Datasets:

afg1

/

litscan-epmc-subset

like 0

PMC9649614

Jin-Xiu Pan,Daehoon Lee,Dong Sun,Kai Zhao,Lei Xiong,Hao-Han Guo,Xiao Ren,Peng Chen,Raquel Lopez de Boer,Yuyi Lu,Helena Lin,Lin Mei,Wen-Cheng Xiong

Muscular Swedish mutant APP-to-Brain axis in the development of Alzheimer’s disease

10-11-2022

Alzheimer's disease,Neurodegenerative diseases

Alzheimer’s disease (AD) is the most common form of dementia. Notably, patients with AD often suffer from severe sarcopenia. However, their direct link and relationship remain poorly understood. Here, we generated a mouse line, TgAPPsweHSA, by crossing LSL (LoxP-STOP-LoxP)-APPswe with HSA-Cre mice, which express APPswe (Swedish mutant APP) selectively in skeletal muscles. Examining phenotypes in TgAPPsweHSA mice showed not only sarcopenia-like deficit, but also AD-relevant hippocampal inflammation, impairments in adult hippocampal neurogenesis and blood brain barrier (BBB), and depression-like behaviors. Further studies suggest that APPswe expression in skeletal muscles induces senescence and expressions of senescence-associated secretory phenotypes (SASPs), which include inflammatory cytokines and chemokines; but decreases growth factors, such as PDGF-BB and BDNF. These changes likely contribute to the systemic and hippocampal inflammation, deficits in neurogenesis and BBB, and depression-like behaviors, revealing a link of sarcopenia with AD, and uncovering an axis of muscular APPswe to brain in AD development.

Muscular Swedish mutant APP-to-Brain axis in the development of Alzheimer’s disease Alzheimer’s disease (AD) is the most common form of dementia. Notably, patients with AD often suffer from severe sarcopenia. However, their direct link and relationship remain poorly understood. Here, we generated a mouse line, TgAPPsweHSA, by crossing LSL (LoxP-STOP-LoxP)-APPswe with HSA-Cre mice, which express APPswe (Swedish mutant APP) selectively in skeletal muscles. Examining phenotypes in TgAPPsweHSA mice showed not only sarcopenia-like deficit, but also AD-relevant hippocampal inflammation, impairments in adult hippocampal neurogenesis and blood brain barrier (BBB), and depression-like behaviors. Further studies suggest that APPswe expression in skeletal muscles induces senescence and expressions of senescence-associated secretory phenotypes (SASPs), which include inflammatory cytokines and chemokines; but decreases growth factors, such as PDGF-BB and BDNF. These changes likely contribute to the systemic and hippocampal inflammation, deficits in neurogenesis and BBB, and depression-like behaviors, revealing a link of sarcopenia with AD, and uncovering an axis of muscular APPswe to brain in AD development. Alzheimer’s disease (AD) is the most common form of dementia. Pathologically, it is characterized by cortical and cerebrovascular β-amyloid (Aβ) plaques, phospho-tau containing neurofibrillary tangles, reactive glial cell-associated chronic inflammation, and neuronal loss [1, 2]. Interestingly, in addition to the brain pathology, AD patients often have sarcopenia, a condition characterized by the loss of skeletal muscle mass and function. Clinically, the sarcopenia appears to be tightly associated with the dementia and AD progression, correlating well with the severity of AD [3, 4]. A significant higher prevalence rate of sarcopenia in AD (early to moderate) patients than that of same-aged population with normal cognition has been reported; [3–5] and the poor muscle functions or lower muscle mass in patients with sarcopenia have been implicated as a driver for their association with later-life cognitive impairment [6, 7]. Etiologically, both AD and sarcopenia disorders share several common environmental risk factors, including aging and chronic inflammation, and a few genetic risk genes, such as ApoE [7–14]. While these clinical, genetic, and epidemic studies indicate a strong association of AD with sarcopenia, it remains possible that their association is a random coincidence due to their shared environmental risk factors, and it remains unclear exactly how they are linked and what are their relationship(s) are. We chose Swedish mutant APP (APPswe) to address above questions for the following reasons. First, APPswe is one of the earliest mutants identified in patients with early-onset AD (EOAD). Although APPswe is detected in small fractions of EOAD patients, its functions in Aβ production in the brain and in promoting AD pathogenesis have been well studied in multiple animal models (e.g., Tg2576 and 5XFAD, both well-characterized AD animal models that express APPswe under the control of prion and Thy1 promoter, respectively). Second, much AD research has been focused on the impact of Aβ on the brain, even though App or APPswe is known to be expressed not only in the brain, but also in periphery tissues [15, 16], including skeletal muscles [17]. While investigating phenotypes in APPswe-based animal models have provided valuable insights into Aβ pathology in the brain and impairments in mouse cognitive functions, the functions of APPswe in periphery tissues, such as muscles, remain poorly understood. Third, APP’s physiological function in muscles has been emerged. In addition to its age-dependent expression in muscles and NMJs (neuro-muscular junction) [17], mice with APP knocking out show dysfunctional NMJs with aberrant localization of presynaptic proteins, reduced synaptic vesicles, and abnormal postsynaptic AChR clusters [18]. Fourth, altered expression or increased cleavage of APP appears to be involved in multiple types of human muscle degenerative diseases (see Supplemental Table 1). For examples, muscle fibers of patients with inclusion-body myositis (IBM) have intra-fiber “plaque-like” Aβ accumulation [19], which are believed to promote myofiber degeneration, atrophy, and death [20]. Aβ accumulation in strophic muscles fibers is a key factor in GNE myopathy [21]. Aβ accumulation is also detected in muscles of patients with ALS (amyotrophic lateral sclerosis) and ALS mouse models [22]. In AD patients, Aβ levels are elevated not only in the brain, but also in muscles (e.g., temporalis) [23]. Finally, examinations of muscle structures in Tg2576, the well-characterized AD animal model that expresses APPswe ubiquitously, have revealed early-onset sarcopenia-like deficits, months before any brain-pathologic defect that can be detected [24, 25]. Taken together, these observations argue against the view for sarcopenia-like deficits as a consequence of neurodegeneration, or a random coincidence, implicate dysfunctional APP or Aβ as a potential common denominator for AD and muscle degenerative diseases, and raise additional question- could problems in muscles contribute to AD pathology in the brain? Here, we addressed this question by use of a newly generated APPswe-based animal model, TgAPPsweHSA mice, which express APPswe specifically in muscles. Investigating phenotypes in TgAPPsweHSA mice showed not only earlier onset sarcopenia-like muscle deficits, but also age-dependent depression-like behaviors and brain pathology (largely in the hippocampus), such as increased glial activation, impaired BBB, and elevated pro-inflammatory cytokines. Further studies demonstrate increased senescence and SASPs in APPswe expressing muscles or C2C12 cells; and inhibition of senescence by its inhibitors diminishes or abolishes nearly all the phenotypes in TgAPPsweHSA mice. These results thus demonstrate a contribution of muscular APPswe to the development of both AD and sarcopenia, revealing a link between sarcopenia and AD, and uncovering a muscle-to-brain crosstalk for AD development. The LSL-APPswe mice were generated using the pCCALL2 plasmid as described previously [26]. TgAPPsweHSA mice were generated by crossing LSL-APPswe with HSA-Cre mice (purchased from Jackson laboratory, which is donated by Dr. IMR Colony, stock #006149) [27]. This study was conducted in accordance with the National Institutes of Health (NIH) guidelines and the Institutional Animal Care and Use Committee at Case Western Reserve University approved protocols (IACUC, 2017–0121 and 2017-0115). The other material and methods including Animals, Reagents, Behavioral tests, Stereological cell counting, Histologic staining [H&E, NADH-TR (transferase), SDH (succinate dehydrogenase), COX (Cytochrome c oxidase), PAS (Periodic Acid Schiff), and Gomori-trichrome], Immunofluorescence staining and image analysis, Western blotting, EdU injection and labeling, L-Series label-multiplex antibody arrays, SA-β-gal staining, Elisa assays, RNA isolation, and RT-qPCR were described in supplemental information. To investigate possible skeletal muscular APPswe’s effect in AD and sarcopenia development, we generated TgAPPsweHSA or LSL-APPswe:HSA-Cre mice by crossing LSL(Loxp-Stop-Loxp)-hAPPswe (human APPswe) with HSA-Cre (human skeletal a-actin promoter driven Cre) mice (Fig. S1A). The APPswe expression in TgAPPsweHSA mice is thus under the control of both the CAG promotor (a chicken β-actin promotor with a CMV enhancer to express its mRNAs) and the HSA-Cre dependent removal of LSL in LSL-hAPPswe mice (Fig. S1A) [26]. The specific expression of the HSA-Cre activity in the skeletal muscles was verified in HSA-Cre: Ai9 mice (where the tdTomato expression depends on Cre activity) (Fig. S1B), in line with literature reports [27]. We further examined hAPPswe’s expression in TgAPPsweHSA mice. RT-qPCR analysis using specific primers for human APP detected hAPP’s transcripts only in skeletal muscles, but un-detectable in the brain-hippocampus or cortex, nor other tissues/organs (such as heart, lung, liver, and kidney) of TgAPPsweHSA mice (Fig. S1C). Among different muscles, the hAPPswe’s transcripts were abundantly expressed in tibialis anterior (TA) (a distal fast twitch type), Quadriceps (a proximal fast twitch type), and soleus (a slow twitch type) in TgAPPsweHSA mice (Fig. S1C). Western blot analysis also showed selectively expression of hAPPswe protein in TA muscles, but not in the cortex nor hippocampus in TgAPPsweHSA mice (Fig. S1D, E), verified the RT-qPCR results. Given the abundant expression of APPswe in skeletal muscles, we wondered whether such muscular APPswe expression could induce sarcopenia-like deficits, such as decreased muscle fiber size and increased muscle fiber degeneration [28, 29]. H&E histologic staining analysis of TA muscle fibers showed normal or comparable morphology in 3-MO TgAPPsweHSA mice to those of control mice (Fig. 1A, B). However, at 6-MO, the mutant TA muscles exhibit sarcopenia-like deficits, showing smaller muscle fiber area with increased fibers containing central nuclei (Fig. 1A, B), a feature of muscle fiber degeneration [30]. We then characterized the phenotypes in other type of muscles, including quadricep and soleus. While the mutant quadricep, a proximal fast twitch type of muscles, showed similar age-dependent deficits to those of mutant TA muscles (Fig. 1C, D), the mutant soleus, a slow twitch type of muscles, exhibited an earlier onset degenerative phenotype, exhibiting a reduction in their fiber size, and an increase in fibers with central nuclei distribution at age of 3-MO (Fig. 1E, F), which were un-detectable in the mutant TA nor quadriceps at this age (Fig. 1). These results thus suggest muscle fiber type- and age-dependent sarcopenia-like deficits in TgAPPsweHSA mice. This view was further tested by additional histologic staining analyses, including Gomori-trichrome, PAS (Periodic Acid Schiff), NADH-Transferase, and COX (cytochrome c oxidase), and SDH (succinate dehydrogenase), in both 3- and 6-MO TA and soleus muscles. Indeed, at both ages of 3- and 6-MO, the mutant soleus muscles showed increases in fibers with Gomori-trichrome positive staining, decreased COX+, but increased COX-:SDH+ fibers, and elevated cytoplasmic PAS+ fibers (Fig. S2), suggesting mitochondrial myopathy, fibrosis, and glycogen overload in the mutant soleus. The mutant TA muscles also showed decreased COX+ and increased COX-:SDH+ fibers at both 3- and 6-MO, elevated cytoplasmic PAS+ fibers at 6-MO, but little changes by Gomori-trichrome and NADH-TR staining (Fig. S2), suggesting a relatively weaker myopathy in TA than those in soleus muscles from the mutant mice. Together, these results provide additional support for earlier onset of myopathies in the mutant muscles, which resemble the features of sarcopenia-like myopathy. We then asked whether TgAPPsweHSA mice exhibit any brain pathology similar to those of APPswe-based AD animal models (e.g., Tg2576) [31–33]. It is known that APPswe-based AD animal models (e.g., Tg2576) exhibit not only increased Aβ40 and Aβ42 levels in the brain, but also elevated glial activation, inflammation, and reduced neuronal synapses [31–33]. We thus first measured both Aβ40 and Aβ42 levels in muscles, serum samples, and brain tissues in TgAPPsweHSA mice (at 6-MO), compared with those of same aged LSL-APPswe (a negative control) and Tg2576 (a positive control) mice. ELISA analyses of Aβ40 or Aβ42 levels showed little to no Aβ increase in the hippocampus, cortex, or serum samples, but a slight increase in the TA muscles, of TgAPPsweHSA mice (6-MO), as compared with those of the negative control mice (Fig. S1F, G). In contrast, Tg2576 mice showed marked increases of Aβ40 and Aβ42 levels in their brain tissues and serum samples, and a comparable level of Aβ40 in TA muscles to that of TgAPPsweHSA mice (Fig. S1F, G). We second examined neuronal distribution patterns and densities in the hippocampus and cortex of TgAPPsweHSA mice (at 6-MO) through a co-immunostaining analysis using antibodies against NeuN (a marker for all neurons) and Ctip2 (a marker for neurons in the Layers V-VI cortex and CA1-2 and DG hippocampus). Little to no changes in the NeuN+ and Ctip2+ neuron distribution patterns and densities were detected in TgAPPsweHSA brains, as compared with those of controls (Fig. S3). Third, we assessed the morphologies and densities of glial cells, including Olig2+ oligodendrocytes, S100β+ ependymal cells, GFAP+ astrocytes, and IBA1+ microglial cells, in the brain sections of control (LSL-APPswe) and TgAPPsweHSA mice. Again, little to no changes in the Olig2+ oligodendrocytes or S100β+ ependymal cells were detected in the brain of TgAPPsweHSA mice (Fig. S4). However, both GFAP+ astrocytes and IBA1+ microglial cells were increased in the 6-MO TgAPPsweHSA brain, particularly in the hippocampus at both dorsal and ventral regions (Fig. 2A–H), suggesting an activation of these glial cells. In line with this view, the increased GFAP and IBA1 protein levels were also detected in 6-MO TgAPPsweHSA hippocampus using Western blot analysis (Fig. 2I, J). Notice that GFAP+ astrocytes and IBA1+ microglial cells were slightly increased in the cortex layer I-III of 6-MO TgAPPsweHSA (Fig. S5A–C), but the protein levels remained unchanged in the 6-MO TgAPPsweHSA cortex by Western blot analysis (Fig. S5D, E). This suggests that the hippocampus appeared to be more vulnerable than the cortex in the mutant mice. Negligible changes of these glial cells at 3-MO were observed in the mutant brain (Fig. S6), indicating an age-dependency of these phenotypes. Considering the tight association of glial cell activation with brain inflammation [13, 34], we examined expressions of inflammatory cytokines (e.g., Il1b, Il6, Il10, and Tnfa), chemokines (e.g., Ccl3, 5, 12, 17), and growth factors (e.g., Pdgfb, Bdnf, Tgfb1 and Csf2) in the hippocampus of both control and TgAPPsweHSA mice (at 3/6-MO) using RT-qPCR analysis (Fig. S7A, B). Among 12 genes examined in 3-MO TgAPPsweHSA mice, only Bdnf was decreased, as compared with that of control mice (Fig. S7A). However, among the 61 genes examined in 6-MO TgAPPsweHSA mice, 17 were increased and 8 were decreased in the mutant hippocampus (Fig. S7B). Interestingly, hippocampus of Tg2576 mice (6-MO) showed a similar inflammation phenotype to that of TgAPPsweHSA mice, displaying increased expression of Il6, Il15, Cxcl10, Lif, and Vegfd, but decreased expression of Pdgfb and Bdnf (Fig. S7C, D). Notice that PDGF-BB is a key growth factor for the development of pericytes, a blood vessel associated cells that regulate BBB integrity [35, 36]. The reduction of Pdgfb in the mutant hippocampus led to a speculation for a deficit in PDGFRb+ pericytes. Indeed, co-immunostaining analysis showed decreased PDGFRβ+ pericytes, but little to no changes in the CD31 marked endothelial cells, in the mutant hippocampus (at 6-MO) (Fig. 3A–C). We then examined BBB leakage by tail vein injections of Dextran (3 kDa) into the control and mutant mice (at 6-MO) (Fig. 3D). The dextran signals outside of CD31+ blood vessels were detected in the mutant hippocampus, but not in the control (Fig. 3E). Interestingly, the dextran signals were largely in the mutant Hilus region (Fig. 3D), indicating a selective BBB leakage in this region. Moreover, more IBA1+ microglial cells were associated with CD31+ blood vessels in the mutant Hilus than those of the control mice (Fig. 3F, G), supporting the notion that blood vessel/BBB are damaged in this region. Given the reports that BDNF is a critical growth factor for adult hippocampal/ DG (dentate gyrus) neurogenesis [37–39], and considering the reduction of Bdnf in not only AD animal models [40], but also TgAPPsweHSA hippocampus (Fig. S7), we examined DG neurogenesis in the mutant mice. EdU was injected into the mice (at ages of 1-, 3- and 6-MO, respectively, ~12 h before sacrifice) to label proliferative neural stem cells (NSCs). Hippocampal sections were co-immunostained with EdU and antibody against DCX (doublecortin), a marker for newborn neurons derived from NSCs, as shown in Fig. S8A. Remarkably, TgAPPsweHSA mice at ages of 3-MO and 6-MO, but not 1-MO, displayed significant reductions in EdU+ and DCX+ cell densities at both dorsal and ventral DG (Fig. S8), demonstrating an early onset deficit in the NSC proliferation and thus DG neurogenesis in TgAPPsweHSA mice, exhibiting another similar deficit as AD animal models [40]. We then asked whether TgAPPsweHSA mice show any behavior changes similar to those of AD animal models (e.g., Tg2576), such as age-dependent impairment in cognitive function [32, 33]. We first subjected TgAPPsweHSA and control (LSL-APPswe) mice (at age of 6-MO, both males and females) to Morris water maze (MWM) and Y-maze, for the assessment of mouse spatial learning and memory function, and working memory, respectively [41, 42] (Figs. 4A–C and S9). No obvious difference in MWM or Y-maze task performance was observed between the mutant and control mice (Figs. 4A–C and S9), suggesting little cognitive decline in TgAPPsweHSA mice. Anxiety- and depression-like behaviors are often associated with increased glial activation and inflammatory cytokines, and decreased DG neurogenesis. We thus subjected mice (at ages of 3- and 6-MO) to behavior tests including open field test (OFT) to evaluate TgAPPsweHSA mice’ locomotor activity, anxiety, and willingness to explore environments [43]; elevated plus maze test (EPMT) and light/dark transition test (LDT) to assess mouse anxiety-related behavior [43–45]; sucrose preference test (SPT), force swimming test (FST), and tail suspension test (TST) to examine mouse depression [45, 46]. As shown in Fig. 4D–G, TgAPPsweHSA mice developed age-dependent depression-like behaviors. At age of 3-MO, no obvious difference in all the behavior tests was detected between mutant and control mice (Fig. S10). However, at 6-MO, the mutant male mice exhibited increased immobility times in both FST and TST, and decreased sucrose preference (Fig. 4G), without obvious changes in the performance of OPT, EPMT, and LDT (Fig. 4D–F), suggesting depression-like behavior, but little deficits in exploratory and locomotor activities and anxiety in 6-MO TgAPPsweHSA mice. These depression-like behaviors were detectable not only in male, but also in female mutant mice (Fig. S9). To understand how expression of APPswe in skeletal muscles in TgAPPsweHSA mice affects their brain inflammation and function, we tested a speculation that the inflammation and depression-like behavior phenotypes of TgAPPsweHSA mice may be induced by secreted proteins from APPswe+ muscles. We first screened for altered serum plasma proteins in TgAPPsweHSA mice (at 6-MO) using multiplex antibody-based arrays (Fig. 5A). Among the 90 proteins on the array, only 3 were lower, but 42 were higher in the mutant serum samples than those of control mice (Fig. 5A, B). We second examined 3-MO mutant mice and compared the changes in their serum samples with those of 6-MO mutant mice. Using the same multiplex antibody-based arrays, 24 proteins were increased, and 14 were decreased in 3-MO mutant mice (Fig. 5C). Comparing the changes between 6-MO and 3-MO mutant samples, 17 proteins were increased, and 2 proteins were decreased at both 3- and 6-MO mutant serum samples (Fig. 5D, E). Interestingly, among these 17 increased proteins, 12 of them exhibited more dramatic increases in 6-MO than those of 3-MO mutant mice (Fig. 5E). These results suggest age-dependent changes in serum proteins in TgAPPsweHSA mice. Third, we addressed whether TgAPPsweHSA mice (at 6-MO) exhibit similar changes in their serum samples to those of 6-MO Tg2576 mice. Remarkably, among 42 upregulated proteins in the serum of TgAPPsweHSA mice, 29 were also elevated in Tg2576 serum samples (Fig. 5F). Finally, among the altered serum proteins in both TgAPPsweHSA and Tg2576 mice, two pathways were noted: one is the increased pro-inflammatory cytokine (e.g., IL6 and IL1β) and chemokine pathway, and the other is the decreased growth factor (e.g., PDGF-BB) pathway. We thus used ELISA to further verified changes in IL6, IL1β, and PDGF-BB in serum samples of both mouse lines. Indeed, both IL6 and IL1β cytokines were increased in TgAPPsweHSA and Tg2576 serum samples (Fig. 5G); and PDGF-BB was significant lower in the serum samples of 6-MO TgAPPsweHSA and Tg2576 mice compared to those of control mice (Fig. 5H), but not in 3-MO TgAPPsweHSA mice (Fig. 5H). Together, these results reveal a similar, but not identical, profile change in TgAPPsweHSA serum samples to those of Tg2576 mice, providing evidence for a chronic systemic inflammation in both TgAPPsweHSA and Tg2576. Given APPswe’s specific expression in skeletal muscles in TgAPPsweHSA mice, we speculate that the APPswe+ muscles might be the key source for the increased systemic inflammation. We thus analyzed transcripts of the altered genes in the 6-MO control and TgAPPsweHSA TA muscles. Among the 61 genes examined by RT-qPCR analyses, 37 up-, and 6 down-regulated transcripts were detected in TgAPPsweHSA TA muscles (Fig. S11A). Interestingly, 49 factors were tested in both serum and TA muscle samples. 30 factors in serum and 27 factors in muscle were upregulated in mutant mice, and 19 (~63%) of them were increased in the mutant TA muscles (Fig. S11B). Comparing the altered transcripts between mutant muscles and hippocampus showed 12 upregulated transcripts (e.g., Il6, Lif, Csf1) and 3 downregulated transcripts (e.g., Pdgfb, Bdnf, and Il4) in both tissues (Fig. S11C, D). Whereas these results support the view for a systemic inflammation in the muscle-blood-hippocampus axis, further correlation plots showed a significant correlation of the significant changes of these transcripts between mutant TA muscles vs hippocampus (Fig. S11E), but not mutant TA muscles vs serum (Fig. S11F), nor mutant serum vs hippocampus (Fig. S11G). Notice that the increased cytokines and chemokines in the mutant mouse muscle/serum/hippocampal samples exhibit features of senescence-associated secretory phenotype (SASP) [47, 48]. We thus asked if APPswe+ muscles exhibit increased expressions of senescence marker proteins, p16Ink4a and p53. RT-qPCR analysis showed that both p16Ink4a and p53 were increased in all three muscles (TA, quadricepts, and soleus), but not in other tissues /organs of TgAPPsweHSA mice at age of 3-MO (Fig. 6A). However, at age of 6-MO, in addition to these muscles, the mutant hippocampus, but not other tissues/organs, showed elevated expression of p16Ink4a and p53 (Fig. 6B). These results suggest age-dependent and muscle and hippocampus selective cellular senescence. The increased muscle and hippocampal senescence in the mutant mice were further verified by Western blot analysis (Fig. 6C, D) and co-immunostaining analysis using antibodies against P53 (Fig. 6E–G and Fig. S12). Notice that the increased P53+ immunosignals were selectively detected in the mutant hippocampal hilus region, but not cortex of TgAPPsweHSA mice (Fig. 6E–G and Fig. S12), in line with the view for the hilus region to be a more vulnerable region in response to the stress induced by the muscle APPswe. In addition, we verified the cellular senescence phenotype induced by expressing APPswe in C2C12 cells (Fig. 6H–K). C2C12 cells (a muscle cell line) expressing APPswe-GFP, but not APPwt-GFP nor YFP showed increased SA-β-Gal (another marker for cellular senescence) (Fig. 6H. I), p16Ink4a, and p53 (Fig. 6J, K), indicating the specificity of the detrimental effect induced by the overexpression of APPswe. We next asked whether inhibition of senescence in TgAPPsweHSA mice could diminish the brain and behavior phenotypes. A combination of Dasatinib (D) and Quercetin (Q) was used to inhibit senescence, due to their well examined senolytic effectiveness in animal studies [49, 50]. TgAPPsweHSA mice at 3-MO were administered D + Q to then be subjected to phenotypic examinations at 6-MO (Fig. S13A). We first verified D + Q’s effect on muscle senescence in TgAPPsweHSA mice. As shown in Fig. S13B, C, muscles from TgAPPsweHSA mice treated with D + Q showed reduced expressions of senescence markers, p53 and p16Ink4a, confirming D + Q’s inhibitory effect. We also examined D + Q’s effect on various types of cells in TA muscles of TgAPPsweHSA mice. In addition to muscle fibers, muscle tissue/organ contain nerve terminals (e.g., NMJ-neuromuscular junction), adipocytes, macrophages, and blood vessels [51, 52]. Oil Red O-stained adipocytes in the mutant TA muscles were comparable to that of controls (data not shown); however, CD11b+ macrophages were significantly increased in the mutant muscles (Fig. S13D, E). Interestingly, D + Q treatments abolished the increase of CD11b+ macrophages (Fig. S13D, E), increased muscle fiber size, and reduced central nuclei+ degenerative muscle fibers (Fig. S13F, G). Furthermore, we examined D + Q’s effect on SASP-like factors in muscles of TgAPPsweHSA mice, which were largely reduced by the D + Q treatments (Fig. 7A). Together, these results suggest that muscle SnCs and the activated macrophages may contribute to the expression of these SASP factors as well as the sarcopenia-like muscle deficit. We then determined whether the hippocampal phenotypes in TgAPPsweHSA mice were diminished by D + Q treatments. Remarkably, the phenotypes including glial activation, elevated vessel associated microglia, increased SASP-like factors (e.g., Lif, Il5, Il15, Ccl9, and Cxcl9), and decrease of growth factors (e.g., Pdgfd and Bdnf) and PDGFRβ+ pericytes, in TgAPPsweHSA hippocampus were all diminished by D + Q treatments (Fig. S14A–F and Fig. 7B). Moreover, the impaired hippocampal DG neurogenesis in TgAPPsweHSA mice was restored by D + Q (Fig. S14G, H). We further measured serum SASP-like cytokines and chemokines, and growth factors (e.g., PDGF-BB) in TgAPPsweHSA mice with and without D + Q treatments. Notice that many cytokines (IL1b, IL3, IL4, IL7, IL23, TNFa, and TREM-1) and chemokines (CCL2, 3, 4, 4, 11, 12, 17, and CXCL10, 11, 13) were increased, and PDGF-BB was decreased in the serum samples of TgAPPsweHSA mice treated with Vehicle (Fig. S15 and Fig. 7C). Those changes were all normalized by D + Q treatments (Fig. S15 and Fig. 7C), suggesting that the systemic inflammation in TgAPPsweHSA mice is in large due to the APPswe-induced senescence and SASPs. Finally, we compared behavior responses in TgAPPsweHSA mice treated with or without D + Q. The depression-like behaviors by TST, FST, and SPT were also diminished in the mutant mice by D + Q treatments (Fig. 7D). Taken together, these results suggest that APPswe-induced senescence and SASPs are likely to prompt systemic and hippocampal inflammation, glial activation, and BBB leakage largely in the Hilus-hippocampus, which may underlie the depression-like behavioral phenotypes in 6-MO TgAPPsweHSA mice (Fig. 7E and Table 1). Here, we use the TgAPPsweHSA mouse model that selectively expresses APPswe in skeletal muscles and provide evidence for muscular APPswe’s contributions to sarcopenia-like deficit, as well as AD-relevant brain pathology. We further investigated the mechanisms underlying muscular APPswe’s detrimental functions. Our results, summarized in Table 1, lead to a working hypothesis depicted in Fig. 7E, in which, muscular APPswe promotes sarcopenia-like deficit, AD-relevant hippocampal pathology, and depression-like behavior likely due to the increased senescence and SASPs, which appear to be a driver for the systemic and hippocampal inflammation, and thus expedites hippocampus pathology and depression-like behaviors in TgAPPsweHSA mice. These observations thus reveal a link of sarcopenia with AD, and uncover a muscular APPswe to brain axis in AD development. How does APPswe in muscle cells induce brain/hippocampal pathology? We propose that APPswe-induced muscle senescence and SASPs may underlie its effects on the brain/hippocampus via systemic inflammation (Fig. 7E). Many SASP-like proteins were induced not only in APPswe+ muscles, but also in serum samples and hippocampus of TgAPPsweHSA mice (Figs. 5, S7, and S11). Regarding the systemic inflammation in TgAPPsweHSA mice, our results suggest that APPswe-induced senescence and SASPs in muscles appear to be a key contributor to this event. Many (19 over 30, ~63%) upregulated SASP-like factors are detected in both muscles and serum samples of TgAPPsweHSA mice (Fig. S11B). Many (29 over 42, ~66%) increased serum proteins in TgAPPsweHSA mice were also detectable in the Tg2576 mouse serum samples (Fig. 5F). Inhibition of senescence by its inhibitors (D + Q) abolished most of the increased inflammatory cytokines in the serum samples of TgAPPsweHSA mice (Fig. S15 and Fig. 7C). However, further correlation analyses showed a significant association of SASP-like factors’ expression levels in TgAPPsweHSA TA muscles with their hippocampus (Fig. S11E), but not with their serum samples (Fig. S11F). We speculate that such un-correlation may be due to the different muscle fibers with different phenotypes and different vulnerabilities to the APPswe (Fig. 1 and Fig. S1C), which could express different SASP-like factors or cytokines, and thus make the elevated SASP-like factors more complex in TgAPPsweHSA serum samples than those in TgAPPsweHSA TA muscles. We also speculate that the dramatic effect on the systemic inflammation by APPswe expression in muscles may be due to the abundant muscle tissues in the body, which account for 30–40% of a person’s body weight; and the consideration of muscles as a critical endocrine organ [53]. In addition, the hypothesis is also in line with our results that various types of muscles from TgAPPsweHSA mice showed increased senescence cells as early as 3-MO (Fig. 6 and Fig. S12). Expression of APPswe, but not APPwt, in C2C12 cells also increased senescence cells (Fig. 6H, I). These deficits occurred at the same age or earlier than the brain (largely hippocampus) phenotypes in TgAPPsweHSA mice (Figs. 1, 2 and Fig. S8). Moreover, inhibition of senescence in TgAPPsweHSA mice attenuated nearly all the hippocampal and behavior phenotypes (Fig. 7D and Fig. S14). In addition to the increased SASP-like proteins (largely cytokines and chemokines), there are reductions in a few of growth factors, such as Pdgfb and Bdnf (Fig. S11A). The reduced Pdgfb and Bdnf were detected not only in muscles, but also in the hippocampus, of TgAPPsweHSA mice (Fig. S7B). The decreased PDGF-BB was also observed in the serum samples of TgAPPsweHSA mice (Fig. 5H). Interestingly, the inhibition of senescence restored PDGF-BB levels in TgAPPsweHSA mice (Fig. S15E, F). These results suggest that APPswe induced muscle senescence not only increases SASP-like proteins but also reduces these growth factors, which may also contribute to hippocampal pathology, especially BBB deficit, in the mutant mice. In light of above observations, we speculate that the mutant muscle derived and increase pro-inflammatory cytokines (e.g., IL6) and decreased growth factors (e.g., BDNF and PDGF-BB) may play important roles in inducing cellular senescence, glial cell activation, and BBB deficit in the mutant hippocampus-in particular the Hillus region. Such a brain-region selective effect may be due to the abundantly expression of their receptors in hippocampal neurons, pericytes, and/or glial cells, which make hippocampus-Hillus region more vulnerable to the stress induced by these upregulated cytokines and/or downregulated growth factors. the receptors. In line with this view are reports that BDNF receptor-TrkB [54], PDGFRb [55], and cytokine receptor-check IL6’s receptor [56] are abundantly expressed in hippocampus. In addition, the hippocampal DG area has more neural stem cells, another feature making it more vulnerable to the stress-induced senescence. This view is also in line with multiple literature reports that link cellular senescence with muscle and brain aging, and various degenerative diseases, including AD [57–62]. Several papers also demonstrate the use of senolytic drugs to attenuate the disease progression in several AD animal models [63, 64]. However, how APPswe in muscles induce senescence and SASPs remains unclear. We hope to address this question in future studies. In summary, the results presented in this paper suggest a multi-cell and multi-organ model for AD development in which skeletal muscle cells may serve as a nidus of the disease. This study may reveal an important muscle-to-brain axis, where APPswe-induced muscle cell-senescence accelerates brain cell aging and neurodegeneration, opening new avenues to explore interactions between muscles and brain cells during AD development and progression. Supplemental information Supplemental table 2 Original western blots checklist

PMC9649618

Boris Tabakoff,Paula L. Hoffman

The role of the type 7 adenylyl cyclase isoform in alcohol use disorder and depression 10.3389/fphar.2022.1012013

28-10-2022

type 7 adenylyl cyclase (AC7),alcohol use disorder,depression,disease markers,medication development

The translation of extracellular signals to intracellular responses involves a number of signal transduction molecules. A major component of this signal transducing function is adenylyl cyclase, which produces the intracellular “second messenger,” cyclic AMP. What was initially considered as a single enzyme for cyclic AMP generation is now known to be a family of nine membrane-bound enzymes, and one cytosolic enzyme. Each member of the adenylyl cyclase family is distinguished by factors that modulate its catalytic activity, by the cell, tissue, and organ distribution of the family members, and by the physiological/behavioral functions that are subserved by particular family members. This review focuses on the Type 7 adenylyl cyclase (AC7) in terms of its catalytic characteristics and its relationship to alcohol use disorder (AUD, alcoholism), and major depressive disorder (MDD). AC7 may be part of the inherited system predisposing an individual to AUD and/or MDD in a sex-specific manner, or this enzyme may change in its expression or activity in response to the progression of disease or in response to treatment. The areas of brain expressing AC7 are related to responses to stress and evidence is available that CRF1 receptors are coupled to AC7 in the amygdala and pituitary. Interestingly, AC7 is the major form of the cyclase contained in bone marrow-derived cells of the immune system and platelets, and in microglia. AC7 is thus, poised to play an integral role in both peripheral and brain immune function thought to be etiologically involved in both AUD and MDD. Both platelet and lymphocyte adenylyl cyclase activity have been proposed as markers for AUD and MDD, as well as prognostic markers of positive response to medication for MDD. We finish with consideration of paths to medication development that may selectively modulate AC7 activity as treatments for MDD and AUD.

The role of the type 7 adenylyl cyclase isoform in alcohol use disorder and depression 10.3389/fphar.2022.1012013 The translation of extracellular signals to intracellular responses involves a number of signal transduction molecules. A major component of this signal transducing function is adenylyl cyclase, which produces the intracellular “second messenger,” cyclic AMP. What was initially considered as a single enzyme for cyclic AMP generation is now known to be a family of nine membrane-bound enzymes, and one cytosolic enzyme. Each member of the adenylyl cyclase family is distinguished by factors that modulate its catalytic activity, by the cell, tissue, and organ distribution of the family members, and by the physiological/behavioral functions that are subserved by particular family members. This review focuses on the Type 7 adenylyl cyclase (AC7) in terms of its catalytic characteristics and its relationship to alcohol use disorder (AUD, alcoholism), and major depressive disorder (MDD). AC7 may be part of the inherited system predisposing an individual to AUD and/or MDD in a sex-specific manner, or this enzyme may change in its expression or activity in response to the progression of disease or in response to treatment. The areas of brain expressing AC7 are related to responses to stress and evidence is available that CRF1 receptors are coupled to AC7 in the amygdala and pituitary. Interestingly, AC7 is the major form of the cyclase contained in bone marrow-derived cells of the immune system and platelets, and in microglia. AC7 is thus, poised to play an integral role in both peripheral and brain immune function thought to be etiologically involved in both AUD and MDD. Both platelet and lymphocyte adenylyl cyclase activity have been proposed as markers for AUD and MDD, as well as prognostic markers of positive response to medication for MDD. We finish with consideration of paths to medication development that may selectively modulate AC7 activity as treatments for MDD and AUD. “In the beginning….”, Blenkinsopp (2011) (which for cyclic adenosine 3′-5′ monophosphate (cyclic AMP, cAMP) and adenylyl cyclase, by the Western calendar, was 1957), Sutherland and Rall identified a chemical which was produced during incubation of liver “particles” (homogenates) with ATP, magnesium, glucagon, and epinephrine (Sutherland and Rall, 1957). They reported that “similar or identical” compounds could be isolated from heart, skeletal muscle, and brain. This was the first identification of the “second messenger” (Sutherland et al., 1968) molecule formally known as adenosine 3′-5′ monophosphate or cyclic AMP. In 1962, Sutherland, Rall, and Menon described “adenyl cyclase” the enzyme that catalyzed the synthesis of cyclic AMP from ATP (Sutherland et al., 1962). It soon became obvious that adenyl cyclase (now referred to as adenylyl cyclase) and cyclic AMP were critical intermediaries in the actions of a plethora of hormones and other first messengers which interacted with their cognate receptors and modulated adenylyl cyclase activity. The receptors that can modulate adenylyl cyclase activity are within the family of G-protein-coupled receptors (GPCRs). The discovery of guanine nucleotide binding proteins (G-proteins), which couple the GPCRs to adenylyl cyclase, is ascribed to Alfred Gilman and Martin Rodbell. Martin Rodbell showed in 1971 that the relay of a signal from a receptor on the exterior of a cell to the cell interior requires three functional units: 1) the receptor, 2) a “transducer” that utilizes GTP and, 3) an “amplifier” that generates a second messenger (Rodbell et al., 1971a; Rodbell et al., 1971b). The character of the transducer that interacts with adenylyl cyclase (the “amplifier”) was then described by the laboratory of Alfred Gilman and colleagues in 1980 (Schleifer et al., 1980). They isolated trimeric proteins (“G proteins” consisting of α, β and γ subunits) from brain that could restore coupling of receptors to adenylyl cyclase in mutant leukemia cells which lacked such proteins (Schleifer et al., 1980). Through such reconstitution, one could restore the responses of the mutant cell to hormones. Interestingly, in the Nobel Lecture by Rodbell on the occasion of the Nobel Prize to Gilman and Rodbell in 1994, it was stated, “In some common disease states the amounts of G-proteins in cells are altered. There can be too much or too little of them. In for example diabetes and in alcoholism there may be some symptoms that are due to altered signaling via G-proteins” (Rodbell, 1995). This statement bears some truth, but there is more to the story. Before proceeding, one should touch on how the second messenger cyclic AMP produces its effects within the cell. The initial discovery of an effector mediating the signal initiated by intracellular levels of cyclic AMP garnered yet another Nobel prize. This prize went to Edwin G. Krebs and Edward Fischer for their discovery and characterization of protein kinase A (Krebs et al., 1964; Walsh et al., 1968), and the description of protein phosphorylation cascades that are the final mediators of much of cellular function, from energy metabolism to gene transcription, to cell survival (Ahn et al., 1991). Although it was initially thought that protein kinase A was the sole mediator of cyclic AMP action, it is now evident that cyclic AMP acts through several effectors. The three most studied effectors are: 1) protein kinase A, 2) the exchange proteins activated by cyclic AMP (Epac) (Robichaux and Cheng, 2018), and 3) cyclic nucleotide-gated ion channels (CNG channels (van der Horst et al., 2020). More recently additional cAMP effector proteins have been identified including hyperpolarizing activated cyclic nucleotide-gated potassium channels (HCN1-4) (Brand, 2019; Santoro and Shah, 2020); the Popeye domain-containing proteins (POPDC proteins) (Brand, 2019); and cyclic nucleotide receptors involved in sperm function (CRIS) (Krähling et al., 2013). In addition, some isoforms of phosphodiesterases (PDE) which degrade cAMP are also regulated allosterically by cAMP (Omori and Kotera, 2007). In the late 1970s and early 1980s, the proposal that ethanol produced its neurobiological effects by perturbing the physical structure of neuronal membranes held dominance in the area of alcohol research. It struck us (Tabakoff and Hoffman, 1979) that the adenylyl cyclase system of brain would be a good test of how, and if, in neuronal membrane preparations, a disruption of membrane structure would translate into a perturbation of an important signal transduction system (i.e., adenylyl cyclase). By then the work of Sutherland, Rall and Gilman had demonstrated that at least three different membrane-bound protein components had to act in concert to modulate the production of cyclic AMP (this is not counting the fact that G-proteins were trimers of α, β and γ protein subunits). Thus we thought that the adenylyl cyclase system would be excellent for reflecting ethanol’s lipid perturbing properties. This review puts in historical context the work that established that ethanol, at concentrations found in the brain of inebriated individuals, can significantly alter adenylyl cyclase activity and the adaptive responses seen in the adenylyl cyclase signaling system with chronic exposure of the brain to ethanol. The discoveries of multiple isoforms of adenylyl cyclase disclosed that one particular isoform was most sensitive to ethanol’s actions, and genetic manipulation of the expression of this isoform revealed the biological context for this isoform’s actions within particular neuronal systems, such as the GABA neurons of the amygdala and nucleus accumbens, and the corticotropes of the pituitary. The effects of genetic manipulation of this isoform on the behavioral repertoire of a genetically modified animal, indicated that alcohol consumptive behavior, and behaviors associated with animal models of MDD, were related to the levels of expression of the alcohol sensitive adenylyl cyclase in brain, but these effects were influenced by the sex of the animal. Extrapolating from mouse to human, the levels of expression of the alcohol-sensitive adenylyl cyclase in brain or cells in blood, established that measures of this enzyme’s expression or activity could be considered state or trait markers of AUD or MDD. A more recent finding, demonstrating that the alcohol-sensitive adenylyl cyclase is the primary cyclase in cells of the immune system, opens another aspect in the biology of this enzyme isoform, and its relation to ethanol’s action in humans. An added facet to this observation is the dominant presence of this isoform in microglia in brain, and the possible implications of this fact on microglial activation, and the effects of ethanol on this aspect of brain function. We finish with some observations on the prospects of isoform-specific modulators of adenylyl cyclase activity, and the possibility of their use as medications in treating AUD and MDD. In our initial studies, we used cell membrane preparations of the striatum of mouse brain and measured the production of cyclic AMP (Luthin and Tabakoff, 1984). We found that concentrations of ethanol up to 500 mM had no effect on basal adenylyl cyclase activity. Only after the addition of Gpp (NH)p, the non-hydrolysable analog of GTP, to the assay mixture did ethanol, at concentrations as low as 50 mM, produce significant increases in production of cyclic AMP. This was the first indication that ethanol’s actions on adenylyl cyclase were related to the presence of the activated G protein in the assay, but the low ethanol concentration necessary to produce the measurable effect did not well support the hypothesis that ethanol was acting by perturbing membrane lipid structure. We followed these studies with the examination of the acute and chronic effects of ethanol administration on dopamine-stimulated adenylyl cyclase activity in striatal membrane preparations (Tabakoff and Hoffman, 1979). Ethanol, at concentrations of 50 mM, added to striatal membrane preparations from “ethanol naive” mice increased dopamine-stimulated adenylyl cyclase activity without changing the potency of dopamine. Mice were then chronically treated with ethanol to produce physical dependence, and sacrificed at various times over the ensuing 7 days following withdrawal. We noted that after ethanol was cleared from their systems (8–24 h), the response of striatal membrane adenylyl cyclase to dopamine was reduced in a time-dependent manner. The reduction in response became evident at a time when withdrawal signs were reaching their peak, and continued to decrease through the initial 24 h of the withdrawal period. The phenomenon was reversible, with the responsiveness to dopamine beginning to return to control levels by 36 h post withdrawal. The responsiveness to exogenously added ethanol remained intact during this period. Thus, ethanol could still increase the response of adenylyl cyclase to addition of dopamine, and at particular concentrations (50 mM), produce a response to dopamine that equaled the response to dopamine of the membranes from control mice, measured without the addition of ethanol. From these studies, we surmised that the withdrawal from a chronic ethanol feeding paradigm generated the diminution of the response of striatal adenylyl cyclase to dopamine, and the reintroduction of ethanol could “normalize” this function of dopamine. Rabin and colleagues (Rabin et al., 1980; Rabin and Molinoff, 1981) followed with an attempt to replicate our results. Their work demonstrated that the addition of ethanol (≥ 68 mM) to incubations containing striatal homogenates increased dopamine stimulated adenylyl cyclase activity, but when they treated mice chronically with an ethanol-containing liquid diet to produce physical dependence, and isolated striatal membranes from control and ethanol-fed mice 24 h after withdrawal, they found no difference in response to dopamine between the preparations from the ethanol-treated and control mice (Rabin et al., 1980). Interestingly, they replicated additional aspects of our prior studies such as the increase in levels of muscarinic cholinergic receptors in the striatum of the ethanol-fed and withdrawn mice (Tabakoff et al., 1979). Further research using rats chronically treated with intraperitoneal (ip) injections of ethanol, produced ambiguous results regarding dopamine-stimulated adenylyl cyclase activity in striatum of the ethanol-treated rats during withdrawal. Seeber and Kuschinsky (1976) found that 15 h after withdrawal, there was a “slight postjunctional subsensitivity to dopamine,” but the differences were not statistically significant. One of the variables contributing to the disparate findings regarding the effects of chronic treatment with ethanol and withdrawal on dopamine-stimulated adenylyl cyclase activity in the striatum, is the time of measurement of the enzyme activity after withdrawal. Our studies demonstrated that the differences become evident after some period after withdrawal, are evident during the first 24 h, and begin to disappear by 72 h after withdrawal (Tabakoff and Hoffman, 1979). The measurements at a single timepoint during the first 24 h after withdrawal in mice (Rabin et al., 1980), or rats (Seeber and Kuschinsky, 1976), may miss the optimal time to demonstrate the changes in the striatum. Such studies that will have to be better designed in the future, including a careful preparation of the cell membranes used in the analysis. As will be discussed later, the phosphorylation state of particular isoforms of the adenylyl cyclase is important in the measure of G-protein stimulated activity. Thus, a more careful assessment of the time course of changes in striatal dopamine-stimulated adenylyl cyclase activity, and more concern about factors that may contribute to ethanol-induced changes, continue to be warranted. It should be noted that there is consistent evidence that the changes in the dopamine-stimulated adenylyl cyclase activity in the striatum are not due to changes in the D1 dopamine receptors (Tabakoff and Hoffman, 1979; Rabin et al., 1980), and probably not due to changes in the quantity of the stimulatory G-protein α subunit (Tabakoff et al., 1995). A reason for giving emphasis to changes in dopamine function in the striatum of ethanol withdrawn animals, is the currently popular concept regarding allostasis and reward deficit that drive withdrawal-induced alcohol consumption (Koob and Volkow, 2016). If dopamine function in the striatum is compromised, and ethanol can normalize (increase) the response to dopamine effects, these factors could put biological context to the behavioral phenomena. Receptor-activated adenylyl cyclase activity has also been measured in other brain areas of animals chronically treated with ethanol (Saito et al., 1987). Isoproterenol (β-adrenergic receptor agonist)-stimulated adenylyl cyclase activity in cerebral cortical membranes was shown to be reduced after chronic treatment of mice with ethanol. These changes were normalized within 24 h after withdrawal (Saito et al., 1987). Rabin (1990a), however, found that there were no differences in isoproterenol-stimulated adenylyl cyclase when measured 20 h after withdrawal. Again, the changes seen in the cortical tissue follow a particular time course, and studies of the changes need to include measurement at several time points after withdrawal. Interestingly, Rabin (1990a), Rabin (1990b) did find that ethanol treatment of cells (cerebellar granule cells and PC12 cells) in culture for several days decreased the maximum activation of adenylyl cyclase in these cells by isoproterenol or 2-chloroadenosine. Studies with HEL cells in cultures containing ethanol also demonstrated ethanol-dependent reduction in PGE1-stimulated adenylyl cyclase activity (Rabbani and Tabakoff, 2001). It may be concluded that a down-regulation of adenylyl cyclase activity in certain areas of brain, and in immune/platelet cell precursors, does take place in animals or cells chronically treated with ethanol, but a careful and more extensive monitoring of the time course of events is necessary to further substantiate such phenomena. Given this observation, the mechanism of this phenomenon becomes of interest. Forskolin is a diterpene alkaloid which has been shown to bind to most forms of adenylyl cyclase, and radioactively-labeled forskolin has also been used to quantify adenylyl cyclase protein levels (Insel and Ostrom, 2003). Measurement of forskolin-stimulated adenylyl cyclase activity in cerebral cortical membranes of chronically ethanol-treated mice, demonstrated that the stimulation by forskolin had a similar potency in tissue from control and ethanol-treated mice, but the maximal effects were significantly lower in the ethanol-fed animals (Valverius et al., 1989). Autoradiographic analysis of 3H-forskolin binding across the various areas of mouse brain revealed differences between control and ethanol-fed animals in several brain areas (Valverius et al., 1989). It should be noted that in these studies, the animals were sacrificed while still intoxicated. Lower levels of forskolin binding were found in areas such as the cortical areas, nucleus accumbens, amygdala, hippocampus, and globus pallidus, while no significant differences were noted in the caudate putamen or cerebellum at this point in time. One can come to a conclusion that chronic ethanol administration to rodents produces diminutions in the expression of adenylyl cyclase protein in certain areas of brain, and the physiological phenomena accompanying the withdrawal from ethanol may produce changes in adenylyl cyclase activity in other brain areas. Overall, the early studies demonstrated that ethanol’s stimulation of adenylyl cyclase activity was dependent on the presence of G protein, and chronic exposure of cells in culture to ethanol resulted in a down-regulation of GPCR-mediated activation of adenylyl cyclase. The results with brain tissue taken from animals that had been chronically treated with ethanol are somewhat ambiguous, but the ambiguity comes mainly from the fact that the diminution in GPCR-stimulated adenylyl cyclase activity after ethanol treatment and withdrawal follows a particular time course, and the phenomenon may be missed if only one timepoint after withdrawal is studied. The down-regulation of GPCR-stimulated adenylyl cyclase activity by chronic exposure of the organism to ethanol may well be related to ethanol craving and CNS hyperexcitability that occur during withdrawal. The work described to this point was performed prior to the discovery that there were multiple forms of adenylyl cyclase. Gilman’s laboratory (Krupinski et al., 1989; Tang et al., 1991) reported on the first isoform of adenylyl cyclase aptly named the Type 1 adenylyl cyclase. Soon after Type 2 adenylyl cyclase was described (Feinstein et al., 1991). Type 3 adenylyl cyclase was first described by Bakalyar and Reed (1990), Type 4 by Gao and Gilman (1991), Type 5 by Ishikawa et al. (1992), Type 6 by Katsushika et al. (1992), Yoshimura and Cooper (1992) and Krupinski et al. (1992), Type 8 by Cali et al. (1994), and Type 9 by Hacker and Storm (1998) and Premont et al. (1996). All of these were membrane-bound forms of adenylyl cyclase and there was one form of adenylyl cyclase that was found to be cytosolic (Type 10) (Buck et al., 1999). Soon after the findings regarding the Type 6 adenylyl cyclase, we isolated a sequence from human erythroleukemia (HEL) cells that showed sequence similarities to other adenylyl cyclases, but also displayed characteristic functional differences (Hellevuo et al., 1993). After cloning the full length sequence, expressing the protein, and characterizing its activity, it became clear that this adenylyl cyclase was unique and was named the Type 7 adenylyl cyclase (Hellevuo et al., 1995). At the same time, Watson et al. (1994), isolated and characterized a similar adenylyl cyclase from rat brain illustrating that the Type 7 adenylyl cyclase (AC7) was present in rodents, as well as in human tissues. The characteristics of the discovered adenylyl cyclases were such that they could be fitted into 4 families (Devasani and Yao, 2022). AC7 joined the family containing the Type 2 and Type 4 adenylyl cyclases. This family was distinguished by its insensitivity to calcium with or without calmodulin, insensitivity to inhibition by the Gαi protein and by the stimulatory effects of phorbol esters acting through PKC, as well as the co-stimulation by the βγ subunits of the G proteins acting simultaneously with the Gαs protein (Yoshimura et al., 1996). The βγ subunits that act in concert with Gαs were found to be derived from Gi/o proteins coupled to GPCRs that are many times thought to be inhibitory to adenylyl cyclase activity (Yoshimura et al., 1996; Rhee et al., 1998). Thus, for example activation of opiate or cannabinoid receptors (coupled to Gi) simultaneously with activation of D1 dopamine receptors (coupled to Gs) further activates AC7 and the other members of its family. We found another feature that distinguished the AC7, i.e., ethanol could stimulate the activity of AC7 to a two-to-three times greater extent than any of the other adenylyl cyclases (Yoshimura and Tabakoff, 1995; Yoshimura and Tabakoff, 1999). An activated Gαs protein was still necessary to witness this effect of ethanol (Yoshimura and Tabakoff, 1995). AC7 was also the most responsive to activation by phorbol esters, in comparison to the other members of its family (Type 2 and Type 4), and the stimulation of AC7 by phorbol esters involved the presence of an activated Gαs (Yoshimura and Cooper, 1993; Hellevuo et al., 1995). It became of interest to consider whether ethanol and phorbol esters may be utilizing a similar pathway to accomplish the activation of AC7. Phorbol esters are known activators of members of the protein kinase C(PKC) family (Reyland, 2009). There are 10 known PKCs and 8 of them are activated by phorbol esters (classical PKCs: α, β1, β2 and γ, which are responsive to phorbol esters and diacylglycerol, and are dependent on calcium binding for their activity; novel PKCs: δ, θ, ε, η which are responsive to phorbol esters and diacylglycerol but insensitive to calcium; and atypical PKCs: ζ, λ/ι, which depend on binding of phosphatidylinositol 3, 4, 5-trisphosphate or ceramide for activation (Reyland, 2009; Newton and Brognard, 2017)). All of the PKC enzymes are processed by a series of ordered phosphorylations and conformational changes to attain a catalytically active form. The enzymes are maintained in an inactive state until the binding of the proper second messenger (in the case of PKCδ, for example, the second messenger is diacylglycerol) and a conformational change leading to a catalytically active, open, form of this enzyme is then attained (Newton and Brognard, 2017). At the time that involvement of PKC in the action of ethanol on adenylyl cyclase was being studied, little of the detail of the activation process for PKCs was known. However, through a series of studies dependent on the process of elimination, the PKC most likely to interact with AC7, and increase its activity, was found to be PKCδ (Nelson et al., 2003). Nelson et al. (2003) using HEL cells which contain predominantly AC7 (Hellevuo et al., 1993), demonstrated that PKCδ could phosphorylate AC7 protein, with the likely site of phosphorylation being the C1b domain of AC7 (Nelson et al., 2003). The catalytic conversion of ATP to cyclic AMP by adenylyl cyclases involves the juxtaposition of two domains of the enzyme protein. The C1a region of the intracellular loop between the membrane spanning domains, M1 and M2 has to align with the C2 region of the C terminal tail of the adenylyl cyclase protein to form the catalytic domain. The addition of Gsα activated by GTPγS to a mixture of the C1a region peptide of the Type 1 adenylyl cyclase and the C2 region peptide of the Type 2 adenylyl cyclase increased the enzymatic activity of this mixture (cyclic AMP production) well over a thousand-fold (Yan et al., 1996). The explanation for this increase in enzymatic activity is that the activated Gsα acts as a link between the two adenylyl cyclase fragments and aligns them into the proper conformation for catalysis. Analysis of the crystal structure of the C1a and C2 regions of adenylyl cyclase in combination with Gαs and forskolin demonstrated the binding of Gαs to the C2 region and interaction with C1a region resulting in a change in orientation of these regions to each other with the resultant increase in catalytic activity (Tesmer et al., 1997). Interposed between the C1a region and the M2 transmembrane domains is a region referred to as C1b and this region has been considered to be important for conferring isoform-specific regulatory properties to members of the adenylyl cyclase family (see references in Beeler et al. (2004)). A particularly interesting function of the C1b region is to modulate the ability of activated Gsα to promote the catalytic function of the C1a•C2 dimers (Scholich et al., 1997). Beeler et al. (2004) generated a recombinant protein representing the C1b region from AC7 (aa 506–584) and examined its effects on the catalytic function of the mixture of C1a and C2 regions from AC7. It was found that the C1b peptide inhibited the activation by Gαs of the mixture of C1a and C2 peptides derived from the AC7. The inhibition was only evident at the lower concentrations of Gαs and no effect was evident at higher concentrations (>2 μM) Gαs. The C1b region can be phosphorylated by PKC. Shen et al. (2012) demonstrated the phosphorylation of serine 490 and 543 in the C1b region of the Type 2 adenylyl cyclase (AC2) by PKC with resultant changes in response of the enzyme to Gαi and βγ. In the AC7 sequence, several PKC phosphorylation sites are evident in the C1b region, but serines 505 and 536 are most interesting since they exist in an area of alignment to a putative binding region for PKCδ on SRBC protein (a PKCδ binding protein) and phosphorylation of these serines by PKCδ (Izumi et al., 1997) in the SRBC protein has been demonstrated. The phosphorylation of serines 505 and 536 may well allow for a more productive interaction between the C1a and C2 domains. One can speculate that the effect of ethanol on the activity of AC7 is mediated by phosphorylation of serines in the C1b region of the AC7 enzyme. The phosphorylation could reduce the inhibition of high affinity Gαs binding by the C1b region, resulting in a greater catalytic response of AC7 to binding of Gαs in the presence of ethanol (Figure 1). There is a caveat to this explanation of ethanol’s actions on AC7. Yoshimura et al. (2006) produced chimeras of different regions of AC2 and AC7. When expressed in HEK 293 cells, which were also transfected with dopamine (D1) receptors, the chimera containing the C1b and M2 region of AC7 with the C1a and C terminal region (C2) of AC2 responded to ethanol potentiation of dopamine-stimulated activity as would AC2, while the chimera containing the C1b and M2 region of AC2 with the C1a and C2 region of AC7 responded to ethanol as would be expected for AC7 (3–4 times greater response). These results led Yoshimura et al. (2006) to conclude that the C1b region of AC7 was not important for ethanol’s action. This conclusion omits consideration of the fact that ethanol’s actions on AC7 are dependent on the presence of the activated form of Gαs and that the effects of the C1b region are not independent of the other domains of the adenylyl cyclase protein. The effect of the AC7 C1b may well be tuned to the specific sequences of the C1a and C2 regions which bind Gαs in particular adenylyl cyclases (Beeler et al., 2004; Shen et al., 2012). It was notable that the chimeras in which the C1b regions of AC7 and AC2 were combined with heterologous C1a and C2 regions of these enzymes had significantly lower (5–8 times lower) dopamine-stimulated activity than in their homologous environment, and the activity in the presence of ethanol was also low. In the end, Yoshimura et al. (2006) concluded that ethanol directly influences the interaction of C1a and C2 regions of particular adenylyl cyclases, but the mechanism of this effect was left open. The control of Gαs binding by phosphorylation of the C1b region of particular adenylyl cyclases, i.e., AC7, thus remains an attractive hypothesis. Finally, it should be noted that there is evidence that ethanol is not simply activating PKCδ to produce its effects on AC7. Rabbani et al. (1999) utilized HEL cells which naturally express AC7 to demonstrate that the effects of phorbol esters and ethanol were additive even though both the ethanol and phorbol ester effects were blocked by PKC inhibitors. This brings forth the possibility that the phorbol esters and ethanol act in a complementary manner with phorbol esters activating PKCδ, and ethanol enhancing the phosphorylation of particular substrates such as AC7. The C1b region of AC7 may be particularly sensitive to ethanol’s amphiphilic properties (Klemm, 1998) which could influence the secondary or tertiary structure of the C1b region (Beeler et al., 2004), and allow phosphorylation of serines in that region (Figure 1). At this point, the best characterization of ethanol’s action on AC7 is that it acts as a “conditional” stimulus, with its actions dependent on the presence of the activated Gαs, and additional work needs to be performed to clarify the molecular events attendant to ethanol’s potentiation of Gαs activating properties. Ethanol’s actions on AC7 activity are evident at concentrations of 50 mM (230 mg%) or higher in cell systems in which AC7 is naturally expressed (e.g., HEL cells), and such pharmacological considerations should be applied when evaluating the physiological implications of the effects of ethanol on adenylyl cyclase-related events. It can be noted that blood alcohol levels of 200 mg% and over are not unusual for individuals coming to emergency rooms or even driving (Maruschak, 1999; Allely et al., 2006). An important issue to consider when one is evaluating the physiological impact of the effects of ethanol on adenylyl cyclase activity is the fact that adenylyl cyclases exist in “microdomains” within a cell and it is the local concentration of cAMP that instigates the downstream consequences (Zaccolo et al., 2021). At this time, the effects of ethanol on levels of cAMP have not taken this fact into account, and the changes in cAMP concentrations have been measured on a whole cell level or within an incubation volume. Localized, and possibly quite significant effects of ethanol may be diluted by such experimental approaches. The generation of AC7 transgenic (TG) and heterozygous (HET) knock- down mice (Yoshimura et al., 2000; Hines et al., 2006), allowed for the qualitative assessment of the behavioral and physiological effects of AC7. The transgene used for generating the TG mice was the human form of AC7 under the control of a synapsin promoter (Yoshimura et al., 2000), while the HET mice were generated by homologous recombination with the deletion of exon 3 of AC7 (Hines et al., 2006). We were not able to produce the homozygous knock-out because the fetuses bearing the homozygous deletion died in utero on GD11 (Hines et al., 2006). The phenomenon of the fetus bearing two copies of the disrupted AC7 gene dying in utero was also noted more recently by Duan et al. (2010), highlighting the importance of AC7 in development. The initial choice of the biological systems, and then the behaviors to be examined, were based on the known involvement of adenylyl cyclase as an effector for dopamine D1 and D2 receptors and the corticotropin-releasing factor (CRF) receptors. An elegant addition to the work on involvement of AC7 in the functions of CRF in brain and pituitary, was the work of Duan et al. (2010) who used genetic manipulation of AC7 expression in cells of the peripheral immune system to demonstrate that AC7 was integral to the innate and adaptive responses of the immune system. CRF acting within the amygdala has been linked to depression and anxiety disorders in humans (Binder and Nemeroff, 2010), and to anxiety-like, and alcohol consumptive behaviors in rodents (Agoglia and Herman, 2018). CRF and CRF1 receptors also appear to be involved in alcohol withdrawal-induced anxiety and increased alcohol consumption in alcohol-dependent animals after withdrawal (craving?) (Agoglia and Herman, 2018). Marissa Roberto and her colleagues have examined the effects of ethanol on CRF-sensitive neurons in the central amygdala (CeA) (Roberto et al., 2021). CRF acting through the CRF1 receptor, which is coupled to Gs protein, can increase GABA release, and activate post synaptic GABA-A receptors. The increased release of GABA can be measured by the increases in inhibitory post-synaptic potentials (IPSPs) (Roberto et al., 2021). Ethanol or CRF added to the CeA slice preparations were shown to significantly increase the GABA-mediated IPSPs (Roberto et al., 2021). Using the CeA slices from the WT and AC7 HET mice, Cruz et al. (2011), showed that the IPSPs measured in the presence of CRF or ethanol were reduced or absent, respectively, in the preparations from the HET mice compared to the WT mice (Cruz et al., 2011). This led to the suggestion that AC7, which in part is located presynaptically (Mons et al., 1998), can be involved in the signaling initiated by the CRF1 receptor and culminating in release of GABA. There is prior evidence that CRF1 receptors couple to both AC7 and Type 9 adenylyl cyclase (Antoni et al., 2003), and the significant diminution of AC7 in brains of the HET knock-down mice may be responsible for the reduced effects of CRF and ethanol in the CeA slice preparations. Work by Bajo et al. (2008), had demonstrated that PKCε was also involved in CRF1 receptor-mediated and ethanol-potentiated GABA release in slices of the CeA. Recording of “basal” IPSP activity attributed to spontaneous GABA release was significantly increased in tissue from animals whose PKCε was disrupted by homologous recombination (PKCε−/−) (Khasar et al., 1999). Additionally, the CRF1 receptor- mediated enhancement of GABA release, as well as ethanol-mediated GABA release in the CeA slices, was blocked in tissue from the PKCε−/− mice (Bajo et al., 2008). There is a significant difference in the results obtained from AC7 HET mouse tissue versus the tissue from the PKCε−/− mice (Bajo et al., 2008; Cruz et al., 2011). The basal GABA release in the slices of the PKCε−/− mice was substantially increased, and thus the stores available for release by CRF or ethanol may have been depleted. In AC7 HET mice, there was no change in the basal release of GABA and thus an explanation based on depletion of GABA stores would not resonate with reduced effects of CRF and ethanol in the HET mice. The evidence for mechanistic differences in PKCε effects and the effects of AC7, thus do not contradict the evidence for PKCδ mediation of the interaction of Gαs and AC7 whether induced by receptor activation or by ethanol. A parsimonious reconciliation (Figure 2) of the involvement of both PKCε and the adenylyl cyclase system can be considered by invoking a cAMP to PKCε communication link. Such a link has already been established for excitatory transmitter release in the CNS (Gekel and Neher, 2008). Hucho et al. (2005) presented evidence that Epac is central for the activation and translocation of PKCε in neurons of the dorsal root ganglion, and that adenylyl cyclase activation via Gαs is the initiator of this cascade. Wang et al. (2022b) further elucidated the role of Epac-PKCε in the facilitation of docking and release of the contents of synaptic vesicles in parallel fibers of the cerebellum. If similar events are evident in GABAergic neurons (Robichaux and Cheng, 2018), then two related pathways (adenylyl cyclase/cAMP/PKA or PKCε-mediated) or one sequential pathway (adenylyl cyclase, Epac, PKC) could explain the effects of both AC7 and PKCε on modulation of CRF-mediated GABA release by ethanol. The effects of ethanol on CRF-mediated signaling (Figure 3) have been further investigated using CRF-mediated ACTH release in the pituitary of the HET knock-down and TG mice overexpressing AC7. Assessment of the forms of adenylyl cyclase present in the mouse pituitary indicated the presence of the Type 2, Type 3, Type 6, and Type 7 (Pronko et al., 2010). It should be noted that Type 9 adenylyl cyclase has been reported to be present in rodent corticotropes (Antoni et al., 2003), but was not found in the mouse pituitary using microarray analysis (Pronko et al., 2010). In the WT, AC7 HET, and AC7 TG mice (Pronko et al., 2010), the most profound differences were noted in the plasma ACTH levels of the male and female mice after injection of ethanol (Pronko et al., 2010). Significant quantitative differences among WT, HET and TG mice were found in both the peak levels and AUC of the ACTH responses to injection of ethanol (these values well surpassed the levels seen after saline injection). The levels of corticosterone correlated in magnitude with plasma ACTH levels after ethanol injection. The rank order of the plasma ACTH and corticosterone levels after ethanol injection was AC7 HET < WT < TG. In all cases, female mice had higher levels of corticosterone than the males of that genotype (Pronko et al., 2010). The results of these studies again establish AC7 as an important component of the link between the CRF1 receptor, and the downstream consequences of its activation, but the differences in the corticosterone response between males and females are not explained by differences between sexes in expression of AC7 in the pituitary. The genetic manipulation of the Adcy7 gene produced similar levels of AC7 RNA in the pituitary of the male and female mice of the HET or TG genotypes (Pronko et al., 2010), and the protein levels for AC7 followed the same pattern as the RNA levels with no statistically significant differences between males and females (Pronko et al., 2010). In the WT and AC7 HET mice, the higher levels of corticosterone in females may reflect higher ACTH levels. However, ACTH levels did not differ significantly between AC TG male and female mice. One explanation of the lack of sex differences in AC7 in the pituitary, but significant sex differences in the corticosterone response of AC7 TG mice, is the observation that the adrenal tissue of females may be more responsive to ACTH than that of males (Rao and Androulakis, 2017), and at a particular level of ACTH more corticosterone would be released from the adrenals of females. The probable involvement of AC7 in the CRF1 receptor-mediated release of ACTH brings into further consideration the importance of microdomains in the actions of ethanol on AC7. AC7 is part of what was referred to as a “signalosome” consisting of Gα12, AC7, PDE3B, PKA and other kinases organized around AKAP13 on the endoplasmic reticulum (Zaccolo et al., 2021). This type of signalosome has been shown to be important in the regulation of secretory function of the endoplasmic reticulum for recently synthesized and properly folded proteins (Subramanian et al., 2019). The presence of Gα12 in this signalosome complex is consistent with the presence of AC7 since Jiang et al. (2008) demonstrated that AC7 is a specific downstream target of the Gα12/13 subunits that produce an increase in AC7 activity (Figure 3). The involvement of the “ethanol sensitive” AC7 in the ACTH/corticosterone response to ethanol administered in vivo, helps explain a seeming enigma with regard to responses to imbibed ethanol. Ethanol is considered an anxiolytic drug, but several reports have provided evidence that ethanol ingestion generates an increase in the circulating levels of cortisol (stress hormone) in humans. Since the anxiolytic and the cortisol elevating effects of ethanol can arise by different mechanisms and involve different areas of brain (Pronko et al., 2010; Olsen and Liang, 2017), these results can be quite compatible. The DARPP-32 signaling pathway has been proposed as a therapeutic target for AUD medication development (Greener and Storr, 2022). The phosphorylation of glutamate receptors (NMDA, AMPA) on the medium spiny neurons of the nucleus accumbens is the major event which controls the strength of excitatory input to these neurons. The actions of DARPP-32 are integral in controlling the phosphorylation state of NMDA and AMPA receptors and DARPP-32 function is itself controlled by phosphorylation/dephosphorylation events (Nishi et al., 2005). The medium spiny neurons of the nucleus accumbens are the integrators of dopaminergic signals from the ventral tegmentum and glutamatergic signals from the pre-frontal cortex (Figure 4) and play an important role in mediating the reinforcing/rewarding effects of addictive drugs (Surmeier et al., 2007; Allichon et al., 2021). An important component of this integration is dopamine D1 receptor-mediated generation of cyclic AMP, the activation of PKA, the phosphorylation of DARPP on residue threonine 34 (T34 Phospho-DARPP), the inhibition of protein phosphatase-1, and the maintenance of the ionotropic glutamate receptors in their phosphorylated state (see Figure 4). Donohue et al. (2005) used the AC7 TG mice to study the phosphorylation of the DARPP-32 protein on the threonine-34 residue in the nucleus accumbens, caudate/putamen, and amygdala. The effects of ethanol administered in vivo on tissue obtained from these brain areas were also examined. In the brains of AC7 TG mice and WT mice, no differences in total levels of the DARPP-32 protein were evident in any of the tested brain areas (Donohue et al., 2005). In the amygdala and caudate/putamen of saline-treated (control) WT mice, the levels of T34 Phospho-DARPP were significantly lower than those in the saline-treated AC7 TG mice. Interestingly, just the opposite was true in the nucleus accumbens. The acute administration of ethanol, in vivo, increased the levels of T34-Phospho-DARPP in all brain areas of the WT mice. But, only in the amygdala was the effect of the transgene evident. In the amygdala of the TG mice, the administration of alcohol significantly increased the levels of T34-Phospho-DARPP beyond those produced by saline or by the same dose of ethanol in WT mice. A somewhat similar experiment was performed by Bjork et al. (2010), in which ethanol was administered to C57BL/6 mice and levels of T32-Phospho-DARPP were measured in the “striatum,” an area including the nucleus accumbens and amygdala. Ethanol administration produced a “robust” increase in T32-Phospho-DARPP which was blocked by the dopamine D2 receptor antagonist, sulpiride. The effect of ethanol was also blocked by administration of naloxone given prior to the administration of ethanol. The results with the dopamine D2 receptor antagonist, and the opiate receptor antagonist naloxone, do indicate more complexity to the phosphorylation of DARPP in the “striatum” than a simple activation of the D1 dopamine receptor to initiate the phosphorylation cascade. A possibility not considered by Bjork et al. (2010) was that the presence of AC7 in the “striatum” would offer the opportunity for activation of dopamine D2 or opiate receptors to potentiate the activity of AC7 through release of βγ subunits from the Gi/Go trimers (Yoshimura et al., 1996). Naloxone administration would block the opiate/D1 dopamine receptor additive effect (via Gsα and βγ) on AC7. The contention that opiates are acting in the striatum by coupling to Gi/Go proteins to release βγ, and produce additional activation of AC7, is further supported by the work of Karlsson et al. (2016). In this work, the role of melanin-concentrating hormone in modulating ethanol-induced conditioned place preference (CPP) was investigated. This included measurement of DARPP-32 phosphorylation in WT and MCH1 receptor knock-out mice. Administration of ethanol produced a significant increase in T32-Phospho-DARPP in the shell region of the nucleus accumbens in the WT mice, but not in the MCH1 knock-out mice. The MCH1 knock-out mice also showed diminished propensity to develop ethanol-induced CPP compared to WT mice. The results seen with mice carrying the deletion of the MCH1 gene could be replicated in WT mice by the use of a MCH1 receptor antagonist (Karlsson et al., 2016). An important consideration in interpreting these results is that the MCH1 receptor is a Gi/Go-coupled receptor (Lembo et al., 1999), which is co-expressed with dopamine receptors on medium spiny neurons (Zhang et al., 2006). Again, the generation of βγ subunits upon activation of the MCH1 receptor, could act in concert with Gsα to produce an accentuated activation of AC7, and more robust generation of cAMP. An interesting conclusion can arise from data on ethanol’s effects on DARPP phosphorylation, as well as the above-described studies on CRF-mediated GABA release in the central amygdala. One can speculate that, in neurons of the limbic system, AC7, which is responsive to both ethanol and to βγ subunits, would be the mediator of ethanol and βγ subunit effects on phosphorylation cascades in these neurons. The PKA- and Epac-mediated events downstream of the activity of AC7 would set the tone for both metabolic and neurotransmission functions in these neurons. Duan and colleagues (Duan et al., 2010) bred mice in which one allele of the Adcy7 gene was disrupted, and although most of the offspring that were double mutants (AC7 −/−) died in utero, approximately 2–3% survived through birth. The bone marrow from these AC7 knockout animals and bone marrow from their wild type littermates was isolated and transplanted into mice whose immune system had been destroyed by irradiation. The immune system of the recipient mice fully regenerated, producing chimeric mice bearing the donor bone marrow cells. The total number of splenocytes was reduced by more than half in the chimeric mice generated from the bone marrow of the AC7−/− mice, indicating the importance of AC7 in proliferation of both B and T lymphocytes. On the other hand, when challenged with LPS, the chimeric mice with the AC7−/− bone marrow generated a 3–4 times greater TNFα response compared to the mice which received the wild type bone marrow. LPS was also more lethal in the mice carrying the AC7-deficient bone marrow. Macrophages from the AC7-deficient mice produced significantly higher levels of TNFα when challenged with LPS in vitro, compared to mice carrying the wild type bone marrow (this response also involved yet-to-be-identified serum factors). Duan et al. (2010) also demonstrated that AC7 was necessary for an optimal antibody response when mice were exposed to antigens. The deficiency in the AC7−/− chimeric mice was primarily due to AC7-dependent function of the T helper cells, even though B cell function was also disrupted in the animals with AC7−/− bone marrow. In conclusion, Duan et al. (2010) state: “…AC7 is the key AC isoform in mediating cAMP response and its downstream physiological functions in the immune system”. Table 1 summarizes the neurobiological phenotypes elucidated in mice in which the expression of AC7 was manipulated. Given the electrophysiological, neurochemical, and physiological results of studies with the HET, WT, and TG AC7 mice, these animals were used for behavioral measures of ethanol consumption and measures of anxiety-like and depressive phenotypes. In measures of alcohol consumption and preference, AC7 HET mice on two genetic backgrounds (C57BL/6 and 129/SvEv) were used. C57BL/6 mice normally show a high preference for alcohol-containing solutions, and using the HET mice on the C57BL/6 background, no statistically significant differences were found between the HET and WT mice in the quantities of ethanol consumed by males or females. When AC7 was knocked down in the 129/SvEv strain, which drinks low to moderate amounts of ethanol, the females of the HET genotype consumed more ethanol, particularly at the higher concentrations of 10 and 20%. This increase in ethanol consumption resulted in a higher calculated “preference” for ethanol when water intake was taken into account (Desrivières et al., 2011). In the male HET mice on the 129/SvEv background, the amount of ethanol consumed at the highest concentration was actually less than that consumed by the WT mice, and there was no statistically significant change in the preference measure (Desrivières et al., 2011). The measure of what is called “preference” has an important concept attached. A “preference” ratio of 0.5 indicates neutrality of choice between the alternatives of the ethanol solution or plain water, while a preference ratio above 0.5 indicates a greater desire for the ethanol solution, and a preference ratio of less than 0.5 indicates an aversion to the ethanol solution. In the case of the HET female mice on the C57BL/6 background and their corresponding WT littermates, all preference ratios were above 0.8, irrespective of the status of AC7 (even though there was an evident increase to almost 1.0 in the HET mice at the lower ethanol concentrations). In the WT females of the 129/SvEv background, the “preference” for the 10 or 20% ethanol solutions was approximately 0.15 and in the HET females the ratio increased to 0.3. These results can be interpreted as indicating that on the 129/SvEv background, the diminution of AC7 in female mice, diminishes the aversion to consuming ethanol solutions containing the higher concentrations of ethanol. Measures of behavior which is interpreted as “depressive” or “learned helplessness” were also performed in HET, WT, and TG AC7 mice (Hines et al., 2006). Using the forced swim test (FST), female HET mice were shown to exhibit a significantly lesser time of being immobile during the duration of this test compared to the female WT mice (i.e., less “depression”). AC7 TG female mice were, on the other hand, found to exhibit longer periods of immobility than the female WT mice (more “depression”). There were no differences in the immobility time of male WT mice versus male TG, or male HET mice in the FST. In the tail suspension test (TST), female HET mice did not differ in immobility from WT mice, but the female TG mice showed greater periods of immobility. Again, there were no differences in immobility time in the TST between male HET, WT, and TG mice. These results indicate that in females, the overexpression of AC7 results in a higher level of “learned helplessness” (depressive-like behavior), while a reduction in AC7 expression produces greater resilience to depressive-like behaviors. A major caveat to this simple explanation, is the fact that the knock-down of AC7 in the HET female mice resulted in significant changes in gene expression of 30 other transcripts, and there were no changes in other transcript expression in the male mice. One of these transcripts in female strains, peroxiredoxin, has been implicated in behavior in the FST and TST (Scotton et al., 2020). Table 1 summarizes the behavioral and physiological phenotypes elucidated in mice in which the expression of AC7 was manipulated. The results with genetic manipulation of AC7 expression in mice qualified this adenylyl cyclase as a possible candidate gene for a genetic contribution to human AUD (Boezio et al., 2017) and/or MDD. Desrivières et al. (2011) examined single nucleotide polymorphisms (SNPs) within the ADCY7 gene in humans for association with alcohol dependence (defined by DSM-IV and ICD-10 criteria). The subjects consisted of 1,703 individuals classified as alcohol dependent and 1,347 controls, and both men and women were included in this Caucasian population. A SNP (rs2302717) that defined a haplotype across a portion of AC7 gene (ADCY7) was found to be associated with alcohol dependence, but this association was only significant in the females. The minor allele (T) at this locus reduced the risk to develop alcohol dependence (OR = 0.71). Desrivières et al. (2011) noted that the haplotype identified by this SNP extended into the promoter region of ADCY7, and performed an analysis of the quantity of RNA for AC7 that was present in whole blood or adipose tissue from another large sample of Caucasian individuals. These studies revealed that the minor allele of rs2302717 correlated with lower ADCY7 expression in both tissues. This was seemingly at odds with the data from the studies with mice (Desrivières et al., 2011), in which knock-down of Adcy7 and resultant diminution of AC7 RNA in brains of females resulted in less aversion to drinking, while in humans, a polymorphism that was protective against alcohol dependence was also responsible for lower levels of AC7 RNA. An obvious caveat is that the RNA measures in the HET mice were made in brain tissue while the human AC7 RNA was measured in blood and adipose tissues (Desrivières et al., 2011). The other caveat, already mentioned above, is the fact that the knock-down of Adcy7 in female mice results in changes in expression of a number of other transcripts in brains of the HET female mice, and this phenomenon will have to be explored in future studies, possibly in postmortem tissue of humans. Given the finding that an allele that is protective against alcohol dependence in women, is also associated with lower levels of AC7 RNA in human blood, it is instructive to review a number of studies which measured adenylyl cyclase activity in human platelets and lymphocytes of alcoholics and control (non-alcoholic) subjects. The first of such studies (Tabakoff et al., 1988) included 95 alcoholic subjects and 33 controls, and the majority of the alcoholic subjects (all except 5 who had been abstinent by self-report for 12–48 months) were abstinent for 23 ± 16 days. All of the subjects were male. Measures of platelet adenylyl cyclase activity demonstrated no differences in basal activity, but significant differences between alcoholic and control subjects in cesium fluoride-, Gpp (NH)p-, and prostaglandin 1 (PGE1)-stimulated adenylyl cyclase activity, with the alcoholic subjects having lower stimulated adenylyl cyclase activity. At the time of this study, the various isoforms of adenylyl cyclase had not yet been described, but currently, it is known that AC7 is the dominant form of adenylyl cyclase in both platelets and lymphocytes (Hellevuo et al., 1993; Duan et al., 2010). Given the earlier discussion regarding the role of PKCδ in promoting the activation of AC7 by Gs protein, the presence of PKCδ and its significant physiological function in platelets, and the fact that all stimulatory agents used in the study of Tabakoff et al. (1988), were acting via the Gs protein to activate adenylyl cyclase, one cannot distinguish the effects as being related to upstream effects involving PKCδ, or to adenylyl cyclase per se. An additional observation made in this study, was that the five alcoholic individuals who had abstained for over 12 months, still displayed lower cesium fluoride-stimulated adenylyl cyclase activity. This led to the suggestion that the stimulated adenylyl cyclase activity in platelets may be a “trait” rather than a “state” marker in alcoholism (Tabakoff et al., 1988). A different conclusion regarding adenylyl cyclase activity measured in lymphocytes of alcoholic and control subjects was provided by Szegedi et al. (1998). These investigators followed the adenylyl cyclase activity in lymphocytes of 73 alcohol-dependent subjects at admission to the clinic while intoxicated, at the time of maximal withdrawal signs, and after detoxification. Lymphocyte adenylyl cyclase activity of the alcohol-dependent subjects was also compared to control subjects. Their findings indicated that there were no differences in lymphocyte adenylyl cyclase activity between the control subjects and the alcohol-dependent subjects at admission, while the dependent subjects were intoxicated, but 2 days later basal, GTPγS-stimulated, and forskolin-stimulated adenylyl cyclase activity were significantly lower in the alcohol-dependent subjects going through withdrawal. After the withdrawal period, there again was no difference in adenylyl cyclase activity in lymphocytes of the control and alcohol-dependent subjects. The time course of changes in lymphocyte adenylyl cyclase activity in the studies of Szegedi et al. (1998) with humans, mirror the changes described in the striatum (Tabakoff and Hoffman, 1979) and cerebral cortex (Saito et al., 1987) of groups of mice chronically fed ethanol, during the early withdrawal period, and also several days after withdrawal. The earliest publication to note the differences in adenylyl cyclase activity (decreased adenosine (A2) receptor-mediated cyclic AMP production) in lymphocytes was that of Diamond et al. (1987). The alcohol-dependent subjects in that study were individuals described as “actively drinking” but having little or no alcohol in blood when blood was taken for isolation of lymphocytes. Thus, these subjects would resemble the “withdrawal” group in the studies of Szegedi et al. (1998). The changes described by Diamond et al. (1987) were evident in both B and T cells in the lymphocyte fraction, and more recent evidence examining the isoform of adenylyl cyclase in T and B cells, as well as macrophages, has identified the major adenylyl cyclase in these cells to be AC7 (Duan et al., 2010). Overall, the measurement of AC7 in lymphocytes may be advantageous for extrapolating to the activity of AC7 in brain of individuals undergoing withdrawal from chronic heavy alcohol consumption. It should be noted that the lower levels of adenylyl cyclase activity in platelets of alcoholics may also be a result of lowering of AC7 expression. But this change of expression would have to take place in the megakaryocytes, which are the precursors of platelets, since platelets do not contain DNA. Thus, at the least, the platelet measures of adenylyl cyclase activity would follow a time course more related to the time course of platelet turnover in blood, rather than a time course for changes in AC7 expression in cells replete with DNA and expression/translation machinery. The platelet adenylyl cyclase activity measured in alcoholics may be confounded by other variables, particularly by the presence of comorbid MDD (Hoffman et al., 2002). In fact, the platelet adenylyl cyclase activity may be a trait marker for MDD which is in turn confounded by current alcohol use by the depressed subject (Hines and Tabakoff, 2005). The initial studies of platelet adenylyl cyclase activity in depressed subjects indicated that forskolin-stimulated adenylyl cyclase activity was particularly lower in individuals diagnosed with MDD, compared to control subjects (Menninger and Tabakoff, 1997). Since forskolin acts directly on the adenylyl cyclase protein to enhance activity (Seamon et al., 1981), one can surmise that depressed subjects have reduced quantities of adenylyl cyclase protein in platelets, and since there is evidence that the major form of adenylyl cyclase in platelets is AC7, one can go further to consider that depressed subjects have lower levels of AC7 in platelets. This supposition was strengthened by the work of Hines et al. (2006), which also proposed an explanation for the lower levels of AC7 in platelets of humans suffering from depression. Hellevuo et al. (1997) demonstrated that the AC7 gene in humans is characterized by a series of polymorphic repeats in the 3′-UTR. The findings of Hines et al. (2006) indicated that the lowest levels of forskolin-stimulated adenylyl cyclase activity in platelets were in depressed subjects whose DNA in the 3′-UTR harbored the longest stretch (seven repeats) of the tetranucleotide AACA (Hellevuo et al., 1997). Some other observations generated by the work of Hines et al. (2006) were: the most prominent diminution in forskolin-stimulated platelet adenylyl cyclase activity was noted in depressed subjects who also had a family history of depressive illness; females diagnosed with MDD with a family history of depression; and in individuals having a genotype for the seven repeats of AACA. Through a combination of studies on gene expression and informatics using AC7 TG and WT mice, AC7 was linked to function of the proopiomelanocortin (POMC) system and immune system function. Clearly there is a link between the POMC transcript, stress, and the immune system, since POMC is the precursor to ACTH, and pituitary ACTH release, instigated by CRF, stimulates release of adrenal glucocorticoids, and modulates the activity of the immune system (Leistner and Menke, 2018). Chronic stress, in conjunction with childhood trauma, has been considered a significant contributory factor to the development of major depression (Heim and Nemeroff, 2001). Although the relationship of stress and depression has been considered to arise via the activation of corticosteroid receptors in brain (Holsboer, 2000), with polymorphisms in FKBP5 (a co-chaperone for the glucocorticoid receptor) being an important component of this relationship (Binder et al., 2004), the above described function of AC7 in control of CRF-mediated ACTH release (Antoni et al., 2003; Pronko et al., 2010) should also be considered in the etiology of depression. Furthermore, it is now becoming evident that AC7 is the major form of adenylyl cyclase in the immune system, and controls activation of macrophages, as well as B and T lymphocytes (Jiang et al., 2008; Duan et al., 2010). Recent studies also indicate that AC7 is the major form of adenylyl cyclase expressed in mouse and human microglia (Bennett et al., 2016; Galatro et al., 2017). Microglia are considered the “macrophages” of the CNS, and it is not surprising that AC7 is expressed in microglia. (Table 2 shows the expression levels of the various isoforms of adenylyl cyclase in microglia). Cyclic AMP levels are important for conversion of microglia from the M1 to M2 phenotype (M1 describes a proinflammatory phenotype, and M2 an activated but reparatory phenotype) (Ghosh et al., 2016). The relationship of stress to microglial activation is well summarized in Yirmiya et al. (2015) and these authors propose that some forms of MDD may be a “microglial disease,” dependent on microglia transitioning to the M1 phenotype. In all, the involvement of AC7 in the CRF-mediated release of ACTH from the pituitary, and involvement in microglia activation status, may play an important role in the etiology of depression. Assuming that AC7 is mediating the CRF-stimulated ACTH release, and the stress response is of consequence in the etiology of MDD, the sex differences described earlier in the ACTH and glucocorticoid responses in the WT mice, versus those with genetically modified expression of AC7, are notable (Pronko et al., 2010). Further evidence for the involvement of AC7 in depression emanated from the laboratories of Etienne Sibille (Joeyen-Waldorf et al., 2012). This group used mice lacking the serotonin transporter (SERTKO), which have been considered to be a model for studying depressive behaviors and emotionality (Lira et al., 2003), to assess gene expression in amygdala and cingulate cortex. They then compared the differentially expressed transcripts noted between the SERTKO and WT mice to differentially expressed transcripts noted in postmortem samples of amygdala and cingulate cortex from humans with familial MDD, and matched controls. “Conserved changes” were found for 31 transcripts in the amygdala, and 20 transcripts in cingulate cortex in comparisons of the mouse and human brain samples, and the transcript for AC7 was found to be significantly upregulated in the brain tissue from the SERTKO mice, compared to the WT controls, and in the brain tissue of depressed subjects compared to their matched controls. Their further studies examined (using BOLD (MRI)) threat-related amygdala reactivity in two independent samples of human subjects and its association with a single nucleotide polymorphism in the ADCY7 gene. This SNP (rs1064448) has previously been shown to identify a haplotype including a major portion of the ADCY7 and the 3′UTR containing the tetranucleotide repeats (Hines et al., 2006). In both samples, there was a significant association of rs1064448 with greater threat-related amygdala reactivity (Joeyen-Waldorf et al., 2012). These studies illustrate the possible importance of ADCY7 in fear-related amygdala function, and “conserved changes” in the expression of AC7 transcript in brain tissue from mice used as a model of depressive behavior (Joeyen-Waldorf et al., 2012), and in human subjects diagnosed with MDD, amplify the studies of Hines, et al. (2006). The SERTKO mice exhibited higher levels of expression of AC7 in brain tissue, as did the post-mortem tissue of the depressed human subjects, and in the studies of Hines et al. (2006), it was the genetically manipulated mice with lower levels of AC7 in brain that exhibited the lesser depressive-like behavior in the FST, and the female animals with the higher expression of AC7 showed higher immobility in the FST. It is parsimonious to think that deletion or pharmacologic blockade of SERT is coupled to upregulation of AC7 RNA, but this implication of the relationship of SERT and AC7 expression in development or treatment of depression does not appear straightforward. One has to be careful in making generalizations from the results obtained with the SERTKO mice produced on the 129/SvEv genetic background since the same genetic manipulation produced no effect in the C57BL/6 mice (Lira et al., 2003). The effect of antidepressants on the activity and possibly the expression of adenylyl cyclase in brain may or may not be reflected in measures of adenylyl cyclase activity in platelets since platelet adenylyl cyclase activity was found to be lower in depressed subjects compared to controls (Hines et al., 2006). A more recent study of platelet adenylyl cyclase activity stimulated by PGE1 (via the GS-coupled prostaglandin EP1 receptor) also demonstrated that subjects diagnosed with MDD had significantly lower PGE1-stimulated platelet adenylyl cyclase activity than control subjects (Targum et al., 2022). This study, however, followed a subset of subjects through a 6-week period of treatment with antidepressants which were primarily inhibitors of SERT (SSRIs). In the subjects that showed significant improvement in their Hamilton Depression Ratings (Ham D17 and Ham D6), there was also a significant increase in their PGE1-stimulated adenylyl cyclase activity toward levels measured in control subjects. Targum et al. (2022), therefore, replicated the lower adenylyl cyclase activity in platelets of clinically depressed subjects (Hines et al., 2006), but also added the fact that the platelet adenylyl cyclase can be not only a marker for depression, but also for measuring response to antidepressants. Targum et al. (2022) also proposed a mechanism for their observed results to be the sequestration of the Gs protein in lipid rafts (Allen et al., 2009) in the platelets of the depressed subjects and suggested that antidepressant treatment would result in the release of Gs from sequestration to be available for stimulation of adenylyl cyclase. Unfortunately, forskolin-stimulated adenylyl cyclase activity was not measured in the studies of Targum et al. (2022) to distinguish between the proposed mechanism, and the diminution of the adenylyl cyclase protein as proposed in other studies (Hines et al., 2006). There may well be different mechanisms in play which result in higher levels of the RNA for AC7 in brain in conjunction with signs of depression in genetically manipulated mice, and depressed human subjects (Hines et al., 2006; Joeyen-Waldorf et al., 2012), and the lower levels of adenylyl cyclase activity (presumably AC7), activated by various means, in platelets of depressed humans. Although forskolin can enhance adenylyl cyclase activity independent of other factors (Seamon et al., 1981), and forskolin (radioactively labeled), can be used to quantify adenylyl cyclase protein (Insel and Ostrom, 2003), there is clear evidence that Gsα can further activate adenylyl cyclase catalytic function (Insel and Ostrom, 2003). Thus, the lower levels of adenylyl cyclase activity, stimulated by forskolin or agents acting via Gs proteins, in platelets of depressed subjects may be either a result of lower levels of the adenylyl cyclase protein, a sequestration of Gs protein, or both mechanisms. Whether the proposed mechanism involving the sequestration of Gs in platelets (Targum et al., 2022) in depressed subjects extends to brain (Senese and Rasenick, 2021), bears scrutiny. AC7 is a member of the sub-family of adenylyl cyclases (Type 2, 4, and 7) whose activity is insensitive to Giα proteins, is potentiated by the βγ subunits of G proteins in conjunction with Gsα stimulation, and whose responsiveness to Gsα is modulated by the state of phosphorylation catalyzed by PKCδ. This enzyme is also insensitive to calcium in the presence or absence of calmodulin. The distinguishing feature that separates AC7 from the Type 2 and Type 4 adenylyl cyclases is the particularly high level of activation of this enzyme by ethanol when the enzyme activity is also influenced by Gsα. AC7 also has a cellular/tissue distribution that distinguishes it from the other members of its sub-family. Particularly notable is the evidence for its presence in the amygdala, nucleus accumbens, hippocampus, and frontal cortical regions in brains of animals, with evidence for presynaptic and postsynaptic localization (Mons et al., 1998), and its presence in the corticotrophs of the pituitary. The presence of AC7 in the pituitary, and its involvement in the release of ACTH, speaks to the possible importance of AC7 in the hypothalamic/pituitary/adrenocortical response to stress. The presence of AC7 in the amygdala, and possibly in other parts of the striatum, as well as in frontal cortical regions, and its coupling to the CRF1 receptor in the amygdala, as well as in the pituitary, bespeaks a deeper involvement in stress and negative affect. It is, thus, not surprising that associations have been reported between measures of adenylyl cyclase activity in brain, platelets, and lymphocytes of alcoholics, and subjects diagnosed with MDD. This association has been extended to genetic markers which identify the haplotype in which the ADCY7 gene is located. The significant comorbidity that exists between AUD and MDD is well accepted (Hasin et al., 2018). There are two manifestations of the co-occurrence of depression in individuals who fit the criteria for AUD. In one manifestation, the signs of depression are evident only during the initial period of time that an individual dependent on alcohol tried to abstain (i.e., alcohol withdrawal), and once abstinence has been achieved for some period of time, the signs and symptoms of depression abate (Raimo and Schuckit, 1998). In another manifestation, the signs of depression become evident during the initial stages of abstinence but continue to persist throughout sobriety (Raimo and Schuckit, 1998). When one considers the time course of changes in brains of animals that have been chronically fed ethanol, and have undergone forced abstinence, one notes that in brain areas such as the cortex, the activity of adenylyl cyclase is within the normal range while the animal is intoxicated, drops below normal levels during the first days of abstinence, and then returns to normal. One wonders whether the lower levels of adenylyl cyclase activity in brain during the initial stages of withdrawal is a contributing factor to the signs of withdrawal (i.e., depression), or simply a byproduct of the withdrawal hyperexcitability syndrome. It is of interest that measures of adenylyl cyclase in platelets of human alcoholics present a picture resembling the time course of fluctuations in adenylyl cyclase activity seen in brains of alcohol dependent and withdrawing animals. Adenylyl cyclase activity in platelets was in the normal range while the individual was actively consuming alcohol, dropped below normal levels during early stages of withdrawal, and then returned to normal after a period of abstinence. It might seem that the stress of abstaining from alcohol may be a factor in diminishing adenylyl cyclase activity during withdrawal. The alcohol withdrawal-induced changes in brain and platelet adenylyl cyclase activity can be classified as a state marker of withdrawal from chronic use of alcohol. On the other hand, the genetically generated increased expression of adenylyl cyclase in brains of animals is associated with more permanent depressive symptomology. There are a number of missing pieces of evidence that need to be added to assume that increases in mRNA for AC7 are related to higher activity of this enzyme in brain of depressed subjects. Even accurate measures of AC7 protein have not been accomplished (Joeyen-Waldorf et al., 2012). The differences in adenylyl cyclase activity between depressed human subjects and controls, are related to lower levels of adenylyl cyclase activity in depressed subjects in platelets, and activity of this adenylyl cyclase in response to Gsα is enhanced when the subject is being successfully treated with antidepressants (Targum et al., 2022). Even though the exact relationship between expression and activity of AC7 and MDD is still enigmatic, the development of pharmacological tools for isoform-selective manipulation of AC7 would help resolve the enigmatic features of the relationship and may lead to novel therapeutics for depression and/or AUD. A prior review (Price and Brust, 2019) suggested the possibility that AC7 may be “A new target for depression,” but did not propose how to “medicate this target.” An excellent review of molecules that inhibit adenylyl cyclase activity is available (Seifert et al., 2012) including many P-site inhibitors (adenosine analogues) and substituted nucleotides that act at the catalytic site of the adenylyl cyclases. This review emphasizes the problems encountered in trying to generate inhibitors that interact with the catalytic domains of the adenylyl cyclases, and would also have some substantial selectivity among the nine membrane-bound adenylyl cyclases, and would not have off-target effects on ion channels and glucose transporters. The effects of the large number of compounds developed as catalytic site inhibitors have not been tested for activity and selectivity for AC7. The diterpene alkaloids, such as forskolin, have generally been thought of as activators of adenylyl cyclase (Seamon et al., 1981), however, 1-deoxy-forskolin and 1,9-dideoxy-forskolin have been found to inhibit adenylyl cyclase activity (Seifert et al., 2012). One of the most interesting analogues of forskolin is referred to as BODIPY-FS (Pinto et al., 2008). BODIPY-FS is an activator of Type 1 and 5 adenylyl cyclases, has little effect on Type 3 and 6, but inhibits the activity of the Type 2 adenylyl cyclase (Seifert et al., 2012). The other members of the Type 2 adenylyl cyclase family, i.e., Type 4 and Type 7, have not been explored with regard to their responses to BODIPY-FS. The search for isoform-specific inhibitors (Brand et al., 2013) or activators of adenylyl cyclases may benefit from a more thorough consideration of structural/regulatory differences. In this regard, peptide analogues designed to correspond to the regions of the Type 2, 4 and 7 adenylyl cyclases that are phosphorylated by PKC may prove valuable modulators of these isoforms, and may even distinguish between these isoforms. Interestingly, a peptide based on the C1b region of AC7 was found to be an inhibitor of this enzyme (Yan et al., 2001). The other sequence that bears attention in trying to modulate the activity of AC7 is the region that binds the β/γ subunits (Diel et al., 2006; Brand, 2015). The sequences in this region of the isoforms responsive to β/γ need to be carefully examined and information should be gathered on the variants of the β and γ subunits that may have selectivity for the particular cyclase isoforms Diel et al. (2006). Such peptide modulators would have an additional advantage (or possible disadvantage) of being coordinate regulators of, for example, AC7 requiring the binding of Gsα, for evidence of their activity (Wang et al., 2005). Development of peptides as drug molecules is a complicated endeavor and as mentioned above, targeting AC7 should be preceded by a clear knowledge of its role in brain function [e.g. opiate tolerance/dependence (Wang et al., 2005), AUD, depression, etc.] as well as in the periphery [e.g. in the immune system (Duan et al., 2010)]. At this point, the evidence for AC7 activity in depression and AUD is tantalizing, but far from definitive. The actions of lithium (a mood stabilizer) as an inhibitor of AC7 (Mann et al., 2008), and the possible upregulation of AC7 by antidepressants such as SSRIs (Targum et al., 2022) should raise interest in the role of this enzyme in mood disorders. On the other hand, the more immediate use of AC7 activity may be as a peripheral state marker in AUD, state or trait marker in depression and a diagnostic distinguishing MDD from AUD (Tabakoff et al., 1986) or manic-depressive illness (Tabakoff et al., 2010).

PMC9649620

Luzhou Wang,Heba Zabri,Simone Gorressen,Dominik Semmler,Christian Hundhausen,Jens W. Fischer,Katharina Bottermann

Cardiac ischemia modulates white adipose tissue in a depot-specific manner 10.3389/fphys.2022.1036945

28-10-2022

Myocardial infarction,white adipose tissue depots,lipolysis,browning,adipokines,lipogenesis

The incidence of heart failure after myocardial infarction (MI) remains high and the underlying causes are incompletely understood. The crosstalk between heart and adipose tissue and stimulated lipolysis has been identified as potential driver of heart failure. Lipolysis is also activated acutely in response to MI. However, the role in the post-ischemic remodeling process and the contribution of different depots of adipose tissue is unclear. Here, we employ a mouse model of 60 min cardiac ischemia and reperfusion (I/R) to monitor morphology, cellular infiltrates and gene expression of visceral and subcutaneous white adipose tissue depots (VAT and SAT) for up to 28 days post ischemia. We found that in SAT but not VAT, adipocyte size gradually decreased over the course of reperfusion and that these changes were associated with upregulation of UCP1 protein, indicating white adipocyte conversion to the so-called ‘brown-in-white’ phenotype. While this phenomenon is generally associated with beneficial metabolic consequences, its role in the context of MI is unknown. We further measured decreased lipogenesis in SAT together with enhanced infiltration of MAC-2+ macrophages. Finally, quantitative PCR analysis revealed transient downregulation of the adipokines adiponectin, leptin and resistin in SAT. While adiponectin and leptin have been shown to be cardioprotective, the role of resistin after MI needs further investigation. Importantly, all significant changes were identified in SAT, while VAT was largely unaffected by MI. We conclude that targeted interference with lipolysis in SAT may be a promising approach to promote cardiac healing after ischemia.

Cardiac ischemia modulates white adipose tissue in a depot-specific manner 10.3389/fphys.2022.1036945 The incidence of heart failure after myocardial infarction (MI) remains high and the underlying causes are incompletely understood. The crosstalk between heart and adipose tissue and stimulated lipolysis has been identified as potential driver of heart failure. Lipolysis is also activated acutely in response to MI. However, the role in the post-ischemic remodeling process and the contribution of different depots of adipose tissue is unclear. Here, we employ a mouse model of 60 min cardiac ischemia and reperfusion (I/R) to monitor morphology, cellular infiltrates and gene expression of visceral and subcutaneous white adipose tissue depots (VAT and SAT) for up to 28 days post ischemia. We found that in SAT but not VAT, adipocyte size gradually decreased over the course of reperfusion and that these changes were associated with upregulation of UCP1 protein, indicating white adipocyte conversion to the so-called ‘brown-in-white’ phenotype. While this phenomenon is generally associated with beneficial metabolic consequences, its role in the context of MI is unknown. We further measured decreased lipogenesis in SAT together with enhanced infiltration of MAC-2+ macrophages. Finally, quantitative PCR analysis revealed transient downregulation of the adipokines adiponectin, leptin and resistin in SAT. While adiponectin and leptin have been shown to be cardioprotective, the role of resistin after MI needs further investigation. Importantly, all significant changes were identified in SAT, while VAT was largely unaffected by MI. We conclude that targeted interference with lipolysis in SAT may be a promising approach to promote cardiac healing after ischemia. Myocardial infarction (MI) is a pathology with strong systemic implications. Due to various factors as reduced cardiac output, pain and distress as well as dying cardiomyocytes, catecholamine levels during and after cardiac ischemia are elevated (Valori et al., 1967). This systemic adrenergic stimulation does not only increase cardiac performance and blood pressure, but also impacts several other organs as kidney, liver and adipose tissue. Catecholamine stimulation is one of the strongest stimuli for peripheral adipose tissue lipolysis, the process of hydrolysis of triacylglycerides (TAG) into glycerol and fatty acids (Schreiber et al., 2019). It is long known that circulating levels of free fatty acids (FFA) are elevated in plasma samples of patients with myocardial infarction (Vetter et al., 1974) and this could also be shown in small animal models of myocardial infarction (Yue et al., 2003). High levels of FFA are known to be detrimental to the ischemic myocardium (Essop and Opie, 2020). The first and rate limiting enzyme of lipolysis is adiposetriglyceride lipase (ATGL) (Haemmerle et al., 2006) which is activated via catecholamine-induced stimulation of PKA and release of its cofactor CGI-58 from Perilipin 1. It catalyzes the hydrolysis from TAGs to DAGs and free fatty acids. Adipocyte ATGL is a major player of whole-body lipid metabolism and glucose homeostasis (Schreiber et al., 2019) and has been shown to be involved in cardiac pathologies as pressure overload (Parajuli et al., 2018; Salatzki et al., 2018; Bottermann et al., 2022), catecholamine-induced heart failure (Takahara et al., 2021; Thiele et al., 2021) or myocardial infarction (Bottermann et al., 2020). Next to its well-known function in the storage and provision of energy, adipose tissue is more and more recognized also as an important endocrine organ which secretes a plethora of factors as adipokines, cytokines, micro-RNAs (Thomou et al., 2017) or extracellular vesicles (EV) (Crewe et al., 2021; Li et al., 2021). Several of these adipose tissue derived factors were also discovered to be involved in cardiac remodeling after myocardial infarction, for example adiponectin (Shibata et al., 2007), miR30d (Li et al., 2021) or small EVs (Crewe et al., 2021). White adipose tissue consists of several depots throughout the body, which can roughly be divided into visceral and subcutaneous white adipose tissue (VAT and SAT) and are associated with different cardiovascular risks (Oikonomou and Antoniades, 2019). Visceral WAT mainly surrounds the inner organs (pericardial, perigonadal, retroperitoneal, omental, mesenteric) and is negatively correlated with cardiovascular risk. Subcutaneous WAT is composed of inguinal, gluteal and abdominal depots and is considered to be protective in the context of cardiovascular disease (Wronska and Kmiec, 2012; Chusyd et al., 2016). Visceral and subcutaneous WAT differ with respect to adipocyte size and number, adipokine secretion and lipid storage and release capacity. In mice SAT adipocytes are smaller than VAT adipocytes, however the overall adipocyte number is higher in SAT. A simplified view is that VAT seems to be more responsible for energy storage and release, while SAT seems to be the major source of adipokines as leptin and adiponectin (Wronska and Kmiec, 2012; Schoettl et al., 2018). However, this strongly depends on species and metabolic state. Due to its function as regulator of whole body metabolism and its secretory activity, targeting of adipose tissue in cardiac pathologies is seen as a promising therapeutic approach (Smeir et al., 2021). However, lipolysis in response to MI seems to be activated only transiently as catecholamine and free fatty acid levels are upregulated within the first hours after myocardial ischemia and found to be reduced thereafter (Oliver, 2014). It remains unclear, how myocardial ischemia and the post-ischemic remodeling processes impact white adipose tissue and if the main WAT depots are affected differentially, which could open up new therapeutic strategies. We hypothesized that myocardial ischemia induces acute and chronic changes within subcutaneous (inguinal) and visceral (gonadal) WAT at different timepoints of reperfusion and therefore examined in the present study key features of white adipose tissue such as morphology, inflammation and gene expression. 12 weeks old C57Bl/6J male mice (Janvier Labs) underwent 60 min closed chest ischemia (Nossuli et al., 2000) followed by 24 h, 7 days or 28 days of reperfusion (I/R). For induction of closed chest ischemia, mice received surgery 5–7 days before ischemia to introduce a suture around the left ventricular descending artery (LAD). Mice were anesthetized with 90 mg/kg BW Ketamin/15 mg/kg BW Xylazin, intubated and mechanically ventilated. The chest was opened in the 3rd intercostal space, the LAD exposed and a 7/0 prolene suture passed underneath the LAD. A small PE-10 tube was threaded on both ends of the suture, loosely forming a loop around the LAD. The ends of the suture were left in a subcutaneous skin pocket and the ribs and skin closed. Mice were allowed to recover from surgery for 5–7 days and then anesthetized with 2% isoflurane, and ischemia was induced under ECG control by using 5 g weights at each end of the suture. Body temperature was maintained at 37.5°C. Postoperative analgesia was achieved using buprenorphine (0.05 mg/kg BW). Blood was sampled from the tail vein before ischemia and after 30 min of reperfusion. Echocardiography was performed with ultrasound device Vevo 3100 (Visual Sonics) and ultrasound probe MX400. All experiments were in accordance with the local animal regulations and approved by the local authorities (LANUV NRW, AZ: 81-02.04.2019.A397). After the intended reperfusion times mice were sacrificed and inguinal and gonadal fat pads, as representative for subcutaneous and visceral WAT, excised. For both depots, the right pad, used for protein isolation, and half of the left pad, used for RNA isolation, were briefly washed in cold PBS and snap frozen in liquid nitrogen. The other half of the left pad, used for histological analysis, was 4% formalin fixed overnight and paraffin embedded. NEFA measurement was performed in serum using Wako HR NEFA Kit (Fujifilm) according to the manufacturers’ recommendation. Briefly, a standard series composing of 0, 0.125, 0.25, 0.5, and 1 mM NEFA Standard was prepared in advance. 4 µl serum or standard and 200 µl FUJIFILM NEFA (HR) 1 were incubated for 5 min at RT, samples were measured once by SYNERGY microplate reader at a wavelength of 550 nm and 100 µl NEFA (HR) 2 was added, further incubated for another 5 min at RT and measured again at a wavelength of 550 nm. 5 µm paraffin sections were deparaffinized, incubated with hematoxylin (Sigma-Aldrich) for 1 min, followed by a short rinse in tap water and 1% HCl. Bluing was performed under running tap water for 10 min. Afterwards, 1% eosin solution (Carl Roth) was incubated for 1 min, sections were dehydrated and mounted with Roti®-Mount (Carl Roth). Samples were analyzed by microscope Zeiss Imager.M2 using a 10× objective. The size of 200 cells was measured for each animal using ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, United States, https://imagej.nih.gov/ij/, 1997–2018) by manually drawing the outline of each adipocyte. After deparaffinization, antigen retrieval was performed by cooking tissue slices in citrate buffer for 20 min. Blocking was performed with 10% FCS (Thermo Fisher Scientific, #10270106) and 1% BSA (Sigma-Aldrich, #A9418) in TBS (20 mM Tris, 150 mM NaCl) at RT for 1 h. After blocking, each section was incubated with 50–100 µl primary antibody (Anti-Mouse/Human Mac-2, 6.67 μg/mL, Cedarlane, #CL8942AP or anti-UCP1, Abcam ab10983, 2 μg/mL) which was diluted with PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4) containing 1 % BSA at 4°C overnight. After 3 times PBS wash, slices were incubated with 50–100 µl secondary antibody (Alexa FluorTM 647 Goat anti-Rat, 10 μg/mL, Thermo Fisher Scientific, #A21247) in PBS at RT in the dark for 1 h, followed again by 3 washes in PBS. Wheat germ agglutinin (WGA Alexa FluorTM 488 Conjugate, 2 μg/mL, Thermo Fischer Scientific, #W11261) was applied together with secondary antibody. Mounting was performed with ROTI® Mount FluorCare DAPI (Carl Roth). Macrophages in white adipose tissue stained by immunohistochemical staining with anti-MAC-2 antibody were counted manually with the help of software ImageJ. Sections were analyzed by microscope Zeiss Imager. M2 using a 20× objective. All the macrophages in one whole section were counted, the outline of the section was drawn by hand. The final macrophage number of each animal was noted as the number of macrophages in 1 mm2 tissue. White adipose tissue was lysed and homogenized using 500 µl lysis buffer [20 mM Tris-HCl, 1 mM EDTA, 255 mM Sucrose, pH = 7.4, HaltTM Protease & Phosphatase Inhibitor Cocktail (Sigma-Aldrich)] and Qiagen TissueRuptor. Protein concentration was determined using BCA Protein Assay (Thermo Fisher Scientific) and samples were diluted with 4 × Lämmli-buffer (250 mM Tris (pH = 6.8), 8% (v/v) SDS, 20 % (w/v) Glycerol, 0.02 % (v/v) Bromophenol blue, 100 mM DTT). Proteins were separated by SDS-PAGE and semi-dry blotted to nitrocellulose membrane. Before membrane blocking (LI-COR Intercept® Blocking Buffer) RevertTM 700 Total Protein Stain (Licor) was applied according to the manufacturers’ recommendation to determine total protein for normalization. Primary antibody (Anti-UCP1, Abcam ab10983, 1 μg/mL) was incubated in 5% BSA in TBST (2.4% (w/v) Tris, 8.8 % (w/v) NaCl, pH = 7.6, 0.1% Tween) over night at 4°C, followed by 3 × wash in TBST and secondary antibody (IRDye® 800CW Goat anti-Rabbit, 0.1 μg/mL, LI-COR, 926-32211) in blocking solution/0.1 %Tween for 1 h RT in the dark. RNA of peripheral white adipose tissue was isolated using Qiagen RNeasy® Lipid Tissue Mini Kit according to the manufacturers’ recommendation. Reverse transcription was performed according to QuantiTect® Reverse Transcription Quick-Start Protocol (Qiagen). Quantitative real time PCR was performed using Platinum® SYBR® Green qPCR SuperMix-UDG (Invitrogen) and StepOne Plus Real-Time PCR Detection System. Calculation of relative gene expression was performed by means of 2(−ΔCt) using Nudc as reference gene and normalized to sham-operated animals. Sequences of used primers (Sigma-Aldrich, Thermo Fisher Scientific) are listed in Table 1, Harvard PrimerBank ID (Wang and Seed, 2003; Spandidos et al., 2008; Spandidos et al., 2010) is denoted when applicable. Data are presented as mean ± standard error mean (SEM). Statistical analysis was performed using GraphPadPrism 9. For comparison of two groups unpaired, two-tailed t-test or Mann-Whitney test was used. For comparison of groups with two variables 2-way-ANOVA was used followed by Sidak’s multiple comparisons test. Outliers were identified using ROUT test (Figure 3D). p-values below 0.05 are indicated by asterisks (*p < 0.05, **p < 0.01, ***p < 0.001). Trends are shown by p-values as numbers as indicated. C57Bl/6J male mice at age of 12 weeks were subjected to 60 min cardiac ischemia or sham operation. Different cohorts of mice were harvested after 24 h, 7 days and 28 days of reperfusion (rep) (Figure 1A). The 28 days rep group, which underwent echocardiography at baseline, day 7 and day 28, showed a significant reduction in systolic pump function (Day 7: EF: 37 ± 3.2%, FAC: 28.9 ± 1.6%, Day 28: EF: 32.3 ± 3.1%, FAC: 22.9 ± 4.2%) and increase in end diastolic and end systolic volumes (Day 7: EDV: 122.3 ± 6 μl, ESV: 77.5 ± 6.7 µl, Day 28: EDV: 156.2 ± 10 μl, ESV: 106.8 ± 11.1 µl) in the ischemia group compared to sham (Figure 1B). The tissue weight to body weight ratios from gonadal WAT (gWAT) and inguinal WAT (iWAT) were unchanged between the two groups (Figure 1C) at all three timepoints. To assess lipolytic activity, serum levels of non-esterified fatty acids (NEFA) were measured before induction of ischemia and after 30′ of reperfusion. The ischemic animals showed a significantly higher 3-fold increase in serum NEFA levels compared to sham operated animals (1.6-fold) (Figure 1D), indicating a stimulation of lipolysis due to cardiac ischemia. To investigate if the lipolytic response after myocardial ischemia induces morphological changes in the different WAT depots, tissue sections of iWAT and gWAT were H&E-stained and adipocyte size was analyzed. While in gWAT no major changes were observed compared to sham controls, iWAT of infarcted mice showed an increase in the number of smaller adipocytes (300–350 μm2) and a decrease in the number of bigger adipocytes over 1500 μm2 after 24 h of reperfusion. This shift towards smaller adipocyte size was also seen at the later timepoints and was most pronounced 28 days post reperfusion (Figures 2A,B). In line with this, the mean adipocyte size of gWAT was unchanged in comparison to sham controls while iWAT showed a trend towards a reduction of the mean adipocyte size (Supplementary Figure S1), even though statistical significance was not reached. To further investigate the underlying cause for decreased adipocyte size in iWAT after cardiac ischemia, we analyzed gene expression of the main lipolytic enzyme ATGL (Pnpla2). This revealed a significant upregulation of Pnpla2 at the timepoint 28 days of reperfusion only in iWAT, while the gene expression at other timepoints or in gWAT was not affected (Figure 2C). Together, these data validate increased lipolysis in our model of cardiac ischemia and demonstrate changes in adipocyte size in iWAT but not gWAT. Next to adipocyte cell size, H&E staining of ischemic animals revealed further morphological abnormalities in iWAT. Adipocytes appeared rather multilocular (Figure 3A), which may indicate “brown-like” or “brite” adipocytes (Barreau et al., 2016). Accordingly, these structures were analyzed for UCP1-expression, which is a common browning marker (Bargut et al., 2017). The staining demonstrated that the striking structures were positive for UCP1 (Figure 3B). We further evaluated UCP1 protein content in samples from 24 h reperfusion animals and indeed found an increase in UCP1 protein levels in ischemic samples compared to sham samples (Figures 3C,D). As myocardial ischemia induced lipolysis, reduced cell size and increased browning of inguinal WAT, we were interested how adipogenesis and lipogenesis of the different WAT depots were affected. While gene expression of the transcription factor PPARγ (Pparg) itself was not affected in both depots, downstream genes involved in lipogenesis as lipoprotein lipase (Lpl), fatty acid synthase (Fasn) and diacylglycerol acyltranferase 2 (Dgat2) were either downregulated (Lpl) or trended to be downregulated (Fasn, Dgat2) after 24 h of reperfusion in inguinal but not in gonadal WAT (Figure 4A and Supplementary Figure S2). This reduced expression of a lipogenic gene signature seemed to be an acute and transient reaction to myocardial ischemia, as expression levels returned to baseline at 7 days and 28 days of reperfusion (Figures 4B,C). Only Lpl still trended to be downregulated also at these later timepoints, in line with persistent smaller adipocyte size in subcutaneous WAT. On day 7 of reperfusion, H&E stained sections of WAT contained increased numbers of nuclei from cells of non-adipocyte morphology. Assuming that these cells represented infiltrated immune cells, we stained both WAT depots for MAC-2 (Figure 5A). Quantification of MAC-2 revealed a significant increase of total macrophage numbers in iWAT while they were unchanged in gWAT (Figure 5B). Interestingly, the formation of crown-like structures (CLS), i.e., macrophages surrounding dead or dying adipocytes (Strissel et al., 2007), was significantly enhanced in gWAT but not in iWAT (Figure 5B). Adipose tissue derived factors exert a plethora of systemic functions and several adipokines are known to be involved in post-infarct remodeling. We therefore analyzed gene expression of several adipokines and found three of them, namely adiponectin, leptin and resistin to be differentially expressed. Again, the inguinal WAT was more affected as gene expression of adiponectin and resistin was downregulated in ischemic animals after 24 h of reperfusion and of leptin after 7 days of reperfusion (Figure 6), while expression in gWAT was unchanged (Supplementary Figure S3). These changes were transient, as there was no difference observed after 28 days of reperfusion. Increasing evidence suggests that targeting adipose tissue and lipolysis is beneficial in various cardiac pathologies (Smeir et al., 2021). However, dissecting physiological and pathogenic functions of lipolysis has proven challenging and requires a better understanding of tissue-specific effects. Here, we used a mouse model of closed chest myocardial ischemia showing activation of lipolysis as measured by increased circulating NEFA levels after 30 min of reperfusion. In line with published data (Oliver, 2014), this increase was transient but interestingly, resulted in chronic alterations in white adipose tissue in a depot-specific manner. Our main findings are: 1) Myocardial ischemia induces alterations of white adipose tissue morphology, gene and protein expression and inflammation up to 28 days after ischemia. 2) The subcutaneous WAT is more susceptible to myocardial ischemia induced changes than the visceral WAT, as we found a reduction in cell size, an increased browning of white adipocytes, a higher macrophage infiltration and a reduction in adipokine gene expression. The analysis of adipocyte morphology in subcutaneous WAT revealed a sustained reduction of adipocyte size which was associated with an increased gene expression of ATGL at the late timepoint of 28 days of reperfusion. Furthermore, an increased occurrence of multilocular adipocytes which were found to be positive for UCP1 could be observed. Subcutaneous and visceral WAT mainly consist of unilocular white adipocytes with a low number of mitochondria. However, also adipocytes with features of brown adipocytes as multilocularity, high number of mitochondria and UCP1 expression can be found in VAT and SAT. These are called “brown-in-white” or “brite” adipocytes and exhibit these features mainly in response to certain stimuli as cold, catecholamines or exercise (Scheel et al., 2022). In the present study, we detected browning in the subcutaneous depot, in line with observations that browning of white adipose tissue, also in response to other stimuli as cold or β3-agonists, mainly occurs in subcutaneous WAT (Bargut et al., 2017). The fact that the control group underwent anesthesia and sham surgery in the same manner as ischemic animals, showed that the browning effect was indeed due to cardiac ischemia and not to other experimental procedures such as surgical trauma (Longchamp et al., 2016). The most likely explanation for an increased browning in our model of cardiac ischemia and reperfusion is the strong β-adrenergic stimulation, which also activates adipose tissue lipolysis, as seen by an increase in circulating NEFA levels. In addition, also natriuretic peptides, which are elevated after myocardial infarction, were recently identified to promote white adipocyte browning via mTOR (Liu et al., 2018). However, the impact of the observed phenomenon on cardiac remodeling is unclear. The implications of browning are generally considered protective in the setting of obesity, as brite adipocytes increase energy expenditure by increasing fatty acid oxidation (Barquissau et al., 2016) and improve insulin sensitivity. However, in the context of energy wasting and hypermetabolic states as cachexia (Tamucci et al., 2018) or burns (Kaur et al., 2021) these effects might not be favorable and therefore aggravate the disease. The mice used in this study were lean C57Bl/6J mice. Interestingly, several hallmarks of adipose tissue which are found in cancer patients, as increased lipolysis, increased browning and reduced lipogenesis (Weber et al., 2022), were also seen in our mouse model of cardiac ischemia/reperfusion. This might indicate that the observed alterations are rather unfavorable, however further studies are needed to elucidate the impact on cardiac remodeling, also in the setting of obesity. Another finding of the morphological analysis of the different WAT depots, was the increased number of MAC-2+ macrophages in both depots after 7 days of reperfusion. Whereas subcutaneous WAT showed an increase in the overall number of macrophages, visceral WAT contained more crown-like structures, indicating adipocyte apoptosis. Adipose tissue inflammation is a major hallmark of obesity and contributes to adipose tissue and whole-body insulin resistance. Pro-inflammatory M1-like macrophages are the main driver of this low-grade inflammation, which predominantly occurs in visceral WAT depots (Kawai et al., 2021). However, also lipolysis is associated with WAT inflammation (Morris et al., 2011). Lipolytic activation in response to fasting and β3-adrenergic stimulation was shown to increase macrophage infiltration by increasing chemotaxis. Inhibition of lipolysis by knocking out ATGL was able to ameliorate infiltration, indicating a direct link between lipolytic products and macrophage infiltration (Kosteli et al., 2010). Of note, in our mouse model of cardiac ischemia/reperfusion, we found increased macrophage accumulation in subcutaneous WAT despite the fact that circulating free fatty acids were only transiently elevated after ischemia, demonstrating the long-term effects of acute lipolysis activation. Adipose tissue inflammation was also observed in a mouse models of stress induced cardiac hypertrophy and was reduced by ATGL-inhibition (Takahara et al., 2021), indicating that also a more sustained cardiac dysfunction can induce adipose tissue inflammation. Adipose tissue is known as an active endocrine organ, which secretes a plethora of systemically active protein factors, called adipokines. We showed that cardiac ischemia/reperfusion affects gene expression of several of these adipokines acutely and sustained, only in the inguinal, subcutaneous WAT depot. We found gene expression of adiponectin and resistin to be downregulated after 24 h of reperfusion and of leptin after 7 days of reperfusion. As expression and release of adipokines is closely correlated to adipocyte size (Skurk et al., 2007), this is in line with our observation of reduced adipocyte size in SAT. Adiponectin levels are known to be transiently reduced after myocardial infarction in humans during the first 72 h after infarction and are nearly normalized again after 7 days (Kojima et al., 2003). Our present data indicate that this reduction might be due to a reduced secretion of adiponectin from the subcutaneous WAT depot. Adiponectin is generally seen as a cardioprotective and anti-inflammatory adipokine, which reduces infarct size in mice (Shibata et al., 2005) and improves cardiac function after experimentally induced myocardial infarction (Shibata et al., 2007). Expression of adiponectin is negatively regulated by insulin, TNFα and dexamethasone (Fasshauer et al., 2002) but also by catecholamines (Fasshauer et al., 2001). It is suggested that adiponectin is a PPARy target gene (Liu and Liu, 2009), which fits to the observation of reduced gene expression of PPARy target genes as Lpl in our model. Next to adiponectin, also the proinflammatory adipokine resistin was downregulated in iWAT after 24 h of reperfusion. Resistin was discovered (Steppan et al., 2001) as a proinflammatory adipokine involved in glucose metabolism and insulin signaling (Gualillo et al., 2007). Several studies show an impact of resistin on cardiac metabolism and function in the context of cardiac hypertrophy (Chemaly et al., 2011; Kang et al., 2011; Zhao et al., 2022), while its role after myocardial infarction is unclear. Plasma levels of resistin in humans were either increased (Korah et al., 2011) or unchanged (Gruzdeva et al., 2014) after MI and high plasma levels are correlated with a higher risk for myocardial infarction (Weikert et al., 2008). In rodent models of cardiac ischemia, resistin either aggravated cardiac dysfunction (Rothwell et al., 2006) or was protective (Gao et al., 2007). Resistin seems to be coupled to lipolytic activity as a model of hemorrhagic shock increased circulating resistin levels, while inhibiting lipolysis attenuated this increase (Raje et al., 2020). Considering this, we were rather surprised to find resistin downregulated in the early phase after cardiac ischemia. Interestingly, PRDM16, a further marker and inductor of browning in white adipose tissue (Seale et al., 2007), was shown to repress resistin gene expression by interacting with C-terminal binding proteins (CtBPs) (Kajimura et al., 2008). As we observe a “brite” phenotype of the inguinal WAT depot after 24 h of reperfusion, this might, at least in part, explain a reduced resistin expression at the same timepoint. While adiponectin and resistin were altered early after ischemia, leptin was found to be downregulated in SAT after 7 days of reperfusion. In the context of myocardial infarction, leptin was shown to be cardioprotective by direct and indirect effects, for example by protection of cardiomyocytes from apoptosis (Smith et al., 2006; McGaffin et al., 2009) and by improvement of cardiac substrate oxidation (Gava et al., 2021). Circulating leptin levels were found to be elevated in humans in the early phase after myocardial infarction, peaking at day 2 to day 3 (Fujimaki et al., 2001; Khafaji et al., 2012). Our data show a transient decrease at day 7 in leptin gene expression only in subcutaneous WAT, while gene expression in the visceral depot, which has the highest leptin expression in rodents (Trayhurn et al., 1995), is not affected by ischemia/reperfusion. Leptin expression seems to underly a negative feedback regulation (Zhang et al., 1997). Thus, the observed reduced expression at day 7 in our model might be explained by elevated circulating levels before. Taken together, the here presented study shows a major depot-specific impact of cardiac ischemia and reperfusion on the subcutaneous WAT, which occurs not only acutely after a strong induction of lipolysis via catecholamines, but also more prolonged during the post-ischemic remodeling process. These findings contribute to a better understanding of the heart–adipose tissue crosstalk in cardiac I/R and might open up new therapeutic strategies in the prevention and treatment of post-infarction heart failure.

PMC9649622

Hoda T. Amer,Reda A. Eissa,Hend M. El Tayebi

A cutting-edge immunomodulatory interlinkage between HOTAIR and MALAT1 in tumor-associated macrophages in breast cancer: A personalized immunotherapeutic approach 10.3389/fmolb.2022.1032517

28-10-2022

breast cancer,immunotherapy,epigenetics,tumor-associated macrophages (TAMs),MALAT1,HOTAIR,CD80,MSLN

Breast cancer (BC) is one of the most common cancers, accounting for 2.3 million cases worldwide. BC can be molecularly subclassified into luminal A, luminal B HER2-, luminal B HER2+, HER2+, and triple-negative breast cancer (TNBC). These molecular subtypes differ in their prognosis and treatment strategies; thus, understanding the tumor microenvironment (TME) of BC could lead to new potential treatment strategies. The TME hosts a population of cells that act as antitumorigenic such as tumor-associated eosinophils or pro-tumorigenic such as cancer-associated fibroblasts (CAFs), tumor-associated neutrophils (TANs), monocytic-derived populations such as MDSCs, or most importantly “tumor-associated macrophages (TAMs),” which are derived from CD14+ monocytes. TAMs are reported to have the pro-inflammatory phenotype M1, which is found only in the very early stages of tumor and is not correlated with progression; however, the M2 phenotype is anti-inflammatory that is correlated with tumor progression and metastasis. The current study focused on controlling the anti-inflammatory activity in TAMs of hormonal, HER2+, and TNBC by epigenetic fine-tuning of two immunomodulatory proteins, namely, CD80 and mesothelin (MSLN), which are known to be overexpressed in BC with pro-tumorigenic activity. Long non-coding RNAs are crucial key players in tumor progression whether acting as oncogenic or tumor suppressors. We focused on the regulatory role of MALAT1 and HOTAIR lncRNAs and their role in controlling the tumorigenic activity of TAMs. This study observed the impact of manipulation of MALAT1 and HOTAIR on the expression of both CD80 and MSLN in TAMs of BC. Moreover, we analyzed the interlinkage between HOTAIR and MALAT1 as regulators to one another in TAMs of BC. The current study reported an upstream regulatory effect of HOTAIR on MALAT1. Moreover, our results showed a promising use of MALAT1 and HOTAIR in regulating oncogenic immune-modulatory proteins MSLN and CD80 in TAMs of HER2+ and TNBC. The downregulation of MALAT1 and HOTAIR resulted in the upregulation of CD80 and MSLN, which indicates that they might have a cell-specific activity in TAMs. These data shed light on novel key players affecting the anti-inflammatory activity of TAMs as a possible therapeutic target in HER2+ and TNBC.

