structure using homology modeling, which does not rely on conservation of protein sequence [29]. No cellular or viral homologs were identified by any of these bioinformatics methods for any of the UL133-UL138 locus cds with the exception of HHV-5/CMV orthologues (data not shown). Further, no protein structures could be predicted. Due to the lack of identifiable protein structure, we next analyzed these proteins for regions of disorder using Disopred2 (http: //bioinf. cs. ucl. ac. uk/disopred). This algorithm predicted large regions of disorder across pUL133, pUL135, pUL136 and pUL138 suggesting that these are intrinsically disordered proteins (data not shown). Intrinsically disordered proteins typically adapt structure through their interactions and often interact with a large number of proteins [30]. These analyses indicate that the UL133-UL138 locus proteins are unique to CMV and, as such, will require further molecular and biochemical studies to understand their role in infection. To determine the extent of conservation of the UL133-UL138 locus within CMV orthologues, we aligned the ULb′ sequences available from NCBI for HCMV (strain TB40E; Accession: EF99921. 1; GI: 157779983), ChCMV (strain heberling; Accession: NC_003521. 1; GI: 20026600) and RhCMV (strain 68-1; Accession: NC_006150. 1; GI: 51556461) (Figure 10). Of note, sequences with similarity to the ULb′ region were identified only in ChCMV and RhCMV and not in any CMVs of lower mammals for which a sequence is known. Orthologues to each gene encoded within the UL133-UL138 locus are present in ChCMV. The ChCMV orthologues for the strains aligned share 44. 2%, 46. 7%, 53. 8% and 56. 7% similarity at the amino acid level with pUL133, pUL135, pUL136, and pUL138, respectively. In RhCMV, the ULb′ region is positionally conserved. However, few RhCMV genes in this region share substantial sequence identity with HCMV ULb′ genes [31]. RhCMV Rh166 ORF shows similarity to both HCMV pUL133 (26. 6%) and pUL138 (35%). In addition, the RhCMV Rh171 ORF also shows similarity to HCMV pUL133 (27. 6%). Hence, Rh166 and Rh171 might represent the orthologues for HCMV pUL138 and pUL133, respectively, though the exact corresponding homologues are not clear from the analyses performed. No significant identity between UL135 and UL136 of HCMV and RhCMV proteins was observed. We have identified and characterized a novel locus within the ULb′ region of the HCMV genome that encodes three proteins, pUL133, pUL135, and pUL136, in addition to the pUL138 latency determinant [22]. The pUL133, pUL135, and pUL136 proteins have not been previously characterized. Proteins encoded by the UL133-UL138 locus predominantly localize to Golgi membranes (Figure 2,3 and 5) with large C-terminal domains exposed on the cytosolic face of the membranes (Figure 3). UL133-UL138 locus proteins exhibit overlapping localization, but did not completely co-localize (Figure 5). The profile of expression of the individual proteins from the UL133-UL138 locus varied substantially depending on the context of infection (Figure 1C–D, 6A, 8C). As would be expected for ULb′ sequences, the UL133-UL138 locus was dispensable for viral replication in fibroblasts (Figure 4 and S3). Disruption of the UL133-UL138 locus resulted in a virus, UL133-UL138NULL, with increased frequency of infectious centers formation in CD34+ HPCs relative to the wild-type virus (Figure 7B), consistent with a failure to establish a latent infection. Intriguingly, UL133-UL138NULL exhibited a severe replication defect in primary human endothelial cells (Figure 8A and 8B). These three distinct context-dependent phenotypes indicate that the UL133-UL138 locus may mediate context-dependent outcomes of infection. The mechanism by which the UL133-UL138 locus contributes to cell-type dependent outcomes of infection awaits further investigation. Importantly, UL133-UL138NULL exhibited increase replication and dissemination in a NOD-scid IL2Rγcnull-humanized mouse model (Figure 9), indicating a role for this locus in mediating viral replication, latency or dissemination in vivo. Further, as the UL133-UL138 locus is unique to CMV strains of higher order primates, we predict that these proteins represent viral adaptations to infection and persistence in the primate host (Figure 10). The ULb′ region of the HCMV genome was first recognized over a decade ago [15]. As this region is unique to CMV strains infecting primates and is dispensable for viral replication in fibroblasts, the most common model for productive viral replication, it presents significant challenge for research and has been understudied. The genes in this region are postulated to encode viral adaptations to the host, involved in immune evasion, pathogenesis, or viral persistence or latency. The coding potential of genes in the ULb′ region has been shown for only a few of the 20 putative ORFs. These include UL138, UL141, UL142, UL144, and UL146. UL138 encodes a protein that is required, but not sufficient, for the latent infection in HPCs infected in vitro [18], [19]. It has been recently demonstrated that pUL138 functions to regulate cell surface levels of TNFR [20], [21]. UL141 [32] and UL142 [33] encode proteins to evade elimination by natural killer cells. UL142 is expressed late in infection and encodes a MHC class I related molecule, which renders the cells resistant to NK mediated cell lysis. UL144 functions as a tumor necrosis factor (TNF) homolog that activates NFkB, which in turn enhances the expression of CCL22, a chemokine, which attracts Th2 and T regulatory cells. Thus UL144 may help the virus evade immune surveillance by enhancing the Th2 response while subverting the Th1 response [34], [35]. UL146 encodes a viral CXCL chemokine that binds the IL-8 receptor to enhance neutrophil chemotaxis and degranulation [36]. Another putative CXC homologue, UL147, has been identified [36]; however, the coding potential of this ORF and its function has not been explored. Most recently, we have extended our studies on pUL138 to characterize the protein coding capacity of the UL133, UL135, and UL136 ORFs, encoded on polycistronic transcripts with UL138 [22]. The individual functions of pUL133, pUL135, and pUL136 have not yet been determined. The sequence analyses of these proteins do not indicate any obvious sequence motifs or functional domains that suggest protein function. The fact that these proteins share similar cellular and biochemical properties (Figure 2,3 and 5) to that of pUL138 suggests that they may function together during infection. We hypothesize that these proteins may function with pUL138 in modulating the outcome of infection in a context dependent manner. Our ongoing research focuses on determining the functions of each of these proteins during HCMV infection. The proteins derived from the UL133-UL138 locus share identity only with other HHV-5 orthologues. While the proteins encoded by the UL133-UL138 locus are conserved in chimpanzee CMVs, orthologues have not been identified in cytomegalovirus strains infecting lower mammals, including RhCMV and MCMV [31]. In RhCMV, the ULb′ region is positionally conserved, but few proteins have considerable sequence identity (Figure 10) [31]. Both UL133 and UL138 share moderate similarity to rh166 while UL133 also shows weak homology to rh171. We propose, given the position of the Rh166 and Rh171 ORFs and the moderate conservation of sequence, that these ORFs represent orthologues of pUL133 and pUL138 of HCMV. No orthologues were identified for pUL135 or pUL136, as previously reported [31]. These observations suggest that UL133-UL138 locus resulted from co-speciation in higher order primates, and suggest an intriguing possibility that these proteins engage in virus-host interactions that are highly adapted to the host species. Our studies indicate, as would be expected for ULb′ sequences, that the UL133-UL138 locus was dispensable for replication in cultured fibroblasts (Figure 4). However, the locus augmented replication in primary endothelial cells (Figure 8) and impeded replication in CD34+ HPCs (Figure 7B). Further, profiles of gene expression from the UL133-UL138 locus varied depending on the cell type infected (compare Figure 1C and D to 6A and 8C). For example, pUL135 was not detected in CD34+ HPCs, but was expressed efficiently in endothelial cells and fibroblasts, while pUL136 could not be detected in endothelial cells of CD34+ HPCs. This finding suggests that the ultimate outcome of infection may rely on the profile of protein expression from the UL133-UL138 locus in individual contexts of infection. As pUL135 and pUL136 are not expressed in HPCs, they may not be required for establishing or maintaining latency, but may function in some other aspect such as reactivation. Interactions between proteins encoded by the UL133-UL138 locus or with unique cell-type dependent host factors may underlie the role of this locus in mediating cell-type specific infection outcomes. Preliminary studies aimed at understanding the function of UL133-UL138 locus proteins have revealed a complex network of interactions and positive and negative acting proteins (Umashankar, Petrucelli, Rak, and Goodrum, unpublished results). The localization of UL133-UL138 locus proteins to the Golgi (Figure 2 and 5) and their orientation in the membranes (Figure 3) is certainly foretelling of the function of these proteins. Proteins localized in the Golgi may play critical roles in viral assembly and egress, protein trafficking, apoptosis [37], [38], and the cellular stress response [38]. Accordingly, the recently defined role of pUL138 in modulating surface levels of TNFR, suggests that pUL138 may mediate protein trafficking [20], [21] and, therefore, the cellular response to signaling molecules. The conclusions drawn from our in vitro studies are bolstered by our in vivo studies in humanized mice. The increased viral loads of TB40E-UL133-UL138NULL virus in the spleens of NOD-scid IL2Rγcnull-humanized mice following mobilization suggests increased replication, reactivation or dissemination of this virus relative to the WT virus (Figure 9B). This in vivo finding further supports an important role for the UL133-UL138 locus in suppressing replication or reactivation for latency. Mobilization did not significantly increase WT or UL133-UL138NULL viral genome copy number in the bone marrow (Figure 9A), possibly reflecting the fact that mobilized cells quickly exit the bone marrow. The fact that higher genome levels were not measured in the bone marrow of UL133-UL138NULL-infected mice relative to WT-infected mice prior to mobilization, suggests that these viruses may not behave differently in this system in the absence of a reactivation stimulus. The nature of the humanized mice studies is such that the results cannot completely recapitulate our in vitro studies, yet they are highly consistent with our in vitro studies, both studies suggesting an important role for the UL133-UL138 locus in modulating the outcomes of infection. Future studies into the UL133-UL138 locus promise to reveal intriguing virus-host interactions unique to higher-order primates mediating viral persistence. Human cord blood was obtained from donors at the University Medical Center at the University of Arizona using a protocol approved by the Institutional Review Board. These specimens are completely deidentified and provided to our research group as anonymous samples. The studies requiring animals were carried out in strict accordance with the recommendations of the American Association for Accreditation of Laboratory Animal Care (AAALAC). The protocol was approved by the Institutional Animal Care and Use Committee (number IS00001049) at Oregon Health and Science University. Human embryonic lung fibroblasts (MRC5) (purchased from ATCC; Manassas, VA) were cultured at 37 °C in Dulbecco' s modified Eagle' s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 10 mM HEPES, 1 mM sodium pyruvate, 2 mM L-Glutamine, 0. 