A cutting-edge immunomodulatory interlinkage between HOTAIR and MALAT1 in tumor-associated macrophages in breast cancer: A personalized immunotherapeutic approach 10.3389/fmolb.2022.1032517 Breast cancer (BC) is one of the most common cancers, accounting for 2.3 million cases worldwide. BC can be molecularly subclassified into luminal A, luminal B HER2-, luminal B HER2+, HER2+, and triple-negative breast cancer (TNBC). These molecular subtypes differ in their prognosis and treatment strategies; thus, understanding the tumor microenvironment (TME) of BC could lead to new potential treatment strategies. The TME hosts a population of cells that act as antitumorigenic such as tumor-associated eosinophils or pro-tumorigenic such as cancer-associated fibroblasts (CAFs), tumor-associated neutrophils (TANs), monocytic-derived populations such as MDSCs, or most importantly “tumor-associated macrophages (TAMs),” which are derived from CD14+ monocytes. TAMs are reported to have the pro-inflammatory phenotype M1, which is found only in the very early stages of tumor and is not correlated with progression; however, the M2 phenotype is anti-inflammatory that is correlated with tumor progression and metastasis. The current study focused on controlling the anti-inflammatory activity in TAMs of hormonal, HER2+, and TNBC by epigenetic fine-tuning of two immunomodulatory proteins, namely, CD80 and mesothelin (MSLN), which are known to be overexpressed in BC with pro-tumorigenic activity. Long non-coding RNAs are crucial key players in tumor progression whether acting as oncogenic or tumor suppressors. We focused on the regulatory role of MALAT1 and HOTAIR lncRNAs and their role in controlling the tumorigenic activity of TAMs. This study observed the impact of manipulation of MALAT1 and HOTAIR on the expression of both CD80 and MSLN in TAMs of BC. Moreover, we analyzed the interlinkage between HOTAIR and MALAT1 as regulators to one another in TAMs of BC. The current study reported an upstream regulatory effect of HOTAIR on MALAT1. Moreover, our results showed a promising use of MALAT1 and HOTAIR in regulating oncogenic immune-modulatory proteins MSLN and CD80 in TAMs of HER2+ and TNBC. The downregulation of MALAT1 and HOTAIR resulted in the upregulation of CD80 and MSLN, which indicates that they might have a cell-specific activity in TAMs. These data shed light on novel key players affecting the anti-inflammatory activity of TAMs as a possible therapeutic target in HER2+ and TNBC. Breast cancer (BC) is one of the most commonly diagnosed cancers in women. It has now exceeded lung cancer as the leading cause of overall cancer incidence in 2020, with 2.3 million new cases; 685,000 deaths occur due to BC, making it the fifth leading cause of cancer mortality worldwide (Sung et al., 2021). Due to the molecular heterogeneity, the subtypes of BC are divided according to the expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), in addition to the percentage of the proliferating index (Ki67). Accordingly, they are classified into five different subtypes: luminal A, luminal B HER2-, luminal B HER2+, HER2+, and TNBC (basal-like) (Kondov et al., 2018). The luminal A subtype is considered the most abundant subtype with 50–60% prevalence among BC patients and is characterized by the expression of ER and/or PR with no expression of HER2 and with Ki67 < 14% (Blows et al., 2010), while luminal B abundance is 10–20% among BC patients with a more aggressive diagnostic profile than luminal A and is further classified into luminal B HER2- and luminal B HER2+. The luminal B HER2- subtype is characterized by the expression of ER and/or PR with no expression of HER2 and with Ki67 ≥ 14% (Blows et al., 2010), while the luminal B HER2+ subtype expresses ER and/or PR+ in addition to HER2+ with the expression of any Ki67 percentage (Blows et al., 2010). The HER2-enriched subtype is hormonal negative, expressing only HER2 receptors with any Ki67 percentage. The HER2-enriched subtype suffers from a worse prognosis than luminals in spite of the availability of its targeted therapy, anti-HER2. Finally, TNBC, ER, PR, and HER2- have the worst prognosis of all subgroups (Blows et al., 2010). Generally, the immune system plays a major role in cancers. Immunity can be subcategorized into innate immunity, which is fast, immediate, and non-specific, and adaptive immunity, which is specific and long lasting (Wills-Karp, 2007). Within innate immunity, myeloid cells are the most abundant hematopoietic cells (Gabrilovich et al., 2012). Tumor-infiltrating myeloid cells include granulocytes (such as basophils, eosinophils, and neutrophils), monocytes, dendritic cells, tumor-associated macrophages (TAMs), immature myeloid cells (IMCs), and MDSCs (6). Recently, the tumor-infiltrating myeloid cells have been reported to have an important role in angiogenesis, invasion, and metastasis, indicating their possible immunosuppressive role (Gabrilovich et al., 2012). Tumors have the ability to recruit stromal cells (e.g., fibroblasts), immune cells, and vascular cells through the secretion of growth factors, cytokines, and chemokines building a tumor microenvironment (TME) by releasing growth-promoting signals and remodeling tissue structure affecting initiation, progression, metastasis, vascularization, and therapy responses (Garofalo et al., 2006). Many treatments focus only on the cancer cell while special attention to the TME is needed since it has the key players of BC development and progression. Tumor not only tries to escape from the host immune system but also benefits from the infiltrating cells by modifying their functions to create the microenvironment that is favorable to its progression (Pollard, 2004). Monocytes play a very critical role in the TME either by themselves or on reprogramming to myeloid-derived suppressor cells (MDSCs) or tumor-associated macrophages (TAMs) (Allaoui et al., 2016). Monocytes are divided into three subsets based on the expression of CD16 and CD14 surface markers (Coillard and Segura, 2019). The CD16 (FcgRIII) molecule was only known initially to be expressed on mature macrophages; however, it is recognized recently as a surface marker for monocytes (Feng et al., 2011). The three subsets of monocytes are “classical” CD14+CD16− monocytes, which make up around 85% of monocytes, “intermediate” CD14+CD16+ monocytes, which account for 5–10% of total monocytes, and finally, “non-classical” CD14−CD16+, which also accounts for 5–10% (Coillard and Segura, 2019). Within the TME, monocytes can differentiate to M1 macrophage, which expresses the CD163 receptor marker at low levels (CD163neg/low macrophages), mediating defensive immune response (Ding et al., 2016). M1 macrophages usually receive stimulation from GM-CSF, LPS, and IFN- γ to produce IL-23 and IL-12 and promote Th1 responses (Rey-Giraud et al., 2012), thus having a pro-inflammatory response. In addition, M1 can secrete IL-6, ROS, and TNF- α (Sousa et al., 2015). Moreover, monocytes can infiltrate the tumor and differentiate into the M2 subtype, which is CD163 high (Feng et al., 2011). M2 can reduce tissue damage caused by inflammatory processes and stimulate their repair, thus having anti-inflammatory responses. M2 is usually activated by M-CSF, IL-4, IL-10, and IL-13 and can produce anti-inflammatory IL-10 and TGF-β. In the presence of tumors, macrophages are plastic and can be reprogrammed to polarize into either M1-like macrophages or M2-like macrophages, according to the cytokines present in the TME (Benner et al., 2019). Interestingly, TAMs contribute 5–40% of tumor mass in solid tumors. In the beginning, TAMs are said to have a pro-inflammatory phenotype (M1-like) and inhibit the tumorigenesis by ROS and TNF- α or even phagocytosis (Zhou et al., 2020). Nevertheless, once the cancer starts progressing, TAMs tend to secrete IL-10, TGF-β, and IL-12, suppressing cytotoxic T lymphocyte (CTL) and NK cells with the upregulation of programmed death ligand-1 (PDL-1), thus having anti-inflammatory action (Zhou et al., 2020). Moreover, TAMs can induce Tregs activity by IL-10, TGF- β, and PDGF-2, thus suppressing T cells. TAMs also promote angiogenesis by releasing VEGF, PDGF, and IL-8. Furthermore, they contribute to invasion and metastasis, particularly in malignant solid tumors, reducing the survival in BC patients and worsening their clinical outcomes (Campbell et al., 2011). TAMs contribute to the extracellular tissue remodeling, thus metastasis via MMPs. It was observed that COX-2 in breast TAMs promotes the metastatic potential of breast cancer cells. COX-2 in TAMs induces MMP-9 expression and enhances epithelial–mesenchymal transition (EMT) in breast cancer cells (Gan et al., 2016). Not only this but also TAMs express VEGF-C/D at the tumor site, which is shown to be related to lymph node metastasis (Schoppmann et al., 2006). TAMs not only play a role in BC but also in other cancers, namely, colorectal cancer. A study was carried out to show the contribution of TAMs in the metastasis of colorectal cancer (CRC). Briefly, the study showed that CRC-conditioned macrophages regulated EMT to enhance migration and invasion by the secretion of IL-6. TAMs-derived IL-6 was shown to activate the JAK2/STAT3 pathway. STAT3 transcriptionally inhibited the tumor suppressor miR-506-3p in CRC cells (Wei et al., 2019). It was also shown in another study that IL-6 secreted by TAMs promote the invasion of the prostate cancer cells and express low levels of the epigenetic factor (SFMBT2) (Gwak et al., 2020). In general, immune cells are known to express a number of immunomodulatory proteins, including CD80 and mesothelin (MSLN) proteins. CD80 (cluster of differentiation 80), a type-1 transmembrane glycoprotein is first identified in Epstein–Barr virus-activated B-cell blasts, B lymphoblastoid cell lines, and Burkitt’s lymphomas (Mir, 2015). It is a costimulatory molecule known to activate T cell and regulate the activity of normal and malignant B cells (Orabona et al., 2004). CD80 is expressed on activated B cells, macrophages, DCs, and cancer cells and binds to CD28 on T cells. It was shown that the loss of CD80 is enough to allow tumor to escape the immune system, thus imparting apoptosis caused by tumor-infiltrating T cells (Lázár-Molnár et al., 2010). Consequently, even if the tumor cell expresses MHC-I molecule while the co-stimulation is absent, the recognition of antigens by CTL cells will not cause any response. Conversely, on transfecting tumor cells with CD80, the tumor cell is found to be more susceptible to the lysis by T cells ex vivo. Moreover, it was evident that CD80 may enhance the memory responses by CTLs (Lázár-Molnár et al., 2010). However, CD80 has also been reported to be overexpressed in a number of cancers including BC (Li et al., 2020). This may be explained by the fact that CD80 is a ligand not only for CD28 but also for CTLA4 (cytotoxic T lymphocyte antigen-4 or CD152). CTLA4 displays important sequence and structure homology with CD28. Conversely to CD28, CTLA4, a negative regulator of T-cell activation, was found to have an anti-inflammatory action and facilitated the escape of tumor immunity (Brahmer et al., 2015). In BC, it was reported that CTLA4 has a higher binding affinity to CD80 than CD28 (Walker and Sansom, 2011). Consequently, blocking CTLA4 is a target for immunotherapy (Walker and Sansom, 2011). As a matter of fact, CD80 is highly expressed on APCs including macrophages, thus it is expected to be highly expressed on TAMs. Upon isolation of TAMs from human renal cell carcinomas, TAMs were shown to induce the CTLA4 expression on T lymphocytes (Daurkin et al., 2011). Mesothelin (MSLN) is another immunomodulatory protein that is present on normal mesothelial cells of the pleura, peritoneum, and pericardium (Hassan et al., 2004). The MSLN gene is shown to be overexpressed in many cancers including ovarian cancer, adenocarcinoma, pancreatic cancer, mesothelioma, lung adenocarcinoma, acute myeloid leukemia (Pastan et al., 1992), endometrial adenocarcinomas, and squamous cell carcinomas of the cervix, lung, head, and neck (Ordóñez, 2003). MSLN is reported to be overexpressed in 67% of TNBC cases and less than 5% of hormonal BC. Not only this but also the presence of MSLN in BC cells is associated with tumor infiltration of the lymph node. MSLN is shown to have a very limited expression in normal tissues, thus making it a very attractive candidate for cancer therapy (Macdonald et al., 2016). Several immunomodulatory proteins are known to be controlled epigenetically. Basically, the epigenetic machinery is composed of three components, namely, DNA methylation, histone modification, and non-coding RNAs. Non-coding RNAs (ncRNAs) are classified into two major classes based on the transcript size, in which 200 nucleotides is the threshold: small ncRNAs (sncRNAs) and long non-coding RNAs (lncRNAs) (Carninci et al., 2005). Metastasis associated in lung adenocarcinoma transcript 1 (MALAT1) is considered as one of the most studied lncRNAs in cancer and is also known as nuclear enrichment autosomal transcript 2 (NEAT2). MALAT1 was observed to be highly expressed in cancer tissues and it was initially observed in the metastatic non-small lung cancer tissues (Guan et al., 2020). Its function is still controversial that whether it acts as oncogenic or tumor suppressor lncRNA, thus understanding MALAT1 better would be important in epigenetics studies. One mechanism of how MALAT1 functions as an oncogene is by interacting with the polycomb repressive complex 2 (PRC2). On interaction, the RNA–protein complex with EZH2 and SUZ12 is formed. EZH2 and SUZ12 are two components of the PRC2 complex, thus facilitating the histone H3K27 trimethylation at the promoters of some tumor suppressor genes such as E-cadherin and N-myc downregulated gene-1 (NDRG1). Consequently, b-catenin and c-myc expression are increased (Chen et al., 2020). On the other hand, MALAT1 can function as a tumor suppressor since its expression was found to be positively correlated with the expression of the tumor suppressor PTEN, and their decreased levels were associated with mortality and poor patient survival in both colorectal cancer and BC patients (Guan et al., 2020). MALAT1 is shown in the literature to have an oncogenic activity enhancing both the progression and metastasis of breast tumors. A research group studied this oncogenic potential of MALAT1 by using MTT and transwell assay to detect proliferation, migration, and invasion. Furthermore, the drug resistance test was performed to assess the sensitivity of BC cells to doxorubicin. The study has shown that silencing of MALAT1 could significantly suppress the proliferation, migration, and invasion of BC cells. Moreover, downregulation of MALAT1 sensitized BC cells to doxorubicin (Yue et al., 2021). Not only this but also another group used the MMTV (mouse mammary tumor virus)-PyMT mouse mammary carcinoma model to assess the proliferative and metastatic potential of MALAT1. The results showed slower tumor growth accompanied by significant differentiation into cystic tumors and a reduction in metastasis upon loss of MALAT1 (Arun et al., 2016). In addition, HOTAIR (HOX transcript antisense RNA) is the first lncRNA found to promote tumor progression. HOTAIR is highly expressed in metastatic BC. It is transcribed from the antisense strand of the HOXC genes and partly overlaps with HOXC11 (Rinn et al., 2007). HOTAIR can function as an oncogene using several mechanisms, especially by the negative regulation of a number of miRNAs. For example, it competes with miR-34a, leading to the upregulation of SOX2, thus cell proliferation. It can also promote cell growth, mobility, and invasiveness by suppressing miR-20a-5p and consequently upregulate HMGA2 (Li et al., 2016). HOTAIR is known in the literature to have a proliferative and metastatic potential. In a research study, CCK-8 and colony formation assays showed that HOTAIR overexpression promoted the proliferation of MCF-7 cells. Furthermore, transwell invasion and migration assays showed that HOTAIR overexpression increased the migration and invasion of BC cells. These results indicate that HOTAIR facilitates the growth and metastasis of BC cells in vitro (He et al., 2022). Another research study confirmed these findings by using qRT-PCR to determine the expression of HOTAIR. CCK-8 and transwell assays were also used to detect the proliferation, migration, and invasion of cells. In addition, animal experiments were conducted to validate the effect of HOTAIR on BC tumor growth in vivo. The results showed that HOTAIR was upregulated in BC tissues and cells, and its knockdown suppressed the proliferation, migration, invasion, and the activity of the AKT signaling pathway of BC cells. Additionally, interference of HOTAIR had impacted BC tumor growth in vivo (Wang et al., 2020). Even though TAMs have a crucial role in tumorigenesis, the epigenetic players controlling and regulating this anti-inflammatory activity have never been studied before through their direct manipulation in TAMs. The aim of this study is to investigate the interlinkage between MALAT1 and HOTAIR and their regulatory activity on immunomodulation by examining the expression profile of CD80 and MSLN in tumor-associated macrophages in the hormonal, HER2+, and TNBC subtypes. In total, 43 blood samples were collected in EDTA tubes from breast cancer (BC) patients, after taking their consents. All patients were females ranging between the age of 34 and 78 years. Clinical features for each patient are represented in Table 1. Gender-matched healthy controls were used. Within 3 or 4 h of sample collection, the Ficoll separation technique was optimized and used to isolate the PBMCs from the whole blood (Jaatinen and Laine, 2007). The PBMCs of each sample were cryopreserved and stored in the −80 C freezer for later use. Detailed clinical features for each patient are mentioned in Supplementary Table S1. Cryopreserved PBMCs are left to melt at room temperature. The melted PBMCs are transferred into a falcon tube containing a 6 ml wash mix and was allowed to shake for 5 min followed by centrifugation for 5 min at 1,500 rpm. A pellet of PBMCs is formed, and the supernatant is discarded. The cells are suspended in an appropriate volume of complete media and counted. Isolation of CD14+ monocytes by negative depletion was performed using MojoSort™ Human CD14+ Monocytes Isolation Kit protocol (Cat. No. 480047), MojoSort™ Buffer (5X) (Cat. No. 480017), and MojoSort™ Magnet (Cat. No. 480019/480020). Isolation of CD8+ T cells by negative depletion was carried out using MojoSort™ Human CD8+ T cell Isolation Kit protocol (Cat. No. 480012), MojoSort™ Buffer (5X) (Cat. No. 480017), and MojoSort™ Magnet (Cat. No. 480019/480020). Freshly isolated CD14+ monocytes were subjected to centrifugation at 1,500 rpm, buffer supernatant was discarded, and CD14+ monocytes were plated in a 48-well plate (10,000 cells per well). Monocytes were cultured in 1:1 ratio of 10% cell culture medium and TCM with the addition ofIL-4 (1 μg/ml) (Schenendoah (United States), ID:100-09), IL-10 (1 μg/ml) (Schenendoah (United States), ID:100-83), and M-CSF (1 μg/ml) (Schenendoah (United States) 100-03). The medium was refreshed every other day. TAMs were harvested on day 7. After incubation of TAMs for 7 days to ensure differentiation (3 × 104 cells per well in a 48-well plate), the supernatant was discarded and refreshed by the transfection media. The appropriate amount of siRNAs (2 ul) was diluted in 60 μL of free culture medium without serum. The appropriate amount of HiPerFect transfection reagent (Qiagen) was added (1 ul) to the diluted siRNAs (MALAT1 siRNA, Qiagen ID: SI04342233) (HOTAIR siRNA, Qiagen ID: SI04446036) and then mixed by vortexing. The siRNAs-HPTR mixture was incubated for 5–10 min at room temperature (15–25°C) to allow the formation of transfection complexes. The complex was then added dropwise onto the cells. The plates were then rotated gently to ensure uniform distribution of the transfection complexes. After 6 h, 140 ul complete culture media of RPMI containing serum and antibiotics were added to the PBMCs and incubated for 48 h for RNA extraction and subsequent analysis of gene silencing or induction. Total RNA was isolated from TAMs, which were previously differentiated from CD14+ monocytes and treated against diluted silencers. Total RNA was extracted using Thermo Scientific GeneJET RNA Purification kit (Catalog no: K0732), according to the manufacturer’s protocol. In brief, lysis buffer and absolute ethanol were added to each sample eppendorf and centrifuged for 1 min at 12,000 × g. The flow-through solution in the collection tube was then discarded. Each sample was washed twice with wash buffer 1 and wash buffer 2. Finally, the collection tube containing the flow-through solution was discarded, and the GeneJet RNA purification column was transferred to a sterile 1.5-ml RNAse-free microcentrifuge tube. Nuclease-free water was finally added, and the centrifugation was repeated. Purified RNA was stored in a −80°C freezer. Total RNA extracted was reverse transcribed into the single-stranded cDNA using the high-capacity cDNA reverse transcription kit (Thermo Fisher, Cat No: K1652), according to the manufacturer’s protocol. Each component of the reverse transcription kit and the extracted RNA of each sample were thawed on ice and mixed by vortexing to ensure appropriate resuspension. Each reaction’s total tube volume was 20 ul with 1:1 ratio (reaction mix: total RNA). Finally, the reaction tubes were placed in a thermo cycler with a heated lid whose thermal profile was adjusted, according to the manufacturer’s protocol. All the cDNA samples were stored in the −20°C freezer until qRT-PCR analysis was performed. The expressions of MALAT1, HOTAIR, CD80, and MSLN mRNA levels were quantified using RT-PCR. Reagents used were the TaqMan, MALAT1, HOTAIR, CD80, and MSLN expression assays (Themo Fisher (United States)-TaqMan MALAT1 assay (Hs00273907), TaqMan CD80 assay (Hs01045161_m1), TaqMan HOTAIR assay (Hs05502358_s1), and TaqMan MSLN assay (ID: Hs00245879) along with B-actin as an endogenous control housekeeping gene to normalize the expression values. Probes used for MALAT1, HOTAIR, CD80, and MSLN were labeled with the FAM reporter dye. B-actin was reported with the VIC reporter dye. Each reaction tube’s total volume was 20 ul with 1:4 ratio (total cDNA:reaction mix). Each reaction mix was composed of nuclease-free water, Premix Ex TaqTM (Probe qPCR), TaqMan target gene assay expression assay (x20), and B-actin (VIC). Upon preparing the reaction tubes, they were placed into the StepOne® real-time PCR instrument and the run was performed in the standard mode, consisting of two stages. A first stage where the Taq-polymerase enzyme is activated followed by the second stage of 40 amplification cycles (each cycle comprises a 15 second denaturation step followed by 60 s of annealing and extension). The StepOne® real-time PCR yields a cycle threshold value (Ct) for each sample, which represents the fractional cycle number at which the fluorescence produced exceeds a threshold line. Each cycle threshold (Ct) obtained was subsequently used for quantification of the amplified target compared to its endogenous control housekeeping gene (B-actin for target genes), yielding a ΔCt value. After transfecting TAMs with silencers against MALAT1 and HOTAIR, VEGF-A protein was quantitatively examined using the VEGF-A Human ELISA Kit (Invitrogen, BMS277-2). The kit is a sandwich ELISA, which is designed to measure the amount of the target bound between a matched antibody pair. Briefly, 400 μL wash buffer per well was used to wash pre-coated microwells from the microplate. After washing, different concentrations of standards were added to the wells, according to the manufacturer’s protocol. The plates were incubated for 2 h at room temperature. After washing, 100 μL of biotin-conjugated antihuman VEGF-A polyclonal antibody (1:100) was then added to the wells and incubated for 1 h at room temperature. After washing, 100 μL of horseradish peroxidase-labeled streptavidin was added to the wells and incubated for 1 h at room temperature. After washing, 100 µL of TMB substrate solution was added to all wells. Color development on the plate is monitored for 30 min at room temperature, then the stop solution was added to terminate the reaction, and the absorbance was analyzed for each microwell for both standards and samples at 450 nm wavelength. The results are calculated by constructing a standard curve plotting the mean OD and concentration for each standard. MDA-MB-231 cell line was purchased from the tissue culture unit, Egyptian Company for Vaccines and Sera, after proper authentication and testing for mycoplasma contamination. MDA-MB-231 cells are ER, PR, and HER2-. MDA-MB-231 cells were cultured using our previous protocol (Hamed et al., 2021). The cell line was cultured in a suitable culture media [high-glucose DMEM supplemented with 10% FBS and 1% penicillin/streptomycin in 10 cm Petri dishes (Gibco, United States)]. Culture media were changed every 2–3 days until the cells reached 80–90% confluency. The cells were washed with PBS, trypsinized, and split into two clean Petri dishes. After the second splitting, the cells were harvested and prepared for further experimentation. The cells were kept in a 37°C, 5% CO2 incubator. Prior to seeding and experimentation, the cells were counted and checked for viability using trypan blue. In a 96-well plate, CD8+ T cells were cultured (2-3 x 104 cells per well) in 200 μL culture media (100 μL of TAMs supernatant +100 μL of full RPMI media) for 24 h. The supernatant of TAMs culture media was isolated after TAM culturing for 7 days under four conditions (untransfected TAMS, siMALAT1 TAMs, siHOTAIR TAMs, and siMALAT1/HOTAIR TAMs). After culturing CD8+ T cells in TAM-conditioned media, LDH toxicity assay was performed to evaluate the cytotoxicity potential of CD8+ T cells upon co-culturing with MDA-MB-231 cell line with and without the addition of PDL-1 inhibitor drug. LDH assay is used as rapid determination of cytotoxicity based on lactate dehydrogenase released into the cell culture medium using the Canvax Biotech protocol (Cat No: CA0020). In the LDH cytotoxicity assay, the lysis control wells are first prepared with the addition of lysis solution and incubating the plate in a 37°C, 5% CO2 incubator for 45 min. Then, 50 μL of culture supernatant from each well is transferred to a new 96-well flat-bottom plate and the reaction mixture is then added (50 μL on each well). After incubation for 30 min, the reaction is terminated by the addition of stop solution (50 μL on each well). Absorbance for all controls and experimental samples is measured at 450 nm wavelength. Data are analyzed by measuring the % relative cytotoxicity. To evaluate the CD8+ T-cell isolation efficiency using flow cytometry [CytoFLEX, Beckman Coulter Life Sciences (United States)], CD8+ was measured using flow cytometry anti-CD8 PE (Biolegend, Cat No: 344706). In brief, the cells were first dissociated, and then, single-cell suspensions were prepared (240,000 cell/tube). The cells were then washed twice with 2 mls (PBS 1% FBS) and centrifuged at 350 × g for 5 min. After washing, 1.2 µg anti-CD8 PE was added (5 µg per one million cells), and the cells were incubated at 4° for 30 min. All data were expressed in relative quantitation (RQ) for RT-qPCR. For the purpose of comparison between two different studied groups, Student’s unpaired t-test was used. One-way Anova followed by Dunette’s test of multiple comparison was used for the comparison between more than two different studied groups. Data were expressed as mean ± standard error of the mean (SEM). A p-value less than 0.05 were considered statistically significant **** = p < 0.0001, *** = p < 0.001, ** = p < 0.01, and * = p < 0.05. Analysis was performed using GraphPad Prism 7.02 software. The expression of MALAT1 was found to be upregulated in TAMs of the hormonal, HER2+, and TNBC compared to healthy donors (p = 0.0002, 0.0001, and 0.0001, respectively). In a similar pattern, the expression of HOTAIR was found to be upregulated in TAMs of the three subgroups (hormonal, HER2+, and TNBC) compared to healthy donors (p=<0.0001, <0.0001, and <0.0001, respectively), as shown in Figure 1. The expression of CD80 was found to have a non-significant difference in TAMs of the hormonal subtype compared to healthy donors. However, it was significantly upregulated in TAMs of both HER2+ and TNBC subtypes (p = 0.0008 and 0.0016 respectively) compared to TAMs of healthy donors. Meanwhile, the expression of MSLN was found to have significant downregulation in TAMs of the hormonal subtype (p = 0.0002) compared to healthy donors. However, it was significantly upregulated in TAMs of both HER2+ and TNBC subtypes (p = 0.0023 and <0.0001, respectively) compared to TAMs of healthy donors, as shown in Figure 2. For transfection efficiency purposes, HOTAIR and MALAT1 expressions were analyzed after transfection with their silencers (siRNAs). The transfection of TAMs with silencers against HOTAIR resulted in downregulation, thus silencing the expression of HOTAIR in hormonal, HER2+, and TNBC (p = 0.0010, <0.0001, and 0.0059, respectively) in comparison to untransfected controls (mocks). Moreover, the transfection of the TAMs with silencers against MALAT1 resulted in downregulation, thus silencing of the expression of MALAT1 in hormonal, HER2+, and TNBC (p = 0.0009, 0.0008, and <0.0001 respectively), as shown in Figure 3. Silencing of HOTAIR in TAMs of different BC subtypes resulted in a decrease in MALAT1 expression in the hormonal, HER2+, and TNBC subgroups (p = 0.0009, <0.0001, and <0.0001, respectively) in comparison to untransfected controls (mocks). Unexpectedly, in the three subtypes, MALAT1 showed a more significant downregulation upon silencing with HOTAIR than upon silencing with MALAT1 itself in hormonal, HER2+, and TNBC (p = 0.0003, 0.0002, and <0.0001, respectively) in comparison to untransfected TAMs (mocks), as shown in Figure 4. Silencing of MALAT1 in TAMs of different subtypes resulted in an increase in the expression of HOTAIR in hormonal, HER2+, and TNBC TAMs (p = 0.0008, <0.0001, and 0.0009, respectively) in comparison to untransfected controls (mocks). HOTAIR mRNA levels were quantified using qRT-PCR and normalized to β-actin as an endogenous control, as shown in Figure 5. CD80 was found to be inversely correlated with MALAT1 in HER2+ and TNBC TAMs, its expression was upregulated by siMALAT1 (p = 0.0478 and 0.0012 respectively), while the upregulation was more remarkable by siHOTAIR in HER2+ and TNBC TAMs (p = 0.0013 and <0.0001, respectively) in comparison to untransfected controls (mocks). Conversely, in hormonal BC TAMs, CD80 was downregulated by siMALAT1 (p = 0.0007) in comparison to untransfected controls (mocks). MALAT1/HOTAIR co-silencing was able to upregulate CD80 in TAMs of hormonal, HER2+ and TNBC (p= <0.0001) in comparison to untransfected controls (mocks), as shown in Figure 6. MSLN was found to have a similar pattern of expression as CD80, where it was inversely correlated with MALAT1 and HOTAIR in HER2+ and TNBC TAMs. Its expression was upregulated by siMALAT1 (p = 0.0367 and 0.0015), while the upregulation was much remarkable by siHOTAIR in HER2+ and TNBC TAMs (p = 0.0066 and 0.0003, respectively) in comparison to untransfected controls (mocks). However, MSLN expression showed no significant change in the expression in TAMs of the hormonal subtype upon silencing of HOTAIR or MALAT1 in comparison to untransfected controls (mocks). Meanwhile the most significant upregulation for MSLN was found upon MALAT1/HOTAIR co-silencing in hormonal, HER2+, and TNBC (p= <0.0001, 0.0002, and <0.0001, respectively) in comparison to untransfected controls (mocks). As a functional analysis, the ELISA technique was used to measure the VEGF-A in the TAMs of the three subgroups (hormonal, HER2+, and TNBC) on silencing MALAT1 or HOTAIR or co-silencing both lncRNAs together in comparison to untransfected TAMs (mocks) corresponding to each subtype. It was shown that VEGF-A was downregulated upon silencing MALAT1 and HOTAIR and co-silencing MALAT1 and HOTAIR in the hormonal BC subtype (p = 0.0155, 0.0330, and 0.0199) compared to mock cells. The same downregulation expression profile was observed in the HER2+ subtype upon silencing MALAT1 and HOTAIR and co-silencing MALAT1 and HOTAIR (p = 0.0006, 0.0015, and 0.0002) compared to mock cells. Finally, TAMs of TNBC also showed a significant downregulation upon silencing MALAT1 and HOTAIR and co-silencing both of them together (p = 0.0009, 0.0025, and 0.0007) compared to untransfected controls (mocks). LDH toxicity assay was performed to evaluate the cytotoxicity potential of CD8+ T cells upon co-culturing with MDA-MB-231 cell line with and without the addition of PDL-1 inhibitor drug. CD8+ T cells were cultured under three conditions of treated TAMs-conditioned media (siMALAT1, siHOTAIR, and siMALAT1+siHOTAIR). The results showed that the cytotoxicity percentage of CD8+ cells upon culturing in siMALAT1, siHOTAIR, and siMALAT1/siHOTAIR TAMs had an average of approximately 76.6% cytotoxicity; however, upon the addition of the PDL-1 inhibitor on CD8+ T cells that is previously cultured in siMALAT1, siHOTAIR, and siMALAT1/siHOTAIR TAMs-conditioned media, the cytotoxicity percentages had an average of approximately 61.3% and upon addition of the PDL-1 inhibitor alone on CD8+/MDA-MB-231 co-culture, the cytotoxicity was 79%. The morphological change was assessed to confirm the differentiation efficiency of CD14+ monocytes to TAMs. CD14+ monocytes were isolated from total PBMCs by negative selection using magnetic nanobeads against antibodies of all cells other than CD14+ monocytes. Freshly isolated CD14+ monocytes were observed under the microscope. Examination showed small, spherical-shaped cells with a smooth surface. However, upon culturing CD14+ monocytes for 7 days in culture media supplemented with anti-inflammatory ILs (IL-10, IL-4, and M-CSF) with the addition of TCM, morphological change was observed where the cells became larger with a non-smooth edgy surface. To assess the isolation purity of CD8+ T cells, the flow cytometry technique was conducted using anti-CD8+ PE after isolating the CD8+ T cells from PBMCS using magnetic nanobeads. The results showed that the isolation purity reached 70.2%. Breast cancer (BC) is one of the most commonly diagnosed cancers in women worldwide, accounting for 11.7% of all cancer cases in 2020. It is classified molecularly according to the expression of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) with the percentage of the proliferating index (Ki67). Accordingly, BC is classified into five major subclasses, which are luminal A, luminal B HER2-, luminal B HER2+, HER2 enriched, and TNBC. Luminal A is the most abundant subtype having a better prognosis than luminal B. Furthermore, HER2+, which is a hormonal negative subtype, has a worse prognosis than the hormonal subtypes despite the presence of anti-HER2-targeted therapy “for example, trastuzumab.” Finally, TNBC (basal-like) has the worst and most aggressive prognostic profile of all the subtypes (Kondov et al., 2018). The most common BC treatment strategies are hormonal therapy, anti-HER2-targeted therapy, and chemotherapy. Both hormonal and anti-HER2-targeted therapies are limited to specific subtypes of BC. Moreover, chemotherapy has the ability to target all BC subtypes but causes profoundly aggressive side effects (Yerushalmi et al., 2009). In this context, targeted immunotherapy can be a new therapy alternative to overcome the limitations of other therapeutic strategies. In general, tumors have the ability to recruit stromal cells (e.g., fibroblasts), immune myeloid cells, and vascular cells by the secretion of cytokines building up a tumor microenvironment (TME) by releasing growth-promoting signals and remodeling the tissue structure affecting initiation, progression, metastasis, vascularization, and therapy responses. During the last decades, cancer treatment strategies focus only on the cancer cell ignoring the TME, which is a key player in the tumor progression (Allaoui et al., 2016). Within the TME, CD14+ monocytes can infiltrate and differentiate into tumor-associated macrophages (TAMs) that can reach 40% of the tumor’s volume. TAMs are known to have pro-inflammatory phenotype (M1) in the early stages of cancer while having anti-inflammatory (M2) phenotype once cancer starts progressing in suppressing many immune cells. The differentiation to M2 takes place in the presence of some anti-inflammatory cytokines in the TME, namely, IL-10, IL-4, IL-13, and M-CSF (Zhou et al., 2020). In addition to TAMs correlation with tumor progression, there are some immunomodulatory proteins that have been reported in the last years to be overexpressed in various types of cancers, especially CD80 and mesothelin (MSLN). Furthermore, in the past few years, the role of non-coding RNAs in the pathogenesis of BC has been of significant importance and their correlation with immunomodulatory proteins, as shown in our previous study (Hussein et al., 2022). Interestingly, HOTAIR and MALAT1 are known to be the master regulators of tumor progression (Aiello et al., 2016). In our previous work, MALAT1 was shown to be upregulated in BC tissues, especially TNBC epigenetically upregulating MSLN; Barsoum et al. (2020) proposed that there may be a possible interlinkage between MALAT1 and immunomodulatory proteins. The aim of the current study is to investigate if MALAT1 will give the same pattern of expression in TAMs of the hormonal, HER2+, and TNBC, highlighting that its role in cancer is still controversial whether it is an oncogenic or a tumor suppressor lncRNA. Additionally, HOTAIR is another lncRNA that is known to function as an oncogenic lncRNA as previously mentioned. Therefore, the aim of this work is to study the interlinkage between MALAT1 and HOTAIR and their effect on the expression of oncogenic immunomodulatory proteins in TAMs of the hormonal, HER2+, and TNBC subtypes. To fulfill this aim, MALAT1 and HOTAIR expressions were first screened. MALAT1 was found to be over expressed in hormonal, HER2+, and TNBC compared to controls, as shown in Figure 1A. This finding aligns with our previous studies confirming that MALAT1 is overexpressed in BC tissues more than healthy tissues (Barsoum et al., 2020). Nevertheless, another research study observed the overexpression of MALAT1 in many BC cell lines (MCF-7, SK-BR-3, MDA-MB-468, MDA-MB-231, T-47d, and MDA-MB-453) compared to normal breast epithelial cell line (MCF-10A) (Yue et al., 2020). Moreover, MALAT1 was shown to be overexpressed in other cancers such as multiple myeloma (Liu et al., 2020) and hepatocellular carcinoma (HCC) (Lai et al., 2012). The screening of HOTAIR expression in TAMs of hormonal, HER2+, and TNBC also showed a significant upregulation in BC samples compared to healthy samples, as shown in Figure 1B. This finding is supported by another study that was conducted on MCF-7 and MDA-MB-231, showing significant upregulation of lncRNA HOTAIR when compared to MCF-10A (Wang et al., 2020). Moreover, HOTAIR is shown to be not only overexpressed in BC but also in ovarian cancer (Qiu et al., 2014); thus, suggesting that HOTAIR and MALAT1 are oncogenic lncRNAs. Therefore, it was tempting to study the impact of lowering the expression of MALAT1 and HOTAIR in TAMs. Our research group has previously studied the role of these lncRNAs along with their candidate downstream targets, immunomodulatory proteins, CD80, and MSLN in TNBC tissues and MDA-MB-231 (Barsoum et al., 2020). However, here, this study is concerned with their expression pattern in TAMs. CD80 and MSLN were found to be overexpressed in HER2+ and TNBC while showing a non-significant change of expression on examining the hormonal subtype, as shown in Figure, suggesting that there is another key player regulating the expression of CD80 in the hormonal subtype and preventing its overexpression. This key player may be estrogen that is known in the literature to downregulate CD80 levels. It was previously observed that estrogen has an inverse correlation with CD80, as shown in a study conducted on RAW264.7 cell line (Yang et al., 2016). Highlighting the fact that RAW264.7 is a murine macrophages cell line, similarity in the expression is expected. Moreover, another research group showed that estrogen promotes the B cell activity in vitro by downregulating CD80 expression (Fu et al., 2011). Thus, concluding that estrogen may be the factor inhibiting CD80 to upregulate. Upon screening of MSLN in TAMs of the three subgroups, it was found that MSLN is also overexpressed in HER2+ and TNBC, while it was significantly downregulated in the hormonal subtype, as shown in Figure 2B, suggesting that estrogen plays a role in this downregulation. This finding is partially supported by a study that screened 99 primary BC samples by immunohistochemistry analysis using formalin-fixed paraffin-embedded archival tumor tissues confirming that MSLN was only overexpressed in TNBC (67%) in contrast to its rare expression in the hormonal and the HER2+ subtypes (Tchou et al., 2012). These data partially support our findings that MSLN is not upregulated in the hormonal subtype (possibly due to the presence of estrogen), however, contradicting with our data that suggest MSLN to be upregulated in HER2+. In another study, MSLN expression was detected in 77 cases out of 482 patients (16.0%) and was the highest in TNBC (31/75; 41.3%), followed by the HER2+ subtype (6/33, 18.2%), and then the luminal subtype (36/374; 9.6%) (Suzuki et al., 2020). Moreover and more interestingly, MSLN was shown in our previous work to be upregulated in BC subtypes, especially TNBC (Barsoum et al., 2020). To observe the impact of knocking down of MALAT1 and HOTAIR in TAMs of the three BC subtypes (hormonal, HER2+, and TNBC), the correlation between MALAT1 and HOTAIR was first observed showing that on the silencing of HOTAIR, MALAT1 was downregulated in all subtypes of BC, as shown in Figure 4, suggesting that HOTAIR may be an upstream regulator for MALAT1. Moreover, upon silencing MALAT1, HOTAIR expression was upregulated, and this was observed in the three subtypes, as shown in Figure 5, suggesting that HOTAIR expression may have increased as a compensatory mechanism for the loss of MALAT1 expression. These findings suggest that MALAT1 expression is directly correlated with HOTAIR expression; however, it was reported that HOTAIR and MALAT1 have opposite expression profiles in estrogen-mediated transcriptional regulation in prostate cancer cells (Rinn et al., 2007). These differences may be due to cancer specificity or tissue specificity as this study observed cancer cell lines in prostate cancer; however, our study is on TAMs in BC. Moreover, the opposite expression profile may be due to the fact that estrogen has an opposite correlation with MALAT1 expression in prostate cancer, as shown in the mentioned study, thus upon downregulation of MALAT1, HOTAIR is overexpressed as a compensatory mechanism, thus having an inverse correlation. Knowing that in the same study upon treatment of the BC cell line with estrogen, no change in the expression of MALAT1 was observed (Aiello et al., 2016), thus no correlation between estrogen and MALAT1 expression in BC supporting our identical finding that MALAT1 and HOTAIR show same overexpression patterns in the three subgroups regardless the hormonal expression. Furthermore, in another study that aligns with our finding states that MALAT1 and HOTAIR have a positive correlation with each other after conducting a correlation analysis between their serum levels in BC (El-Fattah et al., 2021). Unexpectedly, upon comparing MALAT1 expression after silencing its gene vs. silencing HOTAIR, it was shown that the HOTAIR knockdown effect was more significant in downregulation of MALAT1 compared to knocking down of MALAT1 itself, as shown in Figure 4. As a consequence of the CD80 overexpression in HER2+ and TNBC TAMs, thus highlighting the possibility of being an oncogenic immunomodulatory protein (Li et al., 2020), it was worth observing the effect of MALAT1 as an important lncRNA in the regulation of CD80. Unexpectedly, our data showed that upon silencing of MALAT1, CD80 expression was upregulated, thus having an inverse correlation in HER2+ and TNBC, as shown in Figures 6B,C, respectively. This finding aligns with another study showing MALAT1 and CD80 to be inversely correlated in A549 cells (neonatal respiratory distress syndrome) (Juan et al., 2018). MALAT1 and CD80 inverse correlation was also supported in another research study that suppressed MALAT1 in the dendritic cells (Wu et al., 2018). Observing this inverse correlation in dendritic cells validate our data since that macrophage (TAMs) and dendritic cells are both derived from the same progenitor (monocytes), thus having similarities in the expression can be expected. CD80 is a ligand for two receptors on T cells, namely, CD28 and CTLA4. T cells are activated upon binding of CD80 and CD28, thus having a pro-inflammatory activity. However, on binding with CTLA4, T cells were found to be suppressed causing energy and anti-inflammatory activity. Surprisingly, it was found that CTLA4 has a higher affinity to bind with CD80 than CD28 (Lai et al., 2012). Taking this into account, this study proposes that the overexpression pattern of CD80 in TAMs of the three subgroups, as shown in Figure 2, occurs due to the immune-suppressive effect of CD80 due to its binding to CTLA4 on T cells. However, CD80 shifts its binding toward CD28 upon silencing MALAT1, thus its overexpression has immunostimulatory activity, and this may happen due to the downregulation of CTLA4 as a result of downregulation of MALAT1. The relationship between MALAT1 and CTLA4 is not previously studied in cancer; however, its data in asthma showed that MALAT1 sponges miR-155 upregulating CTLA4 (Liang and Tang, 2020). In other words, our study proposes that the overexpression of MALAT1 in TAMs might have upregulated CTLA4, thus upregulating CD80 to enhance its immunosuppressive activity and build up an environment favorable for its tumorigenic role. Furthermore, it was expected that CD80 to be downregulated upon the silencing of MALAT1 but unexpectedly an inverse correlation was observed. In this context, we propose that this might have happened after silencing MALAT1, thus consequently and considerably impacting the oncogenic potential of TAMs making the CD80 regain its potential (in the absence of MALAT1) to function as a pro-inflammatory protein, thus upregulation was observed. Focusing onto the hormonal subtype, CD80 expression was observed to be significantly downregulated upon silencing MALAT1, as shown in Figure 6A. This could be due to the dominancy of the effect of estrogen in downregulating CD80. Estrogen might have the potential to abolish the upregulation effect caused by MALAT1 silencing. This finding supports our previous conclusion that estrogen and CD80 are inversely correlated to each other, and upon analyzing the effect of silencing of HOTAIR on the expression profile of CD80, the same expression pattern was observed, in which CD80 was upregulated in HER2+ and TNBC, as shown in Figures 6B,C, respectively. The correlation between CD80 and HOTAIR is not studied before, thus in this context, this is considered the first research study to be conducted focusing on HOTAIR and CD80 correlation. Focusing on to the hormonal subtype, CD80 expression is shown to have a non-significant change in the expression upon HOTAIR silencing compared to mock, as shown in Figure 6A. MSLN is known to be an oncogenic immunomodulatory protein in many types of cancers, for example ovarian, adenocarcinoma, and most importantly BC (Wang et al., 2020). Our previous work confirmed that MSLN is overexpressed in BC tissues and cell lines, thus functioning as an oncogenic protein (Barsoum et al., 2020). So, it was a good candidate to examine its correlation with MALAT1 and HOTAIR. Upon silencing MALAT1, MSLN expression had the same pattern as CD80 in both HER2+ and TNBC. Unexpectedly, upon silencing MALAT1, MSLN expression was upregulated in both non-hormonal subtypes: HER2+ and TNBC, as shown in Figures 7B,C, respectively. The relationship between MALAT1 and MSLN is not studied except in our previous work on TNBC tissues that showed downregulation of MSLN upon MALAT1 silencing (Barsoum et al., 2020). The contradiction might be due to the difference in the cell type studied as our previous work was on the cancer cell; however, this work focused on TAMs (immune cell). It is important to highlight that MSLN overexpression correlates to the increased levels of soluble MSLN that by its turn binds to CD206 (mannose receptor) via GPI anchor facilitating macrophages polarization to TAMs (Dangaj et al., 2011). In this context and as mentioned before it was unexpected to observe an upregulation in MSLN expression after MALAT1 silencing. MALAT1 silencing consequently and considerably impacted the oncogenic potential of TAMs. The unexpected upregulation of MSLN upon MALAT1 silencing might have happened as a compensatory mechanism performed by TAMs for the loss of MALAT1. Upon upregulation of MSLN, soluble MSLN levels would also increase, thus increasing the binding potential to CD206. As a consequence, the polarization of macrophages to TAMs would be enhanced. Focusing on to the hormonal subtype, MSLN gives a close expression pattern to CD80, its change in expression is non-significant compared to controls upon MALAT1 silencing, as shown in Figure 7A, proposing again that estrogen is a factor playing an important role in downregulating MSLN, thus balancing the upregulation effect caused by MALAT1 silencing in the hormonal subtype leading to a non-significant change in expression of MSLN compared to mocks. Upon silencing HOTAIR, MSLN gave the same pattern of expression on silencing with MALAT1 in HER2+ and TNBC. MSLN was shown to be upregulated in both HER2+ and TNBC, as shown in Figures 7B,C, respectively. Interestingly, the correlation between HOTAIR and MSLN is not been previously studied on any cell type, thus studying this correlation in TAMs highlights the importance of investigating this relationship in other cell types either cancer cells or immune cells. On comparing the MSLN or CD80 expression upon silencing with HOTAIR and upon silencing with MALAT1 in both HER2+ and TNBC, immunomodulatory proteins are found to be upregulated more significantly upon silencing with HOTAIR, as shown in Figures 6B,C and Figures 7B,C. This is may again be due to that HOTAIR is the upstream regulator for MALAT1 as previously mentioned having a greater effect on the downregulation of MALAT1 than knockdown of MALAT1 itself. Thus, upon silencing HOTAIR, MALAT1 is abolished, making the downregulation of MALAT1 significantly remarkable and consequently more significant upregulation for immunomodulatory proteins was observed. Co-transfection of silencers against MALAT1 with silencers against HOTAIR was expected to affect the expression of CD80 and MSLN more significantly than silencing each lncRNA on its own. Thus, upon silencing both lncRNAs, CD80 and MSLN were upregulated more significantly compared to MALAT1 or HOTAIR separately in HER2+ and TNBC, as shown in Figures 6B,C and Figures 7B,C. As for the hormonal subtype, a significant upregulation of CD80 and MSLN was observed upon co-transfection of silencers against MALAT1 and HOTAIR (Figure 6A and Figure 7A). These results were expected due to the dual effect of the simultaneous silencing of HOTAIR and MALAT1 and their success to dominate the effect of downregulation caused by estrogen, confirming that, in general, MALAT1 and HOTAIR are inversely correlated with CD80 and MSLN in the three subtypes, but estrogen can dominate in the hormonal subtype causing the effect of silencing HOTAIR and MALAT1 to be masked. As previously mentioned, there are various studies that tackled the role of MALAT1 and HOTAIR functionally as oncogenic lncRNAs in breast cancer tissues. However, TAMs were still questionable and not confirmed whether they were functioning as oncogenic or tumor suppressor lncRNAs. For this reason, one of the functional analyses for oncogenesis and metastasis was observed to be a confirmatory tool for the role of these two lncRNAs. Vascular endothelial growth factor A (VEGF-A) was chosen as a suitable candidate for the functional analysis for two major reasons: first, to confirm that TAMs have a role in metastasis through VEGF-A, highlighting that the role of TAMs in metastasis is functionally analyzed for the first time, and the second reason was that VEGF-A is considered a reflection for the metastasis thus tumorigenesis, and this will confirm whether MALAT1 and HOTAIR are oncogenic or tumor suppressor. The VEGF-A protein level was shown to be downregulated upon silencing lncRNAs whether separately or simultaneously in the hormonal, HER2+, and TNBC, as shown in Figure 8, confirming the idea that MALAT1 and HOTAIR are oncogenic lncRNAs in TAMs. The regulatory role of TAMs on CD8+ T cells was for the first time assessed in the presence of MDA-MB-231 cell line. The purity of CD8+ T cells isolation was assessed using flow cytometry as shown in Figure 9. CD8+ T cells (previously cultured in transfected TAMs-conditioned media) were co-cultured with MDA-MB-231 cell line. As previously mentioned, TAMs have an inhibitory role on CD8+ T cells, thus our work aims to observe if the silencing of MALAT1 and HOTAIR will impact this immune-suppressive activity. The immune-suppressive activity was assessed by observing the cytotoxic activity of CD8+ T cells on MDA-MB-231 cell line. LDH assay was conducted showing that upon culturing of CD8+ T cells in TAMs-conditioned media previously treated with silencers against MALAT1 and HOTAIR, the cytotoxicity activity of CD8+ T cells is increased, as shown in Figure 10, supporting the fact that MALAT1 and HOTAIR are oncogenic lncRNAs, and upon their knocking down, the T cells restored its immunostimulant cytotoxic activity. Programmed cell death protein 1 (PDL-1) is known to be a negative regulator expressed on the surface of T cells. PDL-1 binds to its ligand programmed death ligand-1 (PD-1) on the tumor cell suppressing the activity of CD8+ T cell (53). Consequently, upon culturing CD8+ T cells with MDA-MB-231 cell line, the cytotoxic activity of T cells is expected to be suppressed, as a result, the PDL-1 inhibitor drug was added to the co-culture media (CD8+ T cells cultured in treated TAM-conditioned media) to observe the impact of adding the PDL-1 inhibitor with oncogenic lncRNAs silencing, and whether this addition will cause a synergetic activity increasing the cytotoxicity of CD8+. Unexpectedly, the cytotoxicity of CD8+ T cells was not increased upon the addition of the PDL-1 inhibitor in comparison to CD8+ cytotoxicity upon only silencing of the oncogenic lncRNAs without the addition of PDL-1, as shown in Figure 10. Our study suggests that this might have happened either due to the PDL-1 inhibitor, which does not have any additional role in increasing the cytotoxicity of CD8+ T cells or that the dose of PDL-1 inhibitor has to be adjusted knowing that the dose used was 200 nM as indicated in the previous literature using same immune cell (CD8+) and same cell line (MDA-MB-231) (Passariello et al., 2019). In all experiments conducted, The efficiency of monocytes differentiation into TAMs was confirmed using microscopic examination as shown in Figure 11. In conclusion, as shown in Figure 12, HOTAIR is suggested to be an upstream regulator for MALAT1 because upon downregulation of HOTAIR, MALAT1 was also downregulated. Supporting this, downregulation of MALAT1 led to upregulation of HOTAIR as a possible compensatory mechanism explaining that both have the same function. Upon downregulation of MALAT1 and HOTAIR in HER2+ and TNBC, unexpectedly, upregulation of CD80 and MSLN was observed with the fact that silencing HOTAIR was more significant, and upon co-silencing HOTAIR and MALAT1, the expressions of both were upregulated more significantly than silencing each lncRNA, separately. Moreover, on silencing MALAT1 and HOTAIR in the hormonal subtype, estrogen played an important role due to its effect in downregulation of CD80 and MSLN expressions, thus masking the effect of MALAT1 silencing in upregulating CD80 or MSLN. Certainly, upon co-transfection, the expressions of CD80 and MSLN were upregulated due to the dual action of silencing both lncRNAs, thus dominating the downregulatory activity of estrogen. It has been concluded that MALAT1 and HOTAIR might be oncogenic lncRNAs in TAMs. This finding was confirmed by the upregulation of VEGF-A protein on silencing MALAT1 and HOTAIR in TAMs of BC and also on assessing the cytotoxicity activity of CD8+ T cells on knocking down of these two lncRNAs where the cytotoxicity activity was increased. Future recommendations involve conducting the dose–response curve for the PDL-1 inhibitor, measuring MSLN and CD80 on the protein level, and examining the interlinkage between MALAT1 and CTLA4 in TAMs of BC.

PMC9649624

Yunxi Liu,R. A. Leo Elworth,Michael D. Jochum,Kjersti M. Aagaard,Todd J. Treangen

De novo identification of microbial contaminants in low microbial biomass microbiomes with Squeegee

10-11-2022

Classification and taxonomy,Metagenomics

Computational analysis of host-associated microbiomes has opened the door to numerous discoveries relevant to human health and disease. However, contaminant sequences in metagenomic samples can potentially impact the interpretation of findings reported in microbiome studies, especially in low-biomass environments. Contamination from DNA extraction kits or sampling lab environments leaves taxonomic "bread crumbs" across multiple distinct sample types. Here we describe Squeegee, a de novo contamination detection tool that is based upon this principle, allowing the detection of microbial contaminants when negative controls are unavailable. On the low-biomass samples, we compare Squeegee predictions to experimental negative control data and show that Squeegee accurately recovers putative contaminants. We analyze samples of varying biomass from the Human Microbiome Project and identify likely, previously unreported kit contamination. Collectively, our results highlight that Squeegee can identify microbial contaminants with high precision and thus represents a computational approach for contaminant detection when negative controls are unavailable.

De novo identification of microbial contaminants in low microbial biomass microbiomes with Squeegee Computational analysis of host-associated microbiomes has opened the door to numerous discoveries relevant to human health and disease. However, contaminant sequences in metagenomic samples can potentially impact the interpretation of findings reported in microbiome studies, especially in low-biomass environments. Contamination from DNA extraction kits or sampling lab environments leaves taxonomic "bread crumbs" across multiple distinct sample types. Here we describe Squeegee, a de novo contamination detection tool that is based upon this principle, allowing the detection of microbial contaminants when negative controls are unavailable. On the low-biomass samples, we compare Squeegee predictions to experimental negative control data and show that Squeegee accurately recovers putative contaminants. We analyze samples of varying biomass from the Human Microbiome Project and identify likely, previously unreported kit contamination. Collectively, our results highlight that Squeegee can identify microbial contaminants with high precision and thus represents a computational approach for contaminant detection when negative controls are unavailable. In recent years, the field of metagenomics has grown at a fast pace thanks to next-generation sequencing technologies. The scale and complexity of metagenomics studies have expanded alongside the volume of the sequencing data. By performing metagenomic sequencing, we are able to analyze the DNA and RNA of the entire microbial community in varying and heterogeneous biomass environments, such as samples from wastewater, soil, or human body sites. One commonly used method is 16S rRNA gene sequencing. The 16S rRNA gene is highly conserved in bacteria and can be amplified and used as a marker gene for taxonomic classification. The other widely used technique is whole-genome shotgun sequencing, where all DNA sequences in the community are fragmented and sequenced. Both methods open the door for identifying members of microbial communities from the sampled environments and estimating the relative abundance of each member. However, the results from both of these methods can be affected by microbial contamination. Microbial contamination occurs when sequences from microbes appear in the data that were not in the original samples. A variety of sources can introduce microbial contamination. External sources include personnel, the laboratory environment, and kits and reagents used for collecting and processing samples. Internal sources of contamination may include human error, such as sample mislabeling or inadvertent mixing. Contaminant sequences have also made their way into public reference databases. Studies have shown that contaminants in DNA extraction kits are ubiquitous, and can bear an impact on metagenomic studies, especially for low-biomass environments if they are not accounted for in the analysis. For example, in a recent nasopharyngeal microbiota study on newborn babies conducted in Thailand, contaminants found in DNA extraction kits resulted in contaminant bias. Extra precautions during sample collection and processing, and well-designed experiments, such as processing samples in a clean, well-structured environment, or using depletion methods to remove host DNA, can help minimize the impact caused by contamination. In addition, computational models have been used to identify and remove contaminants from sequenced datasets. For example, the recently published software Recentrifuge uses a score-oriented comparative approach to identify and remove contaminants from sequencing reads. As is the case with all current computational methods for microbial contaminant detection, performing contamination removal with Recentrifuge requires experimental controls. Another statistical tool for identifying and removing contamination is Decontam. Decontam includes a combination of a frequency-based approach and a prevalence-based approach. Auxiliary DNA quantitation data are required to perform the frequency-based analysis, and standard negative control samples are required to perform the prevalence-based analysis. Experimental negative and/or environmental contaminant controls combined with computational contamination identification and removal is effective. However, the additional costs (both time and resources) to include negative control experiments are often a barrier to utilization. As a result, negative control experiments available for publicly available datasets are often lacking. Although contaminant sequences have been a known issue for some time, negative control data are often unavailable in public databases, making it nearly impossible to perform contamination removal on uploaded data. Since the composition of contaminants within DNA extraction kits and other lab reagents are ubiquitous and can be distinct, our hypothesis is that contaminants from the same sources, such as DNA extraction kits or from a lab environment, will share similar characteristics in the composition of their contaminants. This fact should enable contaminants to be found in the form of shared species in samples taken from sufficiently distinct ecological niches, or in our case, body sites. In particular, this proposed approach is most relevant when the sequencing runs use the same DNA extraction kit and/or are processed in the same lab after reaching sufficient sequencing depth. In this work, we have implemented a de novo computational contamination detection tool, Squeegee, which is able to identify potential contaminants at the species level. Squeegee performs taxonomic classification and searches for shared organisms across multiple samples and sample types. The workflow of the pipeline is shown in Fig. 1. The software takes multiple samples containing sequencing data collected from distinct microbiomes as input and then uses taxonomic classification to search for candidate contaminant species that are shared across samples. By estimating pairwise similarity between metagenomic samples that the candidate contaminant species presents, and calculating breadth and depth of genome coverage by aligning the reads to the reference genome of the candidate contaminant species, Squeegee identifies taxonomic classification errors and makes accurate contaminant predictions at the species rank by filtering false calls from the candidates. We evaluated Squeegee on three datasets, including (i) a simulated dataset with ground truth contaminants, (ii) a real dataset with negative controls, and (iii) HMP samples without negative controls but with associated DNA extraction kit contaminants. Details on the implementation and evaluation of Squeegee can be found in the methods section. The dataset characteristics and parameters used in the study can be found in Supplementary table 1. In order to accurately identify contaminant sequences from external sources, such as lab environments or reagents used during the extraction or sequencing process, stable community members from different sample types must be considered. To assess whether there are ubiquitous genera across body sites comprising the human microbiome, we identified the stable community members across different human niches using Kraken classification results for HMP samples (Supplementary Table 2). By looking at each set of common community members of different body sites, we found no genera present in more than three of the six body sites (oral, nasal, skin, stool, throat, and vaginal). We evaluated Squeegee prediction accuracy at both genus and species rank on the maternal/infant datasets. During this benchmark, Squeegee performed contamination prediction without using the negative control samples, while Decontam took the classification results of the 10 negative control samples as input for contamination identification. A permissive ground truth contaminant set, and a strict version of the ground truth contaminant set, are generated with data from the negative control samples as well to use as a reference for the evaluation, whereas the strict set is generated with more stringent filtering to ensure high confidence. The details of the contaminant ground truth sets can be found in the methods section. Figure 2a shows the precision, recall, and F-score of Squeegee and Decontam at both species and genus rank using the permissive ground truth set. The unweighted precision, unweighted recall, and unweighted F-score for Squeegee are 0.714 (10/14 species), 0.323 (10/31 species), and 0.444 at species rank, and 0.833 (10/12 genera), 0.625 (10/16 genera), and 0.714 at genus rank, respectively. The false positive calls for Squeegee are Rothia mucilaginosa, Staphylococcus cohnii, Staphylococcus haemolyticus, and Streptococcus mitis. The unweighted precision, unweighted recall, and unweighted F-score for Decontam are 0.140, 0.774, and 0.238 at species rank, along with 0.174, 0.750, and 0.282 at genus rank, respectively. We also evaluated both methods with weighted scores, taking into account the abundance of information. Each of the species are first weighted by the mean fraction of reads assigned to those species in the non-negative samples. The weighted precision, weighted recall, and weighted F-score for Squeegee were 0.580, 0.728, and 0.645, respectively, and for Decontam were 0.928, 0.494, and 0.645, respectively, at species rank. The same measurements at genus rank were 0.438, 0.804, and 0.567 for Squeegee, respectively, and 0.947, 0.732, and 0.826 for Decontam, respectively. More importantly, we took a closer look at the predicted contaminants output by each method and evaluated the recall weighted by the relative abundance of the taxa in the negative control samples. Although Squeegee failed to identify several putative low-abundance contaminant species, the 10 correctly predicted species by Squeegee occupy over 0.763 of the cumulative relative abundance from the composition of the putative ground truth contaminants. With the same measurement, the species rank weighted recall under the same criteria for Decontam is 0.645. At genus rank, both methods performed well, with weighted recall for Squeegee scored at 0.892 and Decontam scored at 0.921. Although Decontam mislabeled some of the high abundance contaminants at species rank, it did label some of the closely related species under the same genera as contaminants, resulting in a significant increase of the score at genus rank. Figure 2b shows the accuracy of Squeegee and Decontam using the strict ground truth set. The detailed results can be found in Supplementary Section 1. Figure 3 shows the relative abundance of all contaminant species predicted by Squeegee in each of the non-control samples. The samples are clustered by sample types, designated by their color label on the y-axis. The predicted contaminant species that can be found in the permissive ground truth contaminant set are labeled in black at the top of the figure, and the predicted contaminant species not found in the permissive ground truth contaminant set are labeled in light gray. We evaluated Squeegee prediction accuracy on the HMP datasets as well. Figure 4 shows the precision, recall, and F-score of Squeegee predictions at genus rank. Squeegee has an unweighted precision of 0.667 (16/24 genera), an unweighted recall of 0.262 (16/61 genera), and an unweighted F-score of 0.376. While each taxa is weighted by their relative abundance from the non-control samples, Squeegee achieved a weighted precision of 0.856, a weighted recall of 0.958, and resulted in a weighted F-score of 0.904. Figure 4 also shows the relative abundance of true contaminant genera identified in the MoBio DNA extraction kit. The contaminants successfully predicted by Squeegee are colored orange with stripes, and the contaminants Squeegee failed to predict are colored gray. Low-abundance genera with relative abundance below 1% are combined in the figure. Although only 16 genera were correctly predicted, those genera accounted for the majority of the contaminated reads in the ground truth with a total relative abundance of 0.686. Since we use bacteria identified at the genera level as inherent putative contaminants in the MoBio DNA extraction kit level for our negative control reference, accuracy measurements at the species level do not apply. It is worth noting that more than 81.3% (61 out of 75) of species Squeegee predicted as contaminant species in the HMP datasets fell under the ground truth contaminant genera. Supplementary Fig. 1 shows the prevalence, the breadth of genome coverage, and additional score and filtering information of the top 50 predicted contaminant species after filtering. The first 16 rows show the prevalence of each species among each of the sample types, where zero prevalence is marked in blue. The following 16 rows show the breadth of genome coverage of each species in each sample type. The remaining rows show the prevalence score, the alignment score, the Mash score, and the combined score used to make the final prediction and whether each species passes the filters. The last row of the heat map shows whether the species can be found in the ground truth, with true positives shown in white and false positives shown in black. Detailed information on all candidate contaminant species can be found in Supplementary Fig. 2. To test the contamination limit of detection of Squeegee, we designed a set of simulated datasets based on the taxonomy profile of the real-world metagenomic samples. About 126 samples are simulated and divided into three groups based on the different relative abundance of the spike-in contaminant sequences (0.25, 0.50, and 1.00%). There are 42 simulated datasets in each of the groups, representing microbial communities from seven distinct environments. The details of how those simulated datasets are generated can be found in the methods section. Figure 5 shows the unweighted precision, recall, and F-score of different simulated sample groups at species rank and the same measurement weighted by the relative abundance of the taxa in the non-control samples. The figure also shows the detailed composition of the taxa, their relative abundance in the spike-in contaminant community, and the cumulative relative abundance of the correctly predicted contaminants at different relative abundances of spike-in. At all three different spike-in levels, where contaminant sequences occupied 0.25, 0.50, and 1.00% of the total reads, Squeegee had the perfect precision of 1.0. For unweighted recall, the 0.25% spike-in group scored 0.500, the 0.5% spike-in group scored 0.583, and the 1.0% spike-in group scored 0.750. As a result, the unweighted F-score for the 0.25% group, 0.50% group, and 1.00% group are 0.667, 0.737, and 0.857. When each species is weighted by their relative abundance in non-control samples, the 0.25% spike-in group scored 0.993, 0.5% spike-in group scored 0.990, and 1.0% spike-in group scored 0.989 for the weighted recall, and 0.25% spike-in group scored 0.997, 0.5% spike-in group scored 0.995, and 1.0% spike-in group scored 0.994 for the weighted F-score. When each species is weighted by its relative abundance in the negative control, the cumulative relative abundance of true positive prediction for the 0.25% spike-in group is 0.634. As the spike-in level increases, at 0.5% spike-in abundance, Squeegee scored 0.700, with one additional species, Salmonella enterica, identified as a contaminant. The cumulative relative abundance of true positive predictions continued to increase at a 1.0% spike-in abundance level, and Squeegee scored 0.844 with two more correct contaminant species predicted. In general, the unweighted recall and the cumulative relative abundance of the true positive predictions increase as the number of spike-in contaminant sequences increases since more contaminant sequences provide a stronger signal for Squeegee to pick up on and to make definite calls with respect to contamination. Figure 6a shows Shannon’s diversity index and Simpson’s diversity index for the maternal/infant dataset before and after contamination removal. Both diversity metrics for the samples were evaluated before the contaminant reads were removed (shown in red), after removing species confirmed by the permissive ground truth contaminants (shown in blue), and after removing all species predicted by Squeegee (shown in black). The max removal cutoff is set to 1%, which only removes species with a relative abundance of less than 1%. We observed significant decreases in Simpson’s diversity index in both placental and breast milk groups and significant decreases in Shannon’s diversity index in the placental group. There are also significant decreases in Shannon’s diversity index in the breast milk group if we remove all predicted contaminant species, but no significant decreases are found by only removing contaminant species confirmed by the negative control experiments. For a more strict max removal cutoff of 0.5%, we still found significant decreases in both Shannon’s and Simpson’s diversity index in the placental group (See Supplementary Fig. 3). Figure 6b shows the same alpha diversity analyses performed on the HMP samples with the maximum removal cutoff set to 1%. We observed significant decreases in Shannon’s diversity index values in oral and nasal samples, and a significant decrease in Simpson’s diversity index in oral samples. With the max removal cutoff of 0.5%, there are significant decreases in Shannon’s diversity index and Simpson’s diversity index in the oral samples (See Supplementary Fig. 4). We applied Squeegee to a human-derived RNA-Seq dataset from an index study that aimed to evaluate the potential for contamination arising during sequencing across 6 different sequencing centers in Europe in the GEUVADIS consortium. The resultant generated RNA sequencing data arose from 40 sequencing runs performed at all six sequencing centers on identical sequencing platforms following extraction with the same kits on parallel samples. The prediction from Squeegee indicates that seven species (Human gammaherpesvirus 4, Proteus virus Isfahan, Escherichia coli, Bacillus megaterium, Bacillus cereus, Klebsiella pneumoniae, and Cutibacterium acnes) are reagent specific contaminants and can be found in the sequencing runs across different sequencing centers. (Supplementary Fig. 5). While Human gammaherpesvirus 4 is associated with the cell line used during the sequencing, Escherichia coli, Bacillus megaterium, Bacillus cereus, Klebsiella pneumoniae, Cutibacterium acnes can be putative common “kit contaminants” that have been previously reported. However, E. coli and K. pneumoniae are also prevalent environmental and human commensal microbes or pathobionts. We tested the performance of Squeegee with datasets of different sizes. Table 1 shows the run time and peak memory usage of Squeegee for different sizes of input. The run time of Squeegee (in CPU hours) is primarily determined by the size of the input data and is also affected by the number of potential contaminants identified based on the taxonomic classification results. Since the reads are mapped to reference genomes of each potential contaminant using Bowtie2 (with multi-alignments enabled), more contaminants would increase the CPU time for alignment and the coverage calculation process. The peak memory usage of Squeegee is mainly driven by the size of the database used by Kraken for taxonomic classification. The following runs use the same Kraken database (302 GB) built with NCBI RefSeq (Release 202); thus they have similar peak memory usage. To the best of our knowledge, Squeegee is the first de novo computational tool specifically designed to identify and nominate taxa as potential contaminants in the absence of “kit negative”, environmental contaminant controls, and other auxiliary data. Squeegee is able to mark these taxa contained within metagenomic samples without requiring negative experimental controls, and can identify potentially widespread contaminants in publicly available data. In order to predict contaminant species, multiple pieces of evidence are taken into consideration, including the prevalence rate of species, the metagenomic distance of the samples that contain the species, and how well the genomes of those species are being covered. A recent study has shown that the accuracy of the taxonomic classification algorithm has become a limiting factor of contamination detection due to high levels of sequence similarity at species rank. With a breadth of genome coverage for each contaminant species being calculated, Squeegee also attempts to address taxonomic classification error that might occur during the taxonomic binning process, which is a common issue for k-mer-based methods. Comparisons between Squeegee predictions and experimental control data show that Squeegee is capable of accurately inferring contamination at the species level, especially in regard to contaminants occurring at a relative abundance of 5% or higher. For the maternal/infant dataset, a strict contaminant ground truth and a permissive contaminant ground truth were constructed with the taxonomic assignment of the sequencing data from the negative control experiments with the use of prevalence, relative abundance, and absolute read count filtering, a common practice to minimize the taxonomic assignment error and determine the presence or absence of species, with different filtering parameters. Squeegee predicted most of the putative contaminants found in the strict ground truth contaminant set (see Fig. 2b), including species from contaminating genera (e.g., Methylobacterium, Pseudomonas, and Xanthomonas) that have been previously reported. For the presumptive false negative contaminant species the Squeegee failed to predict, all were of relative abundances below 5% except for Staphylococcus capitis. In addition to other genera and species unique to the low-biomass maternal/infant samples, we also found that Squeegee predicted a number of contaminant species from the genera Staphylococcus, including Staphylococcus haemolyticus and Staphylococcus cohnii, that are not found in the experimental control samples. Staphylococcus species are often found in the normal flora of the skin and have been reported multiple times as contaminants from DNA extraction kits and laboratory environments. Staphylococcus species are also well-known for their highly similar genomes, which creates a big challenge for the taxonomic assignment task. Additionally, Squeegee identified Rothia mucilaginosa, which is a part of the normal oropharyngeal flora, and Escherichia coli as contaminants. Both species may represent bonafide species shared across body niches. It is also possible that the experimental control samples were not sequenced deeply enough to reveal these species or the species were at a low enough relative abundance in the experimental control samples that were filtered out during quality control. We benchmarked Squeegee against a “gold-standard” contamination detection approach in Decontam, with its prevalence-based method requiring negative control samples as input. We note that we view both tools as complementary, especially since using negative controls is recommended best practice for contamination removal. From the results (Fig. 2b), we see that Squeegee is able to achieve performance that meets or exceeds Decontam predictions at species rank using the strict ground truth, with respect to unweighted F-score, weighted F-score by the relative abundance in the non-control samples, and cumulative relative abundance of the putative correctly identified contaminants from the negative control experiment samples. On the other hand, at the genus rank, Squeegee is unable to match Decontam performance when F-score is weighted by the relative abundance in the non-control samples while performing on par with Decontam with respect to the cumulative relative abundance of the correctly identified contaminant genera in negative controls. Although Decontam failed to recognize Pseudomonas tolaasii and Xanthomonas euvesicatoria as putative contaminant species, multiple species under the same genera were successfully identified, increasing its genus rank score. While evaluating using the permissive ground truth contaminant set, Squeegee performed equally well at species rank with respect to weighted F-score (Fig. 2a), with a drop in unweighted recall given Squeegee failed to recognize most contaminant species with mean relative abundance less than 1% in the negative controls. As expected, Decontam with the experimental negative control data performs best in terms of unweighted recall (see Fig. 2a). Decontam identified 24 out of 31 species within the permissive contaminant ground truth set, only missing Pseudomonas tolaasii, Xanthomonas euvesicatoria, Cupriavidus oxalaticus, Staphylococcus aureus, Pasteurella multocida, Klebsiella pneumoniae, and Escherichia coli. It is possible that Decontam did not flag some species as contaminants that are shared between the source of contamination and the sampling environment, such as Staphylococcus aureus and Escherichia coli. At the same time, Squeegee identified all those seven species, which the relative abundance in the permissive ground truth adds up to 35.5%, as contaminants, which completes the entire contaminant ground truth set if we take the union of the predictions made by the two methods. Alternatively, we acknowledge that this may over-call “contamination” by virtue of shared species among body niches. This once again highlights the complementarity of Decontam with negative controls and Squeegee, and also the value of Squeegee either when negative controls are unavailable (existing metagenomic sequence datasets) or for lab contamination that affects both the negative control and samples. In addition to identifying microbial contaminants within microbiome datasets lacking negative controls, Squeegee can also identify contaminants in human RNA-seq data. In our experiments with the GEUVADIS consortium human RNA-Seq dataset, Squeegee predicted that seven species were lab preparation related. In this dataset, since non-human reads are all classified as contamination, by identifying reagent-specific contaminants shared across different sequencing labs, one can backtrack and identify lab-specific contaminants using the classification report provided by Squeegee. Another use case for Squeegee is to detect batch-specific contaminants, as well as cross-contaminants. Suppose negative control samples are not prepared and sequenced for every batch run. In this case, batch-specific contaminants may get mixed into the samples, causing bias in the downstream analysis. A similar scenario is that cross-contamination occurs in a single batch but happens not to affect the negative control sample. Running Squeegee on each individual batch allows the user to detect such batch-specific contaminants or cross-contaminants since the Squeegee detection method does not depend on the negative control profiles. Squeegee is designed for de novo identification of microbial species that are likely contaminants; a higher combined contaminant score indicates the species has a higher potential for being an actual contaminant. However, Squeegee’s failure to flag a microbial species in a sample as a likely contaminant does not mean it is not a contaminant. As mentioned before, one of the limiting factors is the relative abundance of the species within the source of the contamination. Figures 2 and 4 show that contaminant species with low relative abundances in the control samples are more challenging to identify since the sequencing signals of such species become even weaker in the non-control metagenomic samples. Squeegee failed to predict some of the low-abundance genera/species in the simulation dataset due to similar reasons (see Fig. 5). In order to challenge Squeegee, the simulation dataset we designed contains very low proportions (0.25–1%) of spike-in contaminant sequences. Among the 12 spike-in contaminant species, all except Ralstonia pickettii have relative abundance below 0.1 within the spike-in. As the relative abundance of the total spike-in sequences increases, we observed that the unweighted recall increased as well, and Squeegee is able to pick up more and more contaminant species. As shown in our experimental results (both simulated and real), Squeegee can exhibit low recall on low-abundance contaminant species, which means there will be residual reads not able to be characterized by Squeegee (e.g., they could either represent microbial contaminants or bonafide metagenomic signal). In order to detect low-abundance contaminants, using Decontam with negative control samples is recommended. Squeegee tracks contaminants that came from the same source, such as DNA extraction kits or laboratory surfaces, and in order for Squeegee to perform well, the input samples should be collected from different ecological communities, for example, the microbiota of well-distinct body sites (skin, gut, and oral). One of the other limitations of Squeegee is that it cannot trace contaminants originating from the sample collection process since different sample collection operations may introduce different contaminant species. Therefore, further investigation is required to validate whether the species truly originated from the sampled metagenome for species that are not included in the predicted contaminants. A stable community member of a specific body site has the potential also to be a contaminant taxon from an external source. Since Squeegee operates without prior knowledge of the input dataset, ubiquitous species that are commonly found in a wide range of environments could allow Squeegee to make false predictions. Although the Staphylococcus genus has been reported as external contamination from multiple studies, it is hard to ignore that some of the Staphylococcus species may be truly present among multiple body sites, including skin and nasal samples. Such ubiquitous species may introduce noise in Squeegee’s predictions. Combined with the prior knowledge of the input dataset and the comprehensive information that Squeegee outputs, the user may further filter the predicted list of contaminants if needed. For any individual sample type, the user should treat the predicted result with care to avoid potential community members being falsely labeled as contaminants. We looked closer into the low weighted precision of non-control sample abundance at species and genus rank for Squeegee in the maternal/infant dataset. Streptococcus mitis, a common member of the microbial communities from the oral, skin, female genital tract, and gastrointestinal tract, was incorrectly identified as a contaminant. Given the high relative abundance of Streptococcus mitis (Fig. 3), this false positive contaminant species received an abnormally high weight compared to the other true positive contaminant species, lowering Squeegee’s weighted precision. At genus rank, the relative abundance difference between Streptococcus and the true positive putative contaminant genera becomes even greater, which explains why Squeegee genus rank performance is lowered compared to species rank in this dataset. These results highlight areas for future improvement to Squeegee that would allow it to take into account microbial species ubiquitous in many different environments. By no means is Squeegee meant to be a replacement for experimental negative controls. It does not estimate the relative abundance of each predicted potential contaminant since the relative abundance of the contaminants varies in different sample types. Squeegee makes predictions based on the assumption that the input data are sampled from multiple distinct microbiomes, and does not apply to cases where the sequencing data are from similar microbiomes. If possible, performing negative control experiments will likely provide a more accurate profile of the external contaminants. However, as discussed, it is common for experimental negative control samples to be unavailable for publicly available metagenomic datasets. The metagenomic datasets from the Human Microbiome Project is one such high-profile example. When compared to other contamination removal methods, Squeegee is the only existing tool able to predict contamination from multiple sources without experimental negative control samples (see Table 2), and its contaminant predictions can have a significant impact on diversity measures which are often a key part of the results of a vast range of microbial studies. Another possible solution for contamination detection without negative control sequencing data is to use a contaminant database. If a database of genomes containing known contaminant species exists, we could identify the contaminant sequences in the data by mapping reads against this database. Building such a contaminant database can be challenging because it requires sequencing data from all possible sources of contamination. Since Squeegee is a negative control-free tool for identifying novel contaminants, it can also be used as an important step in filling out such a comprehensive database of likely putative contaminants. Over 81% of the contaminant species predicted by Squeegee for the HMP dataset match the bacterial genus described as inherent contaminants of the MoBio DNA extraction kit, which was used for the Human Microbiome Project. The cumulative relative abundance of correct prediction is 68.6%, while Squeegee failed to predict most of the genera from phylum Proteobacteria. This may be due to the fact that the kit used in the Mobio contamination study is closely related to the one used for HMP but not identical. The contamination profile of the same kit might change over time, and samples processed in different labs may also affect the results since contaminants from lab surfaces and lab members can potentially contribute to the composition of the contamination. Finally, though Squeegee was tested and evaluated with metagenomic shotgun sequencing datasets, it could be extended for use on 16S rRNA sequencing data. However, Squeegee wouldn’t be able to use the breadth and depth of genome coverage of the alignment to determine classification errors. Therefore, choosing an accurate taxonomic classifier is critical for running Squeegee on 16S rRNA sequencing data. In summary, as far as we are aware, Squeegee is the first de novo computational method for identifying potential microbial contaminants in microbiome datasets in the absence of environmental negative control samples and auxiliary information such as DNA concentration information. Squeegee predictions on multiple datasets have shown that contaminant sequences from the same source, such as DNA extraction kits and other reagents used during the sample processing and sequencing, can be accurately identified across multiple samples using this computational method without experimental negative controls or DNA quantitation data. Squeegee achieves high weighted recall (weighted by both relative abundance of taxa in negative control and non-control samples) and low false positive rates on real metagenomic datasets, and can help to identify putative contaminant sequences of suspicious taxa for low-biomass microbiome studies, enabling sample-independent and orthogonal approaches aimed at distinguishing true microbiome signals from environmental contamination. In order to generate reproducible estimates of contaminants and their composition among the samples, the user must collect sequencing data from multiple metagenomic samples. The microbial community composition should be largely distinct between any two samples included in the analyses. Here, distinct refers to different metagenomic environments or sample types in which it is rare to observe a given microbial species present across most samples. Each sample should be provided with a tag or descriptor that distinguishes the different types of samples (e.g., oral, vaginal, fecal, soil, ocean, etc.). Squeegee first performs taxonomic classification using Kraken v1.1.1 with default settings (k = 31). The reference database for Kraken was built with complete bacterial/archaeal/viral genomes from NCBI RefSeq (Release 202). A classification report is generated for each of the samples. Based on the classification, Squeegee chooses a set of candidate contaminant species based on the prevalence of the species across the samples. The prevalence score is weighted by the number of samples of the same type to avoid bias introduced by an unbalanced number of samples between sample types. Higher prevalence rates of a species indicate that the species is shared by more samples across more sample types, and it is more likely to be a contaminant. Squeegee also calculates the metagenomic similarity between the samples using Mash v2.2.2, a tool that estimates the Jaccard index using MinHash. This is done by first generating a sketch of each sample (Mash sketch -s 100000 -k 21 -m 2) and then calculating the pairwise Mash distance between all pairs of samples (Mash dist). High Mash distances indicate that the metagenomes of the two samples are more distinct (i.e., there are fewer genera and species shared between the samples). Squeegee weights shared species coming from more distinct samples as more likely to be a contaminant. Squeegee then fetches the representative genomes for each of the candidate contaminant species from the NCBI RefSeq database used to build the Kraken database. These representative genomes are used as references to perform a multi-alignment for all reads in the samples using Bowtie2 v2.3.5 with the multi-alignments enabled (bowtie2 –local -a –maxins 600). To accelerate this process, k-mer-mask from meryl v1.0 is used to filter out reads that do not contain any 28-mers from the reference genomes (k-mer-mask -ms 28 -clean 0.0 -match 0.01 -nomasking). Based on the alignment results, the breadth and depth of genome coverage is calculated for each of the sample type using samtools v1.11 (samtools depth). The breadth and depth of genome coverage are used to determine whether the species is truly present or is a potential misclassification from the taxonomic classifier. A species that is truly present should have a large proportion of its genome covered. On the other hand, a large number of reads covering only a small proportion of the genome often suggests that the species was a misclassification. Since contaminant species are often low in abundance, combining samples from the same type would give us a better indication of the presence of the species. In the last step, Squeegee combines multiple pieces of evidence, including the prevalence score, Mash distance score, and alignment score, and makes a final prediction for contaminant species using the following equation, In Eq. (1) above, Pi is defined as the prevalence score of candidate contaminant species i, which is calculated as the weighted mean prevalence rate of species i among all sample types. Mi is the Mash distance score of candidate contaminant species i. Squeegee takes the Mash distance values (from 0 to 1) of all sample pairs that both contain species i, and calculates Mi by averaging the top 10% of the pairwise Mash distance value. Ai is the alignment score of candidate contaminant species i, which is defined as the mean breadth of genome coverage of species i across sample types with a minimum depth of coverage of 3. min_cov is the minimum coverage threshold defined by the user. While calculating the combined contaminant score of each taxon, both the prevalence score and Mash score are normalized by the mean of those scores for all candidate contaminant species ( and ). The alignment score is capped at 1 for those taxa which have a mean breadth of genome coverage exceeding 5× of the minimum breadth of coverage threshold (min_cov) since the breadth of genome coverage can vary greatly between species. For example, if the minimum breadth of coverage is set to 5%, taxa with a mean breadth of coverage exceeding 25% will receive an alignment score of 1, and taxa with a mean breadth of coverage of 5% will receive an alignment score of 0.2, and taxa with a mean breadth of coverage less than 5% will be classified as false calls and be eliminated by the filter. Details about how each score is calculated can be found in Supplementary Table 3. Such a scoring mechanism allows Squeegee to automatically distribute different weights based on which evidence contributes more to distinction for the candidate contaminants. For example, if all candidate contaminants have similar Mash scores and similar breadth of genome coverage, but distinguishable prevalence across different sample types, then the algorithms for the contaminant predictor would be automatically favoring prevalence over other factors. After the combined contaminant scores are calculated, Squeegee filters out species below a user-defined minimum combined score threshold. The combined score averages all three normalized scores for each piece of contaminant evidence. Candidate contaminants with a low combined score suggest insufficient evidence supporting the argument that the candidate species is both an actual contaminant and present in the samples. Squeegee also provides a comprehensive output for the user if the further downstream analysis is required. The parameter settings retain the potential to affect the precision and recall of Squeegee. Based on the basic understanding of the samples, the user is able to control how likely a taxon is to be recruited as a candidate contaminant by setting a minimum prevalence threshold (Default:0.6) to different values. If the users are processing samples that have similar microbiome communities, increasing the minimum prevalence threshold will reduce the number of false positives caused by shared true community members. Lowering the minimum prevalence threshold allows the program to consider more candidate contaminants, potentially increasing recall but will increase the run time. The minimum read support threshold, minimum abundance threshold, and minimum alignment coverage threshold all contribute to how restrictive a taxon is considered present. Based on different sequencing technologies, more than 5% of the reads may be misclassified by the taxonomic classifier even at the genus level. Increasing those thresholds allows more confident identification of whether a taxon is truly present or not. On the other hand, in a scenario where contaminant species are low abundant, setting those parameters at high values could cause an increase in false negatives. Evaluation of Squeegee predictions was performed by comparing the predicted contaminant species using three datasets: (1) a simulated dataset with ground truth contaminant species, (2) a real dataset with available negative control samples, and (3) a real dataset without a negative control (HMP samples) but with associated kit contaminants. The simulated dataset contains a total of 126 simulated samples representing seven distinct microbial communities. The real dataset contains 344 samples over 9 distinct sample types collected of adult females and infants, as well as sequencing data from ten negative control experiment samples. The HMP dataset includes 749 samples collected and sequenced from healthy individuals across 16 different body sites. The parameters and data characteristics are shown in Supplementary table 1. For (1), the simulated dataset, the contaminant species in the ground truth were generated based on the species of a simulated spike-in of contaminant sequences. In order to simulate a realistic dataset and test the detection limit of Squeegee, 42 real-world metagenomic samples were chosen from seven distinct environments, including six soil samples of mining sites, one soil sample collected from the wetland, six freshwater samples, seven hot spring samples, six skin samples of cows, ten healthy human skin samples, and six healthy human gut samples.. We filtered out the species with relative abundance lower than 0.0005 or with supported read count less than 300 in those samples, and used the remaining species and their relative abundance as a reference to simulate the dataset. We then used the species with relative abundance greater than 0.01 found in the FastDNA SPIN Kit for Soil (MP Biomedicals) from the previous study to simulate contaminant sequences. Each distinct sample was simulated three times with spiked-in contaminant sequences that occupy 0.25, 0.5, and 1% of the total sequences in the sample. A total of 126 simulated samples were generated using CAMISIM and ART, simulating Illumina paired-end reads with an average read length of 150 bp and an average read pair count of 6664348. The simulated samples are grouped into three groups by the spike-in contaminant level, and we evaluate Squeegee on each of the groups individually. For (2), maternal/infant metagenomic datasets, the contaminant species in the ground truth were generated based on the classification of multiple experimental negative controls. To minimize classification errors, we applied a set of criterion to include a species in the contamination ground truth. Species with relative abundance above 0.5% and more than 20 reads assigned in at least half of the negative control samples and species with relative abundance above 10% in a single sample were chosen for inclusion in the ground truth contaminant set, resulting in a strict ground truth set with 15 species from 13 genera. We also applied more permissive filtering and generated a permissive ground truth contaminant set that contains 31 species that belong to 16 genera by lowering the minimum relative abundance threshold to 0.2%. We then aligned the sequencing reads in the experimental control samples to the representative genomes. Reads assigned to the Staphylococcus virus Andhra stacked in a small 449 bp region with an average depth of 1429, indicating a false classification call, so we removed it from both strict and permissive ground truth contaminant sets. Once the ground truth contaminants were identified, the relative abundance of the ground truth contaminants was calculated as the average relative abundance across all negative control samples over the sum of the average relative abundance of each contaminant. For (3) the HMP dataset, which was extracted using the MoBio DNA extraction kit, we used the 61 bacteria genus (excluding lot-dependent organisms), which were identified as inherent contaminants within a latter version of a related MoBio extraction kit, the MoBio PowerMax® Soil DNA Isolation Kit 12,988-10 (MoBio Laboratories, USA), in a recent study as the ground truth contaminants. Relative abundances of each genus were also obtained in the same study. Since Squeegee makes contamination predictions at the species level, predicted contaminant species from the reference genus are counted as true positives. During the evaluation, Cutibacterium acnes (formerly Proprionibacterium acnes) was assigned to the genus Proprionibacterium to keep the ground truth and Squeegee prediction consistent. Performance is evaluated via precision, recall, and F-score. The unweighted precision is calculated as the ratio between the number of predicted contaminants found in the ground truth and the total number of predicted contaminants. The unweighted recall is calculated as the ratio between the number of correctly predicted contaminants and the total number of contaminants in the ground truth. The unweighted F-score is calculated as 2 × (unweighted precision × unweighted recall)/(unweighted precision + unweighted recall). Those three measurements are also calculated using the mean fraction of the reads of each taxon in the non-control samples as weight. The weighted precision is calculated as the average fraction of reads from the non-control samples in the true positive (TP) taxa over those in the true positive and false positive (TP + FP) taxa. The weighted recall is calculated as the average fraction of reads from the non-control samples in the true positive (TP) taxa over those in the true positive and false negative (TP + FN) taxa. The weighted F-score is calculated as 2 × (weighted precision × weighted recall)/(weighted precision + weighted recall). We also evaluated the methods using the cumulative relative abundance of true positive taxa, which is a weighted recall score that is weighted by the mean relative abundance of the taxa in the negative control samples. To further demonstrate the contamination detection capabilities of Squeegee, we leveraged a human-derived RNA-seq dataset from a study performed by the Genetic European Variation in Health and Disease (GEUVADIS) consortium. The dataset contains parallel RNA-Seq samples from Epstein-Barr virus (EBV)-positive lymphoblastoid cell lines that are sequenced across seven different sequencing centers with identical library preparation kits. We used samples from six out of seven sequencing centers that used the Illumina sequencing platform (Illumina Genome Analyzer II) and excluded the ones that used AB SOLiD System 3.0, leaving us with a total number of 40 paired-end sequencing runs. We then mapped each of the sequencing runs with bowtie2 against the human reference genome (Homo sapiens GRCh38.p13) with the parameter –maxins 600 to remove human reads. We gathered the unmapped reads for each of the samples and used them as input for Squeegee. The sample type for each of the samples is labeled by the sequencing center where it was run. The parameter settings for Squeegee and data characteristics are shown in Supplementary Table 1. In order to benchmark the performance of Squeegee, we also ran Decontam v1.10.0 on the maternal/infant metagenomic datasets with the negative control samples. After taxonomic classification with Kraken, all species with at least 30 read support or relative abundance greater or equal to 0.0005 were collected to construct the abundance input table for Decontam. Contamination detection was done using isContaminant function with the prevalence method from the Decontam R package with the default parameters. We categorized the labeled sample types of the maternal/infant dataset and HMP dataset into combined sample types based on the body site. The combined sample types for the maternal/infant dataset include placenta, breast milk, oral, stool, and vaginal. The combined sample types for HMP include vaginal, throat, stool, oral, skin, and nasal samples. Samples from the same combined sample types in each dataset were used for alpha diversity analysis. Both Shannon’s diversity index and Simpson’s diversity index were measured before and after contamination removal. Only reads assigned to the species rank by Kraken were used in calculating Shannon’s diversity index and Simpson’s diversity index. Since contamination originating from external sources can also be actual community members of the metagenomes, we set a max removal cutoff and only remove species with relative abundance below this cutoff. The significance test was done using a two-sided Mann–Whitney U-test for all combined sample types with more than 20 samples. We used the samples from the HMP dataset and their combined sample types to generate a set of stable community members for different human body sites. Stable community members were defined as genera with more than 1% of their reads assigned from Kraken classification in more than 50% of the samples from the same combined sample types. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Reporting Summary Peer Review File

PMC9649626

Jun Wu,Yingtao Xu,Zikai Geng,Jianqing Zhou,Qingping Xiong,Zhimeng Xu,Hailun Li,Yun Han

Chitosan oligosaccharide alleviates renal fibrosis through reducing oxidative stress damage and regulating TGF-β1/Smads pathway

10-11-2022

Renal fibrosis,Drug development,Translational research

Renal fibrosis (RF) is the common pathway for a variety of chronic kidney diseases that progress to end-stage renal disease. Chitosan oligosaccharide (COS) has been identified as possessing many health functions. However, it is not clear whether COS can prevent RF. The purpose of this paper was to explore the action and mechanism of COS in alleviating RF. First, an acute unilateral ureteral obstruction operation (UUO) in male BALB/c mice was performed to induce RF, and COS or fosinopril (positive control drug) were administered for 7 consecutive days. Data from our experiments indicated that COS treatment can significantly alleviate kidney injury and decrease the levels of blood urea nitrogen (BUN) and serum creatinine (SCr) in the UUO mouse model. More importantly, our results show that COS can reduce collagen deposition and decrease the expression of fibrosis proteins, such as collagen IV, fibronectin, collagen I, α-smooth muscle actin (α-SMA) and E-cadherin, ameliorating experimental renal fibrosis in vivo. In addition, we also found that COS suppressed oxidative stress and inflammation in RF model mice. Further studies indicated that the mechanism by which COS alleviates renal fibrosis is closely related to the regulation of the TGF-β1/Smad pathway. COS has a therapeutic effect on ameliorating renal fibrosis similar to that of the positive control drug fosinopril. Taken together, COS can alleviate renal fibrosis induced by UUO by reducing oxidative stress damage and regulating the TGF-β1/Smad pathway.

Chitosan oligosaccharide alleviates renal fibrosis through reducing oxidative stress damage and regulating TGF-β1/Smads pathway Renal fibrosis (RF) is the common pathway for a variety of chronic kidney diseases that progress to end-stage renal disease. Chitosan oligosaccharide (COS) has been identified as possessing many health functions. However, it is not clear whether COS can prevent RF. The purpose of this paper was to explore the action and mechanism of COS in alleviating RF. First, an acute unilateral ureteral obstruction operation (UUO) in male BALB/c mice was performed to induce RF, and COS or fosinopril (positive control drug) were administered for 7 consecutive days. Data from our experiments indicated that COS treatment can significantly alleviate kidney injury and decrease the levels of blood urea nitrogen (BUN) and serum creatinine (SCr) in the UUO mouse model. More importantly, our results show that COS can reduce collagen deposition and decrease the expression of fibrosis proteins, such as collagen IV, fibronectin, collagen I, α-smooth muscle actin (α-SMA) and E-cadherin, ameliorating experimental renal fibrosis in vivo. In addition, we also found that COS suppressed oxidative stress and inflammation in RF model mice. Further studies indicated that the mechanism by which COS alleviates renal fibrosis is closely related to the regulation of the TGF-β1/Smad pathway. COS has a therapeutic effect on ameliorating renal fibrosis similar to that of the positive control drug fosinopril. Taken together, COS can alleviate renal fibrosis induced by UUO by reducing oxidative stress damage and regulating the TGF-β1/Smad pathway. Chronic kidney disease (CKD) is currently considered as a major public health problem. CKD prevalence has been determined to be 8–16% worldwide. Renal fibrosis (RF) is a clinical pathological syndrome characterized by abnormal extracellular matrix components (ECMs) stored in the kidney with many pathological causes. RF is the unifying feature of progressive renal alterations and the common pathway for a variety of CKDs progressing to end-stage renal disease (ESRD). A growing body of research indicates that RF contributes to the progressive and irreversible decline in renal function and is associated with high morbidity and mortality induced by CKD. Therefore, it is urgent to develop an effective drug or method to prevent and treat RF, so as to slow down the ESRD process induced by CKD as much as possible. The renin–angiotensin system (RAS) is actively involved in the development of renal fibrosis, which can trigger the excessive deposition of collagen and change vascular resistance to aggravate RF. This provides an important and rational basis for using angiotensin converting enzyme inhibitors (ACE-Is) to alleviate RF. Fosinopril, an ACE-I, which has been well tested in the treatment of renal injury. Previous investigations have confirmed that fosinopril can prevent inflammation and oxidative stress to inhibit fibrinoid necrosis, focal and segmental hyperplasia and interstitial infiltration in the kidney. Considering the significant effect of fosinopril on RF, fosinopril was used as a positive control drug in most of the efficacy evaluation experiments on RF. Many studies have shown that inflammation and oxidative stress injury have been widely accepted to play an important role in the occurrence and development of RF. Furthermore, a number of studies have also demonstrated that TGF-β/Smads signal transduction is a key pathway in progressive RF. Thus, anti-inflammatory, antioxidative stress injury and TGF-β/Smads pathway regulation have been considered as important strategies for the treatment of RF. Recently, it was discovered that numerous natural polysaccharides with good antioxidant and anti-inflammatory properties and TGF-β/Smads pathway regulation have been discovered to possess potent against RF activities, which have attracted considerable interest as potential candidates for the development of novel RF therapies. Chitosan oligosaccharide (COS), a natural oligomer polysaccharide from the depolymerized product of chitosan, consists of β-1,4-linked d-glucosamine units (Fig. 1A). Studies have shown that COS has higher water solubility, lower viscosity, and a higher rate of absorption through the intestinal epithelia than those of chitosan because it has a degree of polymerization below than 50–55 and an average molecular weight (Mw) of less than 10 kDa. These characteristics of COS make it a preferable candidate for pharmaceutical applications. Even more encouragingly, COS has been proven to possess a variety of beneficial biological effects, including anti-inflammatory, antioxidative, antidiabetic, antihypertensive, and lipid-lowering activities. Furthermore, there is also evidence showing that COS can regulate the TGF-β/Smads pathway to prevent and treat fibrous disease. In view of the previous findings on the biological activities of COS, it is reasonable to speculate that COS may possess a significant on anti-RF effect. However, there are still few reports, and there is a lack of relevant experimental data supporting these hypotheses. Therefore, in the present paper, we will systematically study the amelioration of RF in an acute unilateral ureteral obstruction mouse model by COS supplementation at the present paper, and further reveal its underlying mechanism. In the course of the experiment, it was found that the mice from the normal group subjected to the sham operation still had glossy fur (Fig. 1B), a good mental state, a more agile reaction, flexible movement, no obvious reduction in food and water intake (Fig. 1C,D), and a gradual increase in body weight (Fig. 1E). Compared with the normal group, the mice in the model group showed a series of pathological symptoms as the experiment progressed including a gradually worsening mental state, dull reactions, dark yellow and dull hair (Fig. 1B), significantly reduced movement, prominently decreased food and water intake (Fig. 1C,D), and markedly slower body weight gain (Fig. 1E). However, these RF symptoms in the model group were significantly improved after fosinopril (except for body weight) and COS intervention (Fig. 1). More importantly, these symptoms, especially body weight and food and water intake, showed a significant dose–response relationship in the COS intervention group. The results fully implied that COS could alleviate the clinical symptoms in UUO-induced RF. As shown in Fig. 2, kidney morphology was vastly altered by UUO. Compared with the normal group, the obstructed kidneys of mice in the model group presented with abnormal swelling and a markedly increased volume of renal parenchyma (Fig. 2A). The kidney weight of the obstructed side and the weight ratio of the obstructed/contralateral unobstructed kidney in the UUO model group were also significantly increased compared with those of the normal group (P < 0.01) (Fig. 2B,C). However, COS could observably prevent the pathological morphologic alterations induced by UUO operation (Fig. 2A). After COS or fosinopril treatment, obviously reduced trends in the kidney weight of the obstructed side and the weight ratio of the obstructed/contralateral unobstructed kidney were observed (P < 0.05 or P < 0.01) compared with those of the model group (Fig. 2B,C). More interestingly, COS showed a dose-dependent effect on the reversal of UUO-induced renal pathological changes. The data suggested that COS could mitigate the swelling and weight of the obstructed kidney in UUO-induced RF mice. As shown in Fig. 3, the contents of SCr and BUN in mice of the model group were significantly increased (P < 0.01) relative to those of the normal group, indicating that the normal renal function was destroyed 7 days after UUO. However, the levels of both BUN (Fig. 3A) and SCr (Fig. 3B) from the COS or fosinopril intervention group mice were significantly reduced (P < 0.01 or P < 0.05) compared to those of the model group. The results demonstrated that the renal dysfunction induced by UUO could be significantly alleviated by the treatment with COS or fosinopril. Furthermore, it is encouraging to see the dose-dependent improvement effect of COS on UUO-induced renal dysfunction. The data implied that COS can improve renal dysfunction in mice with UUO-induced RF. Haematoxylin and eosin (H&E) staining showed that the obstructed kidneys of mice from the model group possessed extensively dilated tubules, a large amount of tubular epithelial cell apoptosis with shedding into the lumen, and inflammatory cell infiltration (Fig. 4A). The tubulointerstitial injury indices were noticeably higher than those of the normal group (Fig. 4B). The results suggested that severe structural damage and tubulointerstitial injury were caused by UUO. At the same time, Masson trichrome staining further revealed that a large number of collagen fibers were already present in the obstructed kidney tissue of mice in the model group (Fig. 4A,C). In contrast, UUO-induced kidney obstruction was significantly reduced after the combined application of COS intervention (Fig. 4A,B), and the production of collagen fibers was notably decreased (Fig. 4A,C). Moreover, the improvement effect of COS has an obvious dose–response relationship. The key indicators for kidney injury were also summarized in Table 1. The results fully confirmed that COS can attenuate tubulointerstitial injury and fibrosis in UUO-induced RF mice. We further investigated the effect of COS on ECM in the kidney tissue of UUO-induced RF mice by examining collagen IV, fibronectin and collagen I as the representative indexes. The RT-PCR and immunohistochemical staining results showed that the expression levels of collagen IV, fibronectin and collagen I were significantly increased in the obstructed kidney tissue of mice from model group compared with that of the normal group (Fig. 5A–C). It is surprising that the excessive and abnormal collagen IV, fibronectin and collagen I secretion in obstructed kidney tissue of mice from the model group was significantly inhibited by COS and fosinopril intervention (Fig. 5). Based on these mRNA expression results (Fig. 5A) and the in situ protein expression (Fig. 5B,C), we also clearly observed that the inhibitory effect from the high-dose COS group on collagen IV, fibronectin and collagen I was obviously better than that of the low-dose of COS group. These results suggested that COS can inhibit ECM accumulation to alleviate RF. Elevated α-smooth muscle actin (α-SMA) and depressed E-cadherin are used as the main markers of EMT initiation. Furthermore, the UUO-injured kidneys in mice showed massive inflammatory cell infiltration and proinflammatory cytokine secretion. As one of the important markers of macrophages, CD68 expression was significantly increased in the kidneys of UUO-induced renal fibrosis model mice. As shown in Fig. 6, α-SMA expression in obstructed kidney tissue of mice from the model group was significantly elevated when compared with that of the normal group. Fortunately, the COS intervention exhibited a significant inhibitory effect on abnormal α-SMA expression as well as an elevated effect on E-cadherin expression (Fig. 6A–C). Meanwhile, we also observed that the CD68 expression and proinflammatory factor secretion in the kidneys of the model group were significantly increased compared with those of the normal group (Fig. 6D–F). The abnormal inflammatory cell invasion and secretion of proinflammatory factors in the model group were significantly inhibited by COS intervention in a dose-dependent manner (Fig. 6D–F). These results revealed that the reduction in RF by COS may be related to the inhibition of EMT and inflammation. Oxidative stress is one of the key factors in the formation of RF. Its continuous existence can stimulate a variety of cytokines and activate multiple signal pathways that lead to the occurrence and development of RF. As shown in Fig. 7, the levels of SOD (Fig. 7A), GSH (Fig. 7B) and GSH-Px (Fig. 7C) in obstructed kidney tissue from model mice were significantly lower than those in the normal group, indicating that UUO induced a significant decrease in the anti-free radical ability. At the same time, a significantly higher MDA level than the normal group (Fig. 7D) also demonstrated a remarkable increase in oxygen consumption and free radical production in kidney tissue. These data suggested that the obstructed kidney tissues from model mice already had a higher state of oxidative stress injury. After COS and fosinopril intervention, however, its antioxidant capacity and free radical production were significantly enhanced and decreased, respectively. COS intervention not only increased SOD (Fig. 7A), GSH (Fig. 7B) and GSH-Px (Fig. 7C) levels, but also decreased MDA content (Fig. 7D) in the obstructed kidney tissues in a dose-dependent manner. These results confirmed that COS can prevent and cure RF by reducing oxidative stress injury. A large body of evidences suggests that TGF-β plays a critical role in promoting kidney fibrosis by affecting multiple pathways and cell types, making it a prime therapeutic target. Furthermore, an increasing number of studies have identified that Smad proteins as transmitters of signals in TGF-β cells. Currently, dysregulation of the TGF-β/Smad pathway is an important pathogenic mechanism of RF. As shown in Fig. 7E–G, the significantly upregulated TGF-β1 (Fig. 7E,G), p-Smad2/Smad2 (Fig. 7E,F), p-Smad3/Smad3 (Fig. 7E,F) and Smad4 (Fig. 7E,G) expression, as well as markedly downregulated Smad7 (Fig. 7E,G) expression, were observed in renal tissues obstructed kidney tissue of mice from the model group compared with those of the normal group. The results proved that UUO without any drug interventions could induce serious dysregulation of TGF-β/Smad pathway to promote the formation and development of RF. Amazingly, we can clearly see from Fig. 7E–G that the abnormal expression of TGF-β/Smad pathway proteins induced by UUO was significantly inhibited after the COS or fosinopril intervention. COS intervention not only attenuated the upregulation trend of TGF-β1 (Fig. 7E,G), p-Smad2/Smad2 (Fig. 7E,F), p-Smad3/Smad3 (Fig. 7E,F) and Smad4 (Fig. 7E,G) proteins induced by UUO, but also weakened the decrease in Smad7 (Fig. 7E,G) protein expression levels. More interestingly, the containment effect of COS on TGF-β/Smad pathway protein activation expression also showed an obvious dose-dependent relationship. These data fully prove that the mechanism of COS inhibition of RF may be related to its regulation of the TGF-β/Smad pathway. In this study, we demonstrated that COS protected against RF in a UUO-induced mouse model. We found that COS alleviated clinical symptoms, kidney function, tubulointerstitial injury, ECM accumulation and renal fibrosis in UUO mice. Further studies indicated that COS could inhibit oxidative stress and inflammation and modulate TGF-β/Smad signal to improve renal fibrosis. More importantly, these results showed that COS might be a potential therapeutic agent for renal fibrosis. The model of RF induced by UUO has the advantages of easy replication, reversibility, short time requirement, high success rate and small animal injury. Currently, UUO is the most widely used model for the study of nonimmune renal tubulointerstitial fibrosis. Therefore, this classic model is also used in this study to evaluate the effect of COS on alleviating renal fibrosis. The results of this study showed that the levels of BUN, SCr and the renal organ index after modeling were significantly increased. Renal hypertrophy and hyperplasia appeared in varying degrees. In addition, the kidneys of mice in the UUO model group displayed obvious pathological damage, including obvious expansion and atrophy of renal tubules, a large area of inflammatory cell infiltration, obvious focal fibrosis of the renal interstitium, and extensive shedding of the brush edge of the lumen, which were basically consistent with the UUO model lesions reported in the literature. Interestingly, these indices were significantly reversed after treatment with COS or fosinopril (an angiotensin converting enzyme inhibitor that widely applied in the clinical treatment of chronic kidney disease). Therefore, in this study, fosinopril was used as the reference standard, and COS has a similar effec as fosinopril. Excessive accumulation of ECM is an important pathological basis of RF. A growing body of research has also shown that continuous injury, inflammation and other factors lead to changes in the kidney tissue microenvironment, which stimulate the activation of fibroblasts and myofibroblasts to secrete a large amount of fibrotic-promoting ECM. At the same time, the degradation of ECM was also inhibited. Changes in the composition, content and physical properties of the ECM lead to the shedding, apoptosis, or transdifferentiation into myofibroblasts in the kidney, which further aggravates the deposition of the ECM. The normal tissue structure of the kidney is gradually compressed and eventually replaced by a dense and firm ECM, and renal function is permanently lost. Interventionism against the deposition of different ECM at each stage is expected to be a potential effective means of clinical antifibrosis therapy. It is exciting to see that COS could significantly inhibit the excessive and abnormal secretion of collagen IV, fibronectin and collagen I induced by UUO. In addition, a large number of studies have confirmed that when EMT occurs in renal tubular epithelial cells, the expression of E-cadherin is not only reduced or lost in renal tissues, but α-SMA will be re-expressed or increased. In the present experiment, we found that the expression of α-SMA in renal tubular epithelial cells was significantly reduced by COS intervention. Moreover, E-cadherin expression was observed in most renal tubular epithelial cells. Because oxidative stress can induce an increase in reactive oxygen species (ROS) to aggravate the formation of ECM and EMT, further studies also indicated that COS significantly improved the adverse oxidative stress injury induced by UUO. After COS intervention, increased SOD, GSH and GSH-Px secretion and the decreased MDA level compared with those of the model group were clearly observed. These results provide strong evidence confirming that COS can inhibit ECM and EMT to exhibit an excellent anti-RF effect. Progressive RF is mediated by a variety of mechanisms and mediators, including growth factors, cytokines, metabolic toxins and stress molecules. Among them, TGF-β1 is considered as a key factor in the pathogenesis of RF. Moreover, the Smad protein is the intracellular transduction molecule of the TGF-β1 signal from the receptor to the nucleus, and it plays an important role in the regulation of signal transduction. TGF-β1/Smads belong to the most typical TGF-β signal pathway in RF regulation. When kidney injury occurs, TGF-β1 is transduced by its type I and type II threonine/serine kinase receptors on the cell membrane surface, and activates the downstream Smad2 and Smad3 proteins by phosphorylation. The phosphorylated Smad2 and Smad3 proteins bind to the Smad4 protein and then shuttle to the nucleus, thereby regulating the transcription of RF related response genes. In contrast, Smad7 exhibited antifibrotic effects by inhibiting the phosphorylation of Smad2 and Smad3. In this study, it was observed that the expression of TGF-β1, p-Smad2, p-Smad3 and Smad4 in the obstructed kidneys of model group mice was significantly increased except for the Smad7 reduction compared with that of the normal group, confirming the involvement of the TGF-β1/Smad signal pathway in the UUO-induced RF process. Surprisingly, COS dose-dependently inhibited the expression of TGF-β1, p-Smad2, p-Smad3 and Smad4 and increased the expression of Smad7. Their changing trends were similar to those when RF was inhibited. These results suggested that COS plays a renoprotective role in protecting against RF by inhibiting the classical TGF-β1/Smad signal pathway. In conclusion, COS revealed a significant anti-RF effect. It can not only improve the clinical symptoms and kidney functions of RF, but also can reduce renal inflammation and structural damage and decrease ECM secretion. The improvement effect of COS on RF was closely related to the inhibition of EMT transformation and the reduction of oxidative stress injury. The anti-RF effect of COS was regulated by the TGF-β1/Smad signal pathway. Nevertheless, there were some limitations due to the experimental conditions, and the article length limit should be mentioned in the present paper. First, the current experimental evidence indicates that both COS and fosinopril are effective in treating UUO-induced RF. However, whether they have synergistic enhancement or antagonistic effects is unknown and needs to be further studied. Second, this paper was only grouped based on different doses of COS to observe its dose–effect relationship. Multiple repeated samples were used for the same COS dose intervention. Multiple cohorts need to be adopted to observe the beneficial effect of COS treatment in the future. In addition, the current experimental design used prophylactic administration and only compared the differences in kidney weight and injury between UUO and sham surgery. In further studies, the release of urinary retention should be added for systematic comparison. COS was prepared by Dr. Wu Jun in the laboratory according to the method reported. In order to confirm whether the prepared sample was COS or not, it was characterized and analyzed by Fourier transform infrared spectroscopy and size-exclusion HPLC chromatography (HPGPC). As shown in Fig. 8A, the strong absorption peak at 3361.9 cm−1 was assigned to the stretching vibration of the –O–H and –N–H bond. The characteristic absorption peaks at 2918.4 and 2876.8 cm−1 were identified as the stretching vibration of C–H. Meanwhile, the deformation vibration of C–H was also found at 1423.4, 1379.3, 1324.1 cm−1. The 1654.8 and 1598.7 cm−1 were attributed to the characteristic absorption peaks of amide bond I and II. These results suggested the presence of unremoved acetyl groups in the sample. The 1157.5 and 1075.7 cm−1 were considered to the stretching vibration of –C–C– bond. Furthermore, the absorption peak at 897.4 cm−1 suggested that there was the β-d-glucosamine symmetric stretching vibration. Taking together, these characteristic absorption peaks of sample indicated that it may be COS or chitosan with a partial acetyl group. On this basis, the Mw of the sample was further determined to be 4.21 kDa by HPGPC (Fig. 8B). The results combined with the infrared spectrum analysis data can fully confirm that the sample used in this experiment is indeed COS. In addition, the average deacetylation degree of the sample by acid–base titration was 92.8% in accord with the results of infrared spectrum analysis. Fosinopril was purchased from Bristol-Myers Squibb Co., Ltd. (New York, USA). Assay kits, including serum creatinine (SCr), blood urea nitrogen (BUN), SOD, GSH, GSH-Px, MDA, TNF-α and IL-1β were bought from Jiancheng Biotechnology Co., Ltd. (Nanjing, China). The primary antibodies, GAPDH primary antibodies and goat anti-rabbit secondary antibody were purchased from Abcam Co., Ltd. (Cambridge, United Kingdom). BCA Protein Assay Kit was purchased from Beyotime (Shanghai, China). ECL western blot kit was purchased from Millipore (WBKLS0500, USA). Other reagents used were of the highest commercial grade available. Male BALB/c mice (6–8 weeks-old and weighing 18–22 g) were purchased from the Animal Center of Guangzhou University of Chinese Medicine. Mice were placed in Experimental Animal Room (12-h light–dark cycle, 22 ± 2 °C and 60–70% relative humidity) and provided animal feeds and distilled water by free foraging. They were acclimatized to their housing environment for seven days prior to experimentation and to the experimental room for 1 h before experiments. This study was conducted strictly by the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The related protocol was authorized (No. SCXK2008-0020) and supervised by the Ethics Committee for Animal Experiments of the Guangzhou University of Chinese Medicine. The UUO model was performed by an established protocol described previously. Methods were reported in accordance with the Animal Research: Reporting of in vivo Experiments (ARRIVE) guidelines. Where after, mice were randomly divided into 5 groups of Fig. 8C for the intervention. Mice in the treatment groups were administered by gastric gavage for 7 days, The doses of COS and fosinopril were used in this study according to previous reports, respectively. After the mice were anesthetized by intraperitoneal injection of 10% pentobarbital sodium solution (3 mL/kg) and blood drawn, they were sacrificed and then quickly removed their kidney tissue at the end of 8th day. The kidney tissues were carefully separated with forceps and the mucosal tissue on the surface of the organs was removed. The surface was blotted with filter paper and weighed with an analytical electronic balance. The level of BUN and SCr in serum were detected using corresponding assay kits. The operation procedures were in accordance with the company's instruction manual. Determination of inflammatory cytokines secretion levels was carried out by assay kits. The operation procedures were in accordance with the company's instruction manual. The sections of kidney tissues were stained using hematoxylin–eosin (H&E) and masson’s trichrome hematoxylin according to the reagent manufacturer's standard protocol. All histological data were obtained in a blinded manner by two independent observers. The degree of inflammation was evaluated and scored according to the method reported. The degree of RF was also observed based on the area of fibrotic staining. The paraffin embedded sections of the kidney tissues were dewaxed in xylene, and the antigen was repaired after graded alcohol rehydrated. Sections were incubated with 3% hydrogen peroxide for 10 min to quench the endogenous peroxidase, and blocked with 5% BSA in 20 min. The diluted primary antibodies were added and incubated at 4 °C overnight. Then, the sections were incubated with a secondary antibody at 37 °C for 30 min, followed by incubation with horseradish peroxidase. DAB was used for colouring development, and hematoxylin was used to stain the nucleus. The semi-quantitative analysis of the protein expression region was performed by using image pro plus (IPP) 6.0. Total RNA was extracted by a High Pure RNA Isolation Kit (Tokyo, Japan), and reverse-transcribed by a Transcriptor First Strand cDNA Synthesis Kit as the manufacturer’s instructions (Roche, Germany). The PCR reaction mixture in a 20 µL volume contained 10 µL of SYBR Premix Ex Taq II (Takara Bio, Japan), 1.0 µL of reverse transcription product, 0.4 µL of sense and antisense primer sets and 8.2 µL of double distilled water. The housekeeping gene GAPDH was used as an internal standard. Supplementary Table S1 presented the primers used to amplify the genes. Total protein was fractionated on 8–12% SDS-PAGE gel, and then transferred to a 0.45 μm PVDF membrane. The membrane was cut into pieces based on the molecular weight of the target protein and washed three times with 1 × TBS with 0.1% Tween-20 (TBST). After incubated with 5% of non-fat milk blocking buffer, the membrane was incubated overnight at 4 °C in primary antibody. The membrane was washed in TBST three times, and then incubated in goat anti-rabbit (1:5000, ab6721, Abcam, USA), goat anti-mouse (1:5000, A21010, Abbkine, USA) or rabbit anti-goat (1:5000, A21110, Abbkine, USA) secondary antibodies for two hours. The membrane was visualized by ECL detection reagent (RPN2232, GE Healthcare, USA). Images were acquired using Tanon 6600 Luminescent Imaging Workstation (Tanon Science and Technology Co., Ltd. Shanghai, China) and quantified by ImageJ software (version 1.48v, NIH, Bethesda, MD, USA). Data were presented as mean ± SD unless stated otherwise. Statistical analysis for multiple groups was performed by one-way ANOVA followed by post hoc test when F achieved P < 0.05 and there was no significant variance in homogeneity. Some results were normalized to control to avoid unwanted sources of variation. P < 0.05 was considered statistically significant. Supplementary Information.

PMC9649629

Rory L. Williams,Richard M. Murray

Integrase-mediated differentiation circuits improve evolutionary stability of burdensome and toxic functions in E. coli

10-11-2022

Expression systems,Synthetic biology,Genetic circuit engineering,Computer modelling

Advances in synthetic biology, bioengineering, and computation allow us to rapidly and reliably program cells with increasingly complex and useful functions. However, because the functions we engineer cells to perform are typically burdensome to cell growth, they can be rapidly lost due to the processes of mutation and natural selection. Here, we show that a strategy of terminal differentiation improves the evolutionary stability of burdensome functions in a general manner by realizing a reproductive and metabolic division of labor. To implement this strategy, we develop a genetic differentiation circuit in Escherichia coli using unidirectional integrase-recombination. With terminal differentiation, differentiated cells uniquely express burdensome functions driven by the orthogonal T7 RNA polymerase, but their capacity to proliferate is limited to prevent the propagation of advantageous loss-of-function mutations that inevitably occur. We demonstrate computationally and experimentally that terminal differentiation increases duration and yield of high-burden expression and that its evolutionary stability can be improved with strategic redundancy. Further, we show this strategy can even be applied to toxic functions. Overall, this study provides an effective, generalizable approach for protecting burdensome engineered functions from evolutionary degradation.

Integrase-mediated differentiation circuits improve evolutionary stability of burdensome and toxic functions in E. coli Advances in synthetic biology, bioengineering, and computation allow us to rapidly and reliably program cells with increasingly complex and useful functions. However, because the functions we engineer cells to perform are typically burdensome to cell growth, they can be rapidly lost due to the processes of mutation and natural selection. Here, we show that a strategy of terminal differentiation improves the evolutionary stability of burdensome functions in a general manner by realizing a reproductive and metabolic division of labor. To implement this strategy, we develop a genetic differentiation circuit in Escherichia coli using unidirectional integrase-recombination. With terminal differentiation, differentiated cells uniquely express burdensome functions driven by the orthogonal T7 RNA polymerase, but their capacity to proliferate is limited to prevent the propagation of advantageous loss-of-function mutations that inevitably occur. We demonstrate computationally and experimentally that terminal differentiation increases duration and yield of high-burden expression and that its evolutionary stability can be improved with strategic redundancy. Further, we show this strategy can even be applied to toxic functions. Overall, this study provides an effective, generalizable approach for protecting burdensome engineered functions from evolutionary degradation. As synthetic biology aims to engineer cells with the capacity to regulate and execute increasingly complex and burdensome functions, strategies which address the evolutionary instability of synthetic functions will only become more essential. It has long been observed that cell fitness negatively correlates with heterologous gene expression level, and increased burden results in a shorter evolutionary half-life of engineered functions. Efforts to improve the evolutionary stability of engineered functions have taken a variety of forms, including the most straightforward goals of reducing mutation rate through sequence design or host-strain engineering to delay the appearance of mutations, or reducing burden constitutively or dynamically to mitigate their selective advantage. Additional strategies have improved evolutionary stability by various means of delaying or preventing the selection for mutations, including genomic integration of numerous copies, linking expression of a gene of interest (GOI) to an essential gene, addicting cells to the product of a metabolic pathway, or iteratively displacing populations of cells before mutational escape occurs. While these strategies have indeed made headway, they vary in their ability to be generalized to diverse functions, and in the effort required to do so. Furthermore, each of these strategies requires the GOI to be expressed by every cell in the population, fundamentally restricting their application to GOIs that are non-toxic and of low burden. A homogenous population of cells all performing the same function is largely unique to the laboratory environment, and recent years have seen the merit of breaking with this paradigm by engineering consortia instead of individual strains. With inspiration from microbial communities, there have been several successful implementations of metabolic division of labor for producing biomolecules of interest. This strategy has numerous advantages, including reducing the number of genes and associated metabolic load in each specialized cell type, allowing independent optimization of separate pathways, and spatially separating potentially incompatible functions. While these benefits may be realized by combining in co-culture independently engineered strains or species, additional attractive properties are possible with dynamically regulated division of labor within a single strain that encompasses both metabolic and reproductive functions. The use of differentiation to coordinate such division of labor is a recurring strategy used by microorganisms, but it has not yet been fully explored for addressing the evolutionary constraints of engineered functions. Natural systems that use differentiation to coordinate a division of labor are characterized by the cooperation of specialized cells carrying out distinct functions to realize an inclusive fitness benefit. This is seen with the multicellular cyanobacteria Anabaena, where photosynthetic vegetative cells reproduce and terminally differentiated heterocyst cells fix nitrogen and do not reproduce. The importance of multicellularity for the evolutionary stability of differentiation-facilitated division of labor has been highlighted in both natural and engineered systems, an observation which likely explains the absence of examples of differentiation in unicellular species. Multicellularity is necessary for natural differentiation systems because it minimizes the extent to which non-differentiating mutants can benefit from differentiated cells. However, while natural systems use differentiation to coordinate beneficial or essential functions, here we apply this strategy to functions that are instead both unnecessary for host survival and burdensome to cell growth. This important difference means that multicellularity would not bolster evolutionary stability in this context. This also necessarily means that there is no intrinsic safeguard against the expansion of non-differentiating mutants in our proposed system, a feature that we will later discuss and address. The rational for implementing a differentiation strategy is to allow for a division of labor for the functions of (1) reproducing and (2) executing the function of interest, and critically to prevent the selection for mutations which disrupt this function. To adopt this strategy into a synthetic context, we develop a circuit architecture consisting of two cell types, with progenitor cells being specialized for the faithful replication of an encoded function in the absence of the burden associated with its expression, and differentiated cells for the execution of the encoded function (Fig. 1B). We utilize Bxb1 integrase-mediated recombination to simultaneously activate T7 RNAP-driven expression of a burdensome engineered function and inactivate the expression of π protein (an essential factor for R6K plasmid replication). We describe the strategy in which differentiated cells execute the function and can grow and divide indefinitely as differentiation. As differentiation results in loss of the R6K plasmid through dilution from cell growth and division, the proliferation of differentiated cells can be limited with antibiotic selection. We refer to the strategy in which the proliferation of differentiated cells is limited as terminal differentiation, similar to the use of this term in describing the terminally differentiated heterocyst cells of Anabaena. Because T7 RNAP-driven expression is not activated prior to differentiation, there is no selective pressure for mutations which disrupt this function in the progenitor population. The limitation of differentiated cell growth with terminal differentiation then prevents such mutations from exponentially expanding in the differentiated cell population. Here, we demonstrate computationally and experimentally that terminal differentiation increases the evolutionary stability of burdensome engineered functions, and that this strategy can be improved with strategic redundancy. We show further that terminal differentiation is robust both to mutations which relieve expression burden and to the level of burden. Finally, this robustness to burden is highlighted by the application of terminal differentiation for the production of a functional toxic protein. An ideal strategy for improving evolutionary stability is agnostic to the engineered function being expressed and can be readily implemented without requiring extensive specialization for each use case. As we will demonstrate computationally, terminal differentiation fulfills these criteria by being robust to both burden level and to mutations which disrupt the function of interest. We develop this intuition with cartoons and deterministic modeling by comparing the performance of differentiation architectures to the benchmark case of engineered expression in which every cell in a homogenous population both encodes and expresses the function. In this strategy, which we designate naive expression (Fig. 1A, C), the initial population of producer cells has a reduced growth rate due to the burden associated with the engineered function. Mutations which inactivate the expression of the burdensome engineered function, designated as burden mutations, give rise to non-producers which do not express the function and have a wild-type growth rate. Loss of expression of the function at the population level results from (1) mutations occurring during DNA replication which disrupt the encoded function, and (2) burden associated with the expression of the function providing selective pressure for these mutations during cell growth and division. Recognizing that reproduction and expression of the burdensome function must occur in same cells to select for non-producers, we proposed a strategy of terminal differentiation which segregates these two features through division of labor, thereby preventing evolution from destroying engineered functions (Fig. 1B). Differentiation confines expression of the burdensome engineered function to the differentiated cell population by (1) activating expression through unidirectional differentiation, and terminal differentiation restricts replication and proliferation to the progenitor cell population by (2) limiting growth of differentiated cells. These two components are both necessary to complete the division of labor strategy. With differentiation, a population of non-producing progenitor cells have a wild-type growth rate and differentiate into producer cells which have a reduced growth rate (Fig. 1D). Burden mutations can occur in both differentiated producers (generating non-producers) and in progenitors (generating progenitors*), however, selective pressure for burden mutations only exists in differentiated cells as such mutations do not affect the growth rate of progenitors. Limiting the growth of differentiated cells as in terminal differentiation then fully prevents the expansion of non-producers (Fig. 1E). The differentiation architecture, however, is susceptible to a new category of mutations occurring in progenitor cells which would destroy their capacity to differentiate. We refer to this category of mutations as differentiation mutations, and cells that can no longer differentiate as non-differentiators. While in the differentiation circuit without limited division, both non-producers and non-differentiators have a selective advantage, with terminal differentiation only non-differentiators have a selective advantage (Fig. 1E). We model differentiation, terminal differentiation, and naive expression deterministically with systems of ordinary differential equations describing carrying capacity limited growth in a chemostat with constant dilution. In these simulations, we model the production of an arbitrary protein by producer cells and examine the impact of (1) production burden through adjusting the specific growth rate of producer cells, and (2) the burden mutation rate. For differentiation, progenitor cells have a wild-type growth rate, the rates of differentiation and differentiation mutations are varied, and the number of post-differentiation cell divisions is modeled explicitly for terminal differentiation (Supplementary Note 1 for full description). Our intuition behind these architectures bears out in modeling, revealing terminal differentiation to be uniquely robust to burden mutations and to the degree of burden. Following what has long been observed experimentally, with naive expression non-producers overtake the population more quickly with higher burden expression, resulting in a faster loss of function (Fig. 1F, I, J; Supplementary Fig. 1D). While this is also true for differentiation without limited growth (Fig. 1G, I, J; Supplementary Fig. 1D), the duration of function for terminal differentiation, strikingly, is unaffected by burden (Fig. 1H–J; Supplementary Fig. 1D). This is because burden mutations do not provide a selective advantage, and non-producers do not expand in the population (Fig. 1H). While the performance of naive expression and the differentiation circuit without limited cell division are susceptible to burden mutations and, therefore, sensitive to the rate of burden mutations, terminal differentiation is robust to this rate (Supplementary Fig. 1G). Further, decreasing the rate of the differentiation mutation uniquely improves longevity for terminal differentiation (Fig. 1I–J). As differentiation mutations are the singular Achille’s heel of the terminal differentiation architecture, any decrease in the rate or probability of these mutations will increase the duration of function. Because this circuit is agnostic to the specific function being expressed and robust to the rate of mutations which disrupt that function, reducing the probability of breaking the differentiation mechanism will improve the evolutionary stability of any function regardless of burden. While the strategy of terminal differentiation is robust to burden level and burden mutations, this comes at the cost of reduced expression of the function of interest. This is because (1) the progenitor cell population does not express the function, and (2) limiting the growth of a cell after differentiation also limits the production achieved by its lineage. If the burden is low, the benefit does not overcome the cost and naive production outperforms terminal differentiation. We can make sense of this by understanding the source of selective pressure. With naive expression, the expression burden provides the selective pressure for burden mutations. With terminal differentiation, where differentiation mutations result in circuit failure, the selective pressure instead comes from the rate of differentiation. From the perspective of progenitor cells, differentiation is equivalent to death, and effectively acts to reduce the growth rate of the population. A consequence of this is that lower rates of differentiation allow for longer duration of function, though with a slower rate of production due to a smaller fraction of cells being differentiated producers (Supplementary Fig. 1E). In this tradeoff between production rate and duration of production with varying differentiation rates, an intermediate differentiation rate strikes a balance and achieves the most total production (Supplementary Fig. 1E). With sufficiently low expression burden, the selective pressure for differentiation mutations in terminal differentiation due to the rate of differentiation is greater than that for burden mutations in naive expression. The strategy of terminal differentiation, therefore, is expected to be beneficial only for functions that are sufficiently burdensome. Further, the magnitude of this benefit is expected to increase with the degree of burden, and with decreases in the rate of differentiation mutations. In order to experimentally implement the differentiation strategy, we required (1) burdensome expression to be fully off in the progenitor cell population, (2) irreversible and inducible activation of an arbitrary function through differentiation, and, in the case of terminal differentiation, (3) means of limiting the growth of differentiated cells. To accomplish this, we used Bxb1, a bacteriophage serine integrase which catalyzes unidirectional DNA recombination between specific sequences of DNA. With strategic placement of integrase attachment sites on the genome, a single integrase-mediated recombination event can simultaneously activate and inactivate the expression of the desired genes. To reduce the impact of leaky Bxb1 expression we used the strong ssrA degradation tag LAA, and to allow tuning of Bxb1 expression and, therefore, differentiation rate we used the salicylate-inducible promoter PSalTTC and its cognate transcription factor NahRAM. In order to limit the capacity of differentiated cells to proliferate in the case of terminal differentiation, we take advantage of the reliance of R6K plasmid replication on the π-protein encoded by pir. We used the 3OC12-HSL (Las-AHL) inducible promoter PLasAM and its cognate transcription factor LasRAM to control the expression of π-protein, and placed its expression cassette such that the recombination event results in its excision (Fig. 1B). The π-protein abundance and R6K plasmid copy number at the time of differentiation can therefore be tuned with Las-AHL. As the R6K plasmid encodes the sole source of chloramphenicol resistance (CmR), the induction level of π-protein sets the limit on number of divisions possible upon differentiation in the presence of chloramphenicol selection. In an initial evaluation of integrase differentiation, we demonstrated that the differentiation rate and R6K copy number could be controlled, and the fraction of cells in the progenitor and differentiated state tuned with a combination of chloramphenicol selection and Las-AHL/salicylate induction (Supplementary Fig. 2). Both to allow any arbitrary function to be expressed and to prevent leaky expression of the function in progenitor cells, we selected T7 RNAP, an orthogonal RNA polymerase broadly used in synthetic biology and bioproduction, to be activated by this recombination. Importantly, T7 RNAP can then drive the expression of any protein or burdensome engineered function desired by the user. To allow the expression level and burden to be tuned, the evolved IPTG inducible promoter PTac and associated transcription factor LacIAM were used to control the expression of T7 RNAP. Recombination-activatable T7 RNAP was integrated in a single copy on the E. coli genome, and a high copy ColE1-AmpR plasmid with T7 RNAP-driven sfGFP was used to report T7 RNAP expression. Initial designs which contained a ribosomal binding site (RBS) adjacent to an intact T7 RNAP coding sequence prior to recombination (Supplementary Fig. 3A, B) displayed leaky sfGFP expression in progenitor cells above negative control lacking T7 RNAP (Supplementary Fig. 3F). To address this we relied on previous studies splitting T7 RNAP into functional domains to rationally choose a split site. With this strategy, there is no potential for leaky expression of functional T7 RNAP prior to differentiation, and the full-length coding sequence that is generated upon recombination contains a 17 amino acid insertion from the attL site and additional bases inserted to conserve the reading frame (Fig. 1B, Supplementary Fig. 3C). In the absence of Bxb1 integrase, sfGFP production was equivalent to the control without T7 RNAP present, and induction of Bxb1 allowed high-level T7 RNAP-driven expression (Supplementary Fig. 3F). In addition to the first three necessary criteria that have been fulfilled, the emergence of non-differentiators should be delayed in the terminal differentiation architecture by addressing the rate of differentiation mutations. From our initial deterministic modeling, we observed that decreasing the rate or probability of differentiation mutations improves the functional duration for terminal differentiation by delaying the emergence of non-differentiators. We reasoned that increasing the number of independent mutations required to break the differentiation mechanism would yield more significant improvements than decreasing the rate of mutations by sequence design. To this end, we envisioned an identical circuit design (Fig. 1B) that instead has two T7 RNAP differentiation cassettes. Importantly the recombination of a single cassette should both activate the function and enable limiting the growth of differentiated cells. If a second identical cassette was integrated, recombination of both cassettes would be required to cease replication of the R6K plasmid and allow limitation of growth through antibiotic selection. Therefore, a single mutation preventing the recombination of one cassette would be sufficient to obviate antibiotic selection and allow burden mutations to expand. However, if each differentiation cassette encoded a unique half of the π-protein, a single recombination event would ablate the expression of functional π-protein and with it the replication of the R6K plasmid. In this case, two independent mutations would be required to generate non-differentiators and circumvent antibiotic selection, and burden mutations would have no opportunity to expand. To accomplish this, we split the pir coding sequence and tagged the N- and C-terminal fragments with the N- and C terminal fragments of the Cfa intein, respectively, functionally screened for R6K plasmid replication, and identified a functional split site (Supplementary Fig. 4). Expression of the intein-tagged fragments allows R6K plasmid replication, and inactivation of either the N-terminal fragment or C-terminal’ fragment through integrase-mediated recombination results in loss of the R6K plasmid. This split-π protein design was incorporated into a 2x differentiation circuit in which cells have two integrase-activatable T7 RNAP cassettes, with each encoding a separate intein-tagged π-protein fragment. Both the 2x differentiation strain and the corresponding 1x differentiation strain (having a single cassette encoding the full-length π-protein) were genomically integrated with two salicylate-inducible integrase (Bxb1-LAA) cassettes (Fig. 1A, B; Supplementary Fig. 5). In characterizing these circuits in ranging induction conditions, we noted that the R6K copy number and performance of the 2x differentiation circuit was more sensitive to the concentration of Las-AHL (controlling production of π-protein/split-π protein) than the 1x differentiation circuit (Supplementary Method 1; Supplementary Figs. 6–9), and selected 10 nM Las-AHL as the appropriate concentration for subsequent experiments. The rate of differentiation is important as it determines the fraction of producer cells and, therefore, the rate of production, as well as the duration of function as we previously discussed. Accordingly, we characterized the dose-response of Bxb1 induction with salicylate on differentiation rate for both the 1x differentiation circuit and the 2x differentiation circuit, in both the non-terminal differentiation (-chlor) and terminal differentiation contexts (+chlor) at two burden levels (Supplementary Figs. 10–11). This revealed very little leaky differentiation for both circuits (~1% or less differentiated cells as determined by GFP + fraction with flow cytometry without salicylate) regardless of burden (IPTG induction) or chloramphenicol selection. We also observed a good dynamic range between 10 μM and 30 μM salicylate for both circuits, though generally, 2x differentiation was somewhat more sensitive than 1x differentiation to salicylate concentration. Specifically, after 8 h growth following a 1:50 dilution, 1x differentiation achieved ~0.2%, ~18%, ~64%, and ~94% differentiated producer cells (GFP + ) with 0, 10, 20, and 30 μΜ salicylate in the lower burden condition (10 μM IPTG), with similar though slightly lower percentages in the higher burden condition (~0.2%, ~13%, ~56%, and ~90%); and 2x differentiation achieved ~1%, ~30%, ~76%, and ~97% in the lower burden case, and ~1%, ~43%, ~74% and ~97% in the higher burden condition (Supplementary Fig. 10). In order to assess the benefit of the differentiation and terminal differentiation architectures, it was important to experimentally compare these to an equivalent naive circuit in which all cells were expressing the burdensome function. We, therefore, constructed 1x and 2x naive T7 RNAP expression strains in which the genomically integrated cassettes were identical to the sequence produced upon Bxb1 recombination (with the exception of not containing LasRAM and NahRAM, the transcription factors that were unnecessary for naive expression), and characterized the dose-response of IPTG induction on T7 RNAP-driven GFP expression and its associated impact on growth rate (Fig. 2C, Supplementary Fig. 5C, D, Supplementary Fig. 12). With this characterization, we see a dose-response between IPTG induction and both GFP production and burden as inferred through growth rate. For 1x naive there was a ~11%, ~31%, ~50%, and ~57% growth penalty with 10, 20, 30, and 50 μM IPTG, respectively, which yielded ~2×, ~3.8×, ~4.9×, and ~6.5× the OD normalized GFP production relative to uninduced at 8 h. Similarly, for 2x naive there was a ~27%, ~54%, ~64%, and ~65% growth penalty, with ~3.1×, ~4.8×, ~6×, and ~7.6× the OD normalized GFP production relative uninduced 1x naive, respectively (Supplementary Fig. 12). To assess the capacity of differentiation strategies to improve the evolutionary stability of burdensome functions, we performed long-term experiments with T7 RNAP-driven expression of a fluorescent protein, which here serves as a model burdensome engineered function. We compared the duration and total amount of production achieved in cells with one or two copies of inducible T7 RNAP (1x and 2x naive) to our single-cassette and two-cassette differentiation circuits (1x and 2x differentiation). Both the single-cassette differentiation and two-cassette differentiation strains have two copies of inducible integrase, and critically all components in the naive and differentiation circuits were genomically integrated, ensuring precise copy number control and preventing effects due to plasmid partitioning (Fig. 2A–C, summary of integration cassettes in Supplementary Table 1). All genomic insertions were verified by whole genome sequencing (Supplementary Data 1). Experimental comparison of differentiation with terminal differentiation required only including chloramphenicol in the medium in the case of terminal differentiation, as without antibiotic present differentiated cells would grow unhindered after losing the R6K plasmid. Inducer and antibiotic conditions were uniform throughout the duration of the experiment, with the degree of burden tuned with IPTG (PTac T7 RNAP) and differentiation rate tuned with salicylate (PSalTTC Bxb1-LAA). Experiments were ran for a total of 16 consecutive batch growths with 50× dilutions into a total volume of 300 μL every 8 h, for a total of 128 h (~88 doublings). In these long-term experiments, we observe both the benefit of redundancy, and the superiority of terminal differentiation particularly with higher burden T7 RNAP-driven expression. Comparing 1x naive to 2x naive expression, we clearly see the benefit of redundancy in both the duration of production and total production achieved (Fig. 2C, D; Supplementary Fig. 13 for plots of individual biological replicates). As each copy of inducible T7 RNAP is genomically integrated, each copy must independently mutate in order to fully disrupt its expression. The magnitude of this redundancy benefit is reduced in the higher burden case, however, with 2x naive yielding ~2.2× that of 1x naive in the lower burden case (~221000 ± 31000 vs. ~101000 ± 8000), and ~1.6× that of 1x naive in the higher burden case (~101000 ± 30000 vs. ~65000 ± 4000). For both 1x and 2x differentiation and terminal differentiation, the initial rate of GFP production increases with the differentiation rate (Fig. 2A, B; Supplementary Figs. 14–15 for plots of individual biological replicates). However, higher differentiation rates also lead to an earlier decline in production. These two counteracting features result in an intermediate differentiation rate yielding the most total sfGFP production (Fig. 2D). With 1x differentiation, the terminal differentiation condition had a moderate negative effect on total production for both the lower burden (~1.56× 1x naive expression for diff. vs. ~1.37× for term. diff. with 15 μM sal/10 μM IPTG) and higher burden (~2.1× 1x naive expression for diff. vs. ~1.89× for term. diff. with 15 μM sal/10 μM IPTG) conditions. With 2x differentiation, terminal differentiation had minimal benefit in the lower burden condition (~1.53× 1x naive expression for diff. w/ 15 μM sal/10 μM IPTG vs. ~1.66× for term. diff. w/ 20 μM sal/10 μM IPTG), and a larger benefit in the higher burden condition (~2.17× 1x naive expression for diff. w/ 15 μM sal/50 μM IPTG vs. ~2.86× for term. diff. w/ 20 μM sal/50 μM IPTG). As expected, naive expression is more sensitive to burden level, and performs worse at higher burden in comparison to both differentiation and terminal differentiation (Fig. 2D). In interpreting the effect of terminal differentiation for both 1x and 2x differentiation, we note that this has two opposing effects on the amount of production that will be achieved over the lifetime of circuit function. Terminal differentiation decreases output by limiting the growth of producer cells, but increases output by preventing the expansion of non-producers. Because these effects are opposing, one may dominate the other depending on the characteristic parameters. In the case of 1x differentiation, the negative effect dominates at both burden levels, and terminal differentiation performs worse. With 2x differentiation, however, the positive effect dominates, particularly so at higher burden. The critical difference between 1x and 2x differentiation which explains this is that the 1x version requires only one differentiation mutation to yield non-differentiators, while the 2x version requires two mutations for the same effect. This delay in the emergence of non-differentiatiors provides additional time for the positive effect from suppression of non-producers to overcome the negative effect from limitation of producer growth. In order to characterize the mechanisms which caused production of GFP to decrease or cease during the experiment, we performed selective plating assays from glycerol stocks saved after the last plate growth, and used Nanopore sequencing to identify causal mutations in a sample of colonies (Supplementary Fig. 16). We chose the higher burden (50 μM IPTG) and highest differentiation rate (30 μM salicylate) for this characterization as it was the condition which displayed the largest decrease in GFP production during the experiment. Select dilutions were plated from three of the eight replicates each for 1x/2x naive, 1x/2x differentiation, and 1x/2x terminal differentiation. Six colonies total for each strain across replicates were analyzed by sequencing. Sequences for the region of the ColE1 plasmid encoding T7 RNAP-driven GFP and all genomic integrations were obtained, with few exceptions due to failed PCR or insufficient reads. 1x and 2x naive had <1% producers in all replicates, as inferred through the fraction of GFP+ colonies. For 1x naive, sequencing identified causal mutations in the coding sequence of T7 RNAP at the P21 (T) locus in all six colonies, with 3 unique frameshift insertion mutations and 3 unique nonsense mutations, and no mutations were observed in the ColE1 plasmid. For 2x naive, however, no mutations were observed in the T7 RNAP coding sequence at the T or HK022 (H) locus, but instead mutations in the T7 promoter on the ColE1 plasmid were present in 4 of the 6 colonies (Supplementary Data 2). Mutations in the T7 promoter highlight the large contribution of transcriptional burden in this system, and the power of random plasmid partitioning in accelerating the fixation of mutations. As well, that we see these promoter mutations in 2x naive but not 1x naive suggests the aggregate rate of generating and enriching for plasmid mutations through random plasmid partitioning is lower than the rate of genomic mutations which disrupt T7 RNAP expression, but higher than the rate of generating two such mutations. As we discuss the mutations identified in the differentiation and terminal differentiation circuits, we note that no mutations were observed on the ColE1 plasmid apart from those observed with 2x naive expression. For 1x differentiation, 68–77% of colonies were non-differentiators, 23–32% were non-producers or non-differentiators that had lost the R6K plasmid, and none were producers or functional differentiators (mScarlet+, non-fluorescent, and GFP+, respectively, when plated on LB + Kan/Las/Sal). For 1x terminal differentiation, two replicates had 100% non-differentiators, and one replicate had ~99.8% non-differentiators and ~0.2% differentiators/producers (Supplementary Fig. 16). From our intuition and modeling, we know that while non-producers and non-differentiators have selective advantage with the differentiation architecture, only non-differentiators have selective advantage with terminal differentiation. The results of the plating assay reflect this, though the high abundance of non-differentiators in non-terminal differentiation suggests that differentiation mutations are more frequent. We sequenced 3 mScarlet+ non-differentiators, and 3 mScarlet-/GFP- colonies from 1x differentiation. All 3 mScarlet+ non-differentiators had identical inversions at the T locus, strongly suggesting that Bxb1 had catalyzed an inversion between the attB and attP sites rather than an excision. The sequence resulting from this recombination does not contain functional attB or attP sites, but does retain the capacity for π-protein expression. Of the three mScarlet- colonies, one contained a correctly recombined cassette that had a nonsense mutation in T7 RNAP (W221*), one had an inverted cassette as described with intact pir, and one had a large deletion encompassing the attP right half through a portion of NahRAM, also with intact pir (Supplementary Data 2). These data demonstrate that loss of the R6K plasmid occurs without loss of π-protein expression. As well, erroneous recombination by Bxb1, which results in a sequence that is inert to recombination, is likely more frequent than errors of replication having the same affect or solely disrupting the expression of T7 RNAP. For 2x differentiation, ~36%, ~17%, and ~1% of colonies were differentiators/producers; ~64%, ~83%, and ~99% were non-producers or non-differentiators that had lost the R6K plasmid; and ~0.1%, <0.1%, and «0.1% were non-differentiators. For 2x terminal differentiation, ~74%, ~27%, and ~26% were differentiators/producers; ~21%, ~49%, ~71% were non-differentiators (mScarlet+); and ~6%, ~24%, and ~4% were non-producers or non-differentiators that had lost the R6K plasmid (Supplementary Fig. 16). Colonies that were mScarlet-/GFP- were only observed for 2x terminal differentiation when plating on media lacking chloramphenicol, and no such colonies were observed for 1x terminal differentiation. This suggests that the split-π protein design is more susceptible to stochastic loss of the R6K plasmid than the full-length pir design. Though non-differentiators were rare in the non-terminal differentiation condition, we sequenced 3 non-differentiators (mScarlet+) and 3 non-producers (mScarlet-/GFP-). In the non-producers, one colony had an inversion at the H locus as described previously with an intact cfaC-pirR fusion protein, and the cassette at the T locus had an inversion involving the integrase attachment sites but disrupting the pirL-cfaN fusion, ablating functional π-protein expression. The two additional colonies had this same T locus mutation, but high-quality sequences were not obtained for the H locus. For the non-differentiators sequenced, two had inversions maintaining expression of the split π-protein (one with failed sequencing at the H locus). The third non-differentiator sequenced had matching 108 bp deletions in both copies of the Bxb1 integrase, and both cassettes had not been recombined and did not have any mutations. For 2x terminal differentiation, we sequenced 2 colonies which were GFP+/mScarlet+, both of which had an inversion in the T cassette which maintained the intact pirL-cfaN fusion, and a correctly recombined H cassette with no mutations in the T7 RNAP coding sequence. The four non-differentiator colonies (GFP-/mScarlet+) colonies all had mutations which disrupted one or both integrase attachment sites at both the T and H loci but left π-protein expression intact. Simple inversions resulting from erroneous recombination were the most common, but a large deletion from the attP right half through a portion of T7 RNAP, and inversions involving a partial duplication of pirL-cfaN were also observed. (Supplementary Data 2). Importantly, the plating and sequencing performed supports our hypothesis that the terminal differentiation strategy removes selective pressure for burden mutations. We also observed evidence that differentiation mutations occur and are selected sequentially in 2x terminal differentiation, as evidenced through the identification of cells in which one cassette was functional while the second had incurred a differentiation mutation preventing its recombination. It is likely that one such mutation reduces the aggregate rate of differentiation of the cell, and thereby provides a selective advantage. It is also apparent that errors in Bxb1-recombination are frequently the cause of mutations in both differentiation and terminal differentiation, and we speculate that these mutations occur at a higher rate than mutations due errors in DNA replication. This hypothesis is also supported by the experimental observation that 2x differentiation (-chlor) underperforms 2x naive in the low-burden condition (Fig. 2D). As terminal differentiation is uniquely robust to burden mutations, we hypothesized that this robustness extends to other sources, mutational or otherwise, which could impact expression. Because the GOI being expressed by T7 RNAP in these experiments is encoded on a high copy plasmid, plasmid mutations and copy number fluctuations (or even plasmid loss) can impact its expression without requiring mutation of the genomically encoded T7 RNAP. To investigate how such plasmid effects could influence each of our circuit architectures, we expanded the modeling framework previously discussed to reflect the experimental circuit design more accurately, model mutations stochastically, and incorporate plasmid-based effects. We explicitly modeled the genotype of each cassette, incorporated integrase expression cassettes, address integrase mutations and differentiation cassette mutations separately, and modeled the differentiation rate of each cassette as being linearly dependent on the number of non-mutated integrase expression cassettes. Both the total plasmid copy number and the fraction of plasmids with mutations likely can fluctuate widely due to plasmid partitioning effects. However, for simplicity and tractability, we considered plasmid mutation or plasmid loss to be binary, with a single stochastic event either mutating all plasmids, or resulting in complete plasmid loss. Furthermore, though antibiotic selection is used experimentally to ensure plasmid maintenance, communal resistance for certain antibiotics (e.g., β-lactams) can allow antibiotic-sensitive cells to persist in the presence of antibiotic. To capture this effect, we modeled Michaelis-Menten antibiotic degradation by plasmid-containing cells and stochastic plasmid loss. While the growth of plasmid-containing cells was not affected by antibiotic concentration, the growth rate of cells that have lost the plasmid was determined with a Heaviside function. With this the growthrate is 0 if the antibiotic concentration is at or above the minimum inhibitory concentration (MIC), and is that of a non-producer if below the MIC. In simulating cells with one or two copies of inducible T7 RNAP in both the naive and differentiation cases, assumptions were required about the relative burden levels and production rates. In the case of differentiation, the producer growth rate (μP) is the growth rate of a cell with one activated cassette of T7 RNAP, and the burden of the second copy produces a proportionate decrease in growth rate. For example, if a non-producer grows at rate 1 h−1 (μN) and a cell with one cassette active grows at rate 0.5 h−1 (μP), a cell with two cassettes active would grow at rate 0.25 h−1. Production then was assumed to increase linearly with the decrease in growth rate (Supplementary Note 1 for full description of model implementation). From these simulations, we recapitulate several observations from the initial deterministic modeling of the general strategies, and from our experiments. For both differentiation and terminal differentiation with one and two copies, we see that lower differentiation rates result in slower production that lasts longer, high differentiation rates yield faster production that breaks more quickly, and intermediate rates strike a balance and achieve the most total production (Fig. 3C, D). At low burden (higher μP), terminal differentiation is counterproductive, but becomes beneficial as burden increases (Fig. 3C, D). We also see that naive expression performs comparatively well at low burden relative to high burden. For 2x naive expression we model both the case where one cassette alone yields the growth rate μP and its corresponding production rate (2x*), and where the two cassettes together yield the growth rate μP (2x). As expected, at low burden and high differentiation rate, 1x differentiation without growth limitation approximates the performance of both 1x and 2x naive, and 2x differentiation without growth limitation approximates the performance of 2x* (Fig. 3C–E). Further, the redundancy and mutational robustness provided with 2x differentiation improves performance relative to the one cassette case for both differentiation and terminal differentiation. Though we do not experimentally interrogate the impact of number of post-differentiation divisions (ndiv) in the case of terminal differentiation, we do so computationally. While there is a benefit of increasing ndiv at low burden, this effect disappears at higher burden (Supplementary Fig. 17). Critically, this stochastic modeling reveals that the robustness of terminal differentiation to burden mutations generalizes to plasmid-based mechanisms of relieving burden: both plasmid loss compensation by communal antibiotic degradation, and plasmid mutation. Incorporating stochastic plasmid loss in conjunction with antibiotic degradation and growth inhibition negatively affects performance of the naive and non-terminal differentiation architectures in a burden dependent manner, and this effect is much greater in the two-cassette case (Fig. 3C–E). The performance of the terminal differentiation architecture, however, is robust to this plasmid instability. We see this both in the total production achieved (Fig. 3C, D), and in tracking the population of cells which have lost the plasmid. When there is no antibiotic degradation, we see no accumulation of cells lacking the plasmid for any circuit, but with a sufficiently high rate of antibiotic degradation, we see a transitory rise in the fraction of cells that have lost the plasmid for naive (Supplementary Figs. 18–20) and differentiation architectures, but not for terminal differentiation (Supplementary Figs. 21–26). This effect increases with higher rates of antibiotic degradation, is more pronounced at intermediate (50%) burden than very low (10%) or very high (90%) burden for naive expression, and, in the case of differentiation, is influenced by both burden level and differentiation rate. As these plasmid-deficient cells are dependent on communal antibiotic degradation, they do not completely take over the population, but instead are eventually displaced by mutated cells which retain the plasmid. Similarly, when we neglect antibiotic degradation and instead consider plasmid mutation, we see that increasing the rate of plasmid mutation negatively impacts naive and differentiation architectures, but not terminal differentiation (Fig. 3C–E). Further, that experimentally we observed mutations on the ColE1 plasmid disrupting T7 RNAP transcription in the 2x naive, but not 1x naive, corroborates simulation results where the impact of plasmid mutations is more pronounced for 2x than 1x naive expression. Though this effect is similar to plasmid loss in that it affects the redundant architectures more significantly, it is different in that it reveals its effect at lower burden (10%). As well, because there is no dependence on communal antibiotic degradation, the accumulation of cells with plasmid mutations is not transitory (Supplementary Figs. 18–26). Critically, the robustness of terminal differentiation circuits holds true when considering plasmid mutations, and we observe no selection for cells with mutated plasmids or any effect on production (Fig. 3, Supplementary Figs. 21–26). While deliberately varying the rate of plasmid mutations experimentally is difficult, we may instead address plasmid instability through the choice of antibiotic resistance marker. Both to test this mechanism and to select any alternative selectable marker to use for the ColE1 plasmid, we used a plasmid identical to ColE1-KanR-PT7-GFP which instead had ampicillin resistance (ColE1-AmpR-PT7-GFP) and performed co-culture experiments of 1x naive and the parental strain JS006 transformed with these plasmids (Supplementary Fig. 27). As only 1x naive cells produce appreciable levels of GFP from these plasmids, we could observe that cells with only AmpR could allow cells with only KanR to grow in LB with both kanamycin and carbenicillin, while cells with only KanR did not allow cells with only AmpR to grow in the same condition. We therefore concluded that the choice of AmpR on the ColE1 plasmid would likely allow loss of expression through plasmid loss and shared antibiotic resistance, while the same mechanism would not hold, or would be much less apparent, with KanR. Performing the same long-term evolution experiments described in Fig. 2 with the single modification of using AmpR marker instead of KanR corroborates the intuition we gained from modeling. With both the lower and higher burden conditions, changing the selectable marker largely removes the benefit from redundancy with naive expression (Fig. 3F, Supplementary Fig. 28). As well, while production is negatively affected by this change at higher burden for 2x differentiation without limited growth, production is higher with AmpR than KanR for 2x terminal differentiation at both burden levels (Fig. 3F, Supplementary Fig. 28). While we cannot explain this performance benefit from our model, we speculate that the concentration of kanamycin used (50 μg/mL) can negatively affect expression even when KanR is expressed. This stark contrast between terminal differentiation and differentiation or naive expression demonstrates that the robustness of terminal differentiation to burden mutations extends generally to plasmid-based mechanisms of reducing burden. As the terminal differentiation architecture is robust to the level of burden associated with the function of interest, this suggests that even toxic functions could be expressed and made evolutionarily stable. To test this, we aimed to demonstrate that the differentiation circuit we developed could allow the production of a toxic protein that will result in cell death and chose dnaseI as a proof-of-concept example. As progenitor cells do not produce any T7 RNAP, we reasoned that a T7 RNAP-driven dnaseI would not be expressed in the progenitor cells, allowing the encoded function to be replicated without toxicity or selective pressure for mutations. However, construction of a dnaseI expression plasmid identical to that of sfGFP yielded only mutated plasmids. Characterization of leaky expression from a PT7-GFP plasmid in the absence of T7 RNAP revealed fluorescence above background, explaining this inability to isolate functional plasmids (Supplementary Fig. 29). Incorporating two insulating terminators upstream of the T7 promoter mitigated leaky expression in the absence of T7 RNAP (Supplementary Fig. 29), and this insulation in conjunction with reducing the RBS strength allowed construction and isolation of a correctly sequenced dnaseI expression construct. While leak could have also been reduced by using the T7/lacO promoter and an additional source of LacI on the expression plasmid, this would not eliminate leaky expression in progenitor cells upon induction of differentiation and T7 RNAP. Highlighting the importance of preventing leaky expression of toxic functions, transformation of 1x differentiation and 2x differentiation cells with insulated dnaseI plasmid yielded ~600 cfu and ~1000 cfu, respectively, while 1x naive and 2x naive strains yielded 1 and 0 colonies, respectively, compared to >104 cfu for both when transformed with ColE1-AmpR-PT7-GFP control (Fig. 4, Supplementary Table 2). Using the insulated T7 RNAP-driven dnaseI construct, we demonstrated that differentiation could enable expression of this toxic product. To assess the capacity of differentiation to enable functional dnaseI expression, we co-transformed the 2x differentiation strain with the insulated dnaseI expression plasmid, and an R6K-CmR-empty plasmid. After outgrowth without induction, cells were diluted into 25 mL cultures with or without induction with 20 μM salicylate and 10 μM IPTG. After 8 h of growth, un-induced cultures grew to cell densities equivalent to JS006 control (7.2–7.6 g wet weight per liter vs. 7.6 ctrl), while induced cultures reached lower cell densities (2–4.4 gCW/L). Diluting 1:50 into fresh media yielded similar densities for the uninduced cultures after 8 h growth (7.2–7.6 gCW/L), while induced cultures had grown minimally after 8 h (OD600 < 0.05) and to a range of densities after 16 h total growth (2–6.8 gCW/L). The growth of the induced and uninduced cultures indicates that the dnaseI plasmid minimally affected growth when T7 RNAP is not expressed, and that 20 μM salicylate induction likely results in more complete differentiation in a large shaking culture in comparison to small volumes in 96 well microplates in previous experiments. Expression of functional dnaseI was quantified with an activity-based assay on lysate extracted from cell pellets (Supplementary Fig. 30), with activity measured equivalent to ~1.9–4.2 × 104 U/gCW (~3.7 × 104–1.9 × 105 U/L) in the three independent induced cultures and ~65–250 U/gCW (~500−1800 U/L) in the uninduced cultures, compared to ~13 U/gCW (~100 U/L) for the JS006 negative control (Fig. 4B). This yield of dnaseI is on the same order of magnitude as yields reported using T7 RNAP to drive the expression of recombinant dnaseI using the LacI repressible T7 promoter in Bl21(DE3)[pLysS] (1.5 × 104 U/L) and Bl21(DE3)[pLysE] (7.5 × 104 U/L). Here we have developed architectures for implementing differentiation and terminal differentiation in E. coli for the expression of burdensome T7 RNAP-driven functions. Importantly, in our circuit design progenitor cells do not have an intact coding sequence for T7 RNAP, completely eliminating leaky T7 RNAP expression prior to differentiation. We computationally demonstrated that limiting the growth of differentiated cells with terminal differentiation provides robustness to burden level and burden mutations. We further demonstrated that reducing the rate of differentiation mutations delays the emergence of non-differentiators, thereby improving the evolutionary stability of the terminal differentiation architecture and with it any arbitrary function. Experimentally, we developed a differentiation-activated T7 RNAP architecture in which all circuit components that could mutate to disrupt the process of differentiation or expression of T7 RNAP were integrated on the genome, both ensuring exact copy number control and preventing plasmid partitioning effects from accelerating circuit failure. With the goal of delaying the emergence of non-differentiators and thereby increasing the longevity of the terminal differentiation architecture, we developed a split-π protein system using split-inteins. In long-term experiments with repeated dilutions of independent cultures, we compared the performance of 1x and 2x naive T7 RNAP-driven expression to 1x and 2x differentiation and terminal differentiation. We demonstrated that the rate, duration, and total amount of production could be tuned by varying the differentiation rate, with lower differentiation rates enabling longer duration expression but at a slower rate. Differentiation was particularly beneficial in comparison to naive expression with higher burden as expected from modeling, and the redundancy and robustness to differentiation mutations provided by incorporating the split-π protein design into the terminal differentiation architecture proved effective. Though here we demonstrate the effectiveness of terminal differentiation and the stability benefit gained from requiring two mutations instead of one to generate non-differentiators, this can be viewed as demonstration of the power that redundancy can provide in synthetic biology. In considering the scaling of this strategy to longer times and larger population sizes, we expect that higher degree of redundancy will be necessary. As well, addressing errors of Bxb1 recombination which serve as a dominant source of differentiation mutations may improve the performance of terminal differentiation at each level of redundancy. Scaling this specific architecture through further splitting of the π protein may be infeasible, but the toolkit of synthetic biology certainly has means of allowing this strategy to scale further, through the inactivation of essential genes, activation of toxins, or otherwise. We demonstrated both computationally and experimentally that effects due to instability of the T7 RNAP-driven expression plasmid and communal antibiotic resistance can negatively affect the performance of naive expression and differentiation, but that terminal differentiation circuits are robust to this effect. We further showed computationally that the robustness of terminal differentiation circuits to burden mutations extends to the general case of plasmid mutations which disrupt the function of interest. Though genomic integration of functions is more time consuming and cumbersome than plasmid transformation, plasmid instability and cost considerations for antibiotics and inducers in large cultures have often made genomic integration of constitutively expressed functions the preferred method for bioproduction in industry. However, terminal differentiation mitigates effects of plasmid instability, potentially allowing the stability benefits of genomic integration to be obtained with the ease of plasmid transformation. Finally, because there should be no limit on the degree of burden or toxicity of a function expressed with our differentiation system (so long as the toxicity is limited to the cells expressing the function), as a proof of concept we demonstrated that differentiation could enable the expression of functional dnaseI. In the course of this demonstration, we discovered that in the absence of leaky expression of T7 RNAP, non-T7 RNAP sources of leak could prevent isolation of correctly assembled dnaseI expression plasmids. While we mitigated this problem through the incorporation of insulating terminators to prevent transcriptional read-through from upstream of the T7 promoter, reducing the strength of the RBS was still required to isolate correctly sequenced plasmid. Improving this insulation and/or reducing any leaky expression that may be coming directly from the T7 promoter through directed evolution efforts may prove beneficial. Addressing this would allow the use of higher strength RBS sequences without concern for leak, thereby enabling improved yields. While the expression of toxic or highly burdensome products has long been of interest in bioproduction and synthetic biology, and effective strategies have been implemented to accomplish this, to our knowledge all existing strategies only work for single-use batch culture inductions. The critical difference with our strategy of terminal differentiation is that progenitor cells continuously differentiate to replenish the population of cells expressing the toxic function, thereby allowing a toxic product to be produced continuously. We envision this strategy to be readily applied to the expression of burdensome and toxic proteins or metabolic pathways with little to no modification of the system, simply by transformation of plasmid encoding the desired T7 RNAP-driven function. With the performance of the current redundant terminal differentiation architecture, expression of the GOI can continue for 10−16+ plate generations (~55–88+ doublings) depending on differentiation rate. This equates to the number of doublings occurring in ~100–150+ h of continuous culture with a dilution rate of 0.4 h−1, suggesting this could enable continuous bioproduction. Furthermore, in contrast to naive expression where cell growth must be considered in optimizing expression, with terminal differentiation this optimization can be done with production yield as the sole factor, potentially enabling higher per cell production rates. This feature naturally motivates the application of metabolic engineering strategies. While tools like flux balance analysis can inform genetic modification of strains to improve the yield of valuable chemicals, these strategies naturally must be concerned with the growth of the organism. However, with terminal differentiation, genomic and metabolic knobs could be tuned to maximize yield without regard for the long-term viability of the cells. CRISPR/Cas systems have been demonstrated to allow activation and repression in E. coli, have been applied in metabolic engineering efforts, and could be co-opted in this context. We are excited both by these engineering opportunities that are enabled by the ability to neglect cell growth in the producer cell population, and for future development of terminal differentiation to further extend the evolutionary stability of engineered functions in a general manner. The wild-type E. coli strain JS006 was the base strain for the construction of all differentiation and naive circuit strains. Constructs were assembled with a combination of Golden Gate and Gibson assembly using 3G assembly, and were integrated into the E. coli genome using clonetegration. Because the R6K origin used for propagation of pOSIP plasmids from the clonetegration method of genomic integration is the same origin in our differentiation architecture R6K plasmid, we PCR amplified pOSIP backbones in two pieces (RW.posX.FL.F/RW.pos.rmR6K.R and RW.pos.rmR6K.F/RW.pos1.FL.R for pOSIP CT and KH; RW.posX.FL.F/RW.O.s2.R and RW.O.s1.F/RW.pos1.FL.R for pOSIP KO; Supplementary Data 3), removing the R6K origin, for use in Gibson assemblies with desired inserts. For Gibson assembly with linear pOSIP pieces, POS1 and POSX were used as terminal adapters instead of UNS1 and UNSX. The 1x naive and 1x differentiation strains were constructed by integration at the P21 (T) landing site, and the 2x naive and 2x split-π protein differentiation strains by additional integration at the HK022 (H) landing site. 1x differentiation and 2x differentiation strains were integrated two additional times with the inducible Bxb1-LAA expression construct at the primary and secondary phage 186 (O) landing sites (Supplementary Fig. 5, Supplementary Table 1). Following transformation, integrations were checked via colony PCR with the pOSIP p4 primary corresponding to the landing site and a reverse primer common to all pOSIP plasmids (RW.pOSIPchk.rev). Fidelity of integrations was checked with a combination of sequencing and functional screening prior to transformation with pE-FLP to excise the antibiotic resistance cassette and integration module, and integration of subsequent constructs. Final strains (eRWnaive1X, eRWnaive2X, eRWdiff1X, eRWdiff2X) were whole-genome sequenced with MinION using the Rapid Barcoding Kit (Nanopore SQK-RBK004) for verification. Reads were assembled with Flye version 2.8.3 (https://github.com/fenderglass/Flye/) and mapped to reference genomes containing intended genomic insertions in Geneious Prime 2021.1. Modified MoClo compatible parts for T7 RNAP, integrase attachment sites, and terminators were generated with standard molecular biology techniques (PCR, Gibson, oligo annealing, and phosphorylation), and modified UNS adapters used for the construction of polycistronic or inverted transcriptional units. The R6K-CmR backbone was constructed with Golden Gate using an R6K origin amplified from the pOSIP plasmids. Sequences for Bxb1 integrase attachment sites attB and attP were obtained from Ghosh. NahRAM, LasR AM, and LacI AM,, and their corresponding evolved promoters PSalTTC, PLasAM, and PTac were provided by Adam Meyer. The CIDAR MoClo Parts Kit, which includes various promoter, RBS, CDS, and terminator parts used in the constructs described, were provided by Douglas Densmore (Addgene kit 1000000059). A summary of plasmids used in this study is shown in Supplementary Table 3, and primer sequences are provided in Supplementary Data 3. Chemically competent cells were prepared from the naive and differentiation strains grown in LB without selection, with differentiation strains induced with 30 nM Las-AHL to allow π-protein expression for R6K plasmid replication. 1x and 2x naive strains were transformed with ColE1-KanR-PT7-GFP or ColE1-AmpR-PT7-GFP and plated on LB with 50 μg/mL kanamycin or 100 μg/mL carbenicillin, respectively. Differentiation strains were co-transformed with R6K-CmR-mScarletI and ColE1-KanR-PT7-GFP or ColE1-AmpR-PT7-GFP, recovered in SOC with 30 nM Las-AHL, and plated on LB agar with 34 μg/mL chloramphenicol, 30 nM Las-AHL (3OC12-HSL, Sigma O9139), and 50 μg/mL kanamycin or 100 μg/mL carbenicillin, respectively. Eight independent colonies were picked from each transformation and grown in 300 μL LB in 96 square deep well plates (Southern Labware SKU# 502062) sealed with breathable film (Diversified Biotech BERM-2000) for 8 h at 37 °C. Naive strains were grown in LB with the appropriate antibiotic, and differentiation strains were grown in LB with chlor and carb or kan with 10 nM Las-AHL (see Supplementary Method 1). Following outgrowth, cells were diluted 1:50 into experimental conditions with varying concentrations of IPTG (Gold I2481C) and salicylate (Sigma S3007). Cells were diluted every 8 h for sixteen total growths in constant antibiotic and induction conditions, and sfGFP (485/515 nm), mScarlet (565/595 nm), and OD700 measured by taking 50 μL aliquots of endpoint culture and measuring in 384 well black wall clear bottom microplates (Thermo Scientific 142761) on a Biotek Synergy H1 plate reader using Gen5 software version 3.10.06. An average of two reads for each measurement in each well was used. 1X LB media (ThermoFisher 12795084) sterilized by autoclaving was used for all experiments. Chemically competent 1x and 2x naive strains, and 1x and 2x differentiation strains were transformed with 10 ng of Cole1-AmpR-PT7-GFP or 10 ng Cole1-AmpR with insulated PT7 dnaseI, and all or 10 percent plated on LB carb. Plates with more than 1000 colonies on the 10 percent plate were reported as >104 cfu. For dnaseI expression experiments, 2x split-π differentiation cells were co-transformed with an empty R6K-CmR plasmid and the insulated ColE1-AmpR PT7 dnaseI expression plasmid, recovered in SOC with 30 nM Las-AHL, and plated on LB agar with carb/chlor/30 nM Las-AHL. Three independent colonies were inoculated into 3 mL LB cultures with carb/chlor/10 nM Las-AHL, outgrown for 8 h at 37 °C, and diluted 1:50 into 25 mL media with or without 20 μM salicylate and 10 μM IPTG to induce differentiation and T7 RNAP expression. After 8 h of growth, cultures were diluted 1:50 into the same conditions, and the remaining culture harvested. Wet weight of pellets after washing with PBS was recorded before storing at −20C. JS006 parental strain without the dnaseI expression plasmid was grown similarly in LB without antibiotics and inducers for negative control. The second growth of uninduced cultures was harvested after 8 h as before, and induced cultures allowed to grow for an additional 8 h as minimal growth was observed after the initial growth. Pellets were lysed via sonication of a 33 percent (w/v) cell suspension in 10 mM Tris pH 7.5/2 mM CaCl2 with protease inhibitor (Roche, 11836170001), cleared with centrifugation at 4 C, and supernatant collected and kept on ice before assaying dnaseI activity. Buffers used for assay were as described in Kunitz, though to allow simultaneous measurement of many samples and to avoid problems we observed with background absorbance in crude cell lysate when performing the Kunitz assay, we developed a fluorescence-based assay similar to Vogel and Frantz. Briefly, dnaseI assay buffer was prepared by diluting SYBR Safe (Invitrogen, S33102) 1:1000 into a solution of 100 mM sodium acetate/5 mM magnesium sulfate with 26.3 μg/mL calf thymus DNA (Sigma D1501). Assay plate was prepared by aliquoting 190 μL dnaseI assay buffer into 96 round-well clear bottom plates and equilibrating in the dark at 25 °C. Standards were prepared by adding various amounts of dnaseI (Invitrogen AM2222) to JS006 lysate diluted 1:10 in 0.85 percent NaCl. Samples to assay were diluted 1:10 or 1:50 in 0.85 percent NaCl, and 10 μL of sample or standard pipetted with a multi-channel pipette into triplicate wells immediately before assay. The final amount of DNA per well was 5 μg. After shaking briefly fluorescence (487/528 nM) was measured every minute for 2 h at 25 °C on a Biotek Synergy H1 plate reader using Gen5 software version 3.10.06. Fluorescence fold-change over the course of the two-hour assay was used in fitting a standard curve (Supplementary Fig. 30), and dnaseI activity calculated from appropriate dilutions. Simulations were run in Python using systems of ODEs using Euler’s method with a time step of 0.01 h. This custom ODE solver was used to allow choice of modeling mutations deterministically using first-order rate constants for mutations, or stochastically by drawing from a binomial distribution. Production was modeled as being proportional to the ratio of specific growth rate (actual growth rate after accounting for effect due to carrying capacity) to maximum growth rate for the specific cell type, and production rate was 1 arb. unit per cell per hour for all simulations regardless of burden. For terminal differentiation, the number of divisions allowed (n) is explicitly modeled. Immediately after differentiation, a cell has divided i = 0 times. Division of a cell for which i = 0 generates two cells with i = 1. Cells for which i = n, instead of dividing, then die at the same rate. Simulations in Fig. 1 and Supplementary Fig. 1 were modeled deterministically with constant dilution and carrying capacity limited growth for 1000 h of simulated time. For Fig. 3 and Supplementary Fig. 17–26, simulations were of batch growths with 1:50 dilutions every 8 h for 20 total growths. Growth, differentiation, and production were modeled deterministically, while all mutations were modeled stochastically by drawing from a binomial distribution. 8 replicate simulations were run for each condition. Naive 2x indicates that a cell with two functional cassettes (subscript PP) has the production rate (β) and growth rate (μ) of a producer cell as indicated by the burden level being modeled (). The growth and production rates of a cell with one functional and one non-functional cassette (subscript NP) for naive 2x are given as (1) and (2), respectively. Naive 2x* indicates that a cell with one functional cassette has the production rate and growth rate of a producer cell (; ). The growth and production rates of a cell with two functional cassettes are given as (3) and (4), respectively. For full description of model implementation, see Supplementary Note 1. Glycerol stocks from plate generation 16 were saved in 384 well plates in 20% glycerol w/v (1:3 dilution of culture with 30% w/v glycerol) and stored at −80 °C. Glycerol stocks were thawed, and 10-fold serial dilutions of three independent experimental replicates for each strain/experimental condition being assessed were prepared and plated. For 1x and 2x naive, dilutions from replicates from the high burden condition (LB Kan/50 μΜ IPTG) were plated on LB + kanamycin. For 1x and 2x differentiation and terminal differentiation, dilutions from replicates from the high differentiation rate/high burden condition (LB Kan/10 nM Las/50 μM IPTG/30 μM salicylate, +/- chloramphenicol) were plated on LB + Kan/10 nM Las/50 μM salicylate, + Kan/Chlor/10 nM Las, and + Kan/Chlor/10 nM Las/50 μM salicylate. Plates were imaged on a ChemiDoc MP imager for GFP fluorescence (Alexa 488 channel, 532/28 Filter, 0.04 s exposure), mScarlet fluorescence (Alexa 546 channel, 602/50 Filter, 0.3 s exposure), and for visualization of all colonies (UV Trans Illumination, 590/110 Filter. 0.488 s exposure). Multichannel RGB images were created in Image Lab (version 6.1) with equivalently transformed images (Alexa 488: high 30000, low 0; Alexa 546: high 30000, low 0; UV Trans: high 60000, low 30000, inverted), and colonies manually counted using an application (COUNT THINGS) on an iPad pro. For each strain/condition, 6 colonies were analyzed with nanopore sequencing. Colony PCR using barcoded primers (Supplementary Data 3) was performed for all regions of interest with PrimeSTAR HS DNA Polymerase (1 min 98 °C; 30 cycles of 10 s 98 °C, 5 s 55 °C or 60 °C, 1 min/kb 72 °C; 5 min 72 °C final extension; 20 μL rxns). Amplified samples were assessed with electrophoresis (5 μL), and equivalently sized products pooled and gel extracted. Equimolar amounts per amplicon of purified products were pooled and prepared for Nanopore sequencing (Flongle flow cell on MinION) with LSK110 according to the manufacturers protocol, and super accuracy basecalling was used in MinKNOW. The portion of the ColE1 plasmid encoding PT7 GFP and the P21 (T) integration locus were amplified for all samples; the HK022 (H) locus was amplified for 2x naive and 2x differentiation/terminal differentiation; and the primary and secondary phage 186 (O) sites were amplified for all differentiation/terminal differentiation strains. Sequences were demultiplexed and analyzed using Maple (https://github.com/gordonrix/maple) using expected reference sequences (WT amplicons for ColE1, naive T and H integrations, and primary and secondary O integrations; and WT, and recombined (correctly excised, as well as inverted) sequences for T and H differentiation integrations. As this pipeline failed for a subset of differentiation cassettes with unexpected deletion/recombination mutations, basecalled reads were annotated with barcode primer sequences and extracted with in silico PCR in Geneious (Geneious Prime 2022.1), and exported as fastq.gz files for de novo assembly using Flye 2.9 (https://github.com/fenderglass/Flye). This de novo assembly was used to analyze all T and H locus amplicons for differentiation/terminal differentiation by alignment to WT, excised, and inverted reference sequences. Further information on research design is available in the Nature Research Reporting Summary linked to this article. Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary

PMC9649633

Can Lu,Yuan Cai,Wei Liu,Bi Peng,Qiuju Liang,Yuanliang Yan,Desheng Liang,Zhijie Xu

Aberrant expression of KDM1A inhibits ferroptosis of lung cancer cells through up-regulating c-Myc

10-11-2022

Cancer,Cell death

Ferroptosis is a cell death process caused by metabolic dysfunction with the feature of aberrant iron accumulation. Emerging studies have identified that ferroptosis is an important biological function involving in the tumorigenesis, and targeting ferroptosis could provide promising therapeutic targets for lung cancer. However, such therapeutic strategies show limited therapeutic effect owing to drug resistance and other unknown underlying mechanisms. In this study, lysine-specific demethylase 1 (LSD1/KDM1A) was found to be significantly upregulated in lung cancer cells and tissues. The patients with KDM1A downregulation displayed the good prognosis. Using gene set enrichment analysis (GSEA), we demonstrated that KDM1A-associated genes might participate in the regulation of cell ferroptosis and Myc signaling in lung cancer. Knockdown of KDM1A inhibited the level of c-Myc and increased the concentration of malondialdehyde (MDA) and irons in human lung cancer cells H1299 and A549. Downregulation of c-Myc could facilitate KDM1A knockdown-mediated ferroptosis. Our study has elucidated the effect of KDM1A/c-Myc regulatory axis in the ferroptosis resistance of lung cancer cells.

Aberrant expression of KDM1A inhibits ferroptosis of lung cancer cells through up-regulating c-Myc Ferroptosis is a cell death process caused by metabolic dysfunction with the feature of aberrant iron accumulation. Emerging studies have identified that ferroptosis is an important biological function involving in the tumorigenesis, and targeting ferroptosis could provide promising therapeutic targets for lung cancer. However, such therapeutic strategies show limited therapeutic effect owing to drug resistance and other unknown underlying mechanisms. In this study, lysine-specific demethylase 1 (LSD1/KDM1A) was found to be significantly upregulated in lung cancer cells and tissues. The patients with KDM1A downregulation displayed the good prognosis. Using gene set enrichment analysis (GSEA), we demonstrated that KDM1A-associated genes might participate in the regulation of cell ferroptosis and Myc signaling in lung cancer. Knockdown of KDM1A inhibited the level of c-Myc and increased the concentration of malondialdehyde (MDA) and irons in human lung cancer cells H1299 and A549. Downregulation of c-Myc could facilitate KDM1A knockdown-mediated ferroptosis. Our study has elucidated the effect of KDM1A/c-Myc regulatory axis in the ferroptosis resistance of lung cancer cells. Lung cancer has been proved to be the most common cause for cancer-associated death. Nowadays, therapy resistance still occurs and leads to the recurrence of lung cancer patients. Recent studies showed that the cisplatin-resistant cancer cells exhibit less sensitive to ferroptosis inducers. Moreover, the clinical use of ferroptosis induction can improve the therapeutic efficacy in cancer patients. However, the underlying mechanisms and biological functions of ferroptosis in lung cancer pathogenesis and therapeutic response remain unclear. Ferroptosis, a novel type of regulatory cell death, triggered by depletion of glutathione and lipid peroxidation. Emerging evidences have suggested that dysregulation of ferroptosis is closely related to tumorigenesis and treatment, highlighting the regulation of ferroptosis as a promising anticancer therapeutic strategy. Previous studies have suggested the potential roles of oncogenic c-Myc in ferroptosis. In lung cancer cells, high levels of c-Myc significantly inhibits ferroptosis through directly activating lymphoid-specific helicase. Clarifying the c-Myc signaling networks and the associated factors underlying ferroptosis would be critical to effectively sensitize cancer cells to the ferroptosis‐based therapeutic strategies. Emerging studies have shown that epigenetic modifications could regulate the expression of signaling molecules involved in the cell death. As an important epigenetic enzyme, lysine specific demethylase 1 (LSD1/KDM1A) plays a key functional role in mediating cancer cell death. Aberrant KDM1A signaling participates in a range of biological processes, including cell proliferation, epithelial-to-mesenchymal transition (EMT) and malignant transformation. Emerging reports have demonstrated that KDM1A is commonly dysregulated in a variety of cancers, suggesting KDM1A as a probable biomarker for cancer development and treatment. However, the underlying mechanisms of KDM1A on the regulation of cell ferroptosis is relatively limited and remains to be explained. Here, we evaluated the potential roles of KDM1A-c-Myc axis in lung cancer. Based on several public bioinformatic databases, we found that KDM1A was over-expressed in lung cancer tissues. And the patients with upregulated KDM1A displayed unfavorable prognosis. Silencing of KDM1A significantly impeded the cell growth and increased the sensitivity to erastin-induced ferroptosis in A549 and H1299 cells. We also found that blocking KDM1A-c-Myc axis could significantly activate cell ferroptosis. Our data suggested that KDM1A confers resistance to ferroptosis in lung cancer cells through upregulating c-Myc signaling. UALCAN (http://ualcan.path.uab.edu/) and TNMplot (https://tnmplot.com/analysis/), two comprehensive and user-friendly databases, were used to analyze the mRNA and protein levels of KDM1A in cancers. Two survival databases, PrognoScan and Kaplan–Meier plotter, were used to analyze the effect of KDM1A expression in patients’ prognosis, including relapse free survival (RFS), overall survival (OS) and first progression (FP). The expression levels of KDM1A in lung cancer tissues were further analyzed using two datasets from Gene Expression Omnibus (GEO), GSE13213 and GSE31210. The database, cBioPortal, was used to analyzed the co-expressed genes of KDM1A in a lung adenocarcinoma dataset (TCGA, PanCancer Atlas). And the GSEA pathway analysis was performed using Xiantao Xueshu (https://www.xiantao.love/products). Human lung cancer cell lines H1299, A549, H157 and H358 were obtained from the Cancer Research Institute, Central South University, China. Cells were cultured in 37 °C incubator with 5% CO2. The cell culture medium for A549 was RPMI-1640 (Procell, Cat#PM150110P) with 10% FBS (Gbico, Cat#10099141C). And the cell culture mediums for H1299, H157 and H358 were Dulbecco’s modified Eagle’s medium (DMEM) (Procell, Cat#PM150210P) with 10% FBS. H1299-shKDM1A and A549-shKDM1A cells were cultured with respective medium supplemented with 1 μg/ml puromycin (Beyotime, Cat#ST551). The exposing concentrations of erastin (Sigma-Aldrich, Cat#E7781) and ferrostatin-1 (APExBIO, Cat#A4371) in lung cancer cells were 5 μM and 10 μM, respectively. 293 T cells were seeded in 6 mm plates and co-transfected with lentivirus package plasmid Pax2 and VSVG the next day. After 48 h incubation at 37 °C, we harvested the virus-containing supernatants. For cell infection, the mixture of virus liquid and polybrene (5 μg/ml) was added into cell culture medium. The sequences for KDM1A shRNAs (shKDM1As) were obtained as previously described and inserted into the pLKO.1 lentiviral vectors. shKDM1A: GATCCCCAGGAAGGCTCT TCTAGCAATATTCAAGAGATATTGCTAGAAGAGCCTTCCTTTTTTC, TCGAGA AAAAAGGAAGGCTCTTCTAGCAATATCTCTTGAATATTGCTAGAAGAGCCTTCCTGGG; shKDM1A-2: GATCCCCGGAGCTCCTGATTTGACAAAGTTCAAGA GACTTTGTCAAATCAGGAGCTCCTTTTTC, TCGAGAAAAAGGAGCTCCTGA TTTGACAAAGTCTCTTGAACTTTGTCAAATCAGGAGCTCCGGG; shCtrl:GATCCCCAATTGCCACAACAGGGTCGTGTTCAAGAGA, CACGACCCTGCCGTGGCAATTTTTTTC, TCGAGAAAAAAATTGCCA, CAACAGGGTCGTGTCTCTTGAACACGACCCTGCCGTGGCAATTGGG. We purchased c-Myc siRNA from Genepharma (China) according to the previous report. We conducted the transfection of siRNA into cancer cells according to the manufacturer’s introductions of lipofectamine 3000 (Invitrogen, United States). In brief, 1 × 106 cells were seeded in 60 mm plates and performed transfection the next day. After 48 h incubation, the cells were further collected for subsequent experiments. After extraction using Trizol (Invitrogen, Cat#15596018), the total RNA was reverse-transcribed with the reverse transcription kit, PrimeScript 1st strand cDNA synthesis kit (Takara, Cat#6210A). Next, we used real-time polymerase chain reaction (RT-PCR) to analyze the transcriptional levels of KDM1A, c-Myc and HMOX1. The forward and reverse primer sequences are as follows: β-actin: 5′-CATGTACGTTGCTATCCAGGC-3′ and 5′-CTCCTTAATGTCACGCACGAT-3′; KDM1A: 5′-TGACCGGATGACTTCTCAAGA-3′ and 5′-GTTGGAGAGTAGCCTC AAATGTC-3′; c-Myc: 5′-GGCTCCTGGCAAAAGGTCA-3′ and 5′-CTGCGTAGTT GTGCTGATGT-3′. HMOX1: 5′-AAGACTGCGTTCCTGCTCAAC-3′ and 5′-AAAGCCCTACAGCAACTGTCG-3′. We used the 2−∆∆CT method to calculate their relative expression levels. After extraction using lysis buffer (Thermo Scientific, United States), the total protein were resolved on the SDS-PAGE and transferred onto a nitrocellulose membrane (Millipore, United States). Next, the nitrocellulose membrane was blocked in 5% skimmed milk, and incubated with the indicated primary antibodies at 4 °C overnight. The following antibodies were used: anti-KDM1A (Abcam, United States, Cat#ab17721), anti-c-Myc (Santa Cruz, United States,Cat# sc-40), anti-HMOX1 (Proteintech, United States, Cat# 10701-1-AP) and anti-β-actin (Santa Cruz, United States, Cat# sc-69879). The protein levels were determined using the chemiluminescence reagent (Millipore, USA). According to the molecular weight, the nitrocellulose membrane was cut prior to hybridization with antibodies and visualized using the ChemiDoc XRS system (Bio-Rad, Berkeley). We performed the cell proliferation assay according to the manufacturer’s introductions of MTS assay kit (B34304, Bimake, United States). In brief, 1 × 103 cells were seeded in 96-well plate. After incubated with erastin (5 μM) orRSL3 (1 μM),ferrostatin-1 (10 μM) for 72 h and MTS solution for 1 h, the optical density of cells was detected at 450 nm with a spectrometer (PerkinElmer, United States). We detected the concentration of cellular irons according to the manufacturer’s introductions of iron assay kit (Abcam, United States). After treated with ferroptosis inducer erastin (5 μM) or RSL3 (1 μM) for 24 h, cells were rapidly mixed with iron assay buffer. We then removed the insoluble material, and added the iron assay buffer and iron probe into the reaction mixture. At last, the spectrometric absorbance was detected at the wavelength of 593 nm. We detected the concentration of cellular malondialdehyde (MDA) according to the manufacturer’s introductions of lipid peroxidation assay kit (Sigma-Alorich, United States). After treated with ferroptosis inducer erastin (5 μM) or RSL3 (1 μM) for 24 h, cells were lysed on ice by ultrasonication. We then removed the insoluble material, and added thiobarbituric acid (TBA) solution into the samples. At last, the spectrometric absorbance was detected at the wavelength of 532 nm. All data represented in graphs were shown as mean ± standard deviation (SD). Student’s t-test and one way analysis of variance (ANOVA) were used for the difference comparisons or multivariate analysis. The difference has a statistic meaning (*p < 0.05, **p < 0.01 and ***p < 0.001). Pan-cancer analysis from UALCAN database revealed the upregulated expression levels of KDM1A in majority of cancers, including lung cancer (Fig. 1A,B, Table S1). We also utilized the TNMplot database to demonstrated that KDM1A was overexpressed in lung cancer tissues from RNA-Seq data (Fig. 1C) and Gene-chip data (Fig. 1D). In addition, we detected the expression profiles of KDM1A in several lung cancer cells. We found the significantly over-expressed KDM1A in A549 and H1299 cells (Fig. 1E,F). Thus, A549 and H1299 cells were used for subsequent experiments. We then used two survival databases, PrognoScan and Kaplan–Meier plotter, to analyze the effect of KDM1A levels on patients’ prognosis. Of note, the patients with high levels of KDM1A displayed poor OS (Fig. 2A) and RFS (Fig. 2B). Moreover, the patients with low levels of KDM1A displayed slightly favorable FP (Fig. 2C). These findings above-mentioned suggested that the expression of KDM1A was associated with the poor clinical outcomes in patients with lung cancer. The analysis of gene co-expression network is an outstanding approach for identifying the co-expression patterns of candidate genes in different phenotypes, and has been widely used for gene function annotation. The co-expressed genes of KDM1A (Table S2) were downloaded from a lung adenocarcinoma dataset, TCGA Pan-Cancer Atlas, to further evaluate the underlying mechanisms regulated by KDM1A co-expression molecules. The GSEA analysis results showed that KDM1A related genes might involve in the cell ferroptosis process (Fig. 3A). Then, we investigated whether KDM1A mediates ferroptosis of lung cancer cells. Firstly, we established H1299 and A549 stable cells with KDM1A gene knockdown. As shown in colony formation assays (Fig. 3B,C), cell growth was significantly decreased in H1299 and A549 cells with KDM1A knockdown. To further understand the roles of KDM1A in cell ferroptosis, we treated the stable KDM1A knockdown A549 and H1299 cell lines with erastin and RSL3, two ferroptosis inducers, and ferrostatin-1 (Fer-1), a ferroptosis inhibitor. The cell viability assays showed that silencing of KDM1A significantly enhanced the lethality of erastin and RSL3 in H1299 and A549 cells. However, Fer-1 treatment could significantly weaken the lethality induced by erastin or RSL3 (Fig. 3D,E, Fig. S3G,H). Iron accumulation and lipid peroxidation have been proved to be the two primary biochemical characteristics of ferroptosis process. And MDA is one of the end products from lipid peroxidation. Accordingly, erastin treatment increased the concentration of intracellular Fe2+ and MDA levels in KDM1A knockdown H1299 and A549 cells (Fig. 3F–I). Inversely, we obtained the downregulated trend of intracellular Fe2+ and MDA levels in H1299 KDM1A overexpression cells induced by erastin or RSL3 (Fig. S3C–F). Furthermore, after KDM1A knockdown, RT-PCR was used to analyze the expression of several ferroptosis markers. As shown in Fig. S1 and Fig. 3J, we found that ferroptosis-associated molecule, HMOX1, was significantly up-regulated in KDM1A knockdown lung cancer cells H1299 and A549. Western blot also demonstrated inhibition of KDM1A could increase the HMOX1 protein levels (Fig. 3K). In addition, GEPIA2 database showed that KDM1A expression was negatively correlated with HMOX1 expression in lung cancer (Fig. S2A). These data suggested that inhibition of KDM1A could improve the ferroptosis sensitivity of lung cancer cells. The GSEA pathway analysis indicated that KDM1A related genes may participate in Myc activation pathway and Myc pathway (Fig. 4A,B). Then, we investigated whether KDM1A mediates Myc signaling in lung cancer. GEPIA2 database showed that KDM1A expression was positively correlated with Myc expression in lung cancer (Fig. S2B). Depletion of endogenous KDM1A repressed the c-Myc expression levels in H1299 and A549 cells (Fig. 4C,D). Meanwhile, overexpression of KDM1A reversely improved the c-Myc expression levels in H1299 and A549 cells (Fig. S3A,B). Combined knockdown of KDM1A and c-Myc displayed the synergetic inhibitory effect on c-Myc expression (Fig. 4E,F). Additionally, we confirmed the interaction between KDM1A and c-Myc in 293 T cells (Fig. S4). These data collectively suggest that c-Myc might be regulated by KDM1A in lung cancers. Recent studies have displayed the inhibitory effect of c-Myc on ferroptosis. Next, we wanted to elucidate the underlying roles of the KDM1A-c-Myc axis in ferroptosis of lung cancer cells. Combined knockdown of KDM1A and c-Myc synergistically reduced the cell proliferation rates in H1299 and A549 cells (Fig. 5A–D). Moreover, the intracellular Fe2+ and MDA levels were significantly increased upon KDM1A and c-Myc knockdown (Fig. 5E–H). Simultaneously, ectopic expression of c-Myc in KDM1A-depleted cells (Fig. S5A,B) treated with erastin or RSL3 completely reversed the decrease of cell growth (Figs. S5C–F, S6E,F) and increase of intracellular Fe2+ and MDA levels (Fig. S6A–D). Previous reports have identified HMOX1 as a ferroptosis inducer. Accordingly, c-Myc knockdown further increased the expression level of HOMX1 upon KDM1A knockdown both in H1299 and A549 cells (Fig. 5I,J). These data together indicate that KDM1A protected lung cancer cells against ferroptosis by maintaining c-Myc level. Nowadays, the 5-year survival rate of lung cancer patients is still unsatisfactory. Exploring the potential therapeutic and prognostic biomarkers are thus urgently needed. Although ferroptosis cell death has been proposed to provide the new hopes for cancer management, the precise roles of ferroptosis in lung cancer development and progression have not been fully elucidated. Previous studies have confirmed the highly-expressed KDM1A in several cancers. Accordingly, our study displayed the upregulated KDM1A in lung cancer H1299 and A549 cells. Knockdown of KDM1A obviously suppressed the cell growth and induced the cell death. The aberrantly expressed c-Myc, a member of MYC gene family, has been proved to participate in the regulation of multiple biological functions in human cancers. It has been reported that c-Myc could regulate the transcription of targeted genes by binding some histone components. Amente et al. showed that c-Myc could directedly bind and recruit KDM1A to the E-box chromatin, promoting the transcriptional activity of target genes. In our study, GSEA pathway analysis indicated that KDM1A might participate in the activation of Myc signaling. And we demonstrated that knockdown of KDM1A could significantly inhibit c-Myc expression levels in lung cancer cells H1299 and A549. Ferroptosis, a reactive oxygen species-dependent cell death, is characterized with lipid peroxidation and iron accumulation. Regarding iron homeostasis regulation, c-Myc has been reported to suppress the expression of iron regulatory proteins, resulting in the accumulation of intracellular iron pool. Accordingly, we noted that c-Myc knockdown as well as KDM1A knockdown significantly increased the expression levels of ferroptosis marker HMOX1 in H1299 and A549 cells. Moreover, knockdown of KDM1A increased ferroptosis of lung cancer cells through downregulating c-Myc expression. KDM1A knockdown or c-Myc knockdown significantly enhanced the lethality induced by erastin in H1299 and A549 cells. Therefore, our findings collectively demonstrated the important roles of KDM1A-c-Myc signaling in the regulation of cell ferroptosis, and revealed a potential therapeutic strategy against lung cancer. The histone demethylase lysine-specific demethylase 1 (LSD1/KDM1A) can regulate the gene transcription through demethylating the histone H3 lysine 4 (H3K4). In previous studies, KDM1A down-regulates the antagonist of the canonical Wnt pathway, APC, by demethylating H3K4me1/2 of APC2 promotor, promoting the progression of thyroid cancer. KDM1A reinforces the immunosuppression in hepatocellular carcinoma through demethylating MEF2D and activating PD-L1. KDM1A exerts anti-cancer effect in bladder cancer through demethylating MMP9. In addition, overexpression of KDM1A could effectively protect cell against preterm death. In our study, we found that KDM1A knockdown significantly repressed the expression of c-Myc, leading to the increased ferroptosis sensitivity of lung cancer cells. Therefore, KDM1A might mechanistically mediate the demethylation of c-Myc in lung cancer cells. In summary, our work revealed the functional roles of KDM1A-c-Myc axis in the regulation of ferroptosis in lung cancer cells. KDM1A depletion could significantly increase the cellular Fe2+ concentration and MDA levels through downregulating c-Myc expression, resulting in cell growth inhibition. Therefore, targeting KDM1A could be a potential ferroptosis-based treatment strategy for lung cancer patients. Supplementary Legends.Supplementary Figure S1.Supplementary Figure S2.Supplementary Figure S3.Supplementary Figure S4.Supplementary Figure S5.Supplementary Figure S6.Supplementary Figure 6.Supplementary Figure 7.Supplementary Figure 8.Supplementary Figure 9.Supplementary Figure 10.Supplementary Figure 11.Supplementary Figure 12.Supplementary Table S1.Supplementary Table S2.

PMC9649634

Kaijia Tang,Danli Kong,Yuan Peng,Jingyi Guo,Yadi Zhong,Haibing Yu,Zhenhua Mai,Yanling Chen,Yingjian Chen,Tianqi Cui,Siwei Duan,Tianyao Li,Naihua Liu,Dong Zhang,Yuanlin Ding,Jiawen Huang

Ginsenoside Rc attenuates DSS-induced ulcerative colitis, intestinal inflammatory, and barrier function by activating the farnesoid X receptor 10.3389/fphar.2022.1000444

28-10-2022

ulcerative colitis,inflammatory bowel disease,inflammation,ginsenoside Rc,intestinal barriers

Objectives: Farnesoid X receptor (FXR) activation is involved in ameliorating inflammatory bowel disease (IBD), such as ulcerative colitis (UC), and inflammatory regulation may be involved in its mechanism. Ginsenoside Rc (Rc) is a major component of Panax ginseng, and it plays an excellent role in the anti-inflammatory processes. Our aim is to explore the alleviative effect of Rc on dextran sulfate sodium (DSS)-induced inflammation and deficiencies in barrier function based on FXR signaling. Materials and Methods: In vitro, we treated human intestinal epithelial cell lines (LS174T) with LPS to explore the anti-inflammatory effect of Rc supplementation. In vivo, a DSS-induced IBD mice model was established, and the changes in inflammatory and barrier function in colons after Rc treatment were measured using the disease activity index (DAI), hematoxylin and eosin (H&E) staining, immunofluorescence, ELISA, and qPCR. Molecular docking analysis, luciferase reporter gene assay, and qPCR were then used to analyze the binding targets of Rc. DSS-induced FXR-knockout (FXR−/-) mice were used for further validation. Results: Rc significantly recovered the abnormal levels of inflammation indexes (TNF-α, IL-6, IL-1β, and NF-KB) induced by LPS in LS174T. DSS-induced C57BL/6 mice exhibited a significantly decreased body weight and elevated DAI, as well as a decrease in colon weight and length. Increased inflammatory markers (TNF-α, IL-6, IL-1β, ICAM1, NF-KB, F4/80, and CD11b displayed an increased expression) and damaged barrier function (Claudin-1, occludin, and ZO-1 displayed a decreased expression) were observed in DSS-induced C57BL/6 mice. Nevertheless, supplementation with Rc mitigated the increased inflammatory and damaged barrier function associated with DSS. Further evaluation revealed an activation of FXR signaling in Rc-treated LS174T, with FXR, BSEP, and SHP found to be upregulated. Furthermore, molecular docking indicated that there is a clear interaction between Rc and FXR, while Rc activated transcriptional expression of FXR in luciferase reporter gene assay. However, these reversal abilities of Rc were not observed in DSS-induced FXR−/- mice. Conclusion: Our findings suggest that Rc may ameliorate inflammation and barrier function in the intestine, which in turn leads to the attenuation of DSS-induced UC, in which Rc may potentially activate FXR signaling to protect the intestines from DSS-induced injury.

Ginsenoside Rc attenuates DSS-induced ulcerative colitis, intestinal inflammatory, and barrier function by activating the farnesoid X receptor 10.3389/fphar.2022.1000444 Objectives: Farnesoid X receptor (FXR) activation is involved in ameliorating inflammatory bowel disease (IBD), such as ulcerative colitis (UC), and inflammatory regulation may be involved in its mechanism. Ginsenoside Rc (Rc) is a major component of Panax ginseng, and it plays an excellent role in the anti-inflammatory processes. Our aim is to explore the alleviative effect of Rc on dextran sulfate sodium (DSS)-induced inflammation and deficiencies in barrier function based on FXR signaling. Materials and Methods: In vitro, we treated human intestinal epithelial cell lines (LS174T) with LPS to explore the anti-inflammatory effect of Rc supplementation. In vivo, a DSS-induced IBD mice model was established, and the changes in inflammatory and barrier function in colons after Rc treatment were measured using the disease activity index (DAI), hematoxylin and eosin (H&E) staining, immunofluorescence, ELISA, and qPCR. Molecular docking analysis, luciferase reporter gene assay, and qPCR were then used to analyze the binding targets of Rc. DSS-induced FXR-knockout (FXR−/-) mice were used for further validation. Results: Rc significantly recovered the abnormal levels of inflammation indexes (TNF-α, IL-6, IL-1β, and NF-KB) induced by LPS in LS174T. DSS-induced C57BL/6 mice exhibited a significantly decreased body weight and elevated DAI, as well as a decrease in colon weight and length. Increased inflammatory markers (TNF-α, IL-6, IL-1β, ICAM1, NF-KB, F4/80, and CD11b displayed an increased expression) and damaged barrier function (Claudin-1, occludin, and ZO-1 displayed a decreased expression) were observed in DSS-induced C57BL/6 mice. Nevertheless, supplementation with Rc mitigated the increased inflammatory and damaged barrier function associated with DSS. Further evaluation revealed an activation of FXR signaling in Rc-treated LS174T, with FXR, BSEP, and SHP found to be upregulated. Furthermore, molecular docking indicated that there is a clear interaction between Rc and FXR, while Rc activated transcriptional expression of FXR in luciferase reporter gene assay. However, these reversal abilities of Rc were not observed in DSS-induced FXR−/- mice. Conclusion: Our findings suggest that Rc may ameliorate inflammation and barrier function in the intestine, which in turn leads to the attenuation of DSS-induced UC, in which Rc may potentially activate FXR signaling to protect the intestines from DSS-induced injury. Ulcerative colitis (UC) is a type of inflammatory bowel disease (IBD) that is often characterized by diarrhea, rectal bleeding, abdominal pain, and so on. It is also described as a chronic idiopathic inflammatory disease (Sands, 2004). In the past few decades, high morbidity and disability caused by IBD have resulted in a high cost of treatment and care management, which requires effective prevention strategies (Higgins. et al., 2003; Sterne et al., 2011). Many factors can increase the risk of IBD, such as smoking, urban living, antibiotic exposure, and so on (Piovani et al., 2019). Although the exact pathogenesis of IBD is unknown, the current view holds that it is caused by disrupted homeostasis of the mucosal immune system and impaired intestinal epithelial barriers (Kiesler et al., 2015). Clinically, aminosalicylates, glucocorticoids, and immunosuppressants are commonly used in IBD treatment. However, these drugs can have significant disadvantages, such as unsatisfactory efficacy and many adverse reactions (Lim et al., 2016). Thus, immunomodulators have received increasing attention as potential therapeutics for the treatment of IBD. Cytokines and inflammatory mediators are the products of IBD inflammatory reactions, which often lead to damage to intestinal epithelial cells, breaking of cell junctions, and finally result in dyshomeostasis of the intestinal flora (Bischoff et al., 2014; Drury et al., 2021). The Farnesoid X receptor (FXR), which is known as a nuclear bile acid receptor, can not only modulate the metabolic balance of bile acid, lipids, and glucose (Sun et al., 2021) but may also be involved in regulating the inflammatory response (Anderson and Gayer, 2021). FXR activation controls the transcriptional induction of the expression of small heterodimer protein (SHP) (Wang et al., 2002), as well as bile salt export pump (BSEP) proteins (Ananthanarayanan et al., 2001). The recruitment of NF-KB can be directly prevented by SHP, which represses several cytokines, including IL-1β and TNF-α (Yang et al., 2016). Both of these can be considered to be FXR targets, and their expression levels have been inversely correlated with inflammation levels (Huang et al., 2018; Cariello et al., 2021). Interestingly, although FXR can modulate inflammatory signaling, acute inflammatory cytokines can also influence FXR expression. For instance, TNFα was able to decrease binding activity with an FXR response element in DNA and resulted in FXR downregulation (Kim et al., 2003). Recently, an increasing number of studies have focused on pharmacologic FXR agonists and FXR has received enthusiastic attention as a developing therapeutic target (Massafra et al., 2018; Badman et al., 2020). Progressive increases in intestinal permeability have been suggested to be involved in the pathogenesis of IBD (Weber and Turner, 2007; Xavier and Podolsky, 2007). Tight junctions (TJs) are the most important components of intestinal barriers and can be affected by many factors (e.g., immune cells). For example, TJs can be interrupted by TNFα, which downregulates claudin and occludin. This results in increased intestinal permeability (Al-Sadi et al., 2013). It is known that FXR activation can not only inhibit pro-inflammatory cytokine production but can also reduce goblet cell loss and inhibit epithelial permeability (Gadaleta et al., 2011). FXR agonists have been shown to be potential candidates for IBD treatment (Stojancevic et al., 2012). Ginsenoside Rc (Rc) is a major anti-inflammatory component of Panax ginseng, which exhibits anti-oxidative and anti-inflammatory activities through different mechanisms (Yu et al., 2016). A previous study has reported that Rc can reduce inflammatory levels and repair cellular damage in cardiomyocytes (Huang et al., 2021). However, it is unknown whether Rc could ameliorate IBD by reducing the inflammatory response. Furthermore, the association of Rc and FXR has not yet been reported. Here, we hypothesize that Rc may ameliorate IBD symptoms by reducing inflammation and its mechanism may involve the activation of the FXR signaling pathway as an FXR agonist. In this study, we assess the anti-inflammation activity and intestinal damage repair ability of Rc on a dextran sulfate sodium (DSS)-induced IBD mice model. Furthermore, experimental FXR−/−and wild-type (WT) mice were used to assess the key role that FXR plays in Rc treatment of IBD. We aimed to explore the anti-inflammatory effects and the functional mechanism of Rc in IBD treatment. Human intestinal epithelial cell lines LS174T (ATCC) were cultured in fresh DMEM with 10% FBS. Rc (purity >98%, HPLC) was purchased from Shanghai Yuanye Bio-Technology Co., Ltd. (Shanghai, China). HepG2 cells (ATCC) were also maintained in fresh DMEM with 10% FBS. To investigate the cell viability of Rc, LS174T were treated with different concentrations of Rc (0, 6.25, 12.5, 25, 50, 100, 200, and 400 µM). Meanwhile, 24 hours later, cell viability was evaluated via the CCK-8 assay. Moreover, to investigate whether Rc can reduce inflammation induced by LPS, cells were pretreated with Rc for 48 h, incubated with LPS (2000 ng/ml) for 24 h, and collected for further analysis. All of the animal experimental studies were approved by the Animal Ethics Committee of Guangzhou University of Chinese Medicine. In total, 50 male C57BL/6 mice (6 weeks old, 20–25 g) were purchased from the Model Animal Research Center of Guangzhou University of Chinese Medicine (Certificate: SCXK 2018-0034; Guangzhou, China). They were housed in an SPF room (25°C, 12 h day/night cycle, free access to chow and water). After 2 weeks, the mice were randomly divided into five groups (Saline, DSS, DSS Rc 5 mg/kg, DSS Rc10 mg/kg, and DSS Rc 20 mg/kg), with 10 mice in each group. The mice were then administered different concentrations of Rc every day throughout the experimental period (11 days). On the third day, the mice were given a 4% DSS solution (w/v) until the end of the experimental period. FXR-knockout (FXR−/-) mice were kindly provided by Changhui Liu, and were generated and used as previously purchased from the Jackson Laboratory (Bar Harbor, ME, United States) (Liu et al., 2020). The genotyping for FXR identification is shown in Supplementary Figure S1. Similarly, 18 FXR−/- mice were divided into three groups (Saline, DSS, and DSS Rc 20 mg/kg), and were treated in the same way as those mentioned above. All of the blood and tissues were collected from the mice after anesthesia on the last day. The disease activity index (DAI) was measured following the method of a previous study (DAI score = weight loss (%) + stool consistency + rectal bleeding) (Detel et al., 2012). The weight loss, stool consistency, and rectal bleeding of each mouse were observed daily to evaluate the symptoms of colitis. The entire colon was weighed and measured at the end of the experiment. Parts of colons were placed in 4% paraformaldehyde and embedded and cut for H&E staining. Images were collected from a light microscope. The remaining parts were stored for further analysis. Immunofluorescence was performed according to current protocols (Liu et al., 2018). In brief, colon sections were incubated with F4/80, CD11b, and ZO-1 (Affinity, United States) at 4°C overnight. After washing them three times in PBS, sections were incubated with secondary antibodies (Abclonal, Wuhan, China) for 40 min at room temperature. After washing them five times in PBS, images were observed using a fluorescence microscope (Nikon, Japan). LS174T cells were treated with Rc (25 μM) for 24 h after treatment with adnexal LPS (2000 ng/ml), washed twice with PBS, blocked with 4% paraformaldehyde, incubated with NF-KB primary antibody (Abcam, ab16502) at 4°C overnight, followed by incubation with secondary antibody (Abmart, Alexa Fluor 488) for 1 h, washed three times with PBS, and blocked with DAPI blocker for microscopic observation. Serum was collected for TNF-α, IL-1β, and IL-6 measurements. In this experiment, IL-1β and IL-6 were detected by the corresponding ELISA kits from Abclonal (Wuhan, China), while TNF-α was detected by ELISA kits from Ruixinbio (Quanzhou, China), following the manufacturer’s instructions. The TRIzol reagent was used for the total mRNA extraction of mouse colon tissue samples. A high-capacity cDNA reverse-transcription kit (Abclonal, Wuhan, China) was used for reverse transcription. Meanwhile, cDNA was subjected to quantitative PCR (qPCR) analysis with the PowerUpTM SYBRTM Green Master Mix (Abclonal, Wuhan, China). The expression levels of all the genes were standardized with β-actin and the specific primer sequences are shown in Table 1. The chemical composition collection method was as follows. The ChemDraw 14.0 software was used to draw the 2D structure of small molecule ligands (Rc), and then ChemDraw3D was used to transform the 2D structure into a 3D structure and save it as an MOL2 file. The 3D structure was imported into Discovery Studio. The Prepare Ligands module in Molecules was then used to process small molecules. The energy of small molecules was minimized and a CHARMm force field was used to obtain the prepared small molecules and save them in the MOL2 format. The structural acquisition and preprocessing of protein crystals were as follows. The PDB database (https://www.rcsb.org/) contains data on most of the crystals of biological macromolecules reported to date, including crystal complexes of biological macromolecules and small molecules. This docking study was mainly focused on the FXR protein. Discovery Studio software was used to preprocess the protein. We then deleted the water molecules, hydrogenated and charged them, extracted the original ligand from the structure, and used PyMol to process the protein. HepG2 cells were planted in 12-well plates. After 24 h, cells were transfected with 1 μg hFXR-luc and 1 μg Ramlila luciferase expression vector pCMV-RL-TK (Promega) for 36 h. Here, pCMV-RL-TK was used as an internal control. To measure the effect of Rc on FXR activity, cells were incubated with Rc (0, 6.25, 12, 25 µM). After 24 h, cells were collected for luciferase activity assessment using the Dual Luciferase Reporter Assay System (Promega). Relative luciferase activity was corrected for Renilla luciferase activity of pCMV-RL-TK, and normalized to the activity of the control. The 5’ end of the mouse FXR gene extending from position −1838 bp (relative to the transcription start site) to +47 was cloned into the pGL3-Basic (Promega) luciferase reporter plasmid with the MluI/XhoI sites. The primers that were used for plasmid construction are shown in Table 2. All of the results are expressed as the means ± SEMs. The data were evaluated and statistical differences between groups were assessed. For multiple group comparisons, we used Student’s t-test and a one-way analysis of variance (ANOVA), followed by a post hoc Tukey test using GraphPad Prism 8. We used LPS-induced LS174T cells to establish the IBD cell model in vitro to observe the potential therapeutic effect of Rc on UC. In our study, Rc exhibited low cytotoxicity and Rc did not inhibit cell viability up to 400 μM (Figure 1A). According to the result of qPCR analysis, the relative mRNA levels of inflammation, including IL-1β, TNF-α, and IL-6 were upregulated in the LPS-induced group. These indices were downregulated with the Rc treatment (Figures 1B–D). Immunofluorescence analysis showed that Rc could inhibit nuclear translocation of NFκB (Figure 1E). These results indicate that Rc may have a potential effect on suppressing inflammation induced by LPS cells. A DSS-induced mouse model was established to further explore the therapeutic effect of RC on UC. Compared with DSS-treated mice, Rc administration resulted in a dose-dependent weight gain (Figure 2A) and a high dose of Rc slightly reduced DAI scores (Figure 2B). DSS treatment significant shortened colons compared to controls (Figure 2C). H&E staining analysis showed mucosal structural repair, an increased number of crypt structures, and the reduced infiltration of inflammatory cells into the mucosa and submucosa. This indicates that Rc effectively reversed the DSS-induced damage of colon tissues (Figure 2D). Meanwhile, Rc treatment reduced the length of colons and increased the weight of colons in a dose-dependent manner (Figures 2E,F). Taken together, these data suggest that Rc had a very positive impact on the treatment of UC. To investigate whether Rc ameliorated the inflammatory response in colitis by activating FXR, we carried out some further studies. First, at the serum level, we found that IL-1β, IL-6, and TNF-α were significantly increased in the DSS-treated model group, whereas high-dose Rc treatment was able to suppress the inflammatory response (Figures 3A–C). The mRNA expression of FXR and its downstream genes were significantly suppressed when compared with the control group, whereas Rc treatment upregulated their expression levels (Figures 3D,E). Similarly, the mRNA levels of DSS-induced pro-inflammatory factors ICAM1, IL-6, and COX2 were significantly increased when compared with the normal group, whereas low expressions were found in the Rc-treatment group (Figures 3F–H). Similarly, immunofluorescence analysis showed that Rc reduced the expression of F4/80 and cd11b in a dose-dependent manner (Figures 3I,J). These results indicate that Rc was a potent anti-inflammatory agent, which could ameliorate or even suppress the inflammatory response and alleviate the inflammatory symptoms of UC. Damage to intestinal barriers is a considerable feature of DSS-induced UC. In our hypothesis we predicted that Rc could not only suppress inflammation but also repair the damage of intestinal barriers induced by DSS. Thus, we focused on TJ molecules. Our results show that Rc upregulated the mRNA expression of ZO-1, claudin-1, and occludin when compared with the DSS group (Figures 4A–C). Meanwhile, the results of immunofluorescence analysis showed that Rc significantly upregulated the expression of ZO-1 in the inflammatory colon when compared with the DSS group (Figure 4D). Thus, Rc improved the reduction in intestinal permeability brought about by UC. Molecular docking was conducted to investigate whether Rc can activate FXR. The results show that the inner part of FXR’s structural domain could be bound by Rc based on hydrogen, hydrophobic interaction, and the van der Waals coefficient (Figure 5A). Luciferase reporter gene assay showed significant effect of Rc in activating transcriptional expression of FXR in a dose-dependent manner (Figure 5B). Meanwhile, qPCR results demonstrated that Rc increased the mRNA levels of FXR, BSEP, and SHP in a dose-dependent manner when compared with the control group (Figures 5C–E). These results indicate that Rc could activate the FXR signaling pathway in vitro. To investigate the effect of FXR on UC, we found that the weight loss and DAI changes caused by DSS were not alleviated after optimal dose treatment with Rc in the case of the knockdown of FXR (Figures 6A,B). Furthermore, the length and weight of the colon were not improved (Figures 6C–E). HE results also showed that the mucosal structure was not repaired, with no change in the mucosa and submucosa in terms of the amount of inflammatory cell infiltration (Figure 6F). This suggests that the absence of FXR renders the treatment of Rc ineffective. Through further studies, we found that under conditions of FXR deficiency, Rc was not able to exert an inhibitory effect on inflammation. At the ELISA level, the expression of IL-1β, IL-6, and TNF-α was not reduced after Rc treatment when compared with the model group (Figures 7A–C). Meanwhile, the mRNA levels of ICAM1, COX-2, and IL-6 were also not significantly inhibited (Figures 7D–F). Immunofluorescence results also suggested that Rc failed to produce the inhibition of inflammatory expression levels in the presence of the knockdown of FXR (Figures 7G,H). Therefore, FXR plays an important role in Rc’s therapeutic process in relation to UC inflammation. In the absence of FXR, the mRNA expression of claudin, occludin, and ZO-1 did not increase after Rc treatment when compared to the model group (Figures 8A–C). Immunofluorescence of ZO-1 also showed no improvement in intestinal permeability (Figure 8D). Overall, Rc treatment could not repair the intestinal mucosal damage caused by DSS after the deletion of FXR. Therefore, FXR may play an active role in repairing the intestinal damage caused by UC and restoring the intestinal mucosa. Although intestinal barrier disruption is often the final factor causing IBD mortality, intestinal inflammation is often the injury that is observed at the beginning. Consequently, ameliorating inflammation by targeting the FXR signaling pathway represents an attractive concept for combating IBD. Activating FXR might be a therapeutic strategy for treating IBD and its complications. In the present study, we assumed that Rc was an FXR activator, and we verified that Rc ameliorated DSS-induced inflammation and intestinal barrier damage by activating FXR. This therapeutic effect disappeared when FXR was absent (the mechanism diagram is shown in Supplementary Figure S2). Rc is believed to play an anti-inflammatory role in many disease, such as gastritis, hepatitis, arthritis, and pneumonia (Yu et al., 2016; Lee et al., 2018). Based on the excellent anti-inflammatory ability of Rc, we wondered if Rc could play the same anti-inflammatory role in IBD. Pro-inflammatory cytokines—including IL-1β, IL-6, and TNF-α—have been used to measure inflammation levels in many IBD studies (Hall et al., 2017; Breugelmans et al., 2020). As we expected, Rc reduced the mRNA levels of IL-1β, IL-6, and TNF-α in LPS-induced LS174T in vitro, which indicates the potential ability of Rc to attenuate intestinal inflammation. Thus, we then explored its therapeutic effect on an IBD model in vivo. DSS-induced mice are characterized by weight loss, loose stools, diarrhea, and even rectal bleeding. Thus, DSS-induced mice have often been used as an animal model of IBD, including UC and CD (Kim et al., 2012; Liu et al., 2021). Consistently, our study showed symptoms such as decreased body weight, a shortened colon, and increased DAI in DSS-induced mice. In this study, Rc-treated mice exhibited increased colonic weight and length when compared to DSS-induced mice. Furthermore, DSS-induced mice displayed a loss of integrity in their intestinal barriers (e.g., decreased crypt foci and increased inflammatory cell infiltration in the mucosa and submucosa), whereas Rc was able to reverse this damage induced by DSS. Here, we report for the first time that Rc can reverse these UC symptoms that are induced by DSS. It is clear that FXR activation suppresses the inflammatory response and preserves intestinal barrier integrity in IBD (Ding et al., 2015). We thus assess the role of Rc in ameliorating inflammation and the effects of IBD on intestinal barriers in vivo. These pro-inflammatory cytokines can cause intestinal barrier impairment (Nunes et al., 2019). Studies have indicated that IL-1β and TNF-α can cause an obvious increase in intestinal barrier permeability, depending on NF-κB activation (Haines et al., 2016; Zhong et al., 2021). According to our results, Rc suppressed the inflammatory response, which resulted in decreased serum levels of IL-1β, IL-6, and TNF-α. Furthermore, the relative mRNA levels of ICAM1, NF-KB, IL-6, and COX-2 were increased by DSS but reduced by Rc in our study. These results suggest that Rc may play an anti-inflammatory role by activating the NF-κB pathway in the intestines. Further verification for this hypothesis was carried out with the use of immunofluorescence, and low levels of CD11b and F4/80 inflammatory cells in filtration were observed in the Rc-treatment group. Intestinal macrophages are key immune cells in the maintenance of intestinal immune homeostasis in IBD, and FXR is a modulator of intestinal innate immunity (Vavassori et al., 2009; Dharmasiri et al., 2021). As we expected, the relative mRNA levels of FXR and its downstream target BSEP were increased with Rc treatment. Mutual crosstalk between FXR and NF-κB might indicate a potential pathway for the anti-inflammatory effect of FXR (Gai et al., 2018). However, more evidence is needed before we can conclude that the activation of FXR directly inhibited the NF-κB pathway. Taken together, our data show that Rc may act as an FXR agonist to reduce intestinal inflammation in DSS-induced mice. The increasing expression of several types of pro-inflammatory mediators also impaired the intestinal barriers (Al-Sadi et al., 2016). We found that Rc could reduce intestinal inflammation by activating FXR, resulting in NF-κB inhibition. It is known that NF-κB is also a mediator for intestinal barriers, regulating multiple cellular signaling pathways (Cuzzocrea et al., 2000). We then explored its effect on repairing intestinal barriers. As expected, our data showed that Rc improved the relative mRNA levels of claudin-1, occludin, and ZO-1. Further verification was carried out using immunofluorescence, which showed that Rc increased the expression of ZO-1, which indicates that Rc could improve the reduction of intestinal permeability induced by DSS. FXR is believed to play an anti-inflammatory role and participates in a wide range of diseases of the gastrointestinal tract, such as IBD, colorectal cancer, and type 2 diabetes (Ding et al., 2015). Zhao et al. (2020) found that the downregulation of FXR promoted DSS-induced UC. Meanwhile, Gadaleta et al. (2011) and Feng et al. (2021) found that inflammation could be inhibited by activating FXR. These studies pointed out that the expression of FXR was closely related to DSS-induced UC, which indicates that the regulation of FXR may be useful for ameliorating IBD. Recently, molecular docking has become a commonly used component of the drug discovery toolbox (Luo et al., 2022). Therefore, molecular docking was conducted to explore whether Rc and FXR could interact each other. Remarkably, Rc showed a strong binding affinity to FXR, which indicates that Rc directly regulated the FXR-mediated signaling pathway. The result of the luciferase reporter gene assay further indicates that Rc regulated the transcriptional activity of FXR. Our qPCR results show that the mRNA expressions of FXR, BSEP, and SHP were increased with Rc treatment. This result is in accordance with those reported for FXR activating its downstream genes BSEP and SHP (Wang et al., 2017; Fu et al., 2022). To our knowledge, this is first time that Rc has been reported as an FXR regulator. Previously, Verbeke et al. (2015) demonstrated a crucial protective role for FXR in cholestatic rats, which meant that FXR agonists could prevent gut barrier dysfunction, observing upregulated claudin-1 and occludin. Novel drugs have been reported that could improve intestinal barrier function by increasing FXR signaling, which resulted in the alleviation of colitis (Song et al., 2019; Dong et al., 2021). Knowing that the effect of Rc on anti-inflammatory and intestinal barrier repair may occur due to FXR activation, we aimed to explore whether FXR activation is essential for Rc to exert its therapeutic effect. Although FXR was lowly expressed in the intestine in DSS-induced mice, FXR-knockout (FXR−/-) mice did not present high DAI scores, whereas DSS-induced FXR−/- mice showed higher DAI scores, lower colon weights, and shorter colon lengths, as well as rising inflammation levels and damaged intestinal barriers. This indicates that DSS also led to UC in FXR−/- mice. However, Rc did not exhibit its therapeutic effect on UC in FXR−/- mice, with no change in the indexes of inflammation and intestinal barriers. Despite the multiple mechanisms behind UC treatment, here the therapeutic effect of Rc on UC appeared to only be achieved by activating FXR. Our findings explain the protective effect of Rc on UC, and they provide evidence that FXR constitutes a valid therapeutic target activated by Rc in the treatment of IBD. Our study suggests that Rc may ameliorate inflammatory and barrier function in the intestines. This leads to the attenuation of DSS-induced UC, in which Rc may potentially activate FXR signaling to protect the intestines from DSS-induced injury.

PMC9649643

Baoxi Zhu,Songping Wang,Rui Wang,Xiaoliang Wang

Identification of molecular subtypes and a six-gene risk model related to cuproptosis for triple negative breast cancer 10.3389/fgene.2022.1022236

28-10-2022

triple negative breast cancer,cuproptosis,molecular subtypes,risk model,SIX-gene

Background: Breast cancer is the mostly diagnosed cancer worldwide, and triple negative breast cancer (TNBC) has the worst prognosis. Cuproptosis is a newly identified form of cell death, whose mechanism has not been fully explored in TNBC. This study thought to unveil the potential association between cuproptosis and TNBC. Materials and Methods: Gene expression files with clinical data of TNBC downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were included in this study. Consensus clustering was utilized to perform molecular subtyping based on cuproptosis-associated genes. Limma package was applied to distinguish differentially expressed genes. Univariate Cox regression was used to identify prognostic genes. Least absolute shrinkage and selection operator and stepwise Akaike information criterion optimized and established a risk model. Results: We constructed three molecular subtypes based on cuproptosis-associated genes, and the cuproptosis-based subtyping showed a robustness in different datasets. Clust2 showed the worst prognosis and immune-related pathways such as chemokine signaling pathway were significantly activated in clust2. Clust2 also exhibited a high possibility of immune escape to immune checkpoint blockade. In addition, a six-gene risk model was established manifesting a high AUC score over 0.85 in TCGA dataset. High- and low-risk groups had distinct prognosis and immune infiltration. Finally, a nomogram was constructed with strong performance in predicting TNBC prognosis than the staging system. Conclusion: The molecular subtyping system related to cuproptosis had a potential in guiding immunotherapy for TNBC patients. Importantly, the six-gene risk model was effective and reliable to predict TNBC prognosis.

Identification of molecular subtypes and a six-gene risk model related to cuproptosis for triple negative breast cancer 10.3389/fgene.2022.1022236 Background: Breast cancer is the mostly diagnosed cancer worldwide, and triple negative breast cancer (TNBC) has the worst prognosis. Cuproptosis is a newly identified form of cell death, whose mechanism has not been fully explored in TNBC. This study thought to unveil the potential association between cuproptosis and TNBC. Materials and Methods: Gene expression files with clinical data of TNBC downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were included in this study. Consensus clustering was utilized to perform molecular subtyping based on cuproptosis-associated genes. Limma package was applied to distinguish differentially expressed genes. Univariate Cox regression was used to identify prognostic genes. Least absolute shrinkage and selection operator and stepwise Akaike information criterion optimized and established a risk model. Results: We constructed three molecular subtypes based on cuproptosis-associated genes, and the cuproptosis-based subtyping showed a robustness in different datasets. Clust2 showed the worst prognosis and immune-related pathways such as chemokine signaling pathway were significantly activated in clust2. Clust2 also exhibited a high possibility of immune escape to immune checkpoint blockade. In addition, a six-gene risk model was established manifesting a high AUC score over 0.85 in TCGA dataset. High- and low-risk groups had distinct prognosis and immune infiltration. Finally, a nomogram was constructed with strong performance in predicting TNBC prognosis than the staging system. Conclusion: The molecular subtyping system related to cuproptosis had a potential in guiding immunotherapy for TNBC patients. Importantly, the six-gene risk model was effective and reliable to predict TNBC prognosis. Breast cancer is one of the leading cause of cancer death in women, which is the top one diagnosed cancer type with 2,261,419 new cases (11.7% of total cases) in 2020 according to the global cancer statistics (Sung et al., 2021). The overall survival of breast cancer is markedly different in developed and developing countries, with an estimated 5-year survival of 80% and below 40%, respectively (Coleman et al., 2008). The incidence of breast cancer elevates with age but seldomly found before the age of 20 years and breast cancer most commonly occurs in 40–50 aged women (Akram et al., 2017). Although many versions of guidelines for the diagnosis and treatment of breast cancer have been established, such as European Breast Guidelines (Schünemann et al., 2020) and the American Joint Committee on Cancer’s (AJCC) guideline (Plichta et al., 2020), the treatment for triple negative breast cancer (TNBC) still remains a challenge. TNBC is a clinically aggressive type of breast cancer with poor survival, compared with other breast cancer types, including HER2-positive, oestrogen receptor (ER)-positive and progesterone receptor (PR)-positive. Chemotherapy resistance and immune escape common occur in TNBC, which makes an obstacle in TNBC treatment (Wein and Loi, 2017). Therefore, accurate molecular biomarkers or subtypes are strongly needed to guide personalized therapy for TNBC. Programmed cell death is recognized as a promising therapeutic target in cancer therapy, where necroptosis, pyroptosis, and apoptosis are the most studied types (Bertheloot et al., 2021). Cuproptosis is considered as a new form of programmed cell death involved in the proliferation of lung cancer cells (Tang et al., 2022). Copper chelators such as RPTDH/R848 nanoparticles are demonstrated to be able to suppress cancer cell growth and metastasis in breast cancer (Zhou et al., 2019), inspiring a possibility that cuproptosis is a potential target for cancer treatment. Up to now, studies have discovered a series of prognostic signatures related to cuproptosis for different cancer types such as kidney renal clear cell carcinoma (Ji et al., 2022), melanoma (Lv et al., 2022), and hepatocellular carcinoma (Zhang et al., 2022). However, the relation between cuproptosis and TNBC has not been revealed. Therefore, in this study, we aimed to analyze the role of cuproptosis in TNBC, and construct molecular subtypes based on cuproptosis-associated genes by using gene expression data of TNBC obtained from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. By comparing the molecular features of different subtypes, we unveiled the relation between cuproptosis and immune infiltration. Moreover, a risk model related to cuproptosis was established for predicting TNBC prognosis. The risk model was effective to distinguish TNBC patients into different risk types. Notably, the model outperformed the AJCC staging system, which had a potential to be used as a prognostic signature in TNBC. The RNA-seq data of TNBC was downloaded from Genomic Data Commons (GDC) Data Portal by TCGA GDC API (https://portal.gdc.cancer.gov/projects/TCGA-BRCA, named as TCGA dataset). GSE103091 dataset was downloaded from GEO database (https://www.ncbi.nlm.nih.gov/geo/). TNBC samples without progression-free survival (PFS) or survival status were eliminated. TNBC samples with PFS shorter than 30 days or more than 10 years were excluded. In GSE103091 dataset, Ensembl ID was converted to gene symbol and we used the averaged expression level when a gene had multiple Ensembl IDs. Finally, 105 TNBC samples and 113 paracancerous samples were remained in TCGA dataset, and 91 TNBC samples were remained in GSE103091 dataset. Cuproptosis genes were obtained from a previous study (Tsvetkov et al., 2022), and a total of 13 cuproptosis genes were used in the study including FDX1, LIPT1, LIAS, DLD, DBT, GCSH, DLST, DLAT, PDHA1, PDHB, SLC31A1, ATP7A, and ATP7B. Firstly, single sample gene set enrichment anlaysis (ssGSEA) was used to calculate the enrichment score of 13 cuproptosis genes for each sample in TCGA dataset. Limma R package (Ritchie et al., 2015) was applied to screen differentially expressed genes (DEGs) between paracancerous and tumor samples (false discovery rate (FDR) < 0.05 and |log2FC| > 1). Then Pearson correlation analysis was performed to evaluate the correlation between the DEG expression and the ssGSEA score of cuproptosis. DEGs with |correlation coefficient (R)| > 0.4 and p < 0.05 were selected. Next, univariate Cox regression analysis in the survival R package was conducted on the DEGs and DEGs with p < 0.05 as the input for unsupervised consensus clustering. ConsensusClusterPlus R package (Wilkerson and Hayes, 2010) was used for conducting unsupervised consensus clustering to identify molecular subtypes. The expression of prognostic cuproptosis-associated genes were used as a basis for clustering samples. KM algorithm and Euclidean distance were set to carry out 500 bootstraps with each bootstrap consisting of 80% of samples in TCGA dataset. Cluster number k was chosen from 2 to 10. The optimal cluster number was determined according to cumulative distribution function (CDF) and area under CDF curve. Gene set enrichment analysis (GSEA) (Subramanian et al., 2005) was utilized to calculate the enrichment score of functional pathways for molecular subtypes. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were obtained from KEGG database (https://www.genome.jp/kegg/). Firstly, DEGs between different molecular subtypes were identified with limma R package (FDR <0.05 and |log2FC| > 1.5). Least absolute shrinkage and selection operator (LASSO) regression analysis (Friedman et al., 2010) decreased the number of DEGs in glmnet R package. Stepwise Akaike information criterion (stepAIC) was applied for further optimizing the risk model through MASS R package (Zhang, 2016). We determined the risk model according to the formula: , where β indicates the coefficient of prognostic genes and Expi indicates the expression level of prognostic genes. Each sample obtained a risk score, which was subsequently transferred to z-score. Samples were stratified into high-risk and low-risk groups according to the z-score = 0. Kaplan-Meier survival analysis was conducted to evaluate the prognosis of the two risk groups. Estimation of STromal and Immune cells in MAlignant Tumours using Expression data (ESTIMATE) tool was implemented to evaluate stromal and immune infiltration (Yoshihara et al., 2013). Microenvironment Cell Populations (MCP)-counter methodology was applied to assess the enrichment of 10 immune cells (Becht et al., 2016). SsGSEA algorithm in GSVA R package was performed to predict estimated proportion of 28 immune cells (Hänzelmann et al., 2013). The bioinformatics analysis in this study was supported by Sangerbox platform (http://vip.sangerbox.com/) (Shen et al., 2022). R software (v4.1) was used as a platform to conduct all statistical analysis. Log-rank test was performed in Kaplan-Meier survival analysis, univariate and multivariate Cox regression analysis. Student t test was performed to examine the difference between two groups. ANOVA was conducted to test the difference among three groups. p < 0.05 was considered as statistically significant. Firstly, we calculated the ssGSEA score of cuproptosis pathway based on 13 cuproptosis genes for each TNBC sample in TCGA dataset (Supplementary Table S1). Paracancerous samples had obviously higher cuproptosis score than tumor samples (Figure 1A). Then differential analysis was performed to identify DEGs between TNBC and paracancerous samples. A total of 3125 DEGs were screened under FDR <0.05 and |log2FC| > 1 (Figure 1B). Next, we analyzed the relation between the expression of DEGs and ssGSEA of cuproptosis by Pearson correlation analysis. 1,275 DEGs with |R| > 0.4 and p < 0.05 were selected for further univariate Cox regression analysis (Supplementary Table S2). 39 prognostic DEGs were found to be significantly associated with TNBC prognosis in TCGA dataset (p < 0.05, Supplementary Table S3), whose expression levels were significantly different between paracancerous and tumor samples (p < 0.0001, Figure 1C). Based on the expression profiles of the 39 cuproptosis-associated genes, we then constructed molecular subtypes through consensus clustering. According to the CDF curve, cluster number k = 3 was determined as the optimal (Figures 2A–C). Three molecular subtypes (clust1, clust2, and clust3) were distinguished based on the 39 cuproptosis-associated genes, and they showed distinct PFS in both TCGA and GSE103091 datasets (Figures 2D,E; Supplementary Figure S1, log-rank p = 0.0038 and 0.036, respectively). Clust2 had the shortest PFS and the most number of dead samples, while clust1 had the favorable prognosis (Figure 2F), indicating that cuproptosis-associated genes may be involved in the TNBC progression. Next we analyzed the enriched pathways of the three subtypes by GSEA. By comparing clust2 to non-clust2 (clust1 and clust3), we observed that immune-related pathways and tumor-related pathways were obviously activated in clust2, such as cytokine-cytokine receptor interaction, chemokine signaling pathway, MAPK signaling pathway, toll-like receptor signaling pathway, TGF-β signaling pathway, and pathways in cancer (Figure 3A). In clust1 vs non-clust1, the above pathways were significantly suppressed (Supplementary Figure S2), suggesting that cuproptosis-associated genes were involved in the immune regulation. Pathways related to cell proliferation and cell death were evaluated in the three subtypes. Among the six pathways, P53 signaling pathway was the most enriched in clust2 and clust1 had the lowest enrichment of cell death-related pathways including necroptosis, ferroptosis, and apoptosis (Figure 3B, ANOVA p < 0.05). This indicated an interaction of cuproptosis with other cell death pathways. Given that immune-related pathways were differentially enriched in three subtypes, we then assessed the immune infiltration. Not surprisingly, clust1 had the lowest stromal score and immune score, compared with other two subtypes (Figure 2C, ANOVA p < 0.0001). Estimation of 10 immune cell types by MCP-counter also showed a lowest enrichment of them in clust1 such as T cells, monocytic lineage, and myeloid dendritic cells (p < 0.05, Figure 3D). Notably, clust2 had the highest enrichment of fibroblasts (p < 0.01, Figure 3D). Similar results were outputted through ssGSEA that majority of immune cells had a low estimated proportion in clust1 (Supplementary Figure S2B). Furthermore, we also determined the expression of immune checkpoint genes in the three subtypes. The result showed that 22 of 47 immune checkpoints were differentially expressed in the three subtypes (Supplementary Figure S2C). We suspected that cuproptosis-associated genes had an influence in tumor microenvironment and therefore affected the efficiency of immunotherapy in TNBC. TIDE analysis revealed the predicted sensitivity of three subtypes to immune checkpoint blockade therapy (Figure 3E). Clust2 had the highest TIDE score, suggesting a high possibility of immune escape to immunotherapy, which may be resulted from a high enrichment of myeloid-derived suppressor cells (MDSCs), T cell exclusion and T cell dysfunction (Figure 3E). The proportion of responders in clust2 was also the lowest compared with other two subtypes (Figure 3F). As three subtypes performed different molecular signatures, we then identified the DEGs between clust1 vs non-clust1, clust2 vs non-clust2, clust3 vs non-clust3. As a result, 2,723 DEGs were screened (FDR <0.05 and |log2FC| > 1.5). Then univariate Cox regression was used to further filter 1,213 DEGs, and finally 89 DEGs (prognostic genes) with 77 risk genes and 12 protective genes remained (Supplementary Figure S3A). Moreover, LASSO regression was performed on 89 genes to generate an optimal risk model. The model reached the optimal when lambda = 0.057, where 14 prognostic genes remained (Supplementary Figure S3B, C). StepAIC was further performed to optimize the prognostic model, and finally six prognostic genes were remained including PTPRN2, SCARB1, SLC37A2, YES1, LY6D, and NOTCH3 (Supplementary Figure S3D). The risk model was determined according to the following formula: For each sample, a risk score was calculated according to the formula. The risk model showed a favorable performance in predicting one- to 5-year PFS with AUC all over than 0.85 in TCGA dataset (Figure 4A). Determined by the optimal cut-off value of risk score, the samples were classified to different risk types (high-risk and low-risk). Kaplan-Meier survival plot showed that high- and low-risk groups had markedly different PFS (Figure 4B, p < 0.0001). In GSE103091 dataset, a favorable AUC of the risk model and differential prognosis between two risk groups was also observed (Figures 4C,D). In the relation between risk score and clinical features, we found that a difference of risk score was shown between stage Ⅰ+Ⅱ and stage Ⅲ+Ⅳ (Figure 5A). In addition, alive samples had a lower risk score than the deceased samples. Kaplan-Meier survival analysis revealed that the risk model could effectively divide samples into high- and low-risk groups grouping by different clinical features (Figure 5B; Supplementary Figure S4A). To understand whether a difference on tumor microenvironment was shown between two risk groups, we applied different tools, including ESTIMATE, MCP-counter, and ssGSEA, to evaluate the immune infiltration. The three tools showed consistent result that high immune infiltration was displayed in samples with high risk (Figures 5C,D, Supplementary Figure S4B). The above findings further demonstrated that cuproptosis-associated genes were possibly involved in the modulation of tumor microenvironment. Univariate and multivariate Cox regression analysis revealed that stage and risk score were independent risk factors (Figures 6A,B). Consequently, we established a nomogram based on stage and risk score, of which risk score contributed the most to the nomogram (Figure 6C). Calibration curve showed that the predicted PFS was similar to the observed PFS (Figure 6D). Decision curve analysis (DCA) demonstrated the reliability of the nomogram and risk model (Figure 6E). Compared with other clinical characteristics, the nomogram and risk model exhibited a better performance in predicting PFS, especially long-term PFS (Figure 6F). An increased level of copper can result in cell death and the disruption of cupper homeostasis can lead to life-threatening diseases such as Wilson’s disease and neurodegenerative disorders (Gaggelli et al., 2006; Bandmann et al., 2015). Tsvetkov et al. have revealed that copper-induced cell death, which defined as cuproptosis, is mediated by protein lipoylation involved in tricarboxylic acid (TCA) cycle (Tsvetkov et al., 2022). Unlike other cell death forms including apoptosis, ferroptosis, necroptosis, and pyroptosis, cuproptosis functions in a new mechanism through which mitochondrial ferredoxin 1-mediated protein lipoylation leads to proteotoxic stress and ultimately cell death (Tsvetkov et al., 2022). Cupper ionophores and cupper chelators have been explored as potential anti-cancer molecules (O'Day et al., 2013; Cui et al., 2021), which inspires the research on the potential of cuproptosis in cancer treatment. We observed a significant difference of cuproptosis score between normal and TNBC samples, suggesting the instability of cuproptosis homeostasis in cancer cells. Normal samples have a higher cuproptosis score than TNBC samples, indicating a higher activity of cupper dwindling. Therefore, we further explored the association of cuproptosis with TNBC prognosis, functional pathways, and tumor immune microenvironment through constructing molecular subtypes based on cuproptosis-associated genes. The current study have shown that the three molecular subtypes had distinct prognosis and enrichment of activated pathways. Clust2 had the worst prognosis and the highest proportion of dead samples. Notably, immune related pathways were significantly activated in clust2, such as cytokine-cytokine signaling pathway, chemokine signaling pathway, and Toll-like receptor signaling pathway, which drove a possibility that cuproptosis may participate in the modulation of immune microenvironment. Not surprisingly, three molecular subtypes demonstrated different immune infiltration and response to immune checkpoint blockade. Clust2 was predicted to have a great possibility of immune escape in immunotherapy, compared to other two subtypes, which may be resulted from T cell exclusion and T cell dysfunction. Differential analysis on three molecular subtypes illustrated that cuproptosis was involved in cancer progression and immune microenvironment. Furthermore, we established a risk model based on cuproptosis-related genes, where six prognostic biomarkers were included (PTPRN2, SCARB1, SLC37A2, YES1, LY6D, and NOTCH3). Most of these biomarkers have been reported to promote cancer progression. PTPRN2 is a protein tyrosine phosphatase receptor, which was found to be upregulated in metastatic breast cancer and could promote cancer metastasis through PI(4,5)P2-dependent actin remodeling (Sengelaub et al., 2016). Immature isoform of PTPRN2 (proPTPRN2) expression was closely associated with lymph node-positive breast cancer and poor clinical outcome (Sorokin et al., 2015). Scavenger receptor class B member 1 (SCARB1) is a cell-surface glycoprotein mediating low density lipoprotein-cholesteryl ester (LDL-CE), which is involved in lipid internalization (Swarnakar et al., 1999). David de Gonzalo-Calvo et al. suggested that SCARB1 potentially promote CE accumulation and aggressive feature in breast cancer (de Gonzalo-Calvo et al., 2015). Proto-oncogene tyrosine-protein kinase (YES1) has been widely reported to stimulate cancer cell growth and migration in various cancer types such as lung cancer (Garmendia et al., 2019), gastric cancer (Mao et al., 2021), and breast cancer (Takeda et al., 2017), which is therefore considered as a novel therapeutic target for cancer therapy (Garmendia et al., 2022). Targeting YES1 was effective to restore the sensitivity to chemotherapeutic drugs (trastuzumab and lapatinib) in drug-resistance breast cancer cell lines (Takeda et al., 2017). Moreover, downregulation of YES1 via miR-133 was demonstrated to inhibit cancer cell proliferation triple-negative breast cancer cell lines (Zhang et al., 2020). Lymphocyte antigen six superfamily member D (LY6D) has been identified as a biomarker for bladder cancer and a chemoresistance marker laryngeal squamous cell carcinoma (Andersson et al., 2020; Wang et al., 2020). NOTCH3 signaling is a well-known pathway contributing to cancer development (Aburjania et al., 2018). SLC37A2 has not been reported to be involved in cancerigenesis or cancer progression. The risk model manifested a favorable performance in predicting TNBC prognosis in the two independent datasets. Two risk groups also showed different immune infiltration, which was consistent with the result on molecular subtypes. To increase the accuracy of the risk model in predicting TNBC prognosis, we further established a nomogram that exhibited a better performance than the staging system. In conclusion, this study revealed the important role of cuproptosis in TNBC development and its crosstalk with tumor immune microenvironment. We distinguished three molecular subtypes related to cuproprotiss, which had a potential to guide the personalized immunotherapy. In addition, we established a six-gene risk model with robust performance to predict TNBC prognosis.

PMC9649646

Yongchao Zhao,Aichao Xia,Chaofu Li,Xianping Long,Zhixun Bai,Zhimei Qiu,Weidong Xiong,Ning Gu,Youcheng Shen,Ranzun Zhao,Bei Shi

Methyltransferase like 3-mediated N6-methylatidin methylation inhibits vascular smooth muscle cells phenotype switching via promoting phosphatidylinositol 3-kinase mRNA decay

28-10-2022

neointimal hyperplasia,VSMCs phenotype switching,N6-methyladenosine,methyltransferase like 3,mRNA decay

N6-methylatidine (m6A) is involved in post-transcriptional metabolism and a variety of pathological processes. However, little is known about the role of m6A in vascular proliferative diseases, particularly in vascular smooth muscle cells (VSMCs) phenotype switching-induced neointimal hyperplasia. In the current study, we discovered that methyltransferase like 3 (METTL3) is a critical candidate for catalyzing a global increase in m6A in response to carotid artery injury and various VSMCs phenotype switching. The inhibited neointimal hyperplasia was obtained after in vivo gene transfer to knock-down Mettl3. In vitro overexpression of Mettl3 resulted in increased VSMC proliferation, migration, and reduced contractile gene expression with a global elevation of m6A modification. In contrast, Mettl3 knockdown reversed this facilitated phenotypic switch in VSMCs, as demonstrated by downregulated m6A, decreased proliferation, migration, and increased expression of contractile genes. Mechanistically, Mettl3 knock-down was found to promote higher phosphatidylinositol 3-kinase (Pi3k) mRNA decay thus inactivating the PI3K/AKT signal to inhibit VSMCs phenotype switching. Overall, our findings highlight the importance of METTL3-mediated m6A in VSMCs phenotype switching and offer a novel perspective on targeting METTL3 as a therapeutic option for VSMCs phenotype switching modulated pathogenesis, including atherosclerosis and restenosis.

Methyltransferase like 3-mediated N6-methylatidin methylation inhibits vascular smooth muscle cells phenotype switching via promoting phosphatidylinositol 3-kinase mRNA decay N6-methylatidine (m6A) is involved in post-transcriptional metabolism and a variety of pathological processes. However, little is known about the role of m6A in vascular proliferative diseases, particularly in vascular smooth muscle cells (VSMCs) phenotype switching-induced neointimal hyperplasia. In the current study, we discovered that methyltransferase like 3 (METTL3) is a critical candidate for catalyzing a global increase in m6A in response to carotid artery injury and various VSMCs phenotype switching. The inhibited neointimal hyperplasia was obtained after in vivo gene transfer to knock-down Mettl3. In vitro overexpression of Mettl3 resulted in increased VSMC proliferation, migration, and reduced contractile gene expression with a global elevation of m6A modification. In contrast, Mettl3 knockdown reversed this facilitated phenotypic switch in VSMCs, as demonstrated by downregulated m6A, decreased proliferation, migration, and increased expression of contractile genes. Mechanistically, Mettl3 knock-down was found to promote higher phosphatidylinositol 3-kinase (Pi3k) mRNA decay thus inactivating the PI3K/AKT signal to inhibit VSMCs phenotype switching. Overall, our findings highlight the importance of METTL3-mediated m6A in VSMCs phenotype switching and offer a novel perspective on targeting METTL3 as a therapeutic option for VSMCs phenotype switching modulated pathogenesis, including atherosclerosis and restenosis. Atherosclerosis, which includes coronary atherosclerosis, restenosis, and other vascular proliferative diseases, is still the leading cause of morbidity and mortality throughout the world (1, 2). The transition of vascular smooth muscle cells (VSMCs) from a differentiated phenotype (contractile cells) to a dedifferentiated phenotype (synthetic cells) has been shown to play an important role in the pathogenesis of atherosclerosis (3). To maintain vascular tone, VSMCs in normal, healthy vessels have a low proliferative and migratory rate and express a variety of contractile-related genes. VSMCs exhibit highly dedifferentiated phenotypes in response to mechanical or chemical vascular injury, characterized by robust proliferation, migration, and decreased expression of contractile-related genes, such as alpha-smooth muscle actin (α-Sma), smooth muscle protein 22-alpha (Sm22α), etc., which eventually lead to neointimal hyperplasia (4). The underlying mechanism that controls this phenotype switching, however, is not yet fully understood. The epigenetic dimensional modulation of gene expression is centrally involved in regulating VSMCs pathogenesis, according to evidence from numerous studies (5). Abnormalities in DNA methylation (6, 7), histone modification (8), and non-coding RNAs (9, 10) have been studied extensively among the identified epigenetic mechanisms in this process, whereas, studies on post-transcriptional chemical modifiers of RNA have received little attention. N6-methylatidine (m6A) is the most common type of eukaryotic RNA modification among the more than 150 discovered to date (11). The m6A represents a dynamic and reversible process that is activated by methyltransferase 3 (METTL3), methyltransferase 14 (METTL14), methyltransferase 16 (METTL16), and Wilms tumor 1-associated protein (WTAP) defined as “writers,” inactivated by demethylases fat mass and obesity-associated (FTO) and alk homolog 5 (ALKBH5) defined as “erasers,” recognized by YTH domain families (YTHDF1-3) and insulin-like growth factor binding proteins (IGFBP1-3) defined as “readers” (12). While, much is known about how m6A affects various biological functions of post-transcriptional RNA metabolism, such as RNA splicing (13, 14), nuclear processing (15), stability (16), decay (17), translation (18), and the development and progression of various diseases (19–21). However, currently very little is known about the biological function and related molecular mechanism of m6A in VSMCs phenotype switching. The current study investigated the role of METTL3-mediated m6A in VSMCs phenotype switching and the post-transcriptional regulatory mechanism that controls it. METTL3, the core of methyltransferase, was screened out and found to be responsible for the global increase of m6A based on several VSMCs phenotype switching models in vivo, in vitro, and ex vivo. Mettl3 knockdown caused a decrease in m6A, phenotype switching in synthetic VSMCs, and neointimal hyperplasia. Mechanistically, the decreased stability and increased decay of m6A on phosphatidylinositol 3-kinase (Pi3k) mRNA causes the protein kinase B (AKT) pathway to be inactivated, preventing VSMCs phenotype switching. Our research, taken together, sheds new light on the role of post-transcriptional modification in the modulation of vascular proliferative diseases, including atherosclerosis and restenosis. All animal experiments were conducted following the National Academy of Sciences’ “Guide for the Care and Use of Laboratory Animals,” which was approved by the Animal Care and Utilization Committee at Zunyi Medical University. The Supplementary material contains a detailed description of the primers, antibodies, and small interfering RNAs (siRNAs) used in this study. The balloon injuries were performed in rat carotid arteries using male 8-week-old Sprague-Dawley rats weighing 220–260 g (Hunan SJA Laboratory Animal Co., Ltd.) as previously described (22). The left common carotid arteries were fully exposed after the rats were anesthetized. At the distal end of the external carotid artery, the common and internal carotid arteries were clamped and ligated. At the proximal end of the ligation site, an incision of about 0.25 cm in diameter was made. The common carotid artery was inserted into this incision while the vascular clamp on the common carotid artery was removed. To successfully damage the vascular endothelium, a 2.0 mm × 20 mm balloon (Medtronic, Inc.) was inflated to 3.5 atmospheres and rotated five times counterclockwise and five times clockwise. The balloon was then deflated, and the incision was closed. The injured arteries were harvested 14 days after surgery for further testing. The arteries were paraffin-embedded and cut into 5 μm slices. The intimal and medial areas were measured and calculated after the sections were stained with hematoxylin-eosin (HE) staining solution and visualized under a microscope. Deparaffinized sections were deparaffinized and blocked for 1 h with five percent goat serum (Beyotime, #C0265) for immunofluorescence. The primary antibodies made with goat serum were dropped and refrigerated overnight at 4°C. The secondary antibodies were then added dropwise and incubated for 1 h at room temperature, protected from light. After that, the images were visualized in fluorescence microscopy using the DAPI staining solution for another 10 min. Supplementary Table 1 provides a detailed description of the antibodies used in this study. Total RNA was extracted by a columnar RNA extraction kit (Sangon, #B511321-0100) according to the manufacturer’s instructions. After that, 1,000 ng of total RNA was reverse transcribed (Takara, #RR036A). The total system was 20 μl, including 4 μl of 5 × PrimeScript RT Master Mix, 10 μl of 100 ng/μl total RNA, and 6 μl of DEPC solution (Bioshap, #BL510A). Amplification was conducted on a CFX Connect Real-Time System (Bio-Rad, CA, USA) using an amplification kit (Takara, #RR820A). The total system was 10 μl, including 5 μl TB Green Premix Ex Taq II, 0.5 μl forward primers, 0.5 μl reverse primers, 0.8 μl template DNA, and 3.2 μl DEPC. RT-qPCR reaction program settings have been composed of: 95°C for 30 s, 95°C for 5 s, and 60°C for 30 s for 39 cycles. After normalizing the expression of the gene to β-actin, the relative expression of the target gene was calculated using the 2–Δ Δ CT method. Primer sequences are shown in Supplementary Table 2. The DEPC (Bioshap, #BL510A) solution was used to adjust total RNA concentrations to 50 ng/μl and 100 ng/μl, respectively. Total RNA was serially diluted and denatured at 95°C for 3 min before being cooled on ice. Two microliters total RNA was carefully dropped onto a Hyond-N+ membrane (GE Healthcare, #RPN203B), dried at room temperature, and then cross-linked to the Hyond-N+ membrane using the UV cross-linking instrument (1.2 J × 3 min). The membrane was blocked for 1 h at room temperature in a 5% BSA (Epizyme, #PS133) before being incubated overnight at 4°C in a specific anti-m6A antibody dilution buffer (Supplementary Table 1). After that, the secondary goat anti-rabbit antibody (Proteintech, #SA00001-2) was incubated for another 1 h. In a Bio-Rad X-ray developer system, the results were obtained. The membrane was then stained for 5 h at room temperature with 0.1% methylene blue (MB). Images were visualized after washing the membrane with ddH2O to see if total RNA loading was consistent. Image J software was used to perform a quantitative analysis of the bands’ gray values. The proteins were extracted and separated using the Epizyme (#PG112) kit before being transferred to a PVDF membrane (Epizyme, #WJ001). For 10 min, a protein-free rapid-blocking solution (Epizyme, #PS108) was used to block the cells. The primary antibody was incubated at 4°C overnight before being rinsed three times for 10 min with Tris-buffered saline containing Tween 20 (TBST). The secondary antibodies were then incubated at room temperature for another 1 h. The Bio-Rad X-ray development system was used to obtain the results after the ECL luminescent substrate (Tanon, #180-501) was added dropwise to the membrane (Bio-Rad, CA, USA). The loading control was -Actin, and the data were analyzed using Image J software. The antibodies used for Western Blot in this study are listed in Supplementary Table 1. Vascular smooth muscle cells were cultured as described previously (23). In brief, cervical dislocation was performed on SD rats aged 6 weeks. The aortic explants were removed and soaked in a four percent penicillin/streptomycin solution (Solarbio, #P1400) twice. The adventitia was then dissected by the cuff after the perivascular tissue was removed. The medial tissue was cut longitudinally, the intima was gently scraped off, and the cells were then placed in DMEM supplemented with 20% FBS (MRC, #CCS300090). The obtained vascular medial tissue was minced into 1 mm3 piece and transferred to a culture flask containing 3 ml DMEM supplemented with 20% FBS, which was then inverted for 2 h before being flipped. On day 3, cells crawled out around the tissue block, and the solution was changed on day 5. The blocks were washed off and the cells were passaged once they had reached 85% confluence. For the following experiments, passages 3–5 were chosen. For 48 h, cells were treated with 20 ng/ml platelet-derived growth factor (PDGF-BB) (R&D, #520-BB) to induce phenotype switching in VSMCs. To assess PDGF-BB-induced VSMCs phenotype switching, the expression of proliferation, migration and contractile genes were examined. The proliferation of VSMCs was detected using the EdU kit (Beyotime, #C0071S). Briefly, the EdU solution was diluted in a 1,000:1 culture medium before being added to 24-well plates seeded with 300 μl of cells per well. Put it back in the incubator for another 2 h. Fixative (Biosharp, #BL539A) was used to fix the cells, and 0.4% TritonX-10 (Biosharp, #BS084) was used to permeabilize. The reaction solution was added at 500 μl per well after rinsing twice with PBS. Incubate in the dark for another 30 min. A fluorescence microscope was used to photograph and record a 100 μl per well DAPI (Solarbio, #C0065) solution incubated for 5 min. A scratch assay was utilized to assess VSMCs migration. In a 6-well plate, the VSMCs were grown to full confluence before being wounded with a sterile 200 μl pipette tip. 2 ml of DMEM culture medium containing 0.1% FBS was added after rinsing twice with PBS. The plates were then returned to the incubator for another 48 h. Under an inverted microscope, images were captured and the scratch area of VSMCs was calculated using Image J software. A transwell assay was further conducted to assess VSMCs migration. Briefly, a 24-well Boyden chamber with a porous polycarbonate membrane (8 μm pore size; Corning, NY, USA) was used by us to examine cell migration function. The cells treated with different conditions were seeded in the upper chamber (100 μl/well, 5 × 104 cells/ml), and the lower chamber was filled with DMEM culture media containing 10% FBS (500 μl/well). After incubation for 12 h (5% CO2, 37°C), the cells that migrated through the filter membrane were fixed with 4% paraformaldehyde for 15 min and stained with crystal violet (#070920201103) for 30 min. Then the number of migrated cells was photographed and recorded under a microscope. The detailed carotid artery gene delivery method was performed as previously described (24). In brief, an adeno-associated virus (AAV) (Hanbio Co., Ltd.) containing Sm22a, a specific VSMCs promoter, was created to knock-down Mettl3 expression on VSMCs with greater specificity. Then, in a liquid state, 1 × 1012 v.g./ml control AAV (AAV-shCtrl) or AAV to knock-down Mettl3 (AAV-shMettl3) were mixed into 30 percent pluronic F-127 (PF-127) glue (Sigma, #9003-11-6). The carotid arteries of 4-week-old male SD rats weighing 60–80 g were then fully exposed, and PF-127 glue was evenly wrapped around the common carotid artery. The balloon injury was performed 4 weeks later, and samples were taken for further investigation on day 14 after surgery. Vascular smooth muscle cells were seeded into 6-well plates to reach a confluence of 50% before transfection. Then, 5 μl of 20 μM Mettl3 siRNA (siMettl3) (Hanbio, #PK210802121DS) or negative control siRNA (siNC) (#JX211009) was diluted into 200 μl in serum-free medium, and mix well with the RNAFit (HANBIO, #HB-RF-500) at room temperature for 10 min. Subsequently, 1.8 ml of culture medium containing 10% FBS with 200 μl mixture per well were placed into a 6-well plate. The silencing efficiency was tested by RT-qPCR and Western Blot at 24 h after transfection. The sequence of siRNAs used in this study is shown in Supplementary Table 3. The MeRIP-qPCR was carried out according to Millipore’s instructions (#17-10499). VSMCs were lysed with RIP buffer and immunoprecipitated in IP buffer overnight with specific anti-m6A antibody or control IgG antibody-conjugated beads, followed by RNA purification. Finally, RT-qPCR was used to determine the enrichment of m6A immunoprecipitated RNA, which was normalized to the input control. Supplementary Table 4 lists the primers used in this study for MeRIP-qPCR. Vascular smooth muscle cells were seeded in 6-well plates and each well was treated with a 5 μl stock solution (1 mg/ml) of actinomycin D (Sigma, # 50-76-0). Following that, total RNA was extracted at time points of 0, 1, 2, 4, 6, and 8 h after adding actinomyces D, with steps referring to total RNA extraction. Following that, the RT-qPCR was carried out, and all of the groups were normalized using a Ct value reference of 0 h. The relative mRNA decay rate and half-life were calculated using GraphPad Prism 9.0 software and non-linear regression curve fitting (single-phase decay). The data were presented as a mean standard ± deviation (SD). When the data satisfies a normal distribution, the Shapiro-Wilk test was used to determine normality, and the differences between the two groups were compared using two-tailed unpaired student’s t-tests; otherwise, the Mann-Whitney non-parametric test was used. For multiple groups, a one-way ANOVA with Bonferroni’s post-hoc test was used to determine statistical significance. GraphPad Prism 9 (San Diego, CA, USA) was used for statistical analysis, and a P-value of less than 0.05 was considered significant. A rat carotid artery balloon injury model was used to induce neointimal hyperplasia to explore if m6A was involved in VSMCs phenotype switching. The intima-to-media area ratio of harvested carotid increased significantly 14 days after balloon injury, according to morphology analysis of HE staining (Supplementary Figure 1A). Further immunofluorescence indicated reduced levels of α-SMA and SM22α (contractile marker proteins) and greater levels of OPN (synthetic marker protein) and PCNA (proliferative marker protein) (Supplementary Figures 1B,C). These findings suggested that a successful neointimal hyperplasia model was created. Total RNAs were extracted after balloon injury, and m6A was detected using a Dot Blot assay. The findings revealed a global rise in m6A modification (Figure 1A). The expression of m6A-related methyltransferases, demethylases, and recognition proteins was comprehensively examined using RT-qPCR to investigate the factors that led to the upward revision of m6A (Figure 1B). Mettl3, Fto, Ythdf2, and Igf2bp2 were found to be differentially expressed in the RT-qPCR analysis. Among these genes, Mettl3, as a protein level further collaborated by Western Blot (Figure 1C), was the most aggressively up-regulated. Considering that METT3 is a methyltransferase and the m6A was found to be up-regulated globally, despite the fact that the expression of the demethylase Fto was also significantly increased. METTL3 was finally chosen as a candidate for further investigations. The location of METTL3 in neointimal hyperplasia was then investigated further (Figure 1D). When METTTL3 was overlaid with α-SMA, another synthetic VSMCs manufacturer, the immunofluorescence image revealed that METTTL3 was highly expressed in the neointima-to-media area (Figure 1E), implying that METTL3 could be a key regulator in VSMCs phenotype switching. Given that PDGF-BB is a well-known platelet-derived bioactive mediator that causes VSMCs to switch from a contractile cell phenotype to a highly synthetic and proliferating cell type that promotes injury repair (25). Thus, the VSMCs were cultured and indentified by immunofluorescence (Supplementary Figure 1D). Next, the proliferation and migration of VSMCs treated by PDGF-BB were also assessed. The EdU assay revealed a higher number of EdU positive VSMCs in PDGF-BB-induced conditions (Supplementary Figure 1E), implying that PDGF-BB-induced VSMCs proliferation. The scratch assay and transwell assay were both utilized to test the VSMCs migratory ability, and the results showed a higher scratch closure and migration in PDGF-BB-treated VSMCs (Supplementary Figures 1F,G), indicating that PDGF-BB-induced VSMCs migration. Finally, the expression of VSMCs marker genes was investigated, and RT-qPCR data revealed a decreased expression of α-Sma, Cnn1, and Sm22α, as well as an increased expression of Opn, Myh10, and Vim, in the PDGF-BB stimulation, compared to the Ctrl (Supplementary Figure 1H). After that, we examined the m6A modification and METTL3 expression in vitro VSMCs phenotype switching induced by PDGF-BB. The m6A was found to be significantly up-regulated in VSMCs after PDGF-BB treatment (Figure 1F), and both Mettl3 expression (Figure 1G) and METTL3 level (Figure 1H) were significantly increased. In conclusion, PDGF-BB-treated VSMCs resulted in an increase in m6A and a promoted switch from contractile to the synthetic phenotype. To further comprehensively elucidate the conserved nature of high Mettl3 expression pattern and enhanced m6A level in VSMCs phenotypic switching. Three phenotypic switching models were constructed and confirmed by RT-qPCR: 20% fetal bovine serum (FBS)-induced rat VSMCs in vitro (Supplementary Figure 2A), rat VSMCs with 3rd to 9th passages of extended culture in vitro (Supplementary Figure 2D), and cultured rat thoracic aortas ex vivo (Supplementary Figure 2G). In either 20% FBS-induced VSMCs (Supplementary Figures 2B,C), continuous passaged VSMCs (Supplementary Figures 2E,F), or cultured aortas (Supplementary Figures 2H,I), these additional findings collectively demonstrated a global increase in m6A and a higher expression of Mettl3. Overall, these findings revealed a conserved signature in vivo, in vitro, and ex vivo for increased m6A modification and METTL3 level, implying a critical regulatory role of METTL3 for VSMCs phenotype switching. To assess the role of METTL3-mediated m6A in neointimal hyperplasia, an adeno-associated virus (AAV) containing Sm22a, a specific VSMCs promoter, was constructed to knock-down Mettl3 expression. The carotid arteries of 4-week-old rats were treated with AAV-shCtrl and AAV-shMettl3 delivered by PF-127. On day 14 after the injury, the balloon injury was operated on and the carotid arteries were harvested for further analysis (Figure 2A). The RT-qPCR and Western Blot results confirmed that both Mettl3 expression and METTL3 levels were significantly down-regulated (Figure 2B and Supplementary Figure 3A). To explore whether m6A was altered in the presence of METTL3 downregulation. Total RNAs of carotid were extracted and the Dot Blot assay revealed a significant reduction in global m6A when compared to the control (Figure 2C). To determine whether there is a link between low METTL3 levels and neointimal hyperplasia. The intima thickness of AAV-shMettl3 treated carotid arteries was significantly inhibited, according to HE staining (Figure 2D). The immunofluorescence results showed that reduced METTL3 expression was linked to thin intima thickness, increased contractile markers level of α-SMA and SM22α and decreased synthetic markers level of OPN and PCNA (Figure 2E). In line with this phenomenon, the RT-qPCR data also found a significant increase in contractile gene expression, including α-Sma, Cnn1 and Sm22α, and a significant decrease in synthetic gene expression, including Opn, Myh10, and Vim (Supplementary Figure 3B). Overall, the evidence strongly suggests that METTL3 deficiency reduced m6A modification and neointimal hyperplasia. To explore the role of METTL3 in VSMCs in vitro, the effects of METTL3 gain-of-function and loss-of-function on VSMCs phenotype switching induced by PDGF-BB were investigated. Firstly, the VSMCs were transfected with either control lentivirus (Lv-Ctrl) or lentivirus specific to Mettl3 (Lv-Mettl3). The RT-qPCR (Supplementary Figure 4A) and Western Blot (Supplementary Figures 4B,C) analysis confirmed a stable up-regulation of METTL3 level, which was associated with a significant global increase in m6A (Supplementary Figures 4D,E) in either vehicle or PDGF-BB-treated cells. The proliferation and migration of VSMCs were also assessed. The EdU assay revealed that up-regulated METTL3 was associated with a higher number of EdU positive VSMCs only in PDGF-BB-induced conditions, not in vehicle-induced conditions (Supplementary Figure 4F), implying that Mettl3 overexpression promoted PDGF-BB-induced VSMCs proliferation. The scratch assay and transwell assay were used to test the migratory ability, and the results showed that upregulated Mettl3 expression resulted in a higher scratch closure and migration cells in PDGF-BB-treated VSMCs, indicating that increased METTL3 significantly augment PDGF-BB-induced migration. Interestingly, this phenomenon was not observed in vehicle-treated VSMCs even though Mettl3 expression was up-regulated (Supplementary Figures 4G,H). Finally, the expression of VSMCs-related genes was investigated, and RT-qPCR data revealed a decrease in α-Sma, Cnn1 and Sm22α, as well as an increase in Opn, Myh10, and Vim, in the Lv-Mettl3 group in response to PDGF-BB stimulation, compared to Lv-Ctrl group (Supplementary Figure 4I). In conclusion, METTL3 overexpression resulted in an increase in m6A and a further promoted phenotype switch of VSMCs. The VSMCs were then transfected with either control siRNA (siNC) or siRNA specific to Mettl3 (siMettl3). The RT-qPCR (Figure 3A) and Western Blot (Figures 3B,C) analysis confirmed a stable down-regulation of METTL3 level, which was associated with a significant global decrease in m6A (Figures 3D,E) in either vehicle or PDGF-BB-treated cells. The proliferation and migration of VSMCs were also assessed. The EdU assay revealed that down-regulated METTL3 was associated with a lower number of EdU positive VSMCs only in PDGF-BB-induced conditions, not in vehicle-induced conditions (Figure 3F), implying that Mettl3 knockdown inhibited PDGF-BB-induced VSMCs proliferation. The scratch assay and transwell assay was used to test the migratory ability, and the results showed that knocking down Mettl3 resulted in a lower scratch closure and migration cells in PDGF-BB-treated VSMCs, indicating that decreased METTL3 significantly reduced PDGF-BB-induced migration. Interestingly, this phenomenon was not observed in vehicle-treated VSMCs even though Mettl3 expression was down-regulated (Figures 3G,H). Finally, the expression of VSMCs-related genes was investigated, and RT-qPCR data revealed an increase in α-Sma, Cnn1, and Sm22α, as well as a decrease in Opn, Myh10, and Vim, in the siMettl3 group in response to PDGF-BB stimulation, compared to vehicle-treated VSMCs (Figure 3I). In conclusion, METTL3 deficiency resulted in a decrease in m6A and a promoted switch reversal from synthetic to contractile phenotype in VSMCs. Until now, the mechanical evidence for the effect of METTL3 on VSMCs phenotype switching is still lacking. Numerous previous studies have shown that the PI3K/AKT pathway plays a critical role in VSMCs phenotype switching (26–28). RT-qPCR was used to confirm the exact role of the PI3K/AKT signal in VSMCs switching, and it revealed that Pi3k expression was significantly up-regulated in PDGF-BB-stimulated VSMCs phenotype switching (Figure 4A). AKT phosphorylation at Thr308 (p-AKTT308) and Ser473 (p-AKTS473) were both increased (Figure 4B), indicating that PI3K/AKT signaling was indeed activated upon PDGF-BB-induced phenotype switching. However, it was unclear whether the PI3K/AKT signal is regulated by METTL3-mediated m6A. To test the hypothesis, we used the gene expression omnibus (GEO) database to seek the presence of m6A on Pi3k mRNA. The methylated RNA immunoprecipitation with next-generation sequencing (MeRIP-seq) data from the GES94148 dataset revealed that m6A modifications do distribute on Pi3k mRNA, as shown by the tool of integrated genome browser (IGV) (Figure 4C). As previously reported, although m6A mRNA methylation preferably occurs within the RRACH consensus motif, only a small portion of such RRACH sites are modified (29, 30). Based on the identified mammalian m6A sites in single-nucleotide resolution from the sequence-based RNA adenosine methylation site predictor (SRAMP) online database (31). Three m6A sites with a high level of confidence were screened for the possible presence of Pi3k mRNA (Figure 4D). Three pairs of primers specific to different m6A sites were designed and the methylated RNA immunoprecipitation qPCR (MeRIP-qPCR) assay was established to confirm the effect of METTL3 deletion on m6A modification of Pi3k mRNA. In METTL3 deficient VSMCs, only a decrease in m6A modification at site 3 within Pi3k mRNA was observed (Figure 4E). Because the presence of m6A methylation impacts the post-transcriptional RNA metabolism including RNA stability and decay (16). Actinomycin D, a transcriptional inhibitor cultured with VSMCs for a different time point, was used to test the decay rate of Pi3k mRNA. The RT-qPCR assay revealed that lower METTL3 was associated with a shorter half-life of Pi3k mRNA, implying that the loss of METTL3-mediated m6A deletion of Pi3k mRNA preferred more rapid decay (Figure 4F). This crucial mechanistic evidence explained the characteristic, down-regulated Pi3k mRNA expression detected by RT-qPCR (Figure 4G) and PI3K protein level detected by Western Blot (Figure 4H), resulting in PI3K/AKT pathway inactivity as evidenced by decreased phosphorylation of p-AKT(T308) and p-AKT(S473). Although the deficiency of Mettl3 led to the post-transcriptional decay of Pi3k mRNA, it was unclear whether METTL3-modulated VSMCs phenotype switching was dependent on PI3K. LY294002, a specific PI3K inhibitor, was used to restrain PI3K phosphorylation. Following that, the effect of METTL3-induced phosphorylation on AKT was investigated. The results showed that LY294002 blocked METTL3 overexpression-induced activation of p-AKT(T308) and p-AKT(S473), implying that LY294002 countered PI3K/AKT activation by up-regulating METTL3 (Figures 5A,B). The effects of LY294002 on METTL3-mediated VSMCs proliferation, migration, and relative contractile gene expression were also investigated. The EdU assay demonstrated a significant reduction in the number of proliferative VSMCs (Figure 5C), as well as a reduction in the scratch closure area and the number of migratory VSMCs (Figures 5D,E). Furthermore, the RT-qPCR analysis revealed a significant increase in α-Sma, Sm22α, Cnn1, and a significant decrease in Opn, Myh10, Vim (Figure 5F). Overall, these findings showed that restraining PI3K phosphorylation could block the function of the PI3K/AKT signal and VSMCs phenotype switching, implying that METTL3 modulated VSMCs function is PI3K dependent. To support the necessity of in vivo carotid artery balloon injury, we also examined the expression of Pi3k and phosphorylation of PI3K/AKT signal. The RT-qPCR (Figure 5G) and Western Blot (Figure 5H) confirmed that Pi3k expression was up-regulated and the PI3K/AKT signal was activated upon balloon injury. In the injured carotid artery with AAV-mediated loss of METTL3, the results found that Pi3k was declined (Figure 5I) and p-AKT(T308), p-AKT(S473) were decreased (Figure 5J). Overall, these findings suggest that the PI3K/AKT signal may play an equal role in METTL3-mediated m6A modulating neointimal hyperplasia in vivo, which warrants more investigation. Prior studies have emphasized the importance of VSMCs phenotype switching in modulating neointimal hyperplasia thus protecting the initiation and progression of vascular proliferative diseases, including atherosclerosis and restenosis (3). The m6A methylation serves as an important dimensionality of epigenetic regulation of gene expression networks (32), but its role and related mechanism in VSMCs phenotype switching-induced intimal hyperplasia remain to be illustrated. In the current study, we discovered that the m6A methyltransferase METTL3 plays a critical role in regulating VSMCs phenotype switching and neointimal hyperplasia in vivo, in vitro, and ex vivo. The current findings show that METTL3 is a key m6A candidate involved in VSMCs phenotype switching in a variety of settings, including in vivo carotid artery injury, in vitro PDGF-BB, FBS, and passage aged stimulated VSMCs, and ex vivo isolated cultured thoracic aortas, and that it is responsible for the global increase in m6A modification. In terms of function, Mettl3 deficiency protects against in vivo carotid artery injury and in vitro PDGF-BB-induced VSMCs phenotype switching, as well as a global reduction in m6A methylation. Inversely, Mettl3 overexpression aggravates PDGF-BB-induced VSMCs phenotype switching in vitro, and a global augment likewise in m6A methylation. Pi3k mRNA was shown to be the key downstream substrate regulated by METTL3-mediated loss of m6A and has a shorter half-life post-transcriptionally to the inactive PI3K/AKT pathway, thus mitigating VSMCs phenotype switching induced by PDGF-BB. We also investigated the PI3K/AKT signal changes and discovered that decreased METTL3 mediated by AAV is linked to a lower level of PI3K and AKT phosphorylation. To the best of our knowledge, this is the first study to show the connection between METTL3-mediated m6A and VSMCs phenotype switching. Vascular endothelial cell dysfunction serves as another important cell fraction in response to atherosclerosis (33). Jian et al. (34) found a high level of METTL14 expression in tumor necrosis factor-alpha-induced endothelial cells, indicating that METTL14 may have therapeutic potential in the treatment of endothelial dysfunction. In the rat carotid artery injury model, the current study suggests that METTL14 expression is not significantly altered, but METTL3 expression is significantly up-regulated. Furthermore, as previously mentioned, METTL3 expression is up-regulated in various in vitro and ex vivo VSMCs phenotype switching experiments. In this study, we discovered a global increase in m6A following carotid artery balloon injury. METTL3 was shown to be a significantly up-regulated component among the m6A related-methyltransferases when compared to the other methyltransferases. It is worth mentioning that after carotid artery injury, the m6A demethylase FTO was upregulated, and recognition proteins of YTHDF2 and IGF2BP2 were also differentially expressed. These findings show that m6A-related methyltransferases have variable expression patterns in various tissues or cells. Given the favorable connection in our experimental setting between the most significantly elevated METTL3 level and the m6A mutation. As a result, METTL3 was ultimately screened out as a candidate for further inquiry, while the functions of FTO, YTHDF2, and IGF2BP2 in neointimal hyperplasia were not investigated. This is a significant unmet need that will be further investigated in the near future. Nevertheless, among all the m6A modification-related methyltransferases discovered until recently, the METTL3 and METTL14 were reported to form N6-methyltransferase complex that methylates various RNAs m6A containing N (6) positions (35). Yet structural and functional studies subsequently confirm that METTL3 constitutes the catalytic core in the heterodimer complex formed with METTL14 (36, 37). Together, this evidence combined with our data confirms the core role of METTL3 as a catalytic m6A and provides a side note on the importance of regulating METTL3 in neointimal hyperplasia. Another intriguing finding is that, despite a decrease in global m6A, we do not discover a regulated phenotype switching in non-PDGF-BB treated VSMCs in vitro, including proliferation, migration, and contractile marker gene expression. The specific pathological stimulus could be one explanation for this inconsistency. Phenotype switching does not occur under physiological conditions, even when m6A is altered, suggesting that m6A may regulate vascular biology homeostasis. Unlike METTL3, ALKBH5 is another m6A demethylase that can be up-regulated in hypoxic conditions, as another of our studies has shown (38). The demethylase ALKBH5 requires molecular oxygen to function, and its m6A demethylase activity is dramatically reduced in hypoxic microenvironments (39), which explains its absence under physiological conditions. As a result, it appears that METTL3-modulated VSMCs phenotype switching is specialized in certain pathological states, emphasizing the therapeutic potential of targeting METTL3 in the treatment of diseased states, such as neointimal hyperplasia. Overall, the current study still finds that METTL3 plays a critical regulatory role in neointimal hyperplasia by inhibiting VSMCs phenotype switching via post-transcriptional down-regulation of Pik3 mRNA decay in an m6A-dependent manner (Figure 6). Our findings suggest that targeting METTL3 in vascular proliferative diseases like atherosclerosis and restenosis could be a viable option. More pre-clinical and clinical studies with various VSMCs phenotype insults are needed to validate the therapeutic potential of blocking METTL3 in our settings. Although m6A methylation and METTL3 expression were investigated in other in vitro VSMCs phenotype switching models, the role and mechanism of METTL3-mediated m6A were only examined in PDGF-BB-induced VSMCs phenotype switching in this study. Given that different in vitro stimuli mimic different pathological states of phenotype switching, our findings on METTL3-modulated VSMCs induced by PDGF-BB warrant further research. In the present study, the role of METTL3 was based on the knockdown or overexpression strategies. More helpful, a convincing approach to the catalytic deficient mutant of METTL3 should be applied in the future. Moreover, PI3K is a kinase, although the gene and protein expression were changed through m6A modification, it remains unclear how m6A affected mRNAs and ultimate protein expression on the phosphorylation of PI3K. Given that the p-PI3K is regulated by the post-translational phosphorylation way and Pi3k mRNA is regulated by the post-transcriptional way. Future research is needed to investigate the association between these two-dimensional gene expression networks regulating pathways. Furthermore, once the m6A modification is methylated or demethylated, the biological function of post-transcriptional RNA metabolism is mediated by “reader” proteins selectively recognizing the m6A site containing the “RRACH” motif (40, 41). However, in the current study, it is unclear which “reader” is involved in the recognition of Pi3k mRNA, and future research is needed to clarify this issue. The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material. The animal study was reviewed and approved by the Animal Care and Utilization Committee at Zunyi Medical University. BS, RZ, and YZ designed the study. YZ, AX, CL, and XL completed most of the experimental process. ZB, ZQ, and WX carried out parts of the experiments. NG and YS analyzed the data and performed the statistical analysis. BS, RZ, and YZ wrote and revised the manuscript, with contributions from AX, CL, XL, ZB, ZQ, WX, NG, and YS. All authors reviewed and approved the final manuscript.

PMC9649651

Mohamed S. Hasanin,Mahmoud E. Abd El-Aziz,Islam El-Nagar,Youssef R. Hassan,Ahmed M. Youssef

Green enhancement of wood plastic composite based on agriculture wastes compatibility via fungal enzymes

10-11-2022

Biotechnology,Materials science

This study deals with the production of natural fiber plastic composites (NFPCs) to reduce environmental pollution with agricultural and plastic waste. Where the NFPCs were prepared from waste/pure polyethylene (WPE) (pure polyethylene (50%)/recycled polyethylene (50%)) and modified sunflower waste via an eco-friendly and economic biological process. The sunflower fibers (SF) were treated via whole selective fungal isolate, namely, Rhizopus oryzae (acc no. OM912662) using two different incubation conditions; submerged (Sub), and solid-state fermentation (SSF) to enhance the fibers' compatibility with WPE. The treated and untreated fibers were added to WPE with various concentrations (10 and 20 wt%). The morphology and structure of fibers were characterised by a scanning electron microscope (SEM) and attenuated total reflection-Fourier transform infrared (ATR-FTIR). Furthermore, the mechanical properties, morphology, biodegradation and water vapour transmission rate (WVTR) for the prepared NFPCs were investigated. The results showed that compatibility, mechanical properties and biodegradation of NFPCs were improved by the addition of sunflower fibers treated by SSF conditions.

Green enhancement of wood plastic composite based on agriculture wastes compatibility via fungal enzymes This study deals with the production of natural fiber plastic composites (NFPCs) to reduce environmental pollution with agricultural and plastic waste. Where the NFPCs were prepared from waste/pure polyethylene (WPE) (pure polyethylene (50%)/recycled polyethylene (50%)) and modified sunflower waste via an eco-friendly and economic biological process. The sunflower fibers (SF) were treated via whole selective fungal isolate, namely, Rhizopus oryzae (acc no. OM912662) using two different incubation conditions; submerged (Sub), and solid-state fermentation (SSF) to enhance the fibers' compatibility with WPE. The treated and untreated fibers were added to WPE with various concentrations (10 and 20 wt%). The morphology and structure of fibers were characterised by a scanning electron microscope (SEM) and attenuated total reflection-Fourier transform infrared (ATR-FTIR). Furthermore, the mechanical properties, morphology, biodegradation and water vapour transmission rate (WVTR) for the prepared NFPCs were investigated. The results showed that compatibility, mechanical properties and biodegradation of NFPCs were improved by the addition of sunflower fibers treated by SSF conditions. In the past few years, timber has been the main material for many industries such as the manufacturing of paper, furniture, and packaging boxes, but the increasing need for timbers has become a major contributor to the evanescence of forests. It is known that forests play important role in contributing to climate change mitigation, where they consume approximately 2.6 billion tons of carbon dioxide (CO2), which is one-third of the CO2 released from burning fossil fuels. Also, agricultural waste is considered a cumbersome by-product, where its disposal by landfilling or incineration is seen as a non-valid solution to the environment. However, agricultural residues are rich in cellulose materials that can be used in various applications as an alternative to forest wood. Moreover, the accumulation of plastic waste (e.g. plastic bottles and bags) in the environment adversely affects wildlife and aquatic life as well as humans. It is known that plastics are cheap and durable, which makes them suitable for different uses, but, due to their chemical structure being resistant to many natural processes of degradation, they are non-degradable. Wood-plastic composites (WPCs) are considered a subset of a huge class of matter called natural fiber plastic composites (NFPCs). Wood-plastic composites are produced from the fiber or the flour of wood and thermoplastic, while NFPCs are composites of thermoplastic polymers that contain pulp fibers, straw, peanut hulls, coffee husk, and bamboo as a filler. The former could be manufactured from agriculture and plastic waste, which is considered an environmentally friendly approach to using agricultural waste and recycled plastic material. Adding cellulosic materials obtained from agricultural waste to plastic materials to prepare WPC enhances their mechanical properties, durability and biodegradability. In addition, WPCs have beneficial characteristics such as low density and cost, durability, high strength, as well as excellent sound-absorbing capacity. So it could be utilized in various applications such as railway coaches, packaging and building. Unfortunately, the poor compatibility between fibers (hydrophilic) and polymer (hydrophobic) plays an important role in the properties of produced WPC. Indeed, the compatibility between cellulose and hydrophilic plastics is the main problem facing researchers. They, therefore, tend to apply various strategies to overcome this obstacle, which include specific modifications for cellulosic fiber surfaces to minimize their hydrophilicity either by physical or chemical modifications. The volume of sunflower production is about 11 million tons, and the volume of waste generated from its cultivation is about one million tons. Sunflower seeds are used in the production of oil, while the waste is used in many industries, such as livestock feed, compost (alternative soil), water treatment, biodiesel, etc.. Sunflower waste was riched by lignocellulosic fibers with high content of lignin about 25%, and so it is difficult to be degradable. Some microorganisms are cabaple to production of enzymes that is able to attack the lignin. Mateusz et al. showed that the addition of sunflower husk to epoxy had a huge effect on the deterioration of flexural strength and tensile of the prepared composite. Physical treatment of natural fibers includes grinding, thermal and radiation. Grinding is usually used to cut fibers into very small and homogeneous lengths. The thermal treatment is used to change the fibers chemical composition as well as extract of the active components of fibers as a dual-role treatment method. While the radiation treatment includes ultrasonic, plasma, and irradiation treatments. The chemical treatment removes non-desirable components or adds a functional group to improve the compatibility between the fibers and polymer by decreasing the hydrophilicity of the fibers. The chemical modification can also be carried out using benzene diazonium salt, sodium hydroxide and dodecane bromide, or esterification using different fatty acids. In addition, the biological treatment could be used as a tool to reactivate the lignocellulosic fibers surface via purified or crude enzymes as well as other in-suite fermentation conditions using microbial costive agents directly on the substrate with Sub or SSF. The in-suite treatment offers an economic and eco-friendly process in comparison with other tools with aspects of biological treatment. In the current work, Sub and SSF conditions were used to activate the surface of the sunflower fibers and modified to become suitable for contact and compactable with a hydrophobic plastic surface. The prepared natural fiber plastic composites (NFPCs) could be suitable for many applications such as alternative wood, household equipment, and packaging. All chemicals (Sodium nitrate, dipotassium phosphate, magnesium sulfate, potassium chloride, cetyltrimethylammonium bromide (CTAB) and ferrous sulfate) and media (potato dextrose agar (PDA) medium, potato dextrose broth (PDB) medium and mineral salts medium (MSM)), as well as reagents (3,5-dinitro salicylic acid, pyrogallol, glucose and xylose), were purchased from Sigma-Aldrich in analytical grade without any purification before useing. Pure low density polyethylene (LDPE) was food grade (density 0.93 g/cm3, softening point 87.4 °C, and melt flow index (190 °C, 2.16 kg) 6.0 g/10 min), and was obtained from ExxonMobil Chemical (Kingdom of Saudi Arabia) as pallets with particles size ranged 2–5 mm. The polyethylene waste was obtained as pallets with particles size ranged 4–8 mm from Bekia, Egypt. Sunflowers trimmings, as a source of lignocellulosic fibers, were collected from a farm located in El-Menofia Governorate, Egypt, which has considerable amounts of agricultural waste. They were dried in the oven at 70 °C to attain a constant weight and then ground mechanically and sieved with a 200 mesh sieve. The collection of plant material complies with the guidelines of the Ethics Committee in the National Research Centre. Samples for isolation were compiled from the soil of the sunflower farm in Giza, Egypt. Fungal isolation was performed according to our previous work. The isolated fungal was subjected to the cultivation on the replaced Carbone source Czapic broth medium by the fine powder of sunflowers as the sole Carbone source. The selected isolated fungal (Rhizopus oryzae) was observed with the highest biomass growth. The selected ligninolytic fungus was identified according to its morphological characteristics and 18s ribosomal DNA (18S rDNA) sequence according to our previous work. The morphological characteristics were examined using a light microscope (Olympus cx41) after 4 days of growth on PDA medium plates via a light microscope at a magnification of 40×. For molecular identification, fungal mycelium from a 4-day-old culture in PDB medium was harvested using Whatman No. 1 filter paper. The total genomic DNA was extracted using the CTAB protocol. The identification was achieved by comparing the contiguous 18S rDNA sequence with data from the reference and type strains available in public databases GenBank using the BLAST programme (National Center for Biotechnology Information). The obtained sequences were aligned using the Jukes Cantor model and the isolate was registered in GenBank. The sunflower fibers were treated using isolated fungi according to two different incubation conditions. The Sub and SSF conditions were used with the same condition except for the ratio of nutritional medium and fibers. The fungal isolate was cultivated using modified Czapek broth media. The carbon source of the previous media was changed with the lignocellulosic fibers. The submerged fermentation conditions were carried out using 1:25 fibers to medium. The solid state fermentation ratio is 1:5 fibers to medium. The incubation conditions were used for both types of fermentation as follows; temperature 25 °C, initial pH 5.5, and flask volume (1:5) in static condition for 10 days. The liquid filtrate from different incubation times (7 days) old fungal cultures cultivated on PDB medium or liquid MSM amended with waste as the sole carbon source were collected and applied directly as the crude enzyme in enzymes assay experiments. The activity of cellulase and xylanase enzyme was carried out via reducing sugars estimations using 3,5-dinitro salicylic acid (DNS) assay with glucose for cellulase and xylose for xylanase. Lignin peroxidase activity was determined spectrophotometrically at 420 nm. All appeared values are the average of triplicate experiments. The fibers were dried at 60 °C for 12 h to remove the moisture content. Also, the recycled PE was washed with distilled water and dried at 60 °C overnight. The NFPCs were synthesized from pure LDPE and recycled PE (WPE; 50:50 wt%) which was loaded with untreated fibers, and treated fiber either by Sub and SSF condition at concentration 10 wt% to obtain NFPC10%SF, NFPC10%Sub, and NFPC10%SSF, respectively, and at concentration 20 wt% to obtain NFPC20%SF, NFPC20%Sub, and NFPC20%SSF, respectively. The composition of the prepared samples is shown in Table 1. The PE was melted in a twin-screw extruder (Haake RheomexTW100, USA) at 140 °C and a rotation speed of 60 rpm. After the PE was melted, the sunflower fibers were added and mixing was continued for an additional 10 min. The mixture was collected and left for cooling then cut into small pieces suitable for feeding into a stainless-steel pressure mould to make plates with dimensions 8 × 10 cm and 0.5 mm thick. The moulding was done using pressure (5 MPa) at 150 °C and for 5 min. after that; the plates were left to cool in the piston to room temperature under pressure. Three replicates were performed for each blend. The Attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR) of samples was measured on the Shimadzu 8400S FT-IR Spectrophotometer in the range of 500–4000 cm−1. The samples were measured as films. In addition, a scanning electron microscope (SEM; JSM 6360LV, JEOL/Noran) was used to study the surface morphology of the prepared samples. The mechanical characterization was carried out for three replicates of each prepared NFPCs to gauge the tensile strength, elongation, Young’s modulus of bending, and modulus of rupture according to the ASTM D638-91 utilizing a testing machine LK10k (Hants, UK) fitted with a 1kN load cell and operated at a rate of 5 mm/min. Biodegradation of NFPCs in soil was done a corroding to Dalev et al.. The soil was taken from the surface layer, and then all inert materials were removed to obtain a homogeneous mass. About 100 g of soil was poured into a plastic pot up to a thickness of about 3 cm. The prepared samples were accurately weighed and dried for 24 h at 50 °C, then were buried in the pots to a depth of 1 cm. Water was sprayed once a day to sustain the moisture. The samples were weighed weekly for 4 weeks after being washed with distilled water and dried at 50 °C for 24 h. The biodegradation was carried out for three replicates for each sample. WVTR was carried out using (GBI W303 (B), Water Vapor Permeability Analyzer; China) via the cup method for three replicates of each prepared NFPCs. The WVTR was designed by way of the quantities of water vapor transferred through a unit area of the NFPCs film in a unit time inferior to particular conditions of temperature (38 °C) and humidity (4%) as specified through the following standards (ASTM E96). The isolated fungal was identified via molecular techniques and morphological characterization. Figure 1a showed that the colonies of isolated fungal contained on the surface of the PDA medium a yellow-green spore on the upper surface and reddish-gold on the lower surface. The genomic DNA of the isolated fungal was extracted and identified via molecular techniques where 18s rRNA amplification was applied. The obtained sequences were compared with related sequences on the National Centre for Biotechnology Information (NCBI), Egypt database. It was found that it was closely associated with Rhizopus oryzae (acc no. OM912662) which confirms their morphological identification. Figure 1b illustrated the phylogenetic tree with a high similarity of about 91.37%. The fungal isolate was subject to two different cultivated conditions due to the ratio between fibers and liquid medium. The quantitative activity of lignocellulolytic enzyme clusters was assayed and shown in Fig. 2. All assayed enzymes in the SSF condition were confirmed as high value in comparison with the other ones produced in the submerged conditions. These results are referred to the harsh conditions which made the isolated fungi produce all enzymes possible to provide the carbon requirements needed to grow. On contrary, in the submerged condition, the fungal isolate grows in normal conditions with a high water content which has many soluble simple carbons dissolved from the fibers. These phenomena induced heavy enzyme production in solid-state fermentation conditions as well as feedback on the production of the enzymes in case of submerged incubation conditions. Moreover, the high productivity of the lignocellulolytic enzymes lead to fiber surface modifications and made it more compatible to interact with polymer plastic where these enzymes increase the hydrophobicity of the fibers. The lignin oxidative enzymes, including lignin peroxidase, polyphenol oxidase and laccase act as natural surface activators where the oxidative effect of these enzymes act as a surface modification process as well as increase surface compatibility to plastic polymer. Whereas, the enzymes eliminated the fine terminal of the fiber molecular structure which made fibers smoother as well as reduced the free active hydroxyl groups and so more hydrophobic. The ATR-FTIR spectra of sunflower fibers and treated ones were shown in Fig. 3. In general, all the untreated and treated fibers showed the broad band corresponding to hydroxyl group (–OH) that stretched at 3400 cm−1, C–H stretched at 2900 cm−1, C–O–C stretched at 1125 cm−1 and C=C stretched alkene at 1610 cm−1. While the band of C–OH that stretched at 1435 cm−1 was present in untreated fibers and disappeared completely in SSF. The disappearance of these bands could have been caused by the partial removal of hemicellulose and lignin from fibers during the fungal treatment. The fungal treatment reduces the hydroxyl groups and so high water resistance through a reaction with enzymes. This led to a decrease in hydrogen bond that increased the intensity of the peak between 3300 and 3500 cm−1 bands in treated fibers compared to untreated fibers. Werchefani et al. observed that fibers treated by enzymes have fibers lower in diameter and length as well as high water resistance as a result of elimination of the hydrophilic components (lignin, pectin and hemicellulose). The sunflower fibers consist of many components each of them having a unique performance in SEM imaging. Figure 4 illustrated the effect of the fungal different fermentation conditions on the fibers' surface morphology in comparison with the blank one. The blank and treated fibers observed a significant difference in the surface morphology level which may affect the compatibility of fibers and plastic behaviour. The blank fibers (Fig. 4a) are clear with a typical lignocellulosic fibers performance as smooth surface fibers loaded with some impurities observed as aggregated crystals. Otherwise, the SSF fibers surface in Fig. 4b observed with many pours located overall fibers surface as well as no smooth appearance enough in comparison with a blank one. In addition, the Sub condition fibers are shown as blank fibers with low smoothness, and the diameter of the fibers was clearly reduced. These observations may be related to the effect of SSF conditions in which the productivity of the fungal enzyme is higher than in the Sub condition according to harsh conditions. Moreover, in the SSF condition, the fungal hypha has penetrated the fibers to achieve the nutrients in harsh conditions (Fig. 4c). These results affirmed that the enzyme productivity, as well as fungal hypha biomass, are increased in the harsh conditions in which the fungal strain combat to gain the required nutrients and survive via degradation of the fibers. In contrast, the high humidity condition in the Sub case made fungal strains grow at a normal rate. Overall, the revised conclusion was concise that the treated fibers may affect the compatibility with plastic polymer and made the fibers attached strongly with plastic polymer in comparison with the blank fibers. Table 2 represents the effect of untreated and treated sunflower fibers content in the prepared wood plastic composites on the tensile, elongation, Young’s modulus of bending, and modulus of rupture. In general, the addition of sunflower fibers to NFPCs decreased the tensile and elongation of the prepared wood plastic composites in comparison with pure NFPCs. However, the tensile and the elongation of NFPCs containing sunflower fibers modified by the SSF condition were higher than that modified by Sub condition or unmodified fibers. This may be due to the high compatibility between modified fibers (SSF) and the polymer, which was higher than modified fibers (Sub) than unmodified sunflower fibers and the polymer. It was found that the content of untreated and treated sunflower fibers increases Young’s modulus of bending NFPCs compared to blank PE related to the increase of the elasticity. These results affirmed the elongation results that emphasised that the addition of fibers was lead to decreases in the elasticity where the relation between elongations, Young’s modulus is inverse. In addition, the treated fiber via SSF conditions is more effective on Young’s modulus (1032.1 MPa) of NFPC that contains 20% SSF. Also, the NFPC contains fiber treated by SSF conditions with a content of 20 wt% improves the modulus of rupture to be 18.8 MPa compared to other fibers, which are near to blank PE 19.7 MPa. This may be due to the compatibility between treated sunflower fibers and the polymer higher than that between untreated sunflower fibers and the polymer. In the same way, Sobczak et al. reported that as the degree of surface roughness increases in the fiber, the area available for interactions with the matrix will be increased. This leads to improving the mechanical performances of the prepared composites. The surface morphology of the prepared NFPCs has illustrated in Fig. 5. The surface morphology of blank PE films was smooth and showed high homogeneity between their polymeric chains (Fig. 5a). This homogeneity was decreased in WPE (50% pure PE/50% recycled PE) as shown in Fig. 5b due to the difference in the polymeric chain's length between pure PE and recycled one, in other words, due to the difference in the molecular weight between pure PE and recycled one. Also, the presence of different additives, impurities, or traces of other polymers in the polymer waste significantly affected on the melt flow index and the homogeneity between virgin and recycled polymer as reported by Patrizio et al.. The loading of WPE by untreated and treated sunflower fibers displayed a rough structure which was decreased in the case of treated fibers, especially with that treated by SSF condition (Fig. 5g). This indicates the happening of compatibility between polymer matrix and treated sunflower fibers. Additionally, the content of fiber in the matrix has a great effect on the surface morphology and so their internal construction. Indeed, the surface of untreated fibers is enveloped by lignin and some impurities of terminal groups of cellulose and hemicellulose which is the reason for decreasing the adhesion and so the compatibility between the fibers and WPE (Fig. 5c,d). Whilst, the treated fibers either submerged (Fig. 5e,f) or solid-state fermentation (Fig. 5g,h) showed higher compatibility with polymer. These results may be due to a decrease in the proportion of lignin via oxidization of its phenolic groups via peroxidases enzymes as well as the elimination of terminal side group of cellulose and hemicellulose and their decrease the hydrophilicity via acting the eradication of free hydroxyl groups and these observasions are in a nice agremment with other studies, especially those treated by the SSF method as shown in Figs. 3 and 4. Corradini et al. showed that the compatibility between recycled poly(ethylene terephthalate) and sugarcane bagasse fiber was increased by the addition of ethylene/n-butyl acrylate/glycidyl methacrylate copolymer as compatibilizing agent that worked to raise the interaction between the components. Also, Chen et al. illustrated that the alkali treatment of sugarcane bagasse increases their compatibility with high-density polyethylene than the untreated one. In the last decay, biodegradability is an important feature for any new component especially the synthetic polymer that has taken a century to start degradation. The addition of natural fibers to synthetic polymers made it more acceptable to degrade in the environment. The biodegradation of the WPE and the prepared NFPCs was illustrated in Fig. 6. The obtained results investigated that the WPE is not degraded in the soil after a complete 4 weeks. The biodegradability properties of polymers are a key factor to consider this polymer is environmentally friendly and not accumulated over. PE is not a biodegradable polymer that is not biodegraded in the environment naturally while taken to complete degradation many years. Otherwise, natural fibers are biodegradable and environmentally friendly. In this context, cellulose has a degradation half-life (t1/2) in the soil at 10–20 °C between 30 and 42 days. Moreover, after 2 months cellulose was decomposed into CO2 and water. The fast degradation rate of natural fibers is attributed to the breakdown of cellulose bonds randomly as the effect of the microorganism cleavage. In this work, the effect of the addition of the sunflower fibers on the biodegradability rate of PE in soil was studied and the results are presented in Fig. 6. The obtained results were observed that the sunflower fibers significantly affect the biodegradation of the polymer in soil. Whereas, LDPE offered resistance to biodegradation with lost about 0.18% of its weight after being buried in soil for 4 weeks. Additionally, the sunflower fibers enhanced the composite biodegradability where NFPCs containing 10 and 20 wt% of untreated sunflower fibers (SF) lost about 5.2 and 6.9%, respectively, after the same period buried in the soil. On the other hand, the fungal enzymes treated fibers (Sub) observed a slight increase in biodegradability rate in comparison with blank ones. However, NFPCs containing 10 and 20 wt% of treated sunflower fibers (SSF) lost about 6.4 and 11.4%, respectively, after the same period buried in the soil. These increases in the biodegradability rate of treated fibers composite in soil may be related to the removal of undesirable sunflower fibers constituent that made the fiber easy to microbial attack. These results are in agreement with previous findings on the biodegradability of non-degradable polymer reinforced with sunflower fibers where enhanced tensile strength properties as well as biodegradability. Additionally, the addition of PE did not prevent the biodegradability of sunflower fibers as well as the biodegradation of lignocellulosic material with time which may be made the degradation of PE easy in comparison with the pure one. Food degradation is meaningfully affected by the WVTR of packaging materials. It designates together the solubility of water molecules as well as the transfer of water molecules into packaging materials. Hence, the materials' high permeability to water vapor has a straightforward impact on their usage in packaging applications attributable to their capability to alter the block humidity between products and their neighboring air. Several issues contain chemical structure (high solubility for selecting polymer), size, molecular weight, etc. require an influence on moisture content as well as molecule mobility in the packaging films. Table 3 represents the values of the WVTR of the pure LDPE, WPE, NFPC10%SF, NFPC20%SF, NFPC10%Sub, NFPC20%Sup, NFPC10%SSF, and NFPC20%SSF were shown in (Table 3). The achieved data shows that the WVTR rises in the fabricated NFPCs films with the addition of different ratios of treated fibers (SF, SSF, and Sub fiber) from 10 to 20%. Where the WVTR was increased from 2.73 g/(m2 day) for pure LDPE to 4.99 g/(m2 day) in the case of using (50:50) percentage of pure LDPE to recycled PE (WPE). Moreover, the WVTR was increased from 4.99 g/(m2 day) to 26.37 g/(m2 day) in case of using 10% of (SF) fibers. Also, the WVTR increased from 4.99 g/(m2 day) to 34.64 g/(m2 day) in the case of using 20% of (SF) fibers. By using from 10 to 20% of treated fibers (Sub) for the fabrication of the NFPC film, the WVTR increased from 16.60 g/(m2 day) to 29.67 g/(m2 day). Furthermore, by usage from 10 to 20% of treated fibers (SSF) for fabrication of the NFPC film, the WVTR increased from 12.58 g/(m2 day) to 15.37/(m2 day). The principal causes for the rise of WVTR with the addition of cellulose fibers are the hydrogen bonds formed between (OH groups) of fibers and the polymer matrix. The increase in WVTR at higher SF concentration is related to the pore network and structure of the NFPCs films. It was realized that, generally, the WVTR improved with increasing treated fibers (Sub and SSF) ratios in the NFPCs samples. Moreover, the enhancement in the WVTR at higher humidity levels is associated with the improved passage of moisture. This phenomenon is convinced by the transmission of water molecules in the microscopic pores of the fiber material which are filled with water because of capillary condensation. For ingredients that display hysteresis in their sorption isotherm, it has been described previously that their WVTR is dependent on moisture content. We detected that there is a small increase in water vapor transmission rate as the percentage of different ratios of treated fibers (Sub and SSF) from 10 to 20% increases. This is as the percentage composition of modified cellulose increases, the hydrophilicity of the NFPCs films increases. This phenomenon could be associated with the significant hydrogen bonding interaction with water. The current manuscript established an inexpensive as well as sustainable approach for fabricating wood plastic composites from recycled PE waste as binding matrix and agricultural waste such as modified sunflower waste as filled materials through a biological unusual procedure designated as a green eco-friendly, and economic method. The sunflower fibers were treated via whole selective fungal isolate, Rhizopus oryzae (acc no. OM912662), to modify the fibers' surface and to improve their compatibility with polymer plastic by increasing the hydrophobicity of the fibers. The mechanical properties were improved by the addition of both forms of modified sunflower (Sub), and (SSF). Moreover, the untreated and treated sunflower fibers increase Young’s modulus of bending NFPCs compared to blank PE. Furthermore, the treated fiber via SSF conditions is more effective on Young’s modulus (1032.1 MPa) of NFPCs that contains 20% SSF. As well, the fiber treated by SSF conditions with a content of 20%/wt develops the modulus of rupture to be 18.8 MPa compared to other fibers, which are near to blank PE 19.7 MPa. The addition of modified sunflower waste fiber enhanced the composite biodegradability where NFPCs containing 10 and 20 wt% of untreated sunflower fibers (SF) lost about 5.2 and 6.9%, respectively, while the treated fiber by Sub condition showed a slight increase in biodegradability. However, NFPCs containing 10 and 20 wt% of treated sunflower fibers SSF condition lost about 6.4 and 11.4%, respectively, after the same period buried in the soil. Moreover, WVTR increases in the prepared NFPCs films with the addition of different ratios of various fibers from 10 to 20%. Thus, the fabricated wood plastic composites might be appropriate for many applications, e.g. alternative wood, household equipment, as well as packaging.