1 mM Non-essential amino acids, 100 U/ml penicillin and 100 µg/ml streptomycin. Human cord blood was obtained from donors at the University Medical Center at the University of Arizona using a protocol approved by the Institutional Review Board. These specimens are completely deidentified and provided to our research group as anonymous samples. Mononuclear cells and CD34+ HPCs were isolated and cultured as previously described [19], [22]. CD34+ cells were maintained in long-term culture as described previously [19] but using MyeloCult H5100 (Stem Cell Technologies). The M2-10B4 murine stromal cell line expressing human interleukin-3 (IL-3) and granulocyte-colony stimulating factor (G-CSF) and the S1/S1 murine stromal cell line expressing human IL-3 and stem cell factor (SCF) (kind gift from Stem Cell Technologies on behalf of D. Hogge, Terry Fox Laboratory, University of British Columbia, Vancouver, BC) and cultured as recommended [39]. Primary human microvascular lung endothelial cells (HMVEC), human umbilical vein endothelial cells (HUVEC), human umbilical vein endothelial cells (HUVEC), and human aortic endothelial cells (HUAEC) were purchased from Lonza (Walkersville, MD). HMVEC, HUAEC, HUVEC cells were cultured in EGM-2 MV (Microvasular Endothelial Cell Growth Medium-2; Lonza), EGM-MV (Endothelial Growth Medium- MV; Lonza) and EGM (Endothelial Growth Medium; Lonza). Human aortic endothelial cells (HAEC) were a generous gift from Andrew Yurochko and were cultured in EGM-2 (Endothelial Growth Medium-2; Lonza). Human renal cortical epithelial cells (HRCE) were purchased from Lonza (Walkersville, MD) and were cultured in REGM (Renal Epithelial Cell Growth Medium; Lonza). All cells were maintained at 37°C with 5% CO2. Recombinant bacterial artificial chromosomes (BACs) containing the HCMV genome were constructed in Escherichiae coli (E. coli) by linear recombination in a two-step positive-negative selection method that leaves no trace of the engineering process [19], [40], [41]. The green fluorescent protein (GFP) was engineered between US34 and TRS1 in BAC clones of FIX [42], [43] or TB40E [44] virus strains as a marker of infection. In the first step, the SW102 E. coli strain containing the FIX or TB40E BAC were used to insert a galk cassette between US34 and TRS1 genes. In the second step, the galk cassette was replaced by an SV40-eGFP-BGH Poly-A cassette PCR amplified from the pCMS-eGFP vector (Clonetech) to generate the FIX or TB40E BACs used as the parental wild type strains in all experiments herein. These variants replicate with kinetics and to titers identical to the parental strains (data not shown). Further recombinant viruses were generated by repeating the insertion and substitution of galk using a PCR product flanked by homologous viral sequences as described previously [19], [22]. Oligonucleotide primers used for BAC recombineering are described in Table 1. Recombinant viruses were screened by BAC digestion, PCR, and sequencing. Virus stocks were propagated, stored and titered as described previously [19]. Oligonucleotide primers used for making expression plasmids are described in Table 2. The UL133, UL135 and UL136 ORFs were PCR amplified using ORF specific primers flanked by a NheI site on the forward primer and a BamHI site on the reverse primer. The reverse primer contained the myc epitope tag sequence to generate 3′ tagged versions of each gene. The PCR products were cloned into the NheI and BamHI sites of the pCIG2 vector [22]. These constructs termed pCIG-UL133myc, pCIG-UL135myc, and pCIG-UL136myc, express proteins with a C-terminal myc epitope tag with a 5 amino acid linker between the protein coding sequence and the myc tag. To obtain HA (YPYDVPDYA), 3X-FLAG (DYKDDDDK), or Glu-Glu (E-E) (EYMPME) (at the C-terminus) versions of these proteins, the PCR products containing a specific epitope tag were cloned into NheI and EcoRV sites of pCIG2 vector as above. The resulting plasmids were named pCIG2-UL133FLAG-IRES-BLEO, pCIG2-UL135HA-IRES-HYGRO, pCIG2-UL136myc-IRES-NEO, and pCIG2-UL138EE-IRES-PURO. In these constructs, the eGFP downstream of IRES was replaced by drug resistance markers such as Hygromycin (HYGRO), Neomycin (NEO), Bleomycin (BLEO) or Puromycin (PURO). The plasmids were co-transfected with pLP1, pLP2 and pVSVG plasmids (Invitrogen, CA) at 2∶1∶1∶1 ratio into 293FT cells using Lipofectamine 2000 (Invitrogen, CA). Culture supernatants were harvested 48 hpi and concentrated at 17,000 rpm using a SW28 rotor for 2 h at 4oC. Pellets were resuspended in IMDM containing 10% BIT9500 (Stem Cell technologies). Lentiviruses were titered on fibroblasts using the TCID50 method. Immunoblotting was performed as described previously [19]. Briefly, 10–15 µg of protein lysates were separated on 4–12% NuPAGE Bis-Tris, (Invitrogen, CA) or 11% Bis-Tris gels by electrophoresis and transferred to 0. 45 µm polyvinylidene difluoride (Immobilon-FL, Millipore, MA) membranes. The proteins were immunoblotted using mouse α-myc (Cell Signalling) or rabbit polyclonal antibodies directed against each protein (Open Biosystems) and detected using fluorescently conjugated secondary antibodies and the Odyssey infrared imaging system (Li-Cor, NE). All antibodies used are listed in Table 3. Immunofluorescence to localize viral and cellular proteins in infected cells was performed as described previously [19]. Briefly, fibroblasts (5×104 cells/well in 24-well plates) were mock infected or infected with recombinant viruses encoding myc epitope-tagged pUL133, pUL135, pUL136 and pUL138 at an MOI of 2 for 24 and 48 h. Cells were fixed in 2% paraformaldehyde in PBS and stained with a rabbit antibody specific to myc epitope tag and visualized using a Zeiss 510 Meta confocal microscope (Carl Zeiss Microimaging, Inc. NY). The nucleus was stained with 1 µg/ml DAPI (4′, 6′-diamidino-2- phenylindole) and GM130 was used as a Golgi marker. Fibroblasts were seeded on to 12 mm glass cover slips in 24 well plates one day prior to infection. The next day cells were mock infected or infected with TB40E-UL133-UL138NULL at an MOI of 2. At 6 hpi, cells were transduced with lentiviruses containing pCIG2-UL133FLAG-IRES-BLEO, pCIG2-UL135HA-IRES-HYGRO, pCIG2-UL136myc-IRES-NEO, pCIG2-UL138EE-IRES-PURO alone or in combination in the presence of 8 µg/mL of Polybrene (Sigma-Aldrich, St. Louis, MO). Cells were processed for direct immunofluorescence 48 hours later using the same method used for indirect immunofluorescence except for the primary antibody incubations, nuclear staining, and mounting. Briefly, primary antibodies to the various epitope tags were conjugated to amine derivatized Quantum Dots (Molecular Probes, Invitrogen, Carlsbad, CA) of 525 nm (Anti-HA), 565 nm (Anti-myc), 585 nm (Anti-FLAG), and 625 nm (Anti-EE) emission wavelengths according to manufacturer' s instructions (Table 3). Primary antibodies were incubated in PBS-T + 1% BSA overnight at 4°C. Post staining, cells were washed 3 times in PBS-T and nuclei stained with Qnuclear Deep Red Stain (Molecular Probes, Invitrogen, Carlsbad, CA) according to manufacturer' s instructions. Coverslips were washed 3 times in PBS and mounted using Qmount Qdot mounting media (Invitrogen) according to manufacturer' s instructions. Cells were imaged using a Ziess 510 Meta Confocal Microscope as a lambda stack and unmixed using the Zeiss 510 Meta software version 4. 2 and images were processed using ImageJ software, NIH (http: //rsbweb. nih. gov/ij/download. html). Microsome preparation was done as described previously [19]. Briefly, fibroblasts were either mock infected or infected with FIX-UL136myc virus at an MOI of 1 for 48 h. Cells were treated with 50 µg/ml cycloheximide for 4 h prior to harvesting at 48 h post infection. Cells were resuspended in buffer A (250 mM Sucrose, 50 mM triethanolamine, 1 mM EDTA, 6 mM magnesium acetate, 50 mM potassium acetate and 1 mM dithiothreitol) and gently sonicated 3 times for 10 sec on ice at 30 sec intervals. Membranes were fractionated by differential centrifugation to obtain membrane pellets at 3000×g, 12,000×g, 25,000×g and 100,000×g. The supernatant obtained post 100,000×g was precipitated using trichloroacetic acid (TCA). All pellets were resuspended in identical volumes of buffer A and analyzed by immunoblotting using protein specific antibodies (Table 3). For salt extraction, the 25,000×g membrane fractions were treated with 100 mM Na2CO3 for 1h on ice and centrifuged at 100,000×g in a TLA 100. 3 rotor for 45 min at 4°C to separate the pellet and supernatant, subsequently precipitated using TCA. Equal amounts of input, pellet and supernatant were analyzed by immunoblotting as above. To determine the topology of proteins in the membrane, the 25,000×g membrane fractions were digested with 0. 5 µg/ml proteinase K in the absence or presence of 1% Triton X-100 for 1 hr at 37oC. Reactions were stopped by adding 1 mM PMSF and SDS sample buffer. Input and protease treated samples were analyzed by immunoblotting. Targets were detected by qRT-PCR (Quantitative reverse-transcription PCR) as described previously [22]. RNA was isolated from TB40E-WT (Sample) or TB40E-UL133-UL138NULL (Control) infected cells and DNase treated using the NucleoSpin RNA II kit (Machery-Nagel) and cDNA was generated using the transcriptor first-strand cDNA synthesis kit (Roche) according to the manufacturer' s instructions. qRT-PCR was performed with the LightCycler 480 Probes Master (Roche) according to the manufacturer' s instructions along with the Universal Probe Library (Roche) probes and primers specific for UL133, UL135, UL136 and UL138 [22]. IE1&2 genes were used as controls for infection. The human β-actin gene was used as a reference and the target levels were quantitated by a ΔΔCT method using the following equation [45]. Where, CT is the cycle threshold and E is the efficiency as determined using the Light Cycler 480 software. In our analysis, we considered a 2-fold change to be within the confidence interval for equally abundant target. CD34+ HPCs isolated from human cord blood were infected at an MOI of 2 for 20 h in IMDM supplemented with 10% BIT9500 serum substitute (Stem Cell Technologies, Canada), 2 mM L-Glutamine, 20 ng/ml low density lipoproteins and 50 µM 2-mercaptoethanol. Following infection, pure populations of infected CD34+ HPCs (GFP-positive) were isolated by fluorescence-activated cell sorting (FACSAria, BD Biosciences Immunocytometry Systems, San Jose, CA) using a phycoerythrin-conjugated antibody specific to CD34 (BD Biosciences) and cultured in transwells above an irradiated (4000 rads, 137Cs gammacell-40 irradiator type B, Atomic Energy of Canada LTD, Ottawa, Canada) M2-10B4 and S1/S1 stromal cell monolayer [39] for 10–12 days. The frequency of infectious centers production during the culture period was measured using a limiting dilution assay as described previously [18]. Briefly, infected HPCs were serially diluted 2-fold in LTBMC medium. An aliquot (0. 05mL) of each dilution was added to 12 wells (first dilution corresponds to 40,000 cells per well) of a 96-well tissue culture plates containing MRC5 cells. MRC5 cells were monitored for GFP expression for a period of 14 days. The frequency of infectious centers formed was calculated based on the number of GFP+ cells at each dilution using software, Extreme limiting dilution analysis (ELDA, http: //bioinf. wehi. edu. au/software/elda) [46]. NOD-scid IL2Rγcnull mice were maintained in a specific pathogen free facility at Oregon Health and Science University in accordance with Institutional Animal Care and Use Committee approved procedures. Mice were sublethally irradiated with 250 cGy by 137Cs g-irradiation and then engrafted with approximately 150,000 human CD34+ stem cells (Catalog #CB008F-S, StemCell Technologies; Vancouver British Columbia, Canada) via retro-orbital injection. At 4 weeks post inoculation of stem cells, the level of engraftment (determined as the percentage of human CD45+ present in the blood of total lymphocytes) was assessed by flow cytometry as previously described [27]. At 5 weeks post engraftment, the mice were infected by intraperitoneal injection of 1×107 normal human dermal fibroblasts previously infected with HCMV TB40E-WT or TB40E-UL133-UL138NULL. A third group of engrafted mice were mock infected with uninfected human fibroblasts. At 4 weeks post infection, a group of mice were treated for 7 days with 100 µl of G-CSF (300 µg/ml; Amgen) via a subcutaneous micro-osmotic pump (1007D; Alzet) and AMD3100 (5mg/kg) to mobilize their hematopoietic stem cells. As a direct comparator for the effects of virus reactivation-dissemination following mobilization an additional non-mobilized infected control group was included for each virus (n = 4/group). At 2 weeks post mobilization, mice were euthanized and bone marrow and spleen were harvested and snap frozen for subsequent analysis. Total DNA was extracted from 0. 8 g of mouse spleen and bone marrow via the DNAzol method (Life Technologies) and analyzed by quantitative PCR (TaqMan) for the presence of HCMV genomic DNA. For Q-PCR analysis, 1 µg of total DNA was analyzed using primers and a probe recognizing HCMV US28 sequences (probe = TGA TCC CGC TCA GTG T; forward primer = GAA CTC ATG CTC GGT GCT TTC; and reverse primer = CTT TGT GGC GCG ACT GAG A). The probe contains labeled with a 5′-end FAM reporter molecule and a 3′-end quencher molecule (Applied Biosystems, Foster City, CA). PCR reactions were prepared using TaqMan Universal PCR Master Mix (Applied Biosystems) according to the manufacturer' s instructions. To initiate the reaction the AmpliTaq Gold was activated at 95°C for 10 minutes and then 40 cycles (15 s at 95°C and 1 min at 60°C) were performed using a StepOnePlus TaqMan PCR machine. Results were analyzed using ABI StepOne software. The sensitivity of detection for this assay was approximately 50 HCMV DNA genomic copies as determine by using a plasmid containing the US28 amplicon to develop a standard curve. Data were analyzed using the statistical program GraphPad Prism 5.