PMC9649652

Dingyuan Tu,Chaoqun Ma,ZhenYu Zeng,Qiang Xu,Zhifu Guo,Xiaowei Song,Xianxian Zhao

Identification of hub genes and transcription factor regulatory network for heart failure using RNA-seq data and robust rank aggregation analysis

28-10-2022

heart failure,RNA-seq dataset,RUVSeq,robust rank aggregation,hub gene,biomarker,functional enrichment analysis,transcription factor

Background Heart failure (HF) is the end stage of various cardiovascular diseases with a high mortality rate. Novel diagnostic and therapeutic biomarkers for HF are urgently required. Our research aims to identify HF-related hub genes and regulatory networks using bioinformatics and validation assays. Methods Using four RNA-seq datasets in the Gene Expression Omnibus (GEO) database, we screened differentially expressed genes (DEGs) of HF using Removal of Unwanted Variation from RNA-seq data (RUVSeq) and the robust rank aggregation (RRA) method. Then, hub genes were recognized using the STRING database and Cytoscape software with cytoHubba plug-in. Furthermore, reliable hub genes were validated by the GEO microarray datasets and quantitative reverse transcription polymerase chain reaction (qRT-PCR) using heart tissues from patients with HF and non-failing donors (NFDs). In addition, R packages “clusterProfiler” and “GSVA” were utilized for enrichment analysis. Moreover, the transcription factor (TF)–DEG regulatory network was constructed by Cytoscape and verified in a microarray dataset. Results A total of 201 robust DEGs were identified in patients with HF and NFDs. STRING and Cytoscape analysis recognized six hub genes, among which ASPN, COL1A1, and FMOD were confirmed as reliable hub genes through microarray datasets and qRT-PCR validation. Functional analysis showed that the DEGs and hub genes were enriched in T-cell-mediated immune response and myocardial glucose metabolism, which were closely associated with myocardial fibrosis. In addition, the TF–DEG regulatory network was constructed, and 13 significant TF–DEG pairs were finally identified. Conclusion Our study integrated different RNA-seq datasets using RUVSeq and the RRA method and identified ASPN, COL1A1, and FMOD as potential diagnostic biomarkers for HF. The results provide new insights into the underlying mechanisms and effective treatments of HF.

Identification of hub genes and transcription factor regulatory network for heart failure using RNA-seq data and robust rank aggregation analysis Heart failure (HF) is the end stage of various cardiovascular diseases with a high mortality rate. Novel diagnostic and therapeutic biomarkers for HF are urgently required. Our research aims to identify HF-related hub genes and regulatory networks using bioinformatics and validation assays. Using four RNA-seq datasets in the Gene Expression Omnibus (GEO) database, we screened differentially expressed genes (DEGs) of HF using Removal of Unwanted Variation from RNA-seq data (RUVSeq) and the robust rank aggregation (RRA) method. Then, hub genes were recognized using the STRING database and Cytoscape software with cytoHubba plug-in. Furthermore, reliable hub genes were validated by the GEO microarray datasets and quantitative reverse transcription polymerase chain reaction (qRT-PCR) using heart tissues from patients with HF and non-failing donors (NFDs). In addition, R packages “clusterProfiler” and “GSVA” were utilized for enrichment analysis. Moreover, the transcription factor (TF)–DEG regulatory network was constructed by Cytoscape and verified in a microarray dataset. A total of 201 robust DEGs were identified in patients with HF and NFDs. STRING and Cytoscape analysis recognized six hub genes, among which ASPN, COL1A1, and FMOD were confirmed as reliable hub genes through microarray datasets and qRT-PCR validation. Functional analysis showed that the DEGs and hub genes were enriched in T-cell-mediated immune response and myocardial glucose metabolism, which were closely associated with myocardial fibrosis. In addition, the TF–DEG regulatory network was constructed, and 13 significant TF–DEG pairs were finally identified. Our study integrated different RNA-seq datasets using RUVSeq and the RRA method and identified ASPN, COL1A1, and FMOD as potential diagnostic biomarkers for HF. The results provide new insights into the underlying mechanisms and effective treatments of HF. Heart failure (HF) is a complex clinical syndrome that results from dysfunction of ventricular filling or ejection, characterized by a variety of worsening symptoms and signs, including dyspnea, fatigue, and fluid retention (1). The occurrence of HF is predominantly caused by underlying myocardial diseases, while cardiac lesions from valves, vasculature, pericardium, heart rate/rhythm, or a combination of cardiac abnormalities may also result in cardiac malfunction (2). Despite the development of drug therapy and surgical interventional therapy, the morbidity and mortality of HF are increasing annually worldwide, which seriously threatens human health and quality of life (3, 4). Therefore, to improve the curative efficacy, it remains urgent to investigate the in-depth underlying molecular mechanisms of HF to facilitate its accurate diagnosis, early intervention, and precision therapy. In recent years, the rapid progress of transcriptome sequencing technology provides new directions for the exploration of epigenetic changes and molecular mechanisms in different diseases, including neoplastic and non-neoplastic diseases (5, 6). Accordingly, an increasing volume of RNA sequencing (RNA-seq) and microarray datasets of HF has been uploaded in the Gene Expression Omnibus (GEO) database, providing opportunities for bioinformatics data mining of marker genes associated with HF (7). However, in comparison to cancer-related surgery, the number of heart transplantation surgeries is relatively small, which results in the small sample size and large batch effects of RNA sequencing or microarray datasets of HF. Therefore, to date, the bioinformatics data mining of HF still faces great challenges, especially regarding the integration and analysis of the RNA-seq data (RUVSeq) related to HF. The robust rank aggregation (RRA) method, first proposed in 2012 by Kolde et al., is a rigorous approach using probabilistic models to analyze the significant probability of all elements in different sequencing or microarray datasets (8). Recently, the RRA algorithm has been extensively used to integrate data in different microarray platforms to screen the differentially expressed genes (DEGs) in multiple diseases, including thyroid carcinoma (9), prostate cancer (PCa) (10), and DCM (11). For example, Song et al. utilized the RRA method to integrate 10 eligible PCa microarray datasets from the GEO and identify four candidate biomarkers for prognosis of PCa (10). Ma et al. integrated four eligible dilated cardiomyopathy (DCM) microarray datasets from the GEO database and developed a 7-gene signature predictive of DCM by utilizing the RRA method (11). However, due to the greater difficulty in integrating sequencing data, the application of the previous RRA algorithm was limited to microarray data, and the RRA analysis of RUVSeq was still rarely reported. Removal of Unwanted Variation from RUVSeq, a Bioconductor package that generalizes a linear model to regress variance estimated from the expression of housekeeping genes, has been reported to be used to reduce batch effects due to different sequencing conditions (12), which provides a huge possibility for the combination of RUVSeq and the RRA method in integrating different RUVSeq sets and identifying HF-associated DEGs. In the present study, RUVSeq and RRA analysis were performed for the first time based on four RNA-seq datasets in the GEO database to identify robust DEGs in HF samples and non-failing donor (NFD) samples, followed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis for the DEGs. Moreover, three reliable HF-related hub genes with differential expression and excellent diagnostic efficiency, ASPN, COL1A1, and FMOD, were selected and validated using microarray datasets and human heart tissue assays. Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were further utilized to investigate potential functions of the hub genes. In addition, the transcription factor (TF)–DEG regulatory network was constructed based on the HF datasets and websites. The GEO database was searched to obtain the sequencing datasets based on the search terms of “heart failure” or/and “HF.” The search results and relevant datasets were filtered according to the following inclusion criteria: (i) the organism was filtered by “homo sapiens”; (ii) the study type was set as “expression profiling by high throughput sequencing”; (iii) RUVSeq for both HF samples and NFDs should be included in the dataset; (iv) the total number of samples should not be < 5; and (v) the raw data of the RNA-seq should be provided for reanalysis. Datasets that did not meet the aforementioned criteria were excluded. The selected HF sequencing datasets from the NCBI Sequence Read Archive (SRA) were downloaded as SRA files and converted to FASTQ files via the SRA toolkit. To obtain high-quality reads, raw data from the GEO dataset were pre-processed using the fastp tool (13), and sample quality was assessed by FastQC and MultiQC (14). The sequences were then aligned against the human reference genome hg38 using STAR (15). Furthermore, the expression values (count matrices) for either gene bodies or called peaks were generated by featureCounts (16). For RNA-seq expression analysis, batch effects were adjusted using the R package RUVSeq, which applies a generalized linear model to regress out the variation estimated from the expression of the housekeeping gene. First, the initial DEGs were detected using the edgeR program package within a single RNA-seq dataset. Second, the RUVg function in RUVSeq was utilized to remove additional sources of unwanted variation (parameter k = 1) (17). The remaining non-DEGs were considered as negative control genes and used as housekeeping genes to correct for relative gene expression levels between different samples. Third, based on the corrected gene expression matrix, the corrected DEGs were further obtained by the edgeR package. Fourth, the RRA method-based R package “RobustRankAggreg” was used to integrate the results of RUVSeq analysis of each RNA-seq dataset to identify the final DEGs in patients with HF compared with NFDs. The threshold of DEGs was set as |logFC| > 1, and the significance criterion was set as an adjusted p-value < 0.05. To further investigate the possible functions of DEGs identified by the RUVSeq and the RRA method, GO enrichment and KEGG pathway analyses were performed in the upregulated and downregulated DEGs separately, using the R package “clusterProfiler” (18). The GO term or KEGG pathway with adjusted p < 0.05 was considered with significant enrichment. The results were visualized by dot plots using the “dotplot” function of the R package. The robust DEG list was uploaded to the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (19), and the significant protein interaction was determined at the criterion of confidence (combined score) > 0.4. Next, we used Cytoscape software and cytoHubba (20) plug-in to investigate node composition and pick out hub nodes with a high degree of connectivity in the network. RNA-seq datasets for HF samples are limited due to a small volume of heart transplant surgeries and the difficulty in obtaining human heart samples. Therefore, in our study, the four eligible HF sequencing datasets (GSE46224, GSE116250, GSE133054, and GSE135055), including 95 HF and NFD samples, were all used for the identification of DEGs, hub genes, and functional enrichment analysis. To further validate the analysis results, HF microarray datasets were acquired from the GEO database. The inclusion criterion was identical to the RUVSeq sources, except that the study type was set as “expression profiling by array.” For the study, four microarray datasets were finally included for the validation: GSE16499 (21), GSE26887 (22), GSE57338 (23), and GSE79962 (24). The gene expression profiling was annotated using the annotation document of corresponding platforms, and the gene expression matrices were column-normalized by the R package “limma” (25) and log-transformed, if necessary. Next, the differential expression of the identified hub genes between patients with HF and NFDs in the microarray datasets was validated and visualized by column graphs. For further validation, total RNAs of the heart tissues from patients with HF and NFDs were extracted for the qRT-PCR validation assay. Heart tissues from six patients with HF and eight NFDs were obtained from the Specimen Bank of Cardiovascular Surgery Laboratory and Department of Pathology of Changhai Hospital, Shanghai, China. Written informed consents were obtained from all patients or their family members, and the study was approved by the Institute Ethics Committee of Changhai Hospital. Total RNAs from the heart tissues were isolated using TRIzol reagent (TRIzol™ Reagent, Invitrogen). RNAs were then reverse-transcribed into cDNAs using a TOYOBO ReverTra Ace® qRT-PCR RT Kit (TOYOBO, Japan). SYBR®GREEN (TOYOBO, Japan) was used for qRT-PCR, and the primer sequences used are listed as follows: asporin (ASPN) forward, 5′-GGGTGACGGTGTTCCATATC-3′ and reverse, 5′-TTGGCACTGTTGGACAGAAG-3′; collagen type I alpha 1 chain (COL1A1) forward, 5′-TCG TGGAAATGATGGTGCTA-3′ and reverse, 5′-ACCAGGTT CACCGCTGTTAC-3′; collagen type IX alpha 2 chain (COL9A2) forward, 5′-AAGAGCAACTGGCAGAGGTC-3′ and reverse, 5′-GACCCTCGATCTCCATCCTT-3′; collagen type X alpha 1 chain (COL10A1) forward, 5′-TGGGACCCCTC TTGTTAGTG-3′ and reverse, 5′-GCCACACCTGGTCA TTTTCT-3′; cartilage oligomeric matrix protein (COMP) forward, 5′-CAGGACGACTTTGATGCAGA-3′ and reverse, 5′-AAGCTGGAGCTGTCCTGGTA-3′; and fibromodulin (FMOD) forward, 5′-AGAGAGCTCCAT CTCGACCA-3′ and reverse, 5′-GCAGCTGGTTGT AGGAGAGG-3′. The expression levels of mRNAs relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were detected using the 2–ΔΔCt method. To further explore potential functions of the hub genes in HF, we performed GSEA and GSVA in the microarray dataset with the maximum HF sample size (GSE57338). The flow of GSEA is as follows: First, correlation analyses were conducted between hub genes and other genes in the gene expression matrix of 54 patients with HF, and genes with the absolute value of correlation coefficient > 0.5 and p-value < 0.05 were defined as hub genes-related genes. Then, KEGG pathway enrichment analysis was conducted on these hub genes-related genes using the ClusterProfiler package. For GSVA, 54 patients with HF in the GSE57338 dataset were divided into two groups based on the median expression level of each hub gene (high- and low-expression groups). Then, the ‘‘GSVA’’ package was used to explore the pathways associated with the hub genes. The annotated gene set ‘‘c2.cp.kegg.v7.4.entrez.gmt’’ in the Molecular Signatures Database (MsigDB) was selected as the reference. It has been reported that binding of TFs to the regulatory regions of genes is a key transcriptional regulatory mechanism to control chromatin and transcription, forming a complex system that guides expression of the genome (26). The TF–DEG regulatory network is constructed by using the following methods: First, the NetworkAnalyst database (27) and the TF–gene interactions module from the JASPAR database (28) were utilized to explore the possible TFs that could bind to the RRA-identified DEGs. Second, a novel significant TF–DEG regulatory pair was defined in our study according to the following criteria: (i) both the TF and DEG were present in the TF regulatory network constructed by the JASPAR database, and there was predicted interaction between them; (ii) the TF was differentially expressed in patients with HF and NFD samples in the validation set GSE57338 (p < 0.05); and (iii) there was a statistically significant relationship between the expression level of TF and its target gene in the validation dataset GSE57338 (the absolute value of correlation coefficient > 0.5 and p < 0.05). Third, the constructed TF–DEG regulatory network was visualized using Cytoscape. Independent two-sample t-tests were used to analyze variables with homogeneous variance and normal distribution, whereas Mann–Whitney non-parametric tests were used to analyze variables without homogeneous variance and normal distribution. P-values were adjusted for multiple testing by using the Benjamini–Hochberg method. The DEG threshold was set as |logFC| > 1, and the significance criterion was set as an adjusted p-value < 0.05. The hypergeometric test was used to calculate the statistical significance of enrichment analysis. An absolute value of the correlation coefficient |r| > 0.3 (p < 0.05) indicates a significant interaction relationship (29). All data analyses in the present study were performed by using R (version 3.5.3) and Rstudio (version 1.2.1335). Graphic representations were generated by using GraphPad Prism 9.0 (GraphPad, San Diego, CA, USA) and Cytoscape (Version 3.7.1). Figure 1 depicts the flow diagram of our study. After screening and exclusion according to the aforementioned criteria, six datasets from the GEO database were finally included in this analysis: GSE46224 (30), GSE48166, GSE116250 (31), GSE120852 (32), GSE133054 (33), and GSE135055 (34). The characteristics of these six datasets are summarized in Supplementary Table 1, including the GSE accession number, study country, number of patients with HF and NFDs, and sequencing platform. After the quality-filtering using the fastp tool, the reads with a base quality < 20 or the sequence length ≤ 36 nt were discarded. Then, FastQC was used to assess the sequence quality of the dataset. The final all-in-one quality control report of each dataset was generated using MultiQC. The per base sequence quality and per sequence GC content across all samples of 6 RNA-seq datasets are demonstrated in Figure 2. Reads were mapped to the human genome (UCSC, hg38) using STAR, and the unique alignments were filtered and presented in Supplementary Table 2. Samples from each dataset were characterized by principal component analysis (PCA) after normalization and adjustment for batch effects using the RUVSeq package. 2D plots of PCA distribution showed that complete separation between samples of patients with HF and NFD samples was observed in five datasets (GSE46224, GSE116250, GSE120852, GSE133054, and GSE135055), except GSE48166 (Figure 3). Hence, dataset GSE48166 was excluded from subsequent analysis. Next, the expression difference and diagnostic efficacy of the four cardiac function markers, namely, natriuretic peptide A (NPPA), natriuretic peptide B (NPPB), myosin heavy chain 6 (MYH6), and myosin heavy chain 7 (MYH7), were examined between samples of patients with HF and NFD samples in the five sequencing databases. As shown in Figure 4, the markers showed no differential expression and poor diagnostic performance between the two groups in dataset GSE120852, which was also eliminated from further analysis. Using the RUVSeq package, DEGs (patients with HF vs. NFDs) were screened for adjusted p < 0.05 and |logFC| > 1 in the four identified datasets, respectively, which were visualized by volcano plots (Figures 5A–D). Furthermore, an integrated analysis was performed using the R package “RobustRankAggreg” to generate the differentially expressed ranked gene list. A total of 201 highly ranked DEGs were identified in patients with HF vs. NFD samples, and Supplementary Table 3 exhibits the top 50 upregulated and the top 50 downregulated DEGs. The top 20 upregulated and the 20 most downregulated genes consistently expressed across all datasets were visualized by heatmap, as shown in Figure 5E. To explore the potential biological functions of these DEGs, GO term enrichment and KEGG pathway analyses were performed. The upregulated genes were significantly enriched in extracellular structure organization, skeletal system development, extracellular matrix organization, T-cell activation, and connective tissue development in the biological process (BP) term; the extracellular matrix, collagen-containing extracellular matrix, endoplasmic reticulum lumen, basement membrane, and collagen trimer in the cellular component (CC) term; and extracellular matrix structural constituent, glycosaminoglycan binding, heparin binding, growth factor activity, and extracellular matrix structural constituent conferring tensile strength in the molecular function (MF) term (Figure 6A). For the downregulated genes, the enriched GO functions included purine ribonucleotide metabolic process, coenzyme metabolic process, energy derivation by oxidation of organic compounds, cellular respiration, and citrate metabolic process in the BP category; organelle inner membrane, mitochondrial inner membrane, mitochondrial matrix, mitochondrial protein complex, and mitochondrial membrane part in the CC category; and cofactor binding, coenzyme binding, and NAD binding in the MF category (Figure 6B). Regarding KEGG pathway analysis, the MAPK signaling pathway, TGF-β signaling pathway, T-cell receptor signaling pathway, Th17 cell differentiation, and ECM–receptor interaction were mostly associated with the upregulated genes (Figure 6C), while the downregulated genes were most enriched in the calcium signaling pathway, carbon metabolism, valine, leucine and isoleucine degradation, citrate cycle, and propanoate metabolism (Figure 6D). The PPI network of the 201 DEGs in patients with HF was constructed by using the STRING database (Figure 7A). Next, the PPI network was loaded into Cytoscape to screen hub genes by degree using the cytoHubba plug-in. As shown in Figure 7B, genes in the inner concentric circles have higher degrees, while genes in the outer concentric circles have relatively lower degrees. Therefore, hub genes were the six genes with the highest degree of connectivity (degree ≥ 10) in the innermost concentric circle: COL1A1, COMP, ASPN, COL10A1, FMOD, and COL9A2. Furthermore, the relative expression of the identified hub genes in patients with HF and NFD samples was assessed in the four RNA-seq datasets. The results showed that COL1A1, ASPN, and FMOD were consistently upregulated in the HF samples of the four datasets (Figures 7C–F). In addition, univariate ROC analysis was performed to determine the diagnostic accuracy of independent hub genes, suggesting that COL1A1, ASPN, and FMOD had a good diagnostic value in HF (Figures 7G–J). After normalization, four microarray datasets (GSE16499, GSE26887, GSE57338, and GSE79962) containing human left ventricular samples of HF and NFDs were used to validate the expression of these hub genes (Supplementary Table 4 and Supplementary Figure 1). As shown in Figures 8A–D, the expression of ASPN or FMOD in HF samples was significantly higher than that in the NFD samples in all four datasets, and COL1A1 or COMP showed the similar upregulation in three datasets. However, the expression of COL9A2 or COL10A1 was not statistically different in the HF and NFD samples in these datasets. Consistently, the diagnostic values of the hub genes suggested by the ROC curves revealed the same trend (Figures 8E–H). In addition to the microarray datasets, the expression of hub genes was further validated by qRT-PCR experiments using 14 heart tissues from patients with HF or NFDs. As described in Figure 9, ASPN, COL1A1, and FMOD were significantly upregulated in the six heart tissues of patients with HF compared with NFDs. Taken together, these validation results confirmed the differential expression and diagnostic value of ASPN, COL1A1, and FMOD as reliable hub genes in HF development. To reveal the underlying mechanism of the three reliable hub genes (ASPN, COL1A1, and FMOD) involved in HF, GSEA was conducted to explore significantly enriched pathways associated with the hub genes in the validation dataset GSE57338. As shown in Figures 10A–C, the top three signaling pathways enriched by the DEGs between subgroups were identified, among which citrate cycle (TCA cycle) and propanoate metabolism pathways were significantly enriched in the subgroups of all the three hub genes. In addition, the enrichment in glucose metabolism-related pathways was further confirmed by GSVA (Figures 10D–F). To determine the potential roles of TFs in regulating the transcriptional expression of DEGs, the specific TF–gene regulatory network was established based on the top 20 upregulated and the 20 most downregulated integrated DEGs (Figure 11A). As demonstrated in Figures 11B,C, several TFs, including CEBPB, MEF2A, PPARG, BRCA1, TEAD1, TFAP2A, TP63, SREBF1, and PDX1, showed significant correlation with multiple DEGs and were differentially expressed in patients with HF and NFDs in GSE57338 (p < 0.05). According to the defining criteria of the significant TF–DEG regulation pair, we identified TP63-SERTM1/SYTL5/UNC80, PPARG-XG, BRCA1-NRG1, MEF2A-LSAMP, SREBF1-NPPA/HOOK1/CENPA, TEAD1-CA14/MYH6/PENK, and PDX1-SEC14L5 as significant TF–DEG pairs (Figure 11D). In the present study, four HF RNA-seq GEO datasets (GSE46224, GSE116250, GSE133054, and GSE135055) were finally included, involving a total of 100 patients with HF and 38 NFDs. In total, 201 robust HF-related DEGs were obtained utilizing RUVSeq and RRA method, and ASPN, COL1A1, COL9A2, COL10A1, COMP, and FMOD were identified as hub genes with the highest degree of connectivity using STRING database and cytoHubba plug-in. Among them, ASPN, COL1A1, and FMOD exhibited differential expressions and excellent diagnostic efficiency in all four RNA-seq datasets, which were further validated using data from the four screened HF microarray datasets (GSE16499, GSE26887, GSE57338, and GSE79962). Moreover, the significant upregulation of ASPN, COL1A1, and FMOD was experimentally confirmed by qRT-PCR using the heart tissues of patients with HF and NFD samples. In addition, functional enrichment analysis showed that myocardial fibrosis-related pathways resulted from T-cell-mediated immune response and myocardial glucose metabolism were closely associated with the onset and progression of HF. In addition to this, the TF–DEG regulatory network was established, and 13 significant TF–DEG pairs were identified. Despite the great advancement in HF medical treatment, it remains the major and growing public health problem that leads to considerable morbidity and mortality (35). Robust biomarkers for early diagnosis of HF are the key for novel targeted pharmacological approaches and for improving the prognosis of patients (36). Consistently, serum type B natriuretic peptide (BNP) has been recognized as a well-established biomarker for the diagnosis of HF. However, a recent study reported that a subset (4.9%) of hospitalized patients with confirmed HF had unexpectedly low BNP levels (<50 pg/ml), and a small proportion (0.1–1.1%) had BNP levels even below detection limits (37). Therefore, it remains urgent to explore novel molecules with potentially new mechanisms for the development of HF. Recently, gene mining using microarrays or RNA-seq datasets has been widely used to generate the transcriptomic profiles of HF development. Zhang et al. identified six hub genes (BCL3, HCK, PPIF, S100A9, SERPINA1, and TBC1D9B) as potential biomarkers of HF by using the weighted gene co-expression network analysis (WGCNA) method through three HF datasets, namely, GSE59867, GSE1869, and GSE42955 (38). Tian et al. constructed a random forest algorithm and artificial neural network and detected six hub genes by mining of two HF datasets (GSE57345 and GSE141910) (39). However, the aforementioned studies were based on the integration of DEGs, rather than raw data from different datasets. To date, the inconsistency between different platforms and datasets remains the major hurdle blocking the bioinformatics mining of HF-related genes, especially for RNA-seq datasets. The RRA method, a recently emerging analysis method, has been widely used to integrate different datasets and produce a ranked list of the DEGs (40). For example, Ma et al. utilized the RRA method to integrate four eligible DCM microarray datasets from the GEO and developed a 7-gene signature predictive model of DCM (11). While in the present study, using RUVSeq to substantially decrease batch effects, we integrated, for the first time, the different RNA-seq datasets of the GEO database to explore DEGs and hub genes associated with HF by using the RRA method. Through internal RNA-seq dataset and external microarray dataset validation, ASPN, COL1A1, and FMOD were finally identified as real hub genes of HF, which were further confirmed by qRT-PCR using the heart tissues from patients with HF and NFDs. Interestingly, the identified hub genes ASPN (41), COL1A1 (42), and FMOD (43), all belong to the type I collagen members in the extracellular matrix (ECM) composition and have been reported to play important roles in the development and progression of various diseases, especially malignant tumors. For example, ASPN was reported to enhance tumor invasion and cancer-associated fibroblasts via activation of the CD44-Rac1 pathway in gastric cancer (41). Ma et al. highlighted the role of COL1A1 as a potential diagnostic biomarker and therapeutic target in early development and metastasis of hepatocellular carcinoma (42). Ao et al. revealed that FMOD could promote tumor angiogenesis by upregulating the expression of angiogenic factors in human small-cell lung cancer (43). Regarding the function of hub genes in HF development, a multi-level transcriptomic study conducted by Hua et al. suggested that COL1A1 might be a plasma biomarker of HF and associated with HF progression, especially to predict the 1-year survival from HF onset to transplantation. A COL1A1 content ≥ 256.5ng/ml in plasma was found to be associated with poor survival within 1 year of heart transplantation from HF (34). In the study conducted by Andenæs et al., FMOD was found 3–10-fold upregulated in hearts of patients with HF and mice, and FMOD-KO mice showed a relatively mild hypertrophic phenotype (44). However, to the best of our knowledge, there are no experimental studies focusing on the role of ASPN in HF. Therefore, our multi-dataset RRA analysis, followed by microarray dataset and experimental validation, provides more robust and comprehensive evidence for the value of the three ECM-related genes, namely, COL1A1, FMOD, and ASPN, in HF development. Recent advances have highlighted the crucial role of immune activation in the development and progression of HF. A study by Aghajanian et al. demonstrated that adoptive transfer of T cells that express a chimeric antigen receptor against fibroblast activation protein can inhibit myocardial fibrosis and improve cardiac function in mice (45). Consistently, according to the GO term analysis in our study, the upregulated HF-related DEGs were enriched in T-cell activation of the “BP” term, the extracellular matrix of “CC” terms, and the extracellular matrix structural constituent of “MF” terms. Moreover, regarding the KEGG pathway analysis, the T-cell receptor signaling pathway and ECM–receptor interaction were identified as the significantly enriched pathways of the upregulated DEGs. Considering that all the three hub genes—ASPN, COL1A1, and FMOD—are closely associated with the ECM, we thus speculate a potentially key pathway in the development of HF, that is, T-cell-mediated immune responses lead to the imbalance in ECM anabolism and catabolism, ultimately resulting in myocardial fibrosis and HF. To further explore the potential mechanism of ASPN, COL1A1, and FMOD in HF, we performed GSEA and GSVA on the validation dataset of GSE57338. Results showed that ASPN-, COL1A1-, or FMOD-related DEGs were enriched in the “citric acid cycle (TCA cycle)” and “propionic acid metabolism” pathways, both of which are closely associated with glucose metabolism (46, 47). Notably, targeting cardiac glucose metabolism has been recognized as a promising therapeutic strategy for HF treatment. Liu et al. reported that dichloroacetate, a pyruvate dehydrogenase kinase inhibitor, could alter glucose metabolism in cardiomyocytes by stimulating the activity of pyruvate dehydrogenase complex, thereby improving cardiac efficiency (48). In addition, inhibitors of fatty acid oxidation such as trimetazidine (49), perhexiline (50), and etomoxir (51) can improve cardiac function in patients with HF by increasing glucose oxidation. Aberrant regulation of TFs is strongly associated with the onset and progression of HF (52). Therefore, in our research, we further investigated the TF–gene interactions to detect the transcriptional regulators of the robust DEGs. Among the seven identified significant TFs, MEF2A (53) and PPARG (54) have been reported to play a role in cardiac remodeling and water retention in HF, respectively. Liu et al. found that suppressing expression of TEAD1, the Hippo signaling effector, could activate the necroptotic pathway and induce massive cardiomyocyte necroptosis, ultimately leading to impaired cardiac function (55). Moreover, loss of BRCA1 in mouse cardiomyocytes resulted in adverse cardiac remodeling and poor ventricular function (56). Although the functions of these TFs in HF have been partially reported, the regulatory relationship of the TF–DEG pairs and the in-depth molecular mechanisms remain to be further validated through HF-related experimental studies. Our study has several limitations. First, the sample size of patients with HF is relatively small, although we have included as many datasets that met the criteria as possible. Future studies with larger sample sizes are needed to confirm these findings. Second, this study is mainly based on bioinformatics analysis and qRT-PCR validation of hub gene expression. Further experimental research is needed to clarify the in-depth mechanism of the hub gene-related HF regulation. Third, information about disease grades, treatment methods, and prognosis of patients with HF is not available in the database, leading to the failure to analyze correlation between hub genes and clinical characteristics or prognosis of HF. Fourth, the etiology of HF is complex, involving multiple environmental factors in addition to genetic factors (57), such as behavioral factors, socioeconomic and psychosocial factors, air quality, and meteorological factors (58–60). Horton et al. reported that the influence of modifiable lifestyle factors cannot be ignored in the development of direct-to-consumer (DTC) genetic tests (61). In recent years, emerging evidence has shown that gene–environment interactions play an important role in complex disease progression. Bentley et al. revealed that the genetic associations with lipids could be modified by smoking (62). Therefore, future research needs to further focus on the role of environmental factors and gene–environment interactions in HF. In conclusion, the present study integrated, for the first time, the different RNA-seq datasets of HF from the GEO database and identified robust HF-related DEGs utilizing RUVSeq and the RRA method. Furthermore, three reliable hub genes—ASPN, COL1A1, and FMOD—were screened and validated by bioinformatics and experimental assays. Functional enrichment analysis showed that DEGs and hub genes were associated with T-cell-mediated immune response and the glucose metabolism signaling pathway. In addition, significant TF–DEG regulatory network of HF was further established. However, high-quality basic or clinical research is required to deeply investigate the mechanisms by which these hub genes are involved in HF and to confirm their values as biomarker for HF diagnosis and treatment. The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material. The studies involving human participants were reviewed and approved by the Institute Ethics Committee of Changhai Hospital. The patients/participants provided their written informed consent to participate in this study. ZG, XS, and XZ contributed to the conception of the study. DT and CM performed the study execution and experiments. ZZ and QX contributed to part of bioinformatic analysis. DT, CM, and XS prepared the manuscript. XZ and ZG contributed to the funding of the study. All authors reviewed the manuscript, provided critical revision, and have approved the final version for publication.

PMC9649664

Marko Petek,Maja Zagorščak,Andrej Blejec,Živa Ramšak,Anna Coll,Špela Baebler,Kristina Gruden

pISA-tree - a data management framework for life science research projects using a standardised directory tree

10-11-2022

Data publication and archiving,Research data

We developed pISA-tree, a straightforward and flexible data management solution for organisation of life science project-associated research data and metadata. pISA-tree was initiated by end-user requirements thus its strong points are practicality and low maintenance cost. It enables on-the-fly creation of enriched directory tree structure ( p roject/ I nvestigation/ S tudy/ A ssay) based on the ISA model, in a standardised manner via consecutive batch files. Templates-based metadata is generated in parallel at each level enabling guided submission of experiment metadata. pISA-tree is complemented by two R packages, pisar and seekr. pisar facilitates integration of pISA-tree datasets into bioinformatic pipelines and generation of ISA-Tab exports. seekr enables synchronisation with the FAIRDOMHub repository. Applicability of pISA-tree was demonstrated in several national and international multi-partner projects. The system thus supports findable, accessible, interoperable and reusable (FAIR) research and is in accordance with the Open Science initiative. Source code and documentation of pISA-tree are available at https://github.com/NIB-SI/pISA-tree.

pISA-tree - a data management framework for life science research projects using a standardised directory tree We developed pISA-tree, a straightforward and flexible data management solution for organisation of life science project-associated research data and metadata. pISA-tree was initiated by end-user requirements thus its strong points are practicality and low maintenance cost. It enables on-the-fly creation of enriched directory tree structure (project/Investigation/Study/Assay) based on the ISA model, in a standardised manner via consecutive batch files. Templates-based metadata is generated in parallel at each level enabling guided submission of experiment metadata. pISA-tree is complemented by two R packages, pisar and seekr. pisar facilitates integration of pISA-tree datasets into bioinformatic pipelines and generation of ISA-Tab exports. seekr enables synchronisation with the FAIRDOMHub repository. Applicability of pISA-tree was demonstrated in several national and international multi-partner projects. The system thus supports findable, accessible, interoperable and reusable (FAIR) research and is in accordance with the Open Science initiative. Source code and documentation of pISA-tree are available at https://github.com/NIB-SI/pISA-tree. Compared to a decade ago, present-day life science projects generate vast amounts of data, that are frequently heterogenous. Both researchers and funders are aware that published results often lack important information that would ensure their reproducibility. Raw data supporting researchers’ conclusions are frequently either inaccessible, inadequately organised or not sufficiently annotated to allow straightforward reuse. To alleviate these issues, the FAIR data principles that advocate policies which ensure findability, accessibility, interoperability and reusability of the generated data were proposed. Thus, reporting, format and semantic standards are of utmost importance. Minimum information metadata standards, providing guidelines for standardised reporting of experimental results, have been developed for a large number of wet-lab assays; e.g. plant phenotyping (MIAPPE), quantitative PCR (MIQE), gene expression microarrays (MIAME), proteomics (MIAPE); as well as dry-lab assays, e.g. quantitative modelling (MIRIAM) and simulations in systems biology (MIASE). To organise and keep track of experimental metadata, the ISA framework implemented the ISA model that captures experimental metadata on three hierarchical levels: Investigation, Study and Assay (abbreviated as ISA levels). Each ISA level contains files describing experimental goals, ontologies used, experimental design, protocols, and experimental conditions. ISA framework developers also provide ISA tools, a software suite that, among others, contains a metadata editor and packages for integration with common programming languages. ISA tools use a series of graphical interface input forms to generate metadata in the ISA-Tab or ISA-JSON format and use assay-specific templates covering a wide variety of methodologies and community standards. In parallel, based on minimum information standards, several public repositories have formulated metadata templates, that guide the researchers before submitting the data with publications. Besides methodology-specific repositories, there are general repositories that offer long-term research data and metadata storage following FAIR principles. One such service is Zenodo (www.zenodo.org), an open science repository developed by CERN that accepts datasets in any format. Another is SEEK/FAIRDOMHub, a general repository that adopted the ISA levels to structure project-related data. While these all fullfill the requirements of OpenScience they are difficult to embrace and adopt by the researchers that are involved in data acquisition. For the duration of a research project, datasets are frequently stored and analysed locally, usually on the researcher’s personal computer or institution’s network drive. Institutions have various strategies on how to deal with organisation and medium to long-term storage of project data. These strategies may include the use of a Laboratory Information Management System (LIMS), electronic laboratory notebook (ELN), and either centralised or decentralised storage of raw and analysed data. However, for research projects, such software solutions can have some disadvantages - they might be associated with license costs, are prone to rigidity, can become unsupported over the course of a few years, or are unable to handle diverse and large scale datasets. Several services were also developed for FAIR data managmenet, besides SEEK/FAIRDOMHub, COPO and e!DAL are examples of such services. In principle, those services can also be run locally at an institutional level; however, this requires substantial IT support and discipline in data upload. Project managers are thus more likely to use these services only for data deposition for long-term storage at the end of the project than for immediate data storage. To address all the issues mentioned above, we have developed pISA-tree, an easy-to-use and maintain system for intra-institutional organization and structured storage of research data, with a special emphasis on the generation of adequate metadata. pISA-tree uses the ISA levels, FAIR principles, and minimum information standards for metadata formulation, and encourages the inclusion of metadata on-the-fly, when the experiments are designed. Its complementary R packages enable easy transfer to public (meta)data repositories and (re)use of data in analytical pipelines. We developed pISA-tree and its accompanying R packages (pisar and seekr) to provide a user-friendly system that guides researchers towards organisation of their data without the requirement of any advanced systems maintenance. It was designed by data scientists in constant interaction with reserchers involved in data acquisition to allow for adoption by both. pISA-tree is based on a consecutive set of batch files that generate an enriched directory tree structure for scientific research projects. It uses a familiar directory system with a standardised structure to organise each project hierarchically into three nested ISA levels: Investigations, which consist of multiple Studies, where each individual study can contain multiple Assays conducted on a set of samples from a single coherent experimental setup (Fig. 1, Supplementary File 1). Typically, an investigation aligns well with a defined experimental question or project work package, while a study encompasses a sample collection from one experiment. One of the core components of the system are consistent metadata files, associated with each pISA-tree level (Fig. 1). Prior to folder creation, the user answers a series of questions (keys) with predefined attributes or free-text answers (values), which are then stored in metadata files as tab-delimited key-value pairs. Key-value pairs can follow any controlled vocabulary and can be further edited using any text editor. This encourages users to enter the minimal information at an early stage of research preparation and provides the advantage of finding relevant information later. In addition, updated and consistent metadata can be used in the data analysis phase. While the upper levels, i.e. project, investigation and study, hold a more descriptive role, the assay level is the richest level of the research record. Each data analysis step can be defined as a separate Assay. Moreover, pISA-tree can support two different classes of assays: laboratory experiment (wet) and data analysis (dry). Each assay class contains several pre-prepared assay types, depending on measurement methodology (e.g. RNA isolation, qPCR, GC-MS). These templates were developed by taking into account minimal information standards (e.g. MIQE for qPCR) as well as consulting researchers performing the assays. The assay class and type define the subdirectories and other standardised information that is typically required. For example, besides common subdirectories such as output and reports, the wet-lab assays will have the output\raw subdirectory for storing raw experimental data, whereas the dry-lab assays will have the scripts subdirectory (Table 1). All options and predefined minimal information are kept in text files and can be customised to extend assay classes and types. For each new class or type, the user adjusts the directory structure and defines metadata template files. Before creating any project subdirectories, key-value pairs that should be present in all sublevel metadata files, e.g. organisation or license information, can be added to the common.ini file. This way, key-value pairs will be automatically assigned to all directories within the project and do not need to be re-entered for each new investigation or study. The investigation level specifically holds the master sample description table, namely the phenodata file (phenodata_yyyymmdd.txt), which should be populated with all relevant information on all the experimental samples that will be collected in the associated studies and further analysed in corresponding assays. The phenodata file should therefore contain the experimental design information required for planned data analysis. We advise users to take the requirements of the community-specific minimal information standards into account when preparing the phenodata files, i.e. use MIAPPE for plant samples. More details on how to structure and keep track of the potential changes in the phenodata file can be found in the pISA-tree user manual (Supplementary File 1). Users are also encouraged to create descriptions of all measured variables within assays and store them in a featuredata file (featuredata_yyyymmdd.txt). Certain assay types additionally generate an analyte table (analytes.txt) which serves as a template for input of specific measurements (e.g. sample quality control values), thus helping the wet-lab researchers to input the results of experiments into the system. The metadata files created are machine-readable, enabling interoperability and interconnectedness across research fields. All levels contain three auxiliary batch scripts: showtree.bat displays the complete directory tree structure and allows for a better overview of the information in the system for complex projects; showmetadata.bat displays all recorded metadata thus far; and xcheckMetadata.bat displays missing mandatory metadata fields. These auxiliaries thus help the scientists when reviewing their project status in terms of data management completeness, thus promoting data FAIRness. Because the R environment for statistical computing has a firmly established developer and user community for bioinformatic data analysis, with well-maintained packages, we further complemented pISA-tree with two R packages. The predefined pISA-tree directory structure enables integration into automated data analysis pipelines, while R enables repeatability of analyses, e.g. by including output of the SessionInfo() function in the analysis reports or using R project environment. Therefore, we further complemented pISA-tree with pisar, an R package that enables import of pISA-tree metadata into R workflows (Fig. 2) and thus repeatability of R analyses within the pISA-tree system. Since information is structured in a standardised way and augmented by metadata, connection to other local or cloud-based FAIR data management platforms can be established. The other part of the system enables seamless upload to the project metadata repository FAIRDOMHub using the R package seekr (https://github.com/NIB-SI/seekr) which interacts with the FAIRDOMHub API. When using general data repositories for long term storage, pisar package allows conversion of pISA-tree metadata into ISA-Tab format, which we recommend to upload together with an archived file containing the entire project (Fig. 2). FAIR paradigms are becoming integral to the way we conduct scientific research. We developed pISA-tree for local project data management to facilitate implementation of these paradigms in day-to-day research. We illustrate the usage of pISA-tree with two example life science projects described below, both including diverse assay types. The seekr package was used for the automatic upload of the complete projects to FAIRDOMHub, thus making these projects data easily accessible and reusable. The first example is a multidisciplinary project on insect pesticide development. The project (“Using RNAi and systems biology approaches for validation of insecticide targets in CPB guts”, https://fairdomhub.org/projects/252) has both wet and dry lab assays. The scope of this project was to develop double-stranded RNA (dsRNA) based insecticide against Colorado potato beetle pest. It included identification of Colorado potato beetle gene targets, establishing dsRNA production, and validating the dsRNAs’ insecticidal potential in laboratory and field trials. The directory tree of this project consists of three investigations covering altogether 16 studies with wet-lab and dry-lab assays (Fig. 3). The investigation “_I_01_LabTrials” includes 11 studies, the first three of which contain data on dsRNA design and production, while the rest contain data on feeding trials. The phenodata file in this investigation describes all samples in the underlying assays in detail. The “RNAisol” wet-lab type assays in the feeding trial studies include analyte files that contain information on isolated RNA quantity and quality parameters measured for samples analysed in the particular trial. The second example is an implementation of pISA-tree in a large-scale dry-lab project on de novo potato transcriptome assembly (“Solanum tuberosum Reference Transcriptomes”, https://fairdomhub.org/projects/161). This project’s aim was to construct a pan-transcriptome from high-volume data accessible from publicly available resources and in-house defined workflows in a collaborative way. The project consists of a single investigation, four studies and a plethora of data files (Fig. 4), encompassing four consecutive bioinformatics procedures: (1) high-throughput sequencing acquisition and pre-processing, (2) de novo assembly, (3) cultivar-specific transcriptomes generation and annotation, and (4) pan-transcriptome generation and annotation. Assays nested within the Studies are structured as predefined for the dry-lab assay class (see Table 1), yet no additional specific assay types were developed within this project, as they were custom for each step of the developed assembly and quality control pipeline. To promote the adoption of FAIR principles, many funding organisations and journals are committed to FAIR and open data principles. Often, project proposals are expected to contain at least a draft of data management plan. This plan has to incorporate policies for data collection, organisation and exchange throughout the project duration and also after its end, since it must also take into account local storage and deposition to publicly accessible repositories. The life science community has at large adopted the core FAIR values. Bottlenecks lie in the implementation of said management plans, partially because the data management systems in place are not designed for researchers that typically collect, annotate, and organise the data. As these systems are usually coordinated on an institutional level, their practicality and usefulness for individual researchers or research groups are often not taken into account. In that respect, pISA-tree is distinct, initiated by end-user requirements and implemented by developers with an emphasis on practicality. The development went through several rounds of refinements to accommodate most of the involved life scientists’ needs. Compared to other life science project data management systems, one of the main advantages of pISA-tree is its simplicity. It is based on file system directories, inherent to all operating systems and can therefore be used by scientists with basic computer skills following a short practical course or careful reading of the user manual (Supplementary File 1). Organising metadata is a fairly complex task, thus the process in pISA-tree is to some degree automated by predefined metadata keys and values. We believe that a simple system will be more easily adopted by a wide range of scientists, thus increasing compliance with FAIR principles. For the same reason, pISA-tree does not require professional technical support but we recommend a data steward being nominated to take care of maintaining order, access control, updates, uploading to public repositories, and data backup. Therefore, the data management system we propose here is particularly practical for smaller institutions or research groups with limited resources and staff. Other advantages include open source code, independence from web browsers and third-party dependencies, and straightforward access control. pISA-tree metadata are stored in plain text files that can be edited, conferring to the system high flexibility. In addition, it is not limited to specific laboratory or computational technologies, as the provided wet and dry metadata templates can be applied and adjusted to any new technology. The user is encouraged to create the pISA-tree directory structure and corresponding metadata files before conducting the experiments. However, considering the impracticality of strict requirements for immediate metadata input, which could discourage end-users, pISA-tree flexibility allows the user to complete or correct the metadata at the later stages. This is a very convenient feature trait of pISA-tree compared to other systems such as ISA tools, which assume that the assays have already been carried out and the data was analysed. Compared to ISA tools, in pISA-tree the metadata files are organised somewhat differently. ISA tools does not use a hierarchical directory structure, instead, all metadata files are stored in one directory. With larger projects hierarchical organisation of experiments helps keeping the overview and eases reuse. Another distinction from ISA tools is that in pISA-tree, sample descriptions and experimental design are stored in the phenodata file and the measured variables are stored in featuredata file at the investigation level. The structure of these files are comprehensible for wet-lab researchers. However, their nomenclature was adopted from statistical analysis of microarray data in R. The rationale for putting these files to the investigation level is to have unique sample and measured variable IDs across all studies. This convention can be adapted for multipartner investigations where each partner stores their work within its study and it thus makes sense to store phenodata files at the study level. For easier adaptation for wet lab researchers, pISA-tree assay metadata files allow for input of file paths or URLs for experimental protocols or data instead of providing full text. pISA-tree users and data stewards have to be aware of some of its limitations, inherent to its design. One is the limitation of the maximum path length to 247 characters in Windows operating systems, which can only be circumvented by changing the Windows registry key. To avoid this issue, the usage of short but descriptive directory and file names is recommended. In addition, no automatic file and directory version control is enabled, except by using subsidiary software. In the user manual (Supplementary File 1), we advise the users on how to manage phenodata and featuredata file versioning, however for all other files a different versioning logic can be applied. The lack of version control can also be solved by uploading to FAIRDOMHub promptly after each significant update to the project. Taking into account the accessibility premise from the FAIR data principles, integration with FAIRDOMHub is also an elegant solution for sharing data with collaborators, stakeholders or the public. Although pISA-tree was designed for small to medium sized projects, it has proven to be flexible enough to be adapted for larger projects with multiple independent research partners, which we showed in two ERACoBiotech projects SUSPHIRE (http://susphire.info/) and INDIE (https://indie.cebitec.uni-bielefeld.de/), both including five partners from three different countries and researchers with diverse fields of expertise, and H2020 RIA project ADAPT (https://adapt.univie.ac.at/) with 17 partners from 8 countries. To conclude, we believe that pISA-tree will facilitate adoption of data management practices by individual researchers that will in long term lead to data FAIRness, adoption of open science principles and ultimately, new scientific discoveries. pISA-tree scripts and template files are organised in the main directory and the Templates subdirectory. The Templates subdirectory is further subdivided into x.lib, DRY and WET, which contain templates for step-by-step creation of bat files, and dry-lab and wet-lab assay classes. The batch script named makeProject.bat is stored in the main directory, and is the one to be used at the start of the pISA-tree structure creation (see manual in Supplementary File 1). The main Windows executable batch file, pISA.cmd, is stored in the \Templates\x.lib directory. Depending on the user’s progress in building the pISA-tree structure for their project, this script generates additional makeInvestigation.bat, makeStudy.bat and makeAssay.bat at the appropriate levels within the project and initialises template metadata files. The directory names are automatically given a prefix defining the level i.e. “_p_” for project, “_I_” for investigation, “_S_” for study and “_A_” for assay, which facilitates visual inspection and general usage of the structure. In addition, assay directory names are automatically given an “-<AssayType>” suffix, that are defined in the Templates directory. The level-specific metadata files are text files containing a collection of tab-delimited key-value pairs in the form “ < key: > \t < value > ” and can be extended to suit the user’s needs via any text editor. Another set of scripts can be used to inspect project directories and metadata files. The showTree.bat draws the pISA-tree directory tree structure to the TREE.TXT file, while the showMetadata.bat outputs a file containing all metadata in either plain text or markdown format, depending on the user’s position within the tree. The xcheckMetadata.bat highlights the missing values in the metadata files required by the minimum reporting standards, and kindly reminds the user to fill them in. To set up the pISA-tree, download the GitHub repository (https://github.com/NIB-SI/pISA-tree) into a local or network directory with read/write/execute permissions. This “pISA-tree root directory” ideally contains all pISA-tree projects. The setup does not require any administrative installation privileges of the user and the entire directory tree can be moved freely, if only relative paths are used in the metadata files, which is one of the recommendations given in the manual (Supplementary File 1). The metadata associated with specific assay types is defined in template files in the \Templates\DRY and \Templates\WET class directories. Users can add new classes of assays, extend the existing types, or develop their own. To predefine the questions (and potentially multiple choice answers) that the user will be asked when creating a new assay, one adds rows of “ < key: > \t < value > ” pairs into the < AssayType > _Template.txt file. Here, keys define the questions and ‘/’ separated values provide menu choices. Users can also enter their own answers not provided in the template by selecting the option “other”. Additional pISA-tree assay metadata templates are available at a separate, conceived as a community driven GitHub repository (https://github.com/NIB-SI/pISA-tree-assay-types) to be utilised for open science contribution of templates. R package pisar enables repeatability of R-based analyses within the pISA-tree system. The users can retrieve recorded metadata information and the complete directory structure of the pISA-tree to reuse the dry-lab analysis protocols. The pisa() function retrieves the relative paths of the pISA-tree directory structure up to the project level whereas the getMeta() function retrieves specific fields (key values) from the metadata files. Thus, one can rerun R scripts without changing the code even if the metadata files are changed. As good practice, we recommend listing all metadata used at the end of the report. The seekr R package was developed for communication of the pISA-tree with the SEEK API used by the FAIRDOMHub public repository and enables bulk upload of pISA-tree investigations. Currently, before uploading, the project page has to be created on the FAIRDOMHub web page to define user roles and permissions. The user can specify to exclude certain directories or files within the pISA-tree project directory tree by editing the “seekignore.txt” file. Further, uploading individual levels can be controlled by changing “Upload to FAIRDOMHub” key’s value in the level metadata file, which is, by default, set to “Yes”. In addition, the “Sharing permission” key sets the sharing of an individual level at FAIRDOMHub as either “Private” or “Public”, while the key “License” determines the license under which the data is shared to the public. Bulk investigation upload to FAIRDOMHub using seekr is implemented with the R functions skFilesToUpload() and skUploadFiles(). The first function creates a list of files to be uploaded and the second function uses this list to prepare a JSON object for the specified investigation and all its accompanying files. It then and communicates with the SEEK API to create the corresponding FAIRDOMHub objects and then uploads the selected files. Package seekr uses the package pisar to obtain the local pISA-tree directory structure and metadata. A Jupyter Notebook is provided at seekr GitHub repository (https://github.com/NIB-SI/seekr) demonstrating the upload to FAIRDOMHub. Unlike pISA-tree, FAIRDOMHub only allows linking files to the assay level. Therefore, to avoid having files disconnected from their corresponding levels, skUploadFiles() function creates a FAIRDOMHub study named “Investigation files” and a FAIRDOMHub assay named after the pISA-tree investigation where it stores the pISA-tree investigation metadata files. Likewise, it stores pISA-tree study metadata files in a FAIRDOMHub assay named after the study. To enable interoperability with the ISA framework, we developed a converter from pISA-tree to ISA-Tab metadata format. The converter is implemented in the pisar package as createISAtab() function. It uses XML templates that contain information about the pISA-tree assay types that structurally correspond to the ISA-tools configuration files (v2015-07-02), and a mapping file that specifies how to convert fields from pISA-tree metadata files to ISA-Tab format. For each investigation, the function outputs ISA-Tab formatted metadata files that can be validated using ISAvalidator v1.6.5 from the ISA tools suite. A Jupyter Notebook demonstrating straightforward conversion is available at pISA-tree-assay-types GitHub repository (https://github.com/NIB-SI/pISA-tree-assay-types). ISA model – a metadata framework to manage an increasingly diverse set of life science, environmental and biomedical experiments that employ one or a combination of technologies. It is built around the investigation (the project context), study (a unit of research) and assay (analytical measurements) concepts. ISA level – common term for each of the concepts within the ISA model, either investigation, study or assay. ISA-Tab format – a hierarchical file format that captures the experimental metadata (https://isa-specs.readthedocs.io/en/latest/isatab.html) and is compatible with different spreadsheet-based formats for data sharing. ISA-JSON format – same as ISA-Tab but in JSON format instead of tabular. Data standard – technical specifications or recorded agreements that include models, formats, reporting guidelines, and identifier schemas (https://fairsharing.org/). Minimal information about experiment requirement – standards that describe minimal metadata that is required for a particular type of experiment to be FAIR (https://fairsharing.org/). Data model – a model that organizes data elements and standardizes how the data elements relate to one another. A data model explicitly determines the structure of data (https://cedar.princeton.edu/understanding-data/what-data-model). FAIR principles – guidelines to improve the Findability, Accessibility, Interoperability, and Reuse of digital assets. Supplementary File 1

PMC9649669

Hang Xu,Jiapeng Zhang,Xiaonan Zheng,Ping Tan,Xingyu Xiong,Xianyanling Yi,Yang Yang,Yan Wang,Dazhou Liao,Hong Li,Qiang Wei,Jianzhong Ai,Lu Yang

SR9009 inhibits lethal prostate cancer subtype 1 by regulating the LXRα/FOXM1 pathway independently of REV-ERBs

10-11-2022

Prostate cancer,Drug development

Perturbations of the circadian clock are linked to multiple diseases, including cancers. Pharmacological activation of REV-ERB nuclear receptors, the core components of the circadian clock, has antitumor effects on various malignancies, while the impact of SR9009 on prostate cancer (PCa) remains unknown. Here, we found that SR9009 was specifically lethal to PCa cell lines but had no cytotoxic effect on prostate cells. SR9009 significantly inhibited colony formation, the cell cycle, and cell migration and promoted apoptosis in PCa cells. SR9009 treatment markedly inhibited prostate cancer subtype 1 (PCS1), the most lethal and aggressive PCa subtype, through FOXM1 pathway blockade, while it had no impacts on PCS2 and PCS3. Seven representative genes, including FOXM1, CENPA, CENPF, CDK1, CCNB1, CCNB2, and BIRC5, were identified as the shared genes involved in the FOXM1 pathway and PCS1. All of these genes were upregulated in PCa tissues, associated with worse clinicopathological outcomes and downregulated after SR9009 treatment. Nevertheless, knockdown or knockout of REV-ERB could not rescue the anticancer effect of SR9009 in PCa. Further analysis confirmed that it was LXRα rather than REV-ERBs which has been activated by SR9009. The expression levels of these seven genes were changed correspondingly after LXRα knockdown and SR9009 treatment. An in vivo study validated that SR9009 restrained tumor growth in 22RV1 xenograft models and inhibited FOXM1 and its targeted gene expression. In summary, SR9009 can serve as an effective treatment option for highly aggressive and lethal PCS1 tumors through mediating the LXRα/FOXM1 pathway independently of REV-ERBs.

SR9009 inhibits lethal prostate cancer subtype 1 by regulating the LXRα/FOXM1 pathway independently of REV-ERBs Perturbations of the circadian clock are linked to multiple diseases, including cancers. Pharmacological activation of REV-ERB nuclear receptors, the core components of the circadian clock, has antitumor effects on various malignancies, while the impact of SR9009 on prostate cancer (PCa) remains unknown. Here, we found that SR9009 was specifically lethal to PCa cell lines but had no cytotoxic effect on prostate cells. SR9009 significantly inhibited colony formation, the cell cycle, and cell migration and promoted apoptosis in PCa cells. SR9009 treatment markedly inhibited prostate cancer subtype 1 (PCS1), the most lethal and aggressive PCa subtype, through FOXM1 pathway blockade, while it had no impacts on PCS2 and PCS3. Seven representative genes, including FOXM1, CENPA, CENPF, CDK1, CCNB1, CCNB2, and BIRC5, were identified as the shared genes involved in the FOXM1 pathway and PCS1. All of these genes were upregulated in PCa tissues, associated with worse clinicopathological outcomes and downregulated after SR9009 treatment. Nevertheless, knockdown or knockout of REV-ERB could not rescue the anticancer effect of SR9009 in PCa. Further analysis confirmed that it was LXRα rather than REV-ERBs which has been activated by SR9009. The expression levels of these seven genes were changed correspondingly after LXRα knockdown and SR9009 treatment. An in vivo study validated that SR9009 restrained tumor growth in 22RV1 xenograft models and inhibited FOXM1 and its targeted gene expression. In summary, SR9009 can serve as an effective treatment option for highly aggressive and lethal PCS1 tumors through mediating the LXRα/FOXM1 pathway independently of REV-ERBs. Prostate cancer (PCa) is the most common malignancy in the male population, accounting for 26% of all estimated new cases and 11% of all estimated deaths in the United States in 2021 [1]. Androgen deprivation therapy (ADT) is the most commonly used treatment strategy for hormone-sensitive nonmetastatic and metastatic PCa, but the disease will eventually progress to castration-resistant prostate cancer (CRPC). Although several novel drugs, including enzalutamide [2] and abiraterone acetate [3], have been introduced to treat mCRPC patients, therapy resistance still occurs rapidly [4], hampering the therapeutic options currently available for patients. Therefore, finding new targets and therapeutic drugs for mCRPC patients has become an urgent unmet need. The circadian clock plays a critical role in regulating physiological processes in humans [5]. Disruption of the circadian clock can lead to multiple problems, ranging from sleep disorders to cancer [6]. For example, night-shift work, which perturbs the circadian rhythm, is an important risk factor for the development of PCa [7]. Thus, pharmacological regulation of the circadian clock might be an attractive method for cancer prevention and therapies. At the molecular level, the mammalian circadian clock is composed of multiple genes that form clock-activator and clock-repressor complexes [8]. The nuclear receptors REV-ERBα (NR1D1) and REV-ERBβ (NR1D2) are the core members of the circadian clock and function as repressors of functions such as circadian rhythm (repressing BMAL1/CLOCK complexes) and metabolism [9] in the absence of a transcriptional activation domain. SR9009, a putative synthetic agonist of REV-ERBs, is beneficial for treating obesity, diabetes and circadian rhythm disorders [8] and can be easily acquired (https://www.simplyanabolics.com/sarms/sr9009-stenabolic/). Recent work by Sulli and colleagues suggested that SR9009 has powerful antitumor effects on multiple cancer types, including brain cancer, leukemia, breast cancer, colon cancer and melanoma [10]. The therapeutic effects of SR9009 on glioblastoma [11], hepatocellular carcinoma [12] and lung cancer [13] were subsequently demonstrated. However, whether SR9009 has an antitumor effect on PCa, especially lethal tumors, remains unknown. Although most SR9009-related studies have demonstrated that the therapeutic effect of SR9009 occurs via REV-ERBs, several reports have also revealed that SR9009 might also work in a REV-ERB-independent manner [14, 15]. Herein, we sought to investigate the influence of SR9009 on PCa progression. Our results revealed that SR9009 has cytotoxic effects on PCa cell lines but not on prostate cells. Furthermore, SR9009 could inhibit prostate cancer subtype 1 (PCS1), the most lethal and aggressive PCa [16], by mediating the LXRα/FOXM1 pathway independently of REV-ERBs. The results obtained provide new insight into molecular mechanisms and therapeutic interventions in PCa. RWPE-1, PC3, 22RV1, DU145, LNCaP, and C4-2B cells were purchased from Shanghai Cell Bank Type Culture Collection Committee (CBTCCC, Shanghai, China). HEK293T cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, Virginia, USA). PCa cell lines were cultured in RPMI 1640 medium (HyClone, Utah, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Australia) and 1% antibiotics (penicillin and streptomycin; HyClone) in a humidified incubator containing 5% CO2 at 37 °C. RWPE-1 cells were cultured in keratinocyte serum-free medium. Cells stably transfected with plasmid were cultured in complete culture medium with additional puromycin (2 μg/mL; KEHBIO, Beijing, China). Small interfering RNAs (siRNAs) targeting REV-ERBα, REV-ERBβ, LXRα, and FOXM1 were obtained from RiboBio (Guangzhou, China). The siRNA sequence is shown in Table S1. siRNA (50 nM) was transfected into cells using Lipofectamine 3000 transfection reagent (Invitrogen, CA, USA). FOXM1 overexpression plasmid (pLV.CMV.FOXM1.PGK. Puro) was purchased from PackGene (Guangzhou, China). The lentiviral packaging procedure for the target plasmid has been described previously [17]. Cas9-expressing stable cell lines were constructed by infection with Cas9 lentivirus and further puromycin screening. REV-ERB (NR1D1 and NR1D2)-specific sgRNA oligos were designed and cloned into the pLentiCRISPR V2 plasmid (sequences listed in Table S2). The harvested lentivirus was added to the cell supernatant and centrifuged at 2000 rpm for 1 h, followed by incubation for 1.5 h. SR9009, purchased from MedChemExpress (MCE, NJ, USA), was first dissolved in dimethyl sulfoxide (DMSO) and then diluted to the working concentrations (maximum DMSO concentration < 0.5%). Approximately (2–5) × 103 cells per well were seeded in a 96-well plate. When the cells grew to 70–80% confluence, they were treated with SR9009, DMSO, or siRNAs for 48 h. Thereafter, the culture medium in each well was replaced with 100 μL of fresh complete culture medium containing 10 μL of CCK-8 reagent (Dojindo Molecular Technologies, Rockville, USA). Then, the 96-well plate was placed into an incubator at 37 °C in the dark for 2 h. Finally, the plate was placed in the EonTM Microplate Reader (Bio-Tek, VT, USA) to measure the absorbance at 450 nm. At least 3 duplicate wells were set at the same time. PCa cells were seeded in 6-well plates at 500 cells/well. After 48 h of incubation, the cells were treated with SR9009 (20 μM) or DMSO for another 10–14 days. The cells were washed with phosphate buffered saline (PBS) and fixed with cold methanol for 20 min. After washing with PBS, the cells were stained with crystal violet (Beyotime, Shanghai, China) for 15 min. Subsequently, the cells were washed and imaged using a Celigo Imaging Cytometer. Cells were pretreated with SR9009 (20 μM), DMSO or siRNAs for 48 h. Subsequently, they were digested, centrifuged and collected into flow tubes. Then, 500 μL of 70% ice-cold ethanol was added, and the cells were fixed overnight at 4 °C. Cells were washed and filtered before the PI/RNase A (KeyGen Biotech, Jiangsu, China) dye working solution was added. After 30 min of incubation in the dark, we detected and recorded the cell cycle using a CytoFLEX Research Flow Cytometer (Beckman Coulter, CA, USA). An Annexin V-PE/7-AAD cell apoptosis detection kit was purchased from KeyGen Biotech. Cells were treated with SR9009 (20 μM) or DMSO for 48 h. The cells were then digested and washed twice with cold PBS. Next, 55 μL dye working solution (50 μL binding buffer + 5 μL 7-AAD) was added to the cells and incubated at 37 °C for 5–15 min in the dark. Subsequently, 450 μL binding buffer and 1 μL Annexin V-PE were added and incubated for 5–15 min. Cell apoptosis was assessed by a FACSAria SORP instrument (BD, USA). Cells were digested, suspended and seeded at 3 × 105 cells per well in 6-well plates. When the cells grew to approximately 80–90% confluence, 3 vertical parallel lines were drawn in each well. Cells were washed twice and treated with SR9009 (20 μM) or DMSO for 24 h. Images were immediately taken under an inverted fluorescence Zeiss OBSERVER D1/AX10 CAM HRC microscope (Zeiss), and the sites were recorded. Subsequently, the 6-well plates were placed in the cell incubator for an additional 24 h of incubation and imaged again. ImageJ software (National Institutes of Health, USA; Version 1.48) was applied to calculate the migration distance. Transwell migration assays were conducted using a Transwell chamber (Millipore, Massachusetts, USA). Briefly, Transwell chambers were placed on a 24-well plate. Fresh medium containing 10% FBS and 20 μM SR9009 in 600 μL was added to the lower chambers, and (2–5) × 104 cells in 200 μL of medium containing 20 μM SR9009 without FBS were added to the upper chamber. The 24-well plate was incubated at 37 °C for 48 h. Cells that invaded through the chamber were washed, fixed (20 min with 4% paraformaldehyde) and stained (30 min with crystal violet). Then, the upper chambers were washed, photographed and preserved under an inverted fluorescence OBSERVER D1/AX10 cam HRC microscope (Zeiss). Transferred cells were analyzed using ImageJ software. PC3 cells were treated with 20 μM SR9009 or DMSO for 48 h. Then, the cells were harvested, and RNA was stored using TRIzol (Invitrogen, CA, USA). Novogene (Beijing, China) was entrusted to perform RNA-seq. Briefly, RNA samples were extracted, and RNA sample quantification and qualification were performed. Then, the NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, USA) was selected to generate sequencing libraries following the manufacturer’s recommendations, and index codes were added to attribute sequences to each sample. After clustering and sequencing (Novogene Experimental Department), data analysis was performed through the following steps: quality control, read mapping to the reference, and quantification of the gene expression level (fragments per kilobase million was calculated). Differential expression analysis was performed using the DESeq2 R package (1.16.1). Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genome (KEGG) enrichment analysis were implemented by the clusterProfiler R package. Total RNA was extracted by using the RNeasy Mini Kit (Qiagen, Texas, USA) according to manual protocol. A total of 1 μg of RNA was added to synthesize first-strand complementary DNA (cDNA) using the Thermo Scientific RevertAid RT kit (Vilnius, Lithuania) with Oligo (dT)18. Quantitative PCR (qPCR) was performed using the QuantiNova SYBR Green PCR kit (Qiagen), and reactions were performed on the CFX96 Touch Real-Time PCR System (Bio-Rad, California, USA). The PCR amplification settings were as follows: 50 °C for 2 min and 95 °C for 10 min; 40 cycles of 98 °C for 5 s; and 59 °C for 10 s. β-actin was used for normalization, and each sample was repeated at least three times. The data were analyzed using the 2−ΔΔCt method. Primer sequences were acquired from the PrimerBank website (https://pga.mgh.harvard.edu/primerbank/) and synthesized by Sangon Biotech (Shanghai, China). The primers used in this study are shown in Table S2. Proteins were extracted, and the concentrations were determined using the Pierce™ BCA Protein Assay Kit (Thermo). Proteins were denatured at 100 °C for 10 min. After the proteins were separated by sodium dodecyl sulfate‒polyacrylamide gel electrophoresis (SDS‒PAGE, Epizyme, Shanghai, China), the gels were transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore) and run at 250 mA for 90 min. Subsequently, the membranes were cut, blocked (5% skim milk powder), and incubated with diluted primary antibodies at 4 °C overnight. The primary antibodies were as follows: anti-GAPDH (ZEN-BIO 200306-7E4), anti-β-actin (ZEN-BIO 250132), anti-vinculin (ZEN-BIO R26085), anti-NR1D1 (ab174309), anti-NR1D2 (ab251948 and Protein Tech 13906-1-AP), anti-FOXM1 (CST20459), anti-LXRα (ab41902), anti-CCNB1 (CST12231), anti-CCNB2 (ab185622), anti-CENPA (CST2186), anti-CENPF (CST58982), anti-CDK1 (ZEN-BIO 200544), anti-survivin (CST2808), and anti-ARNTL (Protein Tech 14268-1-AP). The membranes were washed and incubated with secondary antibodies for 1 h according to the primary antibody sources. Immunoreactivity was visualized using enhanced chemiluminescent (ECL) chromogenic substrate (Millipore). The membranes were finally detected by using a ChemiDoc MP Imager System (Bio-Rad). Specimens were fixed in 4% paraformaldehyde at room temperature and embedded in paraffin. Then, tissues were cut into 4 μm thick sections. Subsequently, we dewaxed, hydrated and incubated the tissues with antibodies overnight at 4 °C. After incubation with the corresponding secondary antibodies, the sections were stained with diaminobenzidine and reverse stained with hematoxylin. Male nude BALB/c mice (18–20 g each) at 6 weeks of age were purchased from Chengdu Dossy Experimental Animals Co., Ltd. (Chengdu, China). Mice were castrated with goserelin (MCE, daily for 19 days) subcutaneously. At the same time, cultured 22RV1 cells were collected and suspended in PBS. A 100 μL cell suspension with 5 × 106 cells was subcutaneously injected into the right flank of the mouse to establish a subcutaneous xenograft model. The weight and tumor volume of the mice were measured every 3 days, and the formula for calculating volume was (length × width2)/2 [18]. When the tumor volume increased up to 100–200 cm3, the mice were randomly divided into two groups (with 4 mice in each group). The investigator was not blinded to the group allocation during the whole experiment. SR9009 was dissolved in 15% Cremophor and administered twice daily (100 mg/kg) [10] in the experimental group through intraperitoneal injection, and the control group was given the same volume of Cremophor. When there was a significant difference between the two groups or the tumor volume exceeded 1000 cm3, mice were sacrificed, and subcutaneous tumors were harvested for hematoxylin-eosin staining and IHC. GEPIA2 (http://gepia2.cancer-pku.cn/#index) was used to analyze the gene expression differences between normal and tumor samples in prostate adenocarcinoma in The Cancer Genome Atlas (TCGA-PRAD). GEPIA2 was also applied to analyze the associations of gene expression (median cutoff value) and disease-free survival (DFS). mRNA expression Z scores relative to diploid samples were analyzed using cBioPortal (http://www.cbioportal.org). Protein levels of LXRα and FOXM1 were obtained from The Human Protein Atlas. The level 3 HTSeq-FPKM data in TCGA-PRAD were downloaded from https://portal.gdc.cancer.gov/. RNA-seq data in FPKM format were transformed into transcripts per million reads (TPM) format with log2 transformation. Correlation analyses were performed using the R package ggplot2. Data are presented as the mean ± standard deviation (SD) of at least three independent experiments. Data were statistically analyzed using GraphPad Prism software (Version 6.02; CA, USA). Student’s t test, ANOVA, or Wilcoxon rank sum test were applied as appropriate. Homogeneity of variance was tested using the Shapiro‒Wilk normality test for equality of variances. A two-tailed P value lower than 0.05 indicated statistical significance, which was labeled as follows: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. We assessed the impact of SR9009 on cell viability in three PCa cell lines and 1 normal prostate cell line. SR9009 markedly inhibited 22RV1, PC3 and DU145 cell viability in a dose-dependent manner but had no impact on RWPE-1 cells (Fig. 1A, Fig. S1A–D). We further showed that SR9009 significantly inhibited colony formation (Fig. 1B, C) and induced cell cycle arrest in PCa cells (Fig. 1D, E). SR9009 also promoted apoptosis in PCa cells compared with controls (Fig. 1F, G). Moreover, SR9009 impaired PCa cell migration in vitro in scratch and Transwell assays (Fig. 1H, I, Fig. S1D–E). These results indicated that SR9009 could inhibit PCa cell growth and migration but had no influence on normal prostate cells in vitro. A total of 559 upregulated and 553 downregulated genes were identified through RNA-seq (Fig. 2A, |Log2FC|> 2, adjusted P value < 0.05). KEGG pathway enrichment analysis revealed that SR9009 significantly promoted ferroptosis and glycine, serine and threonine metabolism (Fig. 2B) and inhibited the cell cycle pathway (Fig. 2C). GSEA indicated that SR9009 mainly functioned by regulating the PCa cell cycle (Fig. 2D, E). A recent analysis classified PCa into three distinct subtypes, named the “PCS classification” [16], and this classification performed better in distinguishing luminal and basal PCa than the PAM50 classification [19]. Our subsequent GSEA found that SR9009 markedly inhibited PCS1, the most aggressive and lethal PCa subtype (enrichment score [ES] = −0.95, P < 0.0001), whereas it had no influence on PCS2 (ES = −0.30, P = 0.25) and PCS3 (ES = 0.30, P = 0.09) when compared with the control groups (Fig. 2F–H, Fig. S2A, B). The results indicated that SR9009 significantly reduced 71 of 82 genes in PCS1 (Fig. 2H). The sequencing results were validated at the mRNA and protein levels (Fig. 2I–K). Previous studies have demonstrated that the FOXM1 pathway is the master driver of PCS1 [20]. We found that SR9009 treatment significantly reduced 20 of 36 FOXM1-regulated genes (Fig. 3A). GSEA also validated that SR9009 treatment inhibited the FOXM1 pathway (ES = −0.75, P < 0.0001, Fig. 3B). Subsequently, from the TCGA-PRAD dataset, we found that FOXM1 was significantly overexpressed in PCa samples compared with normal tissues (Fig. 3C), and the high expression of FOXM1 was associated with worse DFS (Fig. 3D). The mRNA expression of FOXM1 was found to be higher in 22RV1 and PC3 cells than in RWPE-1 cells (Fig. 3E). Knocking down FOXM1 in PC3 and 22RV1 cell lines (Fig. 3F–H) significantly induced cell cycle arrest (Fig. 3 I,J) and reduced cell viability (Fig. 3K). FOXM1 overexpression (Fig. 3L) increased cell viability in PC3 cells, and the treatment effect of SR9009 could be partially rescued after FOXM1 overexpression (Fig. 3M). Together, these results indicated that SR9009 mediated PCS1 inhibition through FOXM1 regulation. To further identify the key genes involved in lethal PCS1, we obtained 7 genes, including FOXM1, CENPA, CENPF, CCNB1, CCNB2, CDK1 and BIRC5, after intersection between PCS1 genes and FOXM1 pathway-involved genes (Fig. 4A). The expression of these 7 genes was significantly downregulated after SR9009 treatment (Fig. 4B). Previous studies have validated the important role of these 7 genes in promoting PCa progression [21–25]. Our bioinformatic analyses showed that the high expression levels of these 7 genes were all associated with adverse clinicopathological outcomes in PCa (Fig. S3A-D). We show the expression heatmap of these 7 genes stratified by Gleason score in Fig. 4C (with the Gleason score increased, the expression of these 7 genes was elevated). Moreover, the high expression of these 7 genes was associated with decreased DFS in PCa and other cancer types (Fig. 2D, Fig. 4D, Fig. S3E–J). SR9009 significantly decreased the expression of these 7 genes at both the mRNA and protein levels (Fig. 4E, F). Consistent with the fact that FOXM1 directly regulates the other 6 genes, we observed reduced expression of these 6 genes after FOXM1 knockdown (Fig. 4G). More importantly, the decreased expression of FOXM1 target genes induced by SR9009 could be partially rescued after FOXM1 overexpression (Fig. 4H). These results indicated that SR9009 could inhibit FOXM1-related genes and serve as a FOXM1 pathway inhibitor. Given that SR9009 was reported to be an REV-ERB agonist [10] and that the activation of REV-ERBs could lead to transcriptional repression of target genes [26] (BMAL1 [also known as ARNTL] is an essential feedback loop in regulating circadian rhythms [27]), we first examined the effect of SR9009 treatment on BMAL1 expression. Surprisingly, after analyzing our RNA-seq data, we found that BMAL1 expression was elevated after SR9009 treatment (Fig. S4A). Subsequent experiments revealed that SR9009 treatment increased BMAL1 expression at both the mRNA (Fig. S4B, C) and protein levels (Fig. S4D, E). To confirm that the cytotoxic effect of SR9009 was mediated by REV-ERBs, we examined the mRNA expression of REV-ERBα and REV-ERBβ across eight prostate cell lines (Fig. 5A). We first downregulated REV-ERBα expression (Fig. 5B, C) and observed elevated BMAL1 expression (Fig. S5A). Surprisingly, the SR9009-induced decrease in cell viability was not rescued after REV-ERBα knockdown in PC3 and 22RV1 cells (Fig. 5D, E). Furthermore, we also found that FOXM1 and its target genes were not elevated after REV-ERBα silencing (Fig. S5B). These results indicated that SR9009-induced FOXM1 pathway inhibition was not mediated by REV-ERBα activation. In addition, SR9009-induced cell viability impairment was not rescued after REV-ERBβ knockdown (Fig. 5F–I, Fig. S5C), and FOXM1 expression was decreased (not elevated) after REV-ERBβ knockdown (Fig. S5D), suggesting that SR9009-induced FOXM1 pathway inhibition was not mediated by REV-ERBβ activation. Subsequently, both REV-ERBα and REV-ERBβ were simultaneously silenced (Fig. 5J, Fig. S5E), which did not rescue the SR9009 effect on PCa cells (Fig. 5K, L) and did not elevate FOXM1 expression (Fig. S5F). Furthermore, we conducted CRISPR/Cas9-mediated REV-ERB knockout in PC3 and 22RV1 cells (Fig. S6). The western blot results revealed that REV-ERBs were efficiently disrupted in both PC3 and 22RV1 cells (Fig. 5M). Most importantly, SR9009 treatment could not rescue PCa cell viability after REV-ERB knockout (Fig. 5N, O). Taken together, the above data demonstrated that SR9009-induced FOXM1 pathway inhibition was independent of REV-ERBs. We next sought to explore other possible SR9009 targets that could repress the FOXM1 pathway. Previous literature reported that SR9009 manifested marked selectivity for LXRα (also known as NR1H3) when compared with REV-ERBα [28], and LXRα suppressed FOXM1 expression in hepatocellular carcinoma [29]. Thus, we hypothesized that SR9009-mediated FOXM1 pathway inhibition might involve the activation of LXRα. We found that LXRα mRNA and protein levels were downregulated in TCGA-PRAD tissues compared with normal tissues (Fig. 6A-B). Compared with RWPE-1, the protein expression of LXRα was decreased in PCa cell lines, whereas FOXM1 was increased (Fig. 6C). A negative association was found between LXRα and FOXM1 (Fig. 6D-E). In addition, LXRα knockdown significantly upregulated FOXM1 expression (Fig. 6F, G), suggesting that FOXM1 is regulated by LXRα. More importantly, the SR9009-induced cell viability reduction was partially rescued after LXRα knockdown (Fig. 6H, I), indicating that LXRα rather than REV-ERBs was activated after SR9009 treatment. In addition, we found that LXRα was negatively associated with FOXM1-targeted genes, including CENPA, CENPF, CDK1, CCNB1 and CCNB2, in TCGA-PRAD (Fig. 6J). LXRα knockdown upregulated FOXM1-targeted genes and partially rescued SR9009-induced FOXM1 pathway inhibition (Fig. 6K). We next assessed the impact of SR9009 on PCa progression in vivo. We established 22RV1 xenografts in nude mice and treated them with SR9009, along with continuous ADT (Fig. 7A). Our results indicated that SR9009 treatment significantly decreased tumor volume (Fig. 7B) and tumor growth (Fig. 7C). Tumor weights in the SR9009 group were significantly lower than those in the control group (Fig. 7D). As expected, the mouse weights were decreased after SR9009 treatment (Fig. 7E). Furthermore, by performing IHC staining using mouse tumor tissues, we found that SR9009 treatment reduced the expression of FOXM1 and its target genes CCNB1, CCNB2 and Survivin (Fig. 7F). The findings of the study are summarized in Fig. 7G. The incidence and mortality of PCa have increased in recent decades. Compared with 1990, the incidence of PCa in 2019 increased by 169.11% and caused over 480 thousand male deaths worldwide [30]. Current treatment strategies for mCRPC patients have been developed, such as androgen receptor inhibitors, PARP inhibitors, and prostate-specific membrane antigen (PSMA)-targeting therapies [31], and there is still considerable room to improve the treatment efficacy and find new drugs for PCa treatment. In this study, by conducting a series of experiments, we showed that SR9009 could significantly inhibit PCa cell growth in vitro and in vivo. SR9009 inhibited the most aggressive and lethal PCS1 tumors. Moreover, our mechanistic exploration showed that SR9009 might also function by regulating the LXRα/FOXM1 pathway independently of REV-ERBs. To the best of our knowledge, this is the first study investigating the role of SR9009 in treating PCa. We also proposed a novel SR9009-related pathway and validated that SR9009 could serve as both a PCS1 and FOXM1 inhibitor. Our study might pave the way for lethal PCa treatment. Due to the significant inherent biological heterogeneities in cancers [32], molecular classification has been applied to guide precise treatment. Classifications of luminal A, luminal B, and basal subtypes are widely adopted in breast cancer. Current clinical practice or guidelines do not recommend a recognized classification system for PCa. Most attempts to classify PCa were based on gene signatures [33–36] and were limited by the sample size analyzed. Two transcriptome-based classifications have been reported recently, named PAM50 [37] and PCS [16]. The PAM50 classification (n = 3782) divided PCa patients into luminal A, luminal B, and basal subtypes, and the PCS classification (n > 4600) divided PCa patients into PCS1, PCS2, and PCS3 subtypes. A comparison of PCS and PAM50 showed that they were similar in molecular profiles and clinical outcomes, while the PCS system had greater separation regarding clinical outcomes [19]. Considering that PCS1 represents the most lethal and aggressive PCa subtype and the absence of therapeutic options for PCS1 in current clinical practice, we identified that a small-molecule drug, SR9009, selectively inhibited PCS1 tumors (>85% genes in PCS1 were inhibited) but had no impact on PCS2 and PCS3 tumors. Consequently, SR9009 could serve as a novel PCS1 inhibitor. SR9009 is traditionally considered a putative ligand of REV-ERBs, and its therapeutic effects are mediated via REV-ERBs [10, 11, 13, 38, 39]. Nevertheless, recent studies suggested that the effects of SR9009 might also be independent of REV-ERBs. Dierickx and his colleagues found that the effect of SR9009 on cell proliferation and metabolism was not mediated by REV-ERBs [14]. Ishimaru et al. found that SR9009 had REV-ERB-independent effects in inhibiting mast cell activation [15]. In addition, Gao et al. identified that SR9009 prevented cellular senescence by activating NRF2 independently of REV-ERBs [40]. In line with the above findings, we demonstrated that SR9009 inhibited PCa cell viability and lethal PCS1 tumors through activation of LXRα instead of REV-ERBs. Evidence regarding the role of the REV-ERB-targeted gene BMAL1 in cancer cell biology is conflicting. Although several studies have demonstrated that BMAL1 promotes tumor progression [41, 42], many studies have revealed the tumor-suppressive role of BMAL1 [43–47]. We found in PCa that SR9009 upregulated BMAL1 expression instead of downregulating it, suggesting a potential synergetic role of SR9009 and BMAL1 in combating PCa. The specific mechanisms of SR9009-induced high BMAL1 expression remain unknown and should be investigated in further studies. In addition, previous research validated that SR9009 shows marked selectivity for LXRα over REV-ERBs (EC50 6.3 μM for LXRα and > 50 μM for REV-ERBα) [28]. LXRα is a nuclear receptor that is considered to play a pivotal role in lipid metabolism [48]. Our GSEA also found that SR9009 upregulated the steroid hormone biosynthesis pathway (Fig. 2D). In addition, several studies revealed that LXRα activation by the synthetic LXR agonists T0901317 and 22(R)-hydroxycholesterol inhibited PCa cell proliferation [49, 50]. We provide new evidence that SR9009 cannot be applied solely as a putative REV-ERB agonist but to be served as a novel LXRα agonist. FOXM1 is a crucial transcription factor in cancer and maintains cancer hallmarks by regulating its target gene expression [51]. From TCGA databases, we found that FOXM1 is aberrantly overexpressed in most cancer types and that its high expression was associated with worse survival outcomes. Targeting the FOXM1 pathway is an intriguing strategy for combating cancer. FOXM1 was also reported to be the master driver of lethal PCS1 tumors [20]. We showed that REV-ERBs did not regulate FOXM1 or its target gene expression. Instead, LXRα silencing or pharmacological activation by SR9009 regulated the FOXM1 transcription network. Although the direct regulation of FOXM1 by LXRα was validated in hepatocellular carcinoma and macrophages [29, 52], our study validated this regulation in PCa and provided a novel therapeutic strategy for targeting the FOXM1 pathway and PCS1 subtype. Our study demonstrated that SR9009 can serve as an effective treatment choice for highly aggressive and lethal PCS1 tumors by mediating the LXRα/FOXM1 pathway independently of REV-ERBs. Nevertheless, the other functions of SR9009 in PCa also warrant further exploration, such as the impact of SR9009 on metabolism, ferroptosis and cellular senescence. Although we showed that SR9009 could serve as both a FOXM1 pathway and PCS1 subtype inhibitor, the application of PCS classification in the clinic is still underway, and prospective clinical trials are warranted to validate the efficacy and safety of SR9009 in mCRPC treatment as a monotherapy or combination therapy. In summary, our study demonstrated for the first time that SR9009 could selectively kill PCa cells in a dose-dependent manner but had no impact on normal prostate cells. SR9009 could inhibit the most aggressive and lethal PCS1 tumors through FOXM1 pathway blockade. The effects of SR9009 on PCa cell proliferation are mediated by LXRα activation instead of REV-ERBs. Our study provides new insights into the function of SR9009 and into the discovery of potential useful drugs for PCa therapy. Fig. S1 Fig. S2 Fig. S3 Fig. S4 Fig. S5 Fig. S6 Table S1 Table S2 Table S3 Checklist Full and uncropped WB

PMC9649670

Ana Paula Corrêa Moneda,Lucas Amoroso Lopes de Carvalho,Luis Guillermo Teheran-Sierra,Michelli Inácio Gonçalves Funnicelli,Daniel Guariz Pinheiro

Sugarcane cultivation practices modulate rhizosphere microbial community composition and structure

10-11-2022

Computational biology and bioinformatics,Ecology,Microbiology

Sugarcane (Saccharum spp.) represents a crop of great economic importance, remarkably relevant in the food industry and energy supply chains from renewable sources. However, its conventional cultivation involves the intensive use of fertilizers, pesticides, and other agrochemical agents whose detrimental effects on the environment are notorious. Alternative systems, such as organic farming, have been presented as an environmentally friendly way of production. Still, the outcomes of different cropping systems on the microbiota associated with sugarcane—whose role in its health and growth is crucial—remain underexplored. Thus, we studied the rhizospheric microbiota of two adjacent sugarcane fields, which differ in terms of the type of farming system. For this, we used the sequencing of taxonomic markers of prokaryotes (gene 16S rRNA, subregions V3–V4) and fungi (Internal transcribed spacer 2) and evaluated the changes caused by the systems. Our results show a well-conserved microbiota composition among farming systems in the highest taxonomic ranks, such as phylum, class, and order. Also, both systems showed very similar alpha diversity indices and shared core taxa with growth-promoting capacities, such as bacteria from the Bacillus and Bradyrhizobium genera and the fungal genus Trichoderma. However, the composition at more specific levels denotes differences, such as the separation of the samples concerning beta diversity and the identification of 74 differentially abundant taxa between the systems. Of these, 60 were fungal taxa, indicating that this microbiota quota is more susceptible to changes caused by farming systems. The analysis of co-occurrence networks also showed the formation of peripheral sub-networks associated with the treatments—especially in fungi—and the presence of keystone taxa in terms of their ability to mediate relationships between other members of microbial communities. Considering that both crop fields used the same cultivar and had almost identical soil properties, we conclude that the observed findings are effects of the activities intrinsic to each system and can contribute to a better understanding of the effects of farming practices on the plant microbiome.

Sugarcane cultivation practices modulate rhizosphere microbial community composition and structure Sugarcane (Saccharum spp.) represents a crop of great economic importance, remarkably relevant in the food industry and energy supply chains from renewable sources. However, its conventional cultivation involves the intensive use of fertilizers, pesticides, and other agrochemical agents whose detrimental effects on the environment are notorious. Alternative systems, such as organic farming, have been presented as an environmentally friendly way of production. Still, the outcomes of different cropping systems on the microbiota associated with sugarcane—whose role in its health and growth is crucial—remain underexplored. Thus, we studied the rhizospheric microbiota of two adjacent sugarcane fields, which differ in terms of the type of farming system. For this, we used the sequencing of taxonomic markers of prokaryotes (gene 16S rRNA, subregions V3–V4) and fungi (Internal transcribed spacer 2) and evaluated the changes caused by the systems. Our results show a well-conserved microbiota composition among farming systems in the highest taxonomic ranks, such as phylum, class, and order. Also, both systems showed very similar alpha diversity indices and shared core taxa with growth-promoting capacities, such as bacteria from the Bacillus and Bradyrhizobium genera and the fungal genus Trichoderma. However, the composition at more specific levels denotes differences, such as the separation of the samples concerning beta diversity and the identification of 74 differentially abundant taxa between the systems. Of these, 60 were fungal taxa, indicating that this microbiota quota is more susceptible to changes caused by farming systems. The analysis of co-occurrence networks also showed the formation of peripheral sub-networks associated with the treatments—especially in fungi—and the presence of keystone taxa in terms of their ability to mediate relationships between other members of microbial communities. Considering that both crop fields used the same cultivar and had almost identical soil properties, we conclude that the observed findings are effects of the activities intrinsic to each system and can contribute to a better understanding of the effects of farming practices on the plant microbiome. Sugarcane (Saccharum spp.) stands out in Brazil as one of the most important socioeconomic crops. The agro-industrial system related to sugarcane processing in biorefineries is responsible for supplying the market with ethanol biofuels and sugar. Furthermore, sugarcane biorefineries can also perform the co-production of large amounts of useful products from wastes, such as bagasse, molasses, cane trash, filter mud, and vinasse. All of them have significant value-added based on the concept of circular bioeconomy and sustainability production. These concepts are congruent with the consumer’s concerns for healthier lifestyles and environmental care. In order to meet rising consumer demands for higher-quality products from a more sustainable production, there is a growing interest in the development and adoption of agricultural models that aims to conserve and enhance the quality of the soil leading to higher yields but taking into account the protection of the local environment and ecosystem services. These agroecological practices are the basis of organic farming systems, which prohibit the use of synthetic inputs, such as pesticides or fertilizers. Organic farming highlights the essential role of humus and organic matter for soil fertility and plant nutrition. The great challenge for organic production is to achieve higher crop yield stability. To overcome that, it is necessary a deep and integrated understanding of climate and biogeochemical cycles, pollination, soil structure and protection, water absorption, and biological interactions, among other processes. In this context, the soil microbiota can also play a fundamental role, especially those that inhabit the rhizosphere and colonize the plant tissues. Therefore, the soil and plant-associated microbiota has become the target of studies to identify the driving factors that shape microbial assemblage composition and structuration. In this sense, culture-independent methods for the investigation of plant microbiota, such as the sequencing of amplicons (or metabarcoding) from taxonomic marker regions, such as the 16S rRNA gene and the Internal Transcribed Spacer (ITS) region, for prokaryotes and fungi, respectively, have enabled to advance the understanding of microbial communities and their relationships. The research on sugarcane-associated microbiota aims to investigate the microbial diversity reservoir still unexploited to acquire knowledge about its role in modulation of plant development, pathogen defense, nutrient uptakes, and stress resistance. This is essential to constitute the foundation for the development of solutions to equilibrate higher productivity with sustainability for this crop. Recent studies indicate that different sugarcane genotypes can shape the associated microbiota by changing from keystone species to the richness of bacterial and fungal communities. The sugarcane rhizosphere is a very relevant ecosystem for deepening knowledge about the plant-associated microbiota. It is characterized as an environment of interaction events between the plant and microorganisms, through intensive chemical signaling exchange⁠. The assemblage of microorganisms in there are part of a complex interaction network, allowing the plant to modulate the microbiome for its own benefit, through the selection of microorganisms with suitable feature to meet its needs, favoring its healthy growth and development. Among these, there are the so-called plant growth promoters, which provide an increase in absorption of minerals, mobilization of nutrients, and a decrease in pathogens activity. Few studies have comparatively evaluated the impact of the organic versus conventional farming systems on the structure and composition of the rhizosphere microbiota in agricultural crops, especially on sugarcane. There is evidence that organic agriculture has positive effects on the microbial community, with increased richness and diversity. On the other hand, recent advances show that the practice of sugarcane monoculture following the precepts of conventional fertilization points to a significant impact on its associated microbiota in the short and long term. These impacts include the depletion of beneficial taxa, such as Rhizobium and Sphingomonas, while potentially phytopathogenic genera are enriched. However, by comparing organic and conventional systems, Orr et al. did not obtain any results that suggest significant effects on microbial communities, concluding that the environmental and chemical variables of the soil are the ones that really govern the present biodiversity. To evaluate and compare the microbial community of the sugarcane rhizosphere under two contrasting cropping systems, those being organic and conventional, we adopted the metabarcoding approach, which allowed us to find shifts in its composition, diversity, and structuring. The sugarcane rhizosphere samples for metagenomic DNA extraction were collected in 2018 between February and March at the São José Farm (organic farming system) and at the São Sebastião Farm (conventional farming system), both in Jaboticabal, São Paulo State, Brazil. The selected sugarcane field stands for sampling were very close (≈100 m; Supplementary Fig. 1), they comprised plants of the same cultivar (CTC9001) in the plant cane cycle and planted at a similar season/date (March 2017). We made the three samplings of rhizosphere material approximately 11 months after planting, near the end of vegetative growth and the beginning of the sugarcane maturation process. We selected the sampled points from a representative portion of the sugarcane fields from equidistant sampling points (≈50 m) (Fig. 1). The samples from the conventional cultivation field were named CRZ, and the samples from the organic cultivation field were named ORZ. The geolocation, environmental conditions, and sample collection dates (Supplementary Table 1). In the São José Farm, the sugarcane had been cultivated in a conventional system until the year 2000, when the conversion process started. During the samplings, the Farm had the Brazilian national certificate of organic production and the American certificate from the United States Department of Agriculture (USDA), besides complying with the regulations defined by the European Union, (EC) n. 834/2007. The procedures for preparing the soil for planting are similar in the sugarcane field stands under organic or conventional farming (i.e., plowing, harrowing, and subsoiling). In the renovation of the fields, leguminous (peanuts, soybeans, or crotalaria) were planted. After harvesting these crops, the remaining plant material is turned along with the soil in preparation. In the area of organic cultivation, for the sugarcane field renovation, there was an application of vinasse in the soil, and specifically in the planting furrow, there was supplementation with organic matter under decomposition process, which were obtained from filter pie, confined cattle manure, and also from the native vegetation of the farm. For replanting, fertilization is initiated with liming and phosphating using mineral fertilizers (limestone and phosphate rock). Commercial products for pest biocontrol and bioinoculants for atmospheric nitrogen fixation were used to promote plant healthy growth (Supplementary Table 2). In the conventional cultivation field, pH soil correction and fertilization were carried out with liming, gypsum, limestone, and NPK (N, P2O5 e K2O) in a proportion of 50-50-25, respectively. In addition, other fertilizers based on organic and inorganic compounds, and insecticides, fungicides, and herbicides were also applied (Supplementary Table 3). The bulk soil around the sugarcane plants was excavated considering a circumference of 0.5 to 0.6 m2 in diameter and a depth of 0.5 m., where the roots and soil adhered to the roots were stored in sterile plastic bags with a capacity of 20 L. The samples were stored in a cold box and taken to the laboratory for processing. The roots were shaken to break the remaining clumps and loosen the excess soil. After this, the soil firmly adhering to the roots was gently brushed out, sieved, fractionated into six (1.5 mL) microtubes, and then frozen in liquid nitrogen and stored at − 80 °C until the DNA extraction. Chemical and physical analysis of bulk soils were performed, considering a pool of 5 soil samples around each sampling point (approximately 2 m radius). For each sampling point, macronutrient analyzes were performed (Potassium [K], Phosphorus [P], Sulfur [S], Magnesium [Mg] and Calcium [Ca]), micronutrients (Boron [B]), Manganese [Mn] , Iron [Fe], Copper [Cu], Zinc [Zn] and Aluminum [Al]), organic matter (OM), as well as analyzes related to soil acidity (pH, potential acidity [H + Al], sum of bases [SB], base saturation [V%], cation exchange capacity [CTC]), Aluminum saturation [m%], according to the methods described by Instituto Agronômico de Campinas – IAC. For granulometric evaluation, the fractions of Clay, Silt and Sand (“Coarse sand” and “Fine sand”) were determined according to the manual of the Brazilian Agricultural Research Corporation—Embrapa. The mean values for each of the physico-chemical soil properties obtained for sugarcane fields under the two contrasting systems (organic vs. conventional) were statistically compared using the “compare_means” package, the “ggpubr” function (v. 0.4.0), of the R statistical program (v 4.0.2) . The extractions of DNA from microbial communities in the rhizosphere soils were performed using the commercial DNeasy PowerSoil Kit (Qiagen®). The manufacturer's standard protocol was used with the modifications made by de Souza et al., which consisted of heating at 65 °C for 10 min. after addition of reagent C1 and two washing steps with ethanol. In order to remove contaminants, such as PCR inhibitors, the DNA was submitted to the purification step with Agencourt AMPure XP kit (A63881, Beckman Coulter®), following the manufacturer's protocol with 1.2 × reagent to sample ratio. The DNA samples were quantified by a spectrophotometer (NanoDrop®) and fluorometer (Qubit®) and stored at − 80 °C. To access the rhizosphere prokaryotic communities, PCR amplification products (amplicons) of the 16S rRNA gene were used, targeting the fragment comprising the V3 and V4 hypervariable regions of the gene (primers 319F [5′-ACTCCTACGGGAGGCAGCAG] and 806R [5′-GGACTACHVGGGTWTCTAAT]; 469 bp). To access the rhizosphere fungal communities, the Internal Transcribed Spacers (ITS) were used, and in this case, the target was the ITS2 region (primers ITS9F [5′-GAACGCAGCRAAIIGYGA] and ITS4R [5′-TCCTCCGCTTATTGATATGC]; variable size). The amplicon libraries for Next Generation Sequencing (NGS) were constructed in two PCR steps. In the first PCR step, target-specific amplification was performed, which adds the Illumina adapter and 4 to 6 random bases adjacent to the forward and reverse primers to incorporate heterogeneity into the read sequences. In the second step, Illumina's N7 and S5 barcodes were incorporated for indexing the reads in the subsequent multiplex sequencing. The PCR steps followed the same protocol described in the Illumina library preparation technical manual for MiSeq, using PCRBIO ULTRA MIX kit for PCRs. The samples were purified with the Agencourt AMPure XP kit (A63881, Beckman Coulter®) and also with the DNA Clean & Concentrator kit—Zymo Research®. In total, 12 libraries were built, 6 for each of the farming systems (i.e. three sample units (replicates) for prokaryotic communities [16S] and three for fungal communities [ITS]). The three sample units correspond to each one of the sampling points (Supplementary Fig. 1). The sequencing of the 16S and ITS amplicon libraries was performed on the Illumina® MiSeq platform. The amplicon fragments were sequenced using the MiSeq Reagent Kit v.2 (600 cycles) producing 300 bp reads in paired-end mode (2 × 300 bp). The demultiplexing of the sequence reads was performed using “bcl2fastq” software (v2.20.0.422) (Illumina®) with default settings. The remaining reads, whose barcodes were not identified, were processed with “deML” program (v.1.1.3) using the -rgqual 90 and -wrongness 80 as parameter settings. The paired-end reads corresponding to the amplicons were merged using the “PEAR” tool (v.0.9.11), with a minimum overlap of 15 bp. The amplicon sequences corresponding exactly to the V3–V4 regions of the 16S rRNA and ITS2 gene were extracted using the “search_pcr2” command of the “USERCH” toolkit (v.11.0.667). For the microbiome analysis, we used the Divisive Amplicon Denoising Algorithm – “DADA2” pipeline (v1.14.1) to infer and quantify Amplicon Sequence Variants (ASVs). The pipeline was implemented in R (v 4.0.2). Both sets of merged read sequences, i.e. from the region V3–V4 of the 16S rRNA gene and from the ITS2 region, were filtered using the “filterAndTrim” function with the following parameter settings: maxN = 0, truncQ = 2 and maxEE = 2. The error rate per sample was estimated based on the error model using the “learnErrors” function and sequence redundancy was removed using the “derepFastq” function. Finally, the sequences were corrected based on the error models obtained previously with the “dada” function and chimeric sequences were removed using a “removeBimeraDenovo” function. The taxonomic assignment of ASVs was performed by the RDP Naive Bayesian classifier (Wang et al. 2007) through DADA2 function “assignTaxonomy” with the following parameter settings: minboot = 80, and refFasta with the file path corresponding to the suitable reference database, i.e. the RDP database training set (v.16) for the 16S dataset and the UNITE (v.8.2) database for the ITS2 dataset. For the alpha-diversity indices were estimated from the absolute counts obtained for ASVs of prokaryotes (mainly bacteria) and fungi of both farming systems were considered. From these data, the parameters of richness (Chao1) and diversity (Shannon and Gini-Simpson indices) were estimated for both datasets. For this, the ASV counts were transformed into a “phyloseq” object (package “phyloseq”) and subsequently submitted to the “alpha” function of the R package “microbiome”. The diversity measures were statistically compared using the “compare_means” function of the R package “ggpubr” (v. 0.4.0), using the Wilcoxon non-parametric test for comparing means, and considering a p-value ≤ 0.1 as statistically significant. For the beta-diversity analysis, the absolute counts for ASVs in both datasets were transformed to compositional matrices (i.e. normalization by TSS – Total Sum Scaling, or relative abundances), through the “transform” function of the “microbiome” package. From the transformed values, we calculated the distances of the Bray–Curtis dissimilarities (“distance” function of the “phyloseq” package). The distance matrices were used in a statistical comparison between the farming systems by a PERMANOVA (Permutational Multivariate Analysis of Variance) analysis, using the “adonis” function of the “vegan” package, considering a p-value ≤ 0.1 as statistically significant. Then, the matrices were used to obtain a dendrogram, resulting from the hierarchical grouping of the samples, and provided as input for a Principal Coordinate Analysis (PCoA). The recognition of the core microbiome of sugarcane rhizosphere (i.e. the one common for the farming systems) was done through the identification of microorganisms (prokaryotes or fungi) with the high prevalence and abundance in all the samples, independently of the label corresponding to the farming system. For this, we considered those microorganisms present in at least 90% of the samples, with a minimum relative abundance of 1%. The calculations and visualizations of the core microbiome were obtained through the “plot_core” function of the R package “microbiome”. The identification of taxa, from each taxonomic level, which are present in significantly different abundances between farming systems, was done using the “DESeq2” approach. For this, we submit the prokaryotes and fungi datasets to the “MicrobiomeAnalyst” platform. In the web platform, the datasets with ASV counts were normalized using Total Sum Scaling (TSS), and also provided as input for DESeq2 analysis, in which the Wald test was performed to evaluate statistical significance, considering a p-value ≤ 0.01 as statistically significant. The functional capabilities of the microbiomes from the sugarcane rhizosphere under the considered farming systems were predicted using the “PICRUSt2” program (v.2.3.0-b). For this, we used the metabolic pathway database “MetaCyc” (Caspi 2006) as a reference for the functional annotations. A comparison of the annotated pathways between the datasets of corresponding farming systems was performed to identify enriched metabolic pathways associated with one of them. For this purpose, the “STAMP” program (v.2.1.3) was used for the application of White’s non-parametric t-test to compare the means, considering a p-value ≤ 0.05 as statistically significant. The co-occurrence networks of the identified genera of fungi and prokaryotes were elaborated based on Pearson's correlation coefficients (r). The coefficients were obtained from the normalized ASVs and were forwarded to the “Correlation Analysis”. The results were filtered using the following criteria: absolute correlation threshold of 0.5 (r ≥ 0.5 or r ≤ − 0.5) and p-value ≤ 0.05. Subsequently, based on the filtered results, the relationships between each genus in addition to their respective correlation coefficients were suitably formatted and used as input for the “Cytoscape” program (v.3.8.0). In this program, in addition to visual representations, the topological network parameters, such as the measures of centrality, were obtained by using the Cytoscape plugin “NetworkAnalyzer”. The physicochemical analysis of the soils showed few significant differences between the crops. Conventional cultivation was slightly more acidic, with lower pH, sum and base saturation values, in contrast to the higher percentage of aluminum saturation when compared to the organic cultivation. All values obtained in soil analysis, as well as statistical comparisons, can be found in Supplementary Table 4. The high-throughput sequencing of sugarcane rhizosphere soil using the Illumina MiSeq instrument resulted in a total of 329,685 paired-end reads from the 16S rRNA gene (V3–V4 region), with an approximate average of 55,000 pairs per library. Of these, about 227,607 were successfully assigned as ASVs. For the ITS2 amplicons, the sequencing resulted in a total of 311,269 paired-end reads, an average of 51,000 pairs per library, of which 276,069 were successfully assigned as ASVs. Among ASVs, 80,465 and 183,705 received a taxonomic assignment at genus level, respectively, for 16S and ITS datasets. (Supplementary Table 5). In the prokaryotic taxonomic composition, most of the most abundant taxa belongs to the phylum Proteobacteria in both management systems, followed by the phyla Actinobacteria, Acidobacteria, Firmicutes, Gemmatimonadetes and Bacteroidetes (Fig. 1A). It is notable the presence of groups categorized as unclassified, grouping taxa in lower abundances, which probably are unknown in the reference database or have sequences with insufficient evidence for the taxonomic definition. Considering the taxonomic level of order, among the most abundant, Rhizobiales, Actinomycetales, Bacillales, and Sphingomonadales stand out (Fig. 1B). In the fungal taxonomic composition, the most abundant phyla are Ascomycota and Basidiomycota, followed by Mucoromycota, with relatively low abundance (Fig. 2A). Among the most abundant orders, Hypocreales, Perosporales, and Sordariales stand out (Fig. 2B). The richness and diversity indices of taxa in the prokaryotes dataset in both farming systems did not show significant differences, through the comparison of means (p-value > 0.1) (Fig. 3; Supplementary Table 6). Regarding the fungal dataset, there was only one difference in diversity, considering the Gini-Simpson index (p-value ≤ 0.1), pointing to a higher diversity in the conventional system. The other indices showed no significant differences between the farming systems, considering the same descriptive level of statistical significance (Fig. 3; Supplementary Table 6). In the two-dimensional Principal Coordinate Analysis (PCoA) plot, based on the Bray–Curtis index (Fig. 4), there was a significant description of the system through a projection of prokaryotic taxa (16S), without major loss of information, preserving 60.46% of the variance in the data. The PcoA 1 index describes most of the variation (accounting for 35.37%). One of the samples from the organic system appears far from the others of the same group, resembling samples from the conventional system, suggesting that it is an outlier. However, even considering this discrepant sample, it is possible to notice the significant separation of the bacterial compositions concerning the farming system factor (PERMANOVA; p-value ≤ 0.1). The PCoA for the fungal dataset (ITS) reveals that the samples present a clustering profile similar to that observed for the prokaryotic dataset, i.e., the samples from the organic and conventional systems are well defined and showed differences between the managements (Fig. 5). The data projection for the fungi dataset preserves 60% of variance in the data, thus enabling a significant description of the system, without major loss of information. The descriptive level of significance by the PERMANOVA analysis was 0.1. The case is similar to the evaluation made for the prokaryotic dataset, thus also representing a significant difference between the systems. For the 16S rRNA data, 7 phyla, 12 classes, 12 orders, 9 families and 5 genera were considered to belong to the essential microbiome (Fig. 6). Among these, the phylum Proteobacteria stands out, with a relative abundance of 35% in all samples, followed by Actinobacteria, with a relative abundance of 15% in the samples. The core fungal microbiome comprised 2 phyla, 5 classes, 7 orders, 7 families, 5 genera and 3 species (Fig. 7). The most prevalent essential phylum was Ascomycota, with a relative abundance of 80% in all samples. The phylum with the lowest prevalence was Basidiomycota, with a relative abundance of 8% in the samples. Its only representative was the species Saitozyma podzolica. In total, 14 bacterial taxa were found to be differentially abundant in both farming systems (Fig. 8; Supplementary Table 7). Of these, 8 taxa had significant differences in abundance, considering a p-value ≤ 0.01 in a conventional farming system. Among them, are the phylum Saccharibacteria, the class Flavobacteria, and the order Flavobacteriales. Two families are highlighted, Erythrobacteraceae and Flavobacteriaceae, in addition to three genera, Flavobacterium, Segetibacter, and, to a lesser extent, Devosia. In the organic system, 6 taxa were differentially abundant in relation to the conventional one, considering a p-value ≤ 0.01. Among these, two families are Labilitrichaceae, Cryptosporangiaceae, and, to a lesser extent, Pseudonocardiaceae. The three genera identified, Jatrophihabitans, Pelomonas, and Pseudonocardia, were the most abundant in the organic system, respectively to the referred families. For fungi, 60 differentially abundant taxa were found, demonstrating greater expressiveness when compared to bacterial taxonomy (Fig. 9; Supplementary Table 8). Of these, 36 taxa were differentially abundant in conventional cultivation and 24 taxa in organic cultivation, considering a p-value ≤ 0.01. In addition to the differences observed in the composition and diversity, we evaluate the differences in the predicted functional profiling of microbial communities found in both the sugarcane rhizosphere of the contrasting farming systems. In total, 324 metabolic pathways were identified in the prokaryotic dataset, among which 24 pathways were slightly differentially enriched, considering a p-value ≤ 0.05. Of these, 17 were differentially abundant in the organic system and 7 in the conventional one (Fig. 10). Regarding the predicted functional profiling of fungal communities, 69 metabolic pathways were identified, of which 14 were more prominent within the analyzed samples, considering a p-value ≤ 0.05. Of these, 11 had higher proportions in the organic system and only 3 in the conventional one (Fig. 11). To investigate the structures of microbial communities, co-occurrence networks of genera identified in the prokaryotic and fungal datasets were performed. Among all the prokaryotic genera, 110 passed the filtering criteria (minimum correlation coefficient and significance) and were considered for the build of the co-occurrence networks for the conventional and organic farming systems (Fig. 12). The genus Burkholderia stands out with greater connectivity with other bacterial genera, totaling 25 correlations (Supplementary Table 9). The genera Nitrosospira, Ktedonobacter, and Kribbella had the highest values for betweenness centrality and could be considered key taxa for the prokaryotic network (Supplementary Table 10). Regarding the fungal dataset, 187 taxa passed the filtering criteria and were considered for the build of the fungal co-occurrence network, with Fusarium being the most abundant genus present in the farming systems (Fig. 13). The genus Ascobolus was more connected to other fungal genera, totaling 27 connections (Supplementary Table 11). The genera Rhizopus, Scopulariopsis, Podospora, and Cladophialophora had the highest values for betweenness centrality, and suggestively stand out as key taxa in this network (Supplementary Table 12). The main purpose of this study was to carry out a comparative and investigative analysis of microbial communities present in the sugarcane rhizosphere, under contrasting farming systems, in order to understand the ecological dynamics in terms of composition and diversity. It is possible to note that the physical and chemical parameters of the soil have a great similarity between crops, with a small but statistically significant difference in the acidity parameters (Supplementary Table 4). However, such differences may have not been enough to cause a great impact on the large-scale composition (Figs. 1 and 2) and microbial diversity between the systems (Fig. 3), despite the soil pH may have a strong correlation with microbial diversity, the general composition of the community or for the relative abundance of individual taxonomic groups. The slightly higher acidity in the conventional soil was possibly caused by the application of agrochemical inputs and intensive use of nitrogen fertilization. It is also recognized that long-term consecutive cultivation of sugarcane leads to a decline in soil pH and can cause considerable changes in the composition and function of the microbiota. The alpha diversity indices did not show significant differences between farming systems (considering a p-value ≤ 0.1), except in the case of fungal diversity provided by the Gini-Simpson index, which was higher in the conventional system (Fig. 3). For the beta diversity indices, the multivariate permutational analysis of variance (PERMANOVA) showed that there are significant compositional differences (considering a p-value ≤ 0.1) when evaluating only the treatment (conventional or organic) as a descriptive factor of heterogeneity between both systems (Figs. 4 and 5). The results show that the effects arising from these farming systems occur in specific microbial taxa, and this does not have a systemic impact on the diversity of the entire microbial community, where variations in certain groups can be counterbalanced by opposite variations in others. For sugarcane, genotype influences on the modulation of the associated microbiota are reported. Even so, these results are according to previous studies which placed the soil properties (i.e., soil texture, water content, and soil type) and host plant (i.e., plant species) as the main drivers of the rhizosphere microbiome assembly, since the soils were very similar and the cultivar was the same in both systems. The higher fungal diversity in conventional cultivation suggests the potential of a plant to select fungal communities in the rhizospheric environment through the composition of its root exudates. The collections of rhizospheric material in each crop were carried out at the end of the vegetative growth period (11 months) and the beginning of the maturation of the sugarcane field, and it can be inferred that different stages of plant growth can determine its composition, amount of rhizodeposits present and its associated microbiome. During plant growth, many ecological succession processes can occur, resulting in new habitats and an increase in the breadth of niches, in our study, we characterized the microbiota only at one specific time point. As seen in a previous study by de Souza et al., core microorganisms in sugarcane, regardless of fertilization, can bring benefits and plant vitality. In our study, we identified members of the sugarcane rhizosphere core microbiome (Figs. 6 and 7). Notably, among bacterial genera, Sphingomonas, Gemmatimonas, Gaiella, Bacillus, and Bradyrhizobium, and among the fungal genera, Trichoderma stand out as potential plant growth promoters, phytopathogen inhibitors and participants in soil nutrient cycling. The same was observed for Antennariella placitae and Saitozyma podzolica, underreported fungal species that demonstrate potential for biological control in rice and apple plants. The microbial community structure, even in environments with climatic and soil type similarities, may differ according to the agricultural practice employed, selecting specific microorganisms. In the conventional system, differentially abundant bacteria were found (Fig. 8), such as those of the Flavobacterium genus, which can perform heterotrophic denitrification and degradation of various pesticides. The Devosia genus, also found differently abundant in the conventional system (Fig. 8), was often found in environments contaminated with hydrocarbon pesticides and hexachlorocyclohexane, considered a biodetoxification agent. The increased abundance of specific microbial taxa, possibly caused by long-term fertilization, may demonstrate a direct link with soil nutrients, as seen in organic farming (Fig. 8). In it, the differentially abundant bacterial phyla were classified as Actinobacteria and Betaproteobacteria, which ecologically have a copiotrophic life strategy, with rapid growth in soils with high nutritional availability. It has been reported that microbial communities associated with organic management practices tend to be copiotrophic, due to high concentration of nutrients, greater availability and utilization of nitrogen and organic carbon. Organic soils are major sources of recalcitrant carbon, which explains the high prevalence of Actinobacteria in this system, confirming the importance of this taxon in the carbon cycle and for the decomposition of this element. This confirms that classifications at high taxonomic levels can predictably respond to environmental variables, resulting in high ecological coherence. Ecological coherence demonstrates that the abundance of certain bacterial phyla can change directionally to the type of long-term fertilization employed. In the organic system (Fig. 9), members of the phylum Ascomycota are predominant. Most saprotrophic microfungi falls under this phylum, and stands out for their importance in the decomposition of organic substrates. Also, it is acknowledged that Ascomycota is positively associated with organic matter and nitrogen present in the sugarcane soil. Although differences in these elements are not noticed in our study, it could be an ongoing shift related to the application of organic inputs during the planting period in this system. In the conventional system (Fig. 9), the fungal phyla were more heterogeneous, contrary to what was reported by Lupatini et al.. As seen by Paungfoo-Lonhienne et al., the use of NPK fertilizer has been associated with the presence and increase of fungal biomass in the sugarcane rhizosphere, which may lead to changes in the composition of fungal communities. Long-term application of mineral fertilizers provides large amounts of nutrients to the soil. These introduced nutrients can increase exudation and alter the fungal community present. This is because fertilization directly influences the physiological state of the plant and favors the release of these exudates. The use of inorganic nitrogen can reduce the dependence of rhizosphere communities on the use of plant-derived carbon and activate many dormant fungal species. This applies to the fact that most fungi are heterotrophs and highly dependent on exogenous carbon for their growth. The release of root exudates may gradually decrease or cease as the plant matures and reaches senescence and the microorganisms obtain their nutrients from the soil. The functions and metabolic pathways associated with the rhizosphere microbiota from both systems were predicted and evaluated using an enrichment analysis (Figs. 10 and 11). Although there were a few shifts between the systems, it was not possible to identify notable associations with the influences caused by the type of agricultural management, since the vast majority of enrichments were related to the structural and biological processes of the microbiomes. Despite inoculation with nitrogen-fixing bacteria in the organic system (Supplementary Table 2), pathways related to this process were not affected by treatments. Nonetheless, it is worth mentioning that the process of biological nitrogen fixation catalyzed by nitrogenases is dependent on the micronutrients Iron (Fe), Vanadium (V), and, mainly, Molybdenum (Mo), which has not been evaluated in the present study. As seen by Schmidt et al., the type of management employed can determine the microbial community structure, i.e. the taxa and their interactions in a co-occurrence network, leading to important ecological and agricultural inferences. The analyzes of bacterial and fungal co-occurrence networks in organic and conventional farming systems demonstrate distinct patterns of connections, through different microbial identities and abundances, even though these crops share similar climate and soil conditions (Figs. 12 and 13). These microbial networks comprise parasitic, amensalistic, commensalistic, synergistic, or mutualistic interactions that influence each of their constituents and may produce effects on plant health and soil fertility. Our results suggest, through high values ​​of betweenness centrality, the presence of key interconnected taxa in the network that are highly important for the formation of microbial communities in their host plants, controlling or inhibiting the colonization by other microorganisms. Both systems have highly connected key taxa with different identities, demonstrating that these crops have important taxa that vary considerably. Betweenness centrality is usually described as an indication of key taxa, although this metric can be confirmed only through experimental validation. The highest intermediation centrality in the bacterial network was represented by the genus Nitrosospira, a well-recognized ammonia oxidant, present in high abundance in conventional cultivation. A great abundance of Nitrosospira in soils that receive nitrogen fertilization has been reported, which may lead to a significant increase in the process of soil nitrification compared to organic treatment. This specialized metabolic function present in Nitrosospira may be critical to the stability of the soil microbiome. In the fungal network, Rhizopus (Mucoromycota) present exclusively in conventional cultivation was considered a key taxon, with the highest betweenness centrality value. This genus may have had its abundance favored in a conventional system by the nitrogen fertilization used, which suggestively led to an increase in fungal diversity in this system. The greater diversity may have been driven by the high nutritional increment and the rhizospheric exudation stimulated by the inorganic nitrogen fertilization used. Through this study, we could identify slight variations in the rhizosphere microbiome of sugarcane plants when comparing organic and conventional farming systems. We could not directly associate the identified variation with the physical and chemical properties of the soil, because we do not find substantial evidence indicating that the organic or conventional farming system influenced these soil properties. It is improbable that the slight observed differences in pH have a direct relationship with the differences in microbial composition and diversity observed between these crops. The results show that there are some differences in beta diversity related to the systems. However, such differences could not lead to a substantial effect on the alpha diversity and taxonomic composition at phylum and order levels, according to descriptive levels of statistical significance. Despite this, our study allowed us to recognize that the contrasting systems present the presence of differentially abundant taxa when analyzed at more specific levels, presumably caused by the farming systems. With this, we can assume that agricultural practices can subtly influence the rhizosphere microbiota. The management systems suggestively may have influenced the structure of interactions revealed by the co-occurrence networks of both microbiotas. The rhizosphere involves different types of interactions between microorganisms, through their root exudates that can shape the structure and a large part of the composition and activities of microbial communities. In particular, in the case of fungi, we can clearly observe differences in their structuring due to changes in the abundance of certain genera and increased diversity caused by the conventional cultivation system, which lead to changes in ecological relationships. In addition, the central microbiome of the sugarcane rhizosphere, that is, the microorganisms independent of the adopted cropping system, revealed taxa known as plant growth promoters. We can consider that the understanding of these microbial relationships is fundamental for the development of a more sustainable agriculture. The great diversification of factors that involve and change the composition and structure of communities, in addition to the type of agricultural practice, leads to the need for deeper analysis. Thus, the types of regimens employed and their effects on the microbial community should be analyzed more comprehensively, using molecular approaches and identifying more precisely the proportion of the typical variations. This means that we cannot rule out the possibility of more expressive differences considering other conditions, for example, other plants, soils, climates, handling, collections, preparations, products used, crop rotation, number of consecutive harvests (cuts), in particular the time of conversion to the organic system. Thus, more research is needed to investigate the impact of each of these factors considering long-term agricultural systems. Supplementary Information.

PMC9649678

Tong Liu,Cheng Liu,Meisi Yan,Lei Zhang,Jing Zhang,Min Xiao,Zhigao Li,Xiaofan Wei,Hongquan Zhang

Single cell profiling of primary and paired metastatic lymph node tumors in breast cancer patients

10-11-2022

Breast cancer,Breast cancer,Tumour immunology

The microenvironment of lymph node metastasized tumors (LNMT) determines tumor progression and response to therapy, but a systematic study of LNMT is lacking. Here, we generate single-cell maps of primary tumors (PTs) and paired LNMTs in 8 breast cancer patients. We demonstrate that the activation, cytotoxicity, and proliferation of T cells are suppressed in LNMT compared with PT. CD4+CXCL13+ T cells in LNMT are more likely to differentiate into an exhausted state. Interestingly, LAMP3+ dendritic cells in LNMT display lower T cell priming and activating ability than in PT. Additionally, we identify a subtype of PLA2G2A+ cancer-associated fibroblasts enriched in HER2+ breast cancer patients that promotes immune infiltration. We also show that the antigen-presentation pathway is downregulated in malignant cells of the metastatic lymph node. Altogether, we characterize the microenvironment of LNMT and PT, which may shed light on the individualized therapeutic strategies for breast cancer patients with lymph node metastasis.

Single cell profiling of primary and paired metastatic lymph node tumors in breast cancer patients The microenvironment of lymph node metastasized tumors (LNMT) determines tumor progression and response to therapy, but a systematic study of LNMT is lacking. Here, we generate single-cell maps of primary tumors (PTs) and paired LNMTs in 8 breast cancer patients. We demonstrate that the activation, cytotoxicity, and proliferation of T cells are suppressed in LNMT compared with PT. CD4+CXCL13+ T cells in LNMT are more likely to differentiate into an exhausted state. Interestingly, LAMP3+ dendritic cells in LNMT display lower T cell priming and activating ability than in PT. Additionally, we identify a subtype of PLA2G2A+ cancer-associated fibroblasts enriched in HER2+ breast cancer patients that promotes immune infiltration. We also show that the antigen-presentation pathway is downregulated in malignant cells of the metastatic lymph node. Altogether, we characterize the microenvironment of LNMT and PT, which may shed light on the individualized therapeutic strategies for breast cancer patients with lymph node metastasis. Metastasis is the most prominent cause of cancer morbidity and mortality. Due to the special structure of lymph node (LN) vessels in tumors, tumor cells tend to metastasize to LN tissue, and tumor cell metastasis to LNs is an early manifestation of metastatic tumors. The microenvironment of lymph node metastasized tumors (LNMTs) is considered to be immunosuppressive; however, the characteristics and specific mechanisms of various immune cells are not clear. There is an apparent need for the therapeutic targeting of LNMT, rather than of the PT, to secure proper antitumor T-cell generation and timely tumor infiltration. Studies have found that metastatic LNs can affect the immune response of tumors. For example, targeting metastatic LNs can significantly enhance the therapeutic effect on PTs, which may represent an important strategy for improving patient survival. Therefore, it is important to characterize and differentiate between the LNMT and PT microenvironments. Genomic and transcriptomics technologies can help us to understand the changes of tumor cells during metastasis to distant organs. In recent years, the emergence of and advances in single-cell sequencing technology have provided a new and precise approach for understanding the complexity of genetic heterogeneity in tumor evolution and tumor cell metastatic progression. Tumor cells in a state of partial epithelial–mesenchymal transition located in the periphery of the original tumor are more likely to metastasize than are the cells inside the tumor. It is well known that tumors form as a consequence of interactions between malignant cells and the microenvironment. Breast cancer is the most common cancer in women worldwide and has a high mortality rate. The molecular characteristics of breast cancer vary considerably among the different subtypes, and specific therapeutic approaches are required for each classification. Therefore, understanding the composition of different types of breast cancer has great clinical value. Thus far, several studies have used single-cell sequencing to investigate breast cancer. For instance, one study explored the mechanism of treatment resistance of malignant cells in triple-negative breast cancer (TNBC) by using single-cell copy number variation (CNV) sequencing technology, while another study on TNBC characterized tumor-infiltrating immune cells. Research involving single-cell analysis of breast cancer has mainly centered on immune cells or tumor cells, but there is currently a lack of systematic research on the interactions among various cell types, and the relevant published articles have primarily focused on TNBC. It has further been shown that luminal and human epidermal growth factor receptor 2 (HER2)+ patients have a higher risk of LN metastasis than patients with TNBC. Therefore, we chose to direct the focus of our study on non–basal-like breast cancer. In this work, we aim to investigate the regulatory mechanism of the tumor microenvironment (TME) that may contribute to malignant cell metastasis and the colonization of LNs, with a particular focus on uncovering the differences between the PT and LNMT in non-TNBC patients using single-cell RNA sequencing. Our analyses reveal that the microenvironment of metastatic LNs is more conducive to tumor cell survival than is that of tumors in situ due to the lower immune cell activity of metastatic LNs. Moreover, we identify a type of cancer-associated fibroblast (CAF) expressing PLA2G2A that can interact with immune cells and that is enriched in HER2+ breast cancer. Our data provide insights into the mechanism of LNMT and immune infiltration in tumors and may be a valuable reference for the clinical application of immunotherapy for breast cancer metastasis. To characterize the TME of the PT and LNMT in patients with breast cancer, we collected paired tissues of LNMT and PT from 8 treatment-naïve patients with breast cancer subtypes including luminal A, luminal B, and HER2+. These tissues were separated into single cells, and we obtained a total of 118,845 cells sequenced by using 10x Genomics 5′ mRNA and T cell receptor (TCR) sequencing methods (Fig. 1a). Hematoxylin and eosin (HE) staining showed the gross appearance of metastatic LNs, indicating a high frequency of LNMT among the enrolled patients (Supplementary Fig. 1a). We used BBKNN integration to integrate cells from different patients (Supplementary Fig. 1b). All of the cells could be divided into the following 9 major types according to their canonical markers: B cells (CD3D, CD79A), CD4 T cells (CD3D, CD4), CD8 T cells (CD3D, CD8A), NK cells (GNLY), myeloid cells (LYZ), epithelial cells (EPCAM and KRT19), CAFs (PDGFRA), perivascular-like (PVL) cells (RGS5), and TECs; (PLVAP; Supplementary Fig. 1c–e). We found that the cell types in PT varied largely across the patients, but those in LNMT were similar (Supplementary Fig. 1f). Then, we further clustered and annotated the cells into 40 different cell clusters according to their specific marker genes (Fig. 1b, Supplementary data 1). To investigate the enrichment of cell types in PT and LNMT, we calculated the percentage of the tissues within each cluster. The data showed that B cells and CD4 T cells, which are well-known cellular components of normal LNs, are enriched in the microenvironment of LNMT. In contrast, epithelial cells, CD8 T cells, CAFs, and mast cells were abundant in PT (Fig. 1c, d). The difference in cell types between PT and LNMT was also shown in the UMAP embedding plot which was colored according to tissue type (Fig. 1e). To further prove that the cells we collected were intratumoral, we randomly picked LN tissues of 4 patients from the 8 patients in the present study to perform spatial transcriptomics. The results showed that the ratio of tumor cells was over 50% in all 4 patients and over 75% in 3 of them, suggesting that the samples were from LN metastases (Fig. 1f, Supplementary Fig. 1g). These data indicated that the cell types of the metastatic microenvironment of LNs differed significantly from those of PTs. LNs are central to immune cell circulation and maturation. To determine why malignant cells can survive in LNs without being eliminated by the immune cells, we analyzed the features of immune cells both in PTs and LNMTs. We annotated T cells and NK cells into 15 clusters, including 5 CD8 T cell subsets, 7 CD4 T cell subsets, γδ T cells, and 2 NK cell subsets (Fig. 2a, Supplementary Fig. a & b). The data showed that T cells and NK cells from the 2 tissues were distributed differently and exhibited disparate transcription programs (Fig. 2a). We found that these 2 types of NK cells expressed tumor-suppressing genes (XCL1 and XCL2 for NK-C1-XCL1, cytotoxic genes for NK-C2-GZMH) in higher proportions in PTs than in LNMTs (Supplementary Fig. 2c). To track the development trajectories of CD8 T cells, we employed a diffusion embedding map to visualize CD8 T subsets and found continuous developmental progression (Fig. 2b). CD8-C1-CD8B was present at the initial stage of CD8 T cells differentiation, in which there was also a high expression of CCR7 and SELL (Fig. 2b, Supplementary Fig. 2c), which are the markers of naïve T cells (TN). CD8-C2-CCL5, characterized by higher expression of cytotoxic markers and high expression of HOPX (Supplementary Fig. 2b, d), was present in the next stage after CD8-C1-CD8B. CCL5+ T cells further branched into CD8-C3-GZMK or CD8-C4-HSPA1A (Fig. 2b). CD8-C3-GZMK cells were defined as effect memory T cells (TEM) due to their expression of cytotoxic markers like NKG7, GZMA, and GZMK; meanwhile, CD8-C4-HSP1A1, which expressed cytotoxic markers and a high level of CD69 but a low level of ITGAE, represented CD69+ITGAE- tissue-resident memory T cells (TRM) (Fig. 2b, Supplementary Fig. 2b). CD8-C5-CXCL13, which differentiated from CD8-C3-GZMK, was considered to be the terminal state of differentiation (Fig. 2b). CD8-C5-CXCL13 expressed cytotoxic markers and exhausted makers including CTLA4, PDCD1, and LAG3 (Supplementary Fig. 2b, e), and was characterized as exhausted or pre-exhausted CD8 T cells. We then performed principal component analysis (PCA) to investigate CD8 T cells in the microenvironment of LNMTs and PTs. Principal component (PC) 2 was the most prominent component distinguishing CD8 T cells between PT and LNMT of the first 20 PCs (Supplementary Fig. 2f). We used a PCA embedding map colored according to tissue type to determine the distribution of CD8 T cells (Fig. 2c). The most variant genes contributing to PC2 were CCL5, CCL4, chemokines involved in T cell recruitment, MHC class II genes functioning in antigen presentation, and cytokines such as GZMA and GZMH, which are known for their cytotoxic function in T cells (Supplementary Fig. 2g). The CD8 T cell activation signature also showed CD8 T cells in PTs to have a higher activation score than those in LNMTs (Fig. 2d), supporting the previous findings. Next, we analyzed the characteristics of CD4 T cells. Single-cell sequencing has recently been used to identify the tumor-suppressing functions of CD4 T cells, for example, the function of GZMK+CD4+ T cells in bladder cancer. It has been found that CD4 T cells are more complex than initially believed and need to be further explored. In our study, CD4 T cells were classified into 7 clusters according to their specific gene expressions (Fig. 2a, Supplementary Fig. 2b). According to their markers, CD4-C1-RPL corresponded to naïve CD4 cells, while CD4-C2-ANXA1 and CD4-C3-YPEL5 corresponded to memory or pre-memory CD4 cells. These 3 clusters were enriched in LNMTs (Fig. 2a, Supplementary Fig. 2c). CD4-C6-CXCL13 highly expressed CXCL13, which is reported to attract B cells, and corresponded to Tfh or Tfh-like cells. Interestingly, we found notable heterogeneity among CD4+CXCL13+ T cells across the 2 microenvironments (Fig. 2e, f). To study the heterogeneity of CD4+CXCL13+ cells, we re-clustered CD4+CXCL13+ T cells and found 5 clusters of CD4+CXCL13+ T cells as identified by their typical markers (Fig. 2g–i). We found that the percentage of cluster 2 in PTs was much higher than that in LNMTs (Fig. 2j). Cluster 2 expressed a higher level of interferon-gamma (IFN-γ) than did the other CD4+CXCL13+ clusters, indicating that this type of CD4 T cells has a potential role in killing tumor cells (Fig. 2i). Intriguingly, we found that BHLHE40 was highly expressed in cluster 2 (Fig. 2i), which is consistent with a previous study that reported BHLHE40+CD4+ T cells to have the ability to suppress colon cancer cells, supporting the speculation of the tumor-suppressing function of cluster 2, which need further functional validation. In contrast, LNMTs showed more cluster 1 cells with a high expression of GPR183, which is reportedly expressed in naive CD4 and CD8 T cells (Fig. 2i). Using a diffusion map to infer the CD4+CXCL13+ T cell trajectory, we found that the initial development stage of cluster 1 could differentiate into 3 branches: cluster 2, cluster 3, and cluster 4 (Fig. 2g). We found that cluster 1 in LNMT mainly differentiated into exhausted cluster 3, whereas in PT, it mainly differentiated into tumor-suppressing cluster 2 (Fig. 2k). Differential gene analysis between cluster 2 and cluster 3 confirmed that cluster 2 expressed high tumor-suppressing genes (e.g., GZMA) in tumor (Fig. 2l). The ratio of cluster 3 to cluster 2 could also be used to predict poor prognosis in TCGA-BRCA datasets (Fig. 2m). This finding suggested that CD4 T cells in LNMTs are less mature than those in PTs and CD4+CXCL13+ T cells and are more likely to be reprogrammed into an exhausted state in LNMTs. Altogether, these data demonstrated that the antitumor cytotoxicity of T cells in LNMTs was reduced compared to that in PTs. T clonal expansion and transition are manifestations of immune response and immune activation. Consequently, we analyzed T clonal expansion and transition in CD4 and CD8 T cells with single-cell resolution TCR sequencing. 59,327 immune T cells were captured and 47,803 of them were performed TCR analysis. We found that patient 2 and patient 4 had very rare immune cells, which were removed from the downstream analysis (Supplementary Fig. 3a). Of all the T cells, approximately 18.4% showed clonal expansion (5.01% with 2 cells per clone, and 13.39% with 3 cells per clone; Supplementary Fig. 3b). We found that the proportion of T cells in the PTs with clonal expansion was higher than that in LNMTs (Fig. 3a). We then employed STARTRAC, a method based on Shannon entropy, to quantify the expansion and transition ability of T cell clusters. The transition and expansion ability of CD8 T cells was more powerful than that of CD4 T cells (Supplementary Fig. 3c, d). We found that CD4 T cells in PTs, such as CD4-C3-YPEL5, had stronger transition ability than did those in LNMTs (Fig. 3b). Then, we used STARTRAC-expansion to measure the expansion of each T cell cluster. CD4-C6-CXCL13 had the greatest expansion ability among CD4 T cells, followed by CD4-C5-FOXP3 (Supplementary Fig. 3c). these both clusters showed a greater expansion ability in PTs than in LNMTs (Fig. 3c). CD8-C2-CCL5 and CD8-C3-GZMK demonstrated a significantly more powerful expansion ability in PTs than in LNMTs (Fig. 3c). To investigate the relationship between T cell expansion and T cell activation based on TCR sequencing analysis, we fitted the line of CD4 regulatory T cells (Tregs) score and CD8 activation score with cell expansion in each cluster in different tissues separately to determine the correlation between T cell activation and TCR expansion. T cells showed a weaker expansion ability in LNMTs than in PTs, even when at a similar developmental stage (Fig. 3d, e). Overall, TCR sequencing analysis of CD4 and CD8 T cells further proved that T cells displayed a lower activity of expansion and transition in LNMTs than in PTs. To further demonstrate that T cells are suppressed in LNMTs, we compared the activity of matched T (MT) cells in LNMTs with those in PTs. T cells with identical TCRs located in 2 different tissues were considered to be MT cells, originating from the same progenitor and having a similar development time (Fig. 4a). The results showed that the microenvironment of PTs had a higher percentage of MT cells (25%) than did the LNMTs (5.8%) (Fig. 4b). Matched T cells only occupy lower than 25% in total T cells (Fig. 4b) and may transit between diverse states. To avoid missing the differences caused by different states of matched T, we included all matched CD8 T in DEG analysis and separated matched CD4 T cells into conventional CD4 T (Tconvs) and Tregs for differential gene analysis. The results showed that 128 genes were significantly down-regulated and 63 genes were up-regulated in matched CD8 T cells of LNMT compared with PT (Fig. 4c). These significant genes were then thrown into pathway enrichment analysis to uncover the underlying biological functions. We found that the down-regulated genes of expanded CD8 T cells in LNMT were enriched in T cell activation, which also reflects a lower T cell activity in LNMT than in PT (Fig. 4d & Supplementary Fig. 4a). For CD4 T cells, 460 genes were significantly down-regulated and 21 genes were up-regulated in matched Tconvs of LNMT compared with PT (Fig. 4e). Pathway enrichment analysis revealed that Tconvs in PT are in enriched in the T cell activation pathway and positive regulation of cytokine production (Fig. 4f & Supplementary Fig. 4b). As for Tregs, only 27 genes were down-regulated and 16 genes were up-regulated in the matched Tregs of LNMT compared with PT (Supplementary Fig. 4c). In summary, these data further supported that T cells are less activated in LNMT compared with PT. Myeloid cells play an important role in the TME. We identified 11 clusters of myeloid cells including 4 dendritic cells (DC) subsets (pDC-LILRA4, DC-C1-CD1C, DC-C2-CLEC9A, and DC-C3-LAMP3), 6 macrophage subsets (Macro-C1-APOC1, Macro-C2-SLC40A1, Macro-C3-VCAN, Macro-C4-CXCL11, Macro-C5-SPP1, and Macro-C6-CCL3) and mast cells (Fig. 5a, Supplementary Fig. 5a). The enriched genes of each cell type are shown in Fig. 5b. DC-C1-CD1C and DC-C2-CLEC9A; the high expression of DC1C/CLEC10A and XCR1/CLEC9A correspond to cDC1 and cDC2, respectively, and have been well-characterized. DC-C3-LMAP3, which was characterized in several recent studies, showed a high expression of CCR7, chemokines, and costimulatory genes, and represented activated DCs (Fig. 5b). An embedding plot revealed differences in DCs across the 2 microenvironments (Supplementary Fig. 5b), with pDC-LILRA4, DC-C1-CD1C, and DC-C3-LAMP3 showing a higher preference for LNMTs than for PTs (Fig. 5c). Differential gene analysis showed that DCs in PTs were more enriched in the interferon-stimulated gene (ISG15) and had higher levels of STAT1 expression (Fig. 5d, Supplementary Fig. 5c), a key transcription for DC differentiation and maturation. Moreover, DC-C3-LAMP3 in PTs had a higher MHC II gene expression than that in LNMTs (Fig. 5e, f). DC-C3-LAMP3 in PTs also showed higher enrichment in the glycolysis pathway (Supplementary Fig. 5d), which contributes to DC activation. Because activated DCs play important roles in communicating with other immune cells, we then undertook to determine whether DC-C3-LAMP3 showed a different type of communication with immune cells depending on whether it was in PTs or LNMTs. LAMP3+ DCs highly expressed CXCL9 and CCL19 in PTs and highly expressed CCL17 and CCL22 in LNMTs (Fig. 5g). Cell–cell interaction analysis showed that LAMP3+ DCs in PTs exhibited greater interaction with immune cells through CXCL9:CXCR3, CCL19:CXCR3, or CCL19:CCR7 than in LNMTs. However, LAMP3+ DCs in LNMTs showed a stronger interaction with Tregs through CCL17:CCR4 and CCL22:CCR4 (Fig. 5h), suggesting that LAMP3+ DCs in LNMTs may be more likely to recruit and activate Tregs to enhance immunosuppression. These data suggested that DCs, especially LAMP3+ DCs in PTs, were more mature and had a greater T cell priming and activating ability than those in LNMTs. Subsets with a high expression of CD68 were defined as macrophages. Tumor-associated macrophages (TAMs) are characterized by enrichment in the TME. We subsequently combined the data of a published study and that of our single-cell analysis to clarify the properties of the tissue-resident macrophage subsets (Supplementary Fig. 5e). Consistent with the previous study, we found that the macrophages in the TME showed a greater degree of diversity and complexity. Macro-C4-CXCL11 and Macro-C5-SPP1 were significantly highly enriched in the TME and corresponded to TAMs. Macro-C1-APOC1, Macro-C2-SLC40A1, and Macro-C6-CCL3 were defined as tissue-resident macrophages, the levels of which in tumors were comparable to those in adjacent tissues (Fig. 5i). It is well known that M1 and M2 signatures represent the different functions of macrophages in the tumor. In this study, we found there is no difference in the ratio of these 2 types of macrophages in PTs compared to LNMTs (Supplementary Fig. 5f, g). However, using differential gene analysis, we found that macrophages have a higher expression of IFI27, IFITM1, and IFI44L in PTs than in LNMTs (Fig. 5j), suggesting activation of IFN-γ signaling in PTs. Furthermore, CCL8, which has been reported to promote breast metastasis and progression, was more highly expressed in PTs than in LNMTs, indicating distinct microenvironments in PT and LNMT that could influence tumor cell survival. In our dataset, we identified 3 types of stromal cells: TECs, PVLs, and CAFs (Fig. 6a, Supplementary Fig. 6a). PVLs included 2 subtypes, PVL-C1-MGS5 and PVL-C2-MYH11, both of which exhibited a high expression of actin cytoskeleton gene ACTA2 and GTPase-activating protein RGS5 (Supplementary Fig. 6b). CAFs play a pivotal role in the cancer microenvironment and may suppress the immune response and/or promote tumor metastasis. We found there is no difference in the proportion of total CAFs in PTs versus LNMTs (Fig. 1d). According to their differentially coexpressed markers, fibroblasts were further classified into 3 CAF types: myofibroblast-like phenotype (mCAF) including CAF-C1-POSTN; and 2 inflammatory property fibroblasts (iCAF), including CAF-C2-APOD and CAF-C3-PLA2G2A (Fig. 6a). CAF-C1-POSTN expressed high levels of POSTN and collagen genes (COL1A2, COL3A1), which have been reported to contribute to extracellular matrix remodeling and stiffness of tumor (Supplementary Fig. 6b). CAF-C3-PLA2G2A, which has been reported in other studies, was characterized by a high expression of OGN, which encodes a small leucine-rich proteoglycan protein that functions in T-cell recruitment and immune infiltration in tumors. Importantly, we found CAF-C3-PLA2G2A at a high frequency in HER2+ tumors, while CAF-C2-APOD was enriched in luminal tumors (Fig. 6b, c). Compared with those in luminal breast cancer, HER2+ tumors have a high degree of immune infiltration, and patients with this subtype can benefit from immunotherapy. We thus aimed to determine whether CAF-C3-PLA2G2A enrichment in HER2+ tumors had a connection with immune infiltration. Using cell–cell interaction analysis, we found that, compared with the other 2 types of fibroblasts, CAF-C3-PLA2G2A showed a much stronger interaction with immune cells, including CD4 and CD8 T cells, DCs, and macrophages (Fig. 6d). Mechanistically, PLA2G2A: α4β1 integrin, OGN: HLA-DRB1, VSIR:CCL4L2, FN1:α4β1 integrin, and DPP4:CCL3L1 interaction complexes were found to be responsible for the association of PLA2GA+ CAF with immune cells. HLA-DRB1, α4β1 integrin, CCL4L2, and CCL3L1 were reported to be expressed in immune cells. In our study, we found that CAF-C3-PLA2G2A also exhibited an abundant expression of OGN, VSIR, FN1, and DPP4, which can interact with immune cells. In contrast, the gene expression of OGN, VSIR, FN1, and DPP4 was much lower in CAF-C2-APOD (Fig. 6d, e). To further validate the function of PLA2G2A+ CAFs in immune cells, we treated monocyte THP1 cells with PLA2G2A protein and found that PLA2G2A could promote the migration of THP1 cells (Fig. 6f). These data suggested the potential of PLA2GA+ CAFs to attract immune cells. Furthermore, we found that PLA2G2A, which is exclusively expressed in PLA2G2A+ CAFs (Supplementary Fig. 6c), was highly expressed in HER2+ patients and highly correlated with the immune cell marker CD3E in the TCGA-BRCA dataset (Fig. 6g). PLA2G2A was also positively correlated with the CD45 antigen PTPRC, B-cell marker CD79A, and CD8 T cell marker CD8A in breast cancer patients (Supplementary Fig. 6d). Immunohistochemical (IHC) staining in human breast cancer tissues further revealed that PLA2G2A+ CAFs were more abundant in HER2+ tumors than in luminal tumors (Fig. 6h, i). Immunofluorescence analysis also showed that PLA2G2A+ CAFs had a similar spatial distribution to macrophages and CD8 T cells (Fig. 6j), and confirmed the interactions of PLA2G2A+ CAFs with macrophages and CD8 T cells at the single-cell level (Fig. 6k). In short, we identified 3 CAF subsets in breast cancer patients and demonstrated enrichment of PLA2G2A+ CAFs in HER2+ tumors, and these may be the main microenvironmental factors that determine the immune infiltration in breast cancer. Finally, we wanted to characterize the features of malignant epithelial cells. Epithelial cells were divided into malignant epithelial cells and nonmalignant epithelial cells according to their CNVs. Malignant cells and nonmalignant cells were separated according to the distribution of epithelial cell malignancy scores (Supplementary Fig. 7a, b). Heatmaps of CNVs exhibited a considerable degree of divergence among the patients, and similar patterns were observed in the same patients, even in different tissues. Malignant cells from the same patient showed a similar CNV pattern, suggesting that the malignant cells were derived from the same point of origin (Fig. 7a, b). To further understand the characteristics of malignant cells of lymph node metastasis, we compared transcriptome signatures of the malignant cells between LNMT and PT for patients that have at least 20 cells in each tissue. Thus, 3 patients were excluded and the transcriptome signature of the other 5 patients was calculated (Supplementary data 4). Few significant genes are shared across patients (Supplementary Fig. 7c, d). Interestingly, we found antigen presentation genes, such as CD74, HLA-DRA and B2M, are mostly down-regulated in LNMT compared with PT for patient 8 (Fig. 7c). Similarly, we also found that HLA-B and HLA-C are down-regulated in LNMT of patient 5 (Supplementary Fig. 7e). To character whether these findings are prevalent in most patients, transcriptome signatures from the 5 patients were used for GSEA enrichment analysis (Supplementary data 5). Significant pathways shared by the 5 patients were analyzed and the normalized enrichment score (NES) for each pathway was averaged. The pathways were then ranked according to the numbers of shared patients and the mean of NES. From the top 10 enriched pathways, 4 pathways are related to antigen presentation and these pathways are enriched in 4 of 5 patients (Fig. 7d, e, Supplementary data 6). It is interesting to ask whether the down-regulated antigen presentation pathway is related to CNV clones of malignant cells. Then we clustered malignant cells according to their CNV similarity for each patient and then compared the DEG of each CNV clone in PT vs LNMT (Fig. 7f). We found that malignant cells belonging to different CNV clusters in patient 8 have no difference in metastasis ability (Fig. 7g). And the malignant cells in different CNV clusters of patient 8 are also enriched in the antigen presentation pathway (Supplementary Fig. 7f, g). Antigen presentation genes can be mainly divided into MHC I and MHC II class molecules. We compared the two types of antigen presentation in PT vs LNMT across the different CNV clusters and found that both MHC I and MHC II class molecules are down-regulated in LNMT across different CNV clusters in patient 8 (Fig. 7h). MHC I genes but not MHC II genes are down-regulated in LNMT of Patient 5 (Supplementary Fig. 7h–j). These findings indicate that malignant cells migrated to LNMT may render lower antigen presentation genes, resulting in an immune evasive mechanism, which provide insight into the characterization of malignant cell metastasis in breast cancer. Single-cell sequencing is a powerful tool for analyzing the TME. Several breast cancer studies have used single-cell sequencing to analyze the differences of immune cells among breast cancer in situ, adjacent cancer, blood, and normal LNs. However, an analysis of the metastatic LN microenvironment in breast cancer at the single-cell level has not yet been reported. We thus systematically revealed the characteristics and differences in the microenvironment between LNMTs and paired PTs in breast cancer. The TME is shaped into an environment with characteristics, such as immunosuppression, that are more conducive to tumor growth. For example, we found that the TME induced the transformation of CXCL11+ macrophages with the manifestation of M1-like macrophages into SPP1+ macrophages with a higher M2 signature, promoting cancer progression. However, compared with the TME in situ, the microenvironment of lymphatic metastasis exhibited stronger immunosuppression. Lymphatic vessels play an important role in tumor immunity, aiding the antigen presentation of DCs and the activation of T cells in the TME. The particularity of lymphatic vessel structure and the high expression of CXCR4 in breast tissue cells may be an important reason for metastasis to and colonization of LNs. Our study investigated this mechanism by performing a single-cell analysis and by characterizing the microenvironment of LNMTs. We found that the overall activity of T cells in metastatic LNs was suppressed more than that in the PTs. This conclusion is supported by several lines of evidence: (1) from the transcriptional level, CD8 T cells showed weaker activity and cytotoxic effect in the LNMT microenvironment compared with those in PTs. Although it is well known that CD8 T cells in PTs have higher activation than those in LNMTs, we identified the phenotype and characteristics of CD8 T cells within the two tissues in detail, such as the subpopulation and development path of CD8 T cells in PT vs LNMT, which can be well characterized in single-cell sequencing. (2) CD4+CXCL13+ cells were reprogramed to an exhausted state and expressed lower levels of IFNG, CXCL13, and genes that inhibit tumor growth in the LNMT microenvironment. However, the function of the 5 clusters of CD4+CXCL13+ cells needs to be further verified by experiments. (3) The frequency of CD8+CXCL13+ was remarkably decreased in LNMT. And (4) compared to those in PTs, MT cells in LNMTs exhibited functional inhibition. Collectively, these findings indicate that immune cells in metastatic LNs are reprogrammed in several ways. Lymphatic vessels express several genes that inhibit T cell activity, such as programmed death-ligand 1 (PD-L1), and may mediate immune tolerance, a fact which may be the key to explaining our findings. Furthermore, very interestingly, we found that malignant cells migrated to LNMT downregulate their antigen presentation pathway, which results in the defect of presenting tumor-related epitope to adaptive immune cells. Loss of antigen presentation has been studied in various tumors and was the common way for cancer cells to escape from immune surveillance. Taken together, these findings may help us understand why the tumor cells are easy to survive in LN. Another interesting finding in this study is that activated LAMP3+ DCs showed higher enrichment in LNMTs than in PTs and strongly interacted with Tregs through CCL17 or CCL22, and this may partly account for the suppressed activity of T cells in LNMTs. A recent study revealed that LAMP3+ DCs in pancreatic adenocarcinoma might promote immune tolerance through interacting with tumor-infiltrating Tregs, supporting our conclusions. Given the higher proportion of naïve state T cells in LNMTs compared to in PTs, we speculate that immunotherapy drugs may have very limited antitumor effects on the microenvironment of LNMTs, and this may have ramifications for the clinical treatment of breast cancer with LNMTs. Immune infiltration in tumors is currently an area of research focus, and findings in this area may have important clinical application. Immune infiltration is related to different subtypes of breast cancer. Compared with patients of other subtypes, patients with triple-negative and HER2+ breast cancer exhibit stronger immune infiltration and can benefit from immunotherapy. Although clinical and pathological characteristics have been found to be related to immune infiltration, the current understanding of immune cell infiltration is extremely limited, and the mechanisms underlying the varying degrees of immune infiltration in patients with different subtypes are still unclear. CAFs are important components in the TME by their regulation of immune cell activity and promotion of tumor progression and metastasis. In this study, we found a type of PLA2G2A+ CAFs that was enriched in HER2+ breast cancer and showed high expression levels of genes that can interact with immune cells. PLA2G2A was reported to promote the proliferation of monocytic cells through interacting with αvβ3 and α4β1 integrin. A previous report also found that PLA2G2A is overexpressed in some fibroblast subtypes. We verified in the TCGA-BRCA database that there was a significantly positive correlation between PLA2G2A+ CAF markers and immune cell markers. Furthermore, in our single-cell dataset, PLA2G2A+ CAFs were observed to interact with various immune cells, especially macrophages, which suggests that PLA2G2A+ CAFs may be a factor in tumor immune infiltration, promoting tumorigenesis and tumor development. Another interesting observation is that PLA2G2A+ CAFs were present in both the PT and the LNMT of HER2+ patients; however, they were not present in the LNMT of patients with luminal breast cancer. Therefore, PLA2G2A+ CAFs may be recruited and dominated by HER2+ tumor cells to migrate or differentiate in the process of tumor development. In summary, we demonstrated that the immune cells in LNMTs are less active than those in PTs and clarified the mechanism underlying this difference. Further mechanistic investigations to determine precisely why tumor cells can more easily colonize and proliferate in LNs are thus warranted. Our study identified the unique features of the LNMT and PT microenvironments, the knowledge of which can aid in developing individualized therapy that targets these microenvironments in patients with breast cancer. The study was conducted in accordance with the Declaration of Harbin Medical University complying with all relevant ethical regulations. Written informed consent was obtained from all participants. No compensation was provided for the study participants. Eight treatment-naïve female patients with a pathological diagnosis of invasive ductal carcinoma of the breast with LNMT were enrolled at Harbin Medical Hospital. Their ages ranged from 47 to 66 years, with the median age being 56 years. PT and paired LNMT tissues were surgically resected from each patient. Among the patients, 5 cases were luminal subtypes and 3 cases were the HER2-overexpressing subtype. This study was approved by the Research and Ethical Committee of Harbin Medical University Cancer Hospital (IRB:KY2019-08). Written informed consent was obtained from all participants in the study. PT or paired LNMT tissues were separated and digested into single-cell suspensions. The 10x Genomics Cell Preparation Guide describes best practices and general protocols for washing, counting, and concentrating cells from both abundant (>100,000 total cells) and limited cell suspensions (<100,000 total cells) in preparation for use in 10x Genomics Single Cell Protocols. Formalin-fixed, paraffin-embedded (FFPE) samples passing the RNA quality control were used to prepare for spatial transcriptomic construction and sequencing. The Visium Spatial Gene Expression Slide & Reagent Kit (10x Genomics) was used to construct sequencing libraries according to the Visium Spatial Gene Expression User Guide (CG000239, 10x Genomics). The cell suspension was loaded into Chromium microfluidic chips with 5′ v.1.1 chemistry and barcoded with a 10× Chromium Controller (10x Genomics). RNA from the barcoded cells was subsequently reverse-transcribed, and sequencing libraries were constructed with reagents from a Chromium Single Cell 5′ v1.1 Reagent Kit (10x Genomics) according to the manufacturer’s instructions. Sequencing was performed with an Illumina NovaSeq 6000 PE150 system, depending on the experiment and following the manufacturer’s instructions. Spatial transcriptomic sequencing was performed with a NovaSeq PE150 platform according to the manufacturer’s instructions (Illumina) at an average depth of 300 million read-pairs per sample. Fresh tissues obtained from the patients were embedded in paraffin. The paraffin-embedded tissues were cut into 5-um–thick sections on a glass slide. The sections were infiltrated with fresh xylene 3 times for 10 min each time before being soaked in 100% ethanol, 95% ethanol, and 75% ethanol, in that order, once for 5 min. The sections were then soaked in sterile water 3 times, for 1 min each time, to remove the paraffin. The deparaffinized slides were exposed to antigen in a 100 °C water bath with antigen retrieval solution for 20 min. Next, the slides were blocked with 10% goat serum for 10 minutes. After the removal of the blocking solution, the slides were incubated with the first primary antibody at room temperature for 1 h and then with the secondary antibody at room temperature for 10 min. Finally, the slides were incubated with fluorescent staining amplification solution (Absin multicolor immunohistochemistry kit) for 10 min at room temperature. The second and third primary antibodies were stained following the same steps as those for the first antibody, including antigen retrieval, and incubation with primary antibody, secondary antibody, and fluorescent amplification solution. Primary antibodies used for multicolor immunohistochemistry were rabbit anti-human CD8A (Abcam, catalog: ab17147, clone id: C8/144B, 1:200 dilution), rabbit anti-human CD68 (Abcam, catalog: ab213363, clone id: EPR20545, 1:200 dilution), and anti-human PLA2G2A (Invitrogen, catalog: PA-102403, 1:200 dilution). Droplet-based sequencing data were aligned and quantified with the Cell Ranger Software Suite (version 3.1.0, 10x Genomics) using the GRCh38 human reference genome (official Cell Ranger reference, version 3.0.0). To obtain high-quality cells, every sample underwent filtering as follows: (1) cells were filtered if the number of detected genes (log10 scale) was below the medians of all cells minus 3 × the median absolute deviation; (2) cells were filtered if the proportion of mitochondrial genes was higher than the median of all cells plus 3 × the median absolute deviation; and (3) cells were filtered if their unique molecular identifier (UMI) counts were lower than 300. For spatial transcriptomic sequencing, FASTQ files and histology images were processed by Space Ranger (version spaceranger-1.2.0, 10x Genomics) software with default parameters. The filtered gene-spots matrix and the fiducial-aligned low-resolution image were used for down-streaming data analyses (Seurat). To remove doublets in single-cell RNA sequencing data, cell doublets were identified using the Scrublet package. Briefly, for each sample, the cell count matrix was fed to Scrublet. Then, the “scrub_doublets” function was applied to simulate doublets, the doublet scores were calculated, and doublet calling was performed. To further reduce the false-negative rate in the Scrublet analysis, we over clustered the remaining cells and calculated the average doublet score within each cluster. We removed any clusters that had an average doublet score of more than 0.6 or more than 1 known cell marker (i.e., CD3D for T cells and CD79A for B cells). We observed ambient messenger RNA (mRNA) effects, which are ubiquitous in droplet-based single-cell RNA-sequencing (RNA-seq) experiments. The possible reason for this is that free mRNA released from dying cells and single cells were simultaneously captured by beads in droplets. SCGB2A2 is expressed almost exclusively in the normal breast epithelium and human breast cancer cells; however, in our dataset, we found other cell types. such as immune cells and CAFs, also expressed SCGB2A2. We stained SCGB2A2 and immune cell markers to confirm the absence of these markers. We then used SoupX software to remove or reduce the ambient mRNA effects. Briefly, we regrouped all cells (“sc.tl.leiden” function from the scanpy package; resolution = 0.8) to obtain a rough cluster classification. Raw count matrices with defined clusters were fed into the “SoupChannel” and “setClusters” functions. To estimate the contamination fraction of the dataset, we manually defined the 3 gene sets with the strongest ambient effect: immunoglobulin (IG) genes, human leukocyte antigen (HLA) genes, and breast epithelium genes (SCGB2A2 and KRT19). These gene lists were fed into the “estimateNonExpressingCells” function and the “calculateContaminationFraction” function to calculate the contamination fraction for each cell. Finally, the original count matrix was automatically corrected using the “adjustCounts” function, and the adjusted matrix was further used for downstream biological analysis. To integrate and visualize data, we used the Scanpy python toolkit (version 1.4.1) to analyze our single-cell dataset. Briefly, we performed dimension reduction steps including normalization, logarithmic transformation, highly variable gene calling, data scaling, and PCA calling. To remove the batch effect, data were processed using batch-balanced k-nearest neighbors (BBKNN). BBKNN modifies the neighborhood construction step to produce a graph that is balanced across all batches of the data. This approach treats the neighbor network as the primary representation of the data. For each cell, the BBKNN graph is constructed by finding the k-nearest neighbors for each cell in each user-defined batch independently. As for our data, we implemented BBKNN from the scanpy package by using the “bbknn function”, and the parameters were set as follows: batch_key = “patient”, n_pcs = 35. A batch-corrected neighbor graph was used to find clusters using the Leiden community detection algorithm. To reasonably cluster cells and find their biological markers, the following steps were performed. First, we changed the Leiden resolution parameter from 0.6 to 2 by 0.2 to obtain a collection of cell classifications. Then, we compared the UMAP embedding plot colored by canonical markers (PTPRC, CD3D, CD8A, CD4, CD79A, LYZ, PLVAP, ACTA2, and KRT19) with the UMAP embedding plot colored by the clusters output by different Leiden resolutions to find the smallest suitable Leiden resolution to distinguish canonical markers. Clusters with the same canonical markers were merged. In the first round of clustering, we identified 6 major cell types including immune T and natural killer (NK) cells, epithelial cells, B cells, myeloid cells, CAFs, and tumor endothelial cells (TECs). We observed that immune cell types, including CD4, CD8, and NK cells could not be distinguished well using the Leiden-based classification in the first-round clustering. The reasons for this include the cell types detected by 10x Genomics having limited features and similar transcriptome profiles, and dimension reduction preserving the difference between major cell types but losing some information about major cell types. In the second round of clustering, immune T and NK cells, epithelial cells, myeloid cells, CAFs, and TECs labeled in the first round were further divided into subsets and reclustered into more detailed subclusters. To identify the specific markers of each cluster, differentially expressed genes were identified using the “FindAllMarkers” function of the Seurat R package (v. 3.1.5). Clusters were annotated based on the expression of known marker genes or the most highly expressed genes (Supplementary data 1). The third round of clustering was performed on the annotated clusters in which we were interested, including the group of FOXP3+ expressed CD8A+ markers that were distinguished when we reanalyzed CD4-C5-FOXP3 and re-clustered CD4-C6-CXCL13 to find heterogeneity and continuous biological states. The second and third rounds of clustering were performed following the same steps as those in the first round, starting from the adjusted count matrix and including normalization, logarithmic transformation, variable gene calling, scaling, PCA calculation, and batch correlation. To compare the relative preference of each cell type in different classifications (e.g., PT vs. LNMT and HER2+ vs. luminal), we made a double table of the number of cells according to the corresponding classification. To reduce the sample size effects, we calculated sample size scaled proportion as follows: number of cells of a specific type in a category/total number of cells in the category. The category-normalized numbers were used to calculate the proportions of categories in a specific cell type. To calculate the cellular composition in a specific patient, we defined broad cell types (e.g., B, CD4, CD8, DC, macrophage, CAF, and epithelial cell). We calculated the number of cells of a specific type as defined above and divided this number by the total numbers of cells from a specific patient. As the tumor cell suspensions were thoroughly mixed and captured without bias, this ratio reflected the natural cellular composition within the tumor. The Seurat package was used to perform gene expression normalization, dimensionality reduction, spot clustering, and differential expression analysis. To integrate the data, single-cell RNA-seq transcriptome and spatial transcriptome were preprocessed by the “SCTransform” function and PCA analysis. Then “FindTransferAnchors” and “TransferData” functions were used to measure each cluster score for each spot. We downloaded count matrix files (GSE114727) from the Gene Expression Omnibus (GEO) database and mapped labels and embeddings from reference data to the new datasets. In brief, raw count matrix data from the new dataset were imported into scanpy and subjected to preprocessing, including filtering, normalization, and logarithmic transformation. We then used the “ingest” function in scanpy to integrate embeddings and annotate the new datasets through projection onto a PCA that was fit our reference data. The mapping labels from the new datasets were output for enrichment analysis. We inferred the CNV of epithelial cells using transcriptomic profiles, to determine the malignant epithelial cells. CNV was estimated based on 2 major steps: initial CNV (CNVi) calculation and final CNV (CNVf) estimation. Genes were first sorted according to their genomic location at each chromosome, and then the CNVi was derived by applying a sliding window of 100 genes to calculate the average relative expression values within each chromosome. In this way, gene-specific patterns could be eliminated, and the derived profiles (i.e., moving average) largely reflected the CNV. We also restricted the relative expression values to [–3, 3] (values beyond the bounds were replaced with bound values) to avoid any genes with extreme expression influencing the moving average. We defined the CNV score of every single cell as the sum of the square of the CNVf across all windows. The malignancy score of each single-cell was defined as the mean of the square of the CNVf minus 1 across all windows. The smooth distribution curve of the malignancy score was fit using bimodal methods to estimate the threshold for malignancy. This function was provided by the “getBimodalThres” function in the scCancer package. Cells with a malignancy score that exceeded the malignant threshold were determined to be malignant epithelial cells. Cell-cell interactions were analyzed using the cellphoneDB python package. To reduce the computational burden and represent different cell types, we downsampled the dataset by randomly sampling 1000 cells of each cell type. The strength of interactions was computed based on the expression of a receptor by 1 cell type and a ligand by another cell type. Only receptors and ligands expressed by more than 30% of the cells in a specific cluster were considered. The cluster labels of all cells were randomly permuted 1000 times to calculate the P value for the likelihood of paired interactions. Only paired interactions with a P value of less than 0.05 were considered. To obtain different paired interactions between different types of CAFs and other cell types, we calculated the differentially expressed genes using the “FindAllMarkers” function. Only genes meeting the criteria of LogFC threshold > 0.5 and min.pct > 0.25 were used to filter paired interactions. We implemented a diffusion map on cell types that had the same lineage, such as CD8 T cells. A diffusion map function was implemented in scanpy packages. The BBKNN batch-corrected expression matrix was used for calculating the neighbor matrix with the following parameters: n_neighbors = 10, n_pcs = 15, and method = “gauss”. The neighbor matrix was then used to calculate the diffusion map. We found that changing the number of neighbors did not impact the relative position of cell types in the diffusion map. We constructed the single-cell trajectory of epithelial tumors by using a reversed graph embedding method implemented in the R Monocle 2 package (v. 2.6.3). To increase the efficiency of the operation, we randomly selected 500 cells annotated as epithelial cells from each sample; for sample sizes of fewer than 500 cells, all the cells were taken. We integrated the expression matrix of epithelial cells using BBKNN, to neutralize patient-specific effects, including different patients and disease subtypes. After this, cell clusters were determined by a Leiden function with a resolution of 0.8 in batched-removed epithelial cells. We compared each cell cluster with other clusters using the “FindAllMarker” function in the Seurat package to determine the batch-removed differential expressed genes, and the top 20 differential expressed genes per cluster were used to order cells in Monocle to construct the epithelial cell DDRtree trajectory plot. The signature of each state was calculated based on differentially expressed genes over the other states. To compare breast cancer epithelial cells with normal epithelial cells, we first obtained signature genes of normal epithelial developmental states from normal epithelial datasets. Then the average of the signatures was calculated for breast cancer epithelial cells. Alignment and quantification of 10× VDJ sequencing data were performed with the Cell Ranger software using the GRCh38 human VDJ reference genome (official Cell Ranger reference, v. 3.1.0). VDJ sequencing information was extracted from the output file using the Scripy python package (v. 0.3). We used the “chain_pairing” function to summarize TCR compositions. Cells with the same α or β chains were defined as clonotypes. According to their unique α and β chains, TCR chains can be classified into 7 types: single pair, orphan beta, orphan alpha, extra alpha, extra beta, 2 full chains, and multichain. Only the single pair type was used in the downstream analysis. To integrate transcriptome data, we included cells annotated as CD4 or CD8 T cells to visualize our TCR data on embeddings. We used STARTRAC packages to analyze the behavior of T cells. To obtain good quality data, we excluded patient 2 and patient 4, who had low T-cell capture rates, from the downstream analysis. STARTAC-expansion and STARTAC-transition were used to separately analyze T cells in PTs and LNMTs. STARTRAC-expansion is usually used in the standard TCR clonality measurement but was specifically applied to different T cell clusters in our analyses. Normalized Shannon entropy was used to calculate the evenness of the TCR repertoire of the given T cell cluster and then define the STARTRAC-expansion index as 1-evenness. STARTRAC-expansion ranged from 0 to 1, with 0 representing no clonal expansion for each clonotype and 1 representing a cluster composed of only 1 clonally expanded clonotype. A high STARTRAC-expansion indicated high clonality. STARTRAC-transition was used to quantify the extent of state transition of each clonotype within a given cluster. The STARTRAC-transition index at the cluster level was defined as the weighted average of all TCR clonotype state transition indices contained in the cluster. THP-1 (human acute monocytic leukemia cell line, 1101HUM-PUMC000057) was purchased from Cell Resource Center of Peking Union Medical College (Beijing, China). THP-1 was authenticated using short tandem repeat analysis. No mycoplasma contamination was detected. THP-1 cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum, penicillin (100 μg/ml) and streptomycin (100 μg/ml). Cell lines were incubated in a humidified atmosphere of 5% CO2 at 37 °C. 2 × 106 THP1 cells were added into the top chamber of 24-well tissue culture inserts (Costar). PLA2G2A (0.5 μg/ml, 11187-H08H, Sino Biological) was respectively applied to the top chamber, in RPMI-1640, and the bottom chamber, in RPMI-1640 containing 20% serum. After incubation at 37 °C in 5% CO2 for 4 h, cells in the bottom chamber were collected and counted. The Cancer Genome Atlas Breast Invasive Carcinoma (TCGA-BRCA) datasets were used to analyze the correlations of selected genes with patient survival. Gene expression as well as clinical and survival data were downloaded from UCSC Xena (http://xena.ucsc.edu/). The signature scores of the TCGA-BRCA patients were calculated as the mean expression of genes in the signature. The signature scores were grouped into high and low expression groups by the 55th and 45th quantile values, respectively. We used the survival packages to calculate the impact of genes or signatures on patient survival and plotted Kaplan–Meier survival curves using ggsurvplot in R. For correlation analysis, gene or signature scores were calculated by applying Spearman’s correlation coefficient using the cor function in R. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Reporting Summary

PMC9649682

Dharmar Manimaran,Namasivayam Elangovan,Panagal Mani,Kumaran Subramanian,Daoud Ali,Saud Alarifi,Chella Perumal Palanisamy,Hongxia Zhang,Kowsalya Rangasamy,Vasan Palanisamy,Renuka Mani,Kavitha Govarthanan,Wilson Aruni,Rajeshkumar Shanmugam,Guru Prasad Srinivasan,Aruncahllam Kalirajan

Isolongifolene-loaded chitosan nanoparticles synthesis and characterization for cancer treatment

10-11-2022

Drug delivery,Pharmacology

Recent breakthroughs in the field of nanoparticle-based therapeutic delivery methods have changed the standpoint of cancer therapy by effectively delaying the process of disease development. Nanoparticles have a unique capacity of good penetrating ability than other therapeutic leads used in traditional therapeutics, and also, they have the highest impact on disease management. In the current study isolongifolene-loaded Chitosan nanoparticles have been formulated, synthesized and then characterized by the use of Fourier Transform Infrared Spectroscopy, X-ray Diffraction, Scanning Electron Microscopy and Transmission Electron Microscopy. Further, the characterized chitosan nano formulation was evaluated for hemocompatibility, plasma stability, and in-vitro release. Isolongifolene-loaded chitosan nanoparticles were found to be compatible with plasma and also, they exhibited a constant release pattern. Hence, chitosan-loaded nanoparticles could be employed as an excellent adjuvant in cancer therapeutic, to combat the multi-drug resistance in solid tumors.

Isolongifolene-loaded chitosan nanoparticles synthesis and characterization for cancer treatment Recent breakthroughs in the field of nanoparticle-based therapeutic delivery methods have changed the standpoint of cancer therapy by effectively delaying the process of disease development. Nanoparticles have a unique capacity of good penetrating ability than other therapeutic leads used in traditional therapeutics, and also, they have the highest impact on disease management. In the current study isolongifolene-loaded Chitosan nanoparticles have been formulated, synthesized and then characterized by the use of Fourier Transform Infrared Spectroscopy, X-ray Diffraction, Scanning Electron Microscopy and Transmission Electron Microscopy. Further, the characterized chitosan nano formulation was evaluated for hemocompatibility, plasma stability, and in-vitro release. Isolongifolene-loaded chitosan nanoparticles were found to be compatible with plasma and also, they exhibited a constant release pattern. Hence, chitosan-loaded nanoparticles could be employed as an excellent adjuvant in cancer therapeutic, to combat the multi-drug resistance in solid tumors. The route of drug delivery to the targeted population is meant for effective illness management and therapy. The main problem with the current conventional way of drug delivery system is that the medicine does not reach the target as they cannot pass the micro-capillaries. Another method is the use of liposomes as potential carriers has the fascinating benefit of protecting pharmaceuticals from degradation, promoting focused action, and reducing systemic toxicity. However, several drawbacks, including limited encapsulation efficiency, quick leakage of water-soluble drugs in the presence of blood components, and poor storage stability, have led us to seek a better alternative to liposome-based administration. So, as an alternate, nanoparticles can be used for delivery processes, which are particulate dispersions or solid particles with a size range of 1–100 nm that could be used as a matrix to help transport the therapeutic lead molecule either (via) dissolved, entrapped (or) attached mechanisms. These nanoparticle-based therapeutics have the potential to revolutionize the treatment of cancer. The nanoparticles could be either synthesized as nano-spheres or nano-capsules depending on the method of preparation. Nano-capsules are systems in which the drug is incorporated within the system's core and protected by a unique polymer membrane. While the nanospheres are matrix systems that produce homogenous drug dispersions. Currently, multidimensional nanotechnology platforms in the development or clinical stages are being investigated to generate more effective and safer treatments with lesser side effects. Polymers are the most promising materials for making numerous molecular patterns and integrating them as distinct nanoparticle constructions for biological purposes, particularly cancer treatments. Intravenous injections have been reported by researchers. Biodegradable polymeric nanoparticles, particularly those coated with hydrophilic polymers such as poly ethylene glycol (PEG), are useful as drug delivery devices that target a specific organ. Polymeric nanoparticles have reportedly emerged as a viable alternative to the aforementioned liposome method, owing to their enhanced drug/protein stability and long-term drug release properties. The size of polymeric nanoparticles has a significant impact on medication retention in the bloodstream. A previous study was published that used a sterically stabilized ligand nanoparticle formulation for the targeted administration of antisense oligodeoxynucleotides and small interfering RNA into lung cancer cells. So, it is clearly evident that the carrier or the delivery system via a polymeric nanoparticle is highly achievable and target specific. Chitosan is a polymer formed from partial deacetylation of chitin that is common in crustacean and insect shells. It is made up of repeating units of glucosamine and N-acetylglucosamine, and the quantities control the degree of polymer deacetylation. At neutral pH, chitosan is insoluble, whereas it is soluble and positively charged at acidic pH. Chitosan comes in a variety of molecular weights. In comparison to their high-molecular-weight relatives, low-molecular-weight and low-deacetylating chitosans have greater solubility and faster breakdown. Chitosan has also been shown to have antimicrobial, antifungal, and wound-healing effects. A previous study using the chitosan for assessing the LD (lethal dose) showed an oral LD50 of over 16 g/kg body weight in mice proving it to be non-toxic and most significantly biodegradable. Sodium alginate is a polymer made up of alginic acid and D-mannuronic acid and L-guluronic acid residues. To create a chain, these units are joined together by − 1, 4, and − 1, 4 glycosidic bonds. Gelatin is a biopolymer with a fine biocompatibility and biodegradability making it an ideal choice for pharmaceutical and medical applications. In a previous finding, a size of 100 nm gelatin particles after delivery were found to be preferentially clustered around the leaky tumor vasculature and failed to permeate the interstitial space's strong collagen matrix. Furthermore, the MMP-2 activity destroyed the gelatin core of the 100 nm particle, allowing smaller 10 nm particles to emerge from the surface. Because of their reduced size, these MMP-2 modified particles can penetrate deep into tumors. Isolongifolene (ILF), a carbazole alkaloid isolated from the curry leaf plant—M. koenigii, an Indian herb have been proven to possess with a wide range of therapeutic characteristics. Isolongifolene has a woody and amber incense odor and is used as a fragrance in cosmetics, perfumes, space sprays, detergents, deodorants, and fabrics. Isolongifolene is a commercially accessible sesquiterpene hydrocarbon with an Isolongifolene skeleton. In an in-vitro model of Parkinson’s disease (PD), it was found that ILF has neuroprotective effects against rotenone-induced pathological symptoms such as oxidative stress, mitochondrial malfunction, and apoptosis. In a rat model of rotenone-induced PD, ILF showed improved behavioral impairments and reduced oxidative damage. As a result, ILF appears to be a potential pharmacological candidate for the treatment. Therefore, the need for a nano-sized particle delivery system of medicine is required for the successful delivery of medicinal lead into the host system. In this study, we have formulated, synthesized and validated an isolongifolene-synthesized polymeric nano-formulation for more advantages and also established its compatibility with polymers like sodium alginate, chitosan and gelatin. The efficacy, stability, and cytotoxicity of the improved Nano formulations for cancer therapy were also investigated. As a result, our research focused on isolongifolene-loaded polymeric nanoparticles, which have great adjuvant properties and could be used to treat a variety of diseases. Isolongifolene, Chitosan, and Sodium tri poly-phosphate were procured from Sigma Aldrich (USA). All other chemicals used in this study were of analytical grande. Isolongifolene-loaded Chitosan Nanoparticles (ICN) were fabricated by ionic cross-linking of chitosan with sodium tripolyphosphate (TPP) anions using a reported procedure. In brief, 2 mg/mL of chitosan was dissolved in 0.25%v/v acetic acid under constant stirring conditions at 10 °C for 12 h. About 0.75% w/v of TPP (aqueous solution) was added (2:1 ratio) into chitosan solution containing 100 μg/mL of isolongifolene (dissolved in methanol), with continuous stirring for 6 h at 4 °C. The resultant dispersion was centrifuged at 13,000 × g for 20 min at 4 °C. The supernatant was removed and ICN was carefully collected and stored at − 55 °C. The ICN was characterized using FTIR spectroscopy (Perkin Elmer, Spectrum-RX1, USA) to analyze the chemical interactions between the isolongifolene and chitosan polymer. The scanning range for FTIR wasbetween 4000 and 400 cm−1. An X-ray diffraction study was performed for both chitosan and ICN to confirm the amorphous nature of ICN. This experiment was performed by exposing the samples to Cu-Kα1 radiation of 40 kV and 30 mA. The scanning rate was5°/min over a range of 4–90°with an interval of 0.1°. Different ICN nano formulations were prepared by varying the ratios of chitosan for the fixed amount of isolongifolene. The different concentrations of chitosan (0.1, 0.2, 0.3, 0.4, and 0.5% w/v) were obtained by dissolving in acetic acid (0.25% v/v) under constant stirring (250 rpm) for overnight. About 20 mL of TPP of 0.75% (w/v) was added into chitosan solution (10 mL) containing 0.20 mg of isolongifolene and constantly stirred at 320 rpm for 6 h at 4 °C. Sonication (Sonitvibra cell, UC130, USA, Amplitude-20, Pulser-4 s) was performed for 10 min at 4 °C. Then, the dispersion was washed thrice (2200 × g) and re-dispersed in HPLC grade water (LiChrosolv). Further, the resultant solution was centrifuged at 13,000 × g for 20 min at 4 °C to obtain five different Isolongifolene polymer ratio nano formulations (ICN-K01 (1:0.5), ICN-K02 (1:1), ICN-K03 (1:1.5), ICN-K04 (1:2) and ICN-K05 (1:2.5). The resultant nanoparticles were freeze-dried (Delvac-lyo1550, INDIA) and stored at − 20 °C for further characterization. Process yield was calculated by weighing the freeze-dried nano formulations i.e., ICN-K01 to ICN-K05 using the reported method. The mean values of three replicates were expressed as process yield. The drug encapsulation efficiency (DEE) and drug loading efficiency (DLE) were calculated following the standard method reported earlier with some modifications. Further, ICN was pelletized at 5800 × g and the obtained supernatant was quantified for isolongifolene by HPLC as described earlier. The percentages of DEE and DLE were calculated using the formula given below:- In vitro drug release was studied by plasma simulation and dialysis membrane method with slight modifications. The nanoformulations were redispersed in 10 mL of NaCl (0.9% w/v) with final Isolongifolene concentration of 20 μg/mL. 1 mL of the above-mentioned mixture was added to 10 mL of PBS (0.5 M, pH = 5.5) and 10 mL of plasma. The suspension was kept in an orbital shaker at 37 °C. One milliliter of the released solution was collected at different time intervals and replaced with a fresh solution. The harvested solutions were centrifuged and the supernatant was used to analyze isolongifolene content using the HPLC method. The release rate was calculated using the following equation. In the dialysis membrane method, 2 mg of nanoformulations were redispersed in 10 mL of PBS (0.5 M, pH = 5.5), placed in a dialysis membrane (cut-off 10 kDa) and dialyzed against PBS (0.5 M pH = 5.5). 2 mL of released solution was collected at different time intervals, replaced with fresh solution and analyzed using HPLC. The release rate of Isolongifolenecan is derived from the calibration curve prepared using known concentrations. The size distribution and zeta potential of different nanoformulations (ICN-K01–ICN-K05) were measured by Zetasizer Nano ZSinstrument (Malvern, Mastersizer 2000, UK). The sample was dispersed in water (pH = 5.5) and the nanoparticles were counted in a 4.8 mm calibrated area with a count rate of 210.3 kcps (kilo counts per second) for 70 s. The average hydrodynamic diameter of different nanoformulations was calculated as mean values. The surface morphology of ICN was analyzed using Field Emission-Scanning Electron Microscopy (FE-SEM) (TESCAN, VEGA3 SBU, Czech). The average particle size and shape of ICN were further studied using High Resolution-Transmission Electron Microscope (HR-TEM) (Jeol, JEM2100, Japan). Fresh blood from Wistar rats was collected in heparinized tubes and the plasma was separated by centrifugation (1800 × g for 15 min at 4 °C). The 10 μg conc. of ICN-K04 was added into the separated plasma and further incubated at 37 °C for 30 min in 0.9% (w/v) of NaCl solution. 1 mL plasma solution was collected at regular time intervals and stored at − 20 °C until use. The isolongifolene content was analyzed by the HPLC method after centrifugation at 17,000 × g for 20 min. This study was carried out in compliance with the CPCSEA safety guidelines and following the ARRIVE Guidelines (https://arrivedguidelines.org) for the reporting of animal experiments. The study was approved by the Institutional Animal Ethics Committee (IAEC) of Periyar University Approval No: 1085/ac/07/PU/IAEC/Feb2012/04. The hemocompatibility of ICN was evaluated using the reported procedure. The whole blood was collected from a Wistar rat and anticoagulated with sodium citrate (ratio of blood to anticoagulant taken was 9:1). Erythrocytes were isolated by centrifuging whole blood at 1000 × g for 10 min. The erythrocytes were washed thrice with saline before use. ICN-K04 was mixed with RBCs in different concentrations (2–24 μg/mL) and then incubated for 2 h at 37 °C and the supernatant was collected by centrifugation at 1500 × g for 5 min. Hemoglobin release was monitored spectrophotometrically (Systronics, 2203, INDIA) at 541 nm. The TritonX-100 (1% v/v) and 0.9% (w/v) NaCl were taken as positive and negative controls respectively. The percentage of hemolysis was calculated using the following formula:-where ODtest, ODneg., and ODpos are the absorbance values of the test sample, negative control, and positive control, respectively. Cell viability was analyzed using a conventional MTT reduction assay. Cells were treated with ICN and the viability was assessed based on the detection of mitochondrial dehydrogenase enzyme activity in viable cells. Cells were cultured in a 96-multiwell plate. The 3 × 103 cells were seeded to each well. Initially, the cells in the medium were pre-incubated with or without ICN for 24 h. After 24 h, the cells were incubated with MTT (5 mg/mL) at 37 oC for 4 h. Following incubation, the medium was removed and the formed formazan crystals were dissolved with DMSO. The absorbance of the reduced product, formazan was measured at 570 nm using an ELISA plate reader (Bio-Rad, Hercules, CA, USA). The percentage of cell viability can be determined using the below formula:-where O.D represents Optical Density. The A549 cell lines were procured from ATCC, USA and were used for apoptotic studies. The Apoptosis assay was performed using both acridine orange (AO) and ethidium bromide (EtBr) dyes. Acridine orange, a permeable dye stain all the cells, and ethidium bromide permeate into the cell only when the cell membranes disintegrate. EtBr intercalates with DNA forming an orange-red complex. Once the medium was removed from the plates after treatment, the cells were washed twice with phosphate-buffered saline (PBS) and stained with AO and EtBr dyes. The stained cells were incubated for 20 min at room temperature and washed with warm PBS to remove the excess dye. The cellular morphology was observed using a fluorescent microscope (λex/λem = 490 nm/530 nm) and the images were captured. The fluorescent intensity was recorded at 535 nm using Spectro-fluorimeter. The present study follows institutional guidelines mandatory for human and animal treatment and complies with relevant legislation ethical approval from the institute for conducting the research. The study was conducted according to the ethical norms approved by Institutional Animal Ethical Committee of Periyar University Approval No: 1085/ac/07/PU/IAEC/Feb2012/04. The FT-IR spectrum of isolongifolene polymers such as sodium alginate, gelatin, and chitosan and their interaction with isolongifolene were shown in Fig. 1. The FTIR spectra of isolongifolene shows signature peaks at 1242 cm−1 (C–O stretch), 1407 cm−1 (C–C stretch), 1590 cm−1(C=O stretch), 2965 cm−1(C–H stretch), 3688 cm−1(O–H stretch). ISN spectrum showed significant peaks at 3218 cm−1(O–H vibration) and 1673 cm−1(C=C stretching). However, an ISN spectrum lacks an FT-IR band at 1242 cm−1corresponding to the C–O stretch of isolongifolene (Fig. 2). To find the effective isolongifolene chitosan polymer ratio, different concentrations of chitosan were varied for a fixed amount of isolongifolene. The higher the amount of isolongifolene in polymer, higher the efficiency. Moreover, the increase in the Isolongifolene Chitosan ratio also increases the process yield as observed for the formulation ICN-K01 to ICN-K05. The formulation ICN-K05 which comprises of 1:2.5 ratio of isolongifolene chitosan polymer displayed a higher process yield (79.05 ± 4.60%). The association and entrapment efficiency of formulations increased with an increase in the isolongifolene chitosan ratio from 1:0.5 to 1:2.5. Figure 1 shows the statistical significance level of pharmaceutical characteristics of isolongifolene (ICN). The hydrodynamic diameter of the different nanoformulations (ICN-K01 to ICN-K05 ranging from 322 to 538 nm was observed. Transmission electron microscopic (TEM) analysis revealed the particle size was found within the size range of 200–250 nm as shown in Fig. 4. Increased particle size seen with higher isolongifolene chitosan polymer ratio could be due to the increased swelling and subsequent polymer interaction. The polydispersity index (PDI) of the nanoformulations (ICN-K01 to ICN-K05) indicates the width of particle size distribution and is found to vary between 0.177 and 0.404. Amongst, the formulation K04 exhibited low PDI (0.195) when compared to other isolongifolene polymer ratios studied. The release pattern of nanoformulations ICN-K01 to ICN-K05 was analyzed by plasma simulation and dialysis method. There was no significant difference in the release pattern of nanoformulations studied using plasma simulation and dialysis membrane methods respectively (Fig. 3A,B). The release profile of isolongifolene was directly proportional to the size of ICN. The release pattern of isolongifolene in 25 h was gradually decreased due to the increased size of ICN. About 50% release of isolongifolene was attained in 3–10 h. The XRD spectrum of chitosan has shown a sharp peak at 25 °C which was found to be diminished in the case of ICN as shown in Fig. 4. The formation of Spherical shaped ICN-K04 nanoparticles was confirmed under transmission and scanning electron microscopic studies as shown in Fig. 5. The hemocompatibility of ICN-K04 was studied by varying the concentration of ICN-K04 for a fixed amount of erythrocytes. It was observed that there was no significant damage occurred to erythrocytes till 14 μg/ml of ICN-K04. Above 14 μg/ml, the increase in nanoparticle concentration has shown to incur significant damage to erythrocytes and the phenomena were found to be concentration-dependent (Fig. 6). Genotoxicity was investigated by a plasmid nicking assay. Results showed that the band intensity corresponding to pUC19 plasmid alone in lane 1 was kept as control (100% intensity). However, the plasmid subjected to Fenton's reagent showed moderate DNA fragmentation with the lowest band intensity of 25%. The addition of 5 and 10 µg ICN-K04 did not incur significant damage to the plasmid with band intensities of 85% and 80% respectively (Fig. 7). Plasma stability studies revealed that the formulation ICN-K04 was moderately stable up to 50 min and the stability gradually decreases further, as shown in Fig. 8. Invitro cytotoxicity of the nanoformulation ICN-K04 of varying concentrations (6.25, 12.5,25,50,100 μM) was investigated using an MTT assay on the A549 cell line for 24 h as shown in Fig. 9. Live cells reduced the MTT and the resultant formazan is directly proportional to the cell viability. Results showed that the viability of cancer cells was inhibited in a dose-dependent manner with a 50% inhibitory concentration (IC50) of 13.42 μM. The addition of 25 μM of ICN-K04 decreased the % cell viability to 23.7%. For further studies, the half-maximal concentration (IC50) of ICN-K04 was used unless otherwise mentioned. Morphological alterations in A549 cancer cells such as viability, nuclear morphology, and chromatin condensation were evaluated by AO/EB staining technique. This method differentiates the viable cells from dead cells by monitoring the bright green nuclei and orange to red nuclei, depending on the loss of membrane integrity. Fluorescent microscopic studies of AO/EB stained A549 cells after treatment with ICN-K04 were shown in Fig. 10. Untreated cancer cells (control) displayed green fluorescent nuclei with normal cellular morphology. Whereas, cancerous cells treated with 6.7 and 13.42 μM ICN-K04 showed a significantly (P < 0.05) increased percentage of apoptotic cells. Hitherto, the major clinical challenge in treating the cancer is the relapse prevalence, which occurs due to the failure of the primary treatment regimen in targeting the cancerous cells. Therefore, drugs having the potential of huge accessibility and specifically delivered to the target site hold a great promise in the management of cancer. Nanoparticle-based drug delivery system has gained huge attention due to its versatile approach to accessing the cancerous site and another inflammatory milieu. Thus, the synthesized nanoparticles have a unique potential to overcome the poor penetrating ability and also exert their maximum efficacy to control the proliferation of the disease. The current study investigated the compatibility of Isolongifolene nanoformulation with different polymers. Preliminary studies showed that the chitosan-based Isolongifolene polymeric nanoformulation could act as an excellent adjuvant in therapeutics, mainly treating multi-drug resistance in solid tumors. In the FT-IR spectroscopic study, the peaks revealed the surface chemistry of the presence of functional groups other than the innate molecules as confirmed by the signature FT-bands with different wavenumbers corresponding to different significant changes obtained in shape and position of the absorbance bands. Among the different ratios investigated, the 1:2.5 ratio of isolongifolene and polymer ratio yields a higher process yield, which could be due to greater carbon efficiency. In an experimental condition reduction of pH leads to greater association and loading efficiency. Zeta potential is an important physicochemical property that reflects the physical stability and mucoadhesive properties of nanoparticles. In principle, zeta potential values in the range of <− 30 mV and > + 30 mV are considered stable regimes. In the present study, the ICN-K04 formulation possesses a zeta potential of +39 mV among others, which demonstrates its good interaction stability. Polydispersity Index (or) PDI is the measure of the particle size distribution with values ranging from 0 to 1. In this study, the PDI values close to 0 (zero) presented a homogeneous dispersion and those greater than 0.5 showed high heterogeneit. The prepared nano formulations (ICN-K01 to ICN-K05) showed PDI values in the range of 0.1 to 0.4. Especially, ICN-K04 possesses a PDI of 0.1 which indicates the presence of monodispersed (homogenous) particles. Moreover, the nanoformulation ICN-K04 showed increased association efficiency which was inversely proportional to the isolongifolene release. Subsequently, in vitro release of isolongifolene monitored by plasma simulation and membrane dialysis method showed the lack of burst effect and also confirmed that the interactions between isolongifolene and the nanoparticles were weak. ICN formulation was found to be optimum based on zeta potential, PDI value, and in vitro release methods, which demonstrate that ICN-K04 possesses good stability and homogenous dispersity. Further, the ICN-K04 formulations were used for the anti-carcinogenic study. The ultra-structural image of the formulation revealed the presence of nearly shaped solid particles after the incorporation of isolongifolene. The formation of intermolecular hydrogen bonding did not show any significant shearing effect on the surface of ICN-K04. The biocompatibility of nano formulations used as drug delivery systems is a crucial parameter. The integrity of the hemoglobin structure might function as a key factor to determine the potential biocompatibility of nanoparticle. A few other mechanisms have been proposed whereby hemolysis can contribute to thrombosis in addition to changing the hematocrit and hemorheology, such as the release of erythrocyte-derived macrovesicles, activation of the complement cascade, and the release of free haemoglobin and heme into circulation, which sequesters nitric oxide. Hemolysis is a crucial factor in the hemocompatibility testing of biomaterials and can have a big impact on how well they work in the clinic. A number of NPs, including amorphous silica, tricalcium phosphate, hydroxyapatite, and particularly silver (Ag) NPs, have been discovered to significantly cause hemolysis, endangering their use in biomedical applications. Most NPs have hemolytic activity; however, it depends on concentration, structure, size, and shape. For instance, the amount of reactive silanol groups exposed on the surface of silica NP is directly related to the size and geometry of the NP. Surface charge, shape, porosity, and surface functionalization with certain polymers or functional groups are the most important surface characteristics that determine the hemocompatibility of NPs.. Upon administration, injected particles will most likely interact with red blood cells. Electrostatic interactions between the red blood cells and the nanoparticles can cause perturbation of the membrane, thereby causing hemolysis. In the present experimental study ICN-K04 showed and induced hemolysis at higher concentrations, which could be due to the electrostatic attractive forces between chitosan and erythrocytes and subsequently lead to thrombus formation. Nanoparticle movements were faster as compared to macromolecules. It was also expected that the nano formulations could reach the tissue compartment within 50 min. In cancer therapy, tumor growth can be suppressed by activating the apoptotic machinery in the cell. Many malignant cells, however, are unable to regulate the genes that control apoptosis, rendering them resistant to the induction of apoptosis by a variety of stimuli, including intracellular and extracellular signals chemotherapeutic drugs, and radiotherapy. A previous study has investigated that the SH-SY5Y cells exposed to rotenone caused about 50% cell death, whereas isolongifolene pre-treated cells dose-dependently regulated the toxic effects of rotenone by which it exhibits neuroprotective effect against apoptosis. In our study, we investigated the cytotoxicity of nanoformulation ICN-K04 at different A549 cell lines for 24 h. The result exhibited that the viability of cancer cells was inhibited by 50% in a dose-dependent manner, and also we found that 50% inhibitory concentration (IC50) of ICN-K04 at 13.42 μM among all other concentrations respectively. Isolongifolene loaded chitosan nanoparticles were successfully prepared and characterized for parameters like association efficiency, loading efficiency, zeta potential, entrapment efficiency, and polydispersity index (PdI). Among the nano formulations (ICN-K01to ICN-K05) prepared, the ICN-K04 formulation was found to be optimum. The X-ray diffraction spectrum of ICN-K04 formulation confirms the poor crystalline nature of the nanoparticles. The spherical morphology of ICN-K04 nanoparticles was confirmed by scanning and transmission electron microscopy. More importantly, ICN-K04 nanoparticles have low hemolytic activity and also lack genotoxicity. ICN-K04 nanoparticles show intrinsic plasma stability of more than 50 min. As a prerequisite for a nanoparticle-based drug delivery system, it is imperative to understand the size, morphology, as well as surface charge of nanoparticles which can deliver the loaded drug effectively and specifically to its targeted site. From the results it can be corroborated that the ICN-K04 nanoparticles show promising data in all the physiochemical criteria, so that they can be used as a delivery system for the drug to show conspicuous effects to the targeted site. Further, validation studies of this drug delivery approach using the nanoparticles system are underway.

PMC9649690

Wei Hu,Yangjun Wu,Qili Shi,Jingni Wu,Deping Kong,Xiaohua Wu,Xianghuo He,Teng Liu,Shengli Li

Systematic characterization of cancer transcriptome at transcript resolution

10-11-2022

Cancer genomics,Data mining,Regulatory networks,Cancer therapy,Transcriptomics

Transcribed RNAs undergo various regulation and modification to become functional transcripts. Notably, cancer transcriptome has not been fully characterized at transcript resolution. Herein, we carry out a reference-based transcript assembly across >1000 cancer cell lines. We identify 498,255 transcripts, approximately half of which are unannotated. Unannotated transcripts are closely associated with cancer-related hallmarks and show clinical significance. We build a high-confidence RNA binding protein (RBP)-transcript regulatory network, wherein most RBPs tend to regulate transcripts involved in cell proliferation. We identify numerous transcripts that are highly associated with anti-cancer drug sensitivity. Furthermore, we establish RBP-transcript-drug axes, wherein PTBP1 is experimentally validated to affect the sensitivity to decitabine by regulating KIAA1522-a6 transcript. Finally, we establish a user-friendly data portal to serve as a valuable resource for understanding cancer transcriptome diversity and its potential clinical utility at transcript level. Our study substantially extends cancer RNA repository and will facilitate anti-cancer drug discovery.

Systematic characterization of cancer transcriptome at transcript resolution Transcribed RNAs undergo various regulation and modification to become functional transcripts. Notably, cancer transcriptome has not been fully characterized at transcript resolution. Herein, we carry out a reference-based transcript assembly across >1000 cancer cell lines. We identify 498,255 transcripts, approximately half of which are unannotated. Unannotated transcripts are closely associated with cancer-related hallmarks and show clinical significance. We build a high-confidence RNA binding protein (RBP)-transcript regulatory network, wherein most RBPs tend to regulate transcripts involved in cell proliferation. We identify numerous transcripts that are highly associated with anti-cancer drug sensitivity. Furthermore, we establish RBP-transcript-drug axes, wherein PTBP1 is experimentally validated to affect the sensitivity to decitabine by regulating KIAA1522-a6 transcript. Finally, we establish a user-friendly data portal to serve as a valuable resource for understanding cancer transcriptome diversity and its potential clinical utility at transcript level. Our study substantially extends cancer RNA repository and will facilitate anti-cancer drug discovery. RNA transcripts, as the direct carriers of translational codes of proteins, are generated via diverse regulation and modification in particular contexts. Under different contexts, transcriptional activity shows diversity in that different quantities of transcripts or totally distinct transcripts are produced from the same genes. Transcriptional diversity greatly expands the encoding storage capacity of the genome for proteins in eukaryotes. Along with advancements in high-throughput RNA sequencing (RNA-seq)and the development of computational algorithms, transcriptional diversity has been largely brought to light in human diseases. By reanalyzing RNA-seq data from 32 cancer types, Kahles et al. detected thousands of alternative splicing variants in many tumors. Xiang et al. identified many alternative polyadenylation variants in 6398 tumor patients and 739 cancer cell lines, wherein 1971 variants were found to be clinically relevant. These transcriptional variants generate specific RNA transcripts that may play important roles in tumor development. For example, a LIN28B variant, LIN28B-TST, was found to be specifically expressed in tumor samples and critical for cancer cell proliferation and tumorigenesis. However, an integrative depiction of RNA transcripts in large-scale cancer transcriptomics data has been lacking. RNA-binding proteins (RBPs)have been shown to play crucial regulatory roles in post-transcriptional RNA expression, and aberrantly programmed RBP-RNA interactions modulate cancer initiation and progression. Van Nostrand et al. reported the largest effort to date to systematically study the functions of 356 human RBPs. They combined the RBP binding data (eCLIP-seq)and RBP knockdown followed by RNA-seq data (KD-RNA-seq) to construct the RBP-gene regulation and RBP-splicing association. Furthermore, several RBPs have also been demonstrated to mediate the response of cancer cells to anti-cancer drugs, such as estrogen receptor α (ERα) and eukaryotic translation initiation factor 2 subunit beta (EIF2S2). Nevertheless, RBP regulation at the transcript level and its roles in mediating the response to anti-cancer drugs of cancer cells remain incompletely understood. Human cancer-derived cell lines have been widely used as preclinical cancer models in cancer biology research and anti-cancer drug discovery. Massive efforts have been made to delineate the molecular characteristics across large-scale cancer cell lines. The Cancer Cell Line Encyclopedia (CCLE) project generated high-throughput sequencing data of hundreds of cancer cell lines at various molecular levels, including genomics, transcriptomics, epigenomics, proteomics, and metabolomics. To further expedite drug discovery, the Genomics of Drug Sensitivity in Cancer (GDSC) project provides an unprecedented resource about drug sensitivity for 266 anti-cancer agents across 1,065 different cancer cell lines, while the Cancer Therapeutics Response Portal (CTRP) provides information on responses to 481 compounds in 860 cancer cell lines. Large biological troves and potential clinical applications have been discovered through the multifarious integrative characterization of sophisticated molecular landscapes across cancer cell lines. To comprehensively delineate the transcript atlas in cancer, we carried out a reference-based transcript assembly with RNA-seq data across more than 1000 cancer cell lines. The unannotated transcript AC092803.3-u1 from the AC092803.3 gene was experimentally validated and showed a higher expression level and clinical significance in multiple tumor types than the other transcript, AC092803.3-a1. Furthermore, RBP-transcript regulation and transcript-drug associations in cancer were combined to build RBP-transcript-drug axes, wherein PTBP1 was experimentally validated to affect the sensitivity to decitabine by regulating the expression of the KIAA1522-a6 transcript of the KIAA1522 gene. We also developed a user-friendly data portal to benefit the biomedical research community. To extensively dissect the transcriptional atlas across pan-cancer cell lines at the transcript level, reference-based transcript assembly was performed across pan-cancer cell lines (see “Methods”). Briefly, 1017 transcriptomes of cancer cell lines derived from 25 different lineages (Supplementary Fig. 1a and Supplementary Data 1) were subjected to two-round alignments to identify all possible splicing junctions (Supplementary Fig. 1b). Based on this comprehensive repertoire of splicing junctions, expressed transcripts derived from various genomic regions were assembled and quantified. In total, 498,255 transcripts were detected in at least one cell line. On average, 72.31% of transcripts showed expression levels lower than 0.1 TPM, and 11.64% were found to be expressed at higher than 1 TPM (Supplementary Fig. 1c). Among all detected transcripts, 27.24% were detected in less than 10% of all cell lines, while 19.78% were expressed in more than 90% of cell lines (Supplementary Fig. 1d). Except for non-coding and intergenic RNAs, which are likely uncharacterized or small regulatory RNAs, most transcripts from non-coding genomic regions have a length distribution similar to those from protein-coding regions (Supplementary Fig. 1e). Transcript expression profiles were then adopted to cluster all cell lines, wherein cell lines from the same or close primary sites were clustered closer (Fig. 1a). The numbers of identified transcripts ranged vastly across different cell lines, with the largest median number of 2062 (per million mapped reads) in prostate and the smallest median number of 1472 (per million mapped reads) in biliary tract cell lines (Fig. 1b). Newly assembled transcripts were compared to those annotated in various databases/datasets to filter unannotated transcripts (see “Methods”), removing 35,986 transcripts from the unannotated transcripts (Supplementary Fig. 2a). The vast majority (72.55%) of transcripts were expressed from protein-coding genes (Fig. 1c, Supplementary Data 2). Among all detected transcripts from protein-coding regions, approximately half (50.57%) were unannotated. Moreover, a considerable portion of transcripts are from non-coding genomic regions, such as lncRNAs (13.04%) and pseudogenes (4.81%). The numbers of lineage types that expressed individual transcripts showed a hump distribution, wherein most transcripts were identified in 22 different lineages or one specific cell line lineage (Supplementary Fig. 2b). The lineage specificity scores were then calculated to identify cancer cell line lineage-specific transcripts (see “Methods”). In total, we identified 72,865 lineage-specific transcripts across 22 different lineages. We further evaluated the specificity of host genes that generated these lineage-specific transcripts. We found that the majority of lineage-specific transcripts were generated from non-specific host genes (Supplementary Fig. 2c). The intergenic and long non-coding RNA transcripts showed the highest overall specificity scores, followed by pseudogene and non-coding RNA transcripts (Supplementary Fig. 2d). The numbers of lineage-specific transcripts ranged from 1201 in the pancreas to 12,769 in haematopoietic and lymphoid cell lineages (Fig. 1d, Supplementary Data 3). For example, the LAPTM5-a7 transcript (transcribed from the LAPTM5 gene) and CORO1A-u2 transcript (an unannotated transcript from the CORO1A gene) are exclusively expressed in cancer cell lines derived from haematopoietic and lymphoid lineages (Fig. 1e). One transcript of the MFSD12 gene, MFSD12-a11, and the unannotated AC141557.1-u1 transcript from the AC141557.1 gene showed specific transcriptional activities in skin cell lines (Fig. 1f). We further analyzed available long-read RNA-seq datasets (see “Methods”). In total, 6.23% of our unannotated transcripts were mapped by long-read RNA sequencing reads (Supplementary Fig. 3a). Unannotated transcripts with high expression level were more likely to be mapped by long-read RNA-seq data, wherein 18.23% of the transcripts that showed top 10% expression levels were found to be overlapped by long-read RNA-seq reads (Supplementary Fig. 3b). The coverage of transcriptome by long-read RNA-seq might be lower than that of short-read RNA-seq. In particular, less than 8% of annotated transcripts were covered by long-read RNA-seq reads, while appropriately 40% of them were mapped by short-read RNA-seq reads (Supplementary Fig. 3c). Therefore, the validation percentages of unannotated transcripts by long-read RNA-seq reads in our study were acceptable and reasonable. In a previous study that used short-read RNA-seq data for transcriptome assembly, 7.6% of their newly assembled single-exon transcripts were mapped by long-read RNA-seq reads. To provide more transcription evidence of our unannotated transcripts, we analyzed the CAGE sequencing data from the FANTOM project, and the chromatin states from the Roadmap Epigenomics project. In total, 78.64% of unannotated transcripts were overlapped by transcription evidence (8.28% by only CAGE, 24.08% by only active chromatin states, and 46.29% by both CAGE and active chromatin states), which was comparable with the annotated transcripts, 86.62% of which were mapped by transcription evidence (Supplementary Fig. 3d). In summary, our results presented an extended compendium of the cancer transcriptome and revealed many unexplored RNA transcripts. As unannotated transcripts constituted over half of our cancer transcript atlas, we next investigated whether the expression of these unannotated transcripts was associated with disease progression or prognosis in human cancer. In total, 253,254 unannotated transcripts were identified. Unannotated transcripts from protein-coding genes exhibited relatively lower expression level, while those from the other gene types showed comparable expression levels over annotated transcripts across different expression ranges (Supplementary Fig. 4a). The majority of unannotated transcripts were derived from alternative splicing junctions of multiple exons with at least one annotated junction (Fig. 2a and Supplementary Fig. 4b). In addition, approximately one-third (35.46%) of unannotated transcripts were readthrough transcripts (Fig. 2b). To investigate the possible biological functions that unannotated transcripts may participate in, we performed correlation analysis between unannotated transcripts and transcriptional activities of hallmark biological processes. Significant hallmarks (FDR < 0.05) with the highest correlation were linked to the corresponding transcripts. Epithelial–mesenchymal transition (EMT) was found to be associated with the largest number of unannotated transcripts (Fig. 2c and Supplementary Fig. 5a). The expression levels of individual unannotated transcripts varied widely across different cell lines (Fig. 2d). Compared to paired adjacent non-tumor samples, 75,343 unannotated transcripts showed significantly differential expression in at least one tumor type, wherein LUSC had the largest number of differentially expressed transcripts (Supplementary Fig. 5b). Most of these differentially expressed transcripts showed specific differences in one or two cancer types (Supplementary Fig. 5b). For example, the UBE2C-u5 transcript showed significant upregulation in 11 different cancer types (Supplementary Fig. 5c). Furthermore, 119,212 showed significant association with tumor stages (Supplementary Fig. 6a), and 131,506 unannotated transcripts were found to be significantly associated with the overall survival of tumor patients in at least one tumor type (Supplementary Fig. 6b). Higher expression of UBE2C-u5 also indicated more advanced tumor stages (Supplementary Fig. 7a) and poorer prognosis (Supplementary Fig. 7b) across different cancer types. Unannotated transcripts from protein-coding gene regions exhibited relatively lower expression level, while those from the other gene types showed comparable expression level over corresponding annotated transcripts across different expression ranges (Supplementary Fig. 3a). Comparable expression levels indicated that these unannotated transcripts from non-protein-coding regions might possess alternative or stronger functions over known transcripts of the same host genes in cancer. LncRNAs have been demonstrated to have key roles in human cancer, which are also the vast majority of non-protein-coding regions where our unannotated transcripts are from. To select stable and representative unannotated transcripts from lncRNAs for validation, we first ranked the expression levels of those that have no overlaps with protein-coding genes and ≤1000 bp in length (Supplementary Fig. 8a). We also calculated the number of cancer types that unannotated transcripts had survival significance but the corresponding annotated transcripts didn’t. RACE assays were performed to validate the top 10 unannotated transcripts with high average expression level. The unique sequences or junctions of three unannotated transcripts, including CRIM1-DT-u1, AC107032.2-u1, and AC092803.3-u1, were confirmed by 3′ RACE and Sanger sequencing (Supplementary Fig. 8b). Of the three transcripts, AC092803.3-u1 had the largest number of cancer types where only unannotated transcripts had survival significance. The AC092803.3-u1 transcript was overlapped by 2, 13, and 4 long-read RNA-seq reads in the K562, PC9, and CACO2 cell lines, respectively (Supplementary Fig. 8c) We also validated the expression of AC092803.3-u1 in cancer tissue samples, wherein AC092803.3-u1 showed significantly higher expression than the corresponding annotated transcript AC092803.3-a1 (P = 0.022) (Supplementary Fig. 8d). The AC092803.3-u1 transcript was found to be transcribed from the AC092803.3 gene, which was derived from splicing and joining of 3 exons, two of which were uncharacterized and quite different from the annotated transcript AC092803.3-a1 (Fig. 2e). The junction that joins exon 2 and exon 3 of AC092803.3-u1 was not found in AC092803.3-a1. The junction of uncharacterized exons 2 and 3 was further validated by 3′ RACE and Sanger sequencing, which demonstrated the valid expression of the AC092803.3-u1 transcript (Fig. 2e). The AC092803.3-u1 transcript showed a significantly higher expression level than the other transcript (P = 2.6E−5), AC092803.3-a1, across 1017 cell lines (Fig. 2f). The expression level of AC092803.3-u1 significantly distinguished tumor patients with longer survival times from those with shorter survival times, while AC092803.3-a1 exhibited no association with patient survival in adrenocortical carcinoma (ACC), brain lower grade glioma (LGG), liver hepatocellular carcinoma (LIHC), and ovarian serous cystadenocarcinoma (OV) cohorts (Fig. 2g). Compared to that in paired non-tumor samples, AC092803.3-a1 showed no expression difference in tumor samples, whereas AC092803.3-u1 was found to be significantly differentially expressed in cholangiocarcinoma (CHOL, P = 0.015), kidney renal clear cell carcinoma (KIRC, P = 0.00042), and head and neck squamous cell carcinoma (HNSC, P = 0.0016) cohorts. These results indicate that a large number of transcripts with considerable expression levels and valuable clinical significance remain unexplored in the cancer transcriptome. RBPs have been shown to extensively participate in the regulation of post-transcriptional modifications, thus modulating the expression levels of transcripts. We next aimed to establish RBP regulation relationships at the transcript level. The RBP knockdown (KD-RNA-seq) and binding (eCLIP-seq) data were integrated to identify high-confidence RBP-transcript regulatory pairs (see “Methods”). A high-confidence regulatory network that consisted of 129 different RBPs and 47,667 different transcripts was constructed (Fig. 3a, Supplementary Data 4), wherein the numbers of up- and down-regulated transcripts had no obvious difference for each RBP (Supplementary Fig. 9a, Supplementary Data 4). We further analyzed the essentiality of RBPs in cancer cells (see “Methods”). Over half of these RBPs (70, 53.85%) showed essentiality in no more than 68 of all examined cancer cell lines (Supplementary Fig. 9b). RBPs that are essential genes in most cell lines also showed a broad range of dependency scores across different cancer cell lines (Fig. 3a and Supplementary Fig. 9c). These RBPs might be important targets for cell viability. Our analysis suggested that RBPs could be targeted to modulate cell viability in specific cell types. The numbers of regulated transcripts varied largely among different RBPs, with the largest number for AQR (8,955 transcripts) and the smallest number for SBDS (2 transcripts) (Fig. 3b). RBPs were categorized according to their primary functions, including “spliceosome”, “splicing regulation”, “modification & processing”, “stability & decay”, “other” and “novel RBP”. RBPs in the “spliceosome” and “stability & decay” categories regulate many more transcripts than those in other categories (Supplementary Fig. 9d). The major portion of transcripts were regulated by a very small number of different RBPs; for example, 21,101 (44.27%) transcripts were regulated by only one RBP, and 8754 (18.36%) transcripts were regulated by two different RBPs (Fig. 3c). However, some transcripts were regulated by many different RBPs, such as the FRMD8-u2 transcript, which was regulated by 47 different RBPs. Most of the RBPs regulated transcripts that were involved in proliferation, especially “G2M checkpoint”, “MYC targets”, and “E2F targets” (Fig. 3d). Some RBPs regulated transcripts with specific biological functions; for example, the regulation of NOL12 was significantly enriched in “Coagulation” and “Complement”. Compared to those regulated by RBPs that were related to “splicing regulation” and “modification & processing”, transcripts regulated by “spliceosome”- and “stability & decay”-associated RBPs were significantly enriched in “MYC targets v1” (Fig. 3e). In summary, our analysis revealed a fine RBP-transcript regulatory network that might guide transcript manipulation through specific RBPs. Some human RBPs have been shown to express in a tissue- or cancer-type specific manner, such as ELAVL3 and ELAVL4 in the neuron. Additionally, RBPs that support basic cellular functions are widely expressed across tissues, such as ribosomal and spliceosome RBPs. To extensively examine the expression specificity of RBPs, we collected 1751 RBPs from previous studies, and calculated a specificity score for each RBP in the RNA-seq datasets of the CCLE, TCGA and GTEx project. We employed the Shannon entropy method to calculate specificity score as described in previous studies. In total, 97 (5.57%), 106 (6.09%), and 128 (7.38%) cancer/tissue-type specific RBP genes (specificity score >1) were identified in the CCLE (Supplementary Fig. 10a and Supplementary Data 5), TCGA (Supplementary Fig. 10b), and GTEx (Supplementary Fig. 10c) datasets, respectively. Our results were consistent with previous investigations of human RBP cell/tissue specificity. Of these RBPs involved in our study, LIN28B showed specifically high expression in the liver and autonomic ganglia cancer cell lines, LIN28B and IGF2BP1 exhibited exclusively high expression in the testicular germ cell cancer, LIN28B and IGF2BP1 showed exclusively high expression in the testis tissue, and IGF2BP3 exhibited specifically high expression in the skin and testis tissue. Some RBPs may express specific transcripts in certain cancer/tissue types, which generate tissue/cancer-specific RBP protein isoforms. We next calculated specificity scores of 16,312 transcripts generated from RBP genes. On the whole, 688 (4.82%), 2549 (16.85%), and 1548 (10.36%) cancer/tissue-specific RBP transcripts were identified in the CCLE (Supplementary Fig. 10d), TCGA (Supplementary Fig. 10e), and GTEx (Supplementary Fig. 10f) datasets, respectively. Our analysis revealed that many RBPs might exert tissue/cancer-specific functions by expressing specific transcripts and protein isoforms. To further explore the potential clinical utility of transcripts in cancer, associations between transcript expression and anti-cancer drug sensitivity were evaluated. Based on correlations with transcript expression, anti-cancer drugs were clustered, wherein drugs with targets from similar gene classes were more closely clustered, such as HDAC- and EGFR-targeted drugs (Fig. 4a). To further identify the transcripts that were closely associated with the sensitivity to anti-cancer agents, transcripts were first filtered to keep those with considerable abundance across cancer cell lines (Fig. 4b). Then a preliminary correlation between each transcript and drug was evaluated to select possible transcript-drug pairs (see “Methods”). Significant transcript-drug pairs were subjected to an elastic net regression model with 5 rounds of repeated 10-fold cross-validation to optimize the α and λ parameters. The optimized model was submitted to a bootstrapping procedure to generate a predictive score for each transcript-drug pair. Transcript-drug pairs with passing predictive scores (≥0.7) were retained as high-confidence transcript-drug associations. The transcript-drug association network comprised 43,602 (top 1.92%) transcript-drug pairs (Supplementary Fig. 11a). Most of the anti-cancer agents had larger numbers of positively associated transcripts, for example, the sensitivity of PHA-793,887 was positively associated with 781 transcripts and negatively associated with 442 transcripts (Fig. 4c). Moreover, some drugs had balanced positively and negatively associated transcripts, such as ABT-737, which was positively associated with 328 transcripts and negatively associated with 436 transcripts. Most of the anti-cancer agents-associated transcripts were dispersed across various biological processes, while some were notably enriched in specific processes; for example, transcripts that were associated with “Notch signalling”-targeted drugs were specifically enriched in “EMT”, “Apical junction”, and “UV response down” (Fig. 4d). Only “IGF1R signalling”-targeted drugs were associated with transcripts that were involved in immune-related processes, such as “inflammatory response” and “IL6 JAK STAT3 signalling”. Our analysis revealed a large number of anti-cancer drug-associated transcripts that may be used to modulate the response sensitivity of cancer cells to anti-cancer drugs. The RBP-transcript regulation and transcript-drug association inspired us to propose that RBPs might affect anti-cancer drug sensitivity by regulating drug-associated transcripts. Next, we performed an analysis to integrate the RBP-transcript and transcript-drug networks. Integrative analysis revealed 1,066,380 RBP-transcript-drug axes, bridging 128 RBPs and 430 anti-cancer drugs through 15,511 transcripts (Fig. 5a). The “spliceosome”-, “splicing regulation”-, and “stability & decay”-related RBPs tended to affect sensitivity to more anti-cancer drugs (Supplementary Fig. 11b, c). We next examined the transcripts that were regulated by one specific RBP. Among these transcripts, KIAA1522-a6 had the second largest number of associated anti-cancer drugs, and showed large expression change upon PTBP1 knockdown (Supplementary Fig. 12a). PTBP1 has been demonstrated to be extensively involved in the regulation of alternative splicing. Knockdown of PTBP1 significantly induced upregulation of 175 and downregulation of 552 transcripts in cancer cells (Fig. 5b). Among them, the KIAA1522-a6 transcript from the cancer-related protein-coding gene KIAA1522 had the most connections with anti-cancer drugs, whose higher expression was significantly associated (P = 2.6E−6) with a lower response sensitivity to decitabine (Fig. 5c). In the KD-RNA-seq data, a notable decrease of KIAA1522-a6 was observed upon PTBP1 knockdown (Supplementary Fig. 12b). The eCLIP binding signals were also observed in multiple exons of the KIAA1522-a6 transcript. Two specific siRNAs targeting PTBP1 were designed, siPTBP1-1 and siPTBP1-2, and both showed efficient knockdown of PTBP1 in the A2780 and Huh7 cell lines (Supplementary Fig. 13). Compared to that with only PTBP1 knockdown or decitabine treatment, cell viability notably decreased upon the combination of PTBP1 knockdown and decitabine treatment (Fig. 5d, e). Furthermore, cancer cells died much more quickly with the knockdown of PTBP1 and combinational treatments of decitabine and carboplatin or navitoclax. These results indicated that PTBP1 knockdown could promote the sensitivity of cancer cells to decitabine and combination treatment with decitabine and carboplatin or navitoclax. To further validate that PTBP1 might impact decitabine sensitivity through the KIAA1522-a6 transcript, we first examined the expression levels of KIAA1522-a6, which showed significant downregulation upon PTBP1 knockdown (Fig. 5f). The transcriptional activity of KIAA1522-a6 markedly decreased upon decitabine treatment (Fig. 5g) and combination treatment with decitabine and carboplatin (Fig. 5h) or navitoclax (Fig. 5i). The PTBP1-KIAA1522-a6-decitabine axes was also demonstrated with the siPTBP1-2 (Supplementary Fig. 14a–g), and in the Huh7 cell line (Supplementary Fig. 14h–n). To further explore the causality of the KIAA1522-a6-decitabine axis, we knocked down KIAA1522-a6 in cells treated with 0 µM and 2 µM decitabine. Our results showed that KIAA1522-a6 knockdown significantly increased the sensitivity to decitabine in the A2780 (Fig. 5j) and Huh7 (Fig. 5k) cell lines. Our integrative analysis uncovered RBP-transcript-drug axes that might provide potential treatment strategies in cancer. To promote the exploration of molecular troves across over 1000 human cancer cell lines at transcript-level resolution, we developed a comprehensive and interactive web resource, the Transcript Atlas in Cancer (TAiC, http://www.shenglilabs.com/TAiC/). In this data portal, we provide four interactive modules, including the expression landscape of transcripts, coding potential of unannotated and non-coding transcripts, RBP-mediated transcript expression, and pharmacological and clinical relevance (Fig. 6a). Users can query and visualize the expression level of individual transcripts in multiple cancer cell lines (Fig. 6b). TAiC enables users to examine the RNA sequences and coding potential of unannotated RNAs. TAiC also provides an RBP-transcript regulatory network for users to explore the upstream regulating factors of transcript expression across cancer cell lines. Users can also investigate the associations between transcript expression and anti-cancer drug sensitivity. In addition, users can also explore the expression changes and clinical relevance of transcripts in 33 different cancer types. TAiC will be continuously updated to serve as an instructive resource for researchers to investigate cancer at the transcript level. With the rapid development of high-throughput RNA sequencing techniques and computational algorithms, transcriptome diversity has been gradually realized in complex human diseases, including cancers. By employing reference-based transcript assembly in more than 1000 cancer cell lines, we presented a comprehensive human cancer transcriptome. Due to the limitation of available large-scale RNA-seq data, the majority of non-polyadenylated RNA transcripts were not detected in our study. However, our results covered major RNA biotypes, which also largely complemented other efforts aiming to annotate human transcripts. Lorenzi et al. presented a comprehensive atlas of the human transcriptome through transcriptome assembly of 300 human tissues and cell lines. Our study is quite different from this study. Firstly, our study is more focused on cancer transcriptome. This study only included 89 cancer cell lines, while we examined 1017 cancer cell lines and investigated the clinical significance of transcripts in cancer. Secondly, this study focused on non-coding RNAs, whereas we examined all detected transcripts, including those from protein-coding gene, lncRNAs, and other non-coding regions. Thirdly, we also built the RBP-transcript regulatory network and the associations between transcripts and the response of cancer cells to anti-tumor drugs. Other studies of large-scale transcriptome assembly used other RNA-seq datasets and focused on different aspects of transcriptome. Jiang et al. and Iyer et al. presented expanded landscapes of lncRNAs from thousands of RNA-seq data of tumor, normal, and cell line samples. Based on transcriptome assembly from RNA-seq datasets of 768 patient samples in TCGA datasets, Attig et al. uncovered a large number of long-terminal repeat (LTR)-overlapping transcripts. Pertea et al. constructed an expanded human gene catalog from deep RNA-seq libraries of nearly 10 thousand normal samples. In the present study, approximately 50% of the detected transcripts were unannotated. This may be due to the reference-based transcriptome assembly strategy used in this study. In addition to existing junctions, our analysis identified many unannotated junctions in both known genes and uncharacterized genomic regions. Of note, approximately one-third of unannotated transcripts spanned more than one gene, named readthrough transcripts. Readthrough transcripts are RNA molecules that are generated through the splicing of exons from multiple distinct genes, which is quite common in human transcriptome. Most previous studies have focused on gene levels that were combinations of all transcripts from the same genes, which only quantified reads falling within genomic regions of reference genes. Our previous studies based on RNA-seq data also identified many unannotated transcripts that were further validated to play important roles in cancer. Lorenzi et al. reported thousands of uncharacterized non-coding RNAs from RNA-seq data. These studies demonstrated that a considerable part of the human transcriptome remains uncharacterized. The uncharacterized transcriptome has been concealed, at least in part, by reference-based transcriptome quantification. Together with these reports, our study provides an important complement to the existing human transcriptome. In the RBP-transcript network, 62.43% of RBP-regulated transcripts were unannotated (Supplementary Fig. 15a). The unannotated transcripts occupied over half of regulated transcripts by each RBP (Supplementary Fig. 15b). In the RBP-transcript-drug axes, 53.78% of all transcripts that bridged RBPs and drugs were unannotated (Supplementary Fig. 15c). Furthermore, the unannotated transcripts linked 75.27% RBP-drug connections together with annotated transcripts, and 15.97% of the RBP-drug connections was linked by only unannotated transcripts (Supplementary Fig. 15d). These results showed that unannotated transcripts contributed appropriately half of the links to the RBP and drug networks. Various strategies of selecting unannotated transcripts for validation can be applied to choose transcripts of interest, such as expression levels, specific cell types, clinical significance, specific gene types, or coding potential. We constructed the TAiC data portal to serve the research community to explore potential functions of these unannotated transcripts. A part of the unannotated transcripts might be incomplete or even nonexistent due to the limitations of short-read RNA sequencing and the transcript assembly algorithm, and cannot be experimentally validated. However, to our best knowledge, our study made nonnegligible contribution to the transcript-level exploration of cancer transcriptome. The integration of RBP-transcript regulation and transcript-drug association networks enables the identification of RBPs that could affect the sensitivity to anti-cancer drugs by regulating transcript expression. Our analysis linked RBPs to anti-cancer drugs through transcripts. Several RBPs have been demonstrated to mediate drug sensitivity in cancer, such as ERα in breast cancer, CELF2 in ovarian cancer, and hnRNPA0 in p53-mutant tumors. These studies showed that RBPs could be used to modulate the response sensitivity to anti-cancer drugs of cancer cells. Our study provided a resource documenting thousands of RBP-transcript-drug axes, which is expected to offer alternative strategies to modulate drug resistance in cancer. Third-generation RNA-seq technologies have shown great power to capture and characterize full-length RNA transcripts, such as Nanopore and PacBio long-read RNA sequencing. These long-read RNA technologies could correct a major bias in next-generation RNA-seq data, wherein fragmented sequencing reads were computationally mapped and assembled to refer to original RNA transcripts. With the rapid development of long-read RNA sequencing technology, we believe that more and more unannotated transcripts identified in our study will be functionally validated. The raw RNA-seq data of 1017 cancer cell lines were downloaded from the Sequence Read Archive (SRA, https://www.ncbi.nlm.nih.gov/sra) database with the accession number SRP186687 by utilizing the prefetch tool (version 2.10.8) in the SRA Toolkit (http://ncbi.github.io/sra-tools/). Then, FASTQ files of raw RNA-seq reads were extracted from the SRA by using the fasterq-dump tool (version 2.10.8). Detailed information of these cancer cell lines was supplied in Supplementary Data 1. Raw RNA-seq reads were aligned to the human reference genome (GRCh38, https://www.gencodegenes.org/human/release_35.html) by using STAR software (version 2.7.6a). To achieve the most sensitive unannotated junction discovery, STAR was run in the 2-pass mode, which allowed more mapping of splicing reads to unannotated junctions. In particular, STAR was run with usual parameters in the first-pass run, wherein the junctions were collected. All detected junctions were subjected to second-pass mapping. The alignments obtained from STAR 2-pass mapping were provided as input to StringTie (version 2.1.4) for reference-based transcript assembly. Transcript annotation from GENCODE version 35 (https://www.gencodegenes.org/human/release_35.html) was adopted as the transcript model reference to guide the assembly process with the “-G” option. Transcript assembly was performed separately for each cell line. Then, all transcript assemblies were merged to generate a nonredundant master set of transcripts for all cell lines by using StringTie “–merge” mode. The GffCompare tool (version 0.12.2) was employed to compare newly assembled transcripts with those annotated in various databases/datasets, including ENCODE, UCSC known genes, RefSeq genes, AceView, CHESS, RefLnc, and LTRs assembled by Attig et al.. The transcripts that were not matched in these databases/datasets were defined as unannotated transcripts in the following analysis. StringTie quantification was utilized to reveal both transcript- and gene-level expression for individual cell lines. Expression levels were normalized in TPM (transcripts per million mapped reads). Transcripts with expression levels ≥0.1 TPM in at least one cell line were retained for subsequent analysis. We used different criteria of transcript expression levels (0.1, 0.5, 1, 2, and 3) and cell line numbers (1, 2, 5, 10, and 20). As expected, the number of unannotated transcripts decreased more with the increasing criteria of cell line numbers (Supplementary Fig. 16). The threshold of TPM 0.1 was shown to be a robust and sensitive expression detection threshold for lowly-expressed transcripts, which has been used in many previous studies. Transcripts were named by using their respective gene names followed by an “a” or “u” with numbers for annotated or unannotated transcripts, respectively (Supplementary Data 2). The StringTie names were used for those transcripts that overlapped with no known genes. The available raw long-read RNA-seq data of cancer cell lines were downloaded from the ENCODE data portal, including the A673, CACO2, CALU3, HCT116, HepG2, K562, MCF7, PANC1, PC3, and PC9 cell lines. The raw long-read RNA-seq data were processed by utilizing the FLAIR pipeline (version 1.5). In particular, reads were aligned to the human reference genome (GRCh38) by using minimap2 (version 2.17-r941) in spliced alignment mode with default parameters. High-confidence isoforms were then collapsed by the collapse function implemented in FLAIR. The quantify function was employed to quantify transcripts, wherein only reads with alignment scores >1 were used. The Cap Analysis Gene Expression (CAGE) sequencing results were downloaded from the FANTOM project. The chromatin states of human genome were obtained from the Roadmap Epigenomics project. Chromatin states indicative for active transcription, including transcription, active transcription start site, transcribed and regulatory, bivalent_promoter, transcribed and enhancer, and active enhancer, were used for the integrative analysis with CAGE data. The presence of CAGE sequencing or chromatin state peaks within 500 nt around the transcription start site (TSS) was considered as transcription evidence of corresponding transcripts. The coding potential of each unannotated transcripts was evaluated by using the CPC2 (version 1.01) and CPAT (version 3.0.4) software. CPC2 predicted the coding potential of RNA sequences by a SVM model trained from four intrinsic features, including Fickett TESTCODE score, open reading frame (ORF) length, ORF integrity, and isoelectric point (pI). CPAT uses four sequence features to distinguish coding and non-coding RNA transcripts, including open reading frame (ORF) size, ORF coverage, Fickett TESTCODE statistic, and hexamer usage bias. To estimate the transcriptional activities of hallmark biological processes, gene lists of 50 hallmarks were retrieved from the MSigDB database. For each hallmark, the Gene Set Variation Analysis (GSVA) algorithm (version 1.38.2) was employed evaluate the overall transcriptional activity in single sample mode. The GSVA method estimates variation of activities of individual gene sets over a sample population in an unsupervised manner. Each sample was endowed 50 activity scores of different hallmarks. We performed the RACE analyses to determine full length of the CRIM1-DT-u1 (CRIM1-DT-u1-3′GSP: GGGGCCAGATTGGAGTTCGA), AC107032.2-u1 (AC092803.3-u1-3′GSP: AGGGAAGAGCACTTTGGTCA), and AC092803.3-u1 (AC107032.2-u1-3′GSP: CCTGGTCTGGTCAGGGCTCAGTTAG) transcript by using a SMARTer™ RACE cDNA Amplification Kit (Clontech, California, USA) according to the manufacturer’s instructions. We retrieved paired eCLIP-seq and KD-RNA-seq (knockdown followed by RNA sequencing) datasets from the Encyclopedia of DNA Elements project (ENCODE, https://www.encodeproject.org/), covering 85 and 107 different RBPs in HepG2 and K562 cell lines, respectively. All raw sequencing reads were first subjected to Trimmomatic (version 0.39) to remove adapters and low-quality bases. For eCLIP-seq data, we followed the ENCODE processing pipeline to obtain enriched binding regions for each RBP. Trimmed KD-RNA-seq reads were mapped to the human reference genome (GRCh38) by using STAR (version 2.7.6a) in two-pass mode. The alignments were then subjected to StringTie (version 2.1.4) with our assembled transcriptome to quantify transcript abundance. Finally, DESeq2 (version 1.30.0) was applied to compare transcript differences between RBP-knockdown and control cell lines for each RBP. To obtain lineage-specific transcripts, a specificity score was calculated for each transcript according to a previous study. In particular, the specificity score was equal to the logarithm of the lineage number minus the Shannon entropy of transcript expression. The calculation was conducted as follows:where St represents the specificity score of transcript t, N is the total number of cell lineages, and pit indicates the expression ratio of transcript t in lineage i. One specificity score and N expression ratio were assigned to each transcript. The expression ratio of each transcript across all lineages was calculated as follows:where pit is the expression ratio of transcript t in lineage i, N indicates the total number of lineages, and xit represents the expression value of transcript t in lineage i. When the largest expression ratio was more than two times the second largest expression ratio and the specificity score was larger than 1, the transcript was defined as a lineage-specific transcript in the lineage with the largest expression ratio. The expression specificity of RBPs across cancer cell lines, cancer tissues, and normal tissues was also calculated as described above. The KD-RNA-seq and eCLIP-seq data for individual RBPs were subjected to integrative analysis to identify high-confidence RBP-transcript regulatory relations. For each RBP, transcripts with |fold change| > 1.5 and FDR < 0.05 were considered significantly changed upon RBP knockdown. High-confidence RBP-transcript regulatory relationships were established when RBP binding signals were found in these significantly changed transcripts. The gene dependency scores of 17,386 genes across 1086 cancer cell lines were downloaded from the DepMap data portal (https://depmap.org/portal/), which were determined by using high-throughput CRISPR screening. In this dataset, genes with a dependency score < −1 are considered as essential genes in the corresponding cell lines. We extracted the RBPs and cell lines involved in our study, generating a dependency score matrix of 130 RBPs and 671 cancer cell lines. The aligned RNA-seq reads of 33 TCGA cancer types were downloaded from the Genomic Data Commons data portal (GDC, https://portal.gdc.cancer.gov/) with official authorization, including 10,358 samples across 33 cancer types. All alignments were subjected to StringTie (version 2.1.4) with customized transcript annotation to quantify transcript abundance. Normalized expression matrixes were employed to perform differential expression analysis by adopting paired Student’s t test (as implemented in R software). Only cancer types with no <5 paired tumor and adjacent non-tumor samples were involved in this differential expression analysis. Transcripts that were expressed ≥0.1 TPM in no less than 25% of tumor or adjacent non-tumor samples in each cancer type were kept for downstream analysis. Clinical follow-up information (days to last follow-up and vital status) of tumor patients was retrieved from GDC data portal (https://portal.gdc.cancer.gov/). For each transcript in each tumor type, tumor patients were divided into high- and low-expression groups by using the median expression level of the transcript. Then the overall survival time was compared by using log-rank test implemented in the survival package (version 3.4-0, https://CRAN.R-project.org/package=survival). The survival curves were generated by using the Kaplan-Meier method in the survminer package (version 0.4.9, https://CRAN.R-project.org/package=survminer). The strategy that combines elastic net regression and bootstrapping was used to evaluate the associations between transcript expression and drug sensitivity as described in a previous study. To feed elastic net regression with a stable transcript expression profile, transcripts that were expressed (TPM > 0.1) in less than 20% of the cell lines were filtered out. The sensitivity (AUC values) to anti-cancer drugs was retrieved from the CTRP database (https://portals.broadinstitute.org/ctrp.v2.1/), which included 481 compounds across 887 cancer cell lines. Then correlations of all possible transcript-drug pairs were calculated by Spearman correlation. Transcript-drug pairs with Spearman R > 0.2 and FDR < 0.05 were used to build a prediction matrix, n × t, wherein n is the number of cancer cell lines and t is the number of transcripts. The prediction matrix was first normalized for each transcript to have zero mean and unit standard deviation. The normalized prediction matrix was then fitted for the elastic net regression by utilized the glment R package (version 4.1). To minimize the root mean squared error, the caret package (version 6.0–86) was employed to optimize the α and λ parameters. In particular, 10-fold cross-validation was performed 5 times with 25 possible α and λ values in random search mode. To obtain reliable transcript-drug pairs, the bootstrapping procedure was performed with the optimal α and λ parameters to produce 1000 resampled datasets by sampling with replacement. For each resampled dataset, a list of regression coefficients (β) was generated, which was used to calculate a prediction score as follows:where F[β > 0] represents the frequency of transcripts with positive coefficients in bootstrap datasets, and F[β < 0] represents the frequency of transcripts with negative coefficients in bootstrap datasets. Transcripts with a prediction score ≥0.7 were considered significantly predictive of the sensitivity of specific drugs. A2780 and Huh7 cells were maintained in DMEM medium supplemented with 10% fetal bovine serum, 100 mg/ml penicillin, and 100 U/ml streptomycin. A2780 cell line was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). Huh7 cell line was purchased from the Shanghai Cell Bank Type Culture Collection (Shanghai, Chinese Academy of Sciences, China). SiRNAs and negative control siRNAs were designed and synthesized by RiboBio (RiboBio Biotechnology, Guangzhou, China). Specific RNAi sequences are as follows (5′-3′): siPTBP1-1, CAAAGCCUCUUUAUUCUUU; siPTBP1-2, CUUCCAUCAUUCCAGAGAA; siKIAA1522-a6-1, ACUCACACCACAAGAGGAAG; siKIAA1522-a6-2, GUCCCCGGGUCCGCAGCUUC. Cells were transfected with the siRNAs using Oligofectamine transfection reagent (RNAi MAX, Invitrogen) according to the manufacturer’s instructions. The cells were harvested 48 h after transfection for further analysis. Cell viability was determined by CCK8 assay. Briefly, A2780 and Huh7 cells (5 × 103 cells/well) were seeded into 96-well plates. After 24 h of culture, the cells were treated with carboplatin, decitabine, and navitoclax (MCE, Shanghai, China) at the indicated concentrations for another 24 h. CCK8 solution (10 μl) was added to each well, and the cells were further incubated at 37 °C for 3 h. The absorbance of each well was measured at 450 nm with a spectrophotometer. A2780 and Huh7 cells transfected with siRNAs were seeded into a 12-well plate and incubated with complete medium at 37 °C for 24 h. Then, the cells were treated with different concentrations of carboplatin, decitabine, and navitoclax (MCE, Shanghai, China) for another 10 days. The cells were fixed with 4% paraformaldehyde and stained with 2% crystal violet. Images were obtained, and the number of colonies was counted. Different concentrations of carboplatin, decitabine, and navitoclax (MCE, Shanghai, China) were diluted in dimethyl sulfoxide (DMSO) (Sigma-Aldrich) or PBS. Total RNA was isolated from cells by using TRIzol Reagent (Thermo Fisher Scientific, Massachusetts, USA). Then, the extracted RNAs were reverse transcribed into cDNA by using a Superscript II reverse transcription kit (Takara Bio, Beijing, China) according to the manufacturer’s protocols. Subsequently, qRT-PCR was conducted with a SYBR-Green master kit (Vazyme, Nanjing, China) on a LightCycler 480 II (Roche Diagnostics) instrument according to the manufacturer’s protocols. The primers used to amplify PTBP1 (PTBP1_q_F1: CTCCAAGTTCGGCACAGTGTTG; PTBP1_q_R1: CAGGCGTTGTAGATGTTCTGCC), KIAA1522-a6 (KIAA1522-a6_q_F1: ACTCACACCACAAGAGGAAG; KIAA1522-a6_q_R1: TTTGTCATTCTCAGCCTTGG), and β-actin (β-actin-F: TTGTTACAGGAAGTCCCTTGCC; β-actin-R: ATGCTATCACCTCCCCTGTGTG) were chemically synthesized by TSINGKE (TSINGKE, Beijing, China). All qRT‑PCRs were performed in triplicate. Proteins were subject to SDS-PAGE and transferred to the nitrocellulose membranes (GE, CT, USA). After being blocked by non-fat milk, the membrane was incubated with PTBP1 polyclonal antibody (Proteintech, cat#12582-1-AP) and GAPDH monoclonal antibody (Proteintech, cat#60004-1-Ig). The band density was analyzed using ImageJ and compared with the internal control. The TAiC database was built with the Python FLASK_REST API (https://flask-restful.readthedocs.io/) as a backend web framework. In the TAiC database, MongoDB (https://www.mongodb.com) was adopted for data deposition and management. Angular (https://angular.io/) was utilized to develop web interfaces of TAiC. The frontend framework was constructed by using Bootstrap (https://getbootstrap.com). Data visualization was carried out by Echarts (https://echarts.apache.org/). The TAiC online database was tested and found to be supported in popular web browsers, including Microsoft Edge, Google Chrome, Firefox, and Safari. The TAiC database is publicly accessible at http://www.shenglilabs.com/TAiC/. Statistical analysis and data visualization in this study were performed by using R software (R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org). Unless otherwise specified, all tests were two-tailed, and a P or FDR value <0.05 was considered to indicate statistical significance. All experiments were repeated independently three times. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary figures Description to Additional Supplementary Information Dataset 1 Dataset 2 Dataset 3 Dataset 4 Dataset 5 Reporting Summary

PMC9649691

Thomas Fleming,Yukiko Kikuchi,Mikoto Nakajo,Masaya Tachizawa,Tomoaki Inazumi,Soken Tsuchiya,Yukihiko Sugimoto,Daisuke Saito,Mikita Suyama,Yasuyuki Ohkawa,Takashi Baba,Ken-ichirou Morohashi,Kataaki Okubo

Prostaglandin E2 receptor Ptger4b regulates female-specific peptidergic neurons and female sexual receptivity in medaka

10-11-2022

Sexual behaviour,Animal behaviour,Sexual dimorphism

In vertebrates, female receptivity to male courtship is highly dependent on ovarian secretion of estrogens and prostaglandins. We recently identified female-specific neurons in the medaka (Oryzias latipes) preoptic area that express Npba, a neuropeptide mediating female sexual receptivity, in response to ovarian estrogens. Here we show by transcriptomic analysis that these neurons express a multitude of neuropeptides, in addition to Npba, in an ovarian-dependent manner, and we thus termed them female-specific, sex steroid-responsive peptidergic (FeSP) neurons. Our results further revealed that FeSP neurons express a prostaglandin E2 receptor gene, ptger4b, in an ovarian estrogen-dependent manner. Behavioral and physiological examination of ptger4b-deficient female medaka found that they exhibit increased sexual receptivity while retaining normal ovarian function and that their FeSP neurons have reduced firing activity and impaired neuropeptide release. Collectively, this work provides evidence that prostaglandin E2/Ptger4b signaling mediates the estrogenic regulation of FeSP neuron activity and female sexual receptivity.

Prostaglandin E2 receptor Ptger4b regulates female-specific peptidergic neurons and female sexual receptivity in medaka In vertebrates, female receptivity to male courtship is highly dependent on ovarian secretion of estrogens and prostaglandins. We recently identified female-specific neurons in the medaka (Oryzias latipes) preoptic area that express Npba, a neuropeptide mediating female sexual receptivity, in response to ovarian estrogens. Here we show by transcriptomic analysis that these neurons express a multitude of neuropeptides, in addition to Npba, in an ovarian-dependent manner, and we thus termed them female-specific, sex steroid-responsive peptidergic (FeSP) neurons. Our results further revealed that FeSP neurons express a prostaglandin E2 receptor gene, ptger4b, in an ovarian estrogen-dependent manner. Behavioral and physiological examination of ptger4b-deficient female medaka found that they exhibit increased sexual receptivity while retaining normal ovarian function and that their FeSP neurons have reduced firing activity and impaired neuropeptide release. Collectively, this work provides evidence that prostaglandin E2/Ptger4b signaling mediates the estrogenic regulation of FeSP neuron activity and female sexual receptivity. In many species, females are sexually receptive to males only during a period surrounding ovulation to ensure reproductive success. In vertebrates, this is accomplished by the proper control of sexual receptivity by ovarian secretion of estrogens, progestins, and prostaglandins, in addition to appropriate sensory cues from males. Studies in rodents have revealed that these ovarian hormones act on a hypothalamic-limbic circuit, comprising the arcuate, medial preoptic, and ventromedial nuclei of the hypothalamus and the medial amygdala, to facilitate sexual receptivity to courting males. Because these nuclei are phylogenetically ancient and somewhat molecularly conserved, it is assumed that this behaviorally relevant circuit is shared among vertebrates. However, little information is available on the neural basis of female sexual receptivity in non-rodent species, and moreover, accumulating evidence highlights large variations in the hormonal regulation of vertebrate mating behavior across taxa. For example, estrogens have stimulatory effects on both male- and female-typical mating behaviors in rodents but prevent the execution of male-typical mating behavior in quail and a teleost species, medaka (Oryzias latipes). In several other teleost species, including goldfish (Carassius auratus) and African cichlid (Astatotilapia burtoni), gonadal estrogens are not essential for female sexual receptivity, and instead, prostaglandin F2α (PGF2α) facilitates female receptivity. Although it is not known whether other prostaglandin species, such as prostaglandin E2 (PGE2), also centrally regulate female receptivity in teleosts, PGE2 has been shown to have facilitative effects on female receptivity in rats and hamsters but have inhibitory effects in guinea pigs and anole lizards (Anolis carolinensis). These lines of evidence suggest that there may be a degree of underlying variation—at either the structural or chemical level—in behaviorally relevant circuits and the action of hormonal mediators therein across species. In medaka, we have recently identified a group of estrogen-dependent neurons relevant to female sexual receptivity, which have not been found in rodents or other species. These neurons are present only in females in the preoptic nucleus PMm/PMg (magnocellular/gigantocellular portion of the magnocellular preoptic nucleus), which is considered homologous to the paraventricular nucleus (PVN) in mammals. Ovarian estrogens regulate the number and size of these neurons, as well as their expression of a gene encoding neuropeptide B (npba). Female medaka deficient for npba and its receptor gene, npbwr2, exhibit abnormal sexual receptivity. Moreover, females deficient for esr2b, a subtype of estrogen receptor expressed in these neurons, show a marked decrease in the number and size of these neurons and a complete loss of sexual receptivity. These findings indicate that npba-expressing neurons in the PMm/PMg represent a major site of action for estrogens on the neuronal circuitry governing female receptivity in medaka. In the present study, we defined the ovarian secretion-dependent transcriptome of female-specific npba-expressing neurons in the PMm/PMg, which revealed that they express multiple neuropeptides, in addition to Npba, depending on ovarian secretion. We further found that these neurons express a subtype of PGE2 receptor, Ptger4b, in an ovarian estrogen-dependent manner and that PGE2/Ptger4b signaling has an inhibitory effect on female sexual receptivity. This effect was likely due to modulation of the electrophysiological properties of these neurons by Ptger4b, which ultimately govern neuropeptide release. In order to molecularly define female-specific npba-expressing neurons in the PMm/PMg, we searched for genes that are expressed in these neurons in an ovarian-dependent manner. We isolated and purified these neurons from intact, sham-operated, and ovariectomized females of npba-GFP transgenic medaka and performed RNA sequencing (RNA-seq) (Supplementary Fig. 1a, b). The results for all annotated genes are summarized in Supplementary Data 1. Differential expression analyses using edgeR and cuffdiff (false discovery rate <0.05) identified 171 and 321 differentially expressed genes (DEGs), respectively, both between intact and ovariectomized females and between sham-operated and ovariectomized females (Supplementary Fig. 1c, d). Among these DEGs, 107 showed significant differences in both edgeR and cuffdiff. Of note, these included at least three putative neuropeptide genes (gene ID: XLOC_022840, XLOC_003385, and XLOC_024729), all of which were downregulated by ovariectomy (Fig. 1a). Further annotation revealed that XLOC_022840 is one of six cocaine- and amphetamine-regulated transcript peptide (CARTPT)-like genes identified in medaka (GenBank accession number NM_001204781), and phylogenetic analysis revealed that it encodes the medaka ortholog of zebrafish Cartpt2b (Fig. 1b). XLOC_003385 was found to be the medaka tachykinin 1 gene (tac1) (GenBank accession number AB441191), while XLOC_024729 was identical to a predicted medaka gene in the GenBank database, XM_020705530, which was identified as the medaka ortholog of zebrafish Tac4a by phylogenetic tree analysis (Fig. 1c). Double in situ hybridization confirmed the expression of these neuropeptide genes in npba-expressing neurons in the PMm/PMg. More specifically, cartpt2b and tac4a were mostly expressed in the anterior (PMm) subpopulation of these neurons, while tac1 was exclusively expressed in the posterior (PMg) subpopulation (Fig. 1d). These results demonstrate that female-specific npba-expressing neurons in the PMm/PMg express a multitude of neuropeptide genes depending on ovarian secretions. Considering this, together with the fact that these neurons occur exclusively in females and are highly dependent on gonadal sex steroids, we termed them female-specific, sex steroid-responsive peptidergic (FeSP) neurons. In addition to the aforementioned neuropeptide genes, the DEGs downregulated by ovariectomy included a putative PGE2 receptor gene, XLOC_004257 (Fig. 2a). Considering that PGE2 has been implicated in the regulation of female sexual receptivity across vertebrate phyla, but its neural mechanisms of action remain poorly understood, we selected this gene as the focus of our analysis. Further annotation revealed that XLOC_004257 represents one of the medaka PGE2 receptor 4 (EP4) genes, ptger4b (GenBank accession number NM_001308974), which was confirmed by phylogenetic tree analysis (Fig. 2b). Double in situ hybridization demonstrated that ptger4b is expressed in both the anterior (PMm) and posterior (PMg) subpopulations of FeSP neurons (Fig. 2c). We further validated the ovarian secretion-dependent regulation of ptger4b expression in FeSP neurons by real-time PCR, which was performed on FeSP neurons isolated and purified from females that were sham-operated, ovariectomized, or ovariectomized and treated with estradiol-17β (E2; the major estrogen in vertebrates, including teleosts) or 11-ketotestosterone (KT; the primary, non-aromatizable androgen in teleosts). The results showed that ovariectomy caused a significant decrease in ptger4b expression (p = 0.0381), which was recovered by E2 treatment (p < 0.0001), whereas KT had no effect (p > 0.9999) (Fig. 2d). It can thus be concluded that ovarian estrogens promote the expression of ptger4b in FeSP neurons. This finding, along with the fact that FeSP neurons express estrogen receptors, led us to test the possibility that estrogens directly regulate the transcription of ptger4b. The search for potential estrogen-responsive elements (EREs) in the medaka ptger4b locus identified two canonical bipartite ERE-like sequences in the 5′-proximal region (at positions −1076 and −714 relative to the transcription start site) (Fig. 2e). Luciferase-based transcriptional activity assays using a 5′-proximal fragment of ptger4b containing these two ERE-like sequences revealed that E2 induced a significant increase in luciferase activity in the presence of Esr2a (p = 0.0056, 0.0024, 0.0015, and 0.0024 at 10−9, 10−8, 10−7, and 10−6 M, respectively) and Esr2b (p < 0.0001 at all concentrations tested) (Fig. 2f). Although not significant, a slight inhibition was observed in the presence of Esr1 (Fig. 2f). Next, we introduced point mutations into each of the two canonical bipartite ERE-like sequences in the luciferase reporter construct and examined the resulting change in luciferase activity. Both mutations of the ERE-like sequences at positions −1076 and −714 abolished E2’s induction of luciferase activity in the presence of Esr2b (Fig. 2g). These results suggest that the stimulatory action of estrogens on ptger4b transcription is mediated, at least in part, by Esr2b and that these two EREs act cooperatively with each other. In contrast, the inhibitory and inductive effects of E2 were still observed in the presence of Esr1 and Esr2a, respectively, even with mutations in the ERE-like sequences at positions −1076 (p = 0.0561 and 0.0422, respectively) and −714 (p = 0.0365 and 0.0129, respectively) (Fig. 2g). Next, we determined the spatial and temporal patterning of PGE2/Ptger4b signaling in the male and female brain to provide further insights into its physiological properties. Examination of the distribution of ptger4b expression throughout the entire brain by in situ hybridization revealed that ptger4b was expressed in five brain nuclei: expression in the Vv (ventral nucleus of the ventral telencephalic area) and PPa (anterior parvocellular preoptic nucleus) was common to both sexes; expression in Pbl (basal lateral preoptic nucleus) was male-specific; and expression in the PPp (posterior parvocellular preoptic nucleus) and PMm/PMg, where FeSP neurons reside, was female-specific (Fig. 3a–c). Given that medaka have a 24 h reproductive cycle (where they spawn at the beginning of the light period), we assessed the diurnal fluctuations of brain ptger4b expression and brain levels of PGE2 and its metabolites by real-time PCR and liquid chromatography-tandem mass spectrometry (LC-MS/MS), respectively. In the female brain, there were no significant changes in ptger4b expression throughout the day, while in the male brain, ptger4b expression was transiently elevated at the beginning of the light period (main effect of time, p = 0.0051; main effect of sex, p = 0.0310; interaction between time and sex, p = 0.0050; p < 0.0001 for males versus females at 1.5 h after light onset (halo); p = 0.0017, 0.0384, 0.0020, and <0.0001 for 1.5 halo versus 5.5, 13.5, 17.5, and 21.5 halo, respectively, in males) (Fig. 3d). Brain levels of PGE2 were highest at 0 halo and decreased throughout the day to their lowest at 20 halo in both sexes (main effect of time, p < 0.0001; main effect of sex, p < 0.0001; interaction between time and sex, p = 0.8229; p = 0.0173 and 0.0002 for 0 versus 16 and 20 halo, respectively, in males; p = 0.0156 for 0 versus 20 halo in females) (Fig. 3e). In addition, PGE2 levels were significantly higher in males than in females at 0 halo (p = 0.0277), and a similar, though not significant, male bias was noted at other times. Brain levels of 15-keto-PGE2 (15k-PGE2) were lower than PGE2 and followed a similar pattern of fluctuation; they were highest at the beginning of the light period and decreased throughout the day (main effect of time, p < 0.0001; main effect of sex, p = 0.3877; interaction between time and sex, p = 0.0669; p = 0.0153, 0.0065, 0.0045, and 0.0015 for 0 versus 8, 12, 16, and 20 halo, respectively, in males; p = 0.0409 and 0.0159 for 4 versus 16 and 20 halo, respectively, in males; p = 0.0385, 0.0095, and 0.0200 for 0 versus 8, 12, and 20 halo, respectively, in females) (Fig. 3f). No significant difference in 15k-PGE2 levels was found between sexes. 13,14-dihydro-15-keto-PGE2 (dhk-PGE2) levels did not show clear temporal changes, although a decrease was seen in males at 20 halo (p = 0.0059 and 0.0267 versus 0 and 16 halo, respectively) (Fig. 3g). Taken together, these results indicate that, in the female brain, ptger4b expression is fairly constant throughout the day, and instead, PGE2 levels show diurnal fluctuations, peaking rapidly at the beginning of the light period when medaka spawn. To examine the role of ptger4b in mating behavior, we generated ptger4b knockout medaka by using clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9)-mediated genome editing. Two independent knockout lines (Δ17 and Δ10) were generated for use in this study to ensure the validity of our results and control for possible off-target effects (Fig. 4a). Each line contained a deleterious frameshift mutation, resulting in the early truncation of the Ptger4b protein (Fig. 4b). Previous evidence suggests that Ptger4b plays a critical role in ovulation in medaka: ptger4b expression is highly upregulated in preovulatory follicles, and ovulation can be blocked by an EP4 antagonist. We therefore first examined the effect of ptger4b deficiency on ovarian development and function. We found no difference in adult body weight or gonad/body weight ratio between genotypes in either Δ17 or Δ10 lines (Fig. 4c, d; Supplementary Fig. 2a, b). There was also no observable difference in gross ovarian morphology between genotypes in both lines (Fig. 4e; Supplementary Fig. 2c). In addition, all genotypes had similar numbers of ovulated oocytes and hatching rate of fertilized eggs (Fig. 4f, g; Supplementary Fig. 2d, e). We further investigated the effect of the EP4 antagonist GW 627368X on ovulation using an in vitro assay as previously reported in order to examine the possibility that other EP4 subtypes may compensate for the loss of Ptger4b and help to induce ovulation. The results showed that GW 627368X failed to inhibit ovulation not only in ptger4b−/− females, but also in ptger4b+/+ and ptger4b+/− females, in contrast to what was reported previously (Fig. 4h). Taken together, these results indicate that ptger4b deficiency does not affect ovarian development and function, contrary to expectations based on previous findings. Next, we assessed the effect of ptger4b deficiency on mating behavior. Medaka mating behavior consists of a stereotyped and quantifiable series of actions that begin with the male approaching and closely following the female. The male then performs a courtship display by swimming in a semi-circle in front of the female. If the female is receptive, the male will grasp her with his dorsal and anal fins (termed “wrapping”) and they quiver together (“quivering”) until sperm and eggs are released (“spawning”) (Fig. 5a). If the female is not receptive, she will either assume a rejection posture or swim rapidly away. Most females of both Δ17 and Δ10 lines spawned successfully, regardless of genotype (Fig. 5b; Supplementary Fig. 3a). There were no significant differences between genotypes in the number of courtship displays received and wrapping refusals by females of both lines (Fig. 5c, d; Supplementary Fig. 3b, c). However, ptger4b−/− females of the Δ17 line exhibited significantly lower latencies from the beginning of the interaction to quivering (p = 0.0204), from the first approach to quivering (p = 0.0180), and from the first courtship display to quivering (p < 0.0001) than ptger4b+/+ siblings (Fig. 5h–j). Similarly, ptger4b−/− females of the Δ10 line showed significantly lower latencies from the first approach to quivering than ptger4b+/+ (p = 0.0219) and ptger4b+/− (p = 0.0162) siblings and from the first courtship display to quivering than ptger4b+/+ siblings (p = 0.0144), indicating reproducibility between the independent lines (Supplementary Fig. 3h, i). No significant differences were found in the other latency parameters analyzed. As with females, most males of both Δ17 and Δ10 lines successfully spawned, regardless of genotype (Supplementary Figs. 4 and 5). In males, no significant differences were found between genotypes in either line for any of the behavioral parameters analyzed (Supplementary Figs. 4 and 5). Collectively, these results demonstrate that ptger4b deficiency has a female-specific effect of increasing sexual receptivity towards male courtship, most likely by shortening the process of perceiving and evaluating male courtship or the subsequent motor response. FeSP neurons are characterized by a large cell body/nucleus and a regular pattern of firing, which are highly dependent on ovarian estrogens. The estrogen-dependent nature of ptger4b described above led us to speculate that these cellular phenotypes may be mediated by ptger4b. To test this idea, we first analyzed the cell body/nuclear size of FeSP neurons in ptger4b-deficient females but found no differences by genotype in either Δ17 or Δ10 lines (Fig. 6a–c; Supplementary Fig. 6a–c). We then analyzed the firing rate of FeSP neurons in ptger4b-deficient females by patch-clamp recording. FeSP neurons of ptger4b−/− females exhibited regular firing patterns similar to ptger4b+/+ siblings, but with significantly lower average firing rates in both Δ17 (p = 0.0004) and Δ10 (p = 0.0366) lines, indicating a role for ptger4b in regulating neuronal firing (Fig. 6d, e; Supplementary Fig. 6d, e). Next, we hypothesized that the reduced firing rate of FeSP neurons may diminish neuropeptide release in ptger4b-deficient females. Immunohistochemical staining for Npba showed that its signal intensity in FeSP neurons of ptger4b−/− females was significantly higher than ptger4b+/+ siblings (p < 0.0001), indicating an intracellular accumulation of Npba (Fig. 6f, g). To exclude the possibility that this effect was transcriptional, we analyzed npba mRNA expression in the PMm/PMg of ptger4b-deficient females by in situ hybridization. There was no difference in the total area of npba mRNA signal between ptger4b+/+ and ptger4b−/− females (Fig. 6h, i). Thus, we concluded that ptger4b deficiency causes reduced firing activity and impaired neuropeptide release in FeSP neurons. In rodents, it has been shown that PGE2/EP4 signaling is critical to the establishment of male-typical mating behaviors by sensitizing medial preoptic neurons to glutamatergic sensory afferents. This led us to hypothesize that a similar mechanism may exist in medaka FeSP neurons. To test this hypothesis, we investigated the response of FeSP neurons to glutamatergic stimulation by patch-clamp recordings. We observed a significant increase in the firing rate of FeSP neurons in response to puff glutamate application (p = 0.0006) (Supplementary Fig. 7a, b). However, comparison of the magnitude of glutamatergic response of FeSP neurons between ptger4b+/+ and ptger4b−/− females revealed no significant difference by genotype (Supplementary Fig. 7b, c). These results suggest that ptger4b does not regulate the sensitivity of FeSP neurons to glutamatergic inputs. We additionally assessed the effects of PGE2 and the EP4 antagonist GW 627368X on the firing activity of FeSP neurons. FeSP neurons showed no change in firing rate in response to bath application of either PGE2 or GW 627368X (Supplementary Fig. 7d). These results suggest that the effects of PGE2/Ptger4b signaling on FeSP neurons are constitutive rather than acute. Here, we show by transcriptomic analysis that female-specific npba-expressing neurons in the medaka PMm/PMg express a multitude of neuropeptides, including cartpt2b, tac1, and tac4a, in an ovarian-dependent manner, and we thus termed them FeSP neurons. A recent single-cell transcriptomic study in the mouse preoptic area revealed that many populations of behaviorally relevant neurons co-express multiple neuropeptides. Together with this study, our findings suggest the evolutionary conservation of the multipeptidergic nature of preoptic neurons. We previously showed that npba (most likely npba expressed in FeSP neurons) plays a critical role in female sexual receptivity. It remains to be determined whether other neuropeptides expressed in FeSP neurons are also relevant to female receptivity; however, their marked ovarian dependence, as well as evidence available in other species, suggests that this may be the case. For example, administration of substance P (encoded by Tac1) into the female rat brain facilitates lordosis, a stereotyped mating posture adopted by sexually receptive female rodents, and substance P receptor-deficient female mice display a decreased preference for male sexual pheromones. In addition, knockdown of natalisin, an insect tachykinin-like peptide, suppresses mating behavior in female Drosophila. Collectively, FeSP neurons may integrate ovarian signaling and multiple different neuropeptide signaling pathways to regulate female receptivity. This idea is further corroborated by the finding that FeSP neurons express ptger4b in an ovarian estrogen-dependent manner, as PGE2 signaling has been implicated in female receptivity across vertebrate phyla. Similar induction of EP4 expression by estrogens has been reported in the smooth muscle of the ewe cervix and bovine oviduct. Additionally, in female rats, PGE2 affects the excitability of preoptic gonadotropin-releasing hormone 1 (GnRH1) neurons via EP4, depending on the presence of estrogens, suggesting that the expression of EP4 in these neurons is dependent on estrogens, as is the case in FeSP neurons. Therefore, the enhancement of PGE2/EP4 signaling by estrogens may be a conserved mechanism across species. We also found that the estrogenic induction of ptger4b expression in FeSP neurons is mediated, at least in part, by direct transcriptional activation by Esr2b (which is the main estrogen receptor expressed in FeSP neurons) and that deficiency of ptger4b alters female receptivity. Given that ptger4b-deficient females had no abnormalities in ovarian function and males did not show behavioral defects, the observed alteration in female receptivity is probably due to the loss of female-specific neural ptger4b expression (most likely in PMm/PMg FeSP neurons). Taken together, PGE2/Ptger4b signaling may act downstream of E2/Esr2b signaling in FeSP neurons to regulate female receptivity. While studies in rats have proposed that the site of action of PGE2 on female receptivity is preoptic GnRH1 neurons, our results strongly suggest that FeSP neurons are a major target of PGE2 in medaka via Ptger4b. Of note, in medaka, ptger4b expression was evident in males but not in females in the Pbl, where GnRH1 neurons reside. The varying effect of PGE2 on female receptivity between species may reflect differences in its site of action within behaviorally relevant circuits. While females deficient for esr2b were completely unreceptive to male courtship, deficiency of npba and ptger4b, which act downstream of E2/Esr2b signaling, had much milder impacts on female receptivity. A plausible explanation for this may be that E2/Esr2b signaling controls most, if not all, components of female receptivity by simultaneously regulating multiple behaviorally relevant genes, including npba and ptger4b, and that each of these genes is separately responsible for one or a few components. This assumption is compatible with the view that sexually dimorphic social behaviors are regulated in a modular manner by multiple sexually dimorphic genes acting downstream of sex steroid signaling. It should be noted that, in contrast to npba deficiency, which led to increased latency to quivering/spawning, ptger4b deficiency decreased spawning latency, suggesting that npba and ptger4b have an overall facilitatory and inhibitory effect on female receptivity, respectively. According to their multipeptidergic nature, FeSP neurons may simultaneously activate multiple signaling pathways with opposite effects to fine-tune female receptivity. Facilitatory and inhibitory signaling within single peptidergic neurons has also been demonstrated in kisspeptin/neurokinin B/dynorphin neurons in the arcuate nucleus of mammals (where kisspeptin and neurokinin B activate, whereas dynorphin inhibits, the pulsatile release of GnRH1). Although further studies are clearly needed, it seems plausible that such a regulatory system is a more general feature of peptidergic neurons than has been recognized. We further found that ptger4b deficiency reduced the firing activity of FeSP neurons, suggesting a regulatory effect of PGE2/Ptger4b on the electrophysiological properties of these neurons. Our previous data showed that the firing of FeSP neurons is dependent on ovarian estrogens and that these neurons do not show diurnal changes in firing rate. Considering that ptger4b expression in FeSP neurons requires ovarian estrogens, it is probable that the estrogen-dependent electrophysiological activity of these neurons is mediated, at least in part, by PGE2/Ptger4b signaling. Interestingly, given that the firing rates of FeSP neurons are not altered by acute treatment with PGE2/EP4 antagonist and do not correspond to the diurnal fluctuation of brain PGE2 levels, the effects of PGE2/Ptger4b signaling on the electrophysiological properties of FeSP neurons are seemingly constitutive, rather than acute and transient. These effects are in contrast to those in mammalian peptidergic neurons, where PGE2/EP4 signaling typically mediates acute changes in excitability. The question then arises as to how the firing activity of FeSP neurons affects female receptivity. We observed increased Npba immunolabeling in FeSP neurons of ptger4b-deficient females with no accompanying change in npba mRNA levels, suggesting that loss of ptger4b expression impairs Npba release in these neurons. It thus seems highly likely that increased firing by PGE2/Ptger4b signaling serves to potentiate the release of neuropeptides. Generally, the exocytosis of neuropeptides requires high-frequency burst firing; it is therefore probable that the firing activity of FeSP neurons serves an important priming function, facilitating the release of neuropeptides in response to certain excitatory cues. The observed impaired Npba release (and the consequent likely reduction in Npba signaling) in ptger4b-deficient females is seemingly inconsistent with the opposite behavioral phenotypes observed in ptger4b- and npba-deficient females. As revealed by transcriptomic analysis, however, FeSP neurons are multipeptidergic, expressing many neuropeptides in addition to Npba. Peptidergic neurons also often co-produce classical small-molecule neurotransmitters, such as glutamate and γ-aminobutyric acid (GABA). Hence, the behavioral phenotype observed in ptger4b-deficient females may be due to an overall reduction in neuropeptides, and possibly also small-molecule neurotransmitters, released from FeSP neurons and not exclusively due to reduced Npba signaling. Notably, PGE2/EP4 signaling has been shown to be critical for the induction of male-typical synaptic patterning in the developing medial preoptic neurons of rodents and the consequent manifestation of male-typical mating behavior in adulthood. This mechanism involves PGE2-induced phosphorylation and mobilization of AMPA-type glutamate receptors to the cell membrane of preoptic neurons and their resulting sensitization to glutamatergic inputs. Based on these pieces of evidence, we hypothesized that a similar system may exist in medaka FeSP neurons, involving enhanced sensitivity to glutamate by PGE2/Ptger4b signaling. However, our results showed that ptger4b deficiency did not alter the response of FeSP neurons to glutamate. Given that the effects of PGE2/EP4 signaling in medaka FeSP neurons appear to be constitutive, as described above, unlike those reported in mammals, the mode of action of PGE2/EP4 signaling may differ between medial preoptic neurons in rodents and FeSP neurons in medaka. Our study adds to the prevailing view that prostaglandins are major regulators of female receptivity in teleosts, demonstrating a role for PGE2 in addition to PGF2α. However, these two prostaglandin species differ greatly in their mode of action. PGF2α is essential for female receptivity and is derived from the ovary or oviduct, where the presence of recently ovulated eggs triggers the production and release of PGF2α, which then acts as a hormonal signal of reproductive status to the brain to induce mating behaviors. In contrast, PGE2 levels in the ovary do not show diurnal changes corresponding to reproductive state, meaning that the diurnal fluctuations of PGE2 and its metabolites in the brain—which are highest at the beginning of the light period when females spawn—are not derived from the ovary. Considering this and data from zebrafish (Danio rerio), which showed PGE2 had no effect on mating behavior when administered to the water, it is probable that the import of systemic PGE2 into the brain is limited, and that the brain is the primary source of behaviorally relevant PGE2. Thus, PGE2 most likely acts as a locally synthesized neuromodulator in the brain and inhibits female receptivity when binding to Ptger4b. It is not currently clear what the functional significance of the inhibitory nature of PGE2/Ptger4b signaling is to female mating behavior. It may be that PGE2/Ptger4b delays spawning to allow for the careful assessment of male suitability/mate choice, but further work is required to validate this idea. In summary, here we found that female-specific Npba neurons located in a preoptic nucleus homologous to the PVN are multipeptidergic, expressing a multitude of neuropeptides in an ovarian-dependent manner. We also found that PGE2/Ptger4b signaling in these neurons (termed FeSP neurons) is activated by ovarian estrogens and that PGE2/Ptger4b signaling is inhibitory to female sexual receptivity in medaka. Together, these lines of evidence suggest that FeSP neurons act as a relay point, converting ovarian estrogen signals into various inter- (neuropeptides) and intra- (PGE2/Ptger4b) neural signaling molecules which, presumably, affect female receptivity. Although neurons that correspond to FeSP neurons have not been identified in other species, both NPB and EP4 have been shown to be expressed in the rat PVN. It would therefore be worthwhile to investigate whether the findings in medaka are applicable to other vertebrates, including rodents. Given that the effects of PGE2 on female receptivity vary greatly among species with facilitative effects in many species, including rats and hamsters, and inhibitory effects in guinea pigs and anole lizards, the regulatory mechanism of female receptivity by PGE2 found in medaka may be shared only with some species. Additionally, although PGE2/EP4 signaling has been shown to be essential for the establishment of male-typical mating behaviors in rodents, no abnormalities were observed in the mating behavior of ptger4b-deficient male medaka. This suggests that the effects of PGE2/EP4 signaling on male mating behavior may also differ among species. Our findings provide a possible mechanism by which PGE2 exerts different effects on the evolutionarily conserved behavioral circuitry in different species. However, the findings in medaka alone do not fully explain the divergent effects of PGE2, and the significance of the inhibitory effects of PGE2/Ptger4b in mature females capable of spawning remains unclear. Further comparative studies on species with different reproductive strategies are needed to address these issues. Wild-type medaka of the d-rR strain and npba-GFP transgenic medaka, which express GFP under the control of regulatory regions of npba, were raised at 28 °C with a 14 h light/10 h dark photoperiod. Expression of GFP in npba-GFP transgenic fish is specifically localized in Npba-expressing neurons. Fish were fed 3 to 4 times per day with live brine shrimp and commercial pellet food (Otohime; Marubeni Nisshin Feed, Tokyo, Japan). Sexually mature, spawning adults (aged 3 to 5 months) were used in all experiments and assigned randomly to experimental groups. Fish were sampled at 1 to 3 halo in all analyses except for the diurnal measurements of brain ptger4b expression and brain levels of PGE2 and its metabolites, where fish were sampled 6 times a day at 4 h intervals. All animal procedures were performed in accordance with the guidelines of the Institutional Animal Care and Use Committee of the University of Tokyo. The committee requests the submission of an animal-use protocol only for use of mammals, birds, and reptiles, in accordance with the Fundamental Guidelines for Proper Conduct of Animal Experiment and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science and Technology of Japan (Ministry of Education, Culture, Sports, Science and Technology, Notice No. 71; June 1, 2006). Accordingly, we did not submit an animal-use protocol for this study, which used only teleost fish and thus did not require approval by the committee. The ovary was removed from each npba-GFP transgenic female under tricaine methane sulfonate anesthesia (0.02%) through a small incision in the ventrolateral abdominal wall. After removal of the ovary, the incision was sutured with nylon thread. Sham-operated females underwent the same operation except for removal of the ovary. For RNA-seq, these females were sampled 9 days after the operation. For real-time PCR analysis, ovariectomized females were immersed in water containing 100 ng ml−1 of E2, KT, or vehicle (ethanol) only for 6 days after a 3-day recovering period. Sham-operated females were treated with only vehicle and used as controls. The sex steroid concentrations used were based on previously reported serum steroid levels in medaka. FeSP neurons were isolated and purified from intact, sham-operated, and ovariectomized npba-GFP transgenic females according to the method of Abe and Oka. In brief, the brain was dissected out and placed in a Petri dish containing Leibovitz’s L-15 medium. The PMm/PMg region containing GFP-labeled FeSP neurons was microdissected with fine metal needles (MicroChisel; Eppendorf, Hamburg, Germany) under a fluorescent stereomicroscope. The dissected piece of brain was incubated in Hank’s balanced salt solution containing 10 U ml−1 papain, 0.8 mM ethylene glycol tetraacetic acid, and 20 mM glucose for 30 min at 28 °C. After transferring to Leibovitz’s L-15 medium containing 20 U ml−1 DNase I and 5% fetal bovine serum (FBS), the samples were gently triturated using approximately 470-, 300-, 220-, 150-, and 120 μm diameter fire-polished Pasteur pipettes. Dissociated cells were resuspended in Leibovitz’s L-15 medium containing 5% FBS and poured into a Petri dish. Individual GFP-labeled FeSP neurons were handpicked with a glass pipette mounted on a micromanipulator (Narishige, Tokyo, Japan) under a fluorescence stereomicroscope M165FC (Leica Microsystems, Wetzlar, Germany). RNA-seq was performed on FeSP neurons isolated and purified from intact, sham-operated, and ovariectomized females. RNA extraction, cDNA synthesis, and subsequent amplification were performed as described elsewhere. In brief, FeSP neurons (50 neurons per sample, n = 3 for each experimental group) were lysed with Buffer RLT (Qiagen, Hilgen, Germany) containing 1% β-mercaptoethanol. Total RNA was purified with Agencourt AMPure XP beads (Beckman Coulter, Brea, CA) and reverse transcribed using SuperScript III reverse transcriptase (Thermo Fisher Scientific, Waltham, MA) and the oligo(dT) primer (5′-TATAGAATTCGCGGCCGCTCGCGATAATACGACTCACTATAGGGCG(T)24-3′). The resulting first-strand cDNA was purified with Agencourt AMPure XP beads (Beckman Coulter) and treated with exonuclease I (Takara Bio, Shiga, Japan) for primer digestion. After the addition of a poly(A) tail with terminal deoxynucleotidyl transferase (Roche Diagnostics, Basel, Switzerland), second-strand cDNA was synthesized using MightyAmp DNA polymerase (Takara Bio) and the tagging primer (5′-TATAGAATTCGCGGCCGCTCGCGA(T)24-3′). The cDNA was then amplified by suppression PCR (18 cycles of 98 °C for 10 s, 65 °C for 15 s, and 68 °C for 5 min) using MightyAmp DNA polymerase (Takara Bio) and the 5′-aminated primer (5′-NH2-GTATAGAATTCGCGGCCGCTCGCGAT-3′) and purified using MinElute PCR Purification Kit (Qiagen). Sequencing libraries were prepared using NEBNext Ultra DNA Library Prep Kit for Illumina and NEBNext Multiplex Oligo for Illumina (New England Biolabs, Ipswich, MA) and sequenced with single-end 51 bp reads on a Hiseq 1500 System (Illumina, San Diego, CA). 39–61 million sequences were obtained from each sample (Supplementary Fig. 1b). Sequences were aligned to the reference medaka genome (oryLat2 assembly) using TopHat (ver. 2.0.13) and subsequently Subjunc. Cufflinks (ver. 2.2.1) was then used to assemble the transcripts and estimate their abundance. Pairwise differential expression analysis between samples was performed using the edgeR and cuffdiff (ver. 2.2.1) packages with false discovery rate <0.05. Raw mapped reads were visualized and manually inspected using Integrative Genomics Viewer (IGV). BLAST searches using the sequences obtained by RNA-seq (gene ID: XLOC_022840, XLOC_003385, XLOC_024729, and XLOC_004257) as queries were performed to identify the corresponding cDNAs in the GenBank nucleotide database, whose sequences were further verified by identification of the corresponding expressed sequence tag (EST) clones in the medaka EST database at National BioResource Project (NBRP) Medaka (http://www.shigen.nig.ac.jp/medaka/). Transmembrane domains of Ptger4b were predicted using the TMHMM server (ver. 2.0; https://services.healthtech.dtu.dk). The deduced amino acid sequences of the identified cDNAs from medaka were aligned with those of their putative orthologs/paralogs from different species by using ClustalW. The resulting alignments were used to construct bootstrapped (1000 replicates) neighbor-joining trees (http://clustalw.ddbj.nig.ac.jp/index.php). Medaka Cartpt (ch9), Drosophila tachykinin, and human PTGER3 were used as outgroups to root the trees of CARTPT, tachykinin, and PTGER4, respectively. The species names and GenBank accession numbers of the sequences used are listed in Supplementary Table 1. A 645 bp DNA fragment corresponding to nucleotides 1–645 of the medaka npba cDNA (NM_001308979) was PCR-amplified and transcribed in vitro to generate a fluorescein-labeled cRNA probe using Fluorescein RNA Labeling Mix and T7 RNA polymerase (Roche Diagnostics). DNA fragments corresponding to nucleotides 1–532 (532 bp) of the cartpt2b cDNA (GenBank accession number NM_001204781), 2–393 (392 bp) of the tac1 cDNA (DK021817), 100–814 (715 bp) of the tac4a cDNA (XM_020705530), and 556–1984 (1429 bp) of the ptger4b cDNA (NM_001308974) were PCR-amplified and transcribed to generate digoxigenin (DIG)-labeled cRNA probes using DIG RNA Labeling Mix and T7 RNA polymerase (Roche Diagnostics). The procedure for double-label in situ hybridization has been described previously. In brief, whole brains removed from females were fixed in 4% paraformaldehyde (PFA) and embedded in 5% agarose supplemented with 20% sucrose. Brains were cut into 20 µm frozen sections in the coronal plane and hybridized simultaneously with the DIG-labeled cartpt2b, tac1, tac4a, or ptger4b probe and fluorescein-labeled npba probe. The fluorescein-labeled probe was visualized using a horseradish peroxidase-conjugated anti-fluorescein antibody (RRID: AB_2737388; PerkinElmer, Waltham, MA) and the TSA Plus Fluorescein System (PerkinElmer); the DIG-labeled probes were visualized using an alkaline phosphatase-conjugated anti-DIG antibody (RRID: AB_514497; Roche Diagnostics) and Fast Red (Roche Diagnostics). Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). Fluorescent images were acquired with a confocal laser-scanning microscope (TCS SP8; Leica Microsystems), using excitation and emission wavelengths of 405 nm and 410–480 nm for DAPI, 488 nm and 495–545 nm for fluorescein, and 552 nm and 620–700 nm for Fast Red. For analysis of ptger4b expression in FeSP neurons, GFP-expressing cells in the PMm/PMg were isolated and purified from sham-operated, ovariectomized, ovariectomized plus E2-treated, and ovariectomized plus KT-treated npba-GFP females as described above. RNA extraction, reverse transcription, and cDNA amplification were performed as in RNA-seq. For analysis of diurnal fluctuations in brain ptger4b expression, whole brains were sampled from females every 4 h over a 24 h period, beginning at 1.5 halo. Total RNA was extracted from the brains using the RNeasy Lipid Tissue Mini Kit with DNase treatment (Qiagen), and cDNA was synthesized using the Omniscript RT Kit (Qiagen). Real-time PCR was performed on the LightCycler 480 System II using the LightCycler 480 SYBR Green I Master (Roche Diagnostics). Melt curve analysis was performed to verify that a single amplicon was obtained in each sample. Data from whole brain samples were normalized to the β-actin gene (actb; GenBank accession number NM_001104808). Data from FeSP neurons were normalized to the geometric mean of actb and the elongation factor 1 α gene (eef1a; NM_001104662), according to Vandesompele et al.. The primers used for real-time PCR are listed in Supplementary Table 2. A fosmid clone (golwfno354_c17) containing the medaka ptger4b locus was obtained from NBRP Medaka. The 5′-proximal region of ptger4b (4.0 kb) was analyzed for the presence of potential EREs by using Jaspar (ver. 5.0_alpha) with default settings. A 4675 bp fragment of genomic DNA containing 4422 bp of the 5′-proximal region and the entire 5′-untranslated region was PCR-amplified from the fosmid clone and inserted into the NheI site of the pGL4.10 luciferase reporter vector (Promega, Madison, WI). The resulting luciferase reporter construct was transiently transfected into CHO cells (obtained from and authenticated by Riken BRC Cell Bank) with an expression construct for either medaka Esr1, Esr2a, or Esr2b and an internal control vector pGL4.74 (Promega) using Lipofectamine LTX (Thermo Fisher Scientific). Cells were treated with E2 in phenol red-free Dulbecco’s modified Eagle’s medium supplemented with 5% charcoal/dextran-treated FBS (Cytiva, Marlborough, MA) at concentrations of 0, 10−10, 10−9, 10−8, 10−7, and 10−6 M for 18 h. The luciferase activity of cell lysates was measured on the GloMax 20/20n Luminometer (Promega) using the Dual-Luciferase Reporter Assay System (Promega). Each assay was performed in triplicate and repeated independently three times, except for the assay with Esr2a, which was repeated six times. To determine the EREs responsible for the estrogenic induction of ptger4b transcription, both half-sites of each of the two potential EREs (at positions −1076 and −714 relative to the transcription start site) were mutated into a HindIII site (AAGCTT) using the PrimeSTAR Mutagenesis Basal Kit (Takara Bio). Assays with these mutated constructs were performed as described above, except that a single dose of E2 (10−6 M) was used. Single-label in situ hybridization was performed as described previously. Briefly, whole brains were fixed in 4% PFA and embedded in paraffin. Serial coronal sections of 10 μm thickness were cut and hybridized with the above-mentioned ptger4b or npba probe, which was labeled with DIG using DIG RNA Labeling Mix and T7 RNA polymerase (Roche Diagnostics). Hybridization signals were visualized using an alkaline phosphatase-conjugated anti-DIG antibody (RRID: AB_514497; Roche Diagnostics) and 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT) substrate (Roche Diagnostics). Color development was allowed to proceed overnight (for analysis of ptger4b expression) or was stopped within 15 min to avoid saturation (for analysis of npba expression). Brain nuclei were identified using the medaka brain atlas. For quantification of npba expression in FeSP neurons, images were acquired with a virtual slide microscope (VS120; Olympus, Tokyo, Japan) and the total area of npba expression signal in the PMm/PMg was calculated using Olyvia software (Olympus). Whole brains were sampled from both sexes every 4 h over a 24 h period, beginning at 0 halo (the brains from 3 fish were pooled per sample), rapidly frozen, and stored at −80 °C until analysis. Males and females sampled at 0 halo were separated by a perforated, transparent partition to prevent spawning. At all other time points, males and females were allowed to interact freely. Brain levels of PGE2 and its metabolites, 15k-PGE2 and dhk-PGE2, were measured by LC-MS/MS essentially as described elsewhere. In brief, the brain samples were homogenized with methanol and solid phase extraction was performed using Oasis HLB extraction cartridges (Waters Corporation, Milford, MA). Chromatography was performed with a Nexera X2 high-performance liquid chromatography system (Shimadzu, Kyoto, Japan) connected to a QTRAP 5500 triple quadrupole mass spectrometer (Sciex, Framingham, MA). Analytes were separated on a Kinetex C18 reverse phase column (150 × 2.1 mm, 1.7 μm) (Phenomenex, Torrance, CA). Mobile phase A consisted of water with 0.1% formic acid, and mobile phase B of acetonitrile. The following stepwise gradient was applied: 30% (0–1 min), 80% (5–6 min), 100% (8–9.5 min), and 30% (9.51–12 min). The flow rate was set to 0.3 ml min−1. Electrospray ionization in negative ion mode and multiple reaction monitoring (MRM) were used to achieve high specificity and sensitivity for the simultaneous detection of prostaglandins. The MRM transitions of PGE2, 15k-PGE2, and dhk-PGE2 were m/z 351 > 271, m/z 349 > 235, and m/z 351 > 175, respectively. The amount of each analyte was calculated by reference to the standard curve and corrected by the percent recovery of the deuterium-labeled internal standard (PGE2-d4). Knockout medaka for ptger4b were generated by CRISPR/Cas9-mediated genome editing. A CRISPR RNA (crRNA) was designed to target the fourth predicted transmembrane domain (Fig. 4a, b). crRNA and trans-activating CRISPR RNA (tracrRNA) were synthesized by Fasmac (Kanagawa, Japan). Cas9 mRNA was synthesized by in vitro transcription of the linearized pCS2+hSpCas9 plasmid (Addgene plasmid number 51815; Addgene, Cambridge, MA) using the mMessage mMachine SP6 Kit (Thermo Fisher Scientific). crRNA, tracrRNA, and Cas9 mRNA were co-microinjected into the cytoplasm of embryos at the one-cell stage. Potential founders were screened by outcrossing with wild-type fish and testing progeny for mutations at the target site using T7 endonuclease I assay followed by direct sequencing. Two founders were selected that yielded progeny carrying deleterious frameshifts leading to premature truncation of the Ptger4b protein: the progeny of one founder carried a 19 bp deletion and 2 bp insertion (Δ17) and progeny of the other carried a 10 bp deletion (Δ10). These progeny were intercrossed to establish knockout lines (Δ17 and Δ10 lines). Each line was maintained by breeding ptger4b+/− males and females to obtain ptger4b+/+, ptger4b+/− and ptger4b−/− siblings for experimental use. The genotype of each fish was determined by direct sequencing and high-resolution melt analysis using the primers listed in Supplementary Table 2. In each subsequent experiment, siblings of different genotypes of the same knockout line raised under the same conditions were used as a comparison group to control for genetic and environmental factors. An in vitro ovulation assay was performed according to the method of Fujimori et al.. In brief, ovaries were removed from ptger4b+/+, ptger4b+/−, and ptger4b−/− females of the Δ17 knockout line at 12 halo. All large-sized ovarian follicles were isolated from each ovary, split evenly into control and treatment groups, and incubated in 90% Medium 199 with Earle’s salts (Thermo Fisher Scientific) containing 50 mg ml−1 gentamycin and either vehicle (dimethyl sulfoxide) alone or 20 µM of the EP4 antagonist GW 627368X (Cayman Chemical Company, Ann Arbor, MI), respectively. The number of oocytes that successfully ovulated was counted the following day at 5 halo. The ovulation rate was presented as the percentage of oocytes that ovulated. The mating behavior tests were performed as described elsewhere. In brief, on the day before behavioral testing, each focal male/female (from Δ17 and Δ10 knockout lines) was paired with a stimulus fish of the opposite sex (wild-type d-rR strain) in a 2-litre tank and separated by a perforated, transparent partition. The partition was removed at 1 halo and fish were allowed to interact for 10 min. All interactions were recorded with a digital video camera (iVIS HF S11/S21, Canon, Tokyo, Japan; Everio GZ-G5, Jvckenwood Corporation, Kanagawa Japan; or HC-W870M, Panasonic Corporation, Osaka, Japan). The following parameters were calculated from video recordings: percentage of females that spawned within the test period; number of courtship displays prior to spawning; number of wrapping attempts refused by the female; latency from the beginning of interaction to the first approach and courtship display by the male; latency from the first approach to the first courtship display by the male; latency from the beginning of interaction, first approach, and first courtship display to quivering. The cell body and nuclear sizes of FeSP neurons were measured as described previously. In brief, whole female brains were fixed in 4% PFA, embedded in 5% agarose supplemented with 20% sucrose, and cut into 20 µm frozen sections in the coronal plane. Sections were incubated with a rabbit anti-Npb polyclonal antibody (RRID: AB_2810229), which has been shown to recognize Npba with high specificity, and then incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (RRID: AB_2534114; Thermo Fisher Scientific). Neuronal cell bodies and nuclei were stained with NeuroTrace Fluorescent Nissl Stain (NeuroTrace 530/615; Thermo Fisher Scientific) and DAPI (Roche Diagnostics), respectively. Fluorescent images were acquired with a confocal laser-scanning microscope (TCS SP8; Leica Microsystems), using excitation and emission wavelengths of 405 nm and 410–480 nm for DAPI, 488 nm and 495–545 nm for Alexa Fluor 488, and 552 nm and 620–700 nm for NeuroTrace 530/615. The acquired images of the cell bodies and nuclei of Npb-immunoreactive neurons were converted to black and white binary images by thresholding using Adobe Photoshop (ver. 22; Adobe, San Jose, CA). The total areas of each cell body and nucleus were calculated by using ImageJ (https://imagej.nih.gov/ij/) and compared between genotypes. For patch-clamp recordings, ptger4b knockout lines were crossed with the npba-GFP transgenic line. Neuronal activities were recorded from GFP-labeled FeSP neurons in the PMm of ptger4b+/+ and ptger4b−/− females, as previously reported with some minor modifications. Briefly, whole brains removed from these females were hemisected along the midline and immersed in artificial cerebrospinal fluid (ACSF) comprising 134 mM NaCl, 2.9 mM KCl, 2.1 mM CaCl2, 1.2 mM MgCl2, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), and 15 mM glucose (pH7.4, adjusted with NaOH). FeSP neurons were approached with patch pipettes made from borosilicate glass capillaries of 1.5-mm outer diameter (GD-1.5; Narishige, Tokyo, Japan) and, when filled with ACSF, had a resistance of 5–25 MΩ under an epifluorescent microscope (Eclipse FN1; Nikon, Tokyo, Japan). Targeted loose-patch recordings were performed with an Axopatch 200B amplifier (Molecular Devices, San Jose, CA) and digitized at 10 kHz using a Digidata 1550A and pClamp software (ver. 10.6; Molecular Devices). The hemi-brain preparations were continuously perfused with ACSF throughout experimentation. Action currents were recorded from FeSP neurons in voltage-clamp mode. Glutamate was applied focally to the recorded neuron by pressure ejection (500 kPa, 500 ms) from a glass pipette filled with ACSF containing 1 mM glutamate (L-glutamic acid monosodium salt hydrate; Sigma-Aldrich, St. Louis, MO) placed 10–30 μm from the cell body. The application of glutamate was made once or twice in a single recording, with an interval of at least 5 min between applications. ACSF without glutamate was applied to the control group. For experiments examining the effects of PGE2 and EP4 antagonist, 1 μM of PGE2 or GW 627368X in ACSF was applied by perfusion; ACSF containing only vehicle (ethanol) was used for control. Data were analyzed using Clampfit software (ver. 10.7; Molecular Devices). A high-pass filter was applied to remove low-frequency background noise. Only those neurons that showed currents greater than 10 pA for a minimum of 10 min were included in the analysis. Firing frequency was calculated from 3 to 8 min segments of the recordings. For the analysis of the response to glutamate application, the change in firing rate is expressed as the ratio of the average firing rate calculated from the 10 s period immediately after the glutamate puff application over the 10 s period immediately before. Whole brains of ptger4b+/+ and ptger4b−/− females were fixed in 4% PFA, embedded in paraffin, and cut into 10 μm coronal sections. After blocking with phosphate-buffered saline (PBS) containing 2% normal goat serum, sections were incubated overnight at 4 °C with the anti-Npba antibody described above diluted 1:1000 in PBS containing 2% normal goat serum, 0.1% bovine serum albumin, and 0.02% keyhole limpet hemocyanin. The sections were then incubated overnight at 4 °C with Alexa Fluor 555-conjugated goat anti-rabbit IgG (RRID: AB_2535849; Thermo Fisher Scientific) and DAPI diluted 1:1000 in PBS. Images were captured as described above, using excitation and emission wavelengths of 405 nm and 410–480 nm for DAPI and 552 nm and 562–700 nm for Alexa Fluor 555. The immunofluorescence intensity of Npba was measured using LAS X software (ver. 3.7.4; Leica Microsystems). For continuous data, results are presented as mean ± standard error of the mean (SEM), with individual data points shown as dots except for neuronal size analysis, patch-clamp recordings, and Npba immunohistochemistry, where data are plotted as box-and-whisker plots by the Tukey method for visual clarity. Categorical data are presented as percentages. Statistical analyses were performed using GraphPad Prism (ver. 8; GraphPad Software, San Diego, CA). Continuous data were compared between two groups by the unpaired two-tailed Student’s t-test. Welch’s correction was applied if the F-test indicated a significant difference in variance between groups. Differences in continuous data between more than two groups were evaluated by one-way analysis of variance (ANOVA) followed by either Bonferroni’s (for comparisons among experimental groups) or Dunnett’s (for comparisons of experimental versus control groups) post hoc test. Homogeneity of variance was verified for all data sets by Brown–Forsythe test. Two-way ANOVA followed by Bonferroni’s post hoc test was used for analyses of diurnal fluctuations in brain ptger4b expression, brain levels of PGE2 and its metabolites, and the effect of EP4 antagonist on ovulation. Behavioral time-series data were analyzed using Kaplan–Meier plots with the inclusion of fish that did not exhibit the given behavior within the test period, following Jahn-Eimermacher et al.. Differences between Kaplan–Meier curves were tested for statistical significance using Gehan–Breslow–Wilcoxon test with Bonferroni’s correction. Statistical outliers in the behavioral data were determined with a ROUT test, using a false-positive rate (Q) of 0.1% and removed from the data sets. Fisher’s exact test was used to compare categorical data. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Peer Review File Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Reporting Summary

PMC9649695

36357504

Hamdy A. Hassan,Mousa A. Alghuthaymi

Biotechnology methods for succession of bacterial communities in polychlorinated biphenyls (PCBs) contaminated soils and isolation novel PCBs-degrading bacteria

10-11-2022

Microbiology,Molecular biology

Polychlorinated Biphenyls (PCBs) are persistence in the contaminated sites as a result of lacking PCBs-degrading microorganisms. Cultivation-independent technique called single-strand-conformation polymorphism (SSCP) based on 16SrRNA genes was chosen to characterize the diversity of bacterial communities in PCBs polluted soil samples. The bacterial communities showed an increasing diversity from the genetic profiles using SSCP technique. 51 single products were identified from the profiles using PCR reamplification and cloning. DNA sequencing of the 51 products, it showed similarities to Acidobacteria, Actinobacteria, Betaproteobateria, Gammaproteobacteria and Alphaproteobacteria, the range of similarities were 92.3 to 100%. Pure 23 isolates were identified from PCBs contaminated sites. The identified isolates belonged to genus Bacillus, Brevibacillus, Burkholderia, Pandoraea, Pseudomonas, and Rhodococcus. The new strains have the capability to use PCBs as a source of sole carbon and harbor 2,3-dihydroxybiphenyl dioxygenase (DHBDO) which could be used as molecular marker for detection PCBs-degrading bacteria in the PCBs contaminated sites. This finding may enhance the PCBs bioremediation by monitoring and characterization of the PCBs degraders using DHBDO in PCBs contaminated sites.