Human cytomegalovirus is a ubiquitous herpesvirus that, like all herpesviruses, establishes a life long relationship with its host through a latent infection. The molecular basis of viral latency is poorly understood, in part, because viral determinants of latency and the corresponding virus-host interactions are not well defined. We have identified a polycistronic locus encoding the pUL138 latency determinant, as well as three previously uncharacterized proteins, pUL133, pUL135, and pUL136. We have characterized this novel locus, the proteins it encodes and demonstrated the role of the locus in modulating viral replication depending on the context of infection. While this locus is dispensable for productive replication in fibroblasts, it adversely impacts virus replication in primary hematopoietic cells, suggesting a role in establishing latency. Surprisingly, the locus is required for efficient replication in primary human endothelial cells. To our knowledge this is the first demonstration of a viral locus that can have positive, negative, or null effects on viral replication depending on the context of infection. Our work defines exciting new primate strain-specific determinants mediating viral replication and latency and exemplifies the complex nature of virus-host interactions in cytomegalovirus infection.

Abstract Introduction Results Discussion Materials and Methods

medicine biochemistry infectious diseases biology microbiology molecular cell biology

2011

A Novel Human Cytomegalovirus Locus Modulates Cell Type-Specific Outcomes of Infection

Appropriate diagnostics to monitor disease trends and assess the effectiveness and impact of interventions are essential for guiding treatment strategies at different thresholds of schistosomiasis transmission and for certifying elimination. Field validation of these assays is urgently needed before they can be adopted to support policy decisions of the national programme for control and elimination of schistosomiasis in P. R. China. We compared the efficacy and utility of different immunoassays in guiding control strategies and monitoring the endemic status of S. japonicum infections towards elimination. A cross-sectional survey was conducted in seven villages with different transmission intensities settings to assess the performance and utility of three immunoassays, e. g. , an indirect hemagglutination assay (IHA_JX), an enzyme linked immunosorbent assay (ELISA_SZ), and a dot immunogold filtration assay (DIGFA_SH). 6,248 individuals aged 6–65 years old who gave consent and supplied their stool and blood samples were included for data analysis. Results showed that ELISA_SZ performed significantly higher sensitivity (95. 45%, 95%CI: 92. 94–97. 97%) than IHA_JX (87. 59%, 95%CI: 83. 51–91. 49%) and DIGFA_SH (79. 55%, 95%CI: 74. 68–84. 41%), especially in subgroups with very low infection intensity. The specificity of ELISA_SZ, IHA_JX, DIGFA_SH in 6–9 year olds with occasional exposure was nearly 90%. DIGFA_SH performed the highest screening efficacy for patients among three assays with overall positive predicative value of 13. 07% (95%CI: 11. 42–14. 72%). We found a positive correlation of antibody positive rate of IHA_JX with results of stool examination in age strata (r = 0. 70, P<0. 001). Seropositivity of IHA_JX in children aged 6–9 years old showed an excellent correlation with prevalence of schistosome infection in the seven communities (r = 0. 77, P<0. 05). Studies suggest that ELISA_SZ could be used to guide selective chemotherapy in moderate or low endemic regions. IHA_JX could be used to as a surveillance tool and for certifying elimination of schistosomiasis through monitoring children as a sentinel population. Following great efforts by the Chinese government and professionals in public health, the People' s Republic of China (P. R. China) has became one of the most successful countries in the world for schistosomiasis control [1]–[3]. Transmission had been interrupted in five provinces, Guangdong, Guangxi, Fujian, Shanghai, and Zhejiang by 1995 [1], [4]. The total number of infected individuals decreased by 90% through 50 years endeavour [3], [5], [6]. Due to changes in social and environmental factors including more migration people with changes in economic system, serious flood occurred in 1998 along the Yangtze River, resurgence of schistosomiasis became a public health problem in P. R. China in the beginning of the new century [3], [6], [7]. Since 2004, the central government reinforced the national control programme, and made schistosomiasis one of four priorities for infectious disease control in P. R. China. The new national goals aim to transition of the control strategy from morbidity reduction to transmission control: reduce prevalence of S. japonicum infection below 5% by 2008 (threshold of infection control) and below 1% by 2015 (threshold of transmission control) [6], [7]. To reach these goals, a novel control strategy was initiated in 2004 that aimed at implementation of integrated human and vector control measures to reduce the transmission of S. japonicum by blocking egg contamination of cattle and humans to infect snails [7], [8]. Under policy of the national control programme, stages for schistosomiasis control are classified as morbidity control (prevalence over than 5% as defined by stool examination), infection control (prevalence of 1–5%), transmission control (prevalence lower than 1%), transmission interruption (no case detected in five years successively) and elimination (no case detected in another five years continuously after transmission was interrupted) in P. R. China [6]. Figure 1 shows current diagnostic approaches in relation to prevalence, treatment strategies and different control threshold towards schistosomiasis elimination in P. R. China. As transmission decreases, simple, affordable and accurate diagnostic tests are urgently needed for case detection, surveillance and assessment of the effectiveness of schistosomiasis control interventions. S. japonicum infection is usually determined by the Kato-Katz thick smear method and the miracidium hatching technique in P. R. China [9], [10]. However, stool examination is laborious and insensitive for monitoring in endemic areas characterized by very low prevalence and intensity of the disease, resulting in underestimation of the prevalence of infection in these regions [11]–[14]. Although they cannot be used to distinguish between current and past infection, immunodiagnostic assays based on antibody detection have been extensively applied in schistosomiasis control programme in P. R. China for years because of their advantages over parasitological tests including high sensitivity, rapid time to result, ease of use and ease of batching for population studies [15]–[17]. Three types of immunodiagnostic assays have been developed in P. R. China. Indirect hemagglutination assay (IHA) is currently the most widely used immunodiagnostic assay as a screening tool during World Bank Loan Project (WBLP) period and in national surveillance system on schistosomiasis in P. R. China [4], [18], [19]. Second, antibody-based enzyme linked immunosorbent assay (ELISA) was used to estimate the endemic status in the third nationwide sampling survey of schistosomiasis in P. R. China followed by Kato-Katz examination of seropositive individuals [6]. Third, rapid diagnostic assays such as dipstick dye immunoassay (DDIA) and dot immunogold filtration assay (DIGFA), were developed by Chinese laboratories for detecting cases infected with S. japonicum [20]–[22]. Although laboratory-based or epidemiological studies have been conducted to investigate the characteristics of immunoassays mentioned above for diagnosis of schistosomiasis [23]–[25], field validation of these assays have not been performed. These evaluations are urgently needed before they can be adopted to support policy decisions of the national programme for the control and elimination of schistosomiasis in P. R. China [26]. In this study, we sought to demonstrate that parasite density is low in settings with low prevalence of infection, making it necessary to switch from microscopy directly to serodiagnostic methods for surveillance. We will also explore, in low transmission intensity settings, the relationship between seropositivity and the prevalence of schistosome infection, as defined by stool examination, and with the intensity of transmission, as defined by egg count per gram of stool. As communities approach elimination, schistosome exposure in children as defined by antibody positivity can be used as surrogate of continued transmission. In this study, we explore the relationship between the rates of seropositivity in children and the prevalence of infection in the village. Our study aims to validate immunodiagnostic tools that can be used to guide selective chemotherapy in moderate or low endemic regions to reduce the number of stool examinations required, to identify individuals for treatment in areas with very low endemicity, and to monitor the high risk population and certify the elimination of schistosomiasis. A two-phase approach was used to assess utility of immunoassays to monitor S. japonicum infection. First the performance characteristics of existing and wild-spread used immunodiagnostic assays were assessed in the laboratory using a well-characterized serum panel of positive and negative samples. Second, tests with acceptable performance in phase one and easy to use, were selected for this study at field locations with different transmission intensity levels of S. japonicum infections. The results of the laboratory based evaluation of nine immunoassays have been reported elsewhere [27]. Three tests, an IHA assay (coded as IHA_JX), one DIGFA test (coded as DIGFA_SH) and an ELISA kit (coded as ELISA_SZ), were selected for the field trial based on their high sensitivity and specificity compared to the enzyme-linked immunoelectrotransfer blot assay (EITB) as a reference standard. All three kits detect serum antibodies against schistosome soluble egg antigen (SEA). The field studies were carried out between October to December, 2008. Before the field work and laboratory examination, experienced staff were selected and then trained through a training course by National Institute of Parasitic Diseases. A manual with standard operation procedure for each reagent was prepared and used for performing the test. 5% serum specimens and Kato-Katz slides were rechecked by national institute for quality assurance. Written informed consents were obtained from all adult participants and the parents or legal guardians of children. Field-based researches were approved by ethics committees of National Institute of Parasitic Diseases, Chinese Center for Disease Control and Prevention. And individuals with positive stool examination results were treated with a single oral dose of praziquantel (40 mg/kg). Field studies were conducted in seven villages, distributed in three provinces in the middle and lower reaches of Yangtz River in P. R. China with different estimated prevalence of schistosome infection based on stool examination. Villages included Caohui, Jingtou and Xinhua located in Jiangxi Province, Longshang, Tieguai and Yuye villages in Anhui province, and Hebei village was administrated by Hubei province. We invited all individuals aged from 6–65 years old to participate in our surveys. Over 50 grams of stool sample and 5 ml blood sample were collected from each participant who gave consent. Specimens were taken to local schistosomiasis stations for testing. Identification and quantification of S. japonicum eggs were carried out by the Kato-Katz thick smear method and the miracidium hatching technique in parallel [10]. Briefly, three slides (41. 7 mg per smear) prepared from a single stool specimen were read 12–48 h after the initial preparation by two qualified technicians in a blinded manner. The number of eggs per slide was counted and recorded. The infection intensity of each individual was recorded as the number of egg per gram faces (EPG), calculated by adding up the number of eggs from all three slides and multiplying by eight to give EPG of stool. The left stool specimens were tested by miracidium hatching technique [10]. Observation was made at 4,8, 12 hours after hatching. The result can be defined as negative if no miracidium was observed. A patient is defined as positive if positive results were obtained with any parasitological technique. The infection intensity of individuals who were tested negative by Kato-Katz method but positive by miracidium hatching technique were regarded as one EPG. And participants were regarded as negative if negative results were obtained with both parasitological methods. Blood samples were maintained in a vertical position for several hours to allow the spontaneous separation of the sera and red blood cells. The sera were removed into clear tubes and sealed. Serum specimens were tested by well trained technicians as following described according to the instructions supplied by the manufactures. Technicians performing the immunoassays were unaware of the results of stool examination and any other immunological test. For ELISA_SZ, all serum specimens were diluted 1∶100 using the dilution liquid and transferred into the kit supplied microtiter wells. The incubation procedure, washing steps and detection steps were carried out according to the manufacturer' s instructions. Absorbance was read at 450 nm zeroed by the reagent blank wells. For each run, positive and negative control sera were measured simultaneously. A positive result was defined as an optical density (OD) value greater than 2. 1 times the OD value of the negative control serum provided by the kit [24]. For IHA_JX, 100 µl of normal saline was placed into the first well of the transverse line, whereas 25 µl was added into wells 2 and 3. Then, 25 µl of serum was added to the first well and thoroughly mixed. 25 µl of mixture was transferred to the second well and mixed as before while 75 µl of mixture in the first well was discarded. 25 µl of mixture in the second well was moved into the third well and mixed followed by discarding 25 µl of them. Thus, the concentrations of serum in the first, second and third wells were 1∶5,1∶10 and 1∶20, respectively. Positive and negative control sera samples provided by the company were tested simultaneously on each plate. 25 µl sensitized red blood cell was placed into each well, shaken, kept at 37°C for 30 min. Observations were made by the naked eye. The titer in the test sera was recorded as one dilution before that which yielded a clear, sharp dark spot similar to that in the negative control wells. Titers were expressed as reciprocal values. Titers of ≥10 indicated a positive result [25]. DIGFA_SH was operated as following four steps: (1) Two drops of washing buffer from the buffer bottle was added to the well on the test box and penetrated the membrane completely; (2) 25 µl of serum was added to the same well and allow it to be absorbed completely; (3) The membrane was washed as step one followed by adding four drops of colloidal gold-labeled soluble egg antigen conjugate from the detecting bottle; (4) After the conjugate solution was absorbed completely, two drops of washing buffer were added to well to remove the unbound conjugate and then the result was read with naked eye immediately. The appearance of two red dots in the well indicated a positive reaction, and the appearance of a single red dot indicates a negative reaction [28]. Only data from subject who gave consent and from whom both stool and blood samples were collected were used for analyses. Using stool examination as a gold reference standard, the prevalence of infection in each community was calculated as the percentage of participants who were stool examination positive among the total number of participants tested in that village. The consistency of results determined by stool examination and any immunoassay was measured by Kappa value. Using stool examination as a reference standard, the sensitivity of each test was calculated as the percentage of participants who were seropositive among those who were positive by stool examination. Seropositivity represents exposure to schistosomes and is an aggregate of current and past infection. School aged children are sentinel population for new infection. The specificity of immunological testing in a subgroup population of 6–9 years old was calculated. Specificity of each immunoassay was expressed as the percentage of individuals who were seronegative among those who were stool examination negative. Furthermore, in low transmission intensity settings, it is useful to calculate the “positive predicative value” (PPV) of screening test by determining the percentage of those who were stool examination positive among those who were seropositive. This index gives a measure to evaluate the screening efficacy of immunoassays of what percentage of the schistosome exposed population is still shedding parasite. Comparisons of index between groups were analyzed by Chi-squared analyses. The relationships between the prevalence and intensity of infection in each village, the antibody positive rate determined by the immunoassays and prevalence of schistosome infection were analyzed by Pearson' s correlation analysis. We judged a P value of less than 0. 