Biotechnology methods for succession of bacterial communities in polychlorinated biphenyls (PCBs) contaminated soils and isolation novel PCBs-degrading bacteria Polychlorinated Biphenyls (PCBs) are persistence in the contaminated sites as a result of lacking PCBs-degrading microorganisms. Cultivation-independent technique called single-strand-conformation polymorphism (SSCP) based on 16SrRNA genes was chosen to characterize the diversity of bacterial communities in PCBs polluted soil samples. The bacterial communities showed an increasing diversity from the genetic profiles using SSCP technique. 51 single products were identified from the profiles using PCR reamplification and cloning. DNA sequencing of the 51 products, it showed similarities to Acidobacteria, Actinobacteria, Betaproteobateria, Gammaproteobacteria and Alphaproteobacteria, the range of similarities were 92.3 to 100%. Pure 23 isolates were identified from PCBs contaminated sites. The identified isolates belonged to genus Bacillus, Brevibacillus, Burkholderia, Pandoraea, Pseudomonas, and Rhodococcus. The new strains have the capability to use PCBs as a source of sole carbon and harbor 2,3-dihydroxybiphenyl dioxygenase (DHBDO) which could be used as molecular marker for detection PCBs-degrading bacteria in the PCBs contaminated sites. This finding may enhance the PCBs bioremediation by monitoring and characterization of the PCBs degraders using DHBDO in PCBs contaminated sites. Polychlorinated biphenyls (PCBs) are synthetic organochlorine chemicals that had many industrial purposes in the past, but proved to be very dangerous and cause many health problems for humans, so their use was discontinued, but it still remain in the environment and living organisms. Everyone could be easy exposed by PCBs through animal fats ingestion, inhalation, or skin contact, where these compounds are fat-soluble s. Because of PCBs exposure lead to the immune system suppressed, that increasing the risk of many human diseases especially in oil-producing countries. PCBs considered as carcinomas inducers. Exposure to PCBs especially during a fetus and early in life lowers IQ and changes behavior. It also alters thyroid and reproductive functions in both males and females and increases the risk of cardiovascular, liver and diabetes diseases. PCBs remain dangerous pollutants. PCBs remediation through incineration or transportation to specialized landfills traditional methods are expensive and have a dangerous environmental impact. PCBs- bioremediation using microorganisms is the best solution to get rid of these dangerous compounds. Cultivation of pure bacterial isolates from any ecosystems samples are 0.1 to 1% from the viable bacteria, as a result of that a full description of the microbial diversity will not be obtained any cultivation method. Therefore, adapting or developing dedicated molecular methods for the culture-independent survey of microorganisms and their functional properties.is tremendous significance, this possibility takes advantage of the great advances in molecular biology in the last few decades. The biodiversity studies on bacterial communities from a variety of ecosystems’ samples could be achieved by some popular techniques including denaturing gradient gel electrophoresis (DGGE) and temperature gradient gel electrophoresis (TGGE). Ammonia-oxidizing bacteria were characterized by DGGE and TGGE techniques in composts. DGGE and TGGE techniques have not yet been used at the phylogenetic level of bacterial-community succession in contaminated soil sites, as that provided by 16SrRNA gene clone libraries. DGGE/TGGE is also a limited method for studying the microbial community either from the limited but not specific 500 bp fragments from 16SrRNA or from the intensity of the band that may not give the actual abundance for the microbial communities. Single-strand-conformation polymorphism (SSCP) as an alternative developed protocol was used for the cultivation-independent assessment of bacterial-community diversity based on 16S rRNA as a reporter to generate phylogenetic tree, that showed an essential function, ubiquity, and evolutionary properties, these characteristics gave it the priority in the microbial ecology as a molecular marker. SSCP method is superior over DGGE and TGGE, because of its ease of application and no requirement for GC clamp or construction of gradient gels. In addition, the sharp bands produced by SSCP were profound, whereas the DGGE method gave an identical pattern. SSCP method therefore makes a differentiation in the structures of bacterial community. SSCP was optimized to analyze one complementary single strands by lambda exonuclease, which is preferentially degrading one strand generated with a phosphorylated primer. This method is characterized by avoiding the formations of heteroduplex or overlapping different amplicons (forward-reverse strands) with identical size separation but different in sequence. The modified technique application was focused on the taxonomic shifts studied in different bacterial communities by targeting 16S rDNA genes however, another potential application was foreseen for assessing functional genes diversity. Extradiol dioxygenases are key enzymes that cleave the aromatic ring of different aromatic and polyaromatic hydrocarbon compounds as benzene, toluene, biphenyl and naphthalene by their capability to introduce both atoms of dioxygen into their substrates, resulting in a ring-cleavage meta to the hydroxyl groups. Those enzymes comprise catechol 2,3-dioxygenases, 2,3-dihydroxybiphenyl dioxygenases, protocatechuate 4,5- and 2,3-dioxygenases and 3,4-dihydroxyphenylacetate dioxygenases among others. Harayama and Rekik had proposed that the major type I family of extradiol dioxygenases could be divided into two subfamilies. One preferred monocyclic aromatic compounds as a substrate such as benzene, toluene, ethylbenzene, and xylene (BTEX) and the other preferred bicyclic aromatic compounds such as biphenyl compounds (usually determined as the activity with 2,3-dihydroxybiphenyl (DHB) and thus these enzymes are often referred to as 2,3-dihydroxybiphenyl 1,2-dioxygenases). In the last few years, the description of new members of this family increased significantly, due to the sequencing of PCR products amplified from the environment, to genome sequencing projects, but also due to the cloning of the genes from various isolates either Gram-positive or Gram-negative organisms. In this study we introduce SSCP as a method to characterize the diversity of bacterial communities in the contaminated soil samples with aromatic hydrocarbon and to apply this method for isolating aromatic hydrocarbon bacterial degraders as pure strains including 2, 3-dihydroxybiphenyl dioxygenases production. 16SrRNA gene sequences from bacteria in the soil samples were amplified using primers targeting different regions, and producing complex patterns from SSCP using polyacrylamide gels. The results showed an increasing in the high intensity bands number on soil (II) in comparison to soil (I) and (0). The degree of pollution in the soil coincides with specific products (bands) occurred with similar yields in profiles obtained from the three soils (II, I, 0) (Fig. 1). Comparative cluster analyses for the gel assembled for the major types of three different contaminated sites in Fig. 1, where (II) for SSCP fingerprints from the highly PCBs-contaminated soil samples, (I) for SSCP fingerprints from the slightly PCBs-contaminated soil samples, and (0) for SSCP fingerprints from clean soil (Fig. 2). In site (II) the soil samples 1and 3 with a similarity of more than 80% were clustered very tightly together, while the samples 2 and 4 with its similarities < 70% were clustered together, but the soil samples 5 and 6 clustered different than 1 and 3, 2 and 4 with a similarity of more than 80% (Fig. 2II). In site (I) the soil samples 1and 2 have similarities of about 90% and clustered together, 3 and 4 clustered very tightly together with a similarity about 80%, but the soil samples 5 and 6 clustered different far from 1 and 2, 3 and 4 with a similarity of more than 80% (Fig. 2I). In site (0), the soil samples 1and 2 clustered very tightly together with a similarity about 80%, 3 and 4 clustered very tightly together with a similarity of about 80%, but the soil samples 5 and 6 clustered more than 1 and 2, 3 and 4 with a similarity of < 80% (Fig. 20). The opposite strands were regenerated from highly PCBs- contaminated soil and reamplified using PCR. SSCP gel electrophoresis targeting 16SrRNA genes from different soil samples was used for evaluating the purity and identity of reamplified products (see Supplementary Fig. S1 online). The reamplified products matched mostly to the foreseeable positions in the community patterns without byside products observed. From these reamplified DNA products, cloning and sequencing were applied directly. A total 98 bands with different DNA single strands were extirpated for the assessment of the highly PCBs- contaminated soil samples concerning its taxonomic composition, where 51 phylotypes could be unique were detected. Sequences of the single bands were compared with 16SrRNA gene sequences data available in the GenBank data base for identification of the single members of the microflora in this highly PCBs- contaminated soil (Table 1). Several members of taxonomic groups were obtained from DNA fingerprints, such as Acidobacteria, Actinobacteria, Alphaproteobacteria, Betaproteobacteria, Gammaproteobacteria, and Actinobacteria (Table 1). A broad phylogenetic distribution for phyla like Actinobacteria, Alphaproteobacteria, and Betaproteobacteria, which showed high similarity about 96% to cultured bacteria and 90% of the identified uncultured taxa. These taxa are commonly found in soil (see Supplementary Fig. S1B online). The bacterial groups, which identified in this work were also reported by other investigators. Proteobacteria and unclassified group were reported in all of the four cases, accounting for 12.5–67.0% and 7.0–64.3% respectively and my results were within these ranges There were 18.8% and 5% Acidobacteria in the British Columbia forest soil and the Scotland heavy metal contaminated soil respectively. Spraying bacterial colonies with 2,3-dihydroxybiphenyl, bacterial colonies turned yellow this indicated that these bacterial colonies harbored extradiol dioxygenases. From the three mentioned soil samples, which were diluted and spread on R2A agar plates. The obtained colonies were analyzed for extradiol dioxygenase activity in the upper pathways of PCBs degradation by spraying with 2,3-dihydroxybiphenyl. A subset of colonies with different colony morphotypes and exhibiting yellow coloration upon spraying resulting from 2,3-dihydroxybiphenyl 1,2-dioxygenase activity could be by colour isolated. Whereas there was no significant difference in the number of colony forming units from the differently contaminated soils tested (approximately 5 ± 3 × 106 CFU/g of soil), Different percentages from colonies exhibiting meta-cleavage activity as a result of the presence of 2,3-dihydroxybiphenyl 1,2-dioxygenase activity in the obtained and predicted PCBs- degrading bacteria. The percentage of the obtained varied in range from < 4%, 12% and > 50% of total CFU/g growing on R2A plates obtained from 0, I and II soil samples, respectively (Fig. 3) i.e. there are correlation between presence of extradiol dioxygenase activity and the capability to grow on the PCBs – contaminated soils (Fig. 3). As there were a correlation between presence of Biphenyl 2,3-dioxygenase activity and the presence of PCBs degraders. All colonies were analyzed for the Biphenyl 2,3-dioxygenase activity by spraying 10 mM from 2,3-dihydroxybiphenyl (DHB). The colonies exhibiting yellow coloration after spraying this mean that exhibited extradiol dioxygenase activity. 23 PCBs bacterial degraders were isolated and characterized by partially 16SrRNA sequence. The strains were belonged to the genus Bacillus, Brevibacillus, Burkholderia, Pandoraea, Pseudomonas, and Rhodococcus. The Pseudomonas strains from HA-OP3- HAOP10 grouped into one cluster as a result of their highest similarities with each other (Fig. 4), while the Pseudomonas strains from HAOP15 and HAOP22 were highly similar with the previously reported as chlorobenzene degrader P. veronii UFZ-B54 and hydrocarbon degrader P. sp. BZ27. The remained P. sp. HAOP1 clustered with anthracene degrader P. aeruginosa W3 (Fig. 4) Brevibacillus sp. HAOP26 has high similarity with polyaromatic hydrocarbon degrader B. brevis BEA. Rhodococcus sp. HAOP25 was 97% similar to the phenol degrader Rhodococcus sp. CH9 and R. sp. HAOP30 has high similarity about 97% with R. erythropolis, which has been reported as dibenzofuran degrader. Burkholderia sp. HAOP24 and Pandoraea sp. HPOP28 showed similarity with a polyaromatic hydrocarbon and dibenzothiophene degrader Burkholderia fungorum DBT1. The growth of the isolates on different PCBs as monochlorinated biphenyl 2-Chlorobiphenyl (2-CB) and 4-Chloobiphenyl (4-CB) , dichlorobiphenyl 2,3 Chlorobiphenyl (2,3CB) and 2,4 Chlorobiphenyl (2,4CB) and polychlorinated biphenyl 2,4,5,2′,4′,5′-Chlorobiphenyl as sole carbon sources . All the strains can use monochlorinated biphenyl 2CB and 4CB quite rapidly. Furthermore, the growth were obtained on 2,4,5,2′,4′,5′-Chlorobiphenyl only with two strains Burkholderia sp. HAOP24 and Rhodococcus sp. HAOP30. The colour was changed to yellow in the culture media of Pseudomonas sp. HAOP2, Pseudomonas sp. HAOP4, and Pseudomonas sp. HAOP20, it could be the DHBDO persisted during the incubation of strains on PCBs. Study and analysis the structures of bacterial community were strongly developed in the past twenty years. These methods of cultivation independent, which based on the amplification of DNA from PCBs-contaminated soil followed by acrylamide gel electrophoresis, DNA fragments having the same length and specific were separately sequenced, when band intensities were too strong to clearly differentiate between products in some gel regions. The degree of contamination-stage-related increase of individual bands in the profiles indicated a succession of community members and suggested that the diversity of bacteria increased as a result of increasing PCBs in the soil. In this study, SSCP analysis targeting 16SrRNA of bacterial community in soil sites were amplified from DNA extracted directly from soil samples collected from the sites. SSCP method could be indication for the structures and successions of community. For community profiling, as described in this study, there are important criteria such as the conditions of PCR and the selected primers which may influence on the analysis outcome. The products, which amplified by PCR could be not accurately reflect the microbial diversity in the template mixture due to different 16SrRNA-subunit gene copy numbers or biases during the amplification process. However, since the number of products which could be detected in the profiles was limited to a maximum of approximately 40, it could be anticipated that shifts in the community structure on the basis of eubacterial PCR amplifications would be detected only when quantitatively dominant organisms were affected. If selected primers, which are specific for microbial community groups, were used, it may be more specific detection for bacteria, where specific primer systems designed to characterize the diversity of bacteria in soil such as actinomycetes, and in compost as ammonia-oxidizing proteobacteria. The combination of products succession with band intensities increasing and decreasing in SSCP profiles in this study, indicates the high potential of this technique to monitor microbial communities and their variation qualitatively and quantitatively. As a result of the availability of more gene sequences, and designing optimized primer, the PCR-SSCP-mediated will become even more attractive for monitoring of different subgroups or microorganisms. The numbers of identified products through DNA sequencing will be reduced to only the specific importance by using the genetic profiles. PCR-SSCP in combination with the efficient DNA sequencing, molecular analysis to the microbial communities gains new relevance for applied microbial ecology using biotechnological monitoring processes. Extradiol dioxygenases play a pivotal role in numerous degradative pathways, and catalyze the second step in the aromatic hydrocarbon catabolic pathway. The Transformation of 2,3-dihydroxybiphenyl from colorless substrates to yellow color product (Fig. 5) could be used broadly to screen for diverse microorganisms harboring extradiol dioxygenase genes. As for Rieske non-heme iron oxygenases, knowledge on the diversity of extradiol dioxygenases is rapidly increasing, and novel branches in the phylogeny are continuously being discovered. However, despite their importance, biochemical characterizations and the crucial role of extradiol dioxygenases specifically are still scarce. In the present study 23 PCBs degraders were isolated and characterized. Based on the analysis of PCBs degrading the lower chlorinated congeners could be easily transformed in comparison with highly chlorinated PCB. From the 23 strains only Burkholderia sp. HAOP24 and Rhodococcus sp. HAOP30 have the capability to transform the higher chlorinated biphenyls behave like Burkholderia sp. strain LB400 and Rhodococcus jostii RHA1, which transform up to hexachlorinated biphenyls and showed high activities with DHB, the results indicated that 2,3 dihydroxybiphenyl dioxygenase (DHBO) may be a good marker for PCBs bacterial degraders. Soil microorganisms play important roles in maintaining soil quality and ecosystem health. For a broader characterization of soil quality, an effective methods were developed for studying the composition, diversity, and behavior of microorganisms in soil habitats is essential. SSCP has the potential for accurate comparison of environmental samples in a short period of time. In this research the group accounted for 30%. The difference could be expected due to the difference in sample locations, soil properties, land use patterns, and also taxonomic classification of different databases (NCBI and RDP) used in this different studies. Indeed, SSCP technique gives good identity for the microbial communities in the aromatic hydrocarbon polluted soil sites and could reveal the mystery of the complex relationships between these microbes. The meta-cleavage reaction product is yellow colored providing an easy colorimetric test for a rapid screening of bacterial colonies carrying 2,3-dihydroxybiphenyl 1,2-dioxygenase activity as a result of the transformation of 2,3-dihydroxybiphenyl to the yellow meta-cleavage product 2-hydroxy-6-oxo-6-phenylhexa-2,4-dienoic acid, correlating with the contamination levels (Fig. 5). Soil samples were collected from the upper few centimeters of the soil surface contaminated with aromatic hydrocarbon in Kafr El Ziat Egypt. The sampling sites were designated II (highly contaminated soil with the wastes of some chemical, insecticides, and pesticides producing factories), I (slightly contaminated soil 10 km far from the wastes), and 0 (supposed to be non-polluted soil 50 km far from the wastes). These samples (II, I, and 0) were used for DNA extraction. Total DNA from the PCB-contaminated soil sample (10 g wet weight) was extracted in triplicate using Ultra-CleanTM MegaPrep soil DNA isolation kit (MO BIO Laboratories, Carlsbad, CA) in combination with cell disruption by bead beating for 30 s using a MSK cell homogenizer (Braun, Melsungen, Germany). DNA was precipitated and purified using standard methods, followed by a further purification step with the Wizard DNA clean-up system (Promega, Madison, WI). DNA concentrations from soil extracts were quantified using a PicoGreen doublestranded DNA (dsDNA) quantitation kit (Molecular Probes, Leiden, the Netherlands). The DNA extracts from AH contaminated soil samples containing approximately 2 ng ml-1 DNA were used either directly or tenfold diluted in Tris–HCl buffer (10 mM, pH 8.0), while the DNA extracts from the PCB-contaminated soil containing approximately 380 ng ml-1 DNA were 50- or 100-fold diluted in Tris–HCl buffer (10 mM, pH 8.0) and used as template DNA in PCR. Two different primer systems were used to amplify 16S rRNA gene from total community DNA of soil samples (Table 2). DNA bands were visualized by ethidium bromide staining DNA was used as a template, and serial dilutions were applied in PCR reactions with each primer set. Purification of PCR products was done with either the Qiaquick PCR Cleaning Kit or the Gel Extraction Kit (Qiagen) according to the manufacturer instructions. PCR amplification of 16S rDNA fragments from isolates was performed as previously described. Single-stranded DNA from PCR products was obtained as previously described. Briefly, PCR performed with one of the primers being 5' phosphorylated, after the elution of the PCR products and digestion the phosphorylated strands using lambda exonuclease (NEB). The remaining single-strands were purified with Qiaquick PCR Cleaning Kit (Qiagen), dried by vacuum centrifugation, resuspended in 6 µl of loading buffer (95% formamide. 0.25% bromophenol blue and 0.25% xylene cyanol), followed by denaturation at 94 °C for 5 min, and then 3 min on ice. DCode System for PCR-SSCP (Bio-Rad) coupled to a cooling bath device (Lauda E100) was optimized at 26 °C and 120 V (10 mA) for 18 h for the separation on 20 cm × 20 cm × 0,75 mm 0.6X MDE gels using TBE running buffer 0.7×. The obtained ssDNA from 100 to 400 ng dsDNA in the gel was amplified and analyzed by PCR-SSCP and silver stained. From the dried gel single-strand harbor different conformations were excised, and using the same primers the original dsDNA fragment was generated. DNA sequencing using M13 forward and reverse primers was carried out, and phylogenetic analyses for the obtained sequences were done. Alignments were generated with CLUSTAL X 1.8 windows interface of CLUSTAL W program using default values. Their results were edited and translated using GeneDoc (version 2.6.001). The Sequence Match program was used to find the closest relative to the 16S rDNA sequences obtained. The Sequence Match program to obtain the 16S rDNA sequences and their closest relative, using CLUSTAL program and Kimura Matrix, the Phylogenetic trees were obtained and the distances were generated. To visualize relationships between the sequences retrieved by all the methods, the program Treecon for Windows (1.3b) was used to estimate distances using a Kimura matrix with 500 bootstrap samples and to infer tree topology by UPGMA clustering method. PCBs-degrading bacteria were isolated from the three PCBs contaminated soil samples. 1 g of soil incubated in 100 ml of mineral medium with adding 2 mM of 2-Chlorobiphenyl (2-CB) or 4-Chloobiphenyl (4-CB) each in one flask as sole carbon source. After one month incubation at 30 °C, reculture using 10% from the old cultures to fresh minimal medium for one more month. Organisms in PCBs enriched cultures were obtained by a spray plate technique. 0.1 ml from each culture was spread on mineral medium agar, the ethereal solution of 2-CB or 4-CB was sprayed onto plates. The plates were incubated about one month. As result of presence clear zone around the colony and transformed into yellow color by DHB, PCBs-degrading bacteria could be selected. The colonies were purified on the same media containing 2 mM of either 2-CB or 4-CB. After one month the plates were sprayed with 10 mM from 2,3-dihydroxybiphenyl (DHB) for conformation its PCBs – degrading metabolic pathway. Colonies turning yellow as a result of DHB were purified and identified partially using two primer sets targeting the 16SrRNA. Supplementary Information.

PMC9649697

Zhi-Jie Xia,Sonal Mahajan,Earnest James Paul Daniel,Bobby G. Ng,Mayank Saraswat,Alexandre Rosa Campos,Rabi Murad,Miao He,Hudson H. Freeze

COG4 mutation in Saul-Wilson syndrome selectively affects secretion of proteins involved in chondrogenesis in chondrocyte-like cells 10.3389/fcell.2022.979096

28-10-2022

COG4,Saul-Wilson Syndrome,3D culture,secretome,mass-spectrometry,N-glycan

Saul-Wilson syndrome is a rare skeletal dysplasia caused by a heterozygous mutation in COG4 (p.G516R). Our previous study showed that this mutation affected glycosylation of proteoglycans and disturbed chondrocyte elongation and intercalation in zebrafish embryos expressing the COG4p.G516R variant. How this mutation causes chondrocyte deficiencies remain unsolved. To analyze a disease-relevant cell type, COG4p.G516R variant was generated by CRISPR knock-in technique in the chondrosarcoma cell line SW1353 to study chondrocyte differentiation and protein secretion. COG4p.G516R cells display impaired protein trafficking and altered COG complex size, similar to SWS-derived fibroblasts. Both SW1353 and HEK293T cells carrying COG4p.G516R showed very modest, cell-type dependent changes in N-glycans. Using 3D culture methods, we found that cells carrying the COG4p.G516R variant made smaller spheroids and had increased apoptosis, indicating impaired in vitro chondrogenesis. Adding WT cells or their conditioned medium reduced cell death and increased spheroid sizes of COG4p.G516R mutant cells, suggesting a deficiency in secreted matrix components. Mass spectrometry-based secretome analysis showed selectively impaired protein secretion, including MMP13 and IGFBP7 which are involved in chondrogenesis and osteogenesis. We verified reduced expression of chondrogenic differentiation markers, MMP13 and COL10A1 and delayed response to BMP2 in COG4p.G516R mutant cells. Collectively, our results show that the Saul-Wilson syndrome COG4p.G516R variant selectively affects the secretion of multiple proteins, especially in chondrocyte-like cells which could further cause pleiotropic defects including hampering long bone growth in SWS individuals.

COG4 mutation in Saul-Wilson syndrome selectively affects secretion of proteins involved in chondrogenesis in chondrocyte-like cells 10.3389/fcell.2022.979096 Saul-Wilson syndrome is a rare skeletal dysplasia caused by a heterozygous mutation in COG4 (p.G516R). Our previous study showed that this mutation affected glycosylation of proteoglycans and disturbed chondrocyte elongation and intercalation in zebrafish embryos expressing the COG4p.G516R variant. How this mutation causes chondrocyte deficiencies remain unsolved. To analyze a disease-relevant cell type, COG4p.G516R variant was generated by CRISPR knock-in technique in the chondrosarcoma cell line SW1353 to study chondrocyte differentiation and protein secretion. COG4p.G516R cells display impaired protein trafficking and altered COG complex size, similar to SWS-derived fibroblasts. Both SW1353 and HEK293T cells carrying COG4p.G516R showed very modest, cell-type dependent changes in N-glycans. Using 3D culture methods, we found that cells carrying the COG4p.G516R variant made smaller spheroids and had increased apoptosis, indicating impaired in vitro chondrogenesis. Adding WT cells or their conditioned medium reduced cell death and increased spheroid sizes of COG4p.G516R mutant cells, suggesting a deficiency in secreted matrix components. Mass spectrometry-based secretome analysis showed selectively impaired protein secretion, including MMP13 and IGFBP7 which are involved in chondrogenesis and osteogenesis. We verified reduced expression of chondrogenic differentiation markers, MMP13 and COL10A1 and delayed response to BMP2 in COG4p.G516R mutant cells. Collectively, our results show that the Saul-Wilson syndrome COG4p.G516R variant selectively affects the secretion of multiple proteins, especially in chondrocyte-like cells which could further cause pleiotropic defects including hampering long bone growth in SWS individuals. Saul-Wilson syndrome (SWS) is a rare skeletal dysplasia characterized by a distinct facial phenotype, short stature, brachydactyly, clubfoot deformities, cataracts, and microcephaly (Saul and Wilson, 1990; Ferreira et al., 2020a). Our previous study identified that SWS is caused by a heterozygous, dominant variant (p.G516R) in COG4, a Golgi-associated protein involved in protein trafficking (Ferreira et al., 2018). The COG4p.G516R variant did not show a reduced protein level, in contrast to loss-of-function mutations in COG4, which causes COG4-deficient congenital disorder of glycosylation (CDG) (Reynders et al., 2009; Ng et al., 2011). COG4-CDG is usually lethal and characterized by neurological deficiencies, microcephaly, and impaired N-glycosylation. Although SWS individuals also show microcephaly, their N-glycans and neurological features appear normal (Ferreira et al., 2018; Ferreira, 2020; Ferreira et al., 2020b). Zebrafish models for both COG4-CDG and COG4-SWS have been reported and they all display small body length, abnormal pectoral fins, and abnormal chondrocyte stacking (Ferreira et al., 2018; Clement et al., 2019; Xia et al., 2021). But COG4-null zebrafish exhibit more severely impaired proteoglycan synthesis based on reduced Alcian blue staining of the jaw compared to COG4p.G516R zebrafish. Our previous work also revealed that the COG4p.G516R variant disturbed WNT4 signaling at the embryonic stage in zebrafish development. These differences likely indicate distinctive mechanisms for the dominant verses recessive COG4 disorders (Xia et al., 2021). COG4 is a subunit of the conserved oligomeric Golgi (COG) complex which belongs to complexes associated with tethering containing helical rods (CATCHR) family (Yu and Hughson, 2010; Blackburn et al., 2019). CATCHR complexes play critical roles in vesicle tethering, intra-Golgi trafficking, Golgi homeostasis, and membrane trafficking (Lees et al., 2010; Blackburn et al., 2019; Adusumalli et al., 2021). It is not surprising that the COG4p.G516R variant could disturb the secretion of multiple proteins. Our previous findings of impaired decorin glycosylation and glypican turnover prompted a more thorough study of protein secretion in a suitable cell line carrying the COG4p.G516R variant (Ferreira et al., 2018; Xia et al., 2021). Sumya et al. have shown increased secretion of intracellular protein SIL1 and ERGIC53 in COG4-G516R RPE1 cells compared to WT. But a more relevant, bone-related cell type might reveal different candidates in a secretome study, since mutant RPE1 cells (p.G516R and p. R729W) did not phenocopy patient fibroblasts (Sumya et al., 2021). Abnormal chondrocyte intercalation and elongation in SWS zebrafish prompted us to choose chondrocyte-like cells to examine the overall protein secretion affected by COG4p.G516R variant. Chondrocytes secret extracellular matrix (ECM) proteins, growth factors, and enzymes which further regulate ECM synthesis (Melrose et al., 2016; Chijimatsu and Saito, 2019; Chen et al., 2021). During endochondral ossification, chondrocytes undergo a series of differentiation steps to form the growth plate with the support of the ECM (Harley, 1988; Yang et al., 2014). Three-dimensional (3D) cell culture is an emerging technology that allows a more physiological expansion and differentiation of cells compared to cultivation on conventional 2D systems (Benya and Shaffer, 1982; Wang et al., 2009; Studer et al., 2012). Pellet culture is a scaffold-free 3D culture form, which is commonly used to stabilize the chondrogenic potential of in vitro cultured chondrocytes (Caron et al., 2012; Grigull et al., 2020). Moreover, studies of using tumor cells showed that spheroids formed in in vitro 3D models exhibit physiologically relevant cell-cell and cell-matrix interactions, gene expression and signaling pathway profiles, bridging the gap between 2D culture models and in vivo whole animal systems (Nath and Devi, 2016). Therefore, this paper aims to utilize cell spheroids obtained by 3D culture methods as a quantitative biomarker and mass spectrometry (MS)-based secretome analysis to investigate the specific changes caused by COG4p.G516R variant compared to WT and COG4-KO in chondrocyte-like cells to extend the picture beyond skin fibroblasts. Since SWS is a skeletal disorder, the chondrocyte-like cell line SW1353 was chosen to generate COG4p.G516R and COG4 knock-out (KO) cells by using CRISPR technology. Generated cell lines were verified by sequencing and Western blots (Figures 1A,B). The protein level of COG4p.G516R variant did not decrease, while COG4-KO cells displayed a total loss of COG4, as expected (Figure 1B). We further evaluated whether these mutant cells mimic the features of patient fibroblasts. One of the most significant changes in SWS-derived fibroblasts was altered protein trafficking between the ER and Golgi (Reynders et al., 2009; Ferreira et al., 2018). COG4p.G516R cells show accelerated Brefeldin A (BFA)-induced retrograde transport as seen previously in SWS-derived fibroblasts. In contrast, COG4-KO cells, show decreased retrograde transport as seen in COG4-CDG patient fibroblasts [Figures 1C,D (Ferreira et al., 2018)]. Another change seen in fibroblasts from SWS individuals was an apparently enlarged COG complex hydrodynamic volume based on glycerol gradient ultracentrifugation. We observed an evident shift of COG complex to the heavier fraction in SW1353 cells carrying COG4p.G516R mutation (Figure 1E) as seen in SWS individual’s fibroblasts. Our initial studies did not find significant N-glycan changes in SWS-derived fibroblasts (Ferreira et al., 2018). Here we compared N-glycans in different cell types carrying COG4 mutations to examine possible cell-type dependent changes. Chondrosarcoma cells carrying COG4p.G516R variant and COG4-KO exhibited subtle differences in N-glycans (Figure 2A). There were two specific changes in COG4p.G516R cells showing increased Fuc1HexHexNAc3 and decreased Fuc1Hex5HexNAc4, which were synthesized in medial- and trans-Golgi respectively (Figure 2A, right panel). For comparison purposes, we also generated stable cell lines expressing WT COG4 and COG- COG4p.G516R under pCMV6 promotor in HEK293T COG4-KO cells (a gift from Prof. Lupashin (Blackburn and Lupashin, 2016)), respectively and verified the COG4 expression by Western (Supplementary Figure S1). HEK293T cells expressing COG4p.G516R mutation did not show specific changes in N-glycans (Figure 2B). As expected, COG4-KO HEK293T cells show dramatic changes in multiple N-glycans including decreased sialylated and galactosylated glycans, specifically Mono-sialo, Mono-sialo fucosylated and Fuc1Hex5HexNAc4 (Asialo fucosylated) glycans (Figure 2B, right panel). Interestingly, COG4-KO HEK293T cells displayed altered abundance of some high mannose N-glycans including significant decrease in Man6 and Man9 and increase in Man4 and Man5 (Figure 2B, left panel) which were not seen in COG4-CDG patient serum samples (Reynders et al., 2009). These changes in N-glycans of COG4-KO cells suggested that COG4 KO affect multiple stages of N-glycan processing in the whole Golgi and preferentially, in the early processing steps in the cis-medial Golgi. Overall, N-glycosylation changes in COG-COG4p.G516R are extremely subtle and cell-type dependent in comparison to WT. A scaffold-free 3D culture format, pellet culture, was used to examine the in vitro chondrogenic differentiation of chondrosarcoma SW1353 cells. COG4p.G516R showed significantly reduced spheroid sizes on Day 10 compared to WT and COG4-KO cells (Figures 3A–C). Hematoxylin and eosin (H&E) staining further revealed a necrotic core on day 6 in COG4p.G516R cells (Figure 3D). Increased apoptosis was exclusively seen in COG4p.G516R cells starting at Day 4 (Figures 3E,F), which probably explains the reduced aggregate size. We also examined the aggregate formation of control and SWS-derived fibroblasts using pellet culture. However, we did not see differences in terms of aggregate sizes and Alcian blue staining (Supplementary Figures S2A–S2C). H&E staining did not show significant differences except for slightly reduced fringe cells in SWS and COG4-CDG cells (Supplementary Figure S2D). Mildly increased apoptosis was also seen in SWS-derived fibroblasts, similar to COG4p.G516R chondrosarcoma cells (Figure 3E). To study the interplay between WT and mutant cells, we made mixed cultures in 384-well ultra-low attachment (ULA) plates and evaluated the effect on the spheroid formation. We first fluorescently labeled WT, COG4p.G516R, and COG4-KO chondrosarcoma cells by expressing cytosolic RFP or GFP. Then WT and mutant cells were either cultured alone or mixed in a one-to-one ratio (Figure 4A). When cultured alone in ULA plates, COG4p.G516R cells showed reduced size on Day 2 (Figure 4A, right bottom) compared to WT (Figure 4A, left bottom) with the same number of cells. Interestingly, in mixed cultures of WT and COG4p.G516R cells, the aggregate size was visibly increased (Figure 4A, middle panel). We did not see changes in mixed culture of WT and COG4-KO cells (Figure 4B). We further examined the cell viability by Propidium Iodide (PI) staining and found that mixed culture with WT could significantly reduce cell death of COG4p.G516R cells (Figures 4C,D). Interestingly, the conditioned medium collected from WT cells could partially rescue the spheroid sizes of COG4p.G516R cells by about 25% (Figures 4E,F). This effect was further confirmed by checking the aggregate formation on poly-lysine coated plates (Supplementary Figure S3). In both systems, adding conditioned medium from COG4p.G516R cells to WT cultures did not decrease spheroid size or cause cell death, showing that the WT medium was providing factors that were deficient in the medium of COG4p.G516R cells. Prompted by the WT medium complementation experiments, we examined the proteins in the conditioned medium by Coomassie-blue staining after separating proteins in polyacrylamide gels. We observed some altered protein bands in chondrosarcoma COG4 mutant cells which were not seen in cell lysates (Figure 5A). MS-based proteomics was further performed to identify all proteins in the conditioned medium. The comparisons of COG4p.G516R vs. WT and COG4p.G516R vs. KO were displayed as Volcano plots (Figures 5B,C). To determine the functional meaning of the differentially regulated secreted proteins, we performed Gene Set Enrichment Analysis (GSEA) by using different Human Gene Sets in Molecular Signatures Database. We found a significant over-representation of categories mainly related to extracellular matrix and some others, with few top candidates as “Naba_Matrisome”, “GOCC_Collagen_Containing_Extracellular_Matrix”, “Angiogenesis”, “Complement”, “Coagulation” and “Epithelial_Mesenchymal_Transition” (Figure 5D). We see about 84 proteins with significantly decreased measurement and 48 proteins with significantly increased measurement, belonging to “Naba Matrisome”. The altered Matrisome proteins includes various proteoglycans (Decorin, Versican, Lumican, hyaluronan and proteoglycan link protein 1), glycoproteins (IGFBPs, LTBPs, Fibulins, laminins, PCOLCE, POSTN), various classes of collagens, ECM Regulators (ADAM metallopeptidases, MMPs, serpin peptidase inhibitor, cathepsins), ECM-affiliated Proteins (Glypicans, syndecans, annexins) and secreted factors (BMPs, WNT proteins, growth factors, chemokines). Many of these deficient Matrisome proteins are common regulators of other affected pathways. A list of top 10 candidates which were decreased in secretions of COG4p.G516R as compared to WT is shown in Figure 5E. Our secretome data showed that some normally secreted proteins were deficient in both COG4p.G516R and COG4-KO cells including fibronectin (FN1) and some proteoglycans (versican and decorin), which we further verified in western blots (Figure 5F). Interestingly, we also found several proteins that were exclusively missing only in COG4p.G516R cells (Figure 5E). Several top candidates were verified by Western blotting, including matrix metallopeptidase 13 (MMP13), insulin-like growth factor binding protein 7 (IGFBP7), and proprotein convertase subtilisin/kexin type 9 serine protease (PCSK9) (Figure 5G), confirming the MS findings. A few ER stress markers were also tested because of reduced secretion, although none of them were top candidates in our list. Western blotting of these markers did not show increased protein level indicating no ER stress in COG4p.G516R cells (Supplementary Figure S4). Conditioned medium from control and SWS patient fibroblasts were also analyzed by MS. However, fibroblasts gave many fewer candidates than chondrosarcoma cells (Supplementary Figure S5), indicating the importance of choosing a relevant cell type to study skeletal disorders. MMP13 and proteoglycans are critical for chondrocyte differentiation, especially from proliferating chondrocytes to hypertrophic chondrocytes (Studer et al., 2012; Komori, 2022). Prompted by the MS data, we further examined the mRNA level of differentiation markers, MMP13 and COL10A1 of aggregates in pellet culture treated with TGFβ3. We found that COG4p.G516R cells showed profoundly low levels of MMP13 on Day 0–4 and COL10A1 on Day 2–4 (Figure 6A). We could not detect COL10A1 expression on Day 0. After 8 days of treatment with TGFβ3, mRNA level of MMP13 in COG4p.G516R cells was still significantly lower than WT and COG4-KO cells, while COL10A1 level could catch up on Day 8 (Figure 6A). Using fibroblasts, we did not see impaired COL10A1 and MMP13 mRNA levels over days in SWS patient cells compared to the control (Supplementary Figure S2F). Another chondrogenic factor, bone morphogenetic protein 2 (BMP2) was also tested using 3D and 2D cultures. In pellet culture, COG4p.G516R cells could not form spheroids with the presence of BMP2 indicating disturbed chondrogenic differentiation, opposite to WT and COG4-KO cells (Figure 6B). Moreover, COG4p.G516R cells exhibited lower levels of phosphorylation of Smad1/5/9 proteins in response to BMP2 than did control cells (Figures 6C,D). COG4 is a protein involved in protein trafficking and glycosylation, and its mutations cause both dominant and recessive human genetic disorders, SWS and COG-CDG, respectively. In this study, we aimed to choose a proper cell model to understand how the dominant COG4p.G516R variant causes a primordial dwarfism. Studies in SWS-derived fibroblasts and HEK293T cells provide confidence that the molecular characteristics and trafficking abnormalities are maintained and are an inherent property of the mutated protein, but chondrocyte-like SW1353 has several advantages. SW1353 cells carrying COG4p.G516R mutation not only manifest the distinctive defects seen in SWS-derived fibroblasts, but also synthesize a more disease-relevant phenotype displaying changes of ECM protein secretion, indicating a suitable cell model to study SWS. We previously reported that N-glycans in SWS patient serum samples and fibroblasts were not significantly changed compared to controls (Ferreira et al., 2018). In this study, we applied higher resolution MS technology, but found only very subtle changes in a few N-glycans in chondrosarcoma cells carrying COG4p.G516R mutation, typical of SWS patient samples. Similarly, essentially no N-glycan changes were seen in HEK293T cells expressing COG4p.G516R variant. On the other hand, COG4-null cells showed severe changes in N-glycans. It has been reported that decreased sialylation of N-glycans was a common observation in COG-CDG patient serum samples with COG4-CDG showing the least severely affected (Reynders et al., 2009). Our study confirmed this observation, and we also found dramatic changes in the abundance of high mannose structures, suggesting that COG4 loss impacts the whole Golgi, but primarily the initial mannose processing of N-glycans in the cis-medial Golgi in HEK293T cells. The dramatic N-glycans changes in HEK 293T COG4-KO cells compared to the SW1353 COG4-KO cells further shows cell-type specific impact of COG4 KO on N-glycosylation. One possible explanation could be that the glycosylation is affected by multiple factors that may vary with cell type including the integrity of the Golgi peripheral membrane proteins, the Golgi membrane dynamics, growth factor signaling, and cellular stress in addition to the primary spectrum of glycosyltransferases (Stanley, 2011) and knocking out COG4 may have different effect on trafficking of any of these glycosylation components in different cell types. 3D culture models have gained growing acceptance as in vitro tools to study bone diseases (Salamanna et al., 2016; Langhans, 2018; Bicer et al., 2021). In the presence of TGFβ3 and BMP2, chondrocyte pellets proliferate, producing hypertrophic chondrocytes, and the extraordinary cell volume increase during hypertrophy is accompanied by an up-regulation of COL10A1 and MMP13 (Studer et al., 2012; Yang et al., 2014). The low levels of COL10A1 and MMP13 detected in SW1353 COG4p.G516R spheroids imply impaired hypertrophic differentiation which could further affect long bone growth. In addition to forming no spheroids, impaired response to BMP2 in COG4p.G516R cells was confirmed by the decreased phosphorylation of SMAD1/5/9 which is required for endochondral bone formation (Retting et al., 2009). ULA plate blocks cell attachment and causes cells in suspension to aggregate into visible spheroids. Altered spheroid size and increased cell death rate verified in 384-well ULA plates also provide promising quantitative markers in high-throughput drug screening for SWS in the future. Secretome data confirmed the reduced protein level of MMP13 and COL10A1 and revealed more changes in protein secretion in COG4 mutant cells. GSEA of COG4p.G516R vs. WT data showed the most significant enrichment with “Naba Matrisome” geneset, which is a depository of not only core ECM proteins but also ECM-affiliated proteins, ECM regulators and secreted factors, many of which are known to bind to ECM (Naba et al., 2016). In both COG4p.G516R and COG4-KO chondrosarcoma cells, secretion of some proteoglycans was significantly altered. It is well established that proteoglycans in the ECM play critical roles in skeletal development by interacting with secreted growth factors (Koosha and Eames, 2022). The reduced ECM proteoglycans in both clonal cell lines might explain the common skeletal deficiencies seen in both SWS and COG4-CDG patients. Interesting, we identified some candidates that are only increased or decreased in COG4p.G516R cells. Our top candidates include decreased IGFBP7, an important growth factor involved in osteoblastic reprogramming of fibroblasts (Lu et al., 2020; Koosha and Eames, 2022), COL1A2 and HAPLN1 which are important components of ECM. COL1A2 deficiencies cause Osteogenesis Imperfecta (OI) which manifests overlapping features with SWS as blue sclerae, bone fragility and hearing loss (Ferreira, 2020). Interestingly, we also found exclusively elevated components in the medium of COG4p.G516R cells, including WNT signaling receptor FZD2, glypican 1 (GPC1), and LTBP4. These results reinforce the critical roles of WNT signaling pathway in chondrogenic hypertrophy and SWS pathogenesis (Studer et al., 2012; Xia et al., 2021). We are aware that FZD2 and GPC1 are membrane proteins instead of secreted or ECM proteins. The possible reason of the detection of GPC1 could be proteoglycan shedding which is considered a powerful post-translational modification (Niehrs, 2012; Xia et al., 2021). Also, the detection of some cytoplasmic proteins in the chondrocyte secretome using proteomic techniques has been reported previously and one of the potential explanations is considered as secretion of ECM-derived vesicles by chondrocytes. These vesicles contain enzymes and substrates necessary for mineral formation and believed to be one of the possible reasons for the cytoplasmic/membrane proteins, such as annexins, in the secretome of the chondrocytes (Sanchez et al., 2017). We believe SWS pathogenesis is a result of complex interplay of alterations in the matrix components, growth factors and signaling pathways, which is currently not fully understood. There is a possibility that differential secretion of some components leads to a regulation of multiple pathways. However, the mechanism behind the selectively affected protein trafficking by COG4p.G516R variant and its implication in regulation of other signaling pathways is still unknown. In summary, our findings demonstrate that chondrocyte-like cells are a suitable model to investigate the deficiencies caused by the COG4p.G516R variant. Combined studies of spheroids generated by 3D culture methods and MS-based secretome analysis revealed impaired chondrogenic differentiation as a highly relevant reason contributing to SWS pathogenesis. How COG4p.G516R variant selectively affects protein trafficking is our top interest and relevant studies are ongoing. Primary fibroblasts derived from healthy controls (GM00038 and GM09503) were obtained from Coriell Institute for Medical Research (Camden, NJ). Each SWS-derived fibroblast line was obtained by the referring clinician and grown via a clinical lab service and then sent to us with consent through an approved IRB (Ferreira et al., 2018). SW1353 was obtained from ATCC. Fibroblasts and SW1353 were cultured in Dulbecco’s Modified Eagle’s medium (DMEM) containing 1 g/L glucose supplemented with 10% heat-inactivated fetal bovine serum (FBS), 1X Penicillin-Streptomycin (Corning) and 200 mm L-glutamine (Corning). HEK293T cells were cultured in 4.5 g/L glucose DMEM supplemented with the same components as mentioned above. CRISPR knock out of COG4 in SW1353 was performed as previously described with some slight modifications (Bailey Blackburn et al., 2016; Blackburn and Lupashin, 2016). Cell Line Nucleofector™ Kit V (Lonza) was used for SW1353 transfection according to manufacturer instructions. SW1353 cells carrying COG4-G516R mutation were generated by CRISPR knock in as previously described (Ran et al., 2013; Dong et al., 2019). Primers used were: 5′-CAC​CGG​CGG​CAC​TTG​TCA​CCC​CGC​GC-3′ and 5′-AAA​CGC​GCG​GGG​TGA​CAA​GTG​CCG​CC-3'. ssODN repair template was 5′-CCT​GCC​ACC​ACC​TTC​CAG​GAC​ATC​CAG​CGC​AGA​GTG​ACA​AGT​GCC​GTG​AAC​ATC​ATG​CAC-3'. Generated COG4-G516R clones were verified by PCR with primers 5′-CCA​GTG​CTC​TGG​GGA​ATG​AAT-3′ and 5′-GGC​CAG​TCT​CCC​CTG​TAT​GT-3′ followed by sequencing. BFA-induced protein trafficking assay and gradient ultracentrifugation were performed and analyzed as previously described (Ferreira et al., 2018). N-glycans were extracted from SW1353, HEK293T and fibroblast cells as previously described (Chen et al., 2019). Briefly, N-glycans were released and derivatized using Water’s Rapifluor label, and the derivatized N-glycans were purified and concentrated using HILIC chromatography and were flow-injected into ESI-QTOF Mass Spectrometry to detect glycan intermediates by accurate Mass. An internal standard is used for quantification and percent relative abundance of glycans is reported with around 0.1% sensitivity. In pellet culture, 4 × 105 of SW1353 or fibroblast cells were centrifuged (150 g, 5 min) in a 15-ml polypropylene tube (Becton Dickinson) to form a pellet. The pellets were treated with DMEM (1 g/L glucose) containing ITS supplement, 0.2 mm L-ascorbic acid-2-phosphate, 1 mm sodium pyruvate, 0.35 mm L-proline (all Sigma-Aldrich), and 10 ng/ml TGFβ3 (Biolegend, 585802) or BMP2 (sigma H4791). Medium was changed every two or 3 days (Ullah et al., 2012; Futrega et al., 2021). Alcian blue staining and extraction of aggregates were performed as described (Wehrle et al., 2019). In 384-well ULA plate, 3 × 103 cells were seeded to each well with the medium mentioned above. Lentivirus transduction and PI staining were performed as described (Tambe et al., 2019). RFP or GFP plasmids were obtained from Addgene (#13726 and #13727). Fluorescence images were acquired using an LSM 710 confocal microscope (Zeiss, Germany). H&E and TUNEL assay were performed by Histology core at Sanford Burnham Preby (SBP) following standard protocol. ApopTag® Red In Situ Apoptosis Detection Kit (Millipore, S7165) was used for TUNEL assay. SW1353 cells were grown in 10-cm dishes to 90–100% confluency, rinsed 3 times with DPBS, and one time with F12K medium (Corning™ 10025CV). Cells were incubated in F12K for 24 h, then conditioned media was collected, clarified, and concentrated using a 3 k concentrator (Amicon® Ultra 3 k, Millipore). For fibroblasts, DMEM medium was used as above mentioned and incubated for 48 h before collection. Duplicate analyses were performed for each cell line. Secreted protein profiling was performed at proteomics core at SBP. Briefly, protein samples were exchanged using 10-kDa Amicon filter (Millipore) to 8 M urea, 50 mm ammonium bicarbonate buffer. While in the filter, proteins were reduced with 5 mm tris(2-carboxyethyl) phosphine (TCEP) at 30°C for 60 min, and subsequently alkylated with 15 mm iodoacetamide (IAA) in the dark at room temperature for 30 min. The buffer was then exchanged again to 1 M urea, 50 mm ammonium bicarbonate, followed by overnight digestion with mass spec grade Trypsin/Lys-C mix. Digested peptides were then desalted in the Bravo platform using AssayMap C18 cartridges, and dried down in a SpeedVac concentrator. Dried peptide samples were reconstituted in 2% ACN, 0.1% FA. A Proxeon EASY nanoLC system (Thermo Fisher Scientific) coupled to an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) was used for performing proteomic analysis (LC-MS/MS). An analytical C18 Aurora column (75 µm × 250 mm, 1.6 µm particles; IonOpticks) was used to separate peptides at a flow rate of 300 nL/min using a 100-min gradient: 2%–23% B in 70 min, 23%–34% B in 20 min, 34%–98% B in 2 min, (A = FA 0.1%; B = 80% ACN: 0.1% FA). Precursors were surveyed in MS scan and settings in the Orbitrap were as follows: m/z range 350 to 1,050, 120k resolution at m/z 200, AGC 5e5 with maximum injection time of 50 ms, RF lens 30%. 3-s cycles were specified for the survey and the MS/MS scans. For DIA window size, a 10 m/z isolation window was utilized. All precursors were fragmented with 30% normalized collision energy. Scan range for MS/MS was 200–2000 m/z. AGC settings were specified as 5e5 and maximum injection time of 22 ms. All mass spectrometry data was analyzed with Spectronaut software (Biognosys; 15.6.211,220.50606) using default settings for DierctDIA. Homo sapiens Uniprot protein sequence database (downloaded in February 2020) and GPM cRAP sequences (commonly known protein contaminants) were specified as sequence databases. Fully tryptic peptides with two missed cleavages were allowed with carbamidomethylation (C) as fixed modification and oxidation (M) and acetylation of protein N-terminus as a variable modification. Only the search hits corresponding to less than 1% of Q-value were accepted. MS1 intensity were used for calculating differences between the groups (t-test p < 0.05). Samples were harvested using SDS lysis buffer (62.5 mm Tris-HCl, pH 6.8, 2% SDS, and 10% glycerol) supplemented with protease and phosphatase inhibitors (Sigma-Aldrich) as previously described (Ferreira et al., 2018). Equal amounts of denatured proteins were separated via SDS–polyacrylamide gel electrophoresis followed by transfer and antibody inoculation as described previously (Tambe et al., 2019). Antibodies used were: Smad1(Cell signaling, D59D7), Smad5 (SCBT, sc-101151), PCSK9 (SCBT, sc-515082), IGFBP7 (SCBT, sc-365293), MMP13 (SCBT, sc-515284), VCAN (Invitrogen, MA5-27638), FN1 (Abclonal, A16678), DCN (R&D, MAB143), GAPDH (Invitrogen, A5-15738), PDIA6 (Proteintech, 18233-1), GRP78 (Proteintech, 11587-1), ATF4 (SCBT, sc-390063), ATF6 (SCTB, sc-166659), pSmad159 (Cell signaling, 13820), β-actin (Cell signaling, 4970), Goat Anti-Rabbit HRP (SeraCare, 54500010), and Goat Anti-Mouse HRP (Invitrogen, 31446) diluted according to manufacturer’s instruction. Total RNA was extracted from cell aggregates using TRIzol™ (ThermoFisher, 15596081) reagent according to manufacturer’s protocol. cDNA was synthesized using QuantiTect Reverse Transcription Kit (QIAGEN, 205311). qPCR and data analysis were performed as described previously (Tambe et al., 2019). The mRNA levels were normalized to the levels of housekeeping genes, POLR2A and RPLP0, and 2−ΔΔCt values were calculated and compared. Three biological replicates for each point were analyzed. Fluorescence images were acquired using an LSM 510 confocal microscope (Zeiss, Germany) with a ×20 objective. Digital images were processed with Zeiss ZEN and ImageJ software.

PMC9649702

Kaushal Asrani,Juhyung Woo,Adrianna A. Mendes,Ethan Schaffer,Thiago Vidotto,Clarence Rachel Villanueva,Kewen Feng,Lia Oliveira,Sanjana Murali,Hans B. Liu,Daniela C. Salles,Brandon Lam,Pedram Argani,Tamara L. Lotan

An mTORC1-mediated negative feedback loop constrains amino acid-induced FLCN-Rag activation in renal cells with TSC2 loss

10-11-2022

Cancer metabolism,Renal cell carcinoma,Lysosomes,Phosphorylation

The mechanistic target of rapamycin complex 1 (mTORC1) integrates inputs from growth factors and nutrients, but how mTORC1 autoregulates its activity remains unclear. The MiT/TFE transcription factors are phosphorylated and inactivated by mTORC1 following lysosomal recruitment by RagC/D GTPases in response to amino acid stimulation. We find that starvation-induced lysosomal localization of the RagC/D GAP complex, FLCN:FNIP2, is markedly impaired in a mTORC1-sensitive manner in renal cells with TSC2 loss, resulting in unexpected TFEB hypophosphorylation and activation upon feeding. TFEB phosphorylation in TSC2-null renal cells is partially restored by destabilization of the lysosomal folliculin complex (LFC) induced by FLCN mutants and is fully rescued by forced lysosomal localization of the FLCN:FNIP2 dimer. Our data indicate that a negative feedback loop constrains amino acid-induced, FLCN:FNIP2-mediated RagC activity in renal cells with constitutive mTORC1 signaling, and the resulting MiT/TFE hyperactivation may drive oncogenesis with loss of the TSC2 tumor suppressor.

An mTORC1-mediated negative feedback loop constrains amino acid-induced FLCN-Rag activation in renal cells with TSC2 loss The mechanistic target of rapamycin complex 1 (mTORC1) integrates inputs from growth factors and nutrients, but how mTORC1 autoregulates its activity remains unclear. The MiT/TFE transcription factors are phosphorylated and inactivated by mTORC1 following lysosomal recruitment by RagC/D GTPases in response to amino acid stimulation. We find that starvation-induced lysosomal localization of the RagC/D GAP complex, FLCN:FNIP2, is markedly impaired in a mTORC1-sensitive manner in renal cells with TSC2 loss, resulting in unexpected TFEB hypophosphorylation and activation upon feeding. TFEB phosphorylation in TSC2-null renal cells is partially restored by destabilization of the lysosomal folliculin complex (LFC) induced by FLCN mutants and is fully rescued by forced lysosomal localization of the FLCN:FNIP2 dimer. Our data indicate that a negative feedback loop constrains amino acid-induced, FLCN:FNIP2-mediated RagC activity in renal cells with constitutive mTORC1 signaling, and the resulting MiT/TFE hyperactivation may drive oncogenesis with loss of the TSC2 tumor suppressor. The mechanistic target of rapamycin complex 1 (mTORC1) kinase complex is a central regulator of cellular growth and proliferation, activated in response to nutrients and growth factor stimuli. In current models, mTORC1 kinase activity is tightly regulated by two classes of small GTPases: Rag GTPase heterodimers (RagA/B with RagC/D) and RHEB. In response to nutrients, Rags recruit the key mTORC1 subunit Raptor to the lysosome, where RHEB, an allosteric mTOR activator, resides. In parallel, growth factor signaling leads to inactivation of the RHEB GTPase activating protein (GAP), TSC2, or PRAS40 with subsequent activation of mTORC1. The majority of studies examining mTORC1 activity have focused on two well-characterized substrates: eIF4E-binding protein 1 (4EBP1), which activates cap-dependent translation, and ribosomal S6 kinase 1 (S6K1), which promotes protein synthesis. These substrates are recruited to the lysosome and mTORC1 by binding to Raptor via a five amino-acid-conserved TOR signaling (TOS) motif, and their phosphorylation is the most commonly used readout of mTORC1 activity. However, another key class of mTORC1 substrates are recruited to the lysosome exclusively by Rag GTPases, thus providing a readout of mTORC1 activation in response to nutrient availability. The Microphthalmia family (MiT/TFE) is comprised of four conserved basic leucine zipper transcription factors (MITF/TFE3/TFEB/TFEC) that regulate expression of CLEAR (Coordinated Lysosomal Expression and Regulation) genes. Phosphorylation of TFEB or TFE3 by mTORC1 promotes interaction with 14-3-3 chaperones in the cytosol, and results in the cytoplasmic retention and proteasomal degradation of these transcription factors. Accordingly, short-term pharmacological mTORC1 inactivation with the ATP-competitive inhibitor torin promotes MiT/TFE de-phosphorylation, nuclear localization and activity, although rapamycin does not (likely due to its incomplete allosteric inhibition of mTOR kinase). In contrast to other mTORC1 substrates, TFEB lacks a TOS motif and depends solely on amino-acid-stimulated RagC/D activation for lysosomal recruitment and phosphorylation by mTORC1. Accordingly, loss of folliculin (FLCN; a GAP for RagC/D) decreases TFEB phosphorylation and increases TFEB/TFE3 nuclear localization. Taken together, these findings are consistent with the known role of mTORC1 in the negative regulation of catabolic processes such as autophagy when nutrients are replete. However, a number of studies have emerged recently to indicate that TFEB regulation by mTORC1 is more complicated than previously appreciated. In previous work by our group and others, constitutive mTORC1 hyperactivity due to TSC1/2 loss paradoxically positively regulated TFEB-dependent lysosomal gene expression and promoted MiT/TFE nuclear localization in an mTORC1-dependent manner. Accordingly, in a study published while this manuscript was being drafted, hypophosphorylation of TFEB was confirmed in cells with TSC2 loss. The relevance of these findings for human health is evidenced by molecular pathology data demonstrating that subsets of human tumors driven by TSC1/2 genomic alterations (including perivascular epithelioid cell tumors and eosinophilic renal cell carcinomas) may alternatively be driven by mutually exclusive TFEB/TFE3 gene rearrangements or FLCN loss leading to constitutive MiT/TFE activity. Accordingly, MiT/TFE-target gene products, such as GPNMB protein, provide an excellent biomarker for tumors with TSC1/2 alterations or TFEB/TFE3 gene rearrangements, suggesting high MiT/TFE activity in both types. Taken together, these data suggest that MiT/TFE may be a previously unappreciated driver of tumorigenesis in the setting of TSC1/2 loss. However, the mechanism by which mTORC1 hyperactivation might result in a paradoxical increase in MiT/TFE activity remains unclear. In the current study, we describe a negative feedback loop that constrains amino-acid-induced mTORC1 activity vis a vis TFEB, in the setting of TSC2 loss. We demonstrate that starvation-induced lysosomal localization of the FLCN:FNIP2 complex and downstream Rag activation in response to amino acids are markedly impaired in a rapamycin-sensitive manner in cells with TSC2 loss. Importantly, TFEB phosphorylation remains responsive to inputs that re-localize or destabilize the lysosomal folliculin complex (LFC). These findings begin to elucidate the mechanism of MiT/TFE hyperactivity in the context of TSC2 loss, which may be an important mediator of tumorigenesis. We previously demonstrated that mice with constitutive epidermal mTORC1 activity (via conditional epidermal deletion of Tsc1 or expression of constitutively active RhebS16H) exhibited elevated MiT/TFE transcriptional activity, with concomitantly increased expression of CLEAR target genes and proteins. To determine whether these paradoxical findings in primary keratinocytes might be generalizable to other systems, we examined transformed human embryonic kidney HEK293T cells with or without somatic genomic deletion (KO) of TSC1, TSC2 or TSC1/2 via CRISPR-Cas9 genome editing (gift of TSC Alliance and Dr. Nellist) (Fig. 1a), (hereafter referred to as WT and TSC1, TSC2, or TSC1,2 KO cells, respectively). By RNA-seq, TFEB-regulated lysosomal gene sets were enriched by GSEA in the TSC1,2 KO cells compared to WT controls (Fig. 1b). Multiple CLEAR-regulated transcripts were significantly enriched in TSC2 and TSC1,2 KO cells, compared to WT cells by RT-PCR, with similar, though non-significant, increases in TSC1 KO cells (Fig. 1c, Fig. S1a). Validating the gene expression findings, multiple lysosomal integral proteins and enzymes were upregulated in whole-cell extracts of TSC2 KO and TSC1,2 KO cells, and in lysosomal-enriched fractions of TSC2 KO cells by immunoblotting (Fig. 1d, e). The lysosomes in TSC2 KO cells were functionally active, as measured indirectly by an increase in cathepsin B (CTSB) processing (Fig. 1d; arrow), as well as increased lipidated LC3-II (Fig. 1d, e; arrow), increased LC3-labeled puncta by immunofluorescence (Fig. S1b), and higher autophagic flux demonstrated by increased LC3-II accumulation (arrow) with hydroxychloroquine (HCQ) and Chloroquine (CQ) treatment (Fig. S1c). Finally, there was a significant increase in Cathepsin D enzyme activity in TSC2 KO cells compared to WT (Fig. S1d). To validate these findings in vivo, we also examined lysosomal biogenesis in Tsc2 ± A/J mice. In this murine model of tuberous sclerosis, spontaneous loss of the second Tsc2 + allele results in Tsc2 protein loss and development of mTORC1-driven renal cystadenomas and cystadenocarcinomas at 6–12 months of age with 100% penetrance. We performed laser-capture microdissection (LCM) on renal tumors from Tsc2 ± mice and found that levels of CLEAR transcripts were significantly enriched in tumors compared to matched normal kidney by RT-PCR (Fig. 1f, g) and were validated by immunohistochemistry (IHC)/immunofluorescence (IF) (Fig. 1h). For additional validation, we examined TTJ cells, a Tsc2-null cell line derived from Tsc2 ± C57BL/6 mice, stably transfected with a control vector (TTJ- parental) or wild-type Tsc2 (TTJ-Tsc2). TTJ-parental cells had increased lysosome/autophagosome proteins in whole-cell lysates and in lysosomal fractions by immunoblotting when compared to cells with Tsc2 re-expression. Notably, these cells also had increased processed Cathepsin B and lipidated LC3-II, consistent with a functional increase in lysosomal proteolysis and autophagic flux, respectively (Fig. 1i, j; arrow). To understand the physiological basis for elevated lysosomal content and activity in TSC2 KO cells, we first examined the subcellular localization of TFEB and TFE3, the principal drivers of autophagy and lysosomal biogenesis in mammalian cells. Paradoxically, but in accordance with our previous findings in Tsc1 KO keratinocytes, nuclear localization of endogenous TFEB and TFE3 was increased in TSC2 KO cells compared to WT controls in nutrient-replete conditions (Fig. 2a, b; Fig. S2a, b). Furthermore, transiently overexpressed GFP-tagged TFEB (TFEB-GFP) (Fig. 2c, d) or TFE3 (TFE3-GFP) (Fig. S2c, d) showed increased nuclear: cytoplasmic localization in TSC2 KO cells versus WT controls. Immunoblotting of nuclear/cytoplasmic fractions confirmed these findings for endogenous TFEB and TFE3 (Fig. 2e–g, Fig. S2e) and exogenous TFEB-GFP (Fig. S2f). Allosteric mTORC1 inhibition with rapamycin significantly suppressed the nuclear localization of both TFEB and TFE3, while the ATP-competitive dual mTORC1/2 kinase inhibitor torin did not significantly affect TFEB localization though it did suppress TFE3 nuclear localization (Fig. 2e–g). The increased TFEB and TFE3 nuclear localization in TSC2 KO cells resulted in a significant increase in MiT/TFE transcriptional activity, as measured by 4XCLEAR promoter (containing 4 tandem copies of a CLEAR promoter element), luciferase reporter assays (Fig. 2h), These effects were confirmed in vivo using TSC2 KO cells grown as allogenic xenografts in NSG (NOD scid gamma) mice where there was elevated TFEB and TFE3 nuclear localization compared to WT controls by genetically validated immunohistochemistry (Fig. 2i, j). We then examined tumor xenograft growth in WT and TSC2 KO cells in response to treatment with rapamycin and torin. Rapamycin significantly decreased tumor growth in both WT and TSC2 KO xenografts (Fig. 2k, Figs. S3a, S3b) consistent with decreased TFEB nuclear localization. In contrast, torin did not significantly affect tumor growth (Fig. 2k), consistent with the lack of effect on TFEB nuclear localization, despite suppressed phosphorylation of classic mTORC1 substrates (p-4E-BP1 and p-p70S6K) in TSC2 KO tumor lysates (Fig. 2l and Fig. S3a, b). Finally, Tsc2-null TTJ-parental cells showed an increase in nuclear: cytoplasmic TFEB in nuclear-fraction immunoblots, compared to Tsc2 re-expressing cells (Fig. S2g), thus validating our findings in an orthogonal renal tumor model system. We next examined murine tumors with constitutive mTORC1 activity to determine MiT/TFE nuclear localization in vivo. In murine renal cystadenomas with Tsc2 loss, nuclear localization of TFEB and TFE3 was increased compared to surrounding normal renal tubules (Fig. 3a), consistent with the previously observed increase in MiT/TFE transcriptional activity (Fig. 1f, g) and this difference was statistically significant on digital quantification (Fig. 3b). We examined two additional mouse models of conditional Tsc1/2 loss-induced renal tumorigenesis: (a) renal cystadenomas from Pax8 Cre;Tsc2fl/wt mice, where Tsc2 protein loss was also confirmed, consistent with spontaneous bi-allelic inactivation of Tsc2 (Fig. 3c), and (b) Tsc1−/− renal cystadenomas from Rosa(ER)Cre;Tsc1fl/fl mouse treated with 4-OHT (Fig. 3d). Both these mice models also showed similarly elevated TFEB/ TFE3 nuclear localization. We next directly examined whether elevated MiT/TFE transcriptional activity drives lysosomal biogenesis and/or increases the proliferation of TSC1/2-null cells and tumors. We used CRISPR-Cas9 genome editing to knockout TFEB, TFE3, or both in TSC2 KO cells, with cells expressing nontargeting gRNA as controls. In TSC2 KO cells, knockout of TFE3 (but not TFEB) was sufficient to reduce LAMP1 expression (Fig. 4a). Importantly, inactivation of TFEB, TFE3, or both, decreased the phosphorylation of mTORC1 substrates S6K and 4E-BP1, similar to what has been reported in TSC2-intact systems. Simultaneous knockout of both TFEB and TFE3 (but not either gene alone) consistently decreased the proliferation of the TSC2 KO clones in in vitro confluence and viability assays (Fig. S4a, b), and also reduced in vivo growth of subcutaneous xenografts (Fig. 4b, c, Fig. S5), consistent with a role for MiT/TFE proteins as potential drivers of tumor progression in cells with constitutive mTORC1 activation. The results of the xenograft experiments suggested that TFEB and TFE3 could partially compensate for one another in the context of TSC2 loss, consistent with findings that have been reported in other systems in a context-specific manner. In order to further characterize potential functional redundancy of TFEB and TFE3 in the TSC2 KO background, we performed RNA-seq on tumor xenografts grown from WT, TSC2 KO, and TFEB KO, TFE3 KO and double TFEB/TFE3 KO cells (Fig. 4d). Multiple TFEB- and/or TFE3-regulated lysosomal gene sets were positively enriched by GSEA in TSC2 KO compared to WT xenografts, validating our initial in vitro findings (Fig. 1b). While TFEB/TFE3 double KO xenografts in the TSC2 KO background showed significant negative enrichment of all lysosomal gene sets compared to TSC2 KO xenografts, this effect was partially attenuated by TFEB or TFE3 KO alone. These results are largely in line with observed effects on tumor growth. In contrast to the apparent redundancy of TFEB and TFE3 for tumor growth in vivo, in vitro results (Fig. 4a) suggested that single TFEB or TFE3 KO may be sufficient to dampen the phosphorylation of direct mTORC1 substrates (p-4E-BP1 and p-p70S6K). Notably, a prior study demonstrated that MiT/TFE activation and CLEAR-mediated gene transcription may lead to increased RagD expression (since RRAGD contains a CLEAR element in its promoter) and thus contribute to increased mTORC1 signaling. We examined TSC2 KO cells and xenografts and observed that, consistent with our finding of CLEAR activation in these cells, RRAGD mRNA and RagD protein levels were increased in TSC2 KO and TSC1,2 KO cells and xenografts, compared to their WT counterparts (Fig. 1c, d and Fig. S6a, b). In TSC2 KO cells, compared to scrambled control shRNA-transfected cells, RRAGD shRNA partially suppressed canonical mTORC1 substrate phosphorylation by immunoblotting (Fig. S6c). Significantly, single TFE3 or TFEB KO or double TFEB/TFE3 KO (Fig. S6d), suppressed RRAGD transcript levels in TSC2 KO cells, paralleling the diminished mTORC1 signaling in vitro (Fig. 4a). These results indicate that similar to previously reported findings in WT cells, increased RRAGD transcription downstream of MiT/TFE activity may contribute to mTORC1 signaling in TSC2 KO cells, and the resulting downregulation in RRAGD transcription in the TFEB and TFE3 KO clones may be sufficient to partially dampen mTORC1 signaling in vitro. Taken together, our results suggest that MiT/TFE factors have partially redundant functions in a context-dependent and readout-dependent fashion in TSC2 KO cells. MiT/TFE subcellular localization is primarily regulated by serine phosphorylation mediated by mTORC1 (at S211 and S122 in TFEB and S321 in TFE3), resulting in 14-3-3 binding and cytoplasmic sequestration. TFEB phosphorylation occurs downstream of nutrient inputs to mTORC1, at S211 and S122 (in response to amino acids) and at S122 (in response to serum). Unlike other mTORC1 substrates, TFEB phosphorylation is insensitive to growth factor signaling, does not require RHEB, nor is it increased upon short-term TSC2 knockdown. Indeed, TFEB phosphorylation was responsive to amino acids and unaltered with glucose or serum addback in WT cells (Fig. S7a). Paradoxically, however, we found that endogenous p-TFEB (S211) and p-TFEB (S122) levels were markedly decreased in TSC2 KO cells, comparable to levels seen with amino-acid starvation (Fig. 5a, b) or mTOR kinase inhibition in WT cells (Fig. 5c). Though TFEB phosphorylation in TSC2 KO cells was unresponsive to exogenous amino-acid addback in the starved state (Fig. 5a, b, Fig. S7a), it did respond to cycloheximide (CHX) treatment in the fed state (Fig. S7b), which increases intra-cellular amino-acid levels by decreasing amino-acid utilization in protein synthesis. TFEB hypophosphorylation was additionally observed in vivo in TSC2 KO tumor xenografts (Fig. 5d), Tsc2-deficient, TTJ-parental cells (Fig. S7c), and TSC2-deficient human TRI-102 cells (an E6/E7 and hTERT- immortalized derivative of human TSC2-null, primary renal angiomyolipoma cells,) (Fig. S7d), compared to their TSC2/ Tsc2 expressing, wild-type counterparts, respectively. GFP-tagged, exogeneous TFEB was also hypophosphorylated in TSC2 KO and TSC1/2 KO cells compared to WT (Fig. S7e). TFEB hypophosphorylation was reversed by rapamycin but not torin in TSC2 KO cells (Fig. 5a–c) and xenografts (Fig. S3b), consistent with our findings for TFEB nuclear localization (Fig. 2c–g). Accordingly, immunoprecipitation experiments confirmed decreased interaction of TFEB with cytosolic 14-3-3 proteins in TSC2 KO cells, and there was reversal of this finding with rapamycin but not torin (Fig. 5e, Fig. S7f). Cumulatively, these results indicate that increased TFEB/TFE3 nuclear localization in TSC2 KO cells is due to TFEB hypophosphorylation at mTORC1-sensitive phospho-sites and is paradoxically reversible by treatment with the allosteric mTOR inhibitor, rapamycin. Both TFEB and Raptor are recruited to the lysosome via active Rag GTPases in response to amino-acid stimulation. TFEB hypophosphorylation was reversible with rapamycin treatment (Fig. 5a–c) consistent with the fact that lysosomal recruitment of TFEB and Raptor are both increased by mTOR inhibition. Thus, we hypothesized that RagC/D-mediated recruitment of TFEB to the lysosome might be impaired in cells with TSC2 loss via a mechanism mediated by rapamycin-sensitive mTORC1 activity. By immunoprecipitation, the interaction of TFEB-GFP with endogenous RagA, RagC, and Raptor was decreased in TSC2 KO compared to WT cells, and this was reversible with rapamycin (Fig. 5e). To begin to test our hypothesis that Rag activity might be constrained by an mTORC1-mediated mechanism in TSC2 KO cells, we first examined lysosomal fractions by immunoblotting. Both TFEB and Raptor levels were decreased at the lysosome in TSC2 KO compared to WT cells, and levels were increased upon mTORC1 inhibition with either rapamycin or torin (Fig. 6a, Fig. S8a). Rag GTPases cycle on and off the lysosome rapidly in response to nutrients, in a manner governed by their activation status. In WT cells, lysosomal localization of RagC was strongly increased with amino-acid starvation and decreased by amino-acid addback, as previously described (Fig. S8a). However, this response was significantly blunted in TSC2 KO cells and reversed with rapamycin, consistent with a defect in RagC activation. Next, we examined whether expression of active Rag mutants, many of which are also tethered to the lysosome, might rescue TFEB phosphorylation in TSC2 KO cells. Strikingly, expression of active RagCGDP (RAGC S75L) or active RagDGDP (RAGD S77L), but not inactive RagCGTP (RAGC Q120L), completely rescued TFEB phosphorylation at S211 and S122 in TSC2 KO cells, without substantially increasing S6 phosphorylation in these cells (Fig. 6b, c), and also promoted TFEB phosphorylation in WT cells as expected (Fig. S8b). Interestingly, overexpression of active RagD further boosted 4EBP1 phosphorylation in TSC2 KO cells (Fig. 6b), consistent with its role as a driver of canonical mTORC1 substrate phosphorylation. Overexpression of active RagBGTP (RAGB Q99L) in TSC2 KO cells, also promoted partial rescue of TFEB phosphorylation, albeit to a lesser extent than active RagC (Fig. S8c), but co-expression of active RagB/C heterodimers was not additive over active RagC alone (Fig. S8c). Moreover, co-expression of inactive RagCGTP with active RagBGTP failed to stimulate TFEB phosphorylation, while co-expression of inactive RagBGDP with active RagCGDP still promoted significant TFEB phosphorylation in TSC2 KO cells (Fig. S8c). Taken together, these data underscore the specific dependence of TFEB phosphorylation on RagC/D activation in TSC2 KO cells, as previously described for wild-type cells. Finally, both active RagC and D resulted in cytosolic relocalization of TFEB in cells with TSC2 loss (Fig. 6d). Since active Rag mutants were sufficient to rescue TFEB phosphorylation in TSC2 KO cells, we next asked whether Rag-mediated lysosomal recruitment was required for the observed rescue of TFEB phosphorylation in the context of mTORC1 inhibition with rapamycin. The N-terminal domain of TFEB is essential for Rag binding and lysosomal recruitment. Accordingly, TFEB mutants a) that lack the first 30 amino acids (Δ30-TFEB-GFP), or b) contain point mutations within the first 30 residues (TFEB-S3A, R4A-GFP), do not bind Rags, fail to relocate to the lysosome on mTORC1 inactivation and are constitutively nuclear localized. As expected, Δ30-TFEB and S3A, R4A-TFEB were hypophosphorylated at S211 in WT cells, compared to GFP-WT TFEB (Fig. 6e), and we also found TFEB to be hypophosphorylated at S122 in these cells. Similar to endogenous TFEB, rapamycin increased p-TFEB S211 and S122 on GFP-WT TFEB in TSC2 KO cells, but this effect was not seen for Δ30-TFEB and S3A, R4A-TFEB, indicating that the N-terminal Rag-binding domain is essential for rapamycin-induced TFEB phosphorylation in the context of TSC2 loss. TFEB is an atypical mTORC1 substrate that lacks the conventional TOR signaling (TOS) motif known to mediate mTORC1 substrate recruitment via direct binding to Raptor, and instead relies on the Rag GTPases for lysosomal recruitment. To bypass the requirement of RagC/D for TFEB recruitment in TSC2 KO cells, we leveraged a previously described TFEB substitution-mutant chimera wherein the first 30 amino acids of TFEB are replaced with the first 30 amino acids of S6K containing the TOS motif (GFP-TOS-Δ30-TFEB). In contrast to GFP-WT TFEB, there were similarly high levels of phosphorylation of GFP-TOS-Δ30-TFEB in TSC2 KO compared to WT cells. In contrast, TOS-(F5A)-Δ30-TFEB-GFP (a variant of TOS-Δ30-TFEB-GFP in which a key phenylalanine residue of the TOS motif has been mutagenized to alanine), was equally hypophosphorylated in both WT and TSC2 KO cells (Fig. 6f). Cumulatively, these data indicate that lysosomal recruitment and phosphorylation of TFEB are constrained in a RagC/D- and mTORC1-dependent manner in TSC2 KO cells. The RagC/D GTPase activating protein (GAP) folliculin (FLCN) is essential for MiT/TFE phosphorylation due to their dependence on Rag recruitment. Consequently, FLCN loss promotes TFEB/TFE3 nuclear localization. On nutrient starvation, recruitment of FLCN and its binding partner, FNIP2 (the FLCN:FNIP2 complex), to the inactive Rag heterodimer at the lysosome results in formation of a stable lysosomal folliculin complex (LFC). The RagC GAP activity of FLCN is inhibited within the LFC, and disruption of the LFC upon amino-acid exposure has been shown to be critical for mTORC1 activation of the MiT/TFE factors. We first examined whether destabilization of the LFC was sufficient to restore TFEB phosphorylation in TSC2 KO cells. We leveraged two recently described FLCN mutations with opposing effects on RagC GAP activity: (a) The FLCNF118D mutation at the FLCN-RagA interface, which fails to assemble the LFC and exhibits uninhibited GAP activity, and (b) The FLCNR164A mutation that assembles into a normal LFC but lacks GAP activity. We generated stable cell lines expressing these mutants in WT and TSC2 KO cells depleted for endogenous FLCN with shRNA. Compared to WT cells expressing scrambled control shRNA, FLCN depletion in WT cells (first four lanes of Fig. S9a), was sufficient to de-phosphorylate TFEB, as expected, and similar results were observed with FNIP2 depletion (Fig. S9b) indicating that both components of the dimer are required for TFEB phosphorylation. Stable expression of FLCNF118D in FLCN-depleted, TSC2 KO cells also partially restored TFEB phosphorylation, while FLCNR164A had no effect (lanes 7–10 of Fig. S9a). Thus, TFEB phosphorylation in TSC2 KO cells remains sensitive to destabilization of the LFC in response to diverse stimuli, which suggests that that endogenous FLCN/FNIP2 and RagC activity can be at least partially reactivated in these cells. In response to amino-acid starvation, the FLCN:FNIP2 complex gets recruited to the lysosome, where it interacts with the Rag-Ragulator complex, and this lysosomal relocalization is an essential prerequisite for its role as a RagC/D GAP upon nutrient restimulation leading to subsequent mTORC1 activation. We examined the lysosomal recruitment of FLCN and FNIP2 in TSC2 KO cells via specific interaction of this complex with components of Rag-Ragulator in immunoprecipitation experiments. In WT cells transiently expressing Lamtor1-GFP, the binding of FLCN and FNIP2 was increased in Lamtor1 immune-precipitates, following starvation as expected (Fig. 7a, b). However, this interaction was significantly compromised in TSC2 KO cells and was completely rescued upon mTORC1 inhibition with rapamycin or torin (Fig. 7a–c, Fig. S9c, d). These findings suggested that the failure of FLCN:FNIP2 to localize to the lysosome and activate the Rag heterodimer in TSC2 KO cells could underlie TFEB hypophosphorylation in this context. To test this hypothesis, we examined whether lysosomal relocalization or activation of FLCN:FNIP2 could rescue TFEB phosphorylation in TSC2 KO cells. It has been noted in previous studies that concurrent overexpression of exogenous FLCN and FNIP2 leads to constitutive lysosomal localization of FLCN, independent of nutrient conditions. Similarly, FNIP proteins are also critical for recruitment of HA-tagged FLCN expressed from an endogenous locus. Corresponding to the observed defect in FLCN:FNIP2 recruitment to the lysosome in TSC2 KO cells, transient co-expression of WT FLCN and FNIP2 (but not either one alone), fully rescued TFEB phosphorylation in parental TSC2 KO cells (lanes 3–5 of Fig. 7d). In FLCN-depleted TSC2 KO cells, co-transfection of FNIP2 with stably-expressed WT FLCN or activated FLCNF118D (lanes 8–9 or 12–13 of Fig. 7d, lanes 9–13 of Fig. S9a) led to a rescue of p-TFEB, while a similar rescue was not seen with inactive FLCNR164A as a negative control. To confirm the functional significance of these results, we examined the subcellular localization of TFEB/TFE3 in cells expressing FLCN mutants. TSC2 KO cells depleted for FLCN with shRNA, and stably expressing FLCNF118D, partially suppressed nuclear localization of TFEB/TFE3, and this effect was further enhanced by transient co-expression of FNIP2 in these cells (Fig. 7e, f). Interestingly, total expression of FNIP2 was decreased in vivo in renal tumors in Tsc2 ± and Pax8 Cre; Tsc2fl/wt mice by immunohistochemistry (Fig. 7g). We also examined tumor xenograft growth in FLCN-depleted, TSC2 KO cells stably expressing FLCNF118D or a combination of FLCNF118D/ FNIP2. Tumor growth was modestly, although non-significantly, suppressed in cells stably expressing FLCNF118D/ FNIP2 consistent with the more attenuated rescue of TFEB phosphorylation in vivo (Fig. S9e, f) compared to what was observed in vitro, potentially due to issues with long-term expression of these constructs. Taken together, these data suggest that the FLCN:FNIP2 complex fails to be recruited to the lysosomal membrane during starvation in TSC2 KO cells, in an mTORC1-dependent manner. Exogenous expression of FLCN and FNIP2 restores these proteins to the lysosomal membrane and rescues TFEB phosphorylation and cytosolic localization in cells with TSC2 loss. Nutrient-mediated substrate phosphorylation by mTORC1 is canonically associated with enhanced anabolism and reduced catabolism. However, substrate selectivity by mTORC1 and consequently, the regulation of metabolism associated with constitutive mTORC1 activity in the setting of TSC2 loss is more complex than previously appreciated. Notably, a previous study from 2011 was the first to show that lysosomal V-ATPase genes are upregulated with Tsc2 loss in MEFs. In support of this and our own studies on murine epidermal Tsc1 loss, and findings from other groups, we find that MiT/TFE transcriptional factors TFEB and TFE3 are paradoxically and constitutively activated in three independent murine and human renal cell line models of TSC1/2 inactivation with a resulting increase in lysosomal biogenesis. We validated these findings in a mouse model of spontaneous Tsc2 loss, as well as in two additional mouse models of conditional Tsc1/ 2 deletion-induced renal tumorigenesis. Thus, these findings appear to be highly conserved between MiT/TFE family members and across tissue types and species. TFEB phosphorylation at the canonical mTORC1 site Ser211 results in binding to 14-3-3 chaperone proteins, cytosolic retention and inactivation. De-phosphorylation at another mTORC1 site, S122, is simultaneously essential for TFEB nuclear localization following mTORC1 inhibition. We found that TFEB phosphorylation at S211 and S122 was paradoxically suppressed in murine and human cell line and xenograft models of TSC2 loss. Unexpectedly, mTORC1 inhibition with rapamycin actually promoted TFEB phosphorylation in TSC2 KO cells and xenografts, restored TFEB cytosolic localization and potently inhibited TSC2 KO tumor xenograft growth. A recent landmark study demonstrated that phosphorylation of TFEB unlike other mTORC1 substrates (S6K, 4EBP1, ULK1; all of which are hyperactivated with TSC1/2 loss), exclusively relies on the amino-acid-dependent recruitment by activated RagC/D. However, in contrast to this study where TFEB phosphorylation was shown to be insensitive to short-term TSC2 knockdown, we found that TFEB was hypophosphorylated in cells with genomic deletion of TSC1/2 in an amino-acid-dependent manner. Consistent with this, expression of active GDP bound RagC or RagD mutants completely restored TFEB phosphorylation and cytosolic localization in TSC2 KO cells. Mechanistically, rapamycin also promoted the interaction of TFEB with Rag GTPases and Raptor, restored lysosomal localization of mTORC1 and TFEB, and consequently increased TFEB phosphorylation in TSC2 KO cells, in a manner that required the N-terminal, Rag-binding domain of TFEB. Cumulatively, these data suggest that amino-acid-dependent RagC/D activation and TFEB phosphorylation are constrained via an auto-regulatory, rapamycin-sensitive mTORC1 negative feedback mechanism in cells with TSC2 loss. The tumor suppressor FLCN and its binding partner FNIP2 comprise the primary GAP complex activating RagC/D and are accordingly crucial for the phosphorylation of TFEB/TFE3 by mTORC1. The lysosomal recruitment of the FLCN:FNIP2 dimer, binding to the Rag-Ragulator proteins and formation of the lysosomal folliculin complex (LFC) following amino-acid starvation are critical initial steps in RagC activation. Subsequently, nutrient-induced disruption of the LFC is an essential prerequisite for FLCN-RagC and mTORC1 activation. Significantly, the lysosomal localization of FLCN:FNIP2 in response to nutrient starvation was severely compromised in TSC2 KO cells. As a result, forced lysosomal localization of the FLCN:FNIP2 complex by concurrent overexpression of both proteins, or rescue of endogenous FLCN:FNIP2 lysosomal localization by mTORC1 inhibition, completely restored TFEB phosphorylation and cytosolic localization in these cells. Notably, destabilization of the LFC induced in response to FLCNF118D also partially rescued TFEB phosphorylation, presumably by activating the low endogenous levels of lysosomal FLCN/FNIP2 in TSC2 KO cells. Collectively, these experiments support our current model: in the context of TSC2 loss, constitutive mTORC1 signaling inhibits lysosomal recruitment of FLCN:FNIP2 with nutrient starvation, thus preventing activation of RagC and TFEB recruitment to the lysosome for phosphorylation by mTORC1 once nutrients are replete. Rapamycin is sufficient to restore FLCN:FNIP2 to the lysosome, and does not inhibit mTORC1 activity towards TFEB (due to its incomplete inhibition of mTORC1), thus leading paradoxically to increased TFEB phosphorylation and inactivation. Short-term in vitro torin has similar effects on FLCN:FNIP2 localization but simultaneously potently inhibits mTORC1 activity towards TFEB, thus failing to robustly rescue TFEB phosphorylation in the context of TSC2 KO. Precisely how mTORC1 activity inhibits lysosomal recruitment of FLCN:FNIP2 remains to be elucidated. Notably, phosphorylation of both FLCN and FNIP2 has been implicated as an mTORC1-mediated negative feedback pathway in prior studies of yeast and mammalian cells, and these phosphorylation sites are rapamycin-sensitive. Future studies will examine whether post-translational modifications of FLCN:FNIP2 occur in the context of TSC2 loss and may modulate recruitment of the FLCN:FNIP2 dimer to the lysosome. Beyond expanding our current models of MiT/TFE regulation, our findings have potential clinical relevance since mTORC1 inhibition and MiT/TFE inactivation may show synergy for treatment of some tumor types with mTORC1 activation. MiT/TFE hyperactivity drives kidney cyst and tumor formation in mice and humans as well as in pancreatic cancer and melanoma. Given the weak effects of torin on TSC2 KO xenograft growth and TFEB phosphorylation, it is likely that the more potent effects of rapamycin are attributable to a combination of MiT/TFE inactivation and the expected downstream effects of mTORC1 inhibition. Thus, mTOR kinase inhibitors such as torin, which inhibit mTORC1 but fail to robustly inactivate TFEB, may be less preferable compared to rapalogs for treatment of cells with TSC2 inactivation. Although isolated CRISPR-Cas9-mediated inactivation of TFEB or TFE3 did not affect growth of TSC2 KO xenografts, it is notable that combined inactivation of both factors (similar to that achieved with rapamycin) had a significant effect, consistent with potent suppression of lysosomal genes by RNA-seq. This is largely consistent with previous studies which have suggested that functional redundancy between TFEB and TFE3 is significant, though this may not be the case in all contexts or for all readouts. For example, Tfeb KO alone was sufficient to attenuate renal tumorigenesis driven by FLCN loss. Along the same lines, loss of TFEB or TFE3 individually was sufficient to reciprocally decrease phosphorylation of mTORC1 substrates S6K and 4E-BP1 in cells with TSC2 loss. This was likely partly due to decreased transcription of the CLEAR gene target RRAGD, as has previously been documented in cells with intact TSC1/2, and further suggests that targeting the MiT/TFE pathway might potentially synergize with mTOR inhibitor therapy in this setting. Ultimately, future studies will reveal whether Tfeb and/or Tfe3 deletion is sufficient to reduce renal tumorigenesis in mice with spontaneous or inducible loss of Tsc2, and establish whether MiT/TFE targeting could be therapeutically useful in the setting of human tuberous sclerosis. Human embryonic kidney HEK293T cells with or without somatic genomic deletion (KO) of TSC1, TSC2, and TSC1/2 via CRISPR-Cas9 genome editing were a kind gift of TS Alliance and Dr. Nellist. TRI102 cells derived from a TSC2-null human AML and TRI103 cells derived from TRI102 cells stably transfected with wild-type TSC2 (pcDNA3.1 TSC2-zeo) were obtained from ATCC (Manassas, VI) (Catalog numbers: PTA-7368 and PTA-7369). TTJ cells, a Tsc2-null cell line derived from Tsc2 ± C57BL/6 mice, stably transfected with a control vector (TTJ- parental) or wild-type Tsc2 (TTJ-Tsc2) were a kind gift of Dr. Vera Krymskaya. Cell culture conditions are described in supplemental methods. Plasmids obtained from Addgene are described in supplemental methods. pLJM1 FLCN (WT) FLAG, pLJM1 FLCN (F118D) FLAG and pLJM1 FLCN (R164A) FLAG were a kind gift of Dr. Roberto Zoncu. GFP-TOS-Δ30TFEB and GFP-F5A-Δ30TFEB were a kind gift of Dr. Andrea Ballabio. Cells were transiently transfected using Lipofectamine 3000 reagent (L3000008, Thermo Fisher Scientific) according to the transfection guidelines. Lentiviral infection and generation of clones stably expressing scrambled, FLCN, FNIP2, or RRAGD shRNAs and WT/ mutant FLCN vectors was performed using the PEI method as previously described. We designed single-guide RNA (sgRNA) for 3 target sequences in the human TFE3 gene (GGCGATTCAACATTAACGACAGG, GCGACGCTCAACTTTGGAGAGGG, TCGCCTGCGACGCTCAACTTTGG) and 4 target sequences in the human TFEB gene (CAACCCTATGCGTGACGCCATGG, GCGGTAGCAGTGAGTCGTCCAGG, TGCCTAGCGAAGAGGGCCCAGGG and GAGTACCTGTCCGAGACCTA) and cloned these into the lentiCRISPR v2 vector (Addgene #52961, Watertown, MA, USA). Lentivirus was produced as previously described and TSC2 KO cells were infected for 48 h and selected with puromycin (1 μg/mL) for 10 days, and monoclonal cell colonies established. Immunoblotting was used for confirmation of TFEB and TFE3 KO. Antibodies and reagents used in this study are described in supplemental methods. Animal protocols were approved by the JHU Animal Care and Use Committee. The following strains were used: (1) The Tsc2 ± A/J mice, heterozygous for a deletion in exons 1–2 were a kind gift from David Kwiatkowski (Harvard University, Boston, USA) and have been previously characterized, (2) Mice carrying loxP sites flanking exon 17 and 18 of Tsc1(Stock Number 005680, Tsc1tm1Djk/J) (The Jackson Laboratory). (3) Mice carrying loxP sites flanking exon 2, 3, and 4 of Tsc2(Stock Number 027458, Tsc2tm1.1Mjgk/J) (The Jackson Laboratory). (4) Mice bearing a tamoxifen-inducible Cre recombinase driven by the endogenous mouse Gt(ROSA)26Sor promoter (Stock Number: 004847, R26CreER) (The Jackson Laboratory). (5) Mice heterozygous for the Pax8 cre recombinase knockin gene (Stock Number: 028196, Pax8cre) (The Jackson Laboratory). Conditional deletion of Tsc1 was obtained by tamoxifen treatment of Rosa(ER)-Cre; Tsc1fl/fl mice as previously described. Renal tubular-specific deletion of Tsc2 was obtained by crossing heterozygously-expressing Pax8-cre mice with Tsc2fl/wt mice to generate Pax8-Cre; Tsc2 fl/wt mice. Genotyping primers and xenograft studies are described in supplemental methods. Laser-capture microdissection was performed as described in supplemental methods. Histology and immunostaining was performed as described in supplemental methods. Cell and tissue lysis protocols, immunoblotting, immunoprecipitation, nuclear fractionation assays are described in supplemental methods. Uncropped and unprocessed western blots for all the figures have been provided as a separate PDF within the Source Data file. Lysosomal fractionation assays were carried out as previously described and in supplemental methods. Luciferase reporter assays are described in supplemental methods. RNA isolation and quantitative real-time RT-PCR and primers used are detailed in supplemental methods. Immunofluorescence protocols and quantification techniques are described in supplemental methods. RNA-sequencing of triplicate cell line and xenograft replicates was performed at Novogene and carried out as previously described. Raw RNA-seq counts were FPKM-normalized and used for plotting while DESeq2 was employed to determine differentially expressed genes. Gene Set Enrichment Analysis, GSEA: (http://www.broad.mit.edu/gsea/) was used to evaluate whether lysosomal gene expression was differentially regulated in cell lines and xenografts. Raw RNA-seq counts from our experiments were used as input in GSEA together with the following seven literature-curated gene sets for lysosomal activity: (1) A set of 435 lysosomal genes from The Human Lysosome Gene Database (hLGDB) was used. Separately, hLGDB sets for TFEB binding sites (n = 70 genes) and CLEAR motifs (n = 69 genes) were also employed for GSEA analyses, (2) We then used CLEAR and TFEB gene sets from ref. 31, which were composed of 95 and 299 genes, respectively, (3) A set of 135 genes from ref. 41 was used to investigate whether our expression data was associated with MITF/TFE-related autophagy pathways. Lastly, MITF-related gene sets were obtained from ref. 42. The seven sets were compared to the universe of all the arrays, collapsed to genes, and provided Normalized Enrichment Score (NES) and q-values for each individual set compared with our expression data. Negative NES indicated negatively enriched pathways in our comparison group vs. control. Q-value cutoffs were set to 0.1. Cathepsin D enzyme activity assays were performed as described in supplemental methods. Cell viability assays were performed as described in supplemental methods. Confluence assays were performed as described in supplemental methods. For image analysis, RNA and protein quantification, luciferase assays and xenograft studies, statistical significance was determined using the unpaired, two-tailed Student’s t-test when comparing two experimental groups, or with one-way ANOVA with Dunnett’s or Bonferroni’s correction when comparing 3 or more experimental groups. Mean values were performed in GraphPad Prism (version 8.2.1). p-values of <0.05 were considered statistically significant. All experiments were repeated at least three times (independent biological replicates) with similar results. Additionally, all experiments with the TFEB, TFE3, and dual TFEB/TFE3 KO cells were performed using multiple clones, and multiple orthogonal techniques were utilized to ensure rigor. For example: (a) TFEB/TFE3 nuclear localization was confirmed by immunofluorescence (cells), IHC (mouse renal tumors and xenografts) and immunoblotting of nuclear-cytoplasmic fractions, (b) increased TFEB/TFE3 transcriptional activity was confirmed by qRT-PCR, immunoblotting of lysosomal proteins, RNASeq and 4XCLEAR promoter activity assay. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Peer Review File Reporting Summary

PMC9649703

Guowen Lin,Tianrun Huang,Xiaobo Zhang,Gangmin Wang

Deubiquitinase USP35 stabilizes BRPF1 to activate mevalonate (MVA) metabolism during prostate tumorigenesis

10-11-2022

Prostate cancer,Ubiquitylation

The mutual interplay between epigenetic modifications and metabolic rewiring contributes to malignant features of prostate adenocarcinoma (PRAD). This study aimed to uncover the biological roles of deubiquitylase USP35 in PRAD and find effective epigenetic or metabolic targets. Bioinformatic tools or methods revealed that USP35 is upregulated in PRAD samples and correlates with inferior prognosis. The in vitro and in vivo assays suggested that USP35 could enhance malignant features of PRAD cells. Mechanistically, we found that USP35 could directly deubiquitinate and stabilize BRPF1 proteins. USP35 depends on accumulated BRPF1 proteins to accelerate cell growth, stem-like properties, and migration in vitro and in vivo. Interestingly, high BRPF1 could bind to promoter of SREBP2 and activate the SREBP2 transcriptional capacity. Therefore, USP35/BRPF1 aixs could promote expressions of mevalonate (MVA) metabolism signature in a SREBP2-dependent manner. USP35 depends on BRPF1 to maintain the activity of mevalonate metabolism in PRAD cells. Last of all, we observed that targeting BRPF1 or using MVA inhibitor (atorvastatin) are effective to suppress USP35high PRAD in vivo tumor growth. USP35 is an indicator of MVA metabolic signature in PRAD. Collectively, our study highlighted the USP35/BRPF1/SREBP2 axis in modulating MVA metabolism in PRAD, suggesting the significance of BRPF1 or MVA as the potential therapeutic targets for PRAD treatment.

Deubiquitinase USP35 stabilizes BRPF1 to activate mevalonate (MVA) metabolism during prostate tumorigenesis The mutual interplay between epigenetic modifications and metabolic rewiring contributes to malignant features of prostate adenocarcinoma (PRAD). This study aimed to uncover the biological roles of deubiquitylase USP35 in PRAD and find effective epigenetic or metabolic targets. Bioinformatic tools or methods revealed that USP35 is upregulated in PRAD samples and correlates with inferior prognosis. The in vitro and in vivo assays suggested that USP35 could enhance malignant features of PRAD cells. Mechanistically, we found that USP35 could directly deubiquitinate and stabilize BRPF1 proteins. USP35 depends on accumulated BRPF1 proteins to accelerate cell growth, stem-like properties, and migration in vitro and in vivo. Interestingly, high BRPF1 could bind to promoter of SREBP2 and activate the SREBP2 transcriptional capacity. Therefore, USP35/BRPF1 aixs could promote expressions of mevalonate (MVA) metabolism signature in a SREBP2-dependent manner. USP35 depends on BRPF1 to maintain the activity of mevalonate metabolism in PRAD cells. Last of all, we observed that targeting BRPF1 or using MVA inhibitor (atorvastatin) are effective to suppress USP35high PRAD in vivo tumor growth. USP35 is an indicator of MVA metabolic signature in PRAD. Collectively, our study highlighted the USP35/BRPF1/SREBP2 axis in modulating MVA metabolism in PRAD, suggesting the significance of BRPF1 or MVA as the potential therapeutic targets for PRAD treatment. As the leading reason of cancer-associated death among males worldwide, prostate cancer remains to be the second most commonly diagnosed malignancy [1, 2]. According to the 2022 cancer statistics, the estimated new cases would come up to 268,490, along with cancer-related 34,500 deaths [3]. Although patients with localized tumors in early stages commonly obtain favorable prognosis, the 5-year progression-free survival in metastatic cases dramatically decreases to nealry 30% [4, 5]. The androgen-deprivation (ADT) therapy is regarded to be the standard treatment for advanced prostate cancer, but it has limited efficacy in castration-resistant prostate cancer (CRPC) [1, 6]. Meanwhile, limited treatment efficacy correlates tightly with the development of even more specific subtypes, like neuroendocrine prostate cancer (NEPC) [7]. As a result, how to identify effective targets to treat lethal prostate cancer is a highly desired issue to be elucidated. Normal protein degradations and turnover contribute to the balance of cellular homeostasis and participate in various biological events [8]. As reported, lysosomal-mediated proteolysis and proteasome-mediated degradation are the two essential proteolytic manners which are highly conserved in eukaryotes [9]. The ubiquitin proteasome system (UPS) contributes to destructions of nearly 80% intracellular proteins, thus manipulating a series of biological events, like cell cycle, migration, metastasis, and drug resistance [10, 11]. Intensive studies have reported that aberrant UPS could lead to prostate cancer progression and aggressive features. Of note, the gene encoding the E3 ubiquitin ligase substrate-binding adaptor SPOP is frequently mutated in primary prostate cancer that leads to accumulations of multiple onco-proteins, including BRD4, AR, NANOG, Caprin1, or GLI [12, 13]. Besides, the E3 ubiquitin ligase STUB1 is down-regulated in prostate cancer, leading to activated JMJD1A/AR signaling that enhances cancer progression and enzalutamide resistance [14]. In contrast, a series of deubiquitinating enzymes (DUBs) exert functions in the processes during prostate tumorigenesis that have provided therapeutical targets for treatment [15]. For instance, USP16, known as a deubiquitinase, was strongly associated with the c-Myc gene signature to serve as a novel deubiquitinase of c-Myc, thereby enhancing the castration-resistant prostate cancer cell proliferation [16]. The deubiquitinase BAP1 is reported to physically bind to and deubiquitinate PTEN, playing an essential role in prostate cancer suppression [17]. Furthermore, USP33 is found to be overexpressed in prostate cancer cells that could inhibit the Lys48 (K48)-linked polyubiquitination of DUSP1, mediating the docetaxel resistance of CRPC [18]. Therefore, we speculated that whether there exist other DUB members that influence progression of prostate cancer. In the current study, we found that USP35 is a novel oncogenic DUB in prostate tumorigenesis, which has never been reported. We reported that BRPF1 is a substrate of USP35 and USP35/BRPF1 axis promotes malignant features of prostate cancer via activating the MVA pathway. Our data collectively suggested the possible therapeutic implications of targeting USP35/BRPF1 as an innovative strategy to improve the overall prognosis in prostate cancer patients. First of all, we queried the expression data of USP35 in GDS2545 derived from the Gene Expression Omnibus (GEO) database, observing that USP35 levels were higher in 65 tumor samples as compared to 63 adjacent normal tissues (Fig. 1A). The clinical information of PRAD samples was summarized in Table S1. In line with the findings, we also analyzed that USP35 expressed highly in PRAD samples in TCGA-PRAD and Oncomine datasets, respectively (Fig. 1B, C). Besides, we collected the information of clinical characteristics for the PRAD patients to conduct the correlation analysis. As expected, high USP35 expressions were observed in patients with advanced T or N stages (Fig. 1D, E). Also, patients with a definite biochemical recurrence or ≥8 Gleason scores showed elevated levels of USP35 (Fig. 1F, G). Moreover, Kaplan–Meier survival curves analysis revealed that patients with high USP35 levels bear worse disease-free survival (DFS) in TCGA-PRAD cohort (p < 0.001), worse overall survival (OS) in GSE70769 (p = 0.021), as well as worse progression-free survival (PFS) in GSE1169181 (p = 0.017) (Fig. 1H–J). Lastly, we also conducted the multi-variate Cox regression analysis by integrating several hazard clinical variables in TCGA-PRAD cohort. Compared with other variables, like age, TN stages, or Gleason scores, USP35 is an independent prognostic factor for predicting DFS of PRAD patients (Fig. 1K). The time-dependent receiver operating characteristic curve (ROC) further demonstrated that combination of USP35 and other clinical variables could reach higher predictive efficiency, as compared to USP35 or clinical variables respectively (Fig. 1L). In conclusion, our bioinformatic analysis suggested that USP35 is up-regulated in PRAD samples and possesses a tight correlation with a worse prognosis. Based on the above bioinformatic findings, we intended to assess the USP35 functions in prostate cancer. First of all, we deleted USP35 in two prostate cancer cell lines C4-2b and PC-3 via sgRNA-mediated CRISPR/Cas9 KO technology (Fig. 2A). In contrast, we generated stable USP35-overexpressing cell lines via lentivirus infection (Fig. 2B). Next, we started to determine the roles of USP35 in PRAD proliferation. The colony formation assay exhibited that USP35 overexpression notably promoted the number and sizes of PRAD cell colonies in two cell lines (Fig. 2C). USP35 depletion could decrease PRAD cell proliferation rates relative to parental control cells that could be restored by USP35 overexpression, as suggested by the cell viability assays (Fig. 2D). Meanwhile, we also found that USP35 could further enhance stem cell-like properties in PRAD cells, as evidenced by the sphere formation assays (Fig. 2E). The wound-healing assay also indicated that USP35 deletion resulted in decreased migration rates of cells relative to parental control cells (Fig. 2F). Conversely, ectopic expression of USP35 robustly potentiated the migration efficacy of cells as compared to cells transfected with vector (Fig. 2G). Transwell assays also indicated that ectopic expression of USP35 in USP35-deficient cells could completely rescue the restricted migration abilities (Fig. 2H). Last of all, to further evaluate USP35 roles in prostate tumorigenesis, we generated an orthotopic prostate tumor model in which C4-2b cells were injected into the prostate gland of nude mice. The in vivo bioluminescence (BIL) signals were used to indicate the growth of orthotopic prostate tumors. In accordance to our speculations, USP35 overexpression could result in the formation of larger prostate tumors relative to those in control mice, suggesting that USP35 significantly promoted tumorigenesis (Fig. 2I). Collectively, these data indicated that USP35 promotes the proliferation, migration, and stemness properties of PRAD cells both in vitro and in vivo. To clarify the downstream targets that exert the oncogenic functions of USP35 in PRAD, we generated the stable C4-2b cells with double-tagged FLAG-HA-USP35. Next, the USP35-containing protein complex was purified and isolated via Tandem Affinity Purification (TAP) method. We thus identified a list of peptides in the complex, including FZR1, TNIP2, MAVS, or BRPF1 (Fig. 3A). Considering that epigenetic deregulation contributes to PRAD progression and BRPF1 is less reported in PRAD, we thus decided to confirm the associations between USP35 and BRPF1. First of all, the co-immunoprecipitation (co-IP) analysis with an anti-USP35 indicated that USP35 could directly interact with BRPF1, implicating the endogenous interactions (Fig. 3B). Besides, we also detected that the BRPF1 proteins were decreased in USP35-deleted C4-2b cells relative to parental control cells (Fig. 3C). However, USP35 loss could not induce alterations of BRPF1 mRNA levels (Fig. 3C). Next, we observed an increased levels of BRPF1 proteins when C42-B cells were transfected with elevated doses of Myc-USP35 plasmids, suggesting that USP35 promotes BRPF1 levels in a dose-dependent manner (Fig. 3D). Intriguingly, high USP35 did not alter BRPF1 mRNA levels (Fig. 3D). Consistently, the cycloheximide (CHX) assays also indicated that USP35 deletion resulted in shorter half-life of BRPF1 proteins relative to those in parental control C4–2B cells (Fig. 3E). In contrast, USP35 overexpression significantly prolonged the half-life of BRPF1 proteins in C4-2B cells (Fig. 3F). As shown in Fig. 3G, only wild-type USP35, but not the C450A mutant, could sufficiently catalyze deubiquitination of BRPF1. Accordingly, only wild-type USP35, but not the C450A mutant, could promote the accumulations of BRPF1 proteins, not altering the corresponding mRNA levels of BRPF1 (Fig. 3G). Taken together, these data indicated that USP35 can function as a deubiquitinase for BRPF1. To further determine the biological roles in PRAD, we knocked down BRPF1 in C4-2b and PC-3 cells through lentiviral transduction method using three different BRPF1 short hairpin RNAs (Fig. 4A). The MTT assay showed that BRPF1 knockdown remarkably inhibited prostate cancer cell growth (Fig. 4B, C). However, BRPF1 overexpression could notably enhance cell colony formation ability of PRAD cells (Fig. 4D). In addition, bioinformatic analysis in TCGA-PRAD samples also suggested that high BRPF1 correlated highly with hazard clinical factors, like advanced T stages, high Gleason scores, as well as biological recurrence (Fig. 4E–G). Kaplan-Meier analysis suggested that patients with high BPRF1 levels had shorter DFS months as compared to those with low BRPF1 levels, as indicated by the log-rank test (Fig. 4H). Given that we have already generated USP35-overexpressing PRAD cell lines, we further knocked down BRPF1 in these cells. In line with the previous findings, USP35 is required for the proliferation, migration, and stem cell-like properties of PRAD cells, which could be largely abolished by BRPF1 KD (Fig. 4I–K). Collectively, these data implicated that USP35 depends on BRPF1 to enhance PRAD malignant progression. Given that BRPF1 is an epigenetic regulator that could induce specific genes expressions, we thus conducted the bioinformatic analysis based on sequencing expression data derived from TCGA-PRAD samples. Gene Set Enrichment Analysis (GSEA) implicated the tight interplay between BRPF1 and mevalonate (MVA) crosstalk in PRAD (Fig. 5A). Given that SREBP2 is the master regulator of MVA crosstalk, we thus intended to figure out the relationships between BRPF1 and SREBP2. Interestingly, BRPF1 KD could significantly reduce SREBP2 mRNA levels in PRAD cells (Fig. 5B). Besides, BRPF1 overexpression could promote SREBP2 mRNA levels (Fig. 5C). Meanwhile, we cloned a list of fragments of the SREBP2 promoter, and BRPF1 increased the luciferase activities of promoter fragments of P3, P4, and P5, but not the P1 or P2 (Fig. 5D). Thus, the region, ranging from −250 to −140 in the SREBP2 promoter, is the essential sequence binded and regulated by BRPF1. We also conducted the ChIP-qPCR assay to confirm that BRPF1 could directly bind to the SREBP2 promoter region to sustain the transcriptional activity, as implicated by the active H3K4me3 modification markers (Fig. 5E). Conversely, BRPF1 KD reduced the transcriptional activity at the SREBP2 promoter, as indicated by decreased H3K4me3 enrichment (Fig. 5F). To exclude the possibilities that BRPF1 could regulate other MVA transcriptional factors (TFs) like SREBP2, we detected that BRPF1 could not induce SREBP1a/c levels in PRAD cells (Fig. 5G). Along with these findings, we found that BRPF1 could not alter SREBP1a/c transcriptional activities (Fig. 5H). Given that USP35 could stabilize BRPF1, we thus wondered whether USP35 could rely on BRPF1 to modulate SREBP2 expressions. Apparently, we found that USP35 overexpression could enhance BRPF1 binding enrichment on the SREBP2 promoter region, which were notably suppressed with BRPF1 KD (Fig. 5I). In line with the results, we found that USP35 could enhance SREBP2 levels, and the increase could be completely abolished by BRPF1 KD (Fig. 5J). Conversely, USP35 deletion could suppress SREBP2 mRNA levels, which could be completely restored by BRPF1 overexpression in C4-2b and PC-3 cells (Fig. 5K). Taken together, these data implicated that USP35/BRPF1 axis could modulate MVA crosstalk in PRAD in a SREBP2-dependent manner. Given that SREBP2 regulates a series of enzymes that sustain MVA metabolism activities, we observed that USP35 depletion could down-regulate the mRNA levels of SREBP2 downstream targets, named as MVA signature, including HMGCR, FDFT1, SQLE, MSMO1, FDPS and LSS (Fig. 6A). However, BRPF1 overexpression could completely restore the expressions of MVA signature (Fig. 6A). Meanwhile, USP35 could activate MVA signature and this effect could be completely abolished by BRPF1 KD (Fig. 6B). Consistently, we detected that the free cholesterol content in USP35-OE cells, was about 50% higher than that in controls, and the increase could be completely abolished upon BRPF1 KD (Fig. 6C). In contrast, USP35 loss suppressed the free cholesterol content and the decrease could be largely restored with BRPF1 overexpression (Fig. 6D). In addition, we utilized the atorvastatin to inhibit the MVA pathway and found that atorvastatin alone could restrict the growth of PRAD cells relative DMSO treatment (Fig. 6E). Meanwhile, in line with our biological findings, atorvastatin could further largely suppress the growth of USP35-OE PRAD cells (Fig. 6E). Given that we have found the USP35/BRPF1/SREBP2 axis in regulating MVA pathway during prostate tumorigenesis, we thus intended to further test the translational significance for treating PRAD. We generated the C4-2b-derived tumor model and found that targeting BRPF1 could significantly suppress USP35high in vivo PRAD growth, as shown by tumor volumes and weight (Fig. 6F–H). Lastly, we also found that atorvastatin could also suppress the in vivo growth of tumors derived from USP35high PRAD cells, as compared to those derived from control cells (Fig. 6I). Collectively, these results implicated that suppressing BRPF1-induced MVA crosstalk provides a therapeutical vulnerability for USP35high PRAD. Last of all, we collected some PRAD samples from our center and divided the samples into USP35high and USP35low groups. As shown in Fig. 7A, the positive associations among USP35, BRPF1, and SREBP2 were demonstrated via IHC method. We also assessed the positive relationships between USP35 and MVA signature in TCGA-PRAD samples (Fig. 7B). These data suggested that USP35 is an indicator for clinical PRAD samples with high MVA metabolism activity. We further illustrated the USP35/BRPF1/MVA axis in prostate cancer cells in Fig. 8. Recently, USP35 is emerging as a hotspot target in cancer research. Zheng Tang et al. found that USP35 is abundant in human lung cancer tissues and cell lines that modulates iron homeostasis and ferroptosis through maintaining the stability of ferroportin (FPN) proteins [19]. Besides, deubiquitinase USP35 could restrain STING-mediated interferon signaling in ovarian cancer, highlighting the potential associations between USP35 and CD8+ T cell infiltrations [20]. Moreover, AKT-activated USP35 enhanced ERα stability by interacting and deubiquitinating ERα, suggesting that USP35 may be a therapeutical vulnerability for ER+ breast cancer with endocrine resistance [21]. Considering USP35 correlates tightly with tumor progression, whether USP35 could modulate other oncogenic features, like metabolic remodeling, stem-like properties is still unknown. Besides, the roles of USP35 in PRAD have never been defined. In the current study, we utilized the bioinformatic algorithms to uncover that USP35 is also up-regulated in PRAD samples and high USP35 levels correlated with hazard clinical characteristics, including TN stages, Gleason scores, or biological replase. Patients with high USP35 have poor prognosis. We further demonstrated that USP35 enhanced cancer cell growth, migration ability and stem-like properties in vitro and in vivo. Mechanistically, USP35 interacts with and stabilizes BRPF1 proteins, relying on accumulated BRPF1 to exert its oncogenic functions. Functional enrichment analysis proposed the tight interplay between BPRF1 and MVA signaling in PRAD samples. BPRF1 could bind to promoter of SREBP2 to activate the transcriptional activity. Thus, USP35/BRPF1 axis modulated the MVA crosstalk via induction of SREBP2. Last of all, in vitro and in vivo models suggested that suppressing BRPF1-mediated MVA crosstalk proved to be effective to inhibit USP35high PRAD growth. As reported, epigenetic modifications induce heritable alterations in gene expressions without changes of DNA sequences, including DNA methylation, histone modifications, as well as microRNAs (miRNA) [22]. Previous studies have identifed a series of epigenetic regulators that impact PRAD progression or drug resistance [23, 24]. Huairui Yuan et al. reported that SETD2 methylates EZH2 which promotes EZH2 degradation and SETD2 deficiency promotes a Polycomb-repressive chromatin state that renders cells to obtain metastatic potentiality in prostate cancer [25]. Besides, SPOP-mutant prostate cancers often accumulate BET proteins, including BRD2, BRD3 and BRD4 [26]. Elevated BRD4 could mediate chromatin remodeling effect to amplify androgen receptor (AR) downstream signaling and promote progression of CRPC or BET inhibitor resistance [27]. Recently, researchers also found that bromodomain-containing protein BRD9, one subunit of SWI/SNF complex, could interact with AR and CCCTC-binding factor (CTCF) to control AR-dependent gene expressions, highlighting the translational significance of nontoxic BRD9 inhibitors in PRAD treatment [28]. The bromodomain and PHD finger containing 1 (BRPF1) belongs to an epigenetic reader, and the bromodomain of BRPF1 may contribute to the chromatin binding and target specificity of the MOZ/MORF complex. Previous studies have indicated that BRPF1 is essential for embryonic development. Besides, BRPF1 loss also decreased the expressions of multipotency genes, like Slamf1, Mecom, Hoxa9, Hlf, or Gfi1, suggesting that BRPF1 is essential for the development of fetal hematopoietic stem cells [29]. Additionally, defective mutations of BRPF1 lead to intellectual disability and facial dysmorphisms in humans [30]. However, the associations between BRPF1 and PRAD progression is unknown. In this study, we conducted the bioinformatic analysis to find that BRPF1 is an oncogenic factor that correlated with PRAD progression. High BRPF1 contributes to tumor growth and malignant progression of PRAD, which is in accordance with findings in other solid tumors, like hepatocellular carcinoma (HCC) [31]. In addition, we uncovered the epigenetic regulations between BRPF1 and SREBP2, in which BRPF1 could epigenetically elevate the transcriptional activity of SREBP2. BRPF1 could directly activate the MVA pathway that provides a novel role of BRPF1 in tumor lipid metabolism. As the essential hallmark of solid tumors, elevated cholesterol and lipid synthesis are considered to play important roles in metabolic rewiring during cancerous transformation [32]. As a precursor for bile acid and steroid hormone biosynthesis, cholesterol participated in constructions and function of cell membranes. Of note, the cholesterol-related metabolites play regulatory roles as signaling molecules in tumor progression. As well known, the mevalonate (MVA) metabolism employs acetyl-CoA to produce sterols and isoprenoids that are essential to potentiate tumor progression [33]. The sterol regulatory elementbinding protein (SREBP) family of transcription factors control the MVA pathway [34]. Mechanistically, SREBP2 is cleaved and translocated, induced by cholesterol depletion, to the nucleus to activate the transcription of MVA signature. As reported, the lipogenesis regulator SREBP2 directly induces transcription of the iron carrier Transferrin (TF), reducing reactive oxygen species (ROS), and lipid peroxidation, thereby resulting in resistance to inducers of ferroptosis in melanoma [35]. Besides, ZMYND8 and SREBP2 drive the enhancer-promoter interaction to facilitate the recruitment of Mediator complex, thus upregulating MVA pathway genes and enhancing colon cancer progression [36]. Intriguingly, in line with our results, Donge Tang et al also confirmed that KDM6A drives prostate tumorigenesis via activating SREBP1c-mediated lipid metabolism [37]. In this study, we highlighted the SREBP2-dependent MVA crosstalk is activated by USP35/BRFP1 axis to drive tumor progression. The in vitro and in vivo data suggested that targeting MVA is effective to suppress PRAD progression and further inhibited the USP35high tumor growth. Previous studies have reported that the MVA pathway inhibitors (statins) exert beneficial efficacy in some specific colon cancer patients [36]. Accordingly, we speculated that targeting MVA (statins) would a valuable strategy to ameliorate USP35high prostate cancer progression. We still have some concerns in the current study that need to be further explored. First of all, owing to limited financial foundations, we did not thoroughly assess the inhibitory efficacy of atorvastatin in more pre-clinical models. The patient-derived tumor xenograft (PDX) models with high or low USP35 were needed to assess the efficacy of atorvastatin in patients with distinct genetic backgrounds. Besides, whether USP35 regulates other biological aspects of prostate cancer remains unclear, like immune cells infiltrations or bone metastasis. In addition, the inhibitors for BRPF1 are now available, including GSK-5959, or IACS-9571 [38]. We intended to assess the clinical efficacy of BRPF1 inhibitors in PRAD models in the following researches. Last of all, we are still uncertain about the relationships between USP35/BRPF1/MVA axis and castration resistance in prostate cancer. We would design other assays in the future to confirm whether USP35-mediated MVA metabolism could influence the progression of Castration-Resistant Prostate Cancer (CRPC). We need to collect more PRAD samples to validate the translational significance of USP35/BRPF1/SREBP2 axis in prognosis classification. Taken together, our study revealed the biological roles and prognostic significance of USP35 in PRAD. USP35 deubiquitinates BRPF1 to activate SREBP2 expressions, thereby strengthening the MVA crosstalk in PRAD progression. Thorough understanding of USP35/BRPF1/SREBP2 axis provides valuable strategies for PRAD treatment and prognosis classification. We obtained the 293T, C4-2b, 22Rv1, and PC-3 cells from the American Type Culture Collection (ATCC). The 293T cells were maintained in DMEM with 10%(v/v) fetal bovine serum (FBS), while C4-2b and PC-3 cells were maintained in RPMI 1640 with 10%(v/v) FBS. The 100 clinical prostate cancer specimens from patients were obtained from the department of urology, Huashan hospital (Shanghai, China). All patients underwent radical prostatectomy resection between January 2015 and September 2020. Patients who were previously diagnosed with other cancers and who received neoadjuvant treatments before surgery were not included. Histopathological diagnoses were made according to the WHO criteria. Written informed consent was obtained from each patient in this study for use of their tissues prior to the acquisition of the specimens. Ethical approval was granted by the Research Ethics Committee of Huashan hospital. pX459 plasmid was used to clone guide oligos targeting USP35 gene. C4-2b cells were plated and transfected with pX459 constructs for 24 h. After transfection for one day, 1 μg/ml puromycin was used to screen cells for 3 days. Living cells were seeded in 96 well plate by limited dilution to isolate monoclonal cell line. The knock out cell clones are screened by Western blot. Sequences of specific sgRNAs are listed as the following: sgUSP35#1: F: 5′-CACCGCACACGACTCGCAGTAGTAG-3′, R: 5′-AAACCTACTACTGCGAGTCGTGTGC-3′. sgUSP35#1: F: 5′-CACCGCTACTACTGCTATGCCCGTG-3′, R: 5′-AAACCACGGGCATAGCAGTAGTAGC-3′. For the knockdown assay, BRPF1 shRNAs were purchased from Sigma-Aldrich (MO, USA). shRNA plasmids were co-transfected with packaging constructs according to the manufacturer’s instruction to package the lentivirus. C4-2b and PC-3 cells were incubated with lentivirus for 72 h. The KD efficiency was confirmed by qRT-PCR assay. The shRNA sequences targeting BRPF1 were listed as the following: shBRPF1#1: CCGCATCAGCATCTTTGACAA; shBRPF1#2: CGCTACTTGAACTTTGATGAT; shBRPF1#3: CGTACTTTGAGAGTCACAATA. Cell proliferation was performed using the CCK8 kit following the manufacturer’s instruction (Jiangsu KeyGENBioTECH Corp., Ltd, China). Three thousand to five thousand prostate cancer cells were seeded into each well of a 96-well plate. Ten microliter CCK8 reagent was added to each well at the indicated time and incubated at 37 °C for 2 h. The absorbance at 450 nm was recorded with a 96-well plate reader. For the colony formation assay, thousand prostate cancer cells were seeded into each well of a 6-well plate and incubated at 37 °C for 10–14 days. The cells were then fixed, stained with 0.2% crystal violet, and imaged. Clones that consisted of at least 50 cells were considered as one colony. Total RNA was isolated from cells using the TRIzol reagent (Tiangen), and cDNA was reversed-transcribed using the Superscript RT kit (TOYOBO) following the manufacturer’s instructions. PCR amplification was performed using the SYBR Green PCR master mix Kit (TOYOBO). All quantitations were normalized to the level of endogenous control GAPDH. The 4 × 104 C4-2b or PC-3 cells were seeded in each well of a 6-well plate, and cells were scratched with a 1 ml pipette tip after confluent. After washed with PBS slightly, the images were captured by using a microscope equipped with a digital camera. The images were recorded again using the same microscope after 24 h. For the in vivo ubiquitination assay, HEK293T cells were transiently co-transfected with indicated plasmids. After 24 h, cells were lysed with 100 μl lysis buffer (2% SDS, 150 mM NaCl and 10 mM Tris-HCl, pH 8.0), boiled for 20 min. 900 μl dilution buffer (150 mM NaCl, 1% Triton, 2 mM EDTA and 10 mM Tris-HCl, pH 8.0) was added. The samples were incubated with M2 or HA beads at 4 °C for 90–120 min with rotation. Then the beads were boiled after extensive washing with washing buffer (1 M NaCl, 1% NP40, 1 mM EDTA and 10 mM Tris-HCl, pH 8.0), and resolved via SDS-PAGE gel for immunoblotting analysis. The indicated cultured cells were lysed with RIPA lysis buffer (Beyotime, Shanghai, China) supplemented with 1% of the mixture of protease inhibitor (Sigma, MO, USA). Proteins were isolated and transferred to PVDF membranes (Millipore) in equivalent amounts using 10% SDSPAGE. As loading controls, antibodies against β-actin (60004-1-Ig, Proteintech) were used. Antibodies used were: USP35 (Proteintech, 24559-1-AP); BRPF1 (Abcam, ab259840); SREBP2 (Abcam, ab30682); FLAG (Abcam, ab205606); Myc (Abcam, ab32072); HA (Abcam, ab9110). Prostate cancer cells were co-transfected with SREBP2 firefly luciferase reporter plasmid, Renilla luciferase plasmid, BRPF1 and USP35 expression plasmids by polyethylenimine (PEI) according to the manufacturer’s instruction. After 24 h, cells were lysed and centrifugated, and luciferase activity was measured using the dual-luciferase reporter assay system (Promega, Madison, USA). Relative SREBP2 activity was calculated as firefly luminescence relative to Renilla. All animal experiments were conducted according to the NIH Guide for the Care and Use of Laboratory Animals. All BALB/c nude male mice (4–6 weeks of age) were obtained. All experimental procedures using animals were approved by the Institutional Animal Care and Use Committee of Fudan University. The indicated C4-2b cells (1 × 106) were mixed with matrigel (1:1) and injected subcutaneously into the flanks of BALB/c nude male mice. Tumors were measured using calipers every 5 days and tumor volumes were calculated using length × width × width × 0.52. Tumor tissues were paraffin embedded and H/E stained. For the construction of orthotopic mice model, male Balb/c athymic nude mice were acclimated for 3 weeks before experimental manipulation. Pca cells (C4-2b) were grafted into the mouse left or right DP under a stereoscopic microscope shortly after the cell resuspension in Matrigel. A 1.5 cm transverse incision was made in the lower midline with microscissors above the presumed location of the bladder. Approximately 100 μL intraperitoneal liquid spilled and was sponged with cotton balls. The intestine was pushed upward into the abdominal cavity using a sterile cotton swab. Starting 2 weeks post-injection, tumor growth was monitored by ultrasound (US) every 5 days. Meanwhile, the BIL signals within Pca tumors could be also detected to monitor the growth of orthotopic tumors. For cell proliferation, cell colony formation, and cell migration, data were analyzed by unpaired Student’s t test. Kaplan–Meier plot was used for patients with log-rank test. Two-tailed unpaired Student’s t tests or Mann–Whitney U tests were applied for comparisons between two groups. The differences with * p < 0.05 or ** p < 0.01 were considered statistically significant. Statistical analyses were performed with Prism 8.0 (GraphPad Software) and R studio (Version 3.5.3). Table S1 Original Data File

PMC9649717

Carla Letizia Busceti,Domenico Bucci,Mariarosaria Scioli,Paola Di Pietro,Ferdinando Nicoletti,Stefano Puglisi-Allegra,Michela Ferrucci,Francesco Fornai

Chronic treatment with corticosterone increases the number of tyrosine hydroxylase-expressing cells within specific nuclei of the brainstem reticular formation

28-10-2022

area postrema,retrorubral field,brainstem,Cushing's syndrome,catecholamines,glucocorticoids,glucose tolerance

Cushing's syndrome is due to increased glucocorticoid levels in the body, and it is characterized by several clinical alterations which concern both vegetative and behavioral functions. The anatomical correlates of these effects remain largely unknown. Apart from peripheral effects induced by corticosteroids as counter-insular hormones, only a few reports are available concerning the neurobiology of glucocorticoid-induced vegetative and behavioral alterations. In the present study, C57 Black mice were administered daily a chronic treatment with corticosterone in drinking water. This treatment produces a significant and selective increase of TH-positive neurons within two nuclei placed in the lateral column of the brainstem reticular formation. These alterations significantly correlate with selective domains of Cushing's syndrome. Specifically, the increase of TH neurons within area postrema significantly correlates with the development of glucose intolerance, which is in line with the selective control by area postrema of vagal neurons innervating the pancreas. The other nucleus corresponds to the retrorubral field, which is involved in the behavioral activity. In detail, the retrorubral field is likely to modulate anxiety and mood disorders, which frequently occur following chronic exposure to glucocorticoids. To our knowledge, this is the first study that provides the neuroanatomical basis underlying specific symptoms occurring in Cushing's syndrome.

Chronic treatment with corticosterone increases the number of tyrosine hydroxylase-expressing cells within specific nuclei of the brainstem reticular formation Cushing's syndrome is due to increased glucocorticoid levels in the body, and it is characterized by several clinical alterations which concern both vegetative and behavioral functions. The anatomical correlates of these effects remain largely unknown. Apart from peripheral effects induced by corticosteroids as counter-insular hormones, only a few reports are available concerning the neurobiology of glucocorticoid-induced vegetative and behavioral alterations. In the present study, C57 Black mice were administered daily a chronic treatment with corticosterone in drinking water. This treatment produces a significant and selective increase of TH-positive neurons within two nuclei placed in the lateral column of the brainstem reticular formation. These alterations significantly correlate with selective domains of Cushing's syndrome. Specifically, the increase of TH neurons within area postrema significantly correlates with the development of glucose intolerance, which is in line with the selective control by area postrema of vagal neurons innervating the pancreas. The other nucleus corresponds to the retrorubral field, which is involved in the behavioral activity. In detail, the retrorubral field is likely to modulate anxiety and mood disorders, which frequently occur following chronic exposure to glucocorticoids. To our knowledge, this is the first study that provides the neuroanatomical basis underlying specific symptoms occurring in Cushing's syndrome. Cushing's syndrome is due to increased glucocorticoid levels in the body, and it is characterized by several clinical alterations which concern vegetative and behavioral functions (Starkman et al., 1992). For instance, altered metabolism (including diabetes), increased blood pressure, sleep disorders, increased feeding, aggressiveness, anxiety, and psychosis have been described in patients suffering from Cushing's syndrome (Krieger and Glick, 1974; Shipley et al., 1992; Ntali et al., 2015). Also, increased feeding is often reported in patients with increased glucocorticoid levels, which results in an increase in body weight (Chanson and Salenave, 2010). The anatomical correlates of these effects remain largely unknown. Apart from peripheral effects induced by corticosteroids as counter-insular hormones, only a few reports allow us to hypothesize which neurobiology underlies glucocorticoids-induced vegetative and behavioral alterations. Among these, a seminal article published by Sloviter et al. (1989) showed that adrenalectomy in rats causes profound hippocampal electrophysiological alterations and a nearly complete loss of granule cells in the hippocampal dentate gyrus. These authors demonstrated that corticosterone replacement rescues electrophysiological responses and prevents cell loss of hippocampal dentate granule cells in adrenalectomized rodents. This suggests that glucocorticoids play a fundamental role in maintaining the structural integrity of the normal adult hippocampus (Sloviter et al., 1989). However, it is unlikely that multiple metabolic and behavioral alterations induced by an excess of glucocorticoids may be entirely generated by hippocampal dysfunctions. Thus, it remains to be elucidated which other brain regions may contribute to increased feeding, altered glucose tolerance, anxiety, increased blood pressure, and alterations in the sleep pattern, which characterize Cushing's syndrome. Due to the seminal role of the brainstem reticular formation in promoting the sleep-waking cycle (Moruzzi and Magoun, 1949), alertness, and anxiety, as well as the specific vegetative control of the cardiovascular system and specific abdominal organs, this area deserves specific investigations. The reticular formation contains nuclei that are responsible for sleep-waking cycle, anxiety, aggressiveness, as well as blood pressure control and, in the case of area postrema (AP), specific control of pancreatic secretion (Loewy et al., 1994). The recruitment of these domains in Cushing's syndrome questions whether glucocorticoid may alter the brainstem reticular formation. Among reticular nuclei, these effects are mainly controlled by the lateral, TH-positive column of reticular nuclei (Bucci et al., 2017, 2018). In line with this, in a recent article, we indicated a selective increase of catecholamine cells placed in the caudal part of the lateral column of the brainstem reticular formation. This evidence was limited to investigations carried out in organotypic cell cultures following corticosterone incubation (Busceti et al., 2019). Here, aiming at translating these effects into system neurobiology, we investigated ex vivo, in the whole brain, which nuclei of the brainstem reticular formation may be altered concomitantly with the occurrence of glucose intolerance and increased body weight following chronic exposure to corticosteroids in mice. Such a project encompassing a plethora of behavioral and vegetative domains cannot be solved in a single research article. Due to the recent discovery of a prominent role of AP in controlling the subdivision of the vegetative nervous system innervating the pancreas, the present study mostly focused on correlating altered TH expression in AP with altered glucose tolerance, which typically features Cushing's syndrome. For these experiments, we used 8-weeks-old C57Bl/6J male mice (N = 18) (Charles River, Calco, LC, Italy). All animals were maintained under controlled conditions (room temperature = 22°C; humidity = 40%) on a 12-h light-dark cycle with food and water ad libitum. C57Bl/6J male mice (N = 9) were chronically administered with corticosterone for 5 weeks (Sigma Aldrich, MI, Italy, code: C-2505) in the drinking water (normal drinking water was replaced with a 0.66% ethanol solution containing 100 μg/mL corticosterone). Based on the daily water intake, the daily dose of corticosterone ranges between 1.5 mg/Kg and 2 mg/Kg. Vehicle-treated mice (N = 9) were treated for 5 weeks with a 0.66% ethanol solution in the drinking water. Solutions were freshly prepared. Body weight changes were monitored weekly during the treatment period (5 weeks, Figure 1A). All mice were assessed for glucose tolerance under basal conditions and after 2 or 4 weeks of chronic treatment with corticosterone (Figure 1A), body weight was monitored. At the end of the treatment, all mice were killed and dissected brains were used for the immunohistochemical analysis of TH-positive cells in the whole rostro-caudal extension of the brainstem reticular formation (Figure 1A). All anatomical points of reference were indicated according to the atlas of Paxinos and Franklin (2001) for mice. A glucose solution (20% in 0.9% NaCl) was administered by intraperitoneal (i.p.) injection (100 μl/10 g body weight). Blood glucose was measured at five time points (15, 30, 60, 90, and 120 min after i.p. injection of glucose) during the following 4 h. Blood samples were obtained by a small incision on the paw, and glucose levels were measured by using the blood glucometer One Touch Vita (Johnson & Johnson, NY, USA). Dissected brains were fixed overnight at 4°C in Carnoy's solution (60% ethanol, 10% acetic acid, and 30% chloroform with a ratio of fixing solution to the tissue of 20:1 in weight). After fixing the tissue, brain samples were embedded in paraffin and cut with a rotative microtome (Leica, Wetzlar, Germany, code: RM 2245) to obtain 20 μm thick sections. These slices were sampled along the whole rostro-caudal extent of the brainstem reticular formation. Tissue sections were incubated overnight with a monoclonal mouse anti-TH primary antibody (1:100; Sigma Aldrich, code: T1299) and then for 10 min with a secondary biotin-coupled anti-mouse secondary antibody (1:400; Vector Laboratories, Burlingame, CA, USA code: BA-2000). 3,3-Diaminobenzidine tetrachloride (Sigma Aldrich, code: D4293-50set) was used for detection. Negative control was performed without incubation with primary antibody. Stereological analysis for all catecholaminergic nuclei was carried out on serial coronal slices sampled every 160, 80, and 40 μm for substantia nigra pars compacta (SNC), ventral tegmental area (VTA), and all other nuclei counted, respectively. This sampling paradigm was established based on a pilot analysis carried out by using different inter-slice intervals (160, 80, and 40 μm). This allows us to obtain a CE value which is ~ 0.1. As shown in Table 1, when sampling with high inter-slice intervals, there is relatively high coefficient of error (CE) values for most nuclei assessed, which can be reduced by reducing the inter-slice interval. This condition is intrinsically dependent on the low density of scattered TH-positive cells within most catecholaminergic nuclei of the brainstem reticular formation. Consistently, the CE values cannot be less than 0.1 as in the case of more densely packed neuron nuclei (West et al., 1996; Lewitus et al., 2012; Dell et al., 2016). Stereological counting of TH-positive cells was carried out by using a microscope Zeiss Axio Imager M1 (Zeiss, Wetzlar, Germany) associated with the software Image Pro-Plus 6.2 for Windows (vers. 6.2.1.491, Media Cybernetics, inc., Rockville, MD) equipped with a specific Macro (obtained by Immagine and Computer, Italy, MI) created ad hoc to perform the Optical Dissector technique. This Macro allows the operator to obtain an unbiased and optimized stereological cell count, according to King et al. (2002). All the areas of interest were identified and outlined at 2.5 × magnification. TH-positive cells were then counted at 100X magnification (numerical aperture 1.3) as previously described (King et al., 2002) by using a different dissector grid depending on the volume of the area to be analyzed. At the end of the procedure, a data sheet was produced containing all the data necessary to obtain the number of cells. The total estimation of cell numbers (N) was calculated by using the following equation: Where ssf is the “section sampling fraction,” asf is the “area sampling fraction,” tsf if the “thickness sampling fraction” (thickness of the tissue divided by the dissector height), and ΣQ- is the total number of cells counted within the dissector. The “Section Sampling Fraction” (ssf) is represented by the number of regularly spaced sections used for counts divided by the total number of sections used to collect the entire structure of interest. To sample the whole brainstem area, we collected 261 sections of 20 μm (covering the full extent of the area which is 5,220 μm). Sections were sampled at a ratio of 1:8, 1:4, and 1:2 (ssf) for SNC, VTA, and all other nuclei assessed, respectively. The “Area Sampling Fraction” (asf) represents the area between dissectors, that is, the ratio between the area of counting frames and the Area of Interest (AOI). The remaining value is tsf , the “Thickness Sampling Frequency,” that is, the height sampling fraction and it is calculated as the ratio between the height of the counting frame and the thickness of the tissue. This value is calculated by our system in each counting frame. The Coefficient of error (CE) was calculated according to King et al. (2002). Data are given as the mean ± SEM with statistical significance defined by p < 0.05. Statistical analyses were performed as follows: (i) Unpaired two-tailed Student's t-test (Figures 8A–C, 9A–H, 10A–D); (ii) Two-way RM ANOVA followed by Fisher's LSD (Figures 1B,C); and (iii) Pearson correlation test (Figures 11A–L, 12A–L). GraphPad Prism (Ver 5.01 GraphPad Software, Inc. La Jolla, CA, USA) statistical software was used for analysis. Chronic administration of corticosterone (100 μg/mL in the drinking water for 5 weeks, Figure 1A) to C57Bl/6J male mice induces a progressive increase in body weight (Figure 1B) and occurrence of glucose intolerance 4 weeks after treatment (Figure 1C). Serial sections of the rostral and caudal brainstem were considered to include 12 catecholamine nuclei of the brainstem reticular formation. (i) The dopamine-containing mesencephalic nuclei: A8 (also known as retrorubral field, RRF) (Bregma −3.8/Bregma −4.12); A9 (SNC) (Bregma −2.54/Bregma −3.80); and A10 (VTA) (Bregma −2.92/Bregma −3.88). (ii) TH-positive cells inside the peri-aqueductal gray (PAG) (Bregma −3.8/Bregma −4.28); (iii) the pontine parabrachial nucleus (PB) (Bregma−4.84/Bregma −5.32); (iv) the A7 nucleus (nucleus of lateral lemniscus) (Bregma −5.00/Bregma −5.32); (v) the big pontine noradrenergic nucleus A6 (locus coeruleus, LC) (Bregma −5.34/Bregma −5.82); (vi) the A5 nucleus (Bregma −5.34/Bregma −5.98); (vii) the rostral ventrolateral medulla C1/A1 (Bregma −6.36/Bregma −7.8); (viii) the dorsomedial nucleus of ala cinerea, C2/A2 (Bregma −6.36/Bregma −7.8); (ix) TH-positive cells inside the nucleus of the solitary tract (NTS) (Bregma −7.06/Bregma −7.4); and (x) the area postrema (AP) (Bregma −7.32/Bregma −7.76) (representative pictures of Figures 2–7). This provides a scenario encompassing the lateral column of the brainstem reticular formation ranging from the SNC down to the AP (Figures 2, 7, respectively). These representative pictures aim to provide the anatomical background of brainstem reticular catecholamine nuclei. It is remarkable that, as shown in representative Figure 6, even the undefined A4 region could be well visualized in these slices. When considering altogether these 12 nuclei, no consistent difference was noticeable by comparing corticosterone- and vehicle-treated mice. Only in the most extreme regions (RRF and AP), placed at the rostral and caudal pole of the brainstem reticular formation, respectively, a difference in TH immunostaining was evident (compare representative pictures of Figures 3, 7 as well as Figures 9C, 10D). This was substantiated by the total cell count (Figures 9C, 10D). Consistently with our previous findings obtained in organotypic mouse brainstem cultures (Busceti et al., 2019), stereological counting provides evidence for an increased number of TH-immunopositive cells in the whole rostro-caudal extension of the pons and medulla oblongata (from Bregma = −3.8 to Bregma = −7.64 without including the rostral midbrain) of mice treated with corticosterone compared with control vehicle-treated mice (Figure 8A). To assess the specific placement of increased TH-positive cells, which occurred following corticosterone administration in vivo, a stereological quantification was carried out by differentiating the cranial (Bregma −3.8/Bregma −5.82) from the caudal (Bregma −6.36/Bregma −7.64) part of the brainstem. Differing from in vitro data reporting an increase of TH in the caudal nuclei only (Busceti et al., 2019), the present investigation carried out ex vivo indicates that corticosterone-induced increase in the number of TH-immunopositive cells similarly occurs within cranial and caudal parts of the mouse brainstem (Figures 8B,C, respectively). This suggests that, when administered in vivo, corticosterone significantly increases TH immunostaining in multiple brainstem regions compared with its effects in isolated brainstem slices. In detail, when counted ex vivo, the increase in the rostral brainstem overalls the increase in the caudal brainstem and both express representatively the increase in TH-positive cell bodies, which was measured in the whole brainstem (Figure 8). The previous study using organotypic cell cultures left the increase of TH immunopositive neurons in the caudal brainstem non-defined since stereology could not be carried out and the increase was roughly attributed to catecholamine cell groups within the lower medulla. In contrast, the present study, which is carried out ex vivo identifies specifically the caudal appendix of this region, the AP, as the specific part where the increase in TH takes place. For what concerns the rostral brainstem, the specific nucleus of RRF owns the increase in TH immunostaining. To obtain a more detailed anatomical regional mapping of corticosterone-induced increase in the number of TH-positive cells, a detailed stereological analysis was carried out for each catecholamine nucleus of the brainstem reticular formation. Stereological quantification performed in catecholamine nuclei from the anterior brainstem indicates a significant increase in the number of TH-positive cells in response to treatment with corticosterone within the RRF (A8, as shown in the graph and representative pictures of Figure 9C). In the SNC, VTA, PAG, PB, A7, and A6, an increase was steady although non-significant (Figures 9A,B,D–G, respectively). On the contrary, the A5 shows a decrease in TH-positive cells (Figure 9H). Stereological counting of TH-positive cells in response to treatment with corticosterone in C1/A1 (Figure 10C) shows an increase that remains non-significant. Similarly, non-significant variations are detected within either C2/A2 (Figure 10B) or NTS (Figure 10A). In contrast, we found a significant increase in the number of TH immunoreactive cells in AP of mice chronically administered corticosterone compared with control vehicle-administered mice (Figure 10D). A correlation analysis was carried out between the number of TH-positive cells in each catecholamine nucleus of the brainstem reticular formation and parameters of glucose tolerance [the glucose blood levels detected 120 min after a bolus of glucose administered after 4 weeks of treatment with vehicle or corticosterone (Figure 11)] or body weight changes [body weight changes following 5 weeks of treatment with vehicle or corticosterone compared with respective values measured before the treatment (Figure 12)]. Concerning glucose tolerance, a positive correlation was selectively found between the number of TH-positive cells and glucose blood levels in NTS and AP (Figures 11I,L, respectively), while no significant correlation was detected in SNC, VTA, A8, PAG, A7, PB, A6, A5, C2/A2, and C1A1 (Figures 11A–J,K; respectively). When the correlation analysis was carried out for the body weight changes, we selectively found a positive correlation between the number of TH-positive cells in the A1/C1 group and the increase in body weight (Figure 12K). No correlation was found between the number of TH-positive cells and the body weight changes in SNC, VTA, A8, PAG, A7, PB, A6, A5, NTS, C2/A2, and AP (Figures 12A–H,J,L, respectively). The present ex vivo study indicates that chronic administration of corticosterone significantly increases the number of catecholamine cells within two specific nuclei, the RRF and AP, posed respectively, at the rostral and caudal poles of the brainstem reticular formation. These findings are in line with data showing that glucocorticoid receptor stimulation increases TH gene transcription (Hagerty et al., 2001). The promoter region of the TH gene contains glucocorticoids responsive element (GRE) (Hagerty et al., 2001) and elevated glucocorticoid levels accelerate both release and turnover of brain catecholamines (Abercrombie et al., 1989; Finlay et al., 1995; Sullivan, 2004; Kvetnansky et al., 2009). Chronic stress exposure activates catecholamine neurons (Mamalaki et al., 1992; Watanabe et al., 1995; Rusnak et al., 2001; Flugge et al., 2004) and long-term, reiterated stress exposure sensitizes the increase of TH mRNA levels in response to novel stressors (Serova et al., 1998, 1999; Rusnak et al., 2001; Tumer et al., 2001; Kvetnansky et al., 2003). However, during prolonged stress, the hyperactive state of catecholamine neurons may deplete catecholamine stores, which eventually may reduce the activity of catecholamine neurons (Loughlin et al., 1986; Valles et al., 2006). Moreover, alterations of catecholamine systems during chronic stress may contribute to neurodegeneration. Stress-related catecholamine alterations might accelerate neuronal degeneration by increasing the formation of toxic dopamine and norepinephrine by-products (Martinez-Vicente et al., 2008; da Luz et al., 2015; Sugama et al., 2016; Goldstein and Kopin, 2018; Kang et al., 2020; Fornai and Puglisi-Allegra, 2021). The present findings provide a quantitative measurement of 12 TH-expressing nuclei, being analyzed by automated stereology, in brainstem reticular formation of mice subjected to chronic administration of corticosterone. The increase in TH-positive cells was significantly and selectively identified within two catecholamine cell groups, AP and RRF. While the increase in A1/C1, although non-significant per se, significantly correlates with the increased body weight of C57Bl/6 mice. Remarkably, the increase of TH-positive cells within AP is significantly correlated with the onset of reduced tolerance to glucose in line with specific control of AP on those neurons in the dorsal vagal complex which innervates the pancreas (Loewy et al., 1994). The present study also indicates an increase of TH-positive cells within A1/C1, which, although is non-significant, is significantly correlated with an increase in the body weight of corticosterone-treated mice. The present findings detail and extend previous findings we obtained in vitro, which documented corticosterone-induced increased expression of catecholamine markers within organotypic cell cultures dissected from the caudal brainstem (Busceti et al., 2019). Thus, the present data obtained in the whole brainstem indicates that chronic corticosterone elevates TH-positive cells within the RRF within the rostral brainstem, which differs from organotypic cell cultures where no effect was detected neither in slices from the pons nor the mesencephalon (Busceti et al., 2019). Despite a significant increase in the number of TH-positive cells in A8, the greatest dopaminergic nuclei (A9 and A10) do not show any increase. These data suggest that there is no correlation between the alterations induced in response to chronic exposure to corticosterone and a specific neurotransmitter in all nuclei. The effect is rather confined to specific nuclei. Remarkably, in the present ex vivo study, the increase in TH-positive cell bodies is selective for two nuclei, although it occurs non-significantly for most TH-positive nuclei, where the number of TH-positive cells surpasses those counted from the control brain. This trend is magnified compared to that documented in organotypic slices of the anterior brainstem where the overall amount was not significantly different (Busceti et al., 2019). This widespread increase in TH immunoreactivity in the cranial brainstem does not occur in A5, where the trend is opposite to the other nuclei. Although both AP and the RRF develop a significant increase of TH-positive neurons, only the increase in AP but not the increase in RRF correlates with corticosterone-induced glucose intolerance. It is expected that further studies will allow us to disclose a causal relationship between the increase of TH-positive cells within RRF (A8) and the development of other alterations associated with Cushing's syndrome, such as anxiety, aggressiveness, or sleep-waking cycle dysfunction. RRF is involved in a number of behavioral states and a number of axons connect A8 dopaminergic neurons with neurons in the amygdala (Wallace et al., 1989). The amygdala, which is in close connection with the medial prefrontal cortex (mPFC) controls emotional responses, such as fear and anxiety (Sotres-Bayon and Quirk, 2010; Kumar et al., 2014; Likhtik et al., 2014; Bukalo et al., 2015), which characterize Cushing's syndrome. Corticosterone-induced emotional changes may likely be partly related to its effects on RRF (A8) neurons. In fact, mPFC exerts an inhibitory control on the amygdala activity, thus controlling emotional behaviors (Rosenkranz and Grace, 2001; Quirk et al., 2003; Rosenkranz et al., 2003; Motzkin et al., 2015). Evidence exists that chronic corticosterone treatment in mice produces defective prefrontal inhibitory control of the amygdala (Liu et al., 2020) fostering anxiety and depression (Quirk and Gehlert, 2003; Correll et al., 2005; Rauch et al., 2006). Thus, corticosterone may produce a dual effect via RRF and mPFC, which synergize to produce anxiety and mood disorders. Specific behavioral investigations are needed to address this point. Concerning the caudal part of the brainstem reticular formation, AP features a significant increase in the number of TH-positive in response to chronic treatment with corticosterone. This is fascinating since AP provides a selective control to the vagal efferent, which innervates the pancreas (Loewy et al., 1994). This is in line with altered tolerance to glucose occurring in these mice, which, in turn, significantly correlates with the increase in TH-positive cells within AP. This is in line with the role of TH-positive cells within AP and its rostral branching within NTS as gluco-sensing neurons (Roberts et al., 2017). In detail, high glucose levels increase the firing of catecholamine neurons within AP and NTS by increasing spontaneous glutamate inputs (Roberts et al., 2017). This provides a mechanism by which changes in glucose could impact catecholamine neurons in the medullary reticular formation, thus controlling cardiovascular, respiratory, and gastrointestinal systems (Simon et al., 1985; Kubo et al., 1990; Itoh and Buñag, 1993; Schild et al., 1994; Saper et al., 2002; Olson et al., 2006). The present correlation data suggest that high plasmatic levels of glucocorticoids through increasing the number of catecholamine neurons in the AP may provide a neuroanatomical substrate extending beyond glucose intolerance. In keeping with the caudal brainstem, considering the C1/A1 region, the increase in TH-positive cells is not significant; however, a significant correlation was measured between the number of TH-positive cells and increased body weight. This is very intriguing since these neurons of the ventrolateral medulla project to the paraventricular hypothalamic areas and are involved in feeding behavior (Rinaman, 1999; Gaykema et al., 2007). This suggests a neural basis to explain why glucocorticoids alter feeding activity, which, in turn, may contribute to metabolic alterations, fat re-distribution, and glucose intolerance concomitant to an increase in body weight, which occurs in Cushing's syndrome. Chronic treatment with corticosterone induces in vivo a significant and selective alteration of TH-positive neurons within two nuclei placed in the lateral column of the brainstem reticular formation. These alterations significantly correlate with the selective domain of Cushing's syndrome. To our knowledge, this is the first study that provides a potential anatomical basis that may underlie specific symptoms occurring in Cushing's syndrome. The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. The animal study was reviewed and approved by Neuromed Institute Ethical Committee Ministry of Health (Authorization #1132/2016-PR). CLB performed immunohistochemical analysis, statistical analysis, and wrote the manuscript. DB performed immunohistochemical analysis and stereological counting. PD performed glucose tolerance test. MS and MF revised the manuscript. SP-A, FN, and FF supervised research and revised the manuscript. All authors contributed to the article and approved the submitted version. This work was supported by a grant from the Italian Ministry of Health (Ricerca Corrente to FF through IRCCS Neuromed). The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649726

Wanrong Hu,Wen Cai,Zhaojun Zheng,Yuanfa Liu,Cheng Luo,Fang Xue,Dongliang Li

Study on the chemical compositions and microbial communities of cigar tobacco leaves fermented with exogenous additive

10-11-2022

Biological techniques,Biotechnology,Microbiology

Fermentation process plays an important role in the biochemical properties and quality of cigar tobacco leaves (CTLs). In industry, exogenous additive (EA) was usually adopted for improving the quality of CTLs during fermentation. However, the mechanism of enhanced quality of CTLs fermented with EA was confused. Herein, the chemical compositions and microbial communities of CTLs during fermentation with EA were analyzed. The increased contents of total nitrogen and total sugar, as well as the improved consumption rate of reducing sugar in CTLs were found with the addition of EA. Besides, fermentation with EA reduced the content of total nonvolatile organic acid, especially unsaturated fatty acid. The contents of total and several representative aroma components were improved. Additionally, the increased abundance of Staphylococcus and decreased abundance of Aspergillus were detected. Combined with the changes of chemical compositions and microbial communities, it was confirmed that the carbohydrates and alcohols originated from EA promote the enrichment of Staphylococcus and accelerate biochemical reactions, such as Maillard reaction and esterification reaction, thus improving the contents and quality of aroma components in CTLs. This study demonstrated the mechanism of enhanced quality of CTLs fermented by EA, which provides more ideas for developing novel and efficient EAs.

Study on the chemical compositions and microbial communities of cigar tobacco leaves fermented with exogenous additive Fermentation process plays an important role in the biochemical properties and quality of cigar tobacco leaves (CTLs). In industry, exogenous additive (EA) was usually adopted for improving the quality of CTLs during fermentation. However, the mechanism of enhanced quality of CTLs fermented with EA was confused. Herein, the chemical compositions and microbial communities of CTLs during fermentation with EA were analyzed. The increased contents of total nitrogen and total sugar, as well as the improved consumption rate of reducing sugar in CTLs were found with the addition of EA. Besides, fermentation with EA reduced the content of total nonvolatile organic acid, especially unsaturated fatty acid. The contents of total and several representative aroma components were improved. Additionally, the increased abundance of Staphylococcus and decreased abundance of Aspergillus were detected. Combined with the changes of chemical compositions and microbial communities, it was confirmed that the carbohydrates and alcohols originated from EA promote the enrichment of Staphylococcus and accelerate biochemical reactions, such as Maillard reaction and esterification reaction, thus improving the contents and quality of aroma components in CTLs. This study demonstrated the mechanism of enhanced quality of CTLs fermented by EA, which provides more ideas for developing novel and efficient EAs. Fermentation is widely adopted to adjust the physical and chemical properties of target products based on microbial activity, which has been extensively applied in food processing, medicine and energy production, environmental remediation, and so on. Microorganisms in the fermentation system could utilized substrates, such as protein and carbohydrate, and generate various products. Traditional fermentation is a natural process without manual intervention, which is significantly influenced by the environment. Owing to the rapid technological and industrial development, different approaches were used to control the process, including the regulation of fermentation parameters and introduction of starter cultures. As a result, the fermentation is more likely proceed with a desired direction, and the consistent quality of products could be obtained. Tobacco is considered as one of the important industrial and economic crops. Among the various cigarette products, cigar has attracted extensive attention owing to the characteristics of hand-made and particularity of raw material. Cigar was totally made by cigar tobacco leaves (CTLs) without filter and wrapping paper. Thus, the quality of CTLs determines the sensory of cigars. Generally, the production process of cigar includes cultivation, air-curing, fermentation, and rolling. Besides of variety optimization, fermentation is the most effective approach to improve tobacco quality. According to recent researches, condition optimization has received noticeable attention in the field of CTL fermentation. Several important parameters for CTL fermentation, such as temperature, humidity, time, and initial moisture content of tobacco leaves, were studied systematically. However, effectiveness of enhancing the quality of CTLs by traditional fermentation is limited. Based on this trend, adding exogenous additive (EA) to facilitate the fermentation process was proposed. At present, EA for CTLs fermentation includes plant extracts, bacterium, and enzyme preparation, as well as the mixture of above materials. Additives of green tea infusion, diluted milk and rice wine, or Bacillus cereus were reported to improve the quality of CTLs, especially the flavor components. Theoretically, EA could affect the growth and metabolism of microorganism community, and alter the biochemical reaction pathways, thus improve the quality of cigar. In fact, mechanism of EA on enhancing the cigar quality has been explored by few researches, and most of them only focused on the effectiveness. Fermentation is complex and enzymatic action of microorganisms as well as chemical interaction would affect the quality of cigar. Previous studies indicated that several chemical indexes, such as nitrogen, sugar, organic acids, and aroma components, affected the quality of cigar tobacco directly. The contents of nitrogen, sugar and organic acids are related to the sweetness and mellowness of CTLs. Besides, different aroma components endow cigars with different flavor profiles. Apart from chemical compositions, microbial communities play an important role in improving the quality of CTLs. The changes of microbial community may demonstrate the mechanism responsible for tobacco quality. Liu et al. found that the bacterial and fungal community structure significantly changed in the CTLs fermentation, and the chemical compositions were influenced by microbial activities. Therefore, it is necessary to analyze the change law of chemical compositions and microbial communities of CTLs with EA during the fermentation process. Considering the research gap about mechanism of EA on enhancing the cigar quality, our study showed for the first time the effects of EA on the chemical and microbial properties of CTLs. To the best of our knowledge, no studies have systematically compared the changes on sensory qualities, chemical compositions, and microbial communities of CTLs under two different systems of adding water and EA. Therefore, the EA, which was adopted for handmade cigars in actual production at present, was adopted for CTLs fermentation. The effects of EA on the contents of major conventional chemical components (including total nitrogen, nicotine, total sugar, reducing sugar, K, and Cl), nonvolatile organic acids (NOCs), and aroma components in CTLs were investigated. Furthermore, the relevant analysis about microbial communities was conducted. This study aims to provide a systematic investigation about influencing reasons of EA and fermentation process on cigar quality, and propose a theoretical basis for rational development of new EAs and optimization of fermentation process. A representative kind of CTL sample was used in this study, which was planted in De-yang (Sichuan, China) and named as DX-4. The CTL raw material was obtained after air-curing and possessed the humidity of 17 ~ 20%. The formulated EA for fermentation, which is adopted for handmade cigar products in the Great Wall Cigar Factory (Chengdu, China), was used for this research. Rice wine, fritillaria cirrhosa, and loquat wine were the main components of EA. The reagents used in this study were purchased from Chengdu Kelong Reagent Company and of analytical grade unless stated otherwise. In a typic fermentation procedure, 217.5 g of EA and 282.5 g of de-ionized water were mixed, which was sprayed evenly on the surface of 5000 g of CTLs. The water content of CTLs was 30 ± 2%. Then, the CTLs were transferred to a linen bag and placed in a constant temperature and humidity incubator (Binder, KBF720). The fermentation was performed under the condition of 35 °C and 75% of relative humidity for 35 days, which was set according to the parameters actually used in the Great Wall Cigar Factory. After 0, 7, 14, 21, 28 and 35 days, 800 g of CTLs was withdraw at given times and denoted as CTL_EA_x (x = 0, 7, 14, 21, 28, and 35, denoted the fermented time). For comparison, control samples without EA replaced by de-ionized water were obtained by the same procedure, which were named as CTL_W_x (x = 0, 7, 14, 21, 28, and 35, denoted the fermented time). All samples were stored in a refrigerator at − 80 °C for further use. Sensory quality evaluation of CTL samples was performed by a six members-trained panel. Nine indicators including aroma quality, aroma concentration, smoke concentration, strength, cleanliness, offensive odor, aftertaste, ash color, and combustibility were assessed. Sensory characteristic scales were calculated by using a 9-point hedonic scale, and a higher score represented the better performance of the corresponding index. The scores were discussed and agreed by all panelists. Six conventional chemical components in tobacco, including the total nitrogen, nicotine, total sugar, reducing sugar, K, and Cl, were usually of general concern in the tobacco industry. Thus, contents of the above-mentioned chemical components were analyzed in this study firstly. A continuous flow analytical system according to the Tobacco Industry Standard (YC/T161-2002, YC/T468-2013, YC/T159-2019, YC/T 217-2007, and YC/T 162-2011) was adopted. The composition and content of nonvolatile organic acid (NOC) were determined by gas chromatograph (GC) equipped with hydrogen flame ionization detector (FID), since the CTL sample was treated with sulfuric acid–methanol method in advance. A column DB-5MS (60 m × 0.25 mm I.D. × 0.25 µm film thickness) was used. Helium was used as a carrier gas and the flow rate was 1.5 mL min−1. The temperature of injection port was 280 °C. The temperature program of column was 40 °C (3 min), 40–280 °C (10 °C min−1), and 280 °C (30 min). Gas Chromatography-Mass Spectrometry (GC–MS) was used to determine the contents of aroma components in CTLs. Acetonitrile extraction and internal standard method were adopted. In detail, the CTLs were ground into powder firstly. 2 g of the CTL powder sample was mixed with 10 mL of de-ionized water in a 50 mL centrifuge tube for 10 min. Then, 10 mL of acetonitrile (chromatographical grade) and 50 μL of internal standard working solution (phenethyl acetate: 9.06 mg mL−1) was pipetted into the centrifuge tube, which was shaken at 2000 rpm for 120 min in a vortex oscillator and cooled at 4 °C for 10 min subsequently. Afterwards, 4 g of anhydrous magnesium sulfate, 1 g of sodium chloride, 1 g of sodium citrate, and 0.5 g of disodium hydrogen citrate were added into the cooled cube, which was shaken using a vortex oscillator immediately to prevent anhydrous magnesium sulfate from agglomerating. Then, 1 mL of extract supernatant was pipetted from the centrifuge cube and mixed with 150 mg of anhydrous magnesium sulfate. Finally, the supernatant from magnesium sulfate suspension was collected for GC–MS analysis. The GC–MS system (Agilent, 7890B-5977) was coupled with an Agilent DB-5MS column (60 m × 0.25 mm, 1.00 μm). Chromatographic conditions: the temperature of injection port was 290 °C, the temperature program of column was 60–250 °C (2 °C min−1), 250–290 °C (5 °C min−1), and 290 °C (20 min). The interface was kept at 290 °C. Qualitative analysis was performed in the electron-impact (EI) mode at 70 eV potential using 26–400 amu for qualitative analysis. The DNA extraction steps were performed based on previous reports. Total microbial genomic DNA was extracted from microorganism in CTL samples according to the manufacturer's instruction of E.Z.N.A.® Soil DNA Kit (Omega Bio-Tek, Norcross, GA, U.S.). The integrity of extracted DNA was checked by 1% agarose gel electrophoresis, while the concentration and purity of DNA were determined by NanoDrop 2000 UV–vis spectrophotometer (Thermo Scientific, Waltham, MA, USA). The V3-V4 regions of the bacterial 16S rRNA gene were amplified using the primers 338F (5ʹ- ACTCCTACGGGAGGCAGCAGCAGG -3ʹ) and 806R (5ʹ-GACTACHVGGGTWTCTAAT-3ʹ). The fungal internal transcribed spacer gene was amplified using the primers ITS1F (5ʹ-CTTGGTCATTTAGAGGAAGTAA-3ʹ) and ITS2R (5ʹ-GCTGCGTTCTTCATCGATGC-3ʹ). Then, the PCR products were analyzed using 2% agarose gel electrophoresis and recovered using an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA). Finally, the purified amplicons were sequenced using the Illumina MiSeq platform. Fastp and FLASH software were used for checking and merging the raw gene sequencing reads. The sequences were identified as the same operational taxonomic units (OTUs) with 97% similarity threshold using Usearch (version 7.0 http://drive5.com/uparse/). Besides, each sequence was classified and noted using the RDP Classifier (http://rdp.cme.msu.edu/) and Silva database with a confidence threshold of 70%. Permissions for plant material collection was obtained. The collection of plant material complies with relevant institutional, national, and international guidelines and legislation. As shown in Table 1, the tobacco raw material used in this experiment exhibited medium concentration and strength of smoke, as well as a relatively low aroma quality and concentration. Therefore, it is necessary to improve the quality of tobacco leaves by fermentation with EA. It can be seen that the addition of EA could improve the sensory quality of tobacco leaves, since the sensory scores of EA-group samples were higher than those in water-group. Notably, aroma quality, cleanliness, and aftertaste were enhanced with the addition of EA. Besides, with the extension of fermentation time, the sensory assessment score of CTLs increased first and then decreased, and reached the highest level in 21 days, indicating that excessive fermentation was not conducive to the improvement of sensory quality of cigar. Effects of EA on six major chemical components in tobacco leaves, including total nitrogen, nicotine, total sugar, reducing sugar, K, and Cl, were analyzed firstly. The contents of total nitrogen and nicotine were related to the smoke concentration and pungency. As shown in Fig. 1a, the content of total nitrogen in CTL raw material was about 3 ~ 4% (w/w), which is consistent with previous reports. Total nitrogen content shows an increased trend with the addition of EA and was affected slightly by the fermentation process. In fact, nicotine is one of the main nitrogenous compounds in CTLs. The changes of nicotine content in DX-4 during fermentation are presented in Fig. 1b. The contents of nicotine were about 4 ~ 5% (w/w) and reduced during the fermentation process, which was due to the degradation of nicotine into nornicotine through nitrosation reaction. Besides, with the introduction of EA, the nicotine contents of CTLs were improved. In the smoking process, nicotine would degrade into nitrogen heterocyclic aromatic components, such as pyrrole, pyridine, and pyrazine compounds, which endows the CTLs with fragrant baked and caramel aromas. Therefore, adding EA in the fermentation process could improve the sensory quality of cigar. Sugar presents a significant influence on the sensory quality of tobacco, which could weaken the acrimonious taste of the smoke by producing acids and be convert to various aromatic substances by Maillard reaction, caramelization, and pyrolysis reaction. Additionally, the content of sugar is related to the sweetness of CTLs. The effects of EA on the contents of total sugar and reducing sugar were shown in Fig. 1c,d. About 0.5% (w/w) of total sugar and 0.2% (w/w) of reducing sugar could be detected in CTL raw materials. Compared with flue-cured tobacco, cigar leave presents a relatively low level on sugar content. As the fermentation proceed, the total sugar content increased firstly and then decreased. The hydrolysis of starch, cellulose, and pectin, as well as the consumption of sugars resulted in the insignificant change of total sugar content. Besides, the addition of EA can increase the total sugar and reducing sugar contents in CTLs. The consumption rate of reducing sugar in EA-group was higher than that in water-group, which was due to the acceleration of biochemical process with the abundance of substrates. At the end of the fermentation process, reducing sugar can hardly be detected, because reducing sugar is almost completely consumed by biochemical reactions, such as Maillard reaction. It is reported that K and Cl show the significant influence on the combustibility of tobacco leaves. The weight ratio of K and Cl (K/Cl) and the content of K present a positive correlation with combustibility. As shown in Fig. 2, 3.7 ~ 4.0% of K could be detected in CTLs, which shows no obvious change in the content during fermentation. By contrast, the contents of Cl decreased during the fermentation process, which may due to the volatilization of chlorinated compounds. As a result, the increased value of K/Cl was found as fermentation proceed. Besides, the Cl content in CTL increased with the introduction of EA, since EA was prepared by tap water as solvent and Cl could be found in solvent. Although the value of K/Cl decreased as adding EA, the relatively high content of K and value of K/Cl were still detected, indicating that the EA-treated CTLs possessed a good performance in combustibility. Nonvolatile organic acid (NOC) is an important component in CTLs, which accounts for 70 ~ 80% of organic acid and affects the sensory quality of cigar significantly. NOC plays an important role in changing the pH and fragrance of flue gas, thus regulates the mellowness and irritation of CTLs. In this study, fourteen NOCs were detected, as shown in Table S1. The total content of NOCs was reduced as adding EA, which was resulted by the alcohols from EA (Fig. 3). The consumption of NOCs by EA would weaken the reaction between NOC and NIC, thus the NIC contents increased with the addition of EA, which is consistent with the results of Fig. 1b. The total content of NOCs decreased firstly and then increased, which showed a similar value at the beginning and end of the fermentation process. However, the variation amplitude of NOC content in EA-group was gentler than that in water-group. It means that EA built an environment similar with buffer system, which would be beneficial to maintain the acid–base balance of CTLs and improve the fluentness of cigars. Among the fourteen NOCs, oxalic acid, malic acid, and citric acid accounts for 90% (w/w) of NOCs, which would affect the irritation and smoke concentration obviously. As exhibited in Fig. 3, the contents of these three acids in EA-group were lower than that in water-group at the beginning and end of the fermentation. The change rule of the three acids content of EA-group in 0 ~ 28 days was similar with that of water-group in 7 ~ 35 days, indicating that the addition of EA accelerated the generation and consumption of NOCs. This phenomenon may be related with the changes of microbial activities with EA, which would be discussed combined with microbial community analysis in the following. It was reported that malic acid possessed a positive effect on the quality of tobacco leaves, while oxalic acid and citric acid would damage the taste of tobacco leaves. Accordingly, 21 days may be an optimal fermentation period based on the NOCs analysis for EA-group. Besides, several long-chain fatty acids, including two unsaturated fatty acid (oleic acid and linoleic acid) and four saturated fatty acid (myristic acid, palmitic acid, stearic acid, and arachidic acid), were detected in this study. Unsaturated fatty acid would increase the irritation and reduce the smoothness of smoke, while saturated fatty acid would increase the mellowness of smoke. Therefore, the relatively low content of long-chain fatty acids, especially unsaturated fatty acid in EA-group implies its possibly good performance for sensory. Various aroma components were reported to detect in cigar leaves, which plays an important role in the sensory quality of CTLs. Therefore, effects of EA on the contents of aroma components were analyzed by GC–MS and shown as Table 2 and Table S2. As we can see, the total contents of aroma components increased from 1.6738 mg g−1 to 1.8076 mg g−1 in 35 days as adding EA. In fact, during the fermentation process, the contents of aroma components were improved with the addition of EA. Higher content of aroma components would endow CTLs with more fragrant flavor. In this study, various kinds of aroma components were detected, including alcohols, alkenes, phenolics, ketones, acids, esters, and heterocycles of compounds. As shown in Table 2, the contents of alcohols, alkenes, ketones, acids, and esters were higher in EA-group than that in water-group. Since the sum of alcohols, alkenes, and ketones account for 80 ~ 90% (w/w) of the total aroma components, and ketones contributed mostly to the aroma value of cigar leaves, CTL samples treated with EA would show an improved performance on sensory quality. Additionally, it shows a trend of first increase and then decrease on the contents of aroma components during the fermentation process. In 14 days, a highest value of aroma component content was found. It indicated that the optimal fermentation time may be 14 days. Except for the total content of aroma components, several representative aroma components were also affected by EA and fermentation process. As shown in Table 2, representative aroma substances such as neophytadiene, megastigmatrienone, and dihydroactinidiolide showed an increasing tread with the addition of EA. As a flavoring substance with the highest content and important intermediate compound of biochemical reaction in tobacco, neophytadiene not only directly affects the sensory quality of tobacco leaves, but also affects the formation and consumption of other flavoring substances. Neophytadiene possesses delicate fragrance, which is formed by the degradation of chlorophyll in the processes of maturation and modulation. As a result, the degradation of chlorophyll with green offensive odor, as well as the generation of neophytadiene, could promote the improvement of mellowness and the reduction of stimulation. At the end of fermentation, the content of neophytadiene increased from 0.3267 mg g−1 to 0.4006 mg g−1 with the addition of EA (Table 2). Compared with neophytadiene, megastigmatrienone shows a relatively low content in CTLs. However, megastigmatrienone contributes greatly to the aroma quality of cigar due to the low aroma threshold. Megastigmatrienone possesses fruit and tobacco aroma, which usually exists with multiple isomers. In this study, two isomers of megastigmatrienone were detected. The total contents of megastigmatrienone increased from 0.0244 mg g−1 to 0.0277 mg g−1 in DX-4 as adding EA. The enrichment of megastigmatrienone can effectively improve the mellowness, sweetness, and cleanliness of CTLs. Only light fruit and baked aromas were emanated from dihydroactinidiolide, but it can weaken the pungency well, thus increasing the cleanliness and mellowness of tobacco leaves. Besides of the above-mentioned compounds, the increasement of other aroma components could also lay a material foundation for the quality improvement of cigars (Table S2). In order to investigate the effects of EA and fermentation process on the microbial community of CTLs, the distinct regions of 16S rRNA and ITS genes were amplified for Illumina paired-end sequencing. The coverage of all sample was higher than 99.88% for the bacterial and fungal sequences (Table 3), indicating that the sequencing depth of each sample was sufficient to reflect the bacterial and fungal composition of CTL samples. Alpha diversity metrics, including Shannon, Simpson, Chao, and ace, are analyzed to reflect the diversity and richness of the microbial community in samples. The Chao and ace were used to identify the community richness, while Shannon and Simpson were adopted to evaluate community diversity. As shown in Table 3, the richness and diversity of bacterial community in EA-group were lower than that in water-group at the early stage of fermentation, whereas it showed a contrary law at the late stage of fermentation. Moreover, the influences of EA on the richness (and diversity) of fungi and bacteria were different. It means that EA could accelerate the growth of fungi in the early stage and provide suitable metabolic environment for bacteria in the late stage of fermentation. For EA-group, the richness of bacterial and fungal community was increased and then decreased with prolonging fermentation time. As illustrated in Fig. 4), 57 ~ 253 bacterial OTUs and 50 ~ 171 fungal OTUs with 97% similarity were found in CTL samples. It showed a low proportion of common OTUs in total OTUs, which means that the unique functional flora in the fermentation process showed an obvious succession phenomenon. The quantity of fungal unique OTUs in EA-group was higher than that in water-group at 0 ~ 21 days, indicating that EA could promote the growth of peculiar fungal colonies in CTLs. Besides, the unique bacterial OTUs in CTL_EA_28 was 118, accounting for 46.64% of the total OTUs, indicating that 28 days of fermentation time is beneficial to the bacterial diversity of tobacco leaves. According to the analysis of alpha-diversity, it could be concluded that excessive fermentation is not conducive to the growth and metabolism of microorganisms in CTLs. Figure 5a showed the changes of bacterial communities on phylum level during the fermentation. Four different phyla, including Firmicutes, Cyanobacteria, Proteobacteria, and Actinobacteriota, were identified among these CTL samples. In the water-group, the relative abundance of Firmicutes decreased with prolonging fermentation time, whereas that of Cyanobacteria increased. After fermentation, the relative abundances of Firmicutes and Cyanobacteria were similar. As for the EA-group, Firmicutes (over 90%) was the dominant phylum in 0 ~ 21 days, which was more prominent than water-group. However, after 28 days of fermentation, the relative abundance of Firmicutes was reduced. At 35 days, Cyanobacteria and Proteobacteria were the dominant phyla in EA-group. The effects of EA and fermentation on bacterial community at species level were also analyzed (Fig. 5b). Among these species, Staphylococcus, norank_f_norank_o_choroplast, Pseudomonas, and Ralstonia were the main dominant bacterial species in CTLs. In the early stage of fermentation, Staphylococcus possessed an obvious preponderance with the abundance over 70%, which was consistent with other reports. It is observed that the addition of EA further enhanced the dominance of Staphylococcus. Sugar was one of the main components in EA, which could provide metabolic substrates for the growth of Staphylococcus and promote the abundance of Staphylococcus. Accordingly, the generation of metabolite, i.e. acids, produced by Staphylococcus was accelerated. In the late stage of fermentation, the relative abundances of Staphylococcus decreased due to the consumption of reducing sugar (Fig. 1d). As a result, norank_f_norank_o_choroplast and Pseudomonas were the most frequently detected species in CTL samples. In addition to bacteria, relative abundance of fungal community was also analyzed. As revealed in Fig. 5c, Ascomycota with the abundance over 70% was the principal phylum among the six samples in water-group. Among the EA-group, the decreased relative abundance of Ascomycota was observed in 7 ~ 21 days, since Basidiomycota became the prevalent fungal. In 21 ~ 35 days, Ascomycota regained its position as the dominant fungal. As illustrated in Fig. 5d, six different species, including Aspergillus, Wallemia, Alternaria, Sampaiozyma, Cladosporium, and Colletotrichum, were mainly found in CTLs. Aspergillus is the dominant species in the fungal community in water-group. It is observed that the relative abundance of Aspergillus increased first and then reduced with the extension of fermentation time. Notably, the introduction of EA changed the fungal communities in CTLs significantly. The abundance of Aspergillus in EA-group was lower than that in water-group. Since Aspergillus poses a relatively high mildew-causing risk for CTLs, the cigar products fermented with EA would possess a reliable health for people. Besides, Wallemia was the prevalent fungal in EA-group during 0 ~ 21 days. Wallemia could promote the production of flavonoid which would be decomposed into aroma components in the combustion of cigars. Overall, these results illustrated that the addition of EA and fermentation process influenced the relative abundances of bacterial and fungal communities in CTLs obviously. The changes of microbial communities with EA could promote the quality improvement of CTLs. In this study, the fermentation of CTLs in EA-group shows several main differences with that in water-group. Firstly, the total contents of total sugar and aroma components increased. Secondly, the consumption rate of reducing sugar was improved, and the generation as well as consumption of NOCs was accelerated. Thirdly, the relative abundances of Staphylococcus and Wallemia were enhanced. Accordingly, a possible mechanism of EA improving the CTLs quality was proposed, as shown in Fig. 6. The carbohydrates originated from EA could improve the content of total sugar and reducing sugar in CTLs, thus promoting the enrichment of Staphylococcus in tobacco leaves. The increased abundance of Staphylococcus accelerated the consumption of reducing sugar and production of NOCs. Meanwhile, alcohols from EA were introduced into CTLs, which would promote the formation of esters. Moreover, the relatively high content of total sugar facilitates the proceeding of Maillard reaction. As a result, the contents of aroma components were improved. Overall, the increased contents of total sugar and aroma components would endow the cigar with an improved sensory quality. This study investigated the changes of sensory qualities, major chemical compositions as well as microbial communities of CTLs during fermentation with EA. Systematic studies indicated that the addition of EA could increase the contents of total nitrogen, nicotine, total sugar, and aroma components in CTLs, and accelerate the generation and consumption of NOCs. The increased abundance of Staphylococcus and decreased abundance of Aspergillus were detected. Furthermore, a preliminary exploration about mechanism of the changes in CTLs compositions with EA was conducted. The present study suggested that carbohydrates and alcohols originated from EA promote the enrichment of Staphylococcus and accelerate biochemical reactions, thus improving the contents of aroma components. As a results, the aroma quality and concentration, cleanliness, as well as aftertaste of cigar were improved. Therefore, it is believed that EA rich in carbohydrates and aroma components is effective in improving the quality of tobacco leaves. However, more detailed studies of the changes in activities of relevant enzymes during the fermentation of CTLs are required to determine the precise mechanisms of improved CTLs quality with EA. In conclusion, this study not only demonstrates the reasons of flavor enhancement and quality improvement by industrial EA, but also provides strategy to prepare novel EAs. Supplementary Tables.

PMC9649729

Jennifer Cantley,Xiaofen Ye,Emma Rousseau,Tom Januario,Brian D. Hamman,Christopher M. Rose,Tommy K. Cheung,Trent Hinkle,Leofal Soto,Connor Quinn,Alicia Harbin,Elizabeth Bortolon,Xin Chen,Roy Haskell,Eva Lin,Shang-Fan Yu,Geoff Del Rosario,Emily Chan,Debra Dunlap,Hartmut Koeppen,Scott Martin,Mark Merchant,Matt Grimmer,Fabio Broccatelli,Jing Wang,Jennifer Pizzano,Peter S. Dragovich,Michael Berlin,Robert L. Yauch

Selective PROTAC-mediated degradation of SMARCA2 is efficacious in SMARCA4 mutant cancers

10-11-2022

Cancer genomics,Targeted therapies,Drug development

The mammalian SWItch/Sucrose Non-Fermentable (SWI/SNF) helicase SMARCA4 is frequently mutated in cancer and inactivation results in a cellular dependence on its paralog, SMARCA2, thus making SMARCA2 an attractive synthetic lethal target. However, published data indicates that achieving a high degree of selective SMARCA2 inhibition is likely essential to afford an acceptable therapeutic index, and realizing this objective is challenging due to the homology with the SMARCA4 paralog. Herein we report the discovery of a potent and selective SMARCA2 proteolysis-targeting chimera molecule (PROTAC), A947. Selective SMARCA2 degradation is achieved in the absence of selective SMARCA2/4 PROTAC binding and translates to potent in vitro growth inhibition and in vivo efficacy in SMARCA4 mutant models, compared to wild type models. Global ubiquitin mapping and proteome profiling reveal no unexpected off-target degradation related to A947 treatment. Our study thus highlights the ability to transform a non-selective SMARCA2/4-binding ligand into a selective and efficacious in vivo SMARCA2-targeting PROTAC, and thereby provides a potential new therapeutic opportunity for patients whose tumors contain SMARCA4 mutations.

Selective PROTAC-mediated degradation of SMARCA2 is efficacious in SMARCA4 mutant cancers The mammalian SWItch/Sucrose Non-Fermentable (SWI/SNF) helicase SMARCA4 is frequently mutated in cancer and inactivation results in a cellular dependence on its paralog, SMARCA2, thus making SMARCA2 an attractive synthetic lethal target. However, published data indicates that achieving a high degree of selective SMARCA2 inhibition is likely essential to afford an acceptable therapeutic index, and realizing this objective is challenging due to the homology with the SMARCA4 paralog. Herein we report the discovery of a potent and selective SMARCA2 proteolysis-targeting chimera molecule (PROTAC), A947. Selective SMARCA2 degradation is achieved in the absence of selective SMARCA2/4 PROTAC binding and translates to potent in vitro growth inhibition and in vivo efficacy in SMARCA4 mutant models, compared to wild type models. Global ubiquitin mapping and proteome profiling reveal no unexpected off-target degradation related to A947 treatment. Our study thus highlights the ability to transform a non-selective SMARCA2/4-binding ligand into a selective and efficacious in vivo SMARCA2-targeting PROTAC, and thereby provides a potential new therapeutic opportunity for patients whose tumors contain SMARCA4 mutations. The multi-subunit switch/sucrose non-fermentable (SWI/SNF or BAF) complex facilitates the remodeling of chromatin to regulate key cellular processes including transcriptional regulation and DNA repair. Catalytic function is conferred by one of two mutually exclusive ATP-dependent helicases, SMARCA2 and SMARCA4, which share strong protein sequence homology. In addition to a highly conserved ATPase domain (93% identity), both proteins contain a conserved bromodomain (96% identity) (BD) that can interact with acetylated chromatin. SWI/SNF has gained a great deal of attention in cancer biology, as tumor sequencing studies have revealed that ~20% of human cancers harbor mutations in specific core or accessory components of the complex. In particular, loss of function mutations in SMARCA4 are enriched in subsets of multiple malignancies, with the highest prevalence of homozygous mutations occurring in non-small cell lung cancer (NSCLC). Aberrant chromatin remodeling caused by SMARCA4 mutations can result in the disruption of enhancer accessibility and accumulation of Polycomb repressive complexes 1 and 2 across the genome. Strategies to therapeutically target SMARCA4 mutant (SMARCA4mut) cancers have focused on the identification of vulnerabilities that may be conferred in the context of the mutant state. Most notably, functional genomic screens to identify gene dependencies have identified the paralog, SMARCA2, as a synthetic lethality in cancers with inactivated SMARCA4. Although the ATPase activity of SMARCA2 is required for the proliferation of SMARCA4mut cells, bromodomain function is dispensible, as highlighted by the failure of inhibitors to the SMARCA2/4 bromodomain to suppress cell growth. Such findings have subsequently led to efforts to discover ATPase inhibitors, however the currently described inhibitors are dual inhibitors of SMARCA2/4 and are hampered by dose-limiting tolerability issues, preventing the full in vivo exploration of anti-tumor activity. Preclinical genetic studies would indicate that achieving selective inhibition of SMARCA2 will likely be essential for a successful therapeutic. Whereas the germline knockout of Smarca2 produced viable mice that are slightly larger than control littermates but show no other overt phenotypes, the knockout of Smarca4 is embryonic lethal and conditional deletion of Smarca4 has been associated with multiple tissue-specific phenotypes. More importantly, the co-deletion of Smarca2 and Smarca4 in adult mice was lethal due to vascular defects. Hence, SMARCA2 inhibitors with improved selectively over SMARCA4 will likely be required to achieve safe and maximal inhibition of SMARCA2 in this context. Proteolysis targeting chimeras (PROTACs) represent an emerging therapeutic modality to induce the degradation of target proteins by recruiting the protein of interest to an E3 ubiquitin ligase, leading to the subsequent tagging of the protein for proteasome-mediated destruction through the addition of ubiquitin. PROTACs offer several advantages over classical small molecule inhibitors, as they circumvent the requirement to employ ligands targeting the enzymatic function of the given target protein and they can function in a sub-stoichiometric manner enabling sustained pharmacodynamic effects at lower systemic exposures. Importantly, selective degradation using warhead ligands with nonselective binding properties has been demonstrated with PROTACs. Although the mechanisms underlying selective degradation remain to be fully elucidated, the ability to form protein-protein interactions between the target protein and E3 ligase within the ternary complex can contribute to a more productive and selective degrader. Hence, warhead selection and the choice of E3 ligase will play a critical role in determining whether selective degradation could be achieved using a PROTAC with equivalent binding affinities to multiple substrates. In this work we use otherwise inert ligands with equivalent binding affinities to the bromodomains of SMARCA2/4 and the 5th bromodomain of PBRM1 to develop a VHL-based PROTAC exhibiting potent and moderately selective degradation of SMARCA2. The VHL-SMARCA2 PROTAC elicits enhanced growth inhibitory effects both in vitro and in vivo in SMARCA4mut cancer models relative to SMARCA4 wild-type (SMARCA4wt) models, in the absence of considerable tolerability issues. In contrast to a previously described SMARCA2/4 ATPase inhibitor and PROTAC, these findings provide the an example of a selective SMARCA2 targeting agent and provide pharmacologic support of this previously defined synthetic lethality in SMARCA4mut cancers. To identify potent and selective PROTACs targeting SMARCA2, we linked a small-molecule ligand capable of binding the bromodomains of SMARCA2/4 and PBRM1 (5th BD) to a VHL-targeting moiety (information regarding chemical synthesis can be found in the Methods and as a Supplementary Note in the Supplemental Information). This work led to the identification of PROTAC, A947 (Fig. 1a). No difference in binding affinity to the SMARCA2 and SMARCA4 bromodomains was observed for A947 (Fig. 1b). (SMARCA2 Kd = 93 nM, SMARCA4 Kd = 65 nM) A947 potently degraded SMARCA2 in SW1573 cells with a DC50 (the drug concentration that results in 50% protein degradation) of 39 pM and achieving a maximal degradation of 96% at 10 nM (Fig. 1c, d). In contrast, ~28-fold higher concentration of A947 was needed to achieve a DC50 on SMARCA4 (1.1 nM), with a maximal degradation of SMARCA4 (92%) being achieved at concentrations approaching 100 nM. This degree of degradative selectivity was maintained independent of the specific isoform of SMARCA2/4 evaluated (Fig. 1e, Supplementary Table 1, Supplementary Fig. 1). Furthermore, A947 exhibited similar selectivity on SMARCA2 degradation over PBRM1 (Fig. 1d). The cellular degradation of SMARCA2 by A947 required both SMARCA2 and VHL binding, as loss of SMARCA2 could be mitigated by the addition of excess free SMARCA2/4 or VHL -binding ligands (Fig. 1f). In addition, a hydroxy-proline diastereomer of A947 expected to attenuate VHL binding (A857), as well as an analog lacking a critical phenol group in the SMARCA2-binding fragment (A858), were largely defective in degrading SMARCA2 in cells; with a negligible impact of A857 on SMARCA2 at the highest concentrations tested (Supplementary Fig. 2). The dependence on ubiquitination and proteosome-mediated degradation was demonstrated by the ability of an inhibitor to the ubiquitin activating enzyme (MLN-7243) and a proteasome inhibitor (MG-132) to block A947-mediated degradation of SMARCA2 (Fig. 1f). A947-mediated cellular degradation of SMARCA2 was rapid, with ~93% loss of the nuclear insoluble pool of SMARCA2 observed within 30 min (Supplementary Fig. 3a–c). Finally, A947 was equally efficient in degrading both the murine and rat orthologs of SMARCA2, as assessed by monitoring the cellular degradation of these orthologs ectopically expressed in cells expressing endogenous human or murine VHL (Fig. 1g, h, Supplementary Fig. 3d). To more broadly assess the impact of A947 on the ubiquitylome in cells, we carried out quantitative di-glycine reminant profiling by mass spectrometry following treatment of SW1573 cells with a high (500 nM) concentration of A957 to ensure maximal degradation of both SMARCA2/4 (Fig. 2a, Supplementary Data 1). We observed ubiquitination of multiple lysines on both SMARCA2 and SMARCA4, with the strongest ubiquitination on K1450 mapping to the bromodomain of SMARCA2/4. Based on the recently elucidated cryo-EM structure of the BAF complex, the majority of the ubiquitination occurred on lysines mapping to the ATPase module and HSA domain, with no ubiquitination observed on very N-terminal lysines that are predicted to be anchored within the core complex. Importantly, we observed no ubiquitination of core BAF complex or accessory proteins. Globally, ubiquitination mediated at this high concentration of A947 in cells was specific to SMARCA2/4. In further support of the selectivity, we quantified degradation at the proteome level by mass spectrometry (Fig. 2b, Supplementary Data 2). SMARCA2/4 and PBRM1 represented the only proteins impacted by A947. Taken together, the data indicate that A947 was highly specific for degrading the expected target proteins at high concentrations. We next evaluated the impact of A947 on cell proliferation. In SMARCA4mut NCI-H1944 cells, A947 treatment resulted in the dose-dependent inhibition of growth that was dependent upon SMARCA2 degradation, as the VHL and SMARCA2/4 -binding defective analogs (A857 and A858, respectively) were significantly weaker in cells (Fig. 3a). To more broadly assess cellular activity and determine whether the moderate selectivity in degradation translated to selective effects on cell growth, we profiled a panel of lung cancer models characterized by SMARCA4 mutation status (Fig. 3b, c, Supplementary Data 3). Two additional cell lines that were deficient in SMARCA2/4 expression were included as further controls for any non-specific effect of A947. SMARCA4mut lung cancer cell lines were most sensitive to A947 treatment, with a median IC50 of 7 nM across the panel of cell lines. We did not observe any relationship with the type of SMARCA4 variant and/or the position of the variant with cellular activity of A947 (Supplementary Fig. 4a). In contrast, SMARCA4WT cells were significantly less sensitive to A947, with a median IC50 of 86 nM across the panel of models evaluated. A947 treatment had no impact on growth of cells deficient in SMARCA2/4 expression. The difference in cellular growth inhibition between SMARCA4mut and SMARCA4WT models was not due to differences in the ability of A947 to degrade SMARCA2 (p = 0.52, ns) (Supplementary Fig. 4b, Supplementary Data 3). Furthermore, there was a range of IC50’s for growth inhibition within SMARCA4mut and SMARCA4WT models, however we did not observe a direct correlation with growth inhibition and degradation potency of SMARCA2 nor with degradation of SMARCA4 and/or PBRM1 degradation in SMARCA4WT models (Supplementary Fig. 4b–d, Supplementary Data 3). This differential in cellular sensitivity to A947 between SMARCA4 mutant and WT cells was also recapitulated upon longer treatment periods in clonogenic growth assays (Supplementary Fig. 4e). A947-mediated degradation resulted primarily in G1 arrest across SMARCA4mut models that was not observed in control cell lines (Fig. 3d). We also did not observe evidence for acute cytotoxicity. At the transcriptional level, A947-mediated SMARCA2 degradation resulted primarily in transcriptional suppression in SMARCA4mut cells, consistent with the role of SMARCA2 as a chromatin regulator (Fig. 3e, Supplementary Data 4). Importantly, we observed a strong correlation (r = 0.53, p < 2.2e16) between the transcriptional changes occurring between A947 treatment compared to inducible SMARCA2 knockdown by shRNA, further supporting the on-target effect of A947 in cells. To address whether A947 is active in vivo, we initially evaluated the pharmacodynamic (PD) effect following a single 40 mg per kg intravenous (IV) dose of A947 in SMARCA4mut HCC515 xenografts over a 2 week period (Fig. 4a, Supplementary Fig. 5). In addition to monitoring SMARCA2 protein levels, we evaluated 2 transcriptional target genes that were broadly regulated by SMARCA2 loss across SMARCA4mut models, KRT80 and PLAU. Both transcripts are directly regulated by SMARCA2, as previously defined by ChIP-seq and ATAC-seq studies. The IV administration of A947 resulted in rapid reduction (96% reduction by 4 h) in tumor SMARCA2 protein levels and achieved a maximal reduction by 24 h. Loss of SMARCA2 in situ was additionally confirmed by immunohistochemistry (Supplementary Fig. 5b). Decreases in the transcriptional readouts followed, achieving maximal target gene suppression by 96 h post-dose. A slight rebound in SMARCA2 protein levels was observed over the 14 day period as tumor concentrations of A947 decreased; however due to prolonged A947 tumor exposures, these concentrations never reached baseline levels. Differential re-expression of KRT80 and PLAU transcripts was observed following one week after dose administration with PLAU mRNA levels paralleling SMARCA2 protein levels. We subsequently took the dosing regimen of 40 mg per kg every other week i.v. administration of A947 forward into efficacy studies in two different SMARCA4mut lung cancer xenograft models, HCC515 and HCC2302 (Fig. 4b, c). Statistically-significant decreases in tumor growth were observed in both models, highlighted by near complete growth inhibition in the HCC515 model and 60% tumor growth inhibition in the HCC2302 model. Tumor growth inhibition was dependent upon SMARCA2 degradation and was not due to non-specific effects of the chemical scaffold, as VHL and SMARCA2/4 -binding defective analogs A857 and A858 were not efficacious (Supplementary Fig. 6). Furthermore, tumor growth inhibition was observed in the absence of any appreciable loss in body weight at this dose and regimen, indicating that efficacy was not a consequence of in vivo toxicity (Supplementary Fig. 7a, b). Tumor pharmacodynamic biomarker responses were also measured at end of study; 24 h after administration of a final dose. A947 treatment led to more than a 95% decrease in tumor SMARCA2 protein levels in both models, however a slightly stronger suppression of KRT80 transcript was observed in the HCC515 model (Fig. 4d). To more extensively address the transcriptional impact of A947, we carried out RNAseq on mRNA isolated from these xenografts and monitored transcripts (n = 412) that were determined from in vitro studies to represent acutely-suppressed and sustained targets of SMARCA2 loss common to both models (Fig. 4e, Supplementary Data 4). Consistent with the differential degree of KRT80 suppression, we observe slightly stronger suppression of SMARCA2-regulated genes in the HCC515 xenograft model, suggesting that slight differences in the pharmacodynamic effect may underlie the differences in efficacy between these models. In order to address whether the tumor growth inhibition observed in SMARCA4mut models was due to a tumor cell autonomous effect of SMARCA2 degradation, we evaluated A947 administration in the SMARCA4wt Calu-6 xenograft model (Fig. 4f). Applying the same dose and regimen as used in the SMARCA4mut xenograft studies, A947 did not result in tumor growth inhibition in SMARCA4wt Calu-6 xenografts despite achieving greater than 95% degradation of SMARCA2 protein (Fig. 4g, Supplementary Fig. 8a). Moderate degradation of SMARCA4 and PBRM1 was observed at the 24 h post-last dose timepoint, with 58% and 57% decreases respectively. Although degradation selectivity cannot be addressed in SMARCA4mut models due to the deficiency in human SMARCA4, we were able to monitor murine SMARCA4 protein levels within the tumor microenvironment of HCC515 and HCC2302 xenografts (see Supplementary Fig. 5b for in situ confirmation of stromal SMARCA4 signal). Analogous to the Calu-6 xenografts, 57% and 69% reductions in murine SMARCA4 levels were observed in these studies, suggesting that murine SMARCA4 is degraded similarly to human SMARCA4 (Supplementary Fig. 8b, c). Hence, the moderate selectivity for SMARCA2 over SMARCA4 degradation observed in vitro for A947 translated to a moderate degradative selectivity in vivo at this dose and regimen. Taken together, the data are supportive of a tumor cell intrinsic effect of SMARCA2 degradation and provide pharmacologic support of the synthetic lethal interaction. Although SMARCA2 degraders would have the potential to be developed as single agents in the clinic to treat SMARCA4mut cancers, we have begun to address whether rational and ubiquitously active pharmacologic combinations exist for SMARCA2 degraders in SMARCA4mut cancers. To assess combination effects, we screened a library of 723 experimental and clinically approved agents in combination with A947 across 4 SMARCA4mut lung cancer cell lines (Fig. 5a, Supplementary Data 5). Although A947 treatment sensitized to unique compounds in a given model, MCL1 inhibition represented the only combination exhibiting strong sensitization with A947 in more than one SMARCA4mut model (3 of the 4 models), as well as with multiple inhibitors. To confirm and extend this observation, we carried out a matrix titration of 2 separate MCL1 inhibitors (AMG-176 and S63845) with A947 and evaluated synergistic growth inhibition based upon the Bliss independence model in 5 SMARCA4mut models (Fig. 5b, Supplementary Fig. 9). In all cases, A947-mediated SMARCA2 degradation exhibited a synergistic interaction with MCL1 inhibition. Synergistic growth inhibition was not observed in 2 SMARCA4WT models (Supplementary Fig. 9). Given the cytostatic effect of A947 treatment as a single agent and the anti-apoptotic function of MCL1, we evaluated whether the synergy was due to the ability of the combination to induce apoptosis. Indeed, live cell imaging of activated caspase 3/7 in SMARCA4mut cells revealed that the combination was able drive these cells toward apoptotic cell death (Fig. 5c). These data, combined with previously published work leveraging a genetically-engineered degron tagging approach, support a potential combination of SMARCA2 degraders with MCL1 inhibitors. In this study, we report a moderately selective SMARCA2-targeting PROTAC, A947, that is active both in vitro and in vivo in SMARCA4mut NSCLC models. A947 exhibited a 28-fold selectivity in degradation of SMARCA2 over SMARCA4 and notably exhibited no unexpected off-target effects at high concentration in both global ubiquitinome and proteome studies in cells. The ability of A947 to phenocopy the effect of inducible shRNA-mediated knockdown of SMARCA2 on transcriptome expression further supported the molecule’s on target effect. A947 differs from a previously reported SMARCA2/4-targeting PROTAC, ACBI1. Despite differences in the models tested, A947 exhibited greater SMARCA2 degradation potency (39pM DC50 v. 6 nM DC50) and selectivity over SMARCA4 (28.2-fold v. 1.8-fold shift in DC50). This 28-fold selectivity in cellular degradation achieved with A947 translated to selective tumor growth inhibition both in vitro and in vivo in SMARCA4mut cancers compared to SMARCA4wt cell line models. Pharmacodynamic profiling revealed that moderate in vivo degradation (~60%) of SMARCA4 could occur along with the associated strong SMARCA2 degradation (>95%), however the former activity was not sufficient to either drive efficacy in a SMARCA4wt model nor to impact tolerability at the dosing regimen utilized. Taken together, these data provide pharmacologic support of this paralog-mediated, synthetic lethality; and provide an early indication that a potent SMARCA2 degrader with ~30-fold selectivity could be safely delivered. An important aspect of this effort has been the ability to exploit non-selective SMARCA2/4 binding ligands to achieve selective cellular degradation in the context of the PROTAC. However, despite our ability to achieve selective cellular degradation, the mechanism underlying selectivity remains to be elucidated. Differences in ternary complex affinity have been attributed to several reported examples of selective degraders that utilize ligands with promiscuous binding properties and ternary complex formation has been previously observed with non-selective SMARCA2/4 PROTACs. However, it remains to be determined whether fine differences in cooperativity between SMARCA2 and the VHL complex (VCB: VHL-ElonginC-ElonginB) compared to SMARCA4/VCB underlie the difference in cellular degradation between the SMARCA2/4 paralogs. In vitro measurements of ternary complex formation have relied on the use of isolated bromodomains, which may not simulate the physiologically relevant substrate:PROTAC:VCB interface in the context of the native BAF complex. Selectivity could also be potentially driven by cell-intrinsic factors, as opposed to the biophysical properties of the ternary complex. Nevertheless, understanding the biochemical and/or cellular mechanism of degradation selectivity of A947 remains an active area of research. Due to their high molecular weight and poor physiochemical properties, the oral delivery of VHL-based PROTACs can be problematic. Preclinical efficacy studies have generally required intraperitoneal or subcutaneous administration of VHL-PROTACs at high doses and high frequencies, with few exceptions. While the poor physiochemical properties of VHL-based PROTACs complicate their oral delivery, opportunities exist for alternative parenteral routes of administration. Intravenous (IV) dosing offers some obvious advantages, such as the lack of physiological barriers to absorption, but requires solubility consistent with the target dose and an intravenous half-life (T1/2) sufficiently long to satisfy the PKPD requirements and a likely intermittent dosing schedule. Prolonged IV T1/2 can be achieved by increasing PROTAC affinity to the lipid and phospholipid cell components (either via addition of a positive charge or increased lipophilicity), while decreasing or maintaining intrinsic metabolic stability. The amine containing linker contained in A947 promotes high solubility in the IV formulation, as well as moderate CL (<16 ml/min/kg) and high affinity for body tissues in rodents, resulting in VDss > 6 L/kg and T1/2 > 6 h. Interestingly, in rodents the residence time of A947 in tumors (and presumably in other body tissues) appears to be much higher than in plasma, suggesting that dissociation from tissues is a slower process compared to metabolic elimination from plasma (Fig. 4a, Supplementary Fig. 5a). While these observations were not mechanistically investigated, they are consistent with intracellular lysosomal trapping, a phenomenon frequently observed for lipophilic basic amines. Overall, our data may have future clinical implications, offering a potential therapeutic option for patients harboring SMARCA4mut cancers. Although A947-mediated degradation of SMARCA2 results primarily in cytostasis that is consistent with prior genetic perturbation studies, the overall depth of single-agent efficacy in the context of an immune-competent animal and/or upon prolonged dosing remains to be determined. Nevertheless, we considered whether rational drug combinations exist as a means to potentiate the tumor cell intrinsic, cellular activity of A947. We specifically sought to identify drug synergies that were broadly active and not necessarily cell context specific. Through pharmacologic profiling, we determined that inhibition of the BCL2-family pro-survival protein, MCL1, could synergize with SMARCA2 degradation across multiple SMARCA4mut models. Interestingly, MCL1 was additionally identified in a genome-wide CRISPR knockout screen to identify sensitizers to SMARCA2 loss in a genetically engineered system whereby SMARCA2 was endogenously tagged with the SMASh degron to enable degradation with the NS3 protease inhibitor, asunaprevir. Drug sensitization was observed with additional anti-apoptotic drugs including BCL-XL and IAP antagonists, however these effects were cell line specific. Combined with the prior report, these data may indicate the broader utility combining MCL1 antagonists with a SMARCA2 PROTAC. Multiple MCL1 inhibitors are under early clinical investigation and are being considered in combination with other therapeutics in solid tumors as a means to lower the apoptosis threshold. This study complies with all relevant ethical regulations. Animals were maintained in accordance with the Guide for the Care and Use of Laboratory Animals. Genentech is an AAALAC-accredited facility and all animal activities in the research studies were conducted under protocols approved by the Genentech Institutional Animal Care and Use Committee (IACUC). A947, A857, A858 and the free VHL ligand (A2702) were prepared by Arvinas, Inc., with detailed chemical synthesis described at the end of this Methods section. The chemical compounds MLN-7243 and MG-132 were obtained from SelleckChem. The SMARCA4/2 bromodomain inhibitor (example 47, WO2016138114) was used in competition studies (Fig. 1f). The following antibodies were utilized: SMARCA2 (Cell Signaling, 11966, dilution 1:2000), SMARCA4 (Abcam, ab110641, dilution 1:1000), PBRM1 (Bethyl Labs, A301-591A, dilution 1:1000), SMARCC1 (Cell Signaling, 11956, dilution 1:1000), HDAC1 (Cell Signaling, 34589, dilution 1:1000), VHL (Cell Signaling, 68547, dilution 1:1000), FLAG (Sigma, F3165, dilution 1:1000), Lamin A/C (Cell Signaling, 4777, dilution 1:1000), α−Tubulin (Sigma, T6074, dilution 1:1000), β-Tubulin (Cell Signaling, 2128, dilution 1:1000), β-Actin (Cell Signaling, 3700, 1:1000 and 4970, dilution 1:1000) and total ubiquitin (Cell Signaling, 3936, dilution 1:1000). Cell lines were obtained from the following sources (indicated in Supplementary Data 3): American Type Culture Collection (ATCC), Japanese Collection of Research Bioresources Cell Bank (JHSF), Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ), Riken, or licensed from UT Southwestern (UTSW). 293 T cells were obtained from ATCC. Cells were maintained in RPMI1640 supplemented with 10% Fetal Bovine Serum (FBS) and 2mM L-Glutamine, under 5% CO2 at 37 °C, with the exception of SW1573 (DMEM) and Calu-6 (EMEM). Cell line identity was verified by high-throughput single nucleotide polymorphism (SNP) genotyping using Illumina Golden Gate multiplexed assays. SNP profiles were compared to SNP calls from internal and external databases to determine or confirm ancestry. All cell lines tested negative for mycoplasma contamination prior to storage/use at our institute. AlphaLISA® studies were carried out using histidine (His)-tagged recombinant human SMARCA2 (aa.1377-1486; NP_620614) or SMARCA4 (aa. 1448-1575; NP_003063) proteins expressed in Escherichia coli. Compounds were diluted with 3-fold dilutions (11-point) in 96-well plates with a top concentration of 10 mM in 100% DMSO. Compounds were further diluted tenfold in Alpha LISA buffer consisting of 50 mM HEPES (Life Technologies 15630-080), 50 mM NaCl (Sigma BCBW5699), 69uM Brij (Sigma SLBM8986V), and 0.1 mg/ml BSA (Sigma A7906-100G) brought up to 100 mL final volume in water (Sigma RNBG4333), resulting in a top concentration of 1 mM in 10% DMSO. 3uL of this dilution was spotted into 384 well plate(s) (Perkin Elmer 6007290), final reaction volume of 30 uL, with compounds having a final top concentration of 0.1 mM. The final 30 ul reaction consisted of the following components: Compound, 7 nM His-SMARCA2 (or His-SMARCA4), 20 nM SMARCA2/SMARCA4 Biotin Probe (example 248, WO2016138114), 1:400 Dilution of anti-His Alpha-LISA Acceptor Beads (Perkin Elmer AL128M), and 1:400 Dilution of Streptavidin Alpha-LISA Donor Beads (Perkin Elmer 6760002). To make a working stock of His-SMARCA2/SMARCA4 and biotinylated probe, 1.7 uL of 60uM His-SMARCA2 (or His-SMARCA4) stock and 30 uL of 10 uM biotinylated probe (diluted in Alpha LISA buffer) was added to 20 mL Alpha LISA buffer to give final concentrations of 17.5 nM protein and 50 nM probe. Mixtures were then incubated at room temperature for 5–10 min. Then 12 uL of protein/probe mixture was added to each reaction well in 384 well plate containing compounds (or probe only for background control wells). Plates were incubated for 10–15 min at room temperature. Anti-His6x Acceptor beads were diluted 100× in Alpha-LISA Buffer protected from light and 7.5 uL added to each well. Plates were incubated 10–15 min at room temperature protected from light. Streptavidin donor beads were diluted 100X in Alpha-LISA Buffer protected from light and 7.5 uL added to each well. Plates were incubated for at least 15 min (no more than 4 h) at room temperature protected from light and plates read at 615 nm on a micro-plate reader. For data analysis, the averages of control wells (probe + protein only max signal and probe only background control) were calculated and emission values at 615 nm were used to calculate percent displacement values using this formula: Other values, such as mean and standard deviation, were calculated using GraphPad Prism software package. SW1573 cells were seeded at 8000/well in 96-well black/clear-bottom plates with 180ul DMEM containing 1% pen-strep, 1% HEPES and 10% FBS and incubated overnight at 37 °C to allow adherence. The next morning cells were treated with 20 uL of 10× compound and incubated for an additional 20 h. Cells were then washed with ice cold DPBS, then 50 uL of ice cold 4% PFA (Electron Microscopy Sciences 15711)/DPBS was added, and plates were then incubated at RT for 20 min. PFA was then removed and 200 uL of TBS-T containing 0.5% Triton X 100 (Sigma T8787) was added. Plates were incubated at RT for 30 min. TBS-T/Triton X was then removed, 50 uL of Li-cor blocking solution (Li-Cor 927-50003) was added, and plates were incubated at RT for 1 h. Blocking solution was removed, 50 uL of Li-Cor blocking solution containing primary antibody to SMARCA2 (1:2000) and α-Tubulin (1:2000) or SMARCA4 (1:1000) and α-Tubulin (1:2000) was added, and plates were incubated at 4 °C overnight. The next day, plates were washed 3× with TBS-T and 50uL of secondary antibody cocktail in Li-Cor blocking solution was added (IRDye 800CW Goat anti-rabbit IgG, Li-Cor 926-32211 and IRDye 680RD Goat anti-mouse IgG, Li-Cor 926-68070; both secondary antibodies are diluted 1:5000). Plates were incubated at RT for 1 h protected from light. Then plates were washed twice with TBS-T and excess liquid removed. Plates were read on the Li-Cor Odyssey with default intensity setting of 5.0 for both channels. Li-Cor images were analyzed with the in-cell Western feature of Image Studio Lite. Assays were run with technical duplicates and multiple biological replicates (>2) with error calculated with GraphPad prism using 95% confidence interval. Following treatment of cells as indicated, cells were lysed in RIPA buffer containing 0.5 M NaCl, then homogenized for 3 min at speed 10 (NextAdavance, Bullet BlenderR 24). Proteins (12 ug or 18 ug) were resolved on 4–12% Bis-Tris or 3-8% Tris-Acetate gels and transferred to nitrocellulose membranes by iBlot. Membranes were incubated overnight with primary antibodies as indicated. IRDyeR -conjugated secondary antibodies were used to bind primary antibodies and images were visualized on the Odyssey Imager (LI-COR). The pLenti6.3 vector system was used for all ectopic expression experiments. All DNA constructs were generated with a C-terminal FLAG tag. The human SMARCA2 and SMARCA4 isoform sequences are annotated in Supplementary Table 1. For SMARCA2 orthologs, the following sequences were utilized: human (NM_139045.2), rat (XM_006231227.3), and mouse (NM_011416.2). The packaging and envelope vectors, Δ8.91 and VSV.G were cotransfected with pLenti6.3-based constructs into 293 T cells by using lipofectamine 2000 (Invitrogen). Media containing lentiviral particles was collected 3 days post transfection, filtered through 0.45 μm filters. Cells (TOV112D (human) and/or LA-4 (murine)) were transduced with pLenti6.3- constructs particles, and selected with 8 μg/mL blasticidin (Gibco, A1113903) 3 days after transduction. Inducible SMARCA2 knockdown cell lines were generated using shRNAs directed against SMARCA2 tandemly delivered in a modified pBH1.2 piggy-bac system (Smarca2_iKD_dual62_pBH1.2). The following shRNAs were utlilized: shRNA6:GATCCGTCTCGTCGAGCAATCATTTGGTTGTAGTGAAATAtATATTAAACAACCAAATGATTGCTCGACGTTACGGTAC and shRNA2:GATCCGTCTGACTGTTCACGTTCATCCTGGTAGTGAAATAtATATTAAACCAGGATGAACGTGAACAGTCTTACGGTAC. pBO (Piggybac transposase) were co- transfected with Smarca2_iKD_dual62_pBH1.2 into HCC2302 cells by using lipofectamine 2000 (ThermoFisher Scientific). Cells were selected with 2 μg/mL puromycin (Gibco) 3 days after transfection and subsequently subcloned. Cells were lysed on plate using a lysis buffer consisting of 9 M urea, 50 mM HEPES (pH 8.5), and complete-mini (EDTA free) protease inhibitor (Roche). Protein concentrations were then estimated by BCA assay (ThermoFisher Pierce, Rockford, IL). Disulfide bonds were reduced with 5 mM DTT (45 min, 37 °C), followed by alkylation of cysteine residues by 15 mM IAA (30 min, RT Dark), and finally capped by the addition of 5 mM DTT (15 min, RT Dark). Proteins were then precipitated by chloroform/methanol precipitation and resuspended in digestion buffer (8 M urea, 50 mM HEPES pH 8.5). Samples were diluted to 4 M urea before initial protein digestion was performed by the addition of 1:100 LysC followed by incubation at 37 °C for 3 h. Samples were then diluted to 1.5 M urea with 50 mM HEPES (pH 8.5) before the addition of 1:50 Trypsin and incubation overnight at room temperature. Peptide mixtures were acidified and desalted via solid phase extraction (SPE; SepPak - Waters, Boston, MA). For global proteome analysis, peptides were resuspended in 200 mM HEPES (pH 8.5) and a 100 µg aliquot of peptides was mixed with tandem mass tags (TMT or TMTpro, ThermoFisher Pierce, Rockford, IL) at a label to protein ratio of 2:1. After 1 h of labeling, the reaction was quenched by the addition of 5% hydroxylamine and incubated at room temperature for 15 min. Labeled peptides were then mixed, acidified, and purified by SPE. Samples were separated by offline high pH reversed-phase fractionation using an ammonium formate based buffer system delivered by an 1100 HPLC system (Agilent). Peptides were separated over a 2.1 × 150 mm, 3.5 µm 300Extend-C18 Zorbax column (Agilent) and separated over a 75-minute gradient from 5% ACN to 85% ACN into 96 fractions. The fractions were concatenated into 24 samples of which 12 or 24 were analyzed for proteome quantification. Fractions were concatenated by mixing different parts of the gradient to produce samples that would be orthogonal to downstream low pH reversed phase LC-MS/MS. Combined fractions were dried, desalted by SPE, and dried again. For ubiquitylome analysis, peptides were resuspended in 1X detergent containing IAP buffer (Cell Signaling Technology), cleared by high speed centrifugation (18,000 × g, 10 min) and enriched using an automated procedure previously described. Enriched ubiquitinated peptides were prepared for multiplexed quantitative analysis as previously described except that TMTpro reagents were used. All six fractions were analyzed by LC-MS/MS. Quantitative mass spectrometry analysis was performed on an Orbitrap Fusion Lumos or Orbitrap Eclipse mass spectrometer (ThermoFisher, San Jose, CA) coupled to a Waters NanoAcquity (Waters, Milford, MA) or Thermo Ultimate 3000 RSLCnano ProFlow (ThermoFisher, San Jose, CA) HPLC. Peptides were separated over a 100 µm × 250 mm PicoFrit column (New Objective) packed with 1.7 µm BEH-130 C18 (Waters, Milford, MA) at a flow rate of 450 or 500 nL/min or over a 25 cm IonOpticks Aurora column (IonOpticks, Fitzroy, Australia) at 300 nL/min for a total run time of 180 min. The gradient started at 2–5% B (98% ACN, 1% FA) and ended at 30% B over 140 min and then to 50% B at 160 min. Peptides were surveyed via Orbitrap FTMS1 analysis (120,000 resolution, AGC = 1 × 106, maximum injection time [max IT] = 50 ms) and the most intense 10 peaks were selected for MS/MS ensuring that any given peak was only selected every 35 or 45 s (tolerance = 10 ppm). For all runs, “one precursor per charge state” was ON. For data collected on the Eclipse mass spectrometer, Advanced Precursor Detection (APD), FAIMS (CVs = −50, −70), and the Precursor Fit Filter (70% fit and 0.5 fit window) were employed. For peptide identification, precursors were isolated using the quadrupole (0.5 Th window), fragmented using CAD (NCE = 35 for TMT and NCE = 30 for TMTpro) and analyzed in the ion trap using a Turbo speed scan (AGC = 2 × 104, maxIT = 100 ms) for proteome analysis or an Orbitrap scan at 15,000 resolution (AGC = 7.5 × 104, maxIT = 200 ms) for ubiquitylome analysis. A real-time database search was utilized for both proteome and ubiquitylation quantification on the Eclipse mass spectrometer. The real-time database search performed an in silico trypsin digest with full specificity and 1 missed cleavage and used concatenated decoy proteins to calculate FDR in real time. The precursor PPM tolerance was set to 10 ppm and the real-time search static and dynamic modifications matched the search parameters below. For quantitative SPS-MS3 analysis, the top 8 ions were simultaneously selected (synchronous precursor selection – SPS, AGC = 1.5 × 105 or 3.0 × 105 [proteome] or 4.0 × 105 [ubiquitylome], max IT = 150 ms [proteome] or 400 ms [ubiquitylome]) and fragmented by HCD (NCE = 55 [TMT] or 40 [TMTpro]) before analysis in the Orbitrap (resolution = 50,000). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD036865. All mass spectrometry data was searched using Mascot against a concatenated target-decoy human database (downloaded June 2016) containing common contaminant sequences. For the database search a precursor mass tolerance of 25 ppm (TMTpro) or 50 ppm (TMT), fragment ion tolerance of 0.5 Da (TMTpro) or 0.8 Da (TMT), and 1–2 (proteome) or 3 (ubiquitylome) missed cleavages. Carbamidomethyl cysteine (+57.0214) and TMT labeled n-terminus (+229.1629 for TMT and +304.2071 for TMTpro) were applied as static modifications for all analyses. For proteome analysis, TMT or TMTpro on lysine was also set as a static modification. Methionine oxidation (+15.9949) and TMT or TMTpro on tyrosine were set as a dynamic modifications for all searches. For ubiquitylome searches, TMTpro on lysine and TMTpro-KGG on lysine (+418.2510) were considered as variable modifications. Peptide spectral matches for each run were filtered using line discriminant analysis (LDA) to a false discovery rate (FDR) of 2% and subsequently as an aggregate to a protein level FDR of 2%. Quantification and statistical testing of TMT proteomics data was performed using MSstats. Prior to MSstats analysis, peptide spectral matches (PSMs) were filtered to remove matches from decoy proteins; peptides with length <7; with isolation specificity <50%; with reporter ion intensity <256; and with summed reporter ion intensity (across all channels) <30,000. In addition, to separate peptides shared between SMARCA2 and SMARCA4, peptides matching to either protein were labeled as SMARCA2, SMARCA4, or SMARCA2/4 prior to MSstats analysis. This enables quantification of shared peptides to be performed as if they were a separate protein group. In the case of redundant PSMs (i.e., multiple PSMs in one MS run that map to the same peptide), PSMs were summarized by the maximum reporter ion intensity per peptide and channel and median equalized. In the case of redundant PSMs across fractions (i.e., redundant matching PSMs being found in multiple fractionated runs), PSMs were summarized by selecting the fraction with the maximum reporter ion intensity for each PSM. Protein level summarization was performed using a Tukey median polish approach. Differential abundance analyses between conditions were performed in MSstats based on a linear mixed-effects model per protein, and resulting two-sided p-values adjusted for multiple hypothesis testing by using the Benjamini-Hochberg procedure. For ubiquitylome analysis, the log2 ratio values of each ubiquitinated peptide were normalized to the corresponding protein measurement before visualization; if a protein log2 ratio was not measure the ubiquitylation measurement was carried forward unchanged. Cells were plated in 384 well plates (PhenoPlate™ 384-well microplates, PerkinElmer) at 4000 cells/well overnight. Cells were subsequently treated for 24 h with a dose response of A947 prior to fixation with 4% formaldehyde for 15 min. Plates were washed three time with PBS, incubated with a blocking solution (10%FCS, 1%BSA, 0.1%Triton, 0.01%Azide, X-100 in PBS) for 1.5 h, and subsequently treated with primary antibody diluted 1:1200 in blocking buffer overnight at 4 °C. Following washing (3×) in PBS, cells were incubated with secondary antibodies (rabbit-Alexa 488, ThermoFisher A21206, 1:1000) for 1 h at room temperature in the dark. Hoechst H3570 (ThermoFisher H3570) at 1:5000 was added to the wells and the plates were incubated for an additional 30 min. Plates were wash 3× in PBS and imaged on an Opera Phenix™ High Content Screening System (PerkinElmer). Using Hoechst H3570 nuclear staining as a mask, nuclear SMARCA2 and SMARCA4 mean signal intensity were quantified. Cells were plated in 96-well plates at 500 cells per well and treated with a dose range of A947 starting with a highest concentration of 500 nM. After incubating for 7 days, viability was measured using CellTiter-Glo (Promega) reagent. Reagent was added directly to the cells at a 1/1 ratio of reagent to cell culture medium. Following a 15 min incubation, luminescence was measured using the multimode plate reader EnVision 2105 (PerkinElmer). Viability was normalized to DMSO treated control cells. Cell viability experiments were performed in triplicate cultures. Cells (1500–5000, depending on doubling time) were plated in 12-well plates for 24 h prior to treatment with fresh media containing compounds at indicated concentrations. Fresh media containing compound was replaced every 3–4 days until cells reached confluence to stop culture. Colonies were visualized by staining with 0.5% crystal violet for 20 min at room temperature. Cell lines were treated for 48 h with a dose response of A947 prior to pulsing for 15 min with 10uM EdU. Cells were subsequently fixed in 4% paraformaldehyde for 10 min, washed 3 times with PBS and then blocked and permeabilized in PBS containing 10% FBS, 1% BSA, 0.1% TX-100, and 0.01% NaN3 for 1 h at room temperature. Permeabilization buffer was removed and the cells were washed 3× with PBS. The Click-iT® reaction was perform according to the manufacturer’s (Invitrogen C10337) protocol. Following a 30 min incubation in the dark, the cells were washed 3 times with PBS. For nuclear staining, cells were treated with Hoechst 33342 (ThermoFisher) at 1:10000 for 10 min at room temperature. Cells were then washed again 3 times with PBS and imaged on Opera Phenix Plus High-Content Screening System (PerkinElmer). Image analysis was conducted with MATLAB standard and custom-written scripts (https://github.com/scappell/Cell_tracking). For each cell, the integrated nuclear Hoechst signal and EdU positivity were used to determine cell cycle phase and values were averaged for each treatment. Live cell imaging for activated caspase-3/7 was performed using the IncuCyte® ZOOM (Essen Bioscience). Cells were seeded in 96 well plates and treated the next day with 100 nM A947 and/or 1uM AMG-176 in the presence of Caspase-3/7 Green Detection Reagent (Essen Bioscience). Fluorescence was monitored over a 48 h period, with data collection every 4 h. Five planes of view were collected per well using 10× objective. Both phase contrast and green channel were collected for all wells. Data are presented as fluorescent events per well. For in vitro gene expression studies, HCC2302 or HCC515 cells were treated with DMSO or A947(100 nM) for both 24 h and 96 h prior to isolation of total RNA using the MagMax mirVana total RNA isolation kit (ThermoFisher Scientific, A27828). In addition, HCC2302-shNTC and HCC2302-shSMARCA2 cells were treated for 168 h prior to isolation of RNA. For in vivo gene expression studies, total RNA was isolated from xenograft tissues as above. RNA concentrations were measured by NanoDrop8000 (ThermoFisher). Integrity of RNA was assessed by Bioanalyzer 2100 prior to library generation using 500 ng RNA. Libraries were prepared using the TruSeq Stranaded Total RNA Library Prep Kit (Illumina), multiplexed and sequenced on Illumina HiSeq2500 (Illumina) to generate ~30 M single end, 50 base pair reads. Raw sequencing reads for in vitro samples were mapped to the UCSC human genome (GRCh38/hg38) using GSNAP software. In order to remove potential mouse stromal contamination for in vivo xenograft samples, raw sequencing reads were stripped of reads showing complete alignment to the UCSC mouse genome (mm10) using Xenome software. Remaining reads not showing complete alignment to mm10 were then mapped to GRCh38/hg38 using GSNAP software. Gene expression counts were obtained by quantifying the number of reads uniquely mapping to each gene locus. Lowly expressed genes were removed from all samples using a high-pass filter for genes with at least 15 counts in at least 10% of samples (6 of 54). Quantile normalized Log2 counts per million (LogCPM) of sufficiently covered genes were generated using the voom function of the limma analysis pipeline. Differential gene expression analysis was performed using edgeR. Significantly downregulated or upregulated genes were defined by a log fold change (Log2FC) absolute value >1 and a false discovery rate (FDR) < 0.05. To evaluate in vivo samples, a consensus, putative SMARCA2 target gene set was defined by genes significantly downregulated by A947 treatment in both HCC515 and HCC2302 in vitro samples at both early (24 h) and late (96 h) time points. RNAseq data has been deposited in the Gene Expression Omnibus database under the accession code GSE205542. Female Crl:NU-Foxn1nu (NU/NU Nude) or CB17/Icr-Prkdcscid/IcrIcoCrl (Fox Chase CB17) mice aged 6–8 weeks were purchased for Charles River laboratories. Mice were housed in individually ventilated cages within animal rooms maintained on a 14:10 h, light:dark cycle. Animal rooms were temperature and humidity-controlled, between 68–79 °F and 30–70% respectively, with 10–15 room air exchanges per hour. Mice received food and water ad libitum and were allowed to acclimate for 1–2 weeks before being used for experiments. All animal work was approved and conducted in accordance with the approval from the Institutional Animal Care and Use Committee (IACUC). All animal studies complied with the ethical regulations and humane endpoint criteria according to the NIH Guidelines for the Care and Use of Laboratory Animals. Genentech is an AAALAC-accredited facility and all animal activities in the research studies were conducted under protocols approved by the Genentech Institutional Animal Care and Use Committee (IACUC). The maximal subcutaneous tumor size/burden allowed (2000 mm3) was not exceeded in this study. Euthanasia was carried out by anesthetization through isoflurane inhalation followed by cardiac exsanguination. HCC515 and HCC2302 were propagated in RPMI-1640/10%FBS (Gibco, #61870-036), and Calu-6 in EMEM/10%FBS (ATCC, #30-2003). The cells were lifted with Trypsin (0.25%) (Gibco, #25200-056), spun down at 500 × g, washed at least 3 times with DPBS (Gibco, #14190-144), and the pellet resuspended in 50%Matrigel (Corning, #354234)/50% phenol-red free RPMI-1640 (Gibco, #11835-030). Implant cell numbers for HCC515 and Calu-6 was 5 × 10^6 and for HCC2032 was 10 × 10^6 in a 100 or 200 ul volume. Tumor cells were implanted into the right flank of the NU/NU or FOX CHASE SCID mice. The Tumor growth was monitored daily, and tumors were measured twice a week using digital calipers. Tumor volume was determined using the following formula (width × width × length)/2), where all measurements are in mm and the tumor volume is in mm3. The treatment started once the average tumor volume reached 150–200 mm3. Treatment was started ~3 weeks after cell implantation. The animals were randomly assigned into separate groups (n = 6–10 animals per group) such that each group had nearly equal starting average tumor volume. Treatment groups were randomly assigned into groups treated with vehicle and A947. A947 was dosed 5 mg per kg of body weight into the lateral tail vein intravenously once a week or every other week for tumor growth studies or only once for PK/PD studies. A947 was formulated for intravenous dosing in 10% Hydroxypropyl Beta Cyclodextrin (HP-b-CD) and 50 mM sodium acetate in water (pH 4.0). All dosing solutions were filtered prior to injection using a 0.2 micron filter to endure sterility. Mice were weighed twice a week, and dosing was performed the treatments were given according to the mouse’s individual weight. Mice were euthanized using an IACUC approved method of euthanasia when an individual mouse reached a maximum tumor size humane endpoint, defined according to institutional policy concerning tumor endpoints in rodents. In addition, to prevent excessive pain or distress, the mice were euthanized if the tumors became ulcerated or if the mice showed any signs of ill health. Post euthanasia, blood and various tissues including tumors were collected for further analyses. Analyses and comparisons of tumor growth were performed using a package of customized functions (https://github.com/wfforrest/maeve) in R (Version 3.4.2 and 3.6.2; R Foundation for Statistical Computing; Vienna, Austria), which integrates software from open source packages (e.g., lme4, mgcv, gamm4, multcomp, settings, and plyr) and several packages from tidyverse (e.g., magrittr, dplyr, tidyr, and ggplot2). Briefly, as tumors generally exhibit exponential growth, tumor volumes were subjected to natural log transformations before analysis. Estimates of group-level efficacy were obtained by calculating percent tumor growth inhibition (TGI). This value represents the percent difference between the area under the curves (AUCs) of the treatment and reference group fits which are calculated after back-transforming tumor volumes to the original scale, correcting for starting tumor burden, and averaging over a common time period. Positive values indicate anti-tumor effects, with 100% denoting stasis and values >100% denoting regression (negative values indicate a pro-tumor effect). Xenografts were harvested, divided into pieces, flash frozen and stored at −80C. For transcript-based pharmacodynamic readouts, RNA was isolated using the MagMAX mirVana total RNA isolation kit. For protein-based pharmacodynamic readouts, RIPA + Halt protease inhibitor (Thermo Fisher, #74830) was used at 400 ul per tube, regardless of tumor weight. A steel ball was used in each sample in the TissueLyzer at 26 Hz for 4 min. The homogenization block was stopped half way through the process and the block flipped over for the duration of the process. Lysates were sonicated for 30 s on 20 Hz in an icy bath. The lysates were spun clean at 15,000 RPM for 15 min at 4 C. Samples were then assessed for concentration by BCA at a dilution of 1:25. Western samples were prepared at 1 ug/ul in SDS-PAGE loading buffer/denaturing agent (Life Technologies, #NP0007 + #B0009) and denatured at 95 C for 5 min. Samples were used immediately or frozen at −20C until blotted. Protein (8 ug) was loaded on 4–15% Criterion Tris/Glycine gels (Bio-Rad, #5671085) and run for 60 min at 150 constant volts in 1X Tris/Glycine buffer (Bio-Rad, #1610732). Protein was transferred from gels to nitrocellulose with Bio-Rad Turbo on default setting. All blots were air-dried, rehydrated with TBS and blocked for 1 h at RT in 5% BSA in TBST (0.1%). Blots were exposed to primary antibody in 5% BSA in TBST (0.1%) overnight at 4 C. Blots were washed 3× with TBST (0.1%), 5 min per wash, at RT. Secondary antibody was added at 1:18,000: anti-rabbit-HRP (CST 7074) and/or anti-mouse-HRP (CST 7076) in 5% BSA in TBST (0.1%). Blots were incubated at RT for 1 h. Blots were washed 3 times in TBST (0.1%) for 5 min each wash at RT. All incubations and washing were done while rocking. Signal was developed with 6 ml of Femto Max ECL substrate (ThermoFisher, #34094) for 4 min and blots read on ChemiDoc. Densitometry was performed with ImageLab. Gene expression levels were determined by Taqman using the following Taqman gene expression assays (KRT80, Hs01372363_g1; PLAU, Hs01547051_g1; GUSB, Hs00939627_m1) and the Taqman RNA-to-Ct 1-Step kit (ThermoFisher Scientific). Analysis is performed using QuantStudio™ 7 Flex Real-Time PCR System (ThermoFisher Scientific). Expression levels are normalized (2−ΔCt) to the housekeeping gene, GUSB and presented relative to expression levels in vehicle-treated tumors. Immunohistochemistry was performed on formalin-fixed, paraffin embedded 5 µm thick sections using a DAKO autostainer (Agilent, Santa Clara, CA) and target antigen retrieval (Agilent). A polyclonal rabbit antibody against SMARCA2 (Sigma, Cat# HPA029981) and a rabbit monoclonal antibody against SMARCA4 (AbCam, Clone EPNCIR111A) were used as primary antibodies at a final concentration of 0.5 and 0.11 ug/ml, respectively. Secondary antibodies were a biotinylated goat anti-rabbit antibody (SMARCA4) or a biotinylated donkey anti-rabbit antibody (SMARCA2) and specifically bound antibody was detected using diaminobenzidine and an avidin-biotin-based peroxidase reaction (ABC-Peroxidase Elite, PK-6100, Vector Laboratories). Tissue sections were counterstained with Mayer’s hematoxylin. For Tumor: After the addition of 30 µL of acetonitrile per 10 mg of tumor, tumor samples were homogenized. 100 µL of tumor homogenate was pipetted out for analysis. After the addition of 50 µL of DMSO:acetonitrile 1:1 (v/v), 20 µL of 2 µg/mL propranolol in methanol:water 1:1 (v/v) as internal standard (IS), and 200 µL of chilled acetonitrile, tumor samples were vortexed and centrifuged at 3500 rpm for 30 min. 2 µL of supernatant was injected onto an AB Sciex API 4000 LC-MS/MS system coupled with a Shimadzu Prominence HPLC for analysis. For plasma: 20 µL of DMSO:acetonitrile 1:1 (v/v) and 20 µL of 2 µg/mL propranolol in methanol:water 1:1 (v/v) as internal standard is added into 20 µL of plasma sample, then 200 µL of chilled acetonitrile was added to precipitate protein. Samples were vortexed and centrifuged at 3500 rpm for 10 min. 2 µL of supernatant was injected onto an AB Sciex API 4000 LC-MS/MS system coupled with a Shimadzu Prominence HPLC for analysis. LC separation was performed on a Phenomenex Synergi Polar-RP column (4 µm, 80 Å, 2 × 50 mm) with 0.1% acetic acid 1 mM ammonium acetate in water as mobile phase A and 50 mM acetic acid in acetonitrile as mobile phase B. A gradient elution at 0. 5 mL/min started with 30% B. B component was increased linearly to 75% in 0.5 min. After holding at 75% B for 1.5 min, the column was reequilibrated with 30% B for 0.75 min. Mass spectrometric detection was performed with TurboSpray ionization in positive ion mode. A chemical library comprising 723 compounds arrayed in nine-point dose–response was screened in the absence or presence of a fixed dose of 100 nM A947. Compounds were obtained from in-house synthesis or purchased from commercial vendors. Cells were dispensed using the Multidrop™ Combi Reagent Dispenser (Thermo Scientific; Waltham, MA) into 384-well, black, clear-bottom plates (Corning, Tewksbury, MA) at seeding densities previous determined to achieve ~70–80% confluence at the final time point of the assay. Following overnight culture, compounds were dispensed using the Bravo Automated Liquid-Handling Platform (Agilent; Santa Clara, CA). Following a 5 day culture period, 25 μL CellTiter-Glo® reagent was added using a MultiFlo™ Microplate Dispenser (BioTek). Cell lysis was induced by mixing for 30 min on an orbital shaker prior to incubating plates at room temperature for 10 min to stabilize the luminescent signal. Luminescence was read by a 2104 EnVision® Multilabel Plate Reader (PerkinElmer; Waltham, MA). Data was processed using Genedata Screener®, Version 15 (Genedata; Basel, Switzerland), with a four-parameter Hill equation using compound dose−response data normalized to the median of 42 vehicle-treated wells on each plate. A “Robust Fit” strategy was also employed by Genedata Screener®, which is based on Tukey’s biweight and is resistant to outlier data. The reported absolute IC50 is the dose at which cross-run estimated inhibition is 50% relative to DMSO control wells. Data are plotted as the difference in the IC50 in the presence versus absence of A947. For matrix-based combinations, cells were seeded and assessed for viability in the same manner as described for the chemical library screen. Cells were treated with A947 (top concentration, 5 uM) in combination with AMG-176 (top concentration, 500 nM) or S63845 (top concentration, 500 nM) in a threefold dilution, 9 × 9 matrix. Drug synergy was assessed using the Bliss independence model and data is presented as excess matrix heatmaps that represent differences between the observed and predicted values determined from the Bliss model for each concentration pair. Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5

PMC9649734

Miriam Cerván-Martín,Frank Tüttelmann,Alexandra M. Lopes,Lara Bossini-Castillo,Rocío Rivera-Egea,Nicolás Garrido,Saturnino Lujan,Gema Romeu,Samuel Santos-Ribeiro,José A. Castilla,M. Carmen Gonzalvo,Ana Clavero,Vicente Maldonado,F. Javier Vicente,Sara González-Muñoz,Andrea Guzmán-Jiménez,Miguel Burgos,Rafael Jiménez,Alberto Pacheco,Cristina González,Susana Gómez,David Amorós,Jesus Aguilar,Fernando Quintana,Carlos Calhaz-Jorge,Ana Aguiar,Joaquim Nunes,Sandra Sousa,Isabel Pereira,Maria Graça Pinto,Sónia Correia,Josvany Sánchez-Curbelo,Olga López-Rodrigo,Javier Martín,Iris Pereira-Caetano,Patricia I. Marques,Filipa Carvalho,Alberto Barros,Jörg Gromoll,Lluís Bassas,Susana Seixas,João Gonçalves,Sara Larriba,Sabine Kliesch,Rogelio J. Palomino-Morales,F. David Carmona

Immune and spermatogenesis-related loci are involved in the development of extreme patterns of male infertility

10-11-2022

Genome-wide association studies,Infertility

We conducted a genome-wide association study in a large population of infertile men due to unexplained spermatogenic failure (SPGF). More than seven million genetic variants were analysed in 1,274 SPGF cases and 1,951 unaffected controls from two independent European cohorts. Two genomic regions were associated with the most severe histological pattern of SPGF, defined by Sertoli cell-only (SCO) phenotype, namely the MHC class II gene HLA-DRB1 (rs1136759, P = 1.32E-08, OR = 1.80) and an upstream locus of VRK1 (rs115054029, P = 4.24E-08, OR = 3.14), which encodes a protein kinase involved in the regulation of spermatogenesis. The SCO-associated rs1136759 allele (G) determines a serine in the position 13 of the HLA-DRβ1 molecule located in the antigen-binding pocket. Overall, our data support the notion of unexplained SPGF as a complex trait influenced by common variation in the genome, with the SCO phenotype likely representing an immune-mediated condition.

Immune and spermatogenesis-related loci are involved in the development of extreme patterns of male infertility We conducted a genome-wide association study in a large population of infertile men due to unexplained spermatogenic failure (SPGF). More than seven million genetic variants were analysed in 1,274 SPGF cases and 1,951 unaffected controls from two independent European cohorts. Two genomic regions were associated with the most severe histological pattern of SPGF, defined by Sertoli cell-only (SCO) phenotype, namely the MHC class II gene HLA-DRB1 (rs1136759, P = 1.32E-08, OR = 1.80) and an upstream locus of VRK1 (rs115054029, P = 4.24E-08, OR = 3.14), which encodes a protein kinase involved in the regulation of spermatogenesis. The SCO-associated rs1136759 allele (G) determines a serine in the position 13 of the HLA-DRβ1 molecule located in the antigen-binding pocket. Overall, our data support the notion of unexplained SPGF as a complex trait influenced by common variation in the genome, with the SCO phenotype likely representing an immune-mediated condition. According to recent estimations, the global prevalence of infertility has increased considerably during the last decades regardless of the socio-demographic index. Specifically, up to 50 million couples worldwide currently require medical assistance for reproduction, with around half of such cases being related to male factor infertility. Male infertility can be due either to an obstruction of the post-testicular tract or to non-obstructive causes. Two extreme manifestations of the latter are non-obstructive azoospermia (NOA) and severe oligozoospermia (SO), which are characterised by a severe spermatogenic failure (SPGF) leading to a reduction in the number of spermatozoa in the ejaculate (very low concentration of spermatozoa in SO and complete lack of sperm in NOA). Many SO patients eventually father a biological child following the isolation of viable seminal spermatozoa and subsequent intracytoplasmic sperm injection (ICSI). Although this simple procedure may not be applicable to azoospermic cases, there is still a chance for men suffering from this condition to benefit from the current in vitro fertilisation techniques by undergoing a testicular sperm extraction (TESE) from a testis biopsy. The overall pregnancy outcomes following TESE depend on the degree of histological abnormalities, which include hypospermatogenesis (HS, production of an extremely low number of sperm cells), maturation arrest (MA, incomplete differentiation of the germline), and Sertoli cell-only (SCO, total absence of germ cells in the seminiferous tubules). NOA patients with a histopathological diagnosis of HS have a considerably higher probability of a successful TESE when compared to those diagnosed with MA or incomplete SCO (with the latter having the poorest success rates). TESE is currently regarded as the gold standard procedure not only for sperm cell retrieval in NOA cases but also in order to obtain a conclusive histological diagnosis. However, approximately half of the TESE performed will eventually be unsuccessful in retrieving viable spermatozoa for ICSI. To that extent, having a non-invasive diagnostic test which could be able to predict sperm retrieval outcomes would be beneficial for the clinical management of NOA cases. Known genetic causes of SPGF include karyotype anomalies (e.g. Klinefelter syndrome), microdeletions of the azoospermia factor (AZF [MIM 415000]) regions in the Y-chromosome, and point mutations in master regulator genes for spermatogenesis. However, thus far, a genetic cause can only be established in about 20% of infertile men due to SPGF, being the origin of the infertility of the remaining cases defined as unexplained (idiopathic). In this regard, increasing evidence clearly suggests that common variants in the genome, such as single-nucleotide polymorphisms (SNP), may play a relevant role in the development of this form of male infertility by unbalancing the molecular network that controls the spermatogenic process. Over the past decade, genome-wide association studies (GWASs), in which hundreds of thousands to millions of genetic colourblindnessvariants across the genome are interrogated in a hypothesis-free fashion, have allowed to gain a valuable knowledge about the genetic component of many complex diseases and traits. Nevertheless, the field of SPGF research has yet to have benefited to its fullest potential from the fast progress achieved during the golden era of GWASs, likely due to the fact that most efforts have been dedicated to identifying high-penetrance rare mutations through targeted sequencing methods. In this context, only three GWASs of SPGF have been performed to date, i.e. a pilot study in a population of European descent in 2009 and two well-powered studies in Asians in 2011 and 2012. The first study did not yield consistent results due to the lack of statistical power to detect signals with a robust effect, as only 92 infertile men due to SPGF (including 52 SO and 40 NOA patients) and 80 normozoospermic controls were analysed for 370,000 SNPs. Conversely, the other two Asian GWASs of SPGF, together with an additional follow-up study from one of the research groups (in which thousands of individuals were included), identified several risk variants for NOA susceptibility at the genome-wide level of significance. The SPGF-associated loci known to date at this significance threshold map within eight genomic regions encompassing protein arginine methyltransferase 6 (PRMT6 [MIM 608274]), peroxisome biogenesis factor 10 (PEX10 [MIM 602859]), SRY-box 5 (SOX5 [MIM 604975]), major histocompatibility complex, class II, DR-alpha (HLA-DRA [MIM 142860]), butyrophilin-like protein 2 (BTNL2 [MIM 606000]), CDC42-binding protein kinase, alpha (CDC42BPA [MIM 603412]), interleukin 17A (IL17A [MIM 603149]), and actin-binding LIM protein family, member 1 (ABLIM1 [MIM 602330]). However, most of these genetic associations with NOA have not been replicated in independent studies and the histological phenotypes are yet to be analysed. Considering the above, we established an international collaborative effort with the aim to substantially improve the current knowledge on the genetic basis of SPGF by conducting a powerful GWAS in a large case-control cohort of European ancestry. Likewise, taking advantage of the high SNP coverage that the major histocompatibility region (MHC) has in the current genotyping arrays, we also decided to specifically interrogate this genomic region at the protein sequence level. In a first attempt to identify genetic polymorphisms that could be involved in the development of the different patterns of SPGF, we performed case-control comparisons between the different established study groups and the control population in the Iberian cohort. Association signals at the genome-wide level of significance were detected in two haplotype blocks including the SNPs rs186420734, associated with TESEneg (P = 2.95E−08, OR = 11.34, 95% CI = 4.80–26.76), and rs9271527, associated with SCO (P = 2.41E−08, OR = 2.38, 95% CI = 1.75–3.22) (Table 1 and Supplementary Fig. 1). According to Open Targets, the genes functionally implicated by these variants were follicle-stimulating hormone receptor (FSHR [MIM 136435]) for rs186420734 and several MHC class II genes, including HLA-DRB1 (MIM 142857) and HLA-DRA, for rs9271527. Considering the strong genetic association observed between the MHC system and the SCO phenotype in our discovery cohort, we decided to conduct a more comprehensive analysis of this genomic region by inferring multiallelic SNPs, classical HLA alleles, and polymorphic amino acid positions (Supplementary Data 1). The top SCO-associated peak was observed in the MHC class II, with the SNP rs1136759 showing the strongest signal (P = 3.04E−08, OR = 2.33, 95% CI = 1.73–3.15) (Table 1 and Supplementary Data 2). This SNP is located in the coding region of the HLA-DRB1 gene and it determines a serine in position 13 of the encoded protein (which also showed the same effect and statistical significance in the analysis), which lies in the antigen-binding pocket (Supplementary Data 2 and Fig. 1). This amino acid defines the HLA-DRB1 13 haplotype, which represented the most associated MHC classical allele with SCO in our study cohort (P = 3.86E−05, OR = 2.19, 95% CI = 1.51–3.17) (Supplementary Data 2). No additional associations with any of the SPGF patterns analysed were observed at the genome-wide significance level (Supplementary Fig. 1). In order to evaluate the consistency of our results in Iberians in an independent European population, we generated genome-wide genotyping data in a case-control cohort from Germany. This new analysis yielded no significant genetic association of the FSHR region with TESEneg (rs186420734: P = 0.98, OR = 1.02, 95% CI = 0.22–4.60) (Table 1). Consequently, the significant P-value observed in the TESEneg vs. controls comparison in the Iberian population was lost in the meta-analysis including both studies (rs186420734: PMETA = 1.37E−06, OR = 6.29, 95% CI = 2.98–13.27) (Table 1 and Supplementary Fig. 2), which showed a high heterogeneity between the ORs (Q = 6.5E−03). However, it should be noted that the lowest P-value across this genomic region in the German dataset was observed for rs28410762 (P = 2.79E−04, OR = 0.34, 95% CI = 0.19–0.61), which is located nearby the association peak in Iberians (49,399,835 and 49,429,854 in chromosome 2 for rs186420734 and rs28410762, respectively) and it is in LD with it, according to the 1KGPh3 EUR data (D’ = 1.00, r2 = 0.0027). On the other hand, a second suggestive peak of association with TESE outcome inside the FSHR gene was observed separately in each study as well as in the meta-analysis (top signal: rs77472631, PMETA = 2.95E−05, OR = 3.18, 95% CI = 1.85–5.47) (Supplementary Fig. 2). In this case, the effect size was homogenous between populations (Q = 0.96). On the contrary, the SCO-specific association signal with the MHC class II region observed in the Iberian population was replicated in the German dataset at the nominal level for this phase (rs1136759/HLA-DRβ1 Ser13: P = 2.38E−02, OR = 1.39, 95% CI = 1.05–1.86) (Table 1 and Supplementary Data 2). Although some heterogeneity in the ORs was observed between studies (Q = 0.015), consistent OR directions (towards risk) of the minor allele (G)/associated residue (Ser) were observed in both populations (ORIBERIANS = 2.33, ORGERMANS = 1.39). Therefore, the meta-analysis by the means of the inverse variance method confirmed this association at the genome-wide level of significance (PMETA [INV VAR] = 4.62E−08, OR = 1.78, 95% CI = 1.45–2.19) (Table 1). The lowest P-value in the meta-analysis amongst the classical MHC alleles was also observed for HLA-DRB1 13 (P = 8.07E−07, OR = 1.96, 95% CI = 1.50–2.56) (Supplementary Data 2). In order to carry out dependency analyses in the combined population, we decided to conduct another meta-analysis using logistic regression analysis assuming an additive model adjusted by the 10 first PCs and the country of origin. A slightly more significant association between SCO and rs1136759/HLA-DRβ1 Ser13 was observed with this method (PMETA [LOG REG] = 1.32E−08, OR = 1.80, 95% CI = 1.47–2.21) (Supplementary Data 2). As observed in the discovery phase, conditioning by the top signal also decreased substantially the statistical significance of class II suggestive signals (Fig. 2, Supplementary Data 2). Similarly, when we tested the possible influence of the polymorphic amino acid positions in SCO predisposition in the combined dataset by the means of a likelihood-ratio test, the most associated position was HLA-DRβ1 13 (P = 2.90E−07). The effect sizes of the six possible residues that can be present at this amino acid position are shown in Table 2. Consistent with the above, the statistical significance of most positions was considerably reduced when conditioning on HLA-DRβ1 13, which supported the causality of this amino acid position (Supplementary Data 3 and Supplementary Fig. 3). Taking advantage of the availability of GWAS data for the replication cohort, we aimed to identify possible additional association signals by performing a much more powerful genome-wide combined analysis using the inverse variance method (Supplementary Fig. 1). A new genetic association at the genome-wide level of significance was observed between the SCO phenotype and a group of SNPs in complete LD with rs115054029 (PMETA [INV VAR] = 4.24E−08, OR = 3.14, 95% CI = 2.09–4.74) (Table 1 and Fig. 3). In this case, the ORs were consistent between studies, with no significant heterogeneity observed (ORIBERIANS = 3.05, ORGERMANS = 3.38, Q = 0.82) (Table 1). The nearest gene to this haplotype is vaccinia-related kinase 1 (VRK1, MIM [602168]), which encodes a member of the VRK family of serine/threonine protein kinases playing a crucial role in regulating the cell cycle. Although several suggestive signals were observed, the analyses of the remaining SPGF groups did not produce any additional significant results (Supplementary Fig. 1). The exclusion criteria for participating in this study considered known causes of male infertility that can be assessed during the clinical routine. Regarding the congenital causes, those include karyotype analysis and screening for Y chromosome microdeletions. However, the presence of high-penetrance point mutations in key genes for spermatogenesis is not usually evaluated. As a consequence, it is likely that our study cohort contained some patients of SPGF whose aetiology could be explained by a single-gene mutation. Therefore, in order to evaluate the consistency of the observed SCO genetic associations, we decided to repeat the SCO analysis after removing cases with potential monogenic causes of their infertility. With that aim, we followed a validated workflow to detect the presence of rare coding variants in genes with known mutations associated with SCO, accordingly with both the “Male Infertility Genomic Consortium (IMIGC) database” and the “Infertility Disease Database (IDDB)”. This method allowed us to identify 32 carriers of rare variants located in the exons of 40 SCO-associated genes. Interestingly, despite the evident reduction in the statistical power of this new genetic association test, the analysis of our GWAS data after removing such individuals produced even more significant results for both the HLA region (rs1136759: P = 1.04E−08, OR = 1.90, 95% CI = 1.52–2.36) and the VRK1 locus (P = 3.91E−08, OR = 3.36, 95% CI = 2.18–5.18). According to the variant-to-Gene (V2G) pipeline of Open Targets (which considers evidence of functionality such as QTL experiments, chromatin interaction experiments, in silico functional predictions, and distance between the variant and each gene’s canonical transcription start site), all the SCO-associated SNPs in chromosome 14 were annotated as being functionally implicated in VRK1. To characterise the possible functional impact of this genomic region on SCO susceptibility, we identified all variants in high LD (r2 > 0.8) with the rs115054029 haplotype in the European population of the 1KGPh3 project, considering all proxies equally as candidates for exerting the pathogenic effect, as in the previous studies. Interestingly, overlap with different regulatory marks was observed for most proxies in multiple tissues (Supplementary Data 4). It should be noted that according to the ENCODE testis assays ENCFF651APG and ENCFF300WML, the proxies rs148465384 and rs17770386 (r2 = 1 and 0.97 with the lead SNP rs115054029, respectively) overlap with a protein binding site for the polymerase II, RNA, subunit A (POLR2A [MIM 180660]), and rs78543559 (r2 = 1 with the lead SNP rs115054029) is located in a CCCTC-binding factor (CTCF [MIM 604167]) site in the adult testis. Out of these three SNPs, rs17770386 showed a CADD value = 11.61, which predicts a high probability of deleteriousness. In addition, accordingly, with position weight matrix (PWM) data generated from ENCODE transcription factor binding experiments, rs76150492 (r2 = 1 with the lead rs115054029) was predicted to modify the binding site of the protein encoded by paired box gene 5 (PAX5 [MIM 167414]), which is reported to play a relevant role in spermatogenesis. The possible effect of the rs115054029 haplotypic block on the deregulation of VRK1 function is consistent with the expression data of this gene reported in the Human Protein Atlas portal, which includes data from GTEx and Single Cell Expression Atlas projects, amongst others. In this regard, this gene shows an abundant expression in the testis tissue, specifically within the seminiferous ducts (Supplementary Fig. 4). At the cellular level, spermatogonia and spermatocytes show the most enhanced mRNA expression of VRK1 amongst all cell types analysed (Supplementary Fig. 4), thus suggesting a possible role of its encoded protein in the first stages of the spermatogenic process. Functional annotation enrichment analysis is a powerful strategy to identify relevant cell types or tissues for a particular trait. Therefore, we evaluated the possible enrichment of DHS hotspots within the grey zone of the GWAS results (defined as the signals with P-values ranging from 1E−05 to 5E−08) for SPGF and the different histological subsets/TESE outcomes. No statistically significant enrichment was observed for any of the analysed subgroups either in the Iberian or German cohorts separately. However, the analysis of the summary stats for the meta-analysis showed a significant DHS hotspot enrichment in SCO. Strikingly, such enrichment was specific for blood-related samples, namely CD19+ primary cells, CD20+ cells, foetal spleen, CD19+ primary cells, and GM06990 lymphoblastoid cell line (Fig. 4). The DHS enrichments detected in the analysis of the remaining combined subgroups did not reach the statistical significance (Supplementary Figs. 5–7). Finally, we checked in our dataset the statistical significance of non-MHC loci that have been described to be associated with NOA at the genome-wide level of significance (±0.5 Mbp 3’ and 5’ of the reported SNP) in previous studies performed in populations of Asian descent. The effect size and P-value of both the reported association signals and the top signals observed in our combined GWAS accordingly with NOA and the extreme phenotype SCO for each region are summarised in Supplementary Data 5 and 6, respectively. Regional Manhattan plots of each genomic region are also available in Supplementary Figs. 8–10. Amongst the six analysed SNPs, only the rs13206743 variant, located in the IL17A genomic region at chromosome 6, showed evidence of association with NOA at the 5% significance level under the additive model (P = 2.32E−03), with an effect of the minor allele similar to that reported in the original Chinese study (OR = 1.20 in the present GWAS vs. OR = 1.35 in the study by Hu et al.). However, suggestive P-values were detected across most genomic regions (Supplementary Data 6 and Supplementary Figs. 8–10). We performed a genome-wide screening of around 7 million common variants in a large European cohort of well-characterised infertile men, comprising a total of 1274 patients diagnosed with SPGF of unexplained origin (772 NOA and 502 SO) and 1951 unaffected controls. The only available GWAS of this condition on this ancestry was published in 2009, which included a modest number of genetic variants and a small study cohort. Therefore, we consider that our study provides an important contribution to the current knowledge on the genetic basis of SPGF, since the European population used in the previous study was underpowered, and the data on Asian populations were not analysed according to specific phenotypic patterns. We were able to identify VRK1 as a potential susceptibility locus for SCO, which represents the most severe manifestation of SPGF. However, it is important to note that this association was not detected in the discovery phase but in the meta-analysis of both study cohorts. Consequently, additional replication studies in independent populations are definitively needed before establishing VRK1 as a firm SCO gene. VRK1 encodes a serine/threonine protein kinase that plays a pivotal role in the regulation of the cell cycle by phosphorylating relevant transcription factors for cell proliferation such as the tumour protein p53 (MIM 191170), histones, and different proteins involved in DNA damage response pathways. Indeed, overexpression of VRK1 has been observed in many types of tumours, as it is directly implicated in the entry of the G1 phase of the cell cycle, chromatin condensation, Golgi fragmentation, and assembly of the nuclear envelope. The human testis represents the structure with the highest expression of VRK1 amongst all the tissues analysed in the GTEx project. At the single-cell level, VRK1 expression has been restricted to actively dividing cells of the testis (mainly spermatogonia and primary spermatocytes). In this context, the association between SCO and the VRK1 region described here is consistent with previous studies on animal models of VRK1 defficiency, including Caenorhabditis elegans, Drosophila melanogaster, and Mus musculus, all three characterised by mitotic defects with resultant infertility. Regarding the latter, mice containing hypomorphic alleles of this gene showed reduced testis size with a progressive loss of cellularity within the seminiferous tubules and absence of spermatogenesis with increasing postnatal age. Interestingly, by 11 weeks of age, these Vrk1-deficient mice developed an SCO-like phenotype, with the tubules comprising only one basal layer of Sertoli cells. Therefore, it is likely that the SCO-associated genotypes in the upstream vicinity of the VRK1 locus identified in our study cohort increase SPGF risk by altering the correct regulation of this gene. Functional experiments focused on this genomic region may shed more light on this assumption. On the other hand, our results reinforce the hypothesis of a crucial involvement of the MHC class II region in SPGF predisposition leading to NOA. In this sense, studies performed in Japan at the beginning of the present century reported a strong contribution of the classical MHC alleles HLA-DRB1*1302 and DQB1*0604 to NOA risk, independently from the presence of Y-chromosome microdeletions. Later on, the two GWASs performed in Chinese populations, and the follow-up study of one of them, also highlighted this genomic region as the top associated signal with NOA across the whole genome. Additional evidence of the major involvement of the MHC class II in NOA was also generated by two recent studies, including an independent meta-analysis and a fine-mapping of this region using GWAS data, both from Han Chinese, in which the haplotype HLA-DRB1*1302 was confirmed as a molecular marker for NOA. No previous studies have specifically interrogated the MHC contribution to SPGF susceptibility under a European genetic architecture. With that aim, we inferred classical MHC alleles and polymorphic amino acid positions using an imputation method that has been thoroughly validated during the last decade using different approaches. In fact, this same imputation pipeline was recently used by Huang et al. to fine-map this genomic region using GWAS data from NOA patients of Asian descent. Interestingly, our analysis in Europeans showed a significant association of the MHC region specifically with the most severe NOA phenotype (defined by SCO) instead of with NOA as a whole. The SNP variant rs1136759*G and its encoded residue in position 13 of the HLA-DRβ1 subunit (serine), were significantly overrepresented in the SCO group compared to healthy controls in both the discovery phase and in the meta-analysis. Some heterogeneity in the effect sizes on SCO was observed between the Iberian and German populations. However, in both cases, the reference alleles (rs1136759*G and HLA-DRβ1 Ser13) showed risk ORs, and the combined analysis by logistic regression adjusted by PCs and country of origin (and, thus, controlling for possible population effects) yielded even more significant results (P = 1.32E−08, OR = 1.80) than those obtained by the inverse variance method (P = 4.62E−08, OR = 1.78). Moreover, all of the observed effects on SCO predisposition within the MHC class II region were eliminated after conditioning either on rs1136759*G/HLA-DRβ1 Ser13 in the independent variant test or on position HLA-DRβ1 13 in the omnibus test. All these pieces of evidence point clearly towards a firm association. The amino acid HLA-DRβ1 Ser13 defines the HLA-DRB1*13 classical haplotypes, which also showed a strong genetic effect on SCO in our study. Therefore, the relevant role of the HLA-DRB1 gene in NOA reported in Asians seems to be limited to the SCO phenotype in Europeans. A possible explanation for this observation could be that the NOA cohorts included in the Asian studies were enriched in SCO patients. However, the clinical characterisation of such populations was not included in the original publications and, therefore, we can only speculate at this point. We did not detect any significant genetic effect on SPGF within the MHC class I region, as reported in the Asian population studies, and a power issue could not be ruled out in this case. Moreover, since our subphenotype analyses were performed with considerably lower study cohorts, this may represent the main limitation of our study. Similarly, although most of the SPGF subtypes analysed here were correlated (that is, SPGF comprised all infertile men, NOA included all subgroups except SO, and TESEneg was composed mostly of SCO and MA individuals), we did not account for possible multiple testing effects due to the subphenotype analyses of our cohort, which may also represent a major caveat because the reported associations are close to the genome-wide significance level (P < 5E−08). In any case, the fact that the MHC class II region reached the genome-wide statistical significance when analysing our less-powered SCO group compared to the larger NOA group gives an idea of the high impact of this region on the most extreme SPGF phenotype. In this regard, the position 13 of HLA-DRβ1 associated with SCO in our study is located in the binding groove of the HLA-DR molecule, being directly involved in the molecular interactions with the presented peptide, which implies a functional impact on T cell antigen recognition, either during early thymic development or peripheral immune responses. Recent evidence also suggests that certain HLA-DRβ1 epitopes may increase the risk for autoimmune processes by favouring macrophage polarisation in an antigen-presenting-independent fashion. Strikingly, this same amino acid position also represents one of the most relevant MHC positions in different immune-related diseases, including systemic lupus erythematosus, giant cell arteritis, rheumatoid arthritis, and type I diabetes, amongst others. Indeed, there is firm evidence pointing to the immune response as a possible aetiological factor in SPGF. For example, (1) autoimmune responses against testicular structures and/or germ cells have been found to be associated with cryptorchidism (which may lead to SPGF), and patients of this condition carrying certain HLA-DRB1 haplotypes have been reported to show a higher production of anti-sperm antibodies, (2) infection and inflammation of the male genital tract is frequent in men diagnosed with male infertility, (3) acute or chronic inflammation may impair the testicular function through the inhibition of steroidogenesis and disturbance of the germ cell epithelium, (4) immune cell infiltrates associated with an exacerbated immune response have been observed in testicular biopsies from NOA patients, and (5) an expression signature comprising proinflammatory genes has been correlated with NOA. Therefore, the contribution of autoimmune processes to the extreme forms of SPGF like SCO should not be disregarded. Our data definitively support this idea and are consistent with the aetiological mechanism proposed by Gong et al., in which a chronic subclinical testicular inflammation may produce the release of novel self-peptides triggering autoimmunity through antigen-presentation to Th17 cells. In fact, active chromatin regions in immune-related cell types and tissues are enriched with suggestive genetic associations with SCO (Fig. 4). Under this pathogenic scenario, it could be possible that the presence in the genome of some MHC class II genetic variants, such as rs1136759*G that implies a serine in the position 13 of HLA-DRβ1, may increase the probability of initiating such autoimmune response by favouring the presentation of more immunogenic peptides. Finally, the association signal with unsuccessful TESE at the 5’ upstream region of FSHR detected in our discovery cohort was not replicated in the German population. In males, the follicle-stimulating hormone (FSH) is a major regulator of testis development and spermatogenesis through binding to its receptor (FSHR), which is located in the cell membrane of the Sertoli cells. This pathway is also very relevant in female fertility, as it controls folliculogenesis and drives oocyte maturation. Consequently, increasing evidence highlights these two genes as key players in the development of infertility. Although high-penetrant inactivating mutations of this signalling pathway are scarce, several SNPs in the genes encoding FSHR and the beta subunit of the ligand (FSHB) have been associated with unfavourable reproductive parameters in both female and male subjects (including SPGF cases) in a vast number of studies. In addition, some of those SNPs have been also reported to influence the gene expression of FSHR/FSHB likely by modifying transcription factor binding sites in regulatory regions. Therefore, a combined effect of both genes in male reproductive impairment has been proposed by integrating the transcriptional activity and the receptor sensitivity, which could be affected by common variations of the FSHB and FSHR genes, respectively. Consistent with the above, stratification of patients accordingly to the risk genotypes of this pathway is being considered for improving the current FSH treatments of male infertility patients, which has been shown to improve sperm parameters in SPGF men. Taking all the above into consideration, we are confident in the consistency of the GWAS peak in the FSHR region detected in our analysis in Iberians accordingly with the TESE success. It can be speculated that there might be population-specific LD patterns that may link the associated rs186420734 SNP with the causal variant/s in the Iberian and German genetic backgrounds. Under this assumption, and considering that rs186420734 is a rare variant in the healthy population, a possible different tagger in Germans could not be detected due to a power limitation. This could be also the case with the seven previously reported non-MHC NOA hits at the genome-wide significance level in Asians, from which only IL17A rs13206743 was replicated here at the nominal level. In conclusion, our results support the notion of unexplained SPGF as a complex trait influenced by common variation of the genome, with the added effect of risk genetic variants in an individual (mainly in non-coding regulatory regions) being critical for its development. Moreover, the data presented here also suggest that SPGF (or NOA) is not a single disease from a genetic point of view, but a combination of different phenotypes that have only in common a critical failure of the spermatogenic process at different points; thus underpinning the importance of defining homogeneous study groups for elucidating its genetic basis. Therefore, there is still a long way to go until we may fully characterise the molecular network that underlies SPGF. Male infertility GWAS remain lagging behind many other fields and, indeed, much larger studies focused on specific SPGF phenotypes are still needed. An integrative approach will be also helpful in this challenging endeavour, considering the key role of the non-coding polymorphisms in SPGF predisposition and the intricate haplotype architecture of the genome. Hopefully, with time and effort, the increase in the understanding of these complex processes may help to develop more efficient diagnostic and prognostic tools that could anticipate both the diagnosis and TESE outcome before the analysis of a testis biopsy, thus preventing the NOA patients with extreme phenotypes from undergoing unnecessary surgeries. Two independent case-control cohorts of European descent (including a discovery cohort from the Iberian Peninsula and a replication cohort from Germany) were analysed in this study, comprising a total of 1274 infertile men due to SPGF of unexplained origin (772 NOA and 502 SO patients) and 1951 unaffected controls. Informed written consent was signed by all participants before being enroled in the study and all DNA samples were irreversibly anonymised. The following procedures were in accordance with the tenets of the Declaration of Helsinki and received approval by the Ethics Committee “CEIM/CEI Provincial de Granada” (Andalusia, Spain) at the session held on January 26, 2021 (approval number: 1/21). Besides, each participating centre received ethical approval and complied with the requirements of their local regulatory authorities. SPGF cases were recruited in different public health centres and private fertility clinics from Spain and Portugal, and at the Centre of Reproductive Medicine and Andrology, University Hospital Münster, Germany, following comprehensive selection criteria based on the approved guidelines for the management of infertile men by the American Urological Association (AUA)/American Society for Reproductive Medicine (ASRM), the Canadian Urological Association (CUA), and the World Health Organization (WHO, 2010). These criteria include a physical examination of male patients showing evidence of clinical infertility by revision of the medical history, genetic screening (including Y-chromosome microdeletions and karyotype analysis), endocrine profile (follicle stimulating hormone, luteinizing hormone and testosterone), and semen analysis. Patients with no signs of post-testicular ejaculatory duct obstruction were analysed to establish the diagnosis of SO (<5 million spermatozoa/mL semen) or NOA (total absence of sperm in the ejaculate after two high-speed centrifugation processes in two different semen samples). Patients showing known causes of male infertility were excluded from the study. Consequently, as in other related genetic studies, only those men with a normal history of testicular development with no evidence of either testicular (such as orchitis, testicular malformations, and obstruction of vas deferens) or karyotype/chromosome abnormalities were selected. The non-obstructive primary spermatogenic impairment was subsequently confirmed in around half of our SPGF cohort by the histological analysis of a testicular biopsy from those patients that decided to undergo assisted reproduction treatments involving TESE (including both conventional TESE and micro-TESE). The pathological anatomy results from the biopsy were used to classify the SPGF patients into different subgroups according to the observed histological phenotypes, including HS (extremely low cell counts of the germline but with all stages of spermatogenesis/spermiogenesis observable in few testicular locations), MA (early maturation arrest either at spermatogonia or at primary spermatocyte stages of more than 90% of the germline), and SCO (total absence of germ cells in all seminiferous tubules). Furthermore, two additional subgroups of NOA were established based on the TESE outcome, as follows: TESEneg (if no viable sperm cell was retrieved from the biopsy) and TESEpos (including NOA patients with a successful sperm retrieval). SO patients were not considered for this classification because the TESE success rate associated with this form of infertility is close to 100%. All the available information about the main clinical features of our study cohort is shown in Supplementary Table 1. Genomic DNA samples obtained from peripheral blood mononuclear cells of every participant were genotyped at the genome-wide level using the Infinium™ Global Screening Array-24 v3.0 (GSA) in an iScan System (Illumina, Inc), following the manufacturer’s protocol. This is an advanced high-throughput genotyping platform that allows the genotyping of more than 700,000 carefully selected genetic variants, including tag polymorphisms, relevant markers for clinical research, and variants for quality control (such as ancestry informative markers). Thus, this system delivers a high genomic coverage ideal for imputation methods. The genotyping of the Iberian samples was conducted in the Human Genotyping Unit of the National Genotyping Centre (CEGEN) at the Spanish National Cancer Research Centre (Madrid, Spain), whereas that of the German samples was carried out in the Genomics Unit of the LIFE & BRAIN GmbH Biomedical & Scientific Technology Platform (Bonn, Germany). In both cases, the genotype calling was performed with the Genotyping Module (v.2.0) implemented in the GenomeStudio software (Illumina, Inc), and assigning the chromosome positions according to the Genome Reference Consortium Human Build 38 (GRCh38). The genotype data was subject to stringent quality control (QC) measures using R and PLINK v.1.9. First, we removed all the genetic variants with a cluster separation < 0.4 and filtered out INDELs and rare variants with minor allele frequencies (MAF) < 0.01. Moreover, SNPs with call rates < 0.98 and those whose genotype distributions deviated from Hardy–Weinberg equilibrium (HWE) in controls (P < 0.001) were also excluded from further analyses. Regarding the QC of the recruited individuals, samples with <95% of successfully called SNPs and one subject per pair of first-degree relatives (identity by descent >0.4) were excluded. In addition, principal component (PC) analyses were conducted with a set of 2921 ancestry-informative markers included in the GSA chip, in order to detect and remove population outliers (>4 standard deviations from the cluster centroids of each population) using PLINK, R and the gcta64 software. Supplementary Fig. 11 showed the two first PCs plotted against each other for the samples that remained after the removal of population outliers. To maximise the genetic coverage of our data sets, we conducted SNP genotype imputation for chromosomes 1-22 and X on the genome build GRCh38, using the haplotype data of the ‘NHLBI Trans-OMICs for Precision Medicine’ (TOPMed) programme (freeze 5) as reference panel, in the Next-Generation Genotype Imputation Service of the TOPMed Imputation Server. Eagle v.2.4. for haplotype phasing and minimac4 algorithms were applied for genotype imputation. Due to the lack of a Y chromosome reference panel, we could not impute additional Y chromosome variants; therefore, only the directly genotyped SNPs were analysed. Moreover, considering previously reported evidence regarding the possible role of the major histocompatibility (MHC) system in NOA predisposition, we decided to carry out a more comprehensive interrogation of this genomic region in our study population. With that aim, we extracted the extended MHC region (from 29 to 34 Mbp in chromosome 6) from the non-imputed data and used the SNP2HLA method, with a reference panel collected by the Type 1 Diabetes Genetics Consortium comprising 5,225 individuals of European origin, to impute SNPs, classical MHC alleles at two- and four-digits, and polymorphic amino acid positions. To ensure the high quality of the imputed data, only SNPs with a very reliable imputation quality metric (namely Rsq > 0.9 for minimac4 or posterior probability > 0.9 for SNP2HLA) were analysed (genotypes that did not reach the selected cut-off value were set to missing). Furthermore, the imputed data underwent also rigorous QC filters using PLINK and R, including the removal of singletons, rare variants (MAF < 0.01), and polymorphisms with call rates lower than 98%. SNPs whose genotype frequencies showed evidence of deviation from HWE (P < 0.001) were also excluded from further analyses. Following the QC procedures, the final case-control data sets comprised 627 SPGF patients and 1027 unaffected controls from the Iberian Peninsula and 647 SPGF patients and 924 control participants from Germany. A total of 7,371,432 SNPs were analysed in the Iberian cohort and 7,536,533 SNPs in the German cohort. Regarding the comprehensive interrogation of the MHC region, the imputed data included 7258 SNPs, 424 classical alleles (at 2- and 4-digit coverage), and 1276 polymorphic amino acid variants from the human leucocyte antigen (HLA) genes HLA-A (MIM 142800), HLA-B (MIM 142830), HLA-C (MIM 142840), HLA-DPA1 (MIM 142880), HLA-DPB1 (MIM 142858), HLA-DQA1 (MIM 146880), HLA-DQB1 (MIM 604305), and HLA-DRB1 (MIM 142857). To determine the minimum effect sizes that could be detected in this study based on experimental design, power analysis estimations were calculated with the online tool of the Genetic Association Study (GAS) Power Calculator, which implements the methods described in Skol et al. assuming additive genetic effects (Supplementary Table 2). All the case-control comparisons were performed with PLINK and R. In the first step, we tested for association using the imputed data of the discovery cohort (Iberian). Specifically, we compared all case groups (SPGF, NOA, SO, MA, HS, and TESEneg) against the group of unaffected controls using logistic regression on the best-guess genotypes (Rsq > 0.9), adding the 10 first PCs and the country of origin (Spain or Portugal) as covariates and assuming additive effects. If a subtype-specific genetic association was detected, cases showing such clinical phenotype/TESE outcome were also compared against those not showing it, to check whether the association was maintained after eliminating SPGF as a possible confounding variable. With regards to the analysis of the MHC region, we tested SNPs, classical HLA alleles, and all possible combinations of amino acid residues per position by logistic regression as described above. For the positional model analysis, we established a null generalised linear model for each position including the 10 first PCs and the country of origin as covariates, which was compared on the basis of a χ2-based estimate to an alternative model including such covariates and all the possible residues at those positions (considered as conditioning factors). Besides, considering the extensive linkage disequilibrium (LD) of this genomic region, dependency analyses were performed to identify independent genetic effects by step-wise logistic regression with conditioning by the top association signals (together with the 10 first PCs and the country of origin). After evaluating the relevance of the results of the discovery phase, we decided to analyse an independent replication cohort from Germany following the same workflow described above for the discovery cohort. Finally, since whole-genome genotype data were generated for both the discovery and the replication cohorts, we decided to conduct a combined analysis of both studies by the means of the inverse variance weighted meta-analysis under a fixed effects model; thus, increasing the statistical power to detect additional association signals. In this case, the possible heterogeneity of the effect sizes between the two analysed studies was evaluated using both I2 and Cochran’s Q tests. Additionally, we also performed a combined analysis of the MHC region (including both the discovery and the replication cohorts) by logistic regression on the best-guess genotypes (>0.9 probability) assuming an additive model with the 10 first PCs and the country of origin (Spain, Portugal, and Germany) as covariates, in order to allow an adequate evaluation of the dependency effects in the pooled dataset. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for all the statistical analyses. The statistical significance was set at the genome-wide level (P < 5E−08) in the meta-analysis, provided that the P-value for each study separately was below 0.05 and the directionality of effect presented by the ORs was consistent between studies. The Manhattan plots were generated using an in-house R script, and the zooms of the associated regions were created with LocusZoom.js. The 3D models of the HLA molecules were performed with the UCSF Chimera software. The online tools provided by the GTEx and LDlink portals were used for figure generation together with custom R scripts. In order to shed light on the possible pathogenic mechanisms involved in SPGF susceptibility, we decided to enrich our results with publicly available functional annotation data by using different bioinformatics approaches. With that aim, we first used LDLink to identify all proxies of the associated variants outside the MHC region (r2 > 0.8) in the European population of the 1KGPh3. Then, we queried different databases and online tools to extract all the relevant information that could help us to elucidate the potential molecular and cellular mechanisms underlying the observed associations, including RegulomeDB, Haploreg v.4.1., Open Targets Genetics, SNPnexus, GTEx, Human Protein Atlas, and ENCODE, which integrate the datasets included in Ensembl, SIFT, Polyphen, CpG, Vista enhancers, miRbase, TarBase, TargetScan, miRNA Registry, snoRNA-LBME-DB, Roadmap, Ensembl regulatory build, CADD, DeepSEA, EIGEN, FATHMM, fitCons, FunSeq2 GWAVA, and REMM. The different predictive scores for functionality are described in Supplementary Tables 3 and 4. Furthermore, the possible overlap of the associated variants and their proxies with regulatory regions in the testicular tissue was assessed by analysing the testis-specific assays in ENCODE: DNase-seq hypersensitivity sites (ENCFF323BCL, ENCFF608KRZ); CTCF (ENCODE sample references: ENCFF300WML, ENCFF559LDF, ENCFF644JKD, ENCFF767LMP, ENCFF788RFY, ENCFF855EVV) and POLR2A (ENCFF535DHF, ENCFF651APG) protein ChIP-seqs; H3K4me3 (ENCFF286DAB, ENCFF509DBT), H3K4me1 (ENCFF316MJM), H3K27ac (ENCFF610XSK, ENCFF819NRA), H3K9me3 (ENCFF711LHL), and H3K27me3 (ENCFF881OHS) histone modification ChIP-seqs. Finally, we also assessed the enrichment of the suggestive association signals (P < 1E−05) observed for the analysed phenotypes and the DNase I-hypersensitive sites (DHS hotspots) identified by ENCODE and the Roadmap Epigenomics project for all available cell types using GARFIELD. In brief, GARFIELD performs a greedy LD-prunning, LD-based tagging, and functional annotation of the genetic variants included in the GWAS summary statistics. Functional annotation enrichment is quantified by the means of generalised linear models controlling for distance to the nearest TSS and number of LD proxies, and establishing different genome-wide significance thresholds. According to Bonferroni’s multiple testing correction based on the number of independent tests carried out, the significant threshold for enrichment in DHS hotspots was established at P-value < 2.6E−04, as recommended by Iotchkova et al.. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Peer Review File Supplementary Information Description of Additional Supplementary Data Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Supplementary Data 7 Reporting Summary

PMC9649744

Youwei Zheng,Xinchao Li,Lirun Kuang,Yong Wang

New insights into the characteristics of DRAK2 and its role in apoptosis: From molecular mechanisms to clinically applied potential 10.3389/fphar.2022.1014508

28-10-2022

DRAK2,DAPK,STK17B,apoptosis,targeted therapy

As a member of the death-associated protein kinase (DAPK) family, DAP kinase-associated apoptosis-inducing kinase 2 (DRAK2) performs apoptosis-related functions. Compelling evidence suggests that DRAK2 is involved in regulating the activation of T lymphocytes as well as pancreatic β-cell apoptosis in type I diabetes. In addition, DRAK2 has been shown to be involved in the development of related tumor and non-tumor diseases through a variety of mechanisms, including exacerbation of alcoholic fatty liver disease (NAFLD) through SRSF6-associated RNA selective splicing mechanism, regulation of chronic lymphocytic leukemia and acute myeloid leukemia, and progression of colorectal cancer. This review focuses on the structure, function, and upstream pathways of DRAK2 and discusses the potential and challenges associated with the clinical application of DRAK2-based small-molecule inhibitors, with the aim of advancing DRAK2 research.