05 significant. All analyses were performed with SPSS (Statistical Products & Service Solutions) package for windows (SPSS Inc. , Chicago, USA, version 13. 0). A total of 7,996 people in seven villages voluntarily participated in field survey. 10. 27% (821/7996) of them didn' t offer stool specimens while 6. 64% (531/7996) didn' t offer serum specimens. Of the remaining 6,644 persons, complete data were available on both fecal and serum specimens for 6,248 and were included in the analysis. This population, aged from 6–65 years, with a mean age of 38. 53±17. 76 years, was made up of 48. 13% (3007/6248) men and 51. 87% (3241/6248) women. The type of endemic areas and results of stool examination are shown in Table 1. The prevalence of S. japonicum infection in the seven villages, as defined by stool examination, ranged from 0. 39%–8. 23% with a geometric mean EPG (±S. D.) in the range of 5. 98 (±7. 90) to 45. 04 (±4. 98), indicating that these endemic areas had light infection intensity. Pearson' s correlation coefficients (r) between the prevalence of schistosome infection and geometric mean EPG of positives was 0. 89 (P<0. 001), suggesting that an excellent correlation existed between the prevalence and intensity of schistosome infection at community level (Figure 2). All individuals infected with schistosomes or other parasites were given anthelminthes treatment. The results of three immunoassays compared with those of stool examination based on field survey were shown in Table 2 and Figure S1. The Kappa values of immunological tests compared with stool examination ranged from 0. 08–0. 16, indicating very poor agreement between of immunoassays and parasitological detection (P<0. 001). For the whole population, ELISA_SZ and IHA_JX had sensitivity of 95. 45% (95% CI 92. 94–97. 97%) and 87. 50% (95% CI 83. 51–91. 49%), respectively, which were significantly higher than that of DIGFA_SH (79. 55%, 95% CI 74. 68–84. 41%) (P<0. 05). And also the sensitivity of ELISA_SZ was notable higher than that of IHA_JX (χ2 = 10. 71, P<0. 05). In stratified analysis, the sensitivity of ELISA_SZ and IHA_JX did not differ in age strata except in the group of 20–29 years old while DIGFA_SH presented significant difference in sensitivity among age groups (P<0. 05) (Figure 3A). All assays showed sensitivities equal to or lower than 66. 7% in the 20–29 years old group. These may be caused by the inadequate sample size of patients since there are only three stool positives in this age group and both of them were with EPG no more than 40. ELISA_SZ and IHA_JX also showed stable sensitivity among four intensity categories (P>0. 05) while the sensitivity of DIGFA_SH increased with EPG notably (P<0. 05) (Figure 3B). Chi-square analysis showed that the sensitivity of ELISA_SZ was significantly higher than those of IHA_JX and DIGFA_SH in subgroup with EPG in the range of 1–40 (P<0. 05) while the sensitivity of these three assays didn' t differ in other three infection intensity categories (P>0. 05). Since school-aged children could be used as sentinel population of new infection, the specificity of three immunodiagnostic tests was analyzed in 6–9 years old subgroup compared with stool examination (Figure 3C). ELISA_SZ, IHA_JX and DIGFA_SH showed specificity of 89. 24% (95%CI: 87. 71–90. 77%), 89. 73% (95%CI: 88. 23–91. 23%) and 90. 71% (95%CI: 89. 27–92. 14%) respectively. And no significant difference in specificity was detected among these immunoassays in this subgroup of population (P>0. 05). The screening efficacy of each immunoassay to find the patients still shedding parasites among seropositives for the whole population was evaluated by calculating the PPV (Table 2). DIGFA_SH had the PPV of 13. 07% (95%CI: 11. 42–14. 72%) which were significantly higher than that of other two assays (ELISA_SZ: 8. 13% (95%CI: 7. 17–9. 09%) and IHA_JX: 9. 30% (95%CI: 8. 15–10. 44) with P values less than 0·001 while no difference was detected in infectivity rate between ELISA_SZ and IHA_JX (P>0. 05). Antibody positive rates of whole population determined by ELISA_SZ, IHA_JX and DIGFA_SH were 49. 60% ( (252+2847) /6248), 39. 77% ( (231+2254) /6248) and 25. 72% ( (210+1397) /6248), respectively, which were significantly higher than the prevalence rate of schistosomiasis determined by stool examination (P<0. 05). But the distribution tendencies of positives determined by each immunoassay and stool examination were consistent in age strata for the entire population. Pearson' s correlation coefficient (r) between the prevalence of schistosome infection and antibody positive rates in age strata determined by IHA_JX was detected with r value of 0. 70 (P<0. 001), indicating the significant correlation between of them (Figure 4A). Antibody positive rate determined by IHA_JX in group of 6–9 years old as a sentinel population of schistosome infection also showed a significant correlation with the prevalence of infection amongst the villagers (r = 0. 77, P<0. 05, Figure 4B). The R-values for ELISA_SZ and DIGFA_SH were 0. 44,0. 68 respectively, suggesting a lack of correlation. The control of schistosomiasis was outlined as a series of consecutive phases changing from morbidity control to elimination of schistosome infection [2], [6], [29]. As the goals in different stage of schistosomiasis control varied, the diagnostic approach should be adjusted [26], [29], [30]. In P. R. China, with the ultimate aim to eliminate of schistosomiasis, a national programme with a comprehensive control strategy to reduce transmission had been implemented since 2004 [7], [8], [31]. During the implementation of national programme, appropriate diagnostic tools are important to yield accurate results to guide chemotherapy, monitor the endemic status and evaluate the efficacy of control intervention. It is well known that parasitological diagnosis for schistosomiasis is a poor prognosticator with insensitivity and increased operational time required in areas with low endemicity or intensity of infection [11], [13], [14], [32]. With many advantages over stool examination such as high sensitivity, simple operation and ease of use in the field, immunodiagnostic techniques had been developed and adopted into national programme of schistosomiasis control in P. R. China [6], [16], [18], [19]. Although many immunoassays are used in the field in P. R. China, little is known of serological tests that could be integrated into the elimination strategy, and used in the final stage of schistosomiasis control in P. R. China when prevalence and intensity of schistosome infection is low [3], [5]. In this study, we provided evidence that parasite density decreases as the prevalence of schistosome infection decreases, making it essential to switch to a more sensitive technique than microscopy for guiding treatment, surveillance and monitoring the impact of control interventions [26], [30]. As test-treat is the most cost-effective approach, immunodiagnostic assays was widely used for identifying target for treatment [15]. We found ELISA_SZ was more likely to be positive in individuals who had light infection intensity compared with IHA_JX and DIGFA_SH. This advantage was greater in subgroup with EPG low than 40. Considering results obtained by serologic tests were closer to the real status and sensitivity is given top priority for more efficient coverage with chemotherapy in low endemic areas [17], [29], [33], ELISA_SZ is appropriate as a screening tool for guiding selective chemotherapy with praziquantel in areas where prevalence of schistosome infection is higher than 1% in P. R. China with aim to reduce the number of infected cases, since praziquantel is a safe antischistosomal drug with few side-effects and toxicity [34], [35]. As transmission decreases especially lower than 1%, the treatment strategy needs to transit from selective chemotherapy to individual treatment to avoid over-treatment and praziquantel resistance, since studies have shown that praziquantel resistance against S. mansoni had been induced in lab and presented in humans in Senegal and Egypt [36]–[39]. Definition of individuals for treatment in endemic areas with such low prevalence and infection intensity needs more sensitive and specific diagnosis. Although polymerase chain reaction (PCR) based molecular methods had been reported for their higher sensitivity and specificity [40]–[42], disadvantages including dependence on expensive machines, complex operating procedure and long time reaction time, limit their widespread application for primary health-care settings. A two-step approach is the best choice until now with serology as the primary screening tool followed by stool examinations only for seropositive individuals [16]. The strategy and diagnostics used in the different stages of control in Figure 1 represent the current policy in China. ELISA_SZ assay is superior to other two assays as the screening tool for its high sensitivity and high throughput capacity. In addition, the high variation of geometric mean EPG of infected persons in our studies support that stool examination is inadequate when only one stool sample is detected in low endemic areas especially when the prevalence of schistosome infection was lower than 1% [11], [13]. Multiple stool examination or combination of other diagnostic methods for the positives of serological test would improve its sensitivity for individual treatment [29]. Our study also found that ELISA_SZ can miss some infected individuals especially with the egg count less than 100. This may only stimulate a relatively low antibody response which is not detectable by ELISA. A few individuals with low egg counts may be unlikely to spread disease further in the community. But to target infected individuals for treatment or to eliminate schistosomiasis, more sensitive methods need to be explored and used such as nested PCR. Since specific antibodies against schistosome may last at least two years or even longer time in individuals who had past infection and cured after effective treatment [17], [25], [43], consistency of results determined by any immunological test and stool examination is very poor since the immunoassays also picked up antibodies from old exposures and it is not appropriate to calculate specificity of immunoassays directly by classifying seropositive–stool examination negative specimens as false-positives for the whole population. In our studies, the specificity of three immunoassays was calculated in subgroup of 6–9 years old since they could be regarded as sentinel population indicating new infection [44]. The results showed that these assays all performed high specificity about 90%. Although the antibody positive rate determined by any immunological test was higher than the prevalence of schistosome infection determined by stool examination, we noted that the distribution tendencies of positives determined by immunoassays and stool examination were consistent in age strata for whole population [45]. The strong correlation of antibody positive rate determined by IHA_JX with prevalence of schistosome infection in age strata suggests that IHA_JX, which performed similar sensitivity in laboratory and field settings [27], could be used as a tool for surveillance and epidemiological study to analyze the epidemic characteristics of schistosome infection and estimate the distribution of high-risk infection population in communities [27], [32], [45]. Immunodiagnostic techniques had been shown as effective tools for disease surveillance through testing school-aged children in areas where transmission of schistosome was under control [44], [46]. In our work, the notable correlation between the antibody positive rate in children aged 6–9 years determined by IHA_JX and the prevalence of schistosome infection of villagers support this point and suggest that IHA_JX is an appropriate tool for schistosomiasis surveillance after the transmission of schistosome was under control (prevalence <1%) through measuring antibody response in school-aged children [47], [48]. Further investigation has recently started to explore how to identify the schistosome carriers in community through surveillance of school-aged children in area of low endemicity. Furthermore, in low transmission intensity settings, screening assay which could identify more patients who are still shedding parasites among seropositives would be more efficient than doing stool examination [49], [50]. Here we calculated “PPV” to evaluate the screening efficacy of immunoassays. Our studies showed that DIGFA_SH was the most efficient screening assay among three assays with the “PPV” of 13. 07% (95%CI: 11. 42–14. 72%). And with advantages over other two kinds of assays, such as quickness, easy operation and free of apparatus, DIGFA_SH is more appropriate for application in rural areas with limited resources for schistosomiasis control. Considering its relative low sensitivity especially for low infection intensity in our studies, DIGFA_SH needs to be improved to be more sensitive. This study had several limitations. One limitation in our study is that the villages were not selected in a random manner and only one village with prevalence lower than 1% was included. The performance of immunoassays in other transmission control or transmission interruption areas might be different. The other limitation is that the three immunoassays are antibody-detection based tests which cannot distinguish history of exposure from active infection, it does not allow for estimation of the specificity of each immunoassay since the information of schistosome exposure history is difficult gained accurately. However, these assays could be used for diagnosis in occasionally exposed people such as travelers, migrants etc. , and also could be used for surveillance in low-transmission or post-transmission settings. And also further studies focused on the cost-effectiveness of immunodiagnosis integrated into control strategies needs to be conducted. In conclusion, our findings indicate that, despite good performance in laboratory setting, three immunoassays (ELISA_SZ, IHA_JX and DIGFA_SH) evaluated in our field trials perform differently in areas with different endemicity. The choice of which test to use with the elimination strategy is dependent on the purpose of testing, the endemic status of community and the resources available [29], [30]. ELISA_SZ is more suitable for guiding chemotherapy of schistosome infection in regions with moderate or low endemicity to reduce cases, and could be used as a screening tool for individual treatment followed by sensitive and specific definite diagnosis in areas after transmission is under control, and IHA_JX is easily applied in the risk assessment for epidemiological surveillance in endemic areas and certifying the elimination of schistosomiasis with monitoring the sentinel population while DIGFA_SH is a most efficient screening tools for chemotherapy in low endemic areas but its sensitivity needs further improved.

Immunodiagnostic assays are widely applied in the field to control schistosomiasis in P. R. China as the prevalence and infection intensity of schistosome infections decrease. Field evaluations are urgently needed before they can be adopted to support policy decisions of the national programme for the control and elimination of schistosomiasis in P. R. China. We carried out a large scale cross-sectional survey in field settings with different transmission situations to validate immunodiagnostic tools that can be used to formulate new schistosomiasis elimination strategy in P. R. China. Regarding stool examination as gold reference, the validity and screening efficacy of each immunodiagnostic kit were calculated and compared with each other. The association of the prevalence of schistosomiasis and antibody positive rates determined by immunoassays were analyzed using Pearson' s correlation coefficient values. The study indicates that which test to use with the elimination strategy is dependent on the purpose of testing, the endemic status of community and the resources available. And more sensitive methods need to be explored and used to target infected individuals for treatment or to eliminate schistosomiasis.

Abstract Introduction Materials and Methods Results Discussion

medicine infectious diseases test evaluation schistosomiasis research monitoring diagnostic medicine science policy neglected tropical diseases government laboratories parasitic diseases research assessment research laboratories

2011

Tools to Support Policy Decisions Related to Treatment Strategies and Surveillance of Schistosomiasis Japonica towards Elimination

Fe-S bound proteins are ubiquitous and contribute to most basic cellular processes. A defect in the ISC components catalyzing Fe-S cluster biogenesis leads to drastic phenotypes in both eukaryotes and prokaryotes. In this context, the Frataxin protein (FXN) stands out as an exception. In eukaryotes, a defect in FXN results in severe defects in Fe-S cluster biogenesis, and in humans, this is associated with Friedreich’s ataxia, a neurodegenerative disease. In contrast, prokaryotes deficient in the FXN homolog CyaY are fully viable, despite the clear involvement of CyaY in ISC-catalyzed Fe-S cluster formation. The molecular basis of the differing importance in the contribution of FXN remains enigmatic. Here, we have demonstrated that a single mutation in the scaffold protein IscU rendered E. coli viability strictly dependent upon a functional CyaY. Remarkably, this mutation changed an Ile residue, conserved in prokaryotes at position 108, into a Met residue, conserved in eukaryotes. We found that in the double mutant IscUIM ΔcyaY, the ISC pathway was completely abolished, becoming equivalent to the ΔiscU deletion strain and recapitulating the drastic phenotype caused by FXN deletion in eukaryotes. Biochemical analyses of the “eukaryotic-like” IscUIM scaffold revealed that it exhibited a reduced capacity to form Fe-S clusters. Finally, bioinformatic studies of prokaryotic IscU proteins allowed us to trace back the source of FXN-dependency as it occurs in present-day eukaryotes. We propose an evolutionary scenario in which the current mitochondrial Isu proteins originated from the IscUIM version present in the ancestor of the Rickettsiae. Subsequent acquisition of SUF, the second Fe-S cluster biogenesis system, in bacteria, was accompanied by diminished contribution of CyaY in prokaryotic Fe-S cluster biogenesis, and increased tolerance to change in the amino acid present at the 108th position of the scaffold. Fe-S bound proteins are ubiquitous and involved in a wide variety of cellular processes such as respiration, regulation of gene expression and central metabolism [1,2]. Maturation of Fe-S proteins is an essential cellular process for both eukaryotic and prokaryotic organisms. The mitochondrial ISC Fe-S biogenesis machinery has been proposed to be inherited from a bacterial ancestor, and they function in a similar way by utilizing two major steps: (i) an assembly step in which the cluster forms transiently on a scaffold protein, and (ii) a delivery step in which the cluster is transferred to apotargets via dedicated carriers [3–5]. The ISC scaffold (Isu for eukaryotes / IscU for prokaryotes) contains three conserved cysteine residues that are essential for Fe-S cluster binding and a conserved motif that is specifically recognized by DnaKJ related chaperones/co-chaperones to facilitate cluster release [6–10]. Sulfur is produced from L-cysteine by the cysteine desulfurase, (Nfs1 for eukaryotes / IscS for prokaryotes) a pyridoxal-5’-phosphate (PLP) -dependent enzyme [11–15]. The sulfur is bound in the form of a persulfide to an active-site cysteine residue of the cysteine desulfurase and is subsequently transferred to the scaffold [15–18]. Frataxin (FXN in human, Yfh1 in yeast and CyaY in bacteria) is a protein present in mammals, plants and bacteria [19]. FXN interacts with the cysteine desulfurase/scaffold complex [20–26]. In both prokaryotes and eukaryotes, deficiency of core ISC components including the ISC scaffold or cysteine desulfurase is associated with severely defective Fe-S cluster biogenesis that translates into drastic phenotypes [12–14,26–29]. In contrast, the consequences resulting from deficiency in FXN differ in eukaryotes or prokaryotes. In yeast, deficiency in frataxin (Yfh1) results in defective growth, mitochondrial iron accumulation, decreased heme synthesis, loss of Fe-S cluster protein activity and hypersensitivity to oxidants [30–34]. In humans, altered levels of FXN lead to a drastic decrease in Fe-S protein activities and cause the neurodegenerative disease Friedreich’s ataxia [35–39]. We and others recently established the participation of E. coli frataxin (CyaY) in ISC-assisted biogenesis of Fe-S clusters. Accordingly, ΔcyaY mutants exhibit pleiotropic but mild phenotypes [40–44]. Both the physiological advantage and the molecular reasons underlying this apparent loss in importance of CyaY in prokaryotes remain obscure. Recently, in their analysis in Saccharomyces cerevisiae, the Dancis lab reported that a point mutation in the scaffold protein Isu1 could bypass a Yfh1 deletion [45]. This demonstrated that a single mutation could make survival of a eukaryote independent of FXN. In the present study, we investigated whether the reverse was true, i. e. could E. coli be turned into a CyaY (FXN) -dependent organism. This proved to be possible and required a single amino acid change in the IscU scaffold as well. Genetic, physiological, biochemical, bioinformatic and phylogenomic approaches were carried out to characterize this E. coli variant. The results of these studies led us to propose an evolutionary scenario according to which frataxin is an ISC-associated factor that appeared in Proteobacteria. It was then acquired by eukaryotes via endosymbiotic mitochondrial event where it became essential. Meanwhile its importance diminished in bacteria possibly because these later contained other Fe-S cluster biogenesis systems, such as SUF in many instances. The studies by the Dancis group revealed that the contribution of frataxin to Fe-S cluster biogenesis might depend on the identity of the residue present at position 108 in IscU [45]. To test this hypothesis, we exchanged the 108th ATT Ile codon with an ATG Met codon in the iscU sequence, and the cognate iscUI108M allele was introduced into the E. coli chromosome, giving rise to the BR755 (iscUIM) strain. Moreover, in order to make this strain more eukaryotic-like, we deleted the suf operon encoding the second E. coli Fe-S cluster biogenesis system, giving rise to the BR763 (iscUIM Δsuf) strain. Growth of the BR763 strain in LB or in minimal M9 medium was similar to the reference strain DV901 (Fig 1A and 1B). In contrast, introduction of the cyaY deletion in the iscUIM Δsuf strain had a drastic negative impact on growth in glucose M9 minimal medium (Fig 1B). To test whether the growth defect of the iscUIM Δsuf ΔcyaY strain in minimal medium was related to defects in Fe-S proteins, we tested whether it was auxotrophic for the amino acids Ile, Leu, and Val whose synthesis depends on Fe-S enzymes. Addition of all of the 20 amino acids restored growth, whereas omitting Ile, Leu, and Val failed to rescue growth (Fig 1C). However, adding only Ile, Leu, and Val failed to restore growth showing that Ile, Leu, and Val were necessary but not sufficient. Adding Cys and Met in addition to Ile, Leu, and Val, did not rescue growth of the iscUIM Δsuf ΔcyaY indicating that other processes must also be impaired in this strain. Addition of vitamins improved marginally growth of the iscUIM Δsuf ΔcyaY strain (S1 Fig). In rich medium, the growth defect of the iscUIM Δsuf ΔcyaY strain indicated that, in addition to nutritional requirements, this strain was also impaired in other processes (Fig 1A). A second assay measuring killing efficiency by aminoglycosides (gentamicin, Gm, and kanamycin, Kan) was used. This assay is an indirect read-out of ISC-mediated Fe-S cluster biogenesis efficiency but is independent of SUF functioning. Indeed, uptake of aminoglycosides is dependent upon proton motive force (p. m. f) at the cytoplasmic membrane, which depends upon the activity of Nuo (also called Complex I), a multi-protein complex containing 9 Fe-S clusters, whose maturation depends predominantly on the ISC system and only marginally on the SUF system [46]. The iscUIM strain was found to exhibit wild-type sensitivity to Gm and Kan, whereas the ΔcyaY derivative iscUIM ΔcyaY showed enhanced resistance, again suggesting that ISC dependent Fe-S cluster biogenesis was compromised in the absence of CyaY in this background (Fig 2). As a matter of fact, the iscUIM ΔcyaY strain exhibited a level of resistance similar to that of a ΔiscU strain, illustrating the important contribution of CyaY in a background using a eukaryotic-like IscUIM scaffold. In contrast, the wt strain remained sensitive to Gm and Kan whether or not CyaY was present (Fig 2). Because frataxin deficiency, in yeast, led to hypersensitivity to oxidants, we also tested the importance of CyaY in the “eukaryotized” background. Fig 3 shows that introduction of the cyaY deletion in the iscUIM Δsuf strain led to hypersensitivity to hydrogen peroxide and to paraquat, a superoxide generator. Altogether, these results indicate that an E. coli lacking SUF can be turned into a frataxin-dependent organism simply by changing a single residue in the IscU scaffold. In order to ascertain that the drastic defects observed in the iscUIM ΔcyaY strain were directly due to a dysfunction of Fe-S cluster biogenesis, we tested the activity of several Fe-S cluster-containing proteins. These latter were IscR, a [2Fe-2S] transcriptional regulator, Nuo and Sdh, two multi-protein complexes containing 9 and 3 Fe-S clusters, respectively. In full agreement with the phenotypic tests reported above, introduction of a ΔcyaY mutation in a strain synthesizing the eukaryote-like IscUIM scaffold essentially recapitulated the effect of deleting the scaffold-encoding gene iscU (Fig 4A, 4B and 4C). As a point of comparison, in the iscUIM strain, the IscR, Nuo and Sdh activities were decreased by 1. 5–2 fold when compared to the wt strain (Fig 4A, 4B and 4C). Immunoblot analysis of the IscR, Nuo and IscUIM proteins ruled out that the decreased activities were due to reduced amounts of target or scaffold proteins (Figs 4D and S2). Altogether, these results indicate that even though it is a conservative change, a single Ile-to-Met substitution in the IscU scaffold alters Fe-S biogenesis efficiency. In order to understand the molecular basis for the effect caused by the mutation, the IscUIM protein was submitted to a thorough in vitro analysis. A plasmid encoding a His-tagged IscUIM was constructed, and the tagged protein was purified in large quantities. The CD spectra of IscUIM and IscUWT were similar, indicating that the mutation did not affect the secondary structure of the protein (Fig 5A). Also gel filtration experiments indicated that the IscUIM formed dimers like the IscUWT (S3 Fig). The IscUWT was previously shown to be isolated from complexes together with IscS and IscS-CyaY [23,24]. Therefore, we investigated whether the IscUIM had similar behavior. To this purpose an anion exchange chromatography approach was used. Purified reconstituted IscUWT or IscUIM was mixed anaerobically with molar stoichiometric amount of IscS and CyaY proteins. The mixtures were loaded onto an anion exchange column (QFF), and the collected fractions were analysed by SDS-PAGE. Using IscUWT, a first peak (peak 1), containing IscU and IscS, eluted at 640 mM NaCl while a second major peak (peak 2), which eluted at 780 mM NaCl contained the IscS, IscU and CyaY proteins (Fig 5B left panel). The proteins recovered in peak 1 and peak 2 were part of a complex, since each individual protein, IscUWT, IscS and CyaY eluted from the column at 400,430 and 530 mM NaCl, respectively (S4 Fig). A similar result was obtained when using IscUIM instead of IscUWT (Figs 5B right panel and S4). Thus, these data show that the ability of IscU to associate with IscS and CyaY was not altered by the Ile-to-Met mutation. Lastly, we investigated whether IscUIM could assemble a [2Fe-2S] cluster. Fig 5C shows that after anaerobic Fe-S cluster reconstitution, IscUWT and IscUIM displayed similar UV-vis. spectra characteristic of [2Fe-2S] clusters, with absorption maxima at 320,410 and 456 nm (Fig 5C) [47–49]. However, the rate of Fe-S cluster formation differed between the two. Indeed the rate of Fe-S cluster formation was slowed down by approximately 2-fold when using the IscUIM mutant (Fig 5D). Altogether, these biochemical investigations revealed that the Ile-to-Met mutation specifically altered the efficiency of Fe-S cluster formation on IscUIM, with no major effect on the structure of IscUIM or its capacity to interact with its partners IscS and CyaY. CyaY contains a single domain of ~100 residues referred to as PF01491 in the Pfam database. By using this domain as a query, we detected 598 homologous proteins within 2742 complete prokaryotic genomes available in the local bank of complete genomes (2 March, 2014) (S1 Table). Homologs of CyaY were found in Alpha-, Beta-, Gammaproteobacteria, Acidobacteria and Deltaproteobacteria species, and in one representative of Chlorobi phylum (Chloroherpeton thalassium ATCC 35110) (Fig 6). These data indicate that CyaY is not widely distributed among prokaryotes. The absence of a CyaY encoding gene in the ancestor of most bacterial phyla suggests that a CyaY encoding gene was absent in LBCA. The phylogenetic analysis of CyaY also showed that the representatives of Chlorobi and Acidobacteria phylum, which emerge within the Gammaproteobacteria, have probably acquired cyaY gene by HGT (dotted black arrows) (Figs 6 and S5). Altogether, these results suggest that the CyaY protein originated in the bacterial domain, likely in the common ancestor of the Proteobacteria with massive loss in Delta/Epsilonproteobacteria subdivision. IscU homologs were retrieved using the PF01592 domain and were aligned using the multiple alignment program MAFFT v7. 045b (S1 Table). We imposed some additional criteria in order for a protein to be considered as an IscU homolog: (i) the presence of the three conserved cysteine residues that are required for the scaffold activity of IscU, (ii) the presence of the sequence that is recognized by the chaperone/co-chaperone system of the ISC system (LPPVK in E. coli IscU) (iii) no other additional domain such as those that could be found in NifU, and (iv) at least one other isc-related gene as a neighbor gene. Using these criteria, well studied U-like proteins such as the SufU protein of Bacillus subtilis and the NifU protein of Azotobacter vinelandii and their close homologs were eliminated. We then showed that all the prokaryotic species that possessed CyaY also contained an IscU homolog. However, the reverse situation was not true, since numerous prokaryotic species possessing IscU did not contain a CyaY encoding gene (Fig 6). Sequence alignment of the 429 prokaryotic IscU homologs showed that amino acids at position 108 were mostly (369/429) Ile, Leu or Val (Figs 6 and S6). A few IscU examples exhibited a Met or Asn amino acid. Interestingly, an IscUM protein was found in 3 out of 28 species of the Rickettsiales order (Alphaproteobacteria) (Orientia tsutsugamushi str. Boryong, YP_001248706; Neorickettsia risticii str. Illinois, YP_003081518 and Neorickettsia sennetsu str. Miyayama, YP_506192). An IscUM protein was also found in two Archaea species (2 out of 30 species) Methanosarcina barkeri str. Fusaro (YP_305925), and Methanosarcina acetivorans C2A (NP_617616). ISC machineries from both eukaryotes and prokaryotes are considered to be homologous. They share many components including cysteine desulfurases, scaffolds, dedicated-chaperone proteins and A-type carriers. A defect in any of these conserved components provokes a drastic drop in Fe-S cluster biogenesis in either eukaryotes or prokaryotes [12–14,26–29,50,51]. The case of frataxin is different, however, as a lack of FXN in eukaryotes, humans or yeast, is markedly more detrimental than a lack of CyaY in prokaryotes such as E. coli or Salmonella [30–44]. A possible explanation for the difference is that variation within the Fe-S cluster assembly machineries provides different contexts, which in turn make the contribution of FXN of greater importance than that of CyaY. In this regard, it is important to recall that the core eukaryotic ISC system includes a component, Isd11, which interacts with Nfs1 [52–54]. A model was recently proposed according to which the eukaryotic Nfs1 cysteine desulfurase remains in an OFF state unless it interacts with FXN and Isd11 [55–58]. However, no Isd11-like proteins are present in E. coli and this regulation of IscS activity does not apply [52,53]. Recent genetic analysis by the Dancis group showed that modifying part of the ISC machinery could render it independent of FXN. Indeed, in a search for suppressing mutation that could bypass the lack of FXN, these authors identified a mutation in the scaffold-encoding gene ISU1 [45]. The suppressing mutation allowed Isu1 to activate Nfs1, thereby mimicking FXN [55,59]. Remarkably, this mutation changed a Met residue, conserved in eukaryotes, to an Ile residue, conserved into prokaryote IscU proteins. Although largely speculative, this result may open the way to deciphering the contribution of frataxin in the functioning of ISC machineries, and possibly provide a lead towards understanding the differences between prokaryotes and eukaryotes. In the present work, we carried