New insights into the characteristics of DRAK2 and its role in apoptosis: From molecular mechanisms to clinically applied potential 10.3389/fphar.2022.1014508 As a member of the death-associated protein kinase (DAPK) family, DAP kinase-associated apoptosis-inducing kinase 2 (DRAK2) performs apoptosis-related functions. Compelling evidence suggests that DRAK2 is involved in regulating the activation of T lymphocytes as well as pancreatic β-cell apoptosis in type I diabetes. In addition, DRAK2 has been shown to be involved in the development of related tumor and non-tumor diseases through a variety of mechanisms, including exacerbation of alcoholic fatty liver disease (NAFLD) through SRSF6-associated RNA selective splicing mechanism, regulation of chronic lymphocytic leukemia and acute myeloid leukemia, and progression of colorectal cancer. This review focuses on the structure, function, and upstream pathways of DRAK2 and discusses the potential and challenges associated with the clinical application of DRAK2-based small-molecule inhibitors, with the aim of advancing DRAK2 research. Death-associated apoptosis-inducing protein kinase 2 (DRAK2) (Gozuacik and Kimchi, 2006), also known as STK17B, is a serine/threonine protein kinase and a member of the death-associated protein kinase (DAPK) family. As the name implies, DRAK2 is primarily associated with apoptosis, particularly in pancreatic beta cells (Mao et al., 2008; Wang et al., 2017). Several studies have shown that DRAK2 plays an important role in the development of type 1 diabetes (Edwards et al., 2015). The mRNA levels and protein levels of DRAK2 in pancreatic β-cells are rapidly increased in response to inflammatory lymphokine stimulation, ultimately leading to apoptosis of islet β-cells. Another relatively important and currently recognized function of DRAK2 is its involvement in the activation of lymphoid T-cells (Fracchia et al., 2013). One study showed that DRAK2 is a negative regulator of T-cell receptor (TCR) signaling and sets the threshold for T-cell activation through this pathway (Friedrich et al., 2005). Although DRAK2 is a member of the DAPK family, it has been much less studied than other members of the family, such as DAPK1, DAPK2, and DAPK3 (Dai et al., 2016). While DRAK2 shows similar pro-apoptotic functions to the other members (Rennier and Ji, 2013; Benderska and Schneider-Stock, 2014), it is structurally different from DAPK1/2/3 and only shows structural similarity to DRAK1 (Farag and Roh, 2019). DRAK2 began to receive research attention in the 1990s, with the earliest studies reporting its structure and the relationship with lymphoid T-cell activation. Subsequent studies gradually showed that DRAK2 is involved in the development of many cancers, such as acute myeloid leukemia (Ye et al., 2013) and colorectal cancer (CRC). In addition, researchers have also identified an important role of DRAK2 in organ transplant rejection (Gao et al., 2014), which is likely to be an important target for avoiding immune rejection of transplanted organs in the future. More recently, the research focus on DRAK2 has been renewed with studies demonstrating an important link between DRAK2 and non-alcoholic fatty liver disease (NAFLD) (Li et al., 2021). DRAK2 was found to be expressed to varying degrees in patients and mice with different severities of fatty liver, suggesting that it plays a crucial role in metabolic disease. Current research on DRAK2 is relatively limited, and although DRAK2 has been shown to be involved in many physiological or pathological activities, a large proportion of the underlying specific mechanisms have not been identified. For example, as mentioned above, DRAK2 affects NAFLD, and one study suggested that it may be involved in the development of NAFLD by affecting the splicing mechanism of RNA (Li et al., 2021). Although this study focused on the changes downstream of DRAK2, it did not mention how DRAK2 acts in the early stages of the disease. Similarly, for example, DRAK2 overexpression was found to promote apoptosis in islet β-cells treated with free fatty acid (FFA), but the mechanism by which FFA caused DRAK2 overexpression was not specified. Phorbol myristate acetate (PMA) has been used by many researchers to induce the expression of DRAK2 in various cells (Kuwahara et al., 2008). PMA can also induce the translocation of DRAK2 from the NIH3T3 cytoplasm to the nucleus, but the mechanism by which PMA induces DRAK2 expression has not been explained. Thus, research on DRAK2 has remained superficial and the mechanisms involved have not been clearly investigated. Nevertheless, research on this topic is ongoing, and in the absence of a detailed summary of the literature on this topic, the accumulating literature has become increasingly confusing. To address these issues, this article reviews the structure, function, upstream pathways, and small-molecule inhibitors of DRAK2 on the basis of previous publications. We hope that the manuscript will provide readers a quick overview of DRAK2. Protein kinases are enzymes that catalyze protein phosphorylation. Currently, there are two known types of protein kinases: protein tyrosine kinase (PTK) and serine/threonine protein kinase (STK) (Farag and Roh, 2019). The DAPK family is one of the important STK families and includes five members: DAPK1, DAPK2 (also known as DAPK-related protein 1 [DRP-1]) (Shiloh et al., 2014; Geering, 2015; Yan et al., 2019), DAPK3 (also known as DAP-like kinase [DLK] or zipper-interacting protein kinase [ZIPK]), and DRAK1 and DRAK2 (also known as DARK-related kinase 1 [DRK-1] and DARK-related kinase [DRK-2], respectively). The five members are highly homologous at the N-terminal end of the amino acid sequence and differ at the C-terminal end, which is related to their functions. Among them, DAPK3 has about 83% homology with DAP kinase within the kinase structural domain (Heyerdahl et al., 2010; Cao et al., 2013), while the two most closely related members of the DAPK family, DRAK1 and DRAK2, have 50% homology with DAP kinase within the kinase structural domain. In addition, the DAPK family also belongs to the calmodulin (CaM)-regulated kinase superfamily, but not all five family members contain CaM-regulated regions, with DAPK1 and DAPK2 containing calcium-regulated regions and DAPK3, DRAK1, and DRAK2 not containing these regions. DAPK1 is the largest member of the family, with a molecular weight of 160 kDa and 1430 amino acids (Nair et al., 2013). In addition to containing a catalytic kinase domain (CD) at the N-terminal (Figure 1), the autoregulatory domain (ARD) next to the CD can exert kinase activity by binding Ca2+/CaM through Ser308 in DAPK1 (Temmerman et al., 2013). When Ser308 is phosphorylated, the ARD cannot bind to CaM, leading to inactivation of DAPK1. This process suggests that calcium-activated CaM inhibits catalytic activity by binding to its own regulatory/CaM-binding fragment and the catalytic cleft in this region (Kuczera and Kursula, 2012; Tavares et al., 2017). Although the activation of DAPK1 relies mainly on CaM, even in the absence of CaM, dephosphorylation of DAPK1 by Ser308 can result in low levels of catalytic activity (Simon et al., 2016; Horvath et al., 2021). The structure of DAPK1 also shows eight anchor protein repeats at the right end of the ARD, the ROCO structural domain, the cytoskeleton-binding region, the death structural domain, and a serine-rich C-tail. In addition to containing a catalytic domain 80% homologous to DAPK1, DAPK2 also contains an ARD that binds Ca2+/CaM, but when its internal Ser308 is phosphorylated, DAPK2 loses its kinase activity. It contains an additional dimerization element (Wu et al., 2020; Du and Kong, 2021) at its right end (Figure 1), and similar to the inactivation of DAPK1 by Ser308 phosphorylation, Ser318 dephosphorylation can promote dimerization and increase DAPK2 activity (Shoval et al., 2011). In addition, DAPK1 and DAPK2 show similarities in their regulation of kinase activity within the catalytic structural domain and both show the following two aspects: 1) upon Ser308 dephosphorylation, calcium-activated CaM binds to the self-regulated/CaM-bound fragment pulling the structural domain out of the catalytic cleft; 2) simultaneously, CaM binding promotes the dephosphorylation of Ser308 (Shoval et al., 2011; Simon et al., 2016), and the dephosphorylated Ser308 increases the DAPK2 activity by promoting dimerization like the dephosphorylated Ser318; therefore, the binding of CaM and the dephosphorylation of Ser308 are reciprocal. In addition, a number of studies have confirmed this role of Ser308 in DAPK1 and DAPK2, replacing all Ser308 in both kinases with Ala or deleting the CaM-binding region in both kinases to produce constitutively active kinases that exhibit more potent killing and non-Ca-dependent catalytic activity (Bialik and Kimchi, 2006; Aziz et al., 2009). In addition to an N-terminal catalytic domain that is 83% homologous to DAPK1, DAPK3 also has a nuclear localization signal (NLS) sequence and a leucine zipper domain at the C-terminus (427–441). Although it does not have the CaM regulatory region of DAPK1 and DAPK2, it has three serine/threonine phosphorylation sites within the CD, Thr180, Thr225, and Thr265 (Figure 1), which may be involved in the regulation of DAPK3 activity (Graves et al., 2005). The remaining two members of the DAPK family are less complex in structure than the first three family members, with DRAK1 (also known as STK17A) showing only an N-(Boosen et al., 2009) terminal catalytic domain and a C-terminal regulatory domain containing seven exons and seven introns. DRAK2, also known as STK17B, is mainly expressed in developing and mature lymphocytes (Ramos et al., 2008), and to a lesser extent in the liver and pancreas, in addition to the thymus. DRAK2 contains 372 amino acids, consists of an N-terminal catalytic domain and a C-terminal domain responsible for regulating kinase activity (Boosen et al., 2009), possesses several relatively significant phosphorylation sites such as Ser10, Ser12, Ser351 (Fracchia et al., 2013), and is autophosphorylated. It also contains an NLS at the C-terminus (Figure 1), and Ser350 is a phosphorylation site for protein kinase C (PKC)-g, whose phosphorylation can affect the nuclear localization of DRAK2 (Kuwahara et al., 2008). This enzyme can induce apoptosis and regulate cell differentiation, and overexpression of DRAK2 in cell lines can induce apoptosis (Guo et al., 2009; Manivannan et al., 2019). Each member of the DAPK family is associated with apoptosis by name and can cause some degree of damage when overexpressed in cells, such as rounding of the cell shape, blistering of membrane structures, detachment from the extracellular matrix, and formation of autophagic vesicles (Chen et al., 2006; Bialik and Kimchi, 2010; Levin-Salomon et al., 2014). The most important features of apoptosis, namely, cell rounding and membrane blistering, are caused by phosphorylation of Ser19 within DAPK, which further allows kinases to act on the myosin II light chain (MLC) within the cytoskeleton, leading to cell spreading, cell motility, cytoplasmic disintegration, and cell death (Dos Santos et al., 2018; Markwardt et al., 2018). DAPK is required for multiple death signals to induce cell death (Bialik and Kimchi, 2006). For example, DAPK1 is associated with induction of autophagy during endoplasmic reticulum (ER) stress (Gozuacik et al., 2008; Xiang et al., 2020). In addition, DAPK1 is a tumor suppressor that inhibits the transformation of normal cells into abnormal cells in the early stages of tumorigenesis (Martoriati et al., 2005), and inhibits tumor metastasis through its effect on the cytoskeleton (Ivanovska et al., 2014). For example, in tumor cell lines lacking p53, the pro-apoptotic activity of DAPK1 gradually disappears (Silginer et al., 2014; Vitillo et al., 2016); thus, DAPKs generally work together to prevent tumors during tumorigenesis by both promoting apoptosis and inhibiting the migration of tumor cells. In addition to the abovementioned aspects, DAPK1 is present in high levels in the brain, and some studies have found an association between DAPK1 and neuronal cell death (Fujita and Yamashita, 2014; Shi et al., 2022; Zhang et al., 2022), with deletion of the DAPK1 gene preventing ischemic neuronal death, and the use of DAPK1 inhibitors producing the same results. Activation of DAPK1 induces Ca2+ entry into cells via functional NMDA receptors (NR2B subunits) in CNS neurons, which in turn leads to cell death (Tu et al., 2010). Because DAPK2 belongs to the same family as DAPK1, it functions similarly to DAPK1 and can also inhibit tumorigenesis and migration by the two methods mentioned above. For example, DAPK2 levels are reduced in Hodgkin’s lymphoma that initiates methylation, and its introduction into cells can promote apoptosis and inhibit tumor growth (Xiang et al., 2020). In addition to these functions, DAPK2 has been associated with autophagy, oxidative stress in cancer cells, myeloid differentiation, and erythropoiesis (Rizzi et al., 2007; Fang et al., 2008; Ber et al., 2015; Schlegel et al., 2015). DAPK3 is currently thought to have a primary function in regulating apoptosis and smooth muscle contraction (Usui et al., 2014a; Usui et al., 2014b; Komatsu and Ikebe, 2014), and also shows tumor-suppressive, apoptosis-promoting, and autophagic functions. Interestingly, DAPK3 can also regulate myosin by inhibiting MLC phosphatase or regulating the light chain (LC20), which in turn enhances responsiveness to Ca2+ and induces smooth muscle contraction (Wang L. et al., 2021). In addition, DAPK3 is also associated with cardiovascular diseases (Chang et al., 2010; Carlson et al., 2018; Zhang et al., 2019), and it can regulate myocardial contraction by phosphorylating the MLC at Ser15. In addition to its role in promoting apoptosis, DRAK1 has been most studied for its association with cervical cancer (Manivannan et al., 2019; Chen and MacDonald, 2022). DRAK1 inhibits the growth and metastasis of advanced cervical cancer cells through two pathways: interfering with the homo-oligomerization of tumor necrosis factor (TNF) receptor-associated factor 6 (TRAF6) and specifically reducing the stability of TRAF6 protein through an autophagy-mediated degradation pathway (Park et al., 2020). At the same time, DRAK1 has also been found to function as a novel negative regulator of the transforming growth factor-β (TGF-β) tumor suppressor signaling pathway. DRAK1 can interrupt the formation of the Smad3/Smad4 complex by binding to Smad3, a process that inhibits TGF-β tumor suppressor signaling in head and neck squamous cell carcinoma (HNSCC) and increases tumor activity (Park et al., 2015). In addition, DRAK1 has also been associated with the development of SLE (da Silva Fonseca et al., 2013) and testicular cancer (Mao et al., 2011). The functional aspects of DRAK2 have been evaluated in some studies, but most of them focus on T-cell activation, islet cell function, etc. For example, DRAK2 is a negative regulator of TCR signaling and sets the threshold for T-cell activation through this pathway (Ramos et al., 2008),. Moreover, inhibition of DRAK2 expression can prevent islet β-cell apoptosis (Wang et al., 2017). DRAK2 has been recently suggested to be involved in the development of NAFLD by inhibiting the phosphorylation of serine-arginine-containing splicing factor 6 (SRSF6) by SRSF protein kinase 1 (SRPK1) through binding to SRSF6 (Li et al., 2021). We found that this enzyme plays an important role in apoptosis, immunity, and metabolism. This article compiles and summarizes the existing literature on DRAK2. The upstream pathways and small-molecule inhibitors of DRAK2 are described separately below. MYB is a segment of the proto-oncogene that encodes a protein with three HTH DNA-binding structural domains. This structural domain has a transcriptional regulatory role. MYB is an important transcription factor that is inextricably linked to hematopoietic function (Greig et al., 2008; Shah et al., 2021). MYB gene family members include MYB, MYBL1, and MYBL2 (Ciciro and Sala, 2021). MYB genes contain an N-terminal DNA-binding domain (DBD), a C-terminal negative regulatory domain (NRD) and a central trans-activation domain (TAD) (Figure 2) (Ko et al., 2008); v-myb is derived from mutations in the MYB gene and can cause severe acute myeloid leukemia in vertebrates (Weng et al., 2018; Wang et al., 2020; Smeenk et al., 2021). DRAK2 plays an important role in v-myb-mediated acute myeloid leukemia (Ye et al., 2013), and CHIP indicates that MYB binds to a conserved element upstream of the DRAK2 transcription start site. Knockdown of the MYB gene using MYB shRNA in U937 cells can induce apoptosis in U937 cells by stimulating DRAK2 expression and downstream caspase-9 activity. In addition, in a clinical study of 22 AML patients (Lee et al., 2006), expression profiling of bone marrow samples revealed MYB upregulation and DRAK2 downregulation in seven patients. In other related studies, DRAK2 expression was also found to be twice as high after inhibition of the MYB gene using MYB shRNA in an AML mouse model (Zuber et al., 2011). In conclusion, all of the above studies demonstrated that DRAK2 is a downstream target gene of MYB. Protein kinase D (PKD) belongs to the serine/threonine kinase family and the Ca++-calmodulin-dependent protein kinase (CaMK) family. PKD contains three isoforms: PKD1, PKD2, and PKD3, with PKD1 being called PKDμ in the earliest study in 1994 (Bush and McKinsey, 2009) and PKD3 being called PKDv (Papazyan et al., 2008). All three isoforms perform important roles in cellular functions, including intracellular vesicle transport and maintenance of Golgi function. PKD has been shown to be involved in the activation of T-cells and its main pathway is through the regulation of DRAK2 activity (Newton et al., 2011). T-cell activation is mainly associated with Ca2+ influx; however, when PKD is knocked out, it strongly blocks the TCR signaling pathway and prevents the activation of DRAK2. Moreover, activation of DRAK2 is inhibited by the PKD small-molecule inhibitor Gö6976. In one specific mechanism, when the TCR is stimulated by certain signals, it produces IP3 through activation of phospholipase Cγ (PLCγ), which acts on the IP3 receptor on the ER, causing release of Ca ions from the ER. If the stromal sympathetic molecule 1 (STIM1) on the ER senses Ca ion insufficiency, it stimulates cell surface calcium release to activate calcium channel regulatory molecule (ORAI1) into the CRAC channel, which causes the inward flow of extracellular Ca ions. Ca ions can stimulate mitochondria to produce large amounts of reactive oxygen species (ROS) under certain circumstances (Schieber and Chandel, 2014; Zorov et al., 2014; Hamilton et al., 2020), and the mitochondrial production of ROS activates DRAK2 by stimulating PKD. DRAK2 acts on STIM1 to regulate the concentration of Ca ions, which in turn regulates T-cell differentiation (Figure 3). In addition, the study also confirmed that DRAK2 is a direct substrate of PKD (Newton et al., 2011). The team found reduced basal autophosphorylation levels of DRAK2 and greatly reduced PMA-induced DRAK2 autophosphorylation when using PKD mutants (KD-PKD1; PKD1-K612W) co-expressed with DRAK2. Moreover, infection of JurkaT-cells with PKD2-expressing shRNA revealed that the DRAK2 activation induced by anti-CD3 cross-linking or toxic carotenoids was greatly reduced by PKD2 knockdown. In conclusion, PKD has been shown to be a substrate for DRAK2 by a number of different methods. Cyclooxygenase 2 (COX-2) is an inducible enzyme regulated by many cytokines, such as interleukin (IL)-1β, IL-6, or TNFα (Chan et al., 2019), and is primarily responsible for prostaglandin (PG) biosynthesis (Karpisheh et al., 2019; Frejborg et al., 2020; Ochiai et al., 2022). COX-2 expression in normal cells is rare and almost negligible (Gurram et al., 2018), and its frequent appearance in cancer is considered a marker for cancer detection (Hashemi Goradel et al., 2019). Some studies have shown that COX-2 expression is increased in most CRC patients, suggesting that COX-2 is closely related to CRC (Wang and Dubois, 2010; Su et al., 2016; Dagallier et al., 2021). Interestingly, when the COX-2 selective inhibitor rofecoxib was administered to CRC patients, the DRAK2 level in tumor cells increased 2.5-fold. In addition, inhibition of COX-2 activity in HCA7 cells also enhanced the expression of DRAK2 (Doherty et al., 2009). The COX-2 transcriptional level in CRC patients was found to be 2.4 times higher than normal, while the opposite was true for DRAK2 expression. Other studies have also highlighted the involvement of COX-2 in the negative regulation of DRAK2. Overexpression of the pro-apoptotic gene DRAK2 was also present in studies in COX-2−/− mice treated with adriamycin Dox (Neilan et al., 2006). Interestingly, COX-2 has also been shown to regulate T-cell activation (Li Q. et al., 2022). Therefore, in combination with the abovementioned involvement of DRAK2 in T-cell activation, we speculate that COX-2 and DRAK2 show some similarities or crossover in the pathways that regulate T-cell activation. TGF-β is a multicellular functional factor involved in the regulation of multiple intracellular activities with three ligands: TGF-β1, TGF-β2, and TGF-β3 (Rubtsov and Rudensky, 2007). In the classical TGF-β signaling, TGF-β binds to the type II TGF-β receptor (TβRII) on the cell membrane, followed by recruitment and phosphorylation of the type I TGF-β receptor (TβRⅠ). The phosphorylated TβRⅠ then acts through the Smad protein, and the heterodimeric complex composed of TβRⅡ and phosphorylated TβRⅠ in turn phosphorylates Smad2 and Smad3 downstream, while the phosphorylated Smad2/3 binds to Smad4 (Figure 4) and finally enters the nucleus to act. However, some studies have confirmed that DRAK2 can act as an antagonist of TGF-β signaling induced by TGF-β1 (Yang et al., 2012). In this process, DRAK2 can bind specifically to TβRⅠ, thus blocking the activation of Smad2/3 by phosphorylated TβRⅠ, while the unphosphorylated Smad2/3 cannot bind to Smad4 and thus cannot enter the nucleus for regulation. Another study published in The Lancet found that DRAK2 was abundantly expressed in breast cancer cells and that knockdown of DRAK2 in breast cancer cells enhanced TGF-β signaling (Wang et al., 2005). Therefore, we speculate that in breast cancer cells, DRAK2 is likely to promote tumor growth by blocking TGF-β signaling. However, the ability of DRAK2 to block TGF-β1-induced TGF-β signaling has been questioned (Harris and McGargill, 2015). In that study, T-cells were isolated from wild-type and Drak2−/− mice, and TGF-β signaling was not enhanced in Drak2−/− T-cells after an exogenous increase in TGF-β1 and did not differ significantly from the wild-type T-cells. There are several possible reasons for this phenomenon: 1) The study in which DRAK2 blocked TGF-β1-induced TGF-β signaling was mainly performed in cancer cells rather than in the normal physiological state, while the latter study was performed in normal T-cells. Thus, the differences in results may be attributable to differences in the function or mode of action of DRAK2 in the normal physiological state and in the cancer setting. 2) Although enhanced TGF-β signaling was not found in Drak2−/−T-cells, the level of Smad7, a negative regulator of TGF-β signaling, was higher than that in wild-type T-cells (Itoh and ten Dijke, 2007). This phenomenon may be due to the fact that Drak2 −/− T-cells compensate for the absence of DRAK2 through other alternative pathways, which in turn affects TGF-β signaling. The mechanisms involved in the functioning of DRAK2 in TGF-β signaling in normal versus abnormal environments, in vivo and in vitro, are unclear and need to be further investigated. Interferon-gamma (IFN-γ), TNF-α, and IL-1β are closely associated with the development of type I diabetes (Arif et al., 2021; Janssen et al., 2021; Sims et al., 2021). One of the main reasons for the development of type I diabetes is that cytokines such as IFN-γ, TNF-α, and IL-1β produced by immune cells break down the islet beta cells. Curiously, one study found that only prolonged treatment of cells with IL-1β + IFN-γ and/or TNF-α resulted in islet β-cell dysfunction and eventual apoptosis, but individual cytokines did not produce similar results (Kim et al., 2007). However, one study found that the effects of IFN-γ, TNF-α, and IL-1β on pancreatic islet beta cells were associated with the activation of DRAK2 (Mao et al., 2009). When pancreatic β-cells were treated with IFN-γ plus IL-1β or TNF-α plus IL-1β, they were able to increase the mRNA expression of DRAK2 and induce apoptosis. Interestingly, when pancreatic β-cells were separately treated with the three abovementioned cytokines, they failed to increase the expression of DRAK2. This phenomenon coincides with our previous mention of the inability of the three cytokines to trigger islet β-cell apoptosis when used alone. The destruction of islet β-cells by IFN-γ plus IL-1β is mainly mediated by NO (Cnop et al., 2005), and this process can be blocked by inducible NO synthase (iNOS) inhibitors, which also inhibit the enhancement of DRAK2. A reduction in the number of apoptotic cells could be observed after DRAK2 knockdown, blocking the apoptosis signal mediated by caspase-9 (Mao et al., 2009). Thus, DRAK2 is also involved in this process here and acts downstream of iNOS and upstream of caspase-9. We mentioned above that NO and iNOS are involved in the effects of IFN-γ, TNF-α, and IL-1β on pancreatic β-cells, and that DRAK2 is also involved in between. Some studies have identified the specific mechanisms involved in these effects (Reddy et al., 2021). This process is associated with the activation of the nuclear factor (NF)-kb signaling pathway by IL-1β and the expression of iNOS (Figure 5). The iNOS promoter has two regions, proximal and distal, containing DNA-binding elements for different transcription factors. The proximal region includes an NF-kb and the distal region includes another NF-kb, an IFN-γ activation site (GAS), and two IFN stimulatory response elements (ISRES) right next to each other. First, IL-1β can stimulate both NF-kb sites proximal and distal to the iNOS promoter, of which the site at the far end is the primary site; however, IL-1β alone is not sufficient to activate iNOS as a single cytokine. IFN-γ simultaneously activates tyrosine kinases JAK1 and JAK2 via surface receptors, followed by dimerization of the phosphorylated transcription factor STAT-1 to bind to the GAS (Neagu and Constantin, 2021). In addition, lipopolysaccharide in macrophages can directly induce the STAT-1 isoform STAT-1α and bind to the GAS (Salim et al., 2016). Besides, STAT can also increase the expression of iNOS by inducing the transcription factor IRF-1 (Sudhakar et al., 2013), and the increased expression of iNOS will eventually produce more NO to damage islet β-cells. Thus, in the apoptosis of pancreatic β-cells, DRAK2 acts downstream of NO and upstream of caspase-9. TNF-α likewise assists IL-1β in inducing apoptosis in pancreatic islet β-cells (Kaminitz et al., 2017; Xie et al., 2018; Quattrin et al., 2020; Perdigoto et al., 2022). When TNF-α binds to its receptor, it induces the production of TRAF6, while the IL-1/IL-1R1/IL-1AcP complex induced by IL-1β simultaneously binds to TRAF6 and forms a new complex (Li L. et al., 2022), stimulating iNOS expression via the mitogen-activated protein kinase signaling pathway (MAPK) (Xiong et al., 2017; Sato et al., 2018). In addition, TRAF6 is also involved in NF-xB activation via NF-xB-inducible kinase (NIK) (Li and Wang, 2022). In conclusion, we learned from the above analysis that IFN-γ, TNF-α, IL-1β, NO, and iNOS are all upstream of DRAK2 and can affect the expression of DRAK2 through different pathways. ROS and free fatty acids (FFA) are also upstream of DRAK2, and as we mentioned above, ROS can regulate DRAK2 through PKD during T-cell activation. In addition to the process of T-cell activation, ROS can also exert apoptotic effects through PKD in other environments (Cobbaut and Van Lint, 2018). For example, under oxidative stress, PKD1 in cancer cells can regulate NF-kb through the IKK complex (Yuan et al., 2022). Moreover, in MCF7 cells, H2O2 activates NIK and phosphorylates IKKα via induction, followed by activation of NF-kb via a non-classical pathway (Li and Engelhardt, 2006). This process then intersects with the pathway through which IFN-γ, TNF-α, and IL-1β regulate DRAK2 via NF-kb. FFA has also been shown to induce DRAK2 under certain conditions (Mao et al., 2008; Lan et al., 2018; Li et al., 2021; Wu and Kapfhammer, 2021). In primary mouse hepatocytes, DRAK2 expression at both the protein level and the RNA level increased with increasing concentrations of palmitic acid (PA) and with increasing duration of action (Li et al., 2021). The study also confirmed that the expression of DRAK2 in vivo was associated with the development of NAFLD. DRAK2 is associated with the development of several diseases, such as NAFLD, diabetes mellitus (McGargill et al., 2008), and breast cancer. In addition, it is also involved in important processes such as autophagy and T-cell differentiation. Inhibition of DRAK2 expression can avoid a variety of pathological mechanisms and maintain healthy homeostasis. In addition, DRAK2 protein is mainly expressed in lymphoid organs, mostly in B cells, but also in higher amounts in T-cells and not in natural killer (NK) cells, macrophages, or dendritic cells. One study found that when DRAK2 expression was inhibited, mice develop resistance to a T-cell-mediated autoimmune disease—experimental autoimmune encephalomyelitis (EAE) (McGargill et al., 2008; Ramos et al., 2008), and it is also resistant to type I diabetes. Allogeneic rejection has also been recently shown to involve DRAK2 signaling, and inhibition of DRAK2 may maintain graft activity in the long term (Weist et al., 2012). On the basis of these findings, DRAK2 is likely to be a potential drug target for the treatment of autoimmune diseases and the prevention of graft rejection after organ transplantation. Thus, the presence of small-molecule inhibitors of DRAK2 is particularly important. but research on inhibitors of DRAK2 is currently very limited. Therefore, we have presented a compilation of previously published data on small-molecule inhibitors in the literature. SC82510 was shown to inhibit DRAK2 at very low concentrations (1 nM) and to promote neuronal differentiation and axonal branch growth in pheochromocytoma (PC12) cells; moreover, this process was enhanced by the addition of neuronal growth factor (FGF-2) (Marvaldi et al., 2014). The team first used KINOMEscan™ to determine the binding constants of different compounds to 442 eukaryotic kinases and screened two classes of compounds, after which they tested all compounds for inhibition of DRAK2. The results showed that SC84458 has a high DRAK2 inhibitory activity, but its specificity is poor and it also shows inhibitory effects on other kinases. However, SC82510 is characterized by low activity and high specificity, inhibiting only DRAK1, DRAK2, and RPSK2, and it was able to induce neuronal differentiation of PC12 cells at very low concentrations (1/5 nM). However, the chemical structure of SC82510 was not shown in the article. Indirubin derivatives were identified as novel DRAK2 small-molecule inhibitors in 2016 (Jung et al., 2016); these derivatives are the main ingredients of two Chinese herbal medicines, Ginseng and Deer Antler Pills and Qing Dai, which have been used to treat chronic granulocytic leukemia in China (Wang et al., 2014; Blazevic et al., 2015; Wang H. et al., 2021). In addition to its anticancer properties, indirubin has also been shown to be effective against diseases such as psoriasis (Hsieh et al., 2012; Sun et al., 2021), Alzheimer’s disease (Chen et al., 2017; Du et al., 2018), and autoimmunity. Although monomeric indirubin shows disadvantages such as poor water solubility and poor drug metabolism kinetics (Wang H. et al., 2021)., but because of its anticancer properties, researchers have attempted to develop derivatives based on indirubin to increase the efficacy and improve the metabolic dynamics of the drug and reduce the original limitations. One team used high-throughput screening to show that indirubin derivatives could act as novel inhibitors of DRAK2. The team first screened 11,000 compounds by an in vitro kinase assay using recombinant DRAK2 protein as the zymogen, and finally identified 41 compounds with 70% inhibition effect and strong activity. These compounds were then subjected to high-throughput screening, and those with activity were found to be indirubin or indirubin-3′-monoxime derivatives. A subsequent series of processes yielded 33 compounds, of which compounds 15–19 significantly increased the inhibition of DRAK2 and compounds 22–33 showed only moderate inhibition (detailed information can be found in the corresponding study (Jung et al., 2016)). Among these, compound 16 showed the highest specificity and inhibited DRAK2 in an ATP-competitive manner (Figure 6). This study also demonstrated that 5-N-acyl indirubin-3′-monoxime compounds are novel DRAK2 inhibitors. Benzofuran-3(2H)-one derivatives have been reported to act as inhibitors of DRAK2 to avoid islet β-cell apoptosis (Wang et al., 2017). The team first identified 2-(3,4-dihydroxybenzylidenebenzofuran-3(2 H)-one as a moderate DRAK2 inhibitor with a half-inhibitory concentration (IC50) of 3.15 µM by high-throughput screening. Subsequently, a constitutive relationship (SAR) study was performed, and a total of 36 compounds were synthesized; the addition of methoxy to the 5-, 6-, and 7-positions of benzofuran-3(2H)-one was found to increase the activity of the compounds. The most effective of these were compounds 40 and 41 (Figure 6), which also showed dose-dependent protective effects on pancreatic β-cells from PA-induced apoptosis in the glucose-stimulated insulin secretion assay (GSIS). A thieno[2,3-b]pyridine derivative was identified as an inhibitor of DRAK2 in 2014 (Gao et al., 2014). However, it also lacked selectivity and has the same inhibitory effect on DRAK1 with an IC50 of 0.82 µM. The team first used the KINOMEscan™ kinase platform for screening their proprietary compound library, and screened 150 representative compounds as potential ligands for DRAK2 in the DiscoverX binding assay. Finally, a compound based on an isothiazolo[5,4-b]pyridine scaffold was found with a Kd value of 1.6 µM. Subsequently, the isothiazolo[5,4b]pyridine derivative was used as the starting point to generate thieno[2,3-b]pyridine derivatives by binding to the scaffold-jumping method, which was experimentally shown to yield strong binding to DRAK2 (Kd = 9 nM). In general, the process starts with an isothiazolo[5,4-b]pyridine based hit compound with weak affinity for DRAK2, and its substituents are systematically altered to obtain compounds without inhibitory activity but with a binding affinity of 0.5 µM. The scaffold-hopping strategy subsequently revealed that the thieno[2,3-b]pyrazine derivative could serve as an effective ligand for DRAK2. However, the thieno[2,3-b]pyrazine derivative, while having no inhibitory activity against DAPK1, DAPK2 and DAPK3, was not selective for DRAK1 and DRAK2. Therefore, the thieno[2,3-b]pyrazine derivative was considered to be a dual inhibitor of DRAK1 and DRAK2. Compound 1 (Figure 6) could inhibit the activity of DRAK2 (Farag and Roh, 2019). The compound was mainly obtained as 5-arylthieno[2,3-b]pyridine as a scaffold and by the scaffold-jumping method. Its IC50 value for DRAK2 was 0.86 µM, and the Kd value was 9 nM, indicating high specificity but slightly lower inhibitory activity. It could be the starting point for the synthesis of highly selective DRAK2 inhibitors. It also showed an inhibitory effect on DRAK1 with an IC50 value of 2.25 µM. The team obtained a series of 5-arylthieno[2,3-b]pyridines by performing some substitution patterns on compound 1, with the most potent compound 2 (Figure 6) having an IC50 value of 29 nM and a strong binding affinity (Kd = 0.008 µM) (Leonczak et al., 2014). However, it was a non-selective DRAK2 inhibitor with poor specificity and inhibited DAPK1, DAPK2, DAPK3, and DRAK1, with Kd values of 54 and 99 nM for DAPK1 and DRAK1, respectively. In addition to some of the small-molecule inhibitors mentioned above, other substances also show DRAK2-inhibiting effects, but they have received less attention in the literature. Oximes have been shown to inhibit DRAK2 (Schepetkin et al., 2021), while 2-benzylidenebenzofuran-3-one was shown to inhibit DRAK2 activity. Drugs such as nintedanib, abemaciclib, and baricitinib can also inhibit its activity, of which nintedanib and abemaciclib were approved by the FDA in 2014 and 2017, respectively. Nintedanib is an intracellular inhibitor of tyrosine kinases that inhibits the processes involved in the progression of pulmonary fibrosis (Flaherty et al., 2019). Although not selective for DAPK inhibition, the affinity of nintedanib for DRAK2 was much higher than that for DAPK2/3, with a Kd of 3.2 and 2.1 nM for DAPK2/3 and 110/670 nM for DRAK1/2, respectively. Abemaciclib is an oral, sequentially administered CDK4/6 inhibitor approved for HR+, HER2-advanced breast cancer (ABC) (Goetz et al., 2017; Johnston et al., 2020). However, it is also not selective and is less effective in inhibiting DRAK than DAPK1/2/3 at the same concentration. Baricitinib is an oral, reversible inhibitor of the Janus kinases JAK1 and JAK2 and may have therapeutic value for patients with rheumatoid arthritis. In addition to treating rheumatoid arthritis (Hu et al., 2022; Roskoski, 2022), it has been also shown to inhibit DAPKs, with the percentage of inhibition of DRAK1 and DRAK2 being 99.5% and 98.6%, respectively. In addition to the abovementioned drugs, ISIS Pharmaceutical company also developed antisense oligonucleotides that can inhibit DRAK2. Moreover, Bennett and Dobie in 2011 investigated the selectivity of 72 known kinase inhibitors for 442 kinases in vivo. Five compounds, including KW2449, lestaurtinib, MLN-8054, R406, and TG-101348 (Figure 6), were found to have strong inhibitory effects on DRAK2. However, they shared a common disadvantage with the drugs mentioned above, which is that they are not selective. In addition, one study confirmed that Alstonlarsine A (Figure 6), one of the four indole alkaloids isolated from Alstonia scholaris, showed moderate inhibitory activity against DRAK2 with an IC50 value of 11.65 ± 0.63 μΜ (Zhu et al., 2019). All of the above small molecule inhibitors have shown varying degrees of inhibition of DRAK2 and have some potential for the treatment of related diseases, and are believed to be able to be used in the clinic in the near future. Research on DRAK2 has never ceased, moving from initial studies of DRAK2 structure to studies of its function (Chen and MacDonald, 2022), including the discovery of the first important function of DRAK2, i.e., its ability to regulate T-cell activation via Ca ions (Friedrich et al., 2005) to the discovery that DRAK2 plays an important role in the immune system (Schaumburg et al., 2007; McGargill et al., 2008). Deletion or low expression of DRAK2 significantly increases resistance to autoimmune diseases. In addition, recent studies have found a correlation between DRAK2 and tumorigenesis in diseases such as chronic lymphocytic leukemia, acute myeloid leukemia (Ye et al., 2013), colorectal cancer, and cutaneous T-cell lymphoma (CTCL) (Hartmann et al., 2008). In the field of drug research, there has been ongoing development of small-molecule inhibitors of DRAK2. Since the discovery that DRAK2 is involved in the development of many cancers, numerous researchers have aspired to hinder disease progression with small-molecule inhibitors of DRAK2, such as the compounds mentioned above. Notably, recent studies have found a strong link between DRAK2 and the development of NAFLD. This is the first study to show that DRAK2 is associated with a typical metabolic disease. NAFLD (Byrne and Targher, 2015; Friedman et al., 2018; Younossi et al., 2018) is a disease of the liver characterized by hepatic steatosis after excluding other known causes (e.g., high alcohol intake, viral infections, etc.), which can further progress to hepatitis, cirrhosis, and even liver cancer. The team demonstrated that DRAK2 exacerbates NAFLD through an SRSF6-related RNA alternative splicing mechanism. When DRAK2 is overexpressed, the kinase binds to SRSF6, resulting in its inability to be phosphorylated by SRPK1. This prevents the complex from entering the nucleus and participating in the splicing process of genes involved in mitochondrial function, thus affecting mitochondrial function and causing NAFLD. However, the study focused on the mechanism by which DRAK2 contributes to the development of NAFLD, so what causes the overexpression of DRAK2 in vivo? As mentioned above in relation to the pathways upstream of DRAK2, all of these pathways can stimulate DRAK2 expression under certain conditions. However, the majority of patients with NAFLD are obese (Fan et al., 2017)., and most obese patients show relatively strong levels of oxidative stress and low levels of inflammation (Karam et al., 2017; Perez-Torres et al., 2021). In combination with our previous description, this finding suggests that obese patients are at high risk of NAFLD because they are prone to oxidative stress and inflammation in the body. The ROS produced by oxidative stress and inflammation leads to high expression of DRAK2 via PKD1/2/3. Moreover, overexpression of DRAK2 can damage mitochondrial function and lead to NAFLD, which can eventually further exacerbate oxidative stress and inflammation. Although DRAK2 undoubtedly plays an extremely important role in the early stages of NAFLD development, because of the limitations of research techniques and challenges in clinical application, the mechanisms underlying the promotion of DRAK2 expression by ROS remain to be investigated.

PMC9649745

Dalia Pakalniškytė,Tanja Schönberger,Benjamin Strobel,Birgit Stierstorfer,Thorsten Lamla,Michael Schuler,Martin Lenter

Rosa26-LSL-dCas9-VPR: a versatile mouse model for tissue specific and simultaneous activation of multiple genes for drug discovery

10-11-2022

Transgenic organisms,Gene expression

Transgenic animals with increased or abrogated target gene expression are powerful tools for drug discovery research. Here, we developed a CRISPR-based Rosa26-LSL-dCas9-VPR mouse model for targeted induction of endogenous gene expression using different Adeno-associated virus (AAV) capsid variants for tissue-specific gRNAs delivery. To show applicability of the model, we targeted low-density lipoprotein receptor (LDLR) and proprotein convertase subtilisin/kexin type 9 (PCSK9), either individually or together. We induced up to ninefold higher expression of hepatocellular proteins. In consequence of LDLR upregulation, plasma LDL levels almost abolished, whereas upregulation of PCSK9 led to increased plasma LDL and cholesterol levels. Strikingly, simultaneous upregulation of both LDLR and PCSK9 resulted in almost unaltered LDL levels. Additionally, we used our model to achieve expression of all α1-Antitrypsin (AAT) gene paralogues simultaneously. These results show the potential of our model as a versatile tool for optimized targeted gene expression, alone or in combination.

Rosa26-LSL-dCas9-VPR: a versatile mouse model for tissue specific and simultaneous activation of multiple genes for drug discovery Transgenic animals with increased or abrogated target gene expression are powerful tools for drug discovery research. Here, we developed a CRISPR-based Rosa26-LSL-dCas9-VPR mouse model for targeted induction of endogenous gene expression using different Adeno-associated virus (AAV) capsid variants for tissue-specific gRNAs delivery. To show applicability of the model, we targeted low-density lipoprotein receptor (LDLR) and proprotein convertase subtilisin/kexin type 9 (PCSK9), either individually or together. We induced up to ninefold higher expression of hepatocellular proteins. In consequence of LDLR upregulation, plasma LDL levels almost abolished, whereas upregulation of PCSK9 led to increased plasma LDL and cholesterol levels. Strikingly, simultaneous upregulation of both LDLR and PCSK9 resulted in almost unaltered LDL levels. Additionally, we used our model to achieve expression of all α1-Antitrypsin (AAT) gene paralogues simultaneously. These results show the potential of our model as a versatile tool for optimized targeted gene expression, alone or in combination. Genetically modified mice, as in vivo models to study human disease mechanisms, have a long history that started by the end of the twentieth century. Since that time, considerable technical advances and new technologies have revolutionized our ability to manipulate the mouse genome and enhance the potential of these models to support preclinical drug discovery. In particular, the application of endonucleases has greatly enhanced the feasibility for researchers to manipulate the mouse genome as desired. Amongst them, the most widely used is the RNA-guided endonuclease CRISPR associated protein 9 (Cas9) system, which induces DNA double strand breaks with high specificity. The specificity is provided by a short 20 base pair spacer sequence of a guide RNA (gRNA) that recognizes the target DNA region of interest and directs the nuclease for editing. Over the last few years, the CRISPR-Cas9 system was adopted, and several Cas9 variants have been generated that lack endonuclease activity, while retaining specificity for target DNA, for applications beyond classical endonuclease activity. One of them is the CRISPR activation (CRISPRa) system, where endonuclease dead Cas9 (dCas9) is fused with four tandem copies of Herpes Simplex Viral Protein 16 (VP64), human NF-kB p65 activation domain (p65), and Epstein-Barr Virus-derived R transactivator (Rta) domains to obtain a programmable transcription factor, termed VPR. This hybrid dCas9-VPR was demonstrated to have a highly efficient potential for activating gene transcription of almost any gene of interest in various species and cell types and led to the development of corresponding dCas9-VPR expressing mouse models. Here, the targeted transcriptional regulation of genes is obtained by delivering appropriate gRNAs complementary to the promoter region of the gene of interest. Amongst various delivery methods, recombinant AAV (rAAV) is one of the most investigated and preferred tools, due to its relative safety, low immunogenicity, and ability to transduce a broad range of cells. In addition, rAAV is replication defective, does not integrate into the host genome, and persists in transduced cells in an episomal fashion, thereby providing long-term transgene expression. Moreover, due to the broad range of natural and capsid-engineered rAAV variants that differ in their transduction efficiency and tissue tropism, transgene delivery to specific cell or tissue types can be achieved. One of the most efficient AAV serotypes for liver transduction in mice is AAV8, which was shown to transduce up to 90–95% of hepatocytes subsequent to intraportal vein or intravenous injection. The liver is one of the most important organs in the body, which is directly or indirectly involved in many essential physiological processes, where reduction or loss of liver function can be life threatening. Hence, liver associated enzymes, circulating proteins and cell receptors are popular targets in the focus of ongoing drug discovery approaches. In this context, hepatocyte expressed LDLR plays an important role in plasma cholesterol homeostasis, where dysregulation leads to a higher risk for the development of cardiovascular diseases LDLR is located at the cell surface of hepatocytes, where it interacts with plasma derived LDL. After binding, the LDLR-LDL complex is internalized and transported into endosomes. Once LDL has been released, LDLR can recirculate to the cell surface or is degraded in the lysosomes. The degradation rate of LDLR can be regulated by modifying another enzyme (PCSK9), which is mainly liver expressed. PCSK9 interacts with LDLR on the cell surface and targets LDLR to lysosomes for its degradation. It has been shown that high levels of circulating PCSK9 increase concentrations of plasma LDL, increasing the risk of atherosclerosis development. Another important protein expressed by hepatocytes is AAT, which is encoded by the SERPINA1 gene. Liver secreted AAT circulates in the blood and its main function is to control activity of various serine proteinases. The primary target of AAT is neutrophil elastase. The number of SERPINA1 genes vary among different mammalian species. Primates, including humans, and some mouse strains contain only a single gene copy, while other mouse lines contain multiple paralogues originating from the same ancestral gene. For instance, C57Bl/6 J mice contain five Serpina1 paralogues, namely Serpina1a, Serpina1b, Serpina1c, Serpina1d and Serpina1e. Here, we describe the development and use of a new Cre recombinase-dependent dCas9-VPR mouse model, with the potential of long-lasting transcriptional activation in vivo. This mouse model is applicable for gene induction by gRNAs to target different genes of interest, individually or in combination. Particularly, by using liver tropic AAV8 and specific gRNAs, we demonstrate tissue specific upregulation of LDLR, PCSK9 and AAT in hepatocytes. For AAT, all five Serpina1 gene variants could be simultaneously upregulated in our mouse model using a mix of five AAV8 preparations, each containing specific gRNAs against one of the five Serpina1 gene variants. Conditions of hyper- or hypocholesterolemia were successfully induced in these mice by activating the expression of either hepatic PCSK9 and/or LDLR. Taken together, these studies demonstrate the potential of the new Rosa26-LSL-dCas9-VPR mouse model for targeted transcriptional gene activation, thereby enabling rapid characterization and validation of gene function in basic biological research or drug discovery. We generated the Cre-dependent CRISPR activation mouse line, termed Rosa26-LSL-dCas9-VPR, using the recombinase mediated cassette exchange technology to integrate the dCas9-VPR gene containing cassette into the ROSA26 locus (Fig. 1A). The cloned targeting vector contained a NeoR cassette, the human CAG promoter, and a translation interrupting LSL cassette linked with the dCas9-VPR gene fused to a self-cleaving P2A sequence and Egfp gene (Fig. 1A). To investigate the effectiveness of Cre recombinase-mediated LSL cassette excision, and consequently dCas9-VPR expression activation, in a tissue specific manner, Rosa26-LSL-dCas9-VPR mice were either injected with a preferentially liver transducing AAV8 containing the Cre gene under the control of a liver specific LP1 promoter (AAV8-Cre) or AAV8 carrying U6 promoter driven guide RNAs targeting Pcsk9 (gPcsk9) (AAV8-gPcsk9), where each gRNA is driven by separate U6 promoter (Fig. 1B). Recombination (i.e., excision of the LSL cassette) was only observed in liver tissue of AAV8-Cre administered Rosa26-LSL-dCas9-VPR mice, but not in solely AAV8-gPcsk9 transduced mice (Fig. 1C). Moreover, in AAV8-Cre treated mice, dCas9-VPR expression was restricted to liver, with absent or only minor but not statistically significant increases in all other investigated tissues, including heart, lung, kidney, and spleen (Fig. 1D). These results show that Rosa26-LSL-dCas9-VPR mice in combination with AAV8 and LP1 promoter driven Cre expression allow efficient and liver specific induction of dCas9-VPR expression. To determine whether the Rosa26 knock-in construct provided functional levels of dCas9-VPR expression, we next investigated parallel transduction of CRISPRa mice with AAV8-Cre and a set of five AAV8s containing gRNAs targeting the five Serpina1 paralogues a-e (AAV8-gSerpina1a, AAV8-gSerpina1b, AAV8-gSerpina1c, AAV8-gSerpina1d, AAV8-gSerpina1e) (Supplementary Fig. S1). Each paralogue is targeted in parallel by 6 different gRNAs, where each gRNA is driven by an individual U6 promoter (Fig. 2A). We injected two groups of animals and collected blood samples 10- and 21-days post transduction. While the animals of the first group received only AAV8-Cre, the second group received a combination of AAV8-Cre and all five AAV8-gSerpina1, each targeting one of the five Serpina1 gene variants a-e (AAV8-gSerpina1(a1-6-e1-6)), with the aim to upregulate all 5 liver Serpina1 paralogues simultaneously in each animal (Fig. 2B). 21 days post transduction, Rosa26-LSL-dCas9-VPR mice were investigated for AAT expression in the liver and blood. mRNA analysis of all Serpina1 variants at day 21 showed an increased expression in the liver of mice that received both Cre and Serpina1 gRNAs, demonstrating that dCas9-VPR expression was successfully induced and capable to form a ribonucleoprotein complex (RNP) with the provided gRNAs to facilitate on-target gene over-expression (Fig. 2C). Furthermore, we performed protein analysis of liver samples and found a fivefold overexpression of the SERPINA1A paralogue as compared to animals receiving only AAV8-Cre (Fig. 2D). To confirm this observation, we additionally quantified AAT levels in plasma (Fig. 2E). In line with the results described above, injection of mice with AAV8-Cre and AAV8-gSerpina(n) led to significantly increased AAT plasma levels detected on day 10 post-transduction, which further increased on day 21 post injection, while control levels remained almost unchanged (Fig. 2E). Taken together, these data demonstrate the successful transcriptional induction of all five Serpina1 paralogues, thereby providing evidence for the use of the Rosa26-LSL-dCas9-VPR mice in combination with AAV8 encoded gRNAs for efficient upregulation of Serpina1 transcription resulting in increased protein levels in liver and plasma. A key advantage of CRISPR-based models is their potential for targeting of two or more genes at the same time by combining different gRNAs. To study the possibility of multiple gene upregulation in our Cre-dependent dCas9-VPR mice, we selected the well characterized PCSK9-LDLR-LDL regulatory loop for the next experiment. PCSK9 plays an important role in cholesterol homeostasis by forming a complex with LDLR on the cell surface, thereby inducing LDLR’s internalization and subsequent lysosomal degradation. To induce PCSK9 and LDLR expression, we selected a set of 6 different gRNAs, each targeting either Ldlr or Pcsk9 (Fig. 3A). To study the function of both genes on cholesterol homeostasis, we analyzed liver and blood 21 days post AAV injection. Animals of the first group were injected with AAV8-Cre alone, as a control, whereas the second and third groups additionally received either AAV8-gPcsk9 or AAV8-gLdlr, respectively (Supplementary Fig. S1). The fourth group was injected with a combination of three different viruses: AAV8-Cre, AAV8-gLdlr and AAV8-gPcsk9 (Fig. 3B). As expected, treatment with AAV8-Cre either in combination with AAV8-gLdlr or AAV8-gPcsk9 led to a significant, approximately threefold transcriptional upregulation of either Ldlr or Pcsk9, compared to the control group (Fig. 3C,D). In group four (AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9), the transcriptional upregulation of both Ldlr and Pcsk9 was comparable to the individual induction of Pcsk9 expression in group two (AAV8-Cre + AAV8-gPcsk9) or Ldlr in group three (AAV8-Cre + AAV8-gLdlr) (Fig. 3C,D). The increase in mRNA translated into an even more pronounced upregulation on protein levels, with a ninefold or fourfold overexpression for LDLR or PCSK9, respectively (Fig. 3E,F). Specificity of PCSK9 detection via Wes™ was confirmed using recombinant PCSK9 protein. Virtual blot-like images are shown in Supplementary Fig. S2. Upregulation of PCSK9 reduced LDLR protein in liver tissue by tenfold (Fig. 3E, AAV8-Cre + AAV8-gPcsk9), whereas upregulation of LDLR protein did not affect the PCSK9 protein amount in liver tissue (Fig. 3F, AAV8-Cre + AAV8-gLdlr), but reduced circulating PCSK9 levels in plasma by 13-fold (Fig. 3G, AAV8-Cre + AAV8-gLdlr). Simultaneous upregulation of PCSK9 and LDLR proteins led to an almost twofold increase in LDLR protein amount in liver (Fig. 3E, AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9), whereas the detected PCSK9 amounts were increased almost threefold compared to the control group (Fig. 3F, AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9). However, despite these higher PCSK9 protein amounts observed in the liver tissue lysates, no significant PCSK9 increase was detected in the corresponding plasma samples (Fig. 3G, AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9). To confirm the increase of LDLR on the surface of the hepatocytes from the dCas9-VPR expressing mice, we additionally performed immunohistochemistry analyses on liver sections stained with an anti-LDLR antibody. The antibody dilution was titrated to see a moderate staining in the control group showing faint cytoplasmic staining and, in a fraction of hepatocytes, distinct membrane staining. As expected, upregulation of PCSK9 in the experimental group two led to decreased LDLR staining compared to the control group (Fig. 3H, AAV8-Cre + AAV8-gPcsk9), whereas upregulation of the LDL-receptor resulted in a strong increase in membrane staining and, to a lesser extent, in cytoplasmic staining (AAV8-Cre + AAV8-gLdlr). After simultaneous upregulation of both PCSK9 and LDLR, the staining for LDLR levels is comparable to the control group (Fig. 3H, AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9). To evaluate the effect of hepatic LDLR overexpression, or its reduction by the enhanced PCSK9 expression, on plasma cholesterol levels, we subjected plasma of the AAV8-transduced dCas9-VPR mice to lipoprotein analysis. As expected, treatment with AAV8-Cre in combination with AAV8-gLdlr led to a significant decrease of LDL, HDL as well as cholesterol in plasma compared to the samples from the control group (Fig. 4A–C, AAV8-Cre and AAV8-Cre + AAV8-gLdlr). In detail, the HDL and total cholesterol levels dropped threefold, whereas plasma LDL dropped to an undetectable amount. In line with these data, AAV-mediated overexpression of PCSK9 increased LDL, HDL and total cholesterol concentrations in plasma compared to the control group samples (Fig. 4A–C, AAV8-Cre and AAV8-Cre + AAV8-gPcsk9). In this context, upregulation of PCSK9 resulted in fourfold higher plasma LDL levels and a twofold increase for total cholesterol when compared to the control group (AAV8-Cre), whereas HDL levels were almost unchanged (1.2-fold increase) (Fig. 4B). Notably, simultaneous upregulation of LDLR in parallel to PCSK9 (Fig. 4A–C, AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9) still resulted in a reduction of plasma LDL, HDL and cholesterol levels, but less pronounced than upon upregulation of LDLR alone. This finding is in concordance with a residual increase of LDLR levels of almost twofold compared to control, despite PCSK9 induction (Fig. 3E, AAV8-Cre + AAV8-gLdlr + AAV8-gPcsk9 and AAV8-Cre). In our study, we generated a novel, conditional (Cre-dependent) dCas9-VPR expressing mouse line and demonstrated its utility for tissue specific gene upregulation using AAV-mediated gRNA expression. Our mouse model offers a versatile basis for diverse research applications that require fine-tuning of targeted expression of any gene of interest using either a mixture or a single AAV with varying tissue tropism, thereby providing the opportunity to simultaneously activate multiple genes in vivo. Additionally, our model offers the advantage to restrict dCas9-VPR expression to a desired tissue by placing Cre-recombinase under a tissue-specific promoter, if needed. Alternatively, the mice could also be cross-bred with Cre driver lines to induce tissue- or cell-specific dCas9-VPR expression. The primary goal of our study was to establish and improve a CRISPR activation model to study pathways and molecular interactions in conjunction with tailored disease models to reproduce human pathological conditions for basic research and drug development. Because of their phylogenetic relatedness and physiological similarity to humans, the use of mice as tools in biomedical research is well established. Unfortunately, none of these models precisely mimic the human phenotype exactly enough, leading to variations in efficacy and toxicity of drug candidates compared to humans in the past. Even though the genetic pathways regulating normal and physiological conditions are quite conserved, the intrinsic genetic differences sometimes complicate the direct comparison between the species. One example is the generation of authentic AAT mouse-models, where the establishment of comparable disease models are hampered by the complexity of the murine Serpina1-genes. The need for mouse models to upregulate the expression of all Serpina1 paralogues simultaneously is of particular importance since there is evidence to suggest that overexpression of AAT is most probably involved in cancer related processes. It has been already shown that higher AAT expression promotes invasion and metastasis as well as correlates with poor prognosis in patients with lung, colon, skin, and gastric cancer. The mouse model presented here will therefore be highly relevant to further investigate not only these findings but can be easily adapted to similar genetic conditions. The most important features to reproduce human diseases are the precision of etiology as well as the ability to reproduce the features of the pathological process. The value of our mouse model to induce and study the regulation of complex pathological conditions is, therefore, demonstrated by the successful modulation of cholesterol metabolism by the hepatic overexpression of two system relevant key players, namely LDLR and PCSK9, alone and in combination. In accordance with recent studies, overexpression of circulating PCSK9 led to reduced LDLR, which was accompanied by increased LDL and cholesterol concentrations in plasma and vice versa. In addition to this, we also observed sex dependent differences in serum/plasma levels of alpa-1-antitrypsin (male > female), LDL (female > male) and PCSK9 in control mice in accordance with the literature. This distinction was still visible in AAT plasma levels despite a general 1.6-fold increase after overexpression. In contrast to this, we could not detect animal sex related significant differences in liver dCas9-VPR mRNA expression. As demonstrated with these data, our model is well suited to study PCSK9 function in the liver via LDLR depletion, but moreover, also possible effects on other organs can be easier addressed, e.g., by the investigation of compensatory effects mediated via additional tissue specific LDLR upregulation. Several CRISPRa mouse models have already been generated, which mainly differ in the choice of the transcriptional activator, chosen locus for dCas9 gene knock-in, the target vector design, and whether dCas9 is conditionally or constitutively expressed. We specifically decided to use dCas9 fused to VPR under the control of a CAG promoter within the Rosa26 locus and downstream of an LSL cassette. To allow gene upregulation in any tissue of interest, we followed the strategy to use the "safe harbor" locus Rosa26 as the preferred site for ubiquitous expression of our transgene. By doing so, we made sure to reach similar expression of dCas9-VPR across various tissues, without affecting endogenous gene expression as observed in mouse models in the past. This was also done by Hunt et al., who published a CRISPRa model using the Rosa26 locus. However, in this configuration the dCas9-synergistic activation mediator (SAM) was placed exclusively under the transcriptional control of the Rosa26 promoter, which might limit the transgene expression levels due to its moderate strength. To overcome this limitation, our (and other) CRISPRa mouse model was generated by additionally inserting a strong exogenous CAG-promoter into the Rosa26 locus upstream of the dCas9 transgene. The decision to use dCas9 fused to the VPR activator was mainly fostered by a comprehensive study by Chavez et.al., where they compared different Cas9 activator systems in several human, mouse and fly cell lines. Even though AAV based gRNA delivery can often be sufficient for tissue-specific target gene expression, we aimed to further increase tissue specificity in our model by conditional dCas9-VPR expression. In addition to that, transgene expression needs to be supervised to prevent unwanted side effects, as several publications pointed out a possible dCas9-VPR toxicity. Moreover, by limiting dCas9 expression to certain tissues, also the risk of gRNA off-targeting is reduced. Narrowing down dCas9-VPR expression to defined cell types and tissues in our mouse model can be achieved by combining tissue tropism provided by AAV serotypes with tissue specific promoter driven Cre. This strategy is of special interest since the field of AAV capsid engineering is thriving and a number of additional AAV serotypes have been isolated and new capsid variants have been generated in recent years. Despite these attractive features, the CRISPRa/AAV-guide system still possesses some limitations, such as the limited availability of truly tissue or cell specific promoters for controlled Cre expression, inefficient in vivo transduction of some tissues (e.g., bone marrow, immune cells, kidney) with AAVs, and specificity of gRNAs. Nevertheless, in light of the large body of information gained from studies in mice, where different AAV serotype vectors have been shown to exhibit distinct tropism for various tissues, ongoing efforts to identify novel promoter/enhancer elements for a variety of tissues, and continuous improvements of CRSIPR technology and guide design, our model holds the potential to target genes in hardly accessible tissues or cell types in the future. Finally, we have decided to use 6 gRNA sequences to target one gene, as it was demonstrated that most efficient gene upregulation is reached when more than 3 gRNAs are used in parallel. Although the selection of these gRNAs was based on a prediction algorithm that aims to select target-specific sequences, off-target effects cannot be fully excluded and therefore need to be carefully addressed in any future study using appropriate methods, e.g., ChIP-seq and prior in vitro evaluation to minimize off-target effects that might otherwise falsify data interpretation. Our Rosa26-LSL-dCas9-VPR model can be a used to study and validate pathways/molecular interactions by selected or combined overexpression of genes. It also offers the possibility for concerted overexpression of multiple gene variants in order to study their biological function jointly and/or individually. Additionally, it has the potential to generate mouse disease models by overexpression of endogenous genes based on the sustained and potentially long-lasting expression of AAV8 constructs in mouse liver and by combining multiple genes in order to achieve the expected disease phenotypes. The fast generation, the precise gene targeting, and the versatile combination of multiple genes makes this model highly attractive, not only for academia but also for industry, to support and accelerate drug discovery by providing detailed insights in target pathway biology and to set up new disease animal models with a better match to human pathology. Rosa26-LSL-dCas9-VPR mouse line was generated by recombinase mediated cassette exchange (RMCE) technology (Taconic Bioscience). RMCE vector containing F3 site, neomycin resistance (NeoR) gene, PGK polyadenylation signal, cytomegalovirus (CMV) immediate enhancer/β-actin (CAG) promoter, loxP-STOP-loxP (LSL) cassette, dCas9-VPR gene, P2A sequence, enhanced green fluorescent protein (EGFP) gene, hGH polyadenylation signal, PGK polyadenylation signal and FRT site was cloned, and transfected into a C57BL/6NTac embryonic stem cell (ESC) line containing an RMCE docking site in the Rosa26 locus (Taconic Bioscience). The targeted ESC clone was injected into BALB/c blastocysts. Spermatozoa from high-percentage chimeric male mice was used for in vitro fertilization of superovulated C57BL/6NTac female mice (Taconic Bioscience) to obtain a first colony of transgenic C57BL/6NTac-Gt(ROSA)26Sortm6458 (CAG-LSL-dCas9VPR-EGFP)Tac (termed Rosa26-LSL-dCas9-VPR or Rosa26LSL-dCas9-VPR/+) mice. Mice were housed in groups of 3–5 in individually ventilated cages at 22–25 °C, a humidity of 45–65%, and a 12 h day/night cycle with free access to water and food. Animal experiments were performed in accordance with the German Animal Welfare Act, and the guidelines of directive 2010/63/EU of the European Parliament and the Council 2010 on the protection of animals used for scientific purposes. Animal experiments performed in this study were reviewed and approved by the local authorities (Regierungspräsidium Tübingen, TVV-17-020). We hereby confirm that all methods in the study were carried out in compliance with the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines and regulations. AAV8 particles were resuspended in 0.9% NaCl saline (04671613, Deltamedica). 125 µl were injected into Rosa26LSL-dCas9-VPR/+ mice via tail vein injection at a dose of 1 × 1011 VG/mouse for each AAV variant. For all experiments, male and female mice at the age of 20–26 weeks were used and genders were distributed equally between the experimental groups. gRNA and LP1-Cre sequences were cloned into separate plasmids harboring AAV2-inverted-terminal repeats (ITRs) with one ITR harboring a mutated terminal resolution site, therefore resulting in a self-complementary genome. Each gRNA expression vector contained six consecutives human U6 promoters, each followed by a gRNA coding sequence. The native Streptococcus pyogenes-derived gRNA scaffold structure was optimized based on previous publications. gRNA sequences were taken from a data base, published by Horlbeck et al., and are listed in Supplementary Table S1. The Cre expression vector was composed of an LP1 promoter, Cre gene and an SV40 polyA sequence. Complete plasmid sequences and annotations can be found in Supplementary Table S2. AAVs were produced using frozen high-density HEK293 cell stocks and CELLdiscs (678116, Greiner Bio-one) as previously described. Briefly, AAVs were produced by calcium phosphate transfection in a minimum of three 16-layer CELLdiscs, resulting in total yield of 2.39E+12–6.84E+12 VG, depending on the construct. Freshly thawed high-density HEK293 cell stocks were seeded with a density of 6E+07 cells per disc, followed by incubation at 37 °C for 72 h. For one disc, 1800 µg of plasmid DNA was mixed with 69 ml 300 mM calcium chloride (C7902, Sigma-Aldrich), added dropwise to 69 mL 2× HBS buffer, pH 7.0 (15450257, Thermo Fisher Scientific) and added to the cells after 2 min of incubation. Plasmid DNA contained equimolar amounts of a rep2-cap8 plasmid, pHelper (Applied Biosystems), and either Cre or one of the gRNA expression plasmids. Six hours after transfection, the medium was changed, and cells were further incubated for 72 h before harvesting. Cells were then lysed in lysis buffer, containing 1300 IU (i.e., 0.325 IU/cm2) salt active nuclease (70910-150, Scientifix) and HALT protease inhibitor (78439, Thermo Fisher Scientific). After spinning the cell debris, supernatants were collected and further processed by PEG-precipitation, iodixanol gradient ultracentrifugation and ultrafiltration, as described in detail before. Droplet digital PCR for absolute quantification of viral genomes for gRNA-expressing virus was performed using an U6 promoter specific Custom TaqMan™ Gene Expression Assay (Applied Biosystems), containing U6-Fwd, U6-Rev and U6-probe (Supplementary Table S1). Absolute quantification for Cre-expressing AAVs was performed using primers LP1-Fwd, LP1-Rev and LP1-probe (Sigma-Aldrich) (Supplementary Table S1). For protein analysis, dissected tissues were immediately snap frozen, while for RNA analysis, organs were first submerged in RNAlater™ Stabilization Solution (AM7021, ThermoFisher) for 24 h at 4 °C. Tissues were transferred into Precellys® tubes (Bertin Instruments) together with 50 µl/10 mg RLT buffer (79,216, QIAGEN) containing 1% β-mercaptoethanol (M3148, Sigma-Aldrich) for RNAlater stabilized tissues or 100 µl/10 mg RIPA buffer (R0278, Sigma-Aldrich) containing 1X HALT Protease inhibitor Cocktail (1861279, ThermoFisher) for snap frozen tissues. Tissues were homogenized at 5500 rpm for 20 s using a Precellys® 24 homogenizer (Bertin Instruments). After disruption, protein lysates were incubated for 30 min at 4 °C. Lysates were centrifuged for 20 min at 15,294g to pellet cell debris and supernatant was collected. Protein concentration was measured using Pierce™ BCA Protein Assay Kit (23225, ThermoFisher) according to the manufacturer’s instructions. For total RNA isolation, 650 µl of prepared tissue lysate was transferred along with 325 μl Phenol–chloroform–isoamyl alcohol mixture (77617, Sigma-Aldrich) into 5PRIME Phase Lock Gel Heavy tubes (2302830, Quantabio), followed by vigorous shaking for 15 s and centrifugation at 16,000g for 5 min. Next, 325 µl of Chloroform–isoamyl alcohol mixture (25666, Sigma-Aldrich) was added and the tubes were shaken again for 15 s, followed by 3 min incubation and centrifugation at 16,000g for 5 min. 350 µl of aqueous phase was collected and used to extract total RNA and using AllPrep DNA/RNA 96 Kit (80311, Qiagen). Isolation was performed according to the manufacturer’s instructions with slight modification to remove DNA contamination from the RNA fraction. For this, the RNA fraction was loaded in the RNeasy® 96 Plate and washed with 400 µl RW1 buffer. 80 µl of DNase I (79254, Qiagen) was added to each well and the RNeasy® 96 Plate was incubated for 15 min at room temperature followed by standard protocol starting with washing the 96-well plate with RW1 buffer. Either 500 ng (dCas9-VPR, Ldlr and Pcsk9) or 1 µg (Serpina1(a-e)) of total RNA was reverse-transcribed into copy DNA (cDNA) using High-Capacity cDNA Reverse Transcription Kit (4368813, Applied Biosystems). Quantitative real-time PCR was performed with a QuantiFast Probe PCR Kit (204256, Qiagen) (for Serpina1(a-e)) or TaqMan™ Gene Expression Master Mix (4370074, Applied Biosystems) (for Ldlr, Pcsk9 and dCas9-VPR expression) using the following TaqMan™ Gene Expression Assays (Applied Biosystems): Mm01177349_m1 for Ldlr, Mm02748447_g1 for Serpina1a, Mm04207706_gH for Serpina1b and Serpina1d, Mm00833655_m1 for Serpina1e, Mm00842094_mH for Serpina1d and Mm04207703_mH for Serpina1a, Serpina1b, Serpina1c, Mm01263610_m1 for Pcsk9, Mm00839502_m1 for Polr2A. dCas9-VPR expression was analyzed using a Custom TaqMan™ Gene Expression Assay (Applied Biosystems), containing dCas9-Fwd, dCas9-Rev and dCas9-probe (Supplementary Table S1). Relative Serpina1(a-e), Pcsk9 and Ldlr expressions were calculated using 2−ΔΔCt method in relation to Polr2a. LSL cassette recombination PCR was performed on liver cDNA using Quick-Load® Taq Master Mix (M0271S, NEB) with primers p1, p3, p2 (Supplementary Table S1). The PCR products with 492 bp for floxed LSL-dCas9-VPR and 393 bp for recombined dCas9-VPR products were separated on a 2% E-Gel™ EX Agarose-Gel (G401002, ThermoFisher). For genotyping, the Rosa26 locus was amplified with the PCR primers GenFw1, GenFw2 and GenRev1 (Supplementary Table S1) by using the Taq polymerase High Fidelity (11304011, ThermoFisher). The expected PCR products were 299 bp for wild-type allele and 744 bp long for knock-in allele. Tissue lysates were analyzed using automated Simple Wes system (Protein Simple) with 12–230 kDa Wes Separation Module capillary cartridges (SM-W004, Protein Simple). Anti- mouse (DM-002, Protein Simple), anti-rabbit (DM-001, Protein Simple), anti-goat (DM-006, Protein Simple) detection modules or F(ab’)2 anti-Rat IgG (H + L)-HRPO (1:20, 112-036-062, Jackson Immuno Research) were used, depending on host species of the primary antibodies. The following primary antibodies were used: LDLR (1:50, PAB8804, Abnova), SERPINA1A (1:20, MAB7690, R&D Systems), PCSK9 (1:10, AF3985, R&D Systems) and β-actin (1:20, NB600-501, Novus Biologicals). Specificity of PCSK9 detection was confirmed using recombinant mouse PCSK9 protein (9258-SE-022, R&D Systems). Compass software version 6.0.0 (Protein Simple) was used to analyze the data. Area under the peak of the protein of interest was measured and normalized with respect to the β-actin area under the peak. Proteins were determined in plasma using Mouse Proprotein Convertase 9/PCSK9 Quantikine ELISA Kit (MPC900, R&D Systems) and Mouse A1AT ELISA Kit (E-90A1T, Immunology Consultants Laboratory) according to the manufacturer’s instructions detecting all 5 paralogues of Serpina1. HDL (high-density lipoprotein), LDL and total cholesterol plasma levels were determined using a COBAS INTEGRA® 400 Plus chemistry analyzer (Roche Diagnostics, Germany), according to the manufacturer's instructions. Mouse tissues were dissected and immediately transferred to 10% neutral buffered formalin (HT501128, Sigma-Aldrich). Tissues were fixed for at least 24 h before samples were processed with an automated tissue processor (Tissue-Tek® VIP® 6, Sakura), embedded in paraffin and cut into 3 µm sections. Immunohistochemical (IHC) staining for LDLR was carried out on the automated Leica Bond RX™ platform (Leica Biosystems, Melbourne, Australia) using a monoclonal rabbit anti-LDL receptor antibody (1:1200, clone EP1553Y, ab271846, abcam) after heat-induced epitope retrieval with Bond™ Epitope Retrieval Solution 1 (ER1, pH6, Leica Biosystems, Newcastle, United Kingdom). Antibody dilution was titrated to have a moderate staining signal in livers of AAV8-Cre-only treated mice. Bound antibodies were visualized using the Bond™ Polymer Refine Detection System (Leica Biosystems, Newcastle, United Kingdom). Anti-LDLR stained sections were scanned with the Axio Scan.Z1 (20 × objective, Carl Zeiss Microscopy GmbH, Jena, Germany). Statistical analyses were performed using GraphPad Prism 9 (GraphPad). Significance was determined according to the p values as *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. Results are shown as mean values ± s.d. Comparison between experimental groups was made using a nonparametric Mann–Whitney test. Supplementary Information.

PMC9649747

Changnan Wang,Bingying Wang,Taruna Pandey,Yong Long,Jianxiu Zhang,Fiona Oh,Jessica Sima,Ruyin Guo,Yun Liu,Chao Zhang,Shaeri Mukherjee,Michael Bassik,Weichun Lin,Huichao Deng,Goncalo Vale,Jeffrey G. McDonald,Kang Shen,Dengke K. Ma

A conserved megaprotein-based molecular bridge critical for lipid trafficking and cold resilience

10-11-2022

Transport carrier,Homeostasis,Stress signalling

Cells adapt to cold by increasing levels of unsaturated phospholipids and membrane fluidity through conserved homeostatic mechanisms. Here we report an exceptionally large and evolutionarily conserved protein LPD-3 in C. elegans that mediates lipid trafficking to confer cold resilience. We identify lpd-3 mutants in a mutagenesis screen for genetic suppressors of the lipid desaturase FAT-7. LPD-3 bridges the endoplasmic reticulum (ER) and plasma membranes (PM), forming a structurally predicted hydrophobic tunnel for lipid trafficking. lpd-3 mutants exhibit abnormal phospholipid distribution, diminished FAT-7 abundance, organismic vulnerability to cold, and are rescued by Lecithin comprising unsaturated phospholipids. Deficient lpd-3 homologues in Zebrafish and mammalian cells cause defects similar to those observed in C. elegans. As mutations in BLTP1, the human orthologue of lpd-3, cause Alkuraya-Kucinskas syndrome, LPD-3 family proteins may serve as evolutionarily conserved highway bridges critical for ER-associated non-vesicular lipid trafficking and resilience to cold stress in eukaryotic cells.

A conserved megaprotein-based molecular bridge critical for lipid trafficking and cold resilience Cells adapt to cold by increasing levels of unsaturated phospholipids and membrane fluidity through conserved homeostatic mechanisms. Here we report an exceptionally large and evolutionarily conserved protein LPD-3 in C. elegans that mediates lipid trafficking to confer cold resilience. We identify lpd-3 mutants in a mutagenesis screen for genetic suppressors of the lipid desaturase FAT-7. LPD-3 bridges the endoplasmic reticulum (ER) and plasma membranes (PM), forming a structurally predicted hydrophobic tunnel for lipid trafficking. lpd-3 mutants exhibit abnormal phospholipid distribution, diminished FAT-7 abundance, organismic vulnerability to cold, and are rescued by Lecithin comprising unsaturated phospholipids. Deficient lpd-3 homologues in Zebrafish and mammalian cells cause defects similar to those observed in C. elegans. As mutations in BLTP1, the human orthologue of lpd-3, cause Alkuraya-Kucinskas syndrome, LPD-3 family proteins may serve as evolutionarily conserved highway bridges critical for ER-associated non-vesicular lipid trafficking and resilience to cold stress in eukaryotic cells. Homeoviscous adaptation (HVA) refers to the ability of cells to adjust membrane viscosity by changing cell membrane lipid compositions and unsaturation in response to environmental temperature shifts. For example, exposure to cold temperature in bacteria rigidifies cell membrane, triggering HVA to maintain membrane fluidity within a normal range to ensure proper activity of membrane proteins. Besides bacteria, HVA has been observed in many eukaryotic organisms as an evolutionarily conserved mechanism that enables adaptation to changes in environmental temperature. In both bacteria and the multicellular model organism C. elegans, HVA involves temperature-triggered transcriptional regulation of genes encoding lipid desaturases. While heat down-regulates a fatty acid desaturase-encoding gene fat-7 through acdh-11, cold up-regulates fat-7 through the membrane fluidity sensor PAQR-2 and downstream transcriptional regulators in C. elegans. Temperature-regulated FAT-7 catalyzes chemical C=C double bond formation in fatty acyl chains, leading to membrane lipid unsaturation and increased membrane fluidity. HVA through such regulation of lipid desaturases facilitates cellular adaptation to, and orgasmic survival against, environmental temperature stresses. In eukaryotes, lipid biosynthetic enzymes and lipid desaturases, including FAT-7, are located at the endoplasmic reticulum (ER). The newly synthesized and unsaturated lipids can distribute to other cellular organelles by both well-characterized vesicular transport pathways and less well-understood non-vesicular transport mechanisms. Earlier studies indicate that inhibition of vesicular transport pathways does not substantially decrease transfer of phospholipids, including phosphatidylcholine (PC) and phosphoatidylethanolamine (PE), from ER to plasma membranes (PM). More recent studies suggest that non-vesicular lipid trafficking among various intracellular organelles, including ER, lysosomes and mitochondria, can occur through a conserved family of RBG domain-containing VPS13-like non-vesicular lipid transporters. However, compared to the vesicular lipid transport pathways, the mechanisms of action, physiological regulation and organismic functions of non-vesicular lipid transporters remain still largely unknown. We performed a mutagenesis screen for genetic suppressors of FAT-7 in C. elegans and identified lpd-3, which encodes a 452 kDa megaprotein bridging the ER and PM. AlphaFold2-assisted structural prediction reveals an elongated hydrophobic tunnel in LPD-3 suited for lipid trafficking. We show that LPD-3 is critical for fat-7 expression, normal distribution of phospholipids at the PM, and organismic resilience to severe cold stress. Mutations in KIAA1109/BLTP1, the human orthologue of lpd-3, cause an autosomal recessive disorder, Alkuraya-Kucinsk syndrome. We found that decreased expression of lpd-3 homologues in Zebrafish and mammalian cells elicited similar phenotypes as in C. elegans. Our results suggest evolutionarily conserved roles of the LPD-3 family proteins as megaprotein-based molecular bridges in non-vesicular trafficking of lipids and stress resilience to cold temperature. We have previously discovered components of a genetic pathway in C. elegans that maintains cell membrane fluidity by regulating lipid unsaturation via the fatty acid desaturase FAT-7 in response to temperature shifts. Loss-of-function mutations in the gene acdh-11 cause constitutive FAT-7 up-regulation. In forward genetic screens to isolate mutants with acdh-11-like constitutive expression of fat-7::GFP, we identified several alleles of acdh-11 and two additional genes, cka-1 and sams-1 (Fig. 1a, Supplementary Fig. 1a, b), which are involved in cellular phosphatidylcholine biosynthesis. acdh-11, cka-1 and sams-1 encode negative regulators of fat-7. To identify positive regulators of fat-7, we performed acdh-11 suppressor screens for mutants with diminished fat-7::GFP (Fig. 1a). Unlike loss-of-function mutants of known positive regulators (e.g., nhr-49/80 or sbp-1 with complete loss of fat-7::GFP signals), a rare acdh-11 suppressor mutant dma544 exhibits diminished fat-7::GFP in the anterior intestine and decreased (but still visible) fat-7::GFP in the posterior intestine (Supplementary Fig. 2a, b). By single nucleotide polymorphisms-based genetic mapping and whole-genome sequencing, we identified dma544 as a missense transition mutation of the gene lpd-3. RNAi against lpd-3, an independently derived deletion mutation or another acdh-11 suppressor dma533 recapitulated both fat-7::GFP suppression and the morphological pale phenotype of dma544 (Fig. 1b, c, f). RNAi against lpd-3 also suppressed fat-7::GFP in the cka-1 or sams-1 mutants (Supplementary Fig. 1c). To better understand functions of LPD-3, we assessed how lpd-3 mutations alone might impact gene expression changes and fat-7-related phenotypes without acdh-11 mutations. In the wild type, fat-7::GFP was increased upon exposure to a cold temperature at 15 °C, yet such increase was abolished in lpd-3 mutants (Supplementary Fig. 2c, d). The baseline expression of fat-7::GFP at 20 °C was also abolished in lpd-3 mutants, including that in the posterior intestine (Fig. 1c). We performed RNA sequencing (RNAseq) to compare transcriptomes of wild type versus lpd-3 mutants cultivated at 20 °C. After differential expression analyses of triplicate samples, we identified 6251 genes that are significantly up- or down-regulated in lpd-3 mutants (Fig. 1d, Supplementary data file 1). As expected, fat-7 was one of the most highly down-regulated genes (log2fold change = −5.05, adjusted p value = 2.54E−13), while expression of its upstream regulators including cka-1, sams-1 or acdh-11 remained largely unchanged (Fig. 1e, Supplementary data file 1). Among the genes that were significantly up-regulated (adjusted p value > 0.05), 234 genes were also up-regulated by exposure to 4 °C cold-warming stress, including the previously validated cold-inducible gene asp-17 that we confirmed to increase dramatically in lpd-3 mutants without cold exposure (Supplementary Fig. 3a–c). We used WormExp to compare these 234 genes to expression data from all previously characterized mutant animals and found they were significantly similar (P = 1.7e−107) to the gene set regulated by RNAi against sbp-1 (Supplementary Fig. 3d). RNAi against lpd-3 or sbp-1 has previously been shown to induce the morphological pale and lipid depletion phenotypes. We confirmed such phenotype in lpd-3(dma544) mutants (Fig. 1c, f) and further showed that deletion of LPD-3 caused markedly fewer and smaller lipid droplets using an established lipid droplet marker DHS-3::GFP (Supplementary Fig. 3e). We made similar observation in animals with RNAi against sbp-1, which encodes a master regulator of lipogenesis and fat-7 expression for unsaturated lipid accumulation in C. elegans (Supplementary Fig. 3e). These results identify transcriptomic changes as well as fat-7-related lipid and morphological phenotypes in lpd-3 mutants. We next examined structural features of LPD-3 that may provide insights into its molecular function. LPD-3 is an exceptionally large protein, consisting of predicted 4018 amino acid residues of 452 kDa molecular weight. We sought to obtain a predicted LPD-3 structure by the machine-learning-based AlphaFold2 program. As the program is limited to polypeptides smaller than 2500 amino acids, we segmented the full-length LPD-3 sequence into four overlapping parts that were separately predicted and then rejoined to generate a full-length structure (Fig. 2a). The yielded full-length structure reveals an ~30 nm-long rod-like shape consisting of twisted β-sheets that form a striking tubular cavity and internal hydrophobic tunnel extending along its entire length (Fig. 2a, b, Supplementary Data file 2). The dma544 mutation (G200E) disrupts a highly conserved glycine residue lining up the internal tubular wall while the resulting G200E glutamic acid residue of LPD-3 is predicted to partially block the tunnel entry (Fig. 2c). The N-terminal sequence of LPD-3 forms a putatively hydrophobic transmembrane helix while its C-terminal sequence harbors an amphiphilic patch and a polybasic cluster (KxKK motif that binds to PIP2/PIP3) indicative of association with the cytosolic side of lipid membranes by electrostatic interaction (Fig. 2d). These structural features of LPD-3 are reminiscent of the recently described VPS13/ATG2 family transporters that mediate non-vesicular lipid trafficking across organelle membranes, although these separate families of proteins lack apparent protein sequence similarities. To determine the subcellular localization of LPD-3, we constructed transgenic reporters for both N- and C-termini of LPD-3. A mCherry-tagged N-terminal LPD-3 translational reporter showed prominently discrete intracellular signals in the intestine (Fig. 2e). We crossed this N-terminal reporter into established C. elegans strains expressing bright fluorescent GFP directed to various intracellular organelles including Golgi (mans::GFP), mitochondria (MAI-2::GFP), peroxisome (GFP::DAF-22), lysosome (LMP-1::GFP), endosome (RAB-7::GFP) and ER membranes (GFP::C34B2.10::SP12). We found that the N-terminal LPD-3Nt::mCherry signals were closely surrounded by ER membrane markers (Fig. 2e). By contrast, a mCherry-tagged C-terminal LPD-3 translational reporter displayed both intracellular and plasma membrane (PM) signals in the intestine (Fig. 2f). We found that the PM signal of mCherry-tagged C-terminal LPD-3 co-localized with Akt-PH::GFP, an established reporter that binds to the phospholipid PIP3 of the inner leaflet of PM (Fig. 2f). We also used CRISPR/Cas9 to generate knock-in of seven copies of GFP11 at the C-terminus of endogenous LPD-3 and reconstitution with GFP1-10 revealed that endogenous LPD-3::GFP localized to ER and co-localized with mScarlet-labeled EYST-2 and MAPPER, specific markers for ER-PM junctions, but not with a Golgi marker (Fig. 2g, h and Supplementary Fig. 4). Thus, LPD-3 primarily localizes to the ER, particularly at the ER-PM membrane contact sites that are known to mediate lipid trafficking and integrate phospholipid regulation. The structural features and cellular localizations of LPD-3 indicate an ER-to-PM bridge-like tunnel with plausible roles in mediating non-vesicular ER-to-PM trafficking of lipids. Next, we conducted a series of functional assays and phenotypic analyses to test this idea. First, we examined how LPD-3 may impact phospholipid distribution in the cell. Phospholipids, including phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylethanolamine (PE) and phosphatidylinositol (PI), are newly synthesized at the ER and transported to cytoplasmic membranes through both vesicular and non-vesicular mechanisms. Since probes for live monitoring of most phospholipid distribution are unavailable in C. elegans, we took advantage of the genetically encoded reporter Akt-PH::GFP, which binds to the phospholipid PIP3 (3,4,5-phosphate), to assess the intracellular distribution and abundance of PIP3 species. We found that wild-type animals exhibited Akt-PH::GFP fluorescence enriched along the apical membrane of the intestine (Fig. 3a). By contrast, when crossed into lpd-3 mutants, this same reporter at the same developmental stage (24 or 48 h after L4) showed attenuated overall fluorescence without apparent apical enrichment in the intestine, and more dispersed intracellular distribution compared to that in wild type (Fig. 3a). We also noticed that transgenic expression of mCherry-tagged C-terminal LPD-3 reduced the apical enrichment of Akt-PH::GFP, indicating competition of both reporters for the same substrate (Fig. 2f). As PM-localized PIP2/PIP3 is associated with and often stimulates actin polymerization, we found that a filamentary actin reporter act-5::GFP in the intestine also displayed markedly reduced abundance and apical localization in lpd-3 mutants (Fig. 3b). These results reveal striking defects of PIP3-binding reporter distribution in LPD-3-deficient intestinal cells and support the notion that LPD-3 normally promotes enrichment of phospholipids, at least certain PI species or their precursors, at the PM. Second, we examined how LPD-3 may impact functional consequences of loss of SAMS-1. The S-adenosyl methionine synthetase SAMS-1 is critical for the biosynthesis of phosphatidylcholine in C. elegans, decreased abundance of which in ER membranes activates ER stress response and expression of lipogenic genes including fat-7 via SBP-1 regulation. We found that fat-7::GFP was strongly activated by RNAi against sams-1 in wild type but not in lpd-3 mutants (Fig. 3c, d). We made similar observation on hsp-4p::GFP, an established reporter for ER stress response (Fig. 3e). These results indicate that LPD-3 antagonizes effects of SAMS-1 in PC accumulation in ER membranes, again supporting a physiological role of LPD-3 in facilitating the ER-to-PM trafficking of phospholipids, which reduces PC accumulation in ER membranes. Third, we examined how LPD-3 may impact the nuclear abundance of SBP-1, a master regulator of lipogenesis and fat-7. The C. elegans SREBP homolog SBP-1 promotes lipogenesis and transcriptionally activates fat-7 expression by translocating from ER membranes to nucleus. We found that the abundance of nuclear SBP-1::GFP was markedly decreased by RNAi against lpd-3 (Fig. 3f). By contrast, a transcriptional sbp-1p::GFP reporter was not apparently affected by lpd-3 RNAi (Fig. 3g). These results indicate that LPD-3 promotes fat-7 expression likely through post-transcriptional regulation of SBP-1. Since low PC levels in ER membranes trigger SBP-1 nuclear translocation, these data are consistent with the notion that LPD-3 decreases PC levels in ER membranes by promoting ER-to-PM phospholipid trafficking. Fourth, we examined how LPD-3 may impact cellular membrane integrity. Phospholipids with proper compositions of saturated and unsaturated fatty acyl chains are critical for the maintenance of membrane fluidity and integrity. Using a fluorescein-based SMURF assay to measure membrane permeability, we found that lpd-3 mutants accumulated markedly higher levels of fluorescein in the intestine, and to a lesser extent in amphid sensory neurons, compared to wild type (Fig. 3h, i). These results suggest that loss of LPD-3 may compromise intestinal PM integrity. As insufficient fatty acyl unsaturation of phospholipids causes membrane leakiness via formation of domains with high-order phases that lack plasticity, we predict that loss of LPD-3 may lead to retention of excessively unsaturated phospholipids in ER membranes. Indeed, we found that excessive lipid saturation in ER membranes by RNAi against mdt-15 activated ER stress response in wild type but not lpd-3 mutants (Supplementary Fig. 5a). Despite increased PM leakiness resulting from reduced unsaturated lipids, the intestinal PM morphology and PM-targeted trafficking of proteins with GFP prenylation reporters appeared largely normal in lpd-3 mutants (Supplementary Fig. 5b), suggesting specific roles of LPD-3 in lipid trafficking. Together, these results support that LPD-3 promotes ER-to-PM trafficking of phospholipids, and thereby regulates SBP-1 nuclear abundance and expression of the fat-7 gene. To determine the physiological role of LPD-3 at the organismic level, we characterized lpd-3 mutant phenotypes in development and adult resilience to cold exposure. Compared with wild type, lpd-3 mutants show developmental delay, reaching to the larval L4 stage more slowly (Fig. 4a). In adult stages, lpd-3 mutants are strikingly sensitive to both cold shock (4 °C for 20 h) and short-term freezing shock (−20 °C for 25 min) (Fig. 4b, Supplementary Fig. 6a, b). As lpd-3 mutants showed probable defective ER-to-PM trafficking of phospholipid, we sought to rescue such organismic phenotypes of lpd-3 mutants by supplementation of phospholipids from various sources and individual constituents of phospholipids, including choline, serine, ethanolamine and fatty acids (unsaturated oleic or saturated stearic acids). We found that phospholipids (from soy or egg yolks) or Lecithin (predominantly unsaturated PC-type glycerophospholipids), but not other compounds tested, fully rescued the developmental delay of lpd-3 mutants (Fig. 4a, Supplementary Fig. 6a). Lecithin also markedly rescued adult survival to cold exposure in a dose-dependent manner (Fig. 4b–d). Additional defects of lpd-3 mutants in fecundity, cold or freezing tolerance, locomotory behavior, and intestinal membrane integrity or permeability were also rescued by Lecithin (Fig. 4e, Supplementary Fig. 6b–d). These data further support the functional role of LPD-3 in ER-to-PM phospholipid trafficking (Fig. 4f) and demonstrate a compelling pharmacological means by using Lecithin or phospholipid compounds to rescue defects in lpd-3 mutants. LPD-3 is the sole C. elegans orthologue of a highly evolutionarily conserved protein family including Tweek (Drosophila), Kiaa1109 (Zebrafish and mice) and KIAA1109 (Humans), recently renamed as BLTP1 (Supplementary Fig. 7a). To assess whether roles of LPD-3 in C. elegans are likely evolutionarily conserved in other organisms, we evaluated the consequences of loss of lpd-3 orthologues in both mammalian cells and Zebrafish. We derived and cultured mouse embryonic fibroblast (MEF) cells from Kiaa1109-deficient mouse embryos. Transfection with AKT-PH::GFP reporters showed that PIP3 phospholipids as bound by AKT-PH::GFP were enriched at ruffling membranes of cell periphery in wild-type but not knock-out (KO) MEFs (Fig. 5a). Kiaa1109 KO MEFs also exhibited higher sensitivity to cold stress, and this defect was rescued by supplementation with Lecithin (Fig. 5b). We also used a click chemistry-based method to image phospholipids based on metabolic incorporation of the choline analog propargylcholine into phospholipids (Fig. 5c). We found that both Kiaa1109 KO MEFs and lpd-3 mutant C. elegans exhibited marked reduction of fluorophore-conjugated propargylcholine signals at the PM (Fig. 5d–f), while decreased overall abundance of such signals is consistent with reduced activity of SBP-1 in PC biosynthesis. In HEK293 human cell lines, we co-expressed two plasmids encoding AKT-PH::GFP and shRNA against KIAA1109 and found that knockdown of KIAA1109 led to reduced PM localization of AKT-PH::GFP (Supplementary Fig. 7b), as in MEFs and C. elegans. In addition, CRISPR/Cas9-mediated KO of KIAA1109 in U937 cells also markedly increased cell death after 4 °C cold stress (Supplementary Fig. 7c). In Zebrafish, we used morpholino (MO) to knockdown kiaa1109 and found that kiaa1109 MO caused curved body and head defects as reported (Supplementary Fig. 7d, e). Using a cold sensitivity assay, we found that knockdown of kiaa1109 led to striking reduction of fish survival against cold stress (Supplementary Fig. 7f, g). Unlike C. elegans, Zebrafish embryos did not readily uptake exogenous lipids provided in their diets, thus precluding us from testing Lecithin effects in Zebrafish models. Nonetheless, the convergent phenotypes of phospholipid reporters, metabolic labeling and cold tolerance we have observed in C. elegans, Zebrafish, mouse and human cells strongly support the evolutionarily conserved roles of LPD-3 family proteins in promoting cellular phospholipid trafficking at organelle membrane contact sites of ER and stress resilience to cold temperature. Based on our integrated genetic, protein structural, cell biological and organismic phenotypic analyses, we propose a model for the role of LPD-3 in C. elegans (Fig. 4f). In this model, LPD-3 spans the ER and PM at localized membrane contact sites and acts as a megaprotein-based molecular bridge that mediates non-vesicular ER-to-PM trafficking of phospholipids. Such non-vesicular and rapid mode of lipid trafficking may be particularly important for meeting the demand of membrane expansion during development and for membrane fluidity adjustment during physiological adaptation to cold stress in adulthood. LPD-3-mediated proper flow of phospholipids from the ER to PM may also ensure appropriate levels of PC in ER membranes that, in turn, control the ER-to-nuclear translocation and abundance of SBP-1. In the nucleus, SBP-1 can regulate the expression of genes including fat-7 and others involved in PC biosynthesis, lipogenesis, metabolic homeostasis, membrane property regulation and stress responses (Fig. 4f). We found striking rescue of all examined phenotypic defects of lpd-3 mutants by exogenous supplementation of phospholipids or Lecithin. Ineffective rescue by phospholipid head group constituents or fatty acids indicates that phospholipid/Lecithin may act by incorporating into host membranes rather than providing simple nutritional support. Unlike eukaryote-derived polyunsaturated phospholipids, bacterial phospholipids from E. Coli, which C. elegans feeds on, contain mostly saturated lipids with little PC and PI thus failing to support proper development and adult adaptation to cold in lpd-3 mutants. Although our results suggest critical roles of LPD-3 in ER-to-PM lipid trafficking, LPD-3 may also localize and function at membrane contact sites formed by other ER-associated organelles. Along with the recently described family of VPS13/ATG2 lipid transporters, LPD-3 may represent an emerging class of lipid transporters that serve as molecular “highway bridges” critical for directed non-vesicular trafficking of lipids across different organelle membranes. Although our data strongly support diverse phospholipids with unsaturated acyl chains as transported substrates by LPD-3, the precise substrate specificity and biophysical mechanisms of transport await further investigations. Molecular functions of LPD-3 and its evolutionarily conserved orthologues have remained hitherto unknown. Its yeast orthologue Csf1 has been implicated in cold tolerance. The Drosophila homologue Tweek regulates synaptic functions. Mutations in the human orthologue KIAA1109/BLTP1 cause Alkuraya-Kucinskas syndrome, a neuro- and cardiovascular development disorder with no known medical treatment. Loss of KIAA1109 also impairs phagocytosis of L. pneumophila by macrophages. These divergent phenotypes may be underpinned by a unifying conserved role of this protein family in lipid trafficking. Rescue of lpd-3 mutants in C. elegans by Lecithin suggests a similar route to treat the Alkuraya-Kucinskas syndrome. Potentially conserved roles of KIAA1109/BLTP1 and other mammalian homologues of LPD-3 in regulating lipid trafficking and lipogenesis also raise the possibility of targeting KIAA1109/BLTP1 in diverse lipid metabolic disorders, including fatty liver diseases and obesity. C. elegans strains were maintained with laboratory standard procedures unless otherwise specified. The N2 Bristol strain was used as the reference wild type, and the polymorphic Hawaiian strain CB4856 was used for genetic linkage mapping and SNP analysis. Forward genetic screens for fat-7p::GFP activating or suppressing mutants after ethyl methanesulfonate (EMS)-induced random mutagenesis were performed as described previously. Approximately 25,000 haploid genomes were screened for acdh-11 suppressors, yielding at least 18 independent mutants. Identification of mutations by whole-genome sequencing and complementation tests by crossing EMS mutants with lpd-3(ok2138) heterozygous males were used to determine dma533 and dma544 as alleles of lpd-3. Feeding RNAi was performed as previously described. Transgenic strains were generated by germline transformation as described. Transgenic constructs were co-injected (at 10–50 ng/μl) with dominant unc-54p::mCherry or GFP, and stable extrachromosomal lines of fluorescent animals were established. Genotypes of strains used are as follows: Chr. I: lpd-3(dma533, 544, ok2138), lpd-3(wy1770, knock-in allele of GFP11x7 to lpd-3 C-terminus), Chr. III: acdh-11(n5857); Chr. IV: cka-1(dma550), Chr. V: nIs590[fat-7::GFP], Chr. X: sams-1(dma553, ok3033), dmaEx647 [rpl-28p::lpd-3 Nt::mCherry; unc-54p::GFP], dmaEx648 [ges-1p:: mCherry::lpd-3 Ct; unc-54p::GFP], epEx141 [sbp-1p::GFP::sbp-1 + rol-6(su1006)], wyEx10612 [pDHC301 ges-1p:: mScarlet::CAAX (PM) 3 ng/ul, pDHC304 ges-1p:: GFP(1-10) 10 ng/ul, Pord-1::GFP 50 ng/ul], wyEx10611 [pDHC302 ges-1p:: mScarlet::SP-12 (ER) 3 ng/ul, pDHC304 ges-1p:: GFP(1-10) 10 ng/ul, Pord-1::GFP 50 ng/ul], wyEx10609 [pDHC303 ges-1p::ERSP::mScarlet::MAPPER (ER-PM contacts) 3 ng/ul, pDHC304 ges-1p:: GFP(1-10) 10 ng/ul, Pord-1::GFP 50 ng/ul], pwIs503 [vha-6p::mans::GFP + Cbr-unc-119(+)], hjIs73 [vha-6p::GFP::daf-22 + C. briggsae unc-119(+)], xmSi01[mai-2p::mai-2::GFP]; epIs14 [sbp-1p::GFP + rol-6(su1006)], pwIs890[Pvha-6::AKT(PH)::GFP], jyIs13 [act-5p::GFP::act-5 + rol-6(su1006)], zcIs4 [hsp-4::GFP]. The full-length LPD-3 was split into four fragments, each with ~ 1500 residues. Each fragment has ~500 overlapping residues with the neighboring fragments. Structure prediction of each fragment was generated by AlphaFold v2.0 (https://cryonet.ai/af2/) program. Predicted structures were aligned using Chimera based on the overlapping sequence. Then, the aligned structures of all fragments were merged in Coot to obtain the full-length structure. C-terminal flexible loop was manually adjusted. The structural images were prepared in ChimeraX. To investigate functions of kiaa1109 in affecting cold resistance zebrafish larvae, the morpholino (MO) used to target zebrafish kiaa1109 (E4I4) was obtained from Gene Tools. Fertilized eggs of AB strain zebrafish were obtained as previously described. The kiaa1109 and ctrl MOs were dissolved in ultrapure water (5 ng/nL) and 1-2 nL MO solution was injected into each zebrafish egg at single cell stage using a PICO-LITER injector from WARNER. The injected embryos were incubated in E3 medium at 28 °C. The injected larvae with normal phenotype were selected and exposed to 10 °C at 96 hpf. After 24 h of cold exposure, the larvae were let to recover at 28 °C for 24 h. The fish were checked frequently and the dead ones were removed and counted. At the end of the experiment, the survived fish were classified as abnormal and normal as previously reported. Photographs of the larvae before and after cold exposure were taken using a Zeiss stereomicroscope equipped with a color CCD camera. Body length of the larvae was measured by analyzing the photographs using AxioVision (v-4.8). N2 wild type and lpd-3(ok2138) animals were maintained at 20 °C and washed down from NGM plates using M9 solution and subjected to RNA extraction using TissueDisruptor and the RNeasy Mini Kit from Qiagen. RNA preparations were used for qRT-PCR or RNAseq. For qRT-PCR, reverse transcription was performed by SuperScript III, and quantitative PCR was performed using LightCycler Real-Time PCR Instruments. Relative mRNA levels were calculated by ∆∆CT method and normalized to actin. Primers for qRT-PCR: act-3 (forward, tccatcatgaagtgcgacat; reverse, tagatcctccgatccagacg) and fat-7 (forward, tgcgttttacgtagctggaa; reverse, caccaacggctacaactgtg). RNAseq library preparation and data analysis were performed as previously described. Three biological replicates were included for each treatment. The cleaned RNAseq reads were mapped to the genome sequence of C. elegans using hisat2. Abundance of genes was expressed as FPKM (Reads per kilobase per million mapped reads). Identification of differentially expressed genes was performed using the DESeq2 package. Animals were cultured under non-starved conditions for at least 4 generations before cold and freezing resilience assays. For cold resilience assay, bleach-synchronized L4 populations were kept at 4 °C for 20 h and then recovered for 24 h at 25 °C. For freezing resilience assay, bleach-synchronized L4 populations were kept at −20 °C for 45 min and then recovered for 24 h at 25 °C. For both cold and freezing experiments, NGM plates spread with equal agar thickness seeded with equal amounts of OP50 were used while cold and freezing temperature readings were monitored by thermometers to ensure minimal fluctuation. After cold or freezing shock, animals were moved to 25 °C for recovery and scored as dead if they showed no pumping and movement upon light touch with the body necrosis subsequently confirmed by light microscopy. For phospholipid and Lecithin rescue experiments, phospholipid (11145, Sigma-Aldrich), Lecithin (O3376-250, Fisher Chemical) or PC (P5394-10G, Sigma-Aldrich) was prepared as mixture by dissolving in M9 solution (from 1 to 20 mg/ml) and thorough vortexing. Phospholipids or Lecithin mixtures were then added (200 µl/60 cm plate) on NGM plates with pre-seeded OP50 and dried briefly before placing animals for cold or freezing tolerance assays. To assay the developmental delay of lpd-3 mutants, developmentally synchronized embryos from bleaching of gravid adult wild-type and lpd-3 mutant hermaphrodites were plated on NGM plates and grown at 20 °C. After indicated duration (40, 45 and 50 h), percentages of animals reaching the L4 stage (with characteristic crescent vulvar structures) were quantified. To assay fecundity, single L4 worms were placed to control, phospholipid (20 mg/ml) or Lecithin (20 mg/ml) containing plates (prepared as above). After 72 h, the total numbers of progeny at all stages were scored. For locomotion behavioral assays, the average speed of worms was recorded for synchronized young adult hermaphrodite (24 h post L4) using the WormLab System (MBF Bioscience) based on the midpoint position of the worms. Each experiment was repeated at least 3 times as independent biological replicates with more than 10 animals per group. SPE confocal and epifluorescence compound microscopes (Leica) or LSM confocal microscope (Zeiss) were used to capture fluorescence images. Animals of different genotypes were randomly picked at the same young adult stage (24 hrs post L4) and treated with 1 mM Levamisole sodium Azide in M9 solution (31,742-250MG, Sigma-Aldrich), aligned on an 4% agar pad on slides for imaging. Identical setting and conditions were used to compare genotypes, experimental groups with control. MEFs were derived from Kiaa1109 mutant mice [B6N(Cg)−4932438A13Riktm1b(EUCOMM)Hmgu/J, Stock No.026878] generated by the Knockout Mouse Project (KOMP) at The Jackson Laboratory (Bar Harbor, Maine, USA) using embryonic stem cells provided by the International Knockout Mouse Consortium. Kiaa1109–/– embryos were obtained from interbreeding of heterozygotes. Kiaa1109 mice were genotyped using the following PCR primers: wild-type allele (380 bp) forward GGG ATA TGG CAG AGA AGC TG, reverse AAA ACA ATT GGC TTA GAG ACT TCA; mutant allele forward CGG TCG CTA CCA TTA CCA GT, reverse GAC CAC ACA AAT CCC TTG GT. MEFs were cultured in DMEM (Thermo Fisher Scientific, MT-10–013-CV), supplemented with 10% FBS (Gemini Bio-Products, 900–208) and 1% penicillin/streptomycin and early passages (P2–P5) were used for reporter transfection, PC lipid labeling and cold resilience experiments. For phospholipid reporter transfection, MEFs were seeded at density of 4 × 105 cells/ml in 12-well plates containing glass cover slips and grown to 70–90% confluency. Mixture of DMEM, PloyJet reagent (Signagen Laboratories, MD, US) and CMVp::AKT-PH::GFP plasmids (Addgene) were prepared and added to wild-type and Kiaa1109 KO MEF cultures, followed by imaging with fluorescence confocal microscopy after 48 h. U937 cells (as suspension cultures) from ATCC were cultured in RPMI-40 (Gibco) medium supplemented with 10% heat-inactivated FBS (Hyclone), penicillin (10,000 I.U./mL), streptomycin (10,000 g/mL). HEK293T cells (as adherent cultures) were cultured in DMEM (Thermo Fisher Scientific, MT-10–013-CV), supplemented with 10% FBS (Gemini Bio-Products, 900–208) and 1% penicillin/streptomycin. Both cell lines were maintained in a humidified 5% CO2 incubator at 37 °C. U937 cells expressing lentiCas9-Blast were used to generate clonal lines of KIAA1109 KO with the sgRNAs targeting sequences GCCAGCTACCCCCGAATAtgg and GTTGACATCTACTACTACAtgg. For cold stress experiments, parental control and KIAA1109 KO U937 cells were cold shocked (4 °C for 20 h) and assayed for cell death using CYTOX Green-based flow cytometry. For lipid reporter experiments, HEK293 cells were co-transfected with plasmids with AKT-PH::GFP and shRNA against KIAA1109 (Sigma-Aldrich, TRCN0000263343 with 73% knockdown efficiency), incubated for 48 h and imaged by confocal microscopy for membrane localized GFP. For PC lipid labelling in MEFs, Kiaa1109+/+ and Kiaa1109−/− MEFs were incubated with propargylcholine (100 μM) in media for 24 h, fixed with 4% PFA in PBS for 5 min, reacted with 100 μM Alexa-488 Azide for 30 min. The cells were washed with PBS and imaged with fluorescence confocal microscopy. For PC lipid labelling in C. elegans, wild type and lpd-3 mutants were cultured under non-starved conditions for at least 4 generations. L4-stage animals were incubated with 100 μM propargylcholine in OP50 culture for 24 h at 20 °C, fixed with 4% PFA in PBS for 5 min, reacted with 100 μM Alexa-488 Azide for 30 min, washed with PBS and imaged with fluorescence confocal microscopy. Data were analyzed using GraphPad Prism 9.2.0 Software (Graphpad, San Diego, CA) and presented as means ± S.D. unless otherwise specified, with significance P values calculated by unpaired two-sided t-tests (comparisons between two groups), one-way or two-way ANOVA (comparisons across more than two groups) and adjusted with Bonferroni’s corrections. Representative fluorescence images were shown for results repeated at least three times independently with similar results. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Reporting Summary

PMC9649750

Chunbo Li,Hao Wu,Luopei Guo,Danyang Liu,Shimin Yang,Shengli Li,Keqin Hua

Single-cell transcriptomics reveals cellular heterogeneity and molecular stratification of cervical cancer

10-11-2022

Cervical cancer,Data mining,Cancer microenvironment

Cervical cancer (CC) is the most common gynecological malignancy, whose cellular heterogeneity has not been fully understood. Here, we performed single-cell RNA sequencing (scRNA-seq) to survey the transcriptomes of 57,669 cells derived from three CC tumors with paired normal adjacent non-tumor (NAT) samples. Single-cell transcriptomics analysis revealed extensive heterogeneity in malignant cells of human CCs, wherein epithelial subpopulation exhibited different genomic and transcriptomic signatures. We also identified cancer-associated fibroblasts (CAFs) that may promote tumor progression of CC, and further distinguished inflammatory CAF (iCAF) and myofibroblastic CAF (myCAF). CD8+ T cell diversity revealed both proliferative (MKI67+) and non-cycling exhausted (PDCD1+) subpopulations at the end of the trajectory path. We used the epithelial signature genes derived from scRNA-seq to deconvolute bulk RNA-seq data of CC, identifying four different CC subtypes, namely hypoxia (S-H subtype), proliferation (S-P subtype), differentiation (S-D subtype), and immunoactive (S-I subtype) subtype. The S-H subtype showed the worst prognosis, while CC patients of the S-I subtype had the longest overall survival time. Our results lay the foundation for precision prognostic and therapeutic stratification of CC.

Single-cell transcriptomics reveals cellular heterogeneity and molecular stratification of cervical cancer Cervical cancer (CC) is the most common gynecological malignancy, whose cellular heterogeneity has not been fully understood. Here, we performed single-cell RNA sequencing (scRNA-seq) to survey the transcriptomes of 57,669 cells derived from three CC tumors with paired normal adjacent non-tumor (NAT) samples. Single-cell transcriptomics analysis revealed extensive heterogeneity in malignant cells of human CCs, wherein epithelial subpopulation exhibited different genomic and transcriptomic signatures. We also identified cancer-associated fibroblasts (CAFs) that may promote tumor progression of CC, and further distinguished inflammatory CAF (iCAF) and myofibroblastic CAF (myCAF). CD8+ T cell diversity revealed both proliferative (MKI67+) and non-cycling exhausted (PDCD1+) subpopulations at the end of the trajectory path. We used the epithelial signature genes derived from scRNA-seq to deconvolute bulk RNA-seq data of CC, identifying four different CC subtypes, namely hypoxia (S-H subtype), proliferation (S-P subtype), differentiation (S-D subtype), and immunoactive (S-I subtype) subtype. The S-H subtype showed the worst prognosis, while CC patients of the S-I subtype had the longest overall survival time. Our results lay the foundation for precision prognostic and therapeutic stratification of CC. Cervical cancer (CC) is one of the most frequent female malignancies around the world. CC ranked the fourth of incidence and the fourth of mortality across all cancer types in women. According to the World Health Organization (WHO), an estimated 604,000 new cases and 342,000 deaths of CC were reported around the world in 2020. A large proportion of CC cases are reported to be related with the human papillomavirus (HPV). Accompanying the HPV infections, some genetic factors contribute a lot to the development of CC. Although patients with early CC can survive for years after surgery or radiation therapy, those diagnosed with advanced or metastasized CC is incurable. Therapeutics against advanced or recurrent CC are available, such as anti-angiogenesis and immunotherapy, but the response rate is still low, which is mainly due to the inter-tumor and intra-tumor heterogeneity of CC. Therefore, understanding the heterogeneity of CC in high resolution is crucial for the development of personalized therapeutic strategies. The Cancer Genome Atlas (TCGA) group reported a comprehensive molecular characterization of CC by profiling genomics, transcriptomics and proteomics in 288 CC samples. They revealed highly heterogenous molecular profiles across samples and identified three subtypes, i.e., the keratin-low squamous, keratin-high squamous and adenocarcinoma-rich subtype. Zhu et al. identified two subtypes of HPV+ CC based on the most varied 50 genes across CC samples. Furthermore, an increasing number of studies have identified molecular units (including DNA, RNA, and proteins) as biomarkers for the diagnosis and treatment of CC. But these studies were based on bulk sequencing data, thus overlooking the extensive cellular heterogeneity of CC. Recently, our study based on single-cell RNA sequencing (scRNA-seq) provided a glimpse into the phenotypic diversity and ecosystems of CC microenvironment. In this study, we presented a comprehensive characterization of CC cellular heterogeneity by utilizing scRNA-seq. We identified subpopulations of epithelial cells, fibroblasts, and CD8+ T cells, illustrating the cellular heterogeneity of CC. Based on the signature genes derived from scRNA-seq analysis, we identified four different CC subtypes that exhibited clinical significance. Our study shed light on the cellular heterogeneity and promoted the personalized treatment of CC. To investigate the cellular diversity and distinct molecular signatures in CC, scRNA-seq was performed in three CC cancer and paired NAT samples (Supplementary Fig. 1a). A total of 57,669 cells were obtained after stringent filtering, with specific cell groups of tumor or NAT samples (Supplementary Fig. 1b). These cells were further classified into 16 different clusters (Supplementary Fig. 1c). Marker genes in each cluster were then compared to known markers of cervical cells to determine known cell types (see “Methods”). These 16 cell clusters were assigned to seven different cell types (Fig. 1a), including epithelial cells (20,547 cells, 35.6%, marked with CDKN2A, EPCAM, CD24, and CDH1), endothelial cells (8617 cells, 14.9%, marked with PECAM1, CDH5, and ENG), fibroblasts (15,304 cells, 26.5%, marked with COL1A2, DCN, and APOD), smooth muscle cells (7429 cells, 12.9%, marked with ACTA2 and ACTG2), lymphocytes (4490 cells, 7.8%, marked with CD3E, CD3D, and CD2), macrophage (571 cells, 1.0%, marked with CD68, CD163, and LYZ), and neutrophils (741 cells, 1.3%, marked with CSF3R) (Fig. 1b). In our scRNA-seq data, the majority of smooth muscle cells (6463 cells, 87.0%) and endothelial cells (7118 cells, 82.6%) were derived from normal samples, while most epithelial cells (18,362 cells, 89.4%) were from tumor samples (Fig. 1c). Epithelial cells from tumor samples showed distinct transcriptional features with those from normal samples (Fig. 1d). Compared to those in NAT samples, both non-immune (Supplementary Fig. 2) and immune cell types (Supplementary Fig. 3) showed hundreds of differentially expressed genes that were enriched in specific pathways. All epithelial cells were further classified into seven different subclusters (Fig. 1e). Cells in subcluster C1, C2, C3, C4, and C5 were mainly from tumor samples, whereas those in the C6 and C7 subclusters were from normal samples. Cells in subcluster C1 were characterized by high expression levels of MMP1, SPRR1B, KRT16, CSTA, and S100A9 (Fig. 1f, Supplementary Table 2). Cells in subcluster C2 exhibited high expression levels of immune-associated genes, such as CD74, and IL32. The C3 subcluster showed high expression of CCDC80, IER5, and MAFB. The C4 subcluster showed high expression of UBE2C, TOP2A, and ANLN. Subcluster C5 exhibited high expression levels of normal epithelial markers, such as CLU, SCGB3A1, and MUC5B. In summary, our single-cell transcriptomics analysis revealed cellular heterogeneity of cervical epithelial cells. To further investigate the distinctions between the identified epithelial subpopulations, we inferred the copy number aberration (CNA) of each cell based its gene expression profile (see Methods). To evaluate the malignancy of identified epithelial subclusters, analysis of CNA levels in each cell population were performed according to average expression patterns across intervals of the genome. Remarkably, subcluster C1, C2, C3, and C4 exhibited copy number gains in chromosome 3q and 18, whereas they showed copy number loss in chromosome 3p, 5, and 13 (Fig. 2a). Overall, cells from subcluster C5, C6, and C7 showed low CNA levels, while those from cluster C1, C2, C3, and C4 showed relatively high CNA levels (Fig. 2b, Supplementary Data 1). The low CNA level in subcluster C5, which is mainly from tumor samples, might indicate a well differentiated state (Fig. 1d–f). Our enrichment analysis of high expression genes revealed the enrichment of response to stimulus, response to hypoxia, regulation of angiogenesis and positive regulation of mesenchymal stem cell migration, suggesting that cells in cluster C1 acquired a malignancy behavior (Fig. 2c and Supplementary Fig. 4a). Highly expressed genes in subcluster C2 were enriched in negative regulation of epithelial stem cell proliferation, regulation of cell cycle process and regulation of cell population proliferation (Fig. 2c and Supplementary Fig. 4b). The C3 and C4 subclusters shared similar enrichment of DNA repair, cell cycle, regulation of cell cycle phase transition and DNA damage checkpoint (Fig. 2c, Supplementary Fig. 4c, d). Highly expressed genes in subcluster C5 were enriched in the regulation of stem cell division and cell differentiation (Fig. 2c and Supplementary Fig. 4e). The C6 and C7 subclusters shared the same enriched biological process of normal epithelial biology, such as epithelial to mesenchymal transition (EMT) and positive regulation of epithelial cell proliferation (Fig. 2c, Supplementary Fig. 4f, g). Next, we employed the single-cell regulatory network inference and clustering (SCENIC) method to identify transcription factors that play important regulatory roles in malignant epithelial subclusters. Our analysis revealed many transcription factors in epithelial subclusters, such as HIF1A, TFDP1, and GRHL1 in subcluster C1, STAT1 and FOSL1 in subcluster C2, and XBP1 and NFKB1 in subcluster C5 (Fig. 2d, Supplementary Table 3). Hypoxia-induced factor-1 (HIF-1) is the most critical gene in hypoxic response and is responsible for the upregulation of many downstream effector genes that were collectively known as hypoxia-responsive genes (such as VEGFA, EGF, p53, GLUT1, and GLUT3). These genes govern multiple key biological pathways such as proliferation, energy metabolism, invasion, and metastasis. For example, as the key member of interferon signaling, STAT1 modulates the response to intracellular and extracellular stimulation. STAT1 has been demonstrated to act as a tumor suppressor in many cancer types. Collectively, our analysis revealed genomic and transcriptomic distinctions among epithelial subpopulations. We next investigated the non-immune cells within the tumor microenvironment (TME), including fibroblasts and smooth muscle cells (SMCs). We identified a total of 22,451 fibroblasts and SMCs. Most of these cells were from the cervical NAT samples (Fig. 3a). These cells were then re-clustered based on gene expression profiles, which generated 13 different clusters (Fig. 3b). These cell clusters showed different expression of marker genes DCN, COL1A2, and ACTA2 (Fig. 3c). According to the specific cell markers, we assigned cluster C1, C4, C5, C6, C11, and C12 as fibroblasts and cluster C2, C3, C7, C8, C9, C10, and C13 as SMCs. To further explore how fibroblasts impact CC tumor progression, we examined the transcriptional alterations of tumor-derived fibroblasts. Compared to fibroblasts from normal tissues, the top upregulated genes in tumor-derived fibroblasts were CXCL8, CXCL2, CCL2, CXCL3 and CXCL1, and the top downregulated genes were IGFBP5, PTGDS, CCN5, CFD, and RAMP1 (Fig. 3d, Supplementary Data 2). Functional enrichment analysis revealed that fibroblasts in tumor tissue were associated with IL-17 signaling pathway, antigen process and presentation and INF-signaling pathway, indicating a potential role in immune regulation (Fig. 3e). More importantly, we identified several DEGs in fibroblasts that were significantly associated with patient prognosis, such as CXCL8 and IGF1. The CXCL8 gene was upregulated in tumor fibroblasts and the high expression was associated with poor prognosis (Fig. 3f). The IGF1 gene was downregulated in tumor fibroblasts and the low expression was associated with poor prognosis (Fig. 3g). The CXCL8 gene showed oncogenic, while IGF1 exhibited tumor suppressor features in tumor fibroblasts. In our scRNA-seq dataset, cluster C4, C5, C6, C11, and C12 fibroblasts had high expression of gene IL6, IL8, CXCL1, CXCL2, CCL2, and CXCL12, and was identified as inflammatory cancer-associated fibroblasts (iCAFs). Cluster C1 fibroblasts was identified as myofibroblastic CAFs (myCAFs) with the high levels of SMA (encoded by gene ACTA2). To explore the functional differences between these two fibroblast types, we identified a set of significant DEGs between iCAFs and myCAFs, including the top upregulated genes, such as CXCL14, IFGBP7, PTGDS, CFD and CCN5, and top downregulated genes, such as LEFTY2, MT1X, CLU, MYH11, and RHOB (Fig. 3h, Supplementary Data 3). Functional enrichment analysis revealed upregulated activities of immune-related biological processes, including cytokine-cytokine receptor interaction, primary immunodeficiency, and intestinal immune network for IgA production (Fig. 3i). We further compared the gene expression between cancer cells and tumor-derived fibroblasts. Differential gene analysis revealed 319 genes upregulated in tumor-derived fibroblasts, such as DCN and SFRP4, and 214 genes upregulated in cancer cells, such as SPRR1B and SLURP2 (Supplementary Fig. 5a). We then performed pathway enrichment of these dysregulated genes. Cell proliferation-related functions were more enriched in tumor-derived fibroblasts, such as “epithelial cell migration” and “response to fibroblast proliferation”, whereas cell communication-related functions showed high enrichment in cancer cells, such as “cell-cell junction assembly” and “positive regulation of leukocyte cell-cell adhesion” (Supplementary Fig. 5b). These results indicated that fibroblasts from tumor promoted the tumor progression. With 24,911 cells detected, T cells represented the most prevalent cell type in our scRNA-seq data (Fig. 4a). Our re-clustering analysis revealed eight T cell subclusters, which were designated as CD8+ T cells (CD8A+, cluster C2, C3 and C8), natural killer T cells (NKG7+, cluster C4) and memory T cell (IL7R+, cluster C1), plasma cells (IGHG1+, cluster C7), regulatory T cells (TNFRSF4+, cluster C6) and mast cell (TPSB2+, cluster C5) (Fig. 4b, c, Supplementary Data 4). We observed that CD8+ T cells highly expressed CD8+ T cell markers, but they had almost no expression of the CD4+ T cell markers. Then, we analyzed the difference between CD8+ T cell clusters, which represented a large proportion of CD8+ T cells in both tumor and normal tissues. The C3 cluster CD8+ T cells (CXCR4+) were characterized by the high expression of the GZMK, CXCR2 and CX3CR1 gene, commonly associated with effective T cells and the low expression of check point genes (PDCD1, TIGIT, CTLA4, HAVCR2, LAG3 and CD274), suggesting that these cells are precursors of cytotoxic T cells (Fig. 4d, Supplementary Data 4). In addition, our analysis also revealed the high expression of some CD8+ T cell migration regulators, such as chemokine receptors (CX3CR1, CXCR4, and CXCR2), S1P receptors (S1PR1 and S1PR5), and integrins (ITGA5 and ITGAL). High expressed genes of the C3 cluster CD8+ T cells (CXCR4+) were found to be enriched with such pathways as NK cells medicated cytotoxicity, T cells receptor signaling and Toll-like receptor signaling pathway, which was related to the cytotoxic function (Fig. 4e). The C2 cluster CD8+ T cells (PDCD1+) were characterized by the high expression of immune checkpoint genes (such as PDCD1, TIGIT, CTLA4, HAVCR2, LAG3, and CD274). Meanwhile, the DEG analysis also revealed higher expression of HLA-DPA1, HLA-DRA, and HLA-DRB1, which is beneficial to the antigen presentation and the activation of cytotoxicity T cells. Functional analysis revealed the enrichment of the immune-related pathways, such as cell-adhesion molecular, ABC-transports, Th17 cell differentiation, and complement cascades. The C8 cluster CD8+ T cells (MKI67+) presented proliferative cells as they expressed high levels of gene MKI67, TOP2A, and CCNB1, and low levels of gene TIGIT, CTLA4, PDCD1, HAVCR2, LAG3, and LAYN. Functional enrichment analysis revealed that these cells were enriched with pathways related to cell proliferation (cell cycle, oocyte-meiosis, DNA replication and base excision repair), suggesting limited effective ability of these cells. To further investigate how different CD8+ T cell subtypes developed in CC, we performed pseudo-time trajectory analysis of all CD8+ T cells (see “Methods”). Our analysis showed that the C3 cluster CD8+ T cells (CXCR4+) cells were at the beginning of the trajectory path, whereas the C2 cluster CD8+ T cells (PDCD1+) and the C8 cluster CD8+ T cells (MKI67+) were at a terminal state (Fig. 4f). This was accompanied by the increased expression of exhaustion markers PDCD1, LAG3, and TIM3, and the decrease of effector markers, such as CX3CR1, CXCR4, and CXCR2 (Fig. 4g). Meanwhile, we found that MKI67, TOP2A and CCNB1 increased at one end of the pseudo-temporal trajectory (Fig. 4h), and demonstrated that cell populations with both proliferative and exhausted states were present. The observation suggested that some cells might be reserved to have proliferative ability before being terminally exhausted. Our single-cell transcriptomics analysis revealed highly different subpopulations in many cell types, suggesting a more precision heterogenous property of CC. In CC, three subtypes have been identified largely based on mRNA expression of certain genes, including two squamous subtypes (Keratin-high and Keratin-low), and an adenocarcinoma-rich subtype (adenocarcinoma). However, these subtypes did not reflect the heterogenous TME compositions and the prognostic differences. We further classified CC subtypes by using cell type-specific genes that were highly expressed by the malignant cells. Our classification strategy reduced the effect of non-malignant cells in CC. We used the 705 marker genes of epithelial cells between normal and tumor tissues (log FC > 1.5 and P < 0.05) from scRNA-seq data and divided 253 samples of the TCGA CESC cohort into four major subtypes (Fig. 5a), namely the hypoxia (S-H subtype), proliferation (S-P subtype), differentiation (S-D subtype), and immunoactive (S-I subtype) subtype. The S-H subtype expressed 10 significant genes (MMP1, IGFBP3, ITGA5, CDH3, ICAM1, FLOD2, TGFB1, PLAU, FSCN1, and ITGB4) (Fig. 5b), and presented the enrichment of hypoxia (Fig. 5c). Hypoxia is one of most common tumor characteristics, which is mainly caused by insufficient vascularization. The hypoxic TME condition impedes immune response by recruiting immunosuppressive cells and genes. The S-P subtype expressed 10 significant genes (TUBB, UBE2C, NUSAP1, CKD1, PSMC4, PCNA, DLG1, ATP2C1, MKI67, and TOP2A) and presented enrichment of cell proliferation. The S-D subtype expressed 10 significant genes (KRT6A, SPRR1B, KRT16, AQP3, PERP, CSTA, DSG3, SPRR2D, CAPN1 and CDKN2A) that showed enrichment of cell differentiation. The S-I subtype expressed 10 significant genes (HLA-DMA, CD74, HLA-DMA, HLA-C, CXCL10, PSMB3, IFI6, CXCL17, CST6 and HLA-DQA1), and presented enrichment of immune-related biological processes. Furthermore, we obtained gene expression profiles of 340 cervical cancer samples from the GEO database (GSE15166, GSE29617, and GSE68335) and divided them into four subtypes using the same gene marker set (Supplementary Fig. 6a). These subtypes expressed similar signature genes and biological processes (Supplementary Fig. 6b–f). Then, we compared the prognosis of different subtypes. These four subtypes showed significantly different overall survival times (Fig. 5d). The S-I showed the longest survival time, while the S-H subtype exhibited the worst prognosis. To further explore the differences between the four CC subtypes, we performed GSVA analysis to evaluate scores of 50 hallmarks in each sample (see Supplementary methods). Different subtypes were specifically enriched in different hallmarks (Supplementary Fig. 7a). In particular, the S-D subtype was specifically enriched in “MYC targets v1” and “protein secretion”-related hallmark gene sets. The S-P subtype was highly enriched in “Wnt β catenin signaling”, “HEME metabolism” and “myogenesis”-related hallmark gene sets. The S-I subtype showed high scores of “IL6 JAK STAT3 signaling”, “oxidative phosphorylation”, and “IL2 STAT5 signaling”-related hallmark gene sets. The S-H subtype specifically high enrichment of “reactive oxygen species pathway” and “hypoxia”-related hallmarks. We also performed CIBERSORT analysis to infer the relative abundance of immune cells in each sample (see Supplementary methods). The four CC subtypes showed distinct infiltration of different immune cell types (Supplementary Fig. 7b). For example, the S-I subtype showed higher infiltration of CD8+ T cells and regulatory T cells. In addition, we compared the expression levels of immune checkpoint genes between different CC subtypes. The S-I subtype showed significantly high expression of many immune checkpoints, such as BTLA, CD27, and TIGIT (Supplementary Fig. 7c). In this study, we employed scRNA-seq to comprehensively delineate the cellular heterogeneity of human CCs. By using the signature genes derived from scRNA-seq data analysis, we identified four molecular subtypes of CCs, namely the hypoxia, proliferative, differentiated, and immunoactive subtype. The stratification of CC tumors not only promotes our understanding of its etiologies, but also accelerates the development of personalized therapeutic strategies for CC patients. We calculated the percentages of the subpopulations of epithelial cells, fibroblasts, and T cells in CC tumor and paired NAT samples (Supplementary Fig. 8). Some cell subpopulations showed acceptable variations in different tumor or NAT samples, such as the C1, C3, and C5 T cells in NAT samples, while some showed large variations in different tumor or NAT samples, such as the C1, C2, and C3 epithelial cells in tumor samples. The large percentage variations of some cell subpopulations might be due to heterogeneity between different patients and sample collections. Cell subpopulations are cells with specific status under specific conditions, some of which showed great heterogeneity between different patients. Samples used for scRNA-seq are randomly chosen from pathological or related areas, but cells are not uniformly distributed. Increasing the number of samples could, to some extent, reduce the bias induced by these limitations, but also enlarges the volume of cell subpopulations. Cell subpopulations identified from limited number of samples might show large variations between samples, but reflect the existence of specific cell status. One significant advantage of our CC classification was that our strategy was based on the epithelial cell markers from scRNA-seq data. The strategy could eliminate the influence of other cells such as, stromal and immune cell markers in the categorizing process. For example, in early phase, HGS-OvCa was identified as four molecular subtypes: immunoreactive, differentiated, proliferative and mesenchymal according to the TCGA data. However, a recent study from scRNA-seq data found that the HGSOC classification of immunoreactive and mesenchymal reflected the infiltration of immune cells and fibroblasts, but ignored malignant cells. The scRNA-seq data had the advantage over the bulk RNA-seq in focusing on the tumor cells. Another advantage is that the scRNA-seq data could help understand the potential mechanism of tumor progression. EMT plays key roles in the development and pathological biology of tumor, and understanding its regulation is important for developing new therapeutic intervention for tumor patients. In addition, emerging evidence has shown that hypoxia could affect EMT by regulating the expression of EMT-related transcription factors and signaling genes. Accumulating evidence has demonstrated that immune cells in TME, such as tumor-associated macrophages and T cells, are closely involved in the progression of tumor. The TME, including immune cells with malleable states and their communications with other cells, is a major contributor to regulating immune response against tumor cell behaviors. To our best knowledge, our study presented the first single-cell landscape of infiltrating immune cells in CC. We observed that CD8+ T cells were infiltrating with different status in CC samples, including proliferative and exhausted status, and activated CD8+ T cells were in low abundance. In the TME, CD8+ T cells are the major effector to kill tumor cells. The immunosuppressed state of CD8+ T cells indicated the lack of sufficient activated T cells to kill tumor cells in the TME of CC. We found that both inhibitory receptors (IRs) and activation markers of T cell exhaustion were expressed in some CD8+ T cells. Whether these CD8+ T cells turn into effective or exhausted state was determined by the expression modulation of IRs. We further revealed the differentiation trajectory of different CD8+ T cells in CC wherein CX3CR1+ CD8+ T cells transformed to the PDCD1+ CD8+ T cells. We found that different subtypes presented various infiltration of immune cells, especially for CD8+ T cells. CC patients of the immunoactive subtype might respond to immune checkpoint blockade (ICB) therapy, but patients of other subtypes may not