out a bioinformatic analysis of IscU sequences in prokaryotes. This allowed us to confirm that position 108 was mostly occupied by Ile, as in E. coli, Leu or Val. By contrast, position 108 in prokaryotes was almost never occupied by Met (see below for an exception), which is the situation most frequently encountered in eukaryotes. In an effort to address the importance of this residue experimentally, the Ile residue was changed to Met at position 108 of IscU and expressed into E. coli lacking SUF. The results confirmed the influential role of that position. First, the E. coli strain containing a eukaryotic-like IscUIM became fully dependent on CyaY. Thus, this strain was unable to mature a series of Fe-S cluster containing proteins such the transcriptional regulator IscR, a [2Fe-2S] protein, or Nuo and Sdh, multi-cluster containing enzymes of the electron transport chain. Moreover, such a strain became auxotrophic for various amino acids, including Ile, Leu and Val, the branched amino acids whose synthesis depends on the Fe-S cluster containing proteins, dihydroxy-acid dehydratase (IlvD) and isopropylmalate dehydratase (LeuD). In addition, the strain showed hypersensitivity to oxidative stress, a phenotype linked to FXN deficiency in eukaryotes [39,60,61]. Therefore, a single Ile-to-Met substitution was sufficient to turn E. coli into a frataxin-dependent organism for Fe-S cluster biogenesis. How could a single conservative Ile-to-Met change have such a crucial impact on Fe-S cluster biogenesis? A hypothesis is that the Ile-to-Met mutation alters the IscU protein, diminishing its efficiency in contributing to the overall Fe-S cluster biogenesis process and that in this context, the contribution of CyaY becomes essential. To test this hypothesis, we carried out a thorough biochemical characterization of the IscUIM variant and could rule out structural or stability defects. This fits with the in vivo observation that the IscUIM protein was as abundant as the wt protein and failed to exhibit instability as assessed by immunoblot analysis. Moreover, we observed that the IscUIM protein interacted with its natural partners, IscS and CyaY, in a mode indistinguishable from the wild type. In contrast, in vitro, IscUIM was found to assemble Fe-S clusters at a rate 2-fold slower than the wild type. Interestingly, these data are consistent with the in vivo observation that E. coli containing a chromosomal copy of the mutated iscU allele was 2-fold less efficient in maturing IscR than the wt strain. Hence, altogether these results support the notion that the Ile-to-Met mutation altered the kinetic formation of Fe-S clusters on IscU. A possible structural explanation for the effect might be that the mutation modifies the accessibility of sulfur or iron for Fe-S cluster intermediate formation, as the 108th position is in close vicinity to the Cys106, one of the three Cys residues acting as ligands. Regardless of the structural basis for this effect, the fact is that this analysis revealed that a eukaryotic-like IscU is slightly less efficient than the E. coli one in assembling a cluster and as a consequence, the contribution of frataxin becomes more significant for helping the process to go on. As previously shown, we observed that via its interaction with IscS, CyaY slowed down the kinetic of Fe-S cluster formation on IscU (S7 Fig) [22,62]. At first this contradicts the view of CyaY acting as a positive effector for Fe-S cluster formation and this has already been discussed at length in the literature [22,62,63]. But what matters here is that CyaY also inhibited, and to the same extent, Fe-S cluster assembly by the IscUIM variant (S7 Fig). This indicated that the CyaY action is not strictly connected to the nature of the residue at position 108. Thus one possibility is that CyaY and IscUIM influence the overall Fe-S cluster biogenesis process at different steps. The fact that the CyaY dependency is not bypassed by increasing amount of the IscUIM scaffold, as indicated in vivo by the CyaY-dependent maturation of IscR when IscUIM was overproduced (S8 Fig), is consistent with this hypothesis. Further biochemical analyses are needed to investigate the possible sites of action in the Fe-S cluster assembly process, such as iron donation, control of sulfur flux, Fe-S cluster transfer to downstream recipients, or HscBA-associated steps, for the CyaY and IscUIM effect. The involvement of CyaY in Fe-S cluster biogenesis was proposed in the early 2000’s on the basis of co-occurrence of cyaY and hscBA genes [64]. This led to the belief that CyaY would be as conserved as the other ISC components. The reason why the cognate structural gene was not part of the isc operon in bacteria remained unclear. Here, exploiting the larger number of genomes now available for analysis, we reinvestigated the distribution of CyaY and its co-occurrence with the ISC system. Surprisingly, CyaY was found to be much less conserved in eubacteria than previously thought, as its presence was mostly restricted to Alpha-, Beta-, and Gammaproteobacteria. Interestingly, in these bacteria, none of the genes encoding components related to Fe-S cluster biogenesis were to be found in the vicinity of cyaY. Phylogenic analysis revealed that CyaY originated in the last common ancestor of Proteobacteria. This contrasts with the story for A-type Fe-S cluster carriers, which we previously found to be present in the last bacterial common ancestor [65,66]. Even more surprising was the fact that many genomes contained iscU but not cyaY, suggesting that these bacteria learned how to make Fe-S clusters in an ISC-dependent and CyaY-independent way. In contrast, all genomes containing cyaY also contained iscU. Hence overall this leads to picture CyaY as a Fe-S cluster biogenesis factor associated with the ISC machinery in most eukaryotes and in a restricted number of prokaryotes. Interestingly, not only some lineages such as Deltaproteobacteria, but also some species within the Alpha- and Betaproteobacteria have lost CyaY, indicating that there might have been some evolutionary drift favoring organisms that evolve without it. Amino acids encoded by the codon at position 108 of IscU are essentially Ile, Leu or Val. Methionine appears in only two cases, in Methanobacteria and some Rickettsiae species that also have a cyaY gene. Rickettsiae are thought to have given rise to mitochondria via the first endosymbiosis event. Hence, it is tempting to speculate that the current mitochondrial Isu protein originated from the IscUM version that was already present in the ancestor of Rickettsiae. Based upon the above considerations, one can envision the following scenario: i) Frataxin appeared in the ancestor of the Proteobacteria, and joined the ISC system for Fe-S cluster biogenesis, ii) Mitochondria developed from Proteobacteria by endosymbiosis, in particular from Rickettsiae, acquiring components what would give rise to the actual IsuM and FXN, iii) Proteobacteria acquired SUF, which released the pressure on ISC, and in parallel they explored variation in the ISC scaffold at position 108. In particular, the Met-to Ile, Leu, Val changes happened to improve Fe-S cluster assembly, iv) Frataxin dependency was loosened in Proteobacteria that have a more efficient ISC scaffold and other Fe-S back up system. The E. coli K-12 strain MG1655 and its derivatives used in this study are listed in Table 1. Deletion mutations from the KEIO collection were introduced by P1 transduction [67]. Transductants were verified by PCR, using primer pairs hybridizing upstream and downstream of the deleted gene. Strain BR755 producing the IscUI108M variant from a chromosomal copy was constructed as follows: a DNA fragment carrying the iscUI108M allele was obtained after a mutagenesis procedure by overlap extension PCR reactions using the following primer pairs: IscU-UPBamH1/IscUI108M-DO, IscUI108M-UP/IscU-DOXbaI, IscU-UPBamH1/IscU-DOXbaI (S2 Table). This DNA fragment was introduced in a strain in which the iscU gene had been replaced by a cat-sacB cassette as previously described [68]. The Suc-resistant clones were checked for Cm sensitivity, and the appropriate region was sequenced. The iscUI108M allele was transduced into desired strains by using a KanR-linked marker in the yphD gene, which is located close to the iscU gene. The Δsuf mutation was introduced in the iscUIM background strains that contained the eukaryotic Fe-S cluster independent mevalonate pathway for IPP biosynthesis (MVA), in case the combination of iscUIM ΔcyaY would have been lethal [50,51]. This precaution proved to be unnecessary since the iscUIM Δsuf and iscUIM Δsuf ΔcyaY strains could be obtained without the addition of arabinose and mevalonate. Addition of arabinose and mevalonate did not improve growth of the iscUIM Δsuf and iscUIM Δsuf ΔcyaY strains; therefore, all the experiments have been done without. However, the iscUIM Δsuf and iscUIM Δsuf ΔcyaY strains are auxotroph for tryptophan since the MVA synthetic operon was inserted in the trp operon, therefore when grown in M9 glucose minimal medium tryptophan was added [50]. Oligonucleotides used in this study are listed in S2 Table. Supplementary strains are listed in S3 Table. E. coli strains were grown at 37°C in Luria—Bertani (LB) rich medium or in minimal medium (M9) supplemented with glucose (0. 4%) and MgSO4 (1 mM). Arabinose (0. 2%), amino acids (0. 5 mM), sucrose (5%), thiamine (0. 2 μg/mL) and nicotinic acid (12. 5 μg/mL) were added as required. Solid media contained 1. 5% agar. Antibiotics were used at the following concentrations: chloramphenicol 25 μg/mL, kanamycin 30 μg/mL, tetracycline 25 μg/mL, gentamicin 5 μg/mL and ampicillin 50 μg/mL. Plasmid pIscU was constructed by PCR amplification of the coding region of iscU from E. coli MG1655 chromosomal DNA using the following primer pair: NcoI-IscU/HindIII-IscU (S2 Table). The PCR product was then digested by NcoI and HindIII and cloned into the NcoI/HindIII linearized pBAD24 vector. Production of the IscUIM variant exhibiting a single amino acid substitution isoleucine to methionine at position 108 was obtained by site-directed mutagenesis in the pIscU plasmid to generate pIscUIM using the following primer pair: IscUI108M_for/IscUI108M_rev (S2 Table). Plasmids pETIscUWT and pETIscUIM were constructed by PCR amplification of the coding region of iscU from E. coli MG1655 chromosomal DNA and from the pIscUIM vector, respectively, using the following primer pair: NdeI-IscU/HindIII-IscU (S2 Table). The PCR products were then digested by NdeI and HindIII and cloned into the NdeI/HindIII linearized pET21a+ vector. Plasmids pET22b-CyaY and pQE-IscS for production of recombinant E. coli CyaY and IscS, were described previously [69]. Overnight cultures were diluted and grown aerobically in LB at 37°C to an OD600 of 0. 2. At this point, antibiotics were added to the cells (Gm at 5 μg/mL and Kan at 10 μg/mL). At different incubation times, 100 μL of cells were diluted in PBS buffer, spotted on LB agar and then incubated at 37°C overnight. Cell survival was determined by counting colony-forming units per mL (CFU/mL). The absolute CFU at time-point 0 (used as the 100%) was ≈ 5x107 CFU/mL. Overnight cultures were diluted in sterile PBS and 5 μL were directly spotted onto LB plates containing either paraquat (250 μM) or H202 (1 mM). The plates were incubated overnight at 37°C before growth was scored. Strains were grown at 37°C in LB rich medium, to an OD600 of ~1. 5. β-galactosidase assays were carried out as previously described [70]. Equal quantities of protein were applied to SDS-PAGE and transferred onto nitrocellulose membranes. The membrane filters were incubated with appropriate antibodies (1/200,1/2000,1/2000,1/150 dilutions of the anti-IscU, anti-NuoF, anti-NuoC and anti-IscR serums, respectively). Immunoblots were developed by using horseradish peroxidase-conjugated goat anti-rabbit antibody, followed by chemiluminescence detection. Recombinant CyaY, IscUWT and IscUIM proteins containing a C-terminal His6 tag were expressed in E. coli and purified as follows: E. coli BL21 (DE3) /pETcyaY was grown in LB medium containing 50 μg/mL ampicillin at 37°C. Protein expression was induced for 4 h by the addition of 0. 5 mM isopropyl β-D-thiogalactoside (IPTG) at an OD600 ≈ 0. 5. The bacterial pellet was resuspended in buffer A (0. 1 M Tris-HCl, pH 8,500 mM NaCl, 20 mM imidazole) and disrupted in a French press. After centrifugation (15 min, 11 000 rpm, 4°C), the supernatant was loaded onto a 1-mL HisTrap affinity column (GE Healthcare) equilibrated with buffer A. Proteins were eluted with a gradient of buffer A containing 500 mM imidazole. Protein-containing fractions were desalted with a Nap-25 column (Amersham Biosciences) and then concentrated. A similar procedure was used to purify IscUWT and IscUIM proteins except that protein expression was induced by the addition of 1 mM IPTG. Recombinant E. coli IscS containing an N-terminal His6 tag was expressed and purified as previously described [74]. The protein concentration was estimated by measuring the absorbance at 280 nm with the NanoDrop2000 spectrophotometer and using the calculated molar extinction coefficient. CD spectra were recorded on a Jasco J-815 spectropolarimeter by using Hellma 110-QS cuvettes of 1 mm path length. CD measurements were performed in 50 mM Tris-HCl pH 8,50 mM NaCl using protein concentrations of 2 μM. 20 scans were averaged and the buffer baseline was subtracted. The purified His-tagged IscUWT and IscUIM proteins were obtained in the apo-form. The purified proteins were reconstituted anaerobically in a glove box as described previously [48]. Briefly, 144 μM protein was mixed with 5 mM DTT, 1. 44 μM IscS, 2 mM L-cysteine and 0. 43 mM Fe (SO4) 2 (NH4) 2 in a total volume of 500 μL of buffer A (50 mM Tris-HCl pH 8). Formation of Fe-S clusters on IscU was followed by UV-visible absorption spectroscopy using a Cary 1 Bio spectrophotometer. After 3 h incubation, samples were loaded onto a 1-mL anion exchange column (QFF) (GE Healthcare) equilibrated with buffer A and eluted with a gradient of buffer A containing 1 M NaCl. Protein fractions were concentrated on a Microcon concentrator (Amicon) and each concentrate was analysed for its Fe content, and for its UV-visible spectrum. Purified His-tagged IscUWT or IscUIM, IscS and CyaY proteins were mixed anaerobically in a 1: 1: 1 ratio (144 μM of each protein) for 40 minutes with 4-fold excess of Fe (SO4) 2 (NH4) 2,10-fold excess of L-cysteine and 5 mM DTT in a total volume of 500 μL of buffer A (50 mM Tris-HCl pH 8). The mixture was loaded onto a 1-mL QFF column (GE Healthcare) equilibrated with buffer A and eluted with a gradient of buffer A containing 1 M NaCl. Proteins elution was visualized by SDS-PAGE. To assess kinetics of cluster formation on IscUWT or IscUIM, absorbance at 420 nm was measured as a function of time. 25 μM IscUWT or IscUIM was incubated anaerobically with 100 μM Fe (SO4) 2 (NH4) 2,2 mM DTT in 50 mM Tris-HCl pH 8. Subsequently, 25 μM IscS and 250 μM L-cysteine were added to start the reaction. The 2742 complete prokaryotic proteome (2591 bacterial and 151 archaeal) available at the NCBI in March 03,2014 were downloaded (ftp: //ftp. ncbi. nlm. nih. gov/genomes/). The HMMER package v3. 0b2 and self-written scripts were then used to search for CyaY homologs in these complete genomes, requiring the presence of Frataxin-like domain (PFAM accession number PF01491) [75]. Alignments E-value with the 599 profile less than 0. 1 were considered as significant. To retrieve CyaY sequence we imposed homology with the entire CyaY sequence and an E-value with 1. 7e-7 as threshold. In addition, alignments have been visually inspected. Proteins of the YjbR family, such as YdhG from Bacillus subtilis have not been detected since despite their structural similarity with CyaY they lack similarity at the sequence level [76,77]. The corresponding sequences were subsequently analysed with the same software in order to determine the presence of additional known functional domains. Additional BLASTP/tBLASTN searches were performed in complete genomes to ensure that the CyaY family was exhaustively sampled and in the nr database at the NCBI to retrieve eukaryotic sequences [78]. For each homolog, the gene context, defined as the 5 neighboring genes located upstream and downstream, was investigated using MGcV (Microbial Genomic context Viewer) [79]. The retrieved homologous sequences were aligned using MAFFT v7. 045b [80]. The best resulting alignment was then visually inspected and manually refined using ED program from the MUST package [81]. The regions in a multiple sequence alignment that were suited for phylogenetic inference were selected by using BMGE (BLOSUM30 similarity matrix) [82]. The phylogeny of all the prokaryotic CyaY was constructed using both maximum likelihood (ML) and Bayesien methods. ML analyses were run using PHYML version 3. 1 with the Le and Gascuel (LG) model (amino acid frequencies estimated from the dataset) and a gamma distribution (4 discrete categories of sites and an estimated alpha parameter) to take into account evolutionary rate variations across sites [80]. The robustness of each branch was estimated by the non-parametric bootstrap procedure implemented in PhyML (100 replicates of the original dataset with the same parameters). Bayesian analyses were performed using MrBayes version 3. 2. 2 with a mixed model of amino acid substitution including a gamma distribution (4 discrete categories) and an estimated proportion of invariant sites [83]. MrBayes was run with four chains for 1 million generations and trees were sampled every 100 generations. To construct the consensus tree, the first 1500 trees were discarded as ‘‘burnin”. For the dataset construction IscU homologs was retrieved from complete proteome available in the local databank (see above) using BLASTP. The distinction between homologous and non-homologous sequences was assessed by visual inspection of each BLASTP outputs (no arbitrary cut-off on the E-value or score). We imposed some additional criterion in order for a protein to be considered as an IscU homologs: the presence of the three conserved cysteine residues that are required for the scaffold activity of IscU, no other additional domain such as those that could be found in NifU, and at least one other isc-related gene as a neighbor gene. The IscU homologs were gathered in a dataset and the corresponding sequences were aligned using MAFFT v7. 045b [80]. Sequence-logo of IscU alignment was generated using Phylo-mLogo visualization tool in order to highlight the LPPVK motif and residues in position 108 [84]. Additional materials and methods are mentioned in S1 Text.

Iron sulfur (Fe-S) clusters are ubiquitous cofactors found in proteins which function in very diverse pathways ranging from respiration to DNA repair. The mitochondrial Fe-S biogenesis machinery ISC was inherited from the bacterial ancestor of mitochondria. In both prokaryotes and eukaryotes, deficiency of core ISC components is associated with drastic decrease in Fe-S proteins activities and causes severe phenotypes. In this context, the case of frataxin, an ISC associated component, is surprising since the lack of frataxin in prokaryotes leads to very mild phenotypes in comparison to eukaryotes. Here, we showed that in an E. coli strain, a single mutation in a key component of the Fe-S cluster biogenesis pathway, namely the scaffold protein, was sufficient to impose a strict frataxin dependency. Remarkably, this mutation substituted an Ile residue that is conserved in prokaryotic scaffolds, for one Met residue that is conserved in eukaryotic scaffolds. These results provide a lead towards understanding the differences between otherwise highly related prokaryotic and eukaryotic ISC Fe-S cluster biogenesis machineries, and provide a new entry point into deciphering the molecular role of frataxin.

Abstract Introduction Results Discussion Materials and Methods

2015

Turning Escherichia coli into a Frataxin-Dependent Organism

Organisms evolve two routes to surviving infections—they can resist pathogen growth (resistance) and they can endure the pathogenesis of infection (tolerance). The sum of these two properties together defines the defensive capabilities of the host. Typically, studies of animal defenses focus on either understanding resistance or, to a lesser extent, tolerance mechanisms, thus providing little understanding of the relationship between these two mechanisms. We suggest there are nine possible pairwise permutations of these traits, assuming they can increase, decrease, or remain unchanged in an independent manner. Here we show that by making a single mutation in the gene encoding a protease, CG3066, active in the melanization cascade in Drosophila melanogaster, we observe the full spectrum of changes; these mutant flies show increases and decreases in their resistance and tolerance properties when challenged with a variety of pathogens. This result implicates melanization in fighting microbial infections and shows that an immune response can affect both resistance and tolerance to infections in microbe-dependent ways. The fly is often described as having an unsophisticated and stereotypical immune response where single mutations cause simple binary changes in immunity. We report a level of complexity in the fly' s immune response that has strong ecological implications. We suggest that immune responses are highly tuned by evolution, since selection for defenses that alter resistance against one pathogen may change both resistance and tolerance to other pathogens. Evolutionary theory suggests that a host can protect itself against an infectious pathological threat by evolving two different mechanisms to increase fitness or health. The first is to reduce the fitness of the pathogen, thereby reducing the number of pathogens attacking the host. The second is to limit the health costs to the host. The sum of both these mechanisms defines an individual host' s defensive capabilities. In the plant ecology community, these two mechanisms are defined as resistance and tolerance [1–5]. Dividing the physiological response to infections into these two components is important because it demonstrates that the health of a host cannot be measured solely by its ability to survive an infection and that studying both pathogen clearance and pathology is essential to fully understanding the defensive measures of a host. Typically in animal immunity studies we focus on understanding resistance mechanisms. For example, most work on Drosophila immunity to date has concentrated on pattern recognition pathways that, when mutated, permit overgrowth of bacteria and thus reduce host defenses [6–12]. There is evidence that tolerance properties exist in Drosophila, but the relationship between these mechanisms and resistance mechanisms, as well as their effects on host defense, have not been examined [13–17]. Tolerance as defined by the evolutionary community measures the slope of fitness versus pathogen load [18–20]. These parameters are quite difficult to assay in Drosophila infections; in the fly, morbidity is most easily measured by measuring the mean time to death and therefore we do not, strictly speaking, assay fitness. In addition, it is difficult to measure the number of infecting bacteria in a fly without killing the fly, and thus we cannot easily relate the bacterial number in a given fly with mortality because both assays are destructive. We define tolerance in the fly system by stating that a fly that can survive a given level of microbes better than another fly is better able to tolerate an infection. Changes in tolerance and resistance could have complicated effects on host defenses. We predict that for any given mutation, there are nine qualitatively different potential ways of affecting resistance and tolerance of a host, though the actual number of states is infinite (Figure 1). We imagine that both properties could be enhanced, diminished, or left unchanged by a single mutation. In our fly infection system not all nine of these combinations will be readily distinguishable. We inject flies with a pathogen and then monitor host survival and bacterial growth. Changes in resistance in mutants are detected by measuring pathogen growth and comparing these levels to those observed in infected wild-type flies. Tolerance of mutant flies is measured functionally as a change in survival when pathogen levels resemble that of infected wild-type flies. This mode of measurement prevents us from measuring changes in tolerance when microbe levels are also changing. Therefore, we predict that we should be able to differentiate among only five of these nine classes unless there are special circumstances, as described below. To test this prediction we examined the effects of altering the melanization arm of the Drosophila immune response on fly defense to a variety of pathogens. We chose this immune response because we anticipated that it not only functioned as a resistance mechanism in the fly and directly fights infections but could also cause considerable pathology in the host because it generates reactive oxygen; we expected this pathology would lead to changes in tolerance. Melanization is a presumed immune mechanism in the fly that produces melanin, visible as dark brown deposits, at the site of wounds and infection. Melanin is deposited after a chain of events induced by pattern recognition proteins, propagated by serine proteases and ultimately produced by the enzyme phenoloxidase [21–23]. In Drosophila, one such serine protease is CG3066, which acts in a melanization cascade that is negatively regulated by the serpin Spn27A [24,25]. Conventional wisdom suggests that melanin can sequester microbes to prevent their spread and that reactive oxygen species generated during melanin production can be directly harmful to microbes and possibly the host. There is evidence from work in other invertebrates, such as the crayfish Pacifastacus leniusculus, demonstrating that PPO activity is important for limiting microbial virulence [26–29]; however, the available literature concludes that melanization in Drosophila plays no role in fighting microbial infections, or it plays a redundant role, at best [24,25,30]. Though quite well defined biochemically, the functional contribution of this potential effector pathway to immunity remains in dispute. In the present study we show that by making a single mutation in the melanization arm, specifically CG3066, of the fly innate immune response, we could alter both tolerance and resistance in a microbe dependent fashion. By doing so we observed five of the nine predicted phenotypic classes for changes in resistance and tolerance. Among these five we found two cryptic phenotypes in which there was no change in survival of the mutant flies but bacterial levels were very different from those found in wild-type flies. This suggests that resistance and tolerance had achieved a new balance in these flies. We also found an unanticipated phenotype of CG3066 flies; these flies die significantly faster than wild-type flies when injected with sterile medium. We suggest that, in addition to its effects on the outcome of infections, this protease is important for tolerating some of the pathology that occurs during wounding. Typically in fly immunity, mutations have been reported to produce only two phenotypic classes—the flies either become sensitive to infections or their phenotype is unchanged. This work shows a level of complexity that has been missing in the description of Drosophila immunity. We suggest that these results have important implications about the evolution of immunity and that the equilibrium between resistance and tolerance of a host will be optimized by its interactions with pathogens in the wild. Previous studies examining the contribution of melanization to fly immunity did not test microbes that induce large melanization responses in the fly. Tang et al. observed that flies pierced with a needle containing a mixture of E. coli and Micrococcus luteus caused melanization at the site of needle insertion and concluded that this response was specific to the infection [25]. Although it is possible that the melanization observed was triggered by the infection, this cannot be concluded with confidence because injection with a sterile needle also results in deposits of melanin at the site of wounding approximately 24 h postinjection. Leclerc et al. did not report observations of melanization [24]. In Listeria monocytogenes– and Salmonella typhimurium–infected flies we observed, in addition to melanization at the site of injection, deposits of melanin just underneath the cuticle as well as in deeper tissues. This melanization is easily seen approximately 4 d (L. monocytogenes) or 7 d (S. typhimurium) after infection (Figure 2; unpublished data). We refer to this as a disseminated melanization response. We were curious if other bacteria elicited disseminated melanization during infection. We selected a diverse panel of bacteria and compared the patterns of melanization observed with media-injected control flies (Table 1). Within the first 24 h postinfection, we saw melanin at the site of injection that was comparable to what we observed in flies that received a control injection of media. This was true for all bacteria tested. In addition to the melanization at the injection site, we found that L. monocytogenes, S. typhimurium, and Staphylococcus aureus all elicit a robust disseminated melanization response in infected flies. On average we found that more than 90% of females and more than 70% of males infected with L. monocytogenes exhibit disseminated melanization, and the majority of these flies have spots of melanin deposited along the dorsal and ventral sides of the abdomen (Figures 2 and 3). These deposits can be found on the cuticle of both sexes, and large melanin clots are commonly observed within the ovaries of females. To a lesser extent we also find melanization along the thorax and the head. In S. typhimurium–infected flies, on average, more than 80% of females and 70% of males exhibit a disseminated melanization response over the course of the infection, and we observe similar patterns of melanization to what we see with L. monocytogenes. The majority of flies exhibit melanization in the abdomen on the cuticle and also in the ovaries of females (Figure 2). In contrast to L. monocytogenes infections, we did not see melanization in the thorax or the head segments with S. typhimurium. S. aureus–infected flies exhibit a different pattern of melanization; we found approximately 40%–50% of both infected females and males exhibit a punctuate patterning of melanin deposits localized to the dorsal vessel. On occasion there are large melanized particles deeper in the tissue of the abdomen. No melanin is deposited along the thorax or the head. We did not observe melanization beyond that seen at the injection site in flies infected with the remaining bacteria tested: Enterococcus faecalis, Streptococcus pneumoniae, Escherichia coli, and Burkholderia cepacia (Figure 3). Once we identified bacteria that elicit a disseminated melanization response we wanted to test whether this melanization response was important for a fly' s survival and how this response affects resistance and tolerance during an infection. There are three genes encoding phenoloxidases in the fly, monophenoloxidase (Bc), diphenol oxidase a2, and diphenol oxidase a3 [31]. The Bc gene has received the most attention for its involvement in immunity because of its single characterized mutant, which eliminates circulating phenoloxidase from the hemolymph [32–34]. This allele Bc1 is assumed to map to the monophenol oxidase gene; however, its molecular nature has not been reported [35]. Bc1 is a dominant mutation that appears to prematurely activate phenoloxidase in larval crystal cells. A troubling aspect of this mutation is that it damages crystal cells and causes them to be phagocytosed by plasmatocytes, and the melanized remains of these cells sit undigested in the hemocytes for the life of the fly [36]. We anticipated that this Bc mutation could have pleiotropic effects on the immune response; it blocks phenoloxidase activity, but it is also anticipated to alter the cellular immune response because it destroys one immune cell outright and causes another to be filled with undigestible material. Since the cellular immune response plays an important role in fighting many infections, we sought another way to reduce melanization. Leclerc et al. identified the protease encoded by CG3066 as a prophenoloxidase activating enzyme (PPO), whereas Tang et al. reported that CG3066 enzyme was required for PPO activation but did not directly target PPO. A mutation of this gene was reported to inhibit the immune induced proteolytic cleavage of a Drosophila protein that cross-reacted with a mosquito anti-phenoloxidase antibody [24]. RNAi inhibition of this gene blocked the induction of phenoloxidase activity in fly extracts [25]. Thus CG3066 mutants appeared to be a useful tool for dissecting the role melanization might play in resistance and tolerance. We found that these mutants are capable of producing melanin deposits at the site of injection for both media and microbial challenges comparable to that observed in wild-type flies (Figure 2); however, we did not observe a disseminated response in the CG3066 mutants with L. monocytogenes, S. typhimurium, S. aureus, or any of the other bacteria tested (Figures 2 and 3). To determine how CG3066 affects both tolerance and resistance properties of Drosophila we challenged CG3066 mutant flies with our panel of bacteria and measured survival rates and bacterial loads (Figures 4 and 5). The microbes we tested produced infections that fell into five different phenotypic classes. The first class includes L. monocytogenes and S. typhimurium. These microbes killed CG3066 mutants faster than wild-type flies and showed increased bacterial growth rates. S. typhimurium–infected mutants exhibited a 60% reduction in the median time to death (p < 0. 0001) with respect to wild-type flies, and there was a 50% reduction in survival in L. monocytogenes–infected flies (p < 0. 0001) (Figure 4). Using the UAS-GAL4 system and transgenic flies expressing double-stranded RNA targeting CG3066, we confirmed this reduction in survival by RNAi (p < 0. 0001) (Figure 4). S. typhimurium and L. monocytogenes grew to significantly higher levels at both 24 and 48 h postinfection in CG3066 mutants as compared to isogenic, wild-type parental controls (for both L. monocytogenes and S. typhimurium at 24 h p < 0. 05, for both L. monocytogenes and S. typhimurium at 48 h p < 0. 005) (Figure 5). This demonstrates that CG3066 plays an important and primary role in fighting some bacterial infections in the fly. L. monocytogenes establishes an intracellular infection in wild-type Drosophila. We performed a gentamicin chase experiment to determine the location of the L. monocytogenes in mutant flies. Following infection, flies were injected with gentamicin, which will kill extracellular bacteria, while intracellular bacteria are protected from the antibiotic [16]. Control flies were injected with medium. Following a 3-h chase, flies were homogenized and plated to determine bacteria levels. This allowed us to measure the numbers of both intracellular and extracellular bacteria in the fly and to determine the contribution this protease might have on both populations of bacteria. We found significantly more bacteria in the CG3066 mutant flies that received the medium chase compared to those that received the gentamicin chase, suggesting that there is an extracellular population of L. monocytogenes present in these mutants that is not present in wild-type flies (24 h, p = 0. 0022; 48 h, p = 0. 0043) (Figure 5). Similarly, we found an increase in L. monocytogenes growth when CG3066 expression is knocked down using RNAi (Figure 5). We conclude that CG3066 is normally important in controlling the growth of these microbes by enhancing the resistance properties of the fly, and this is similar to the sort of phenotype that has been seen for most Drosophila immunity mutants. The second class of microbes we found is defined by E. coli, which is a nonmelanizer and showed no change in killing rates or bacterial levels in CG3066 mutants. We define pathogenic bacteria as those that cause disease in wild-type flies; using this criterion, E. coli was the only nonpathogenic microbe we tested. An E. coli infection does not kill wild-type flies any faster than control flies injected with medium. We saw the same result in CG3066 homozygous mutant flies; E. coli infected mutants die at the same rate as medium injected mutants (Figure 4). Colony counts in infected CG3066 mutant flies were the same as seen in wild-type flies with an E. coli infection (Figure 5). This indicates that CG3066 has no effects on either fly resistance or tolerance with this type of infection. S. pneumoniae defines our third class of microbes; CG3066 mutants die significantly slower when infected with S. pneumoniae compared to wild-type flies. The median time to death in CG3066 mutants was extended by 100% (p < 0. 0001) (Figure 4). This increase in survival could have been due to changes in either resistance or tolerance. If resistance was altered, we anticipated that there would be differences in the levels of S. pneumoniae in the mutant flies, while changes in tolerance would leave the bacterial levels constant. We found that S. pneumoniae grew at a slower rate in CG3066 mutants than in wild-type flies, leading us to conclude that CG3066 mutant flies have better resistance against S. pneumoniae infection when melanization is absent. Alternatively, the presence of a functional melanization response could actually promote an S. pneumoniae infection in some manner. These results were surprising because we anticipated that the removal of a resistance response might increase the tolerance of the fly, but did not anticipate that it would increase the resistance of the host. The fourth and fifth classes are cryptic and are defined by E. faecalis and B. cepacia, respectively. Our results with E. faecalis were in some ways similar to what has been published previously; we found that E. faecalis killed wild-type and CG3066 mutant flies at the same rate. This result led Leclerc et al. to the conclusion that this mutation has no net effect on immunity [24]; however, we found that colony counts of the infected flies demonstrated that the story is more complicated than survival rates alone would lead us to believe. Infected CG3066 mutant flies had significantly lower levels of E. faecalis than do wild-type flies at 48 h postinfection (Figure 5). This result suggests that the resistance properties of these flies are increased with respect to E. faecalis because the fly is better able to kill this type of bacteria. Given that the survival rates of these flies is the same as that of wild-type flies, this increase in resistance properties appears to be balanced by a reduction in tolerance. We conclude that a lower number of E. faecalis can cause disease symptoms in CG3066 flies. In contrast to what we observed during E. faecalis infections, we found that although B. cepacia kills wild-type and CG3066 mutant flies at the same rate, there is increased growth of B. cepacia in the mutant flies. B. cepacia–infected flies exhibit a median time to death of 5 d post-infection (Figure 4). By 48 h postinfection, we find there is approximately 25 times more bacteria in the mutant flies compared to wild-type flies (p = 0. 0043) (Figure 5). This increased bacterial growth suggests that the resistance mechanisms in the mutant flies are less effective at fighting a B. cepacia infection. Because this increased bacterial growth is coupled with no change in survival we suggest that the decrease in resistance properties is accompanied by an increase in tolerance, which is the opposite of what we see with an E. faecalis infection. These E. faecalis and B. cepacia experiments highlight the importance of using multiple tests for immunity when studying infections; if we had assayed survival alone we would have been led to the conclusion that CG3066 plays no role in the fly' s interactions with these microbes, when in fact, the gene plays a complicated role in defense. The last microbe we tested was S. aureus. S. aureus infected CG3066 mutants exhibited the most striking difference in survival with approximately an 80% reduction in the mean time to death (p < 0. 0001) (Figure 4). Our results with S. aureus differ from those reported by Leclerc et al. who reported no difference in survival between S. aureus–infected CG3066 mutants and wild-type flies [24]. We did not measure growth in S. aureus–infected flies because S. aureus aggregates when grown in flies and this creates a lot of scatter in colony count experiments [15]. We are therefore unable to determine whether CG3066 mutants die from a S. aureus infection because of defects in resistance and/or tolerance properties and cannot place it in one of our predicted classes. We noted that medium-injected CG3066 mutant flies died faster than similarly treated wild-type flies. This result was missed in past publications because the survival curves were not extended until these control flies died, or these controls were apparently not performed [25,26]. To determine whether this was an effect of wounding on survival of CG3066 mutants or whether these mutant flies were merely shorter lived, we performed lifespan analysis on unmanipulated mutant and wild-type flies. We found that the unmanipulated flies had similar life spans (Figure 4). This suggests that CG3066 is important for tolerating some of the pathology of the wounding response. By testing a panel of bacteria that cause different types of infections, we demonstrated that melanization is activated during infection, and that the degree of activation is dependent on the type of infection. We predict that there are nine potential ways of affecting resistance and tolerance of a host (Figure 1). We found that by mutating a single gene we could alter both the resistance and tolerance properties of the fly and observed up to five of the nine predicted phenotypic classes. Though not all nine possible classes were seen, we did observe the four major changes that we predicted; both tolerance and resistance could be increased or decreased by a single mutation, and these properties were dependent upon the particular microbial challenge. The phenotypes found in CG3066 mutant flies were somewhat surprising. We anticipated that this protease mutant would be less able to kill some bacteria and thus would show decreased resistance. Likewise we predicted that melanization might cause collateral damage and nonmelanizing flies would show increased tolerance. It was a surprise to find that melanizing mutants are more resistant to some microbes. We propose that the microbe may benefit from the damage done by the reactive oxygen because of autoimmune damage to the host; or perhaps when flies lack melanization, other, more effective immune responses show increased activity. It was also a surprise that tolerance would decrease in nonmelanizing flies. We propose that regulation could increase the activity of alternate immunity pathways that cause increased collateral damage or that the bacteria cause a different type of pathology in nonmelanizing flies that is more damaging to the host than we see in wild-type flies. Of the nine proposed phenotypic classes (Figure 6), three should be easy to distinguish; these are cases in which resistance remains unchanged (and thus microbe levels are the same as in wild-type flies) and tolerance varies (Figure 6, right). We saw one of these classes: CG3066 mutant flies infected with E. coli show no change in either resistance or tolerance. We interpret this as meaning that CG3066 has no effect at all on this type of infection. We did not see an example of the second or third class of mutant with our panel of bacteria. The second class would show no change in resistance but an increase in tolerance. The third would have no resistance effect and would reduce tolerance. We note that we have previously identified mutations in the second and third class. We have reported that the fly tumor necrosis factor–related molecule, eiger, is important for fighting infections with extracellular pathogens, and eiger mutants have decreased resistance during these types of infections. Yet during Salmonella infection we found that eiger mutants exhibit similar bacterial burdens to wild-type flies but have an extended life span, indicating that during this type of infection there is an increase in tolerance with no change in resistance. Eiger mutants have a balanced increase in resistance and a decrease in tolerance during a Listeria infection, similarly to what is seen in CG3066 mutant flies infected with E. faecalis. Eiger does not appear to exhibit as many phenotypic combinations of resistance and tolerance as we observe with CG3066 but it offers additional support that a single gene can affect both resistance and tolerance in various ways depending on the microbe [15,16]. In a published genetic screen we identified six mutants, all of which were sensitive to Listeria but exhibited levels of bacteria comparable to that found in wild-type flies [17]. These six mutants represent an additional phenotypic class, a decrease in tolerance with no change in resistance. It is not known if these genes can affect resistance and tolerance in additional ways with different types of infections. Three classes of phenotypes will show a decrease in resistance (Figure 6, middle). This is the phenotype we observed with L. monocytogenes and S. typhimurium infections. These bacteria grow faster in CG3066 mutants and the flies die faster. Typically, we cannot distinguish changes in tolerance here because we do not have a good method of keeping the bacterial levels constant or correlating bacterial load and morbidity. B. cepacia provides a special case where CG3066 mutant flies show no change in survival but show a significant increase in bacterial load when infected with this bacterium. This suggests that there must be a balanced change in resistance and tolerance in these flies. Because we did not determine the growth of S. aureus in CG3066 mutants, we cannot determine whether resistance and/or tolerance is affected with this microbe. If we were to consider the survival and melanization phenotypes in S. aureus-infected CG3066 mutants only, we would suggest that it falls into the same class as L. monocytogenes and S. typhimurium. Yet, because we have seen that survival is not an accurate predicator of bacterial loads we cannot make this claim. Another three classes of phenotypes are expected to show an increase in resistance (Figure 6, left). We saw at least one of these classes. CG3066 mutant flies live longer than do wild-type flies infected with S. pneumoniae and are better at clearing the infection because they have a heightened resistance response. In this case we suggest that when melanization occurs, flies are less able to defend themselves against the infection; perhaps the flies waste energy on a nonproductive immune response or suffer from autoimmune damage. CG3066 flies infected with E. faecalis provides a second special case, where we can determine changes in resistance and tolerance; the survival rate of CG3066 mutants and wild-type parental strains were the same. This means that resistance and tolerance changes must be balanced, and we conclude that since resistance is increased in these flies, tolerance must have decreased. We showed that CG3066 is important in controlling the nature of the infection. In the case of L. monocytogenes, we found that there are more extracellular bacteria present in CG3066 mutants while the number of intracellular microbes remains constant. We suggest two mechanistic explanations for this change in the nature of the infection. First, melanization may be responsible for killing extracellular L. monocytogenes, and a loss in CG3066 results in an increase in extracellular bacteria. Second, CG3066 might be responsible for killing fly cells infected with L. monocytogenes, and if this does not happen, the cells may release larger numbers of bacteria into the circulation. We made assumptions about the shape of tolerance curves when interpreting our data. We determined the life span of uninfected flies and compared this to the lifespan of infected flies and measured their bacterial levels 24 h postinfection. If these data were graphed, they would define two points and a tolerance curve could be interpolated between them. We interpreted our results as if the tolerance curve was a straight line and that each additional microbe would incrementally result in a decrease in survival. This is an assumption and should apply to some infections; however, it is possible to imagine alternative tolerance curves. We made this assumption because we do not know the actual shape of any of these curves and chose the simplest possibility. This raises the point that to truly understand the nature of microbial pathogenesis we will have to precisely define tolerance curves. Resistance and tolerance are predicted to have very different evolutionary outcomes [37]. For resistance, if the benefits of the trait outweigh the cost of the trait then the number of hosts with that resistant trait will become more frequent in a population. As the trait occurs in higher proportions the occurrence of disease will decrease. Eventually the occurrence will become so low that the cost of the trait then outweighs the benefits of the trait and the trait will cease spreading through a population. Therefore a resistance trait is not predicted to become fixed in a population. The dynamics of a tolerance trait should be quite different. As a tolerance trait spreads through a population the occurrence of disease may also increase because more tolerant hosts are available to infect. Because the incidence of disease remains high the benefits of carrying the tolerance trait will always outweigh the costs of having it, so the tolerance trait is predicted to become fixed in a population. Our results have very important implications for the evolutionary dynamics of tolerance traits. We show that a tolerance trait can actually be quite dynamic and predict that it will not reach fixation because the same trait can serve as a resistant trait for a different pathogen. Furthermore, resistant traits are typically highly dynamic because of the coevolutionary relationship they have with pathogens, and this will feed into the system with corresponding changes in tolerance. Our findings suggest that the evolutionary dynamics of resistance and tolerance can be highly fluid even in the absence of pathogens driving such genetic instability in a host. We noted an interaction between wounding, survival, and melanization; CG3066 mutant flies that were given an injection of sterile medium were shorter lived compared to wild-type flies given the same treatment. Unmanipulated CG3066 and isogenic parental lines showed no difference in survival. We have always found that medium-injected flies die faster than unmanipulated flies and do not know the pathology behind this early death. It remains possible that these flies are suffering from a cryptic infection of the native microbiota in the fly vial or that the wound healing process itself causes some pathology. These results demonstrate that CG3066 contributes to a fly' s ability to withstand this pathology. An issue that arises as a result of the difference in the survival rate of CG3066 flies in medium-injected flies versus unmanipulated controls is that if the medium-injected CG3066 flies die faster than do wild-type flies, how can we determine which flies have changes in immunity? We have two answers to this question: The first is that in the three cases where we see CG3066 flies dying faster than wild-type flies during an infection, we found that this is associated with an increased growth of bacteria and conclude that these flies have a resistance defect, in addition to other problems that they might have. The second answer is that we suggest the flies are dying for different reasons in medium-injected versus pathogen-infected flies, and that the two processes do not necessarily correlate with each other. We do not know the cause of pathology in either death by infection or death by wounding and have no reason to suspect that they are identical. Recent studies suggested that these immune mechanisms are dispensable in Drosophila with respect to their importance for survival to microbial infections or, at best, serve a redundant role [24,25]. These experiments were carried out by assaying the phenotypes of CG3066 mutants. We are careful to limit the analysis of our phenotypes to the effects of CG3066 and do not go so far as to state that the loss of melanization is the cause for the phenotypes we observe. It remains possible that CG3066 serves additional roles in fly immunity and does not solely activate phenoloxidase. Here we report that the response initiated by CG3066 is important for surviving infections and that its effects are dependent on the type of infection. This response affects both resistance and tolerance mechanisms in Drosophila. We suggest that the importance of these mechanisms was missed previously because past studies utilized microbes that do not induce a strong melanization response in the fly, did not measure bacterial loads in the infected flies, did not take the survival curves out to completion, and compared mutants to other mutants instead of to wild-type flies. The findings reported here have implications about the evolution of immunity; they suggest that polymorphisms that increase resistance to one pathogen may reduce the resistance or tolerance to other pathogens and thus the defense response of a given fly strain will likely be optimized by its interaction with microbes in the wild. As a result, there is likely no “best” solution that produces a perfect immune system, only an equilibrium that allows the fly to survive the pathogenic threats that its ancestors have faced. This equilibrium will require the balancing of both resistance and tolerance, and thus we can not completely understand the defensive properties of a host unless we measure both of these properties in response to a variety of pathogens. The wild-type parental strain used in all experiments is white1118 (Bloomington stock center, stock 6326) The CG3066 KG02818 allele was obtained from Bloomington stock center and backcrossed onto the white1118 background for four generations. Flies were kept in standard fly bottles containing dextrose medium. All strains used are listed in Table 1. S. pneumoniae cultures were grown standing at 37 °C 5% CO2 in brain heart infusion medium (BHI) to an OD600 of 0. 15 and aliquots were frozen at −80 °C in 10% glycerol. For infection, an aliquot of S. pneumoniae was thawed and diluted 1: 3 with fresh BHI medium and allowed to grow to OD600 of 0. 15 at 37 °C 5% CO2. Bacteria was then concentrated to an OD600 of 0. 3 in PBS. E. coli, E. faecalis, and S. typhimurium cultures were grown overnight at 37 °C in Luria Bertani (LB) medium. E. coli and E. faecalis cultures were shaken, while S. typhimurium cultures were grown standing. E. coli and S. typhimurium cultures were diluted to OD600 of 0. 1 with fresh LB medium prior to injection. E. faecalis cultures were diluted to an OD600 of 0. 05 with medium. B. cepacia cultures were grown standing overnight in LB medium at 29 °C and injected at an OD600 of 0. 001. L. monocytogenes and S. aureus were grown in BHI medium. L. monocytogenes was grown standing and injected at an OD600 of 0. 01. S. aureus was grown shaken and injected at an OD600 0. 001. Male 5- to 7-d-old files were used for injection. Flies were anesthetized with CO2 and injected with 50 nl of culture using a picospritzer and pulled glass needle. Flies were injected in the anterior abdomen on the ventrolateral surface. Flies were then placed in vials containing dextrose medium in groups of twenty and incubated at 29 °C, 65% humidity with the exception of B. cepacia, which was incubated at 18 °C with no humidity control. For each microbe tested, w1118 and CG3066 mutants were injected with the microbe or medium as a control. Flies were placed in dextrose vials in groups of 20 after injection and a total of sixty flies were assayed for each condition. The number of dead flies was counted daily. Using Prism software, Kaplan-Meier survival curves were generated and statistical analysis was done using log-rank analysis. Survival was tested for each microbe listed in Table 1 at least three times and gave similar results for each trial. Infected flies were homogenized in media supplemented with 1% Triton X-100 and serially diluted. S. pneumoniae-infected flies were homogenized in PBS without triton. Dilutions were plated on LB agar plates (blood agar plates for S. pneumoniae) and incubated overnight. The data was plotted as box-and-whisker plots using Graphpad Prism software for three independent experiments. The p-value was determined with a nonparametric two-tailed t-test. For the gentamicin chase experiments, flies were injected with 50 nl of 1 mg/ml gentamicin or water 3 h prior to homogenizing and plating.

To boost its defenses, an organism may increase its resistance to infection by reducing the fitness of the invading pathogen; alternatively, the host may increase its tolerance by reducing the damage caused by a given quantity of pathogen. Melanization is an immune response that has been linked to defense in the fly and other invertebrates. It is expected to cause resistance to infection, as well as host damage mediated by reactive oxygen species generated during melanin production. We demonstrate here that the loss of a gene required for melanization produces a surprisingly complex spectrum of phenotypes, increasing and decreasing both resistance and tolerance to a variety of microbes. For example, increasing resistance to one pathogen can produce corresponding changes in either resistance or tolerance to another pathogen. As a result, there is likely no “best” solution that produces a perfect immune system, only an equilibrium that allows the fly to deal with the pathogenic threats that its ancestors have faced. This equilibrium will require the balancing of both resistance and tolerance, and our study demonstrates that we cannot completely understand the defensive properties of a host unless we measure both of these properties in response to a variety of pathogens.

Abstract Introduction Results Discussion Materials and Methods

infectious diseases immunology

2008

A Signaling Protease Required for Melanization in Drosophila Affects Resistance and Tolerance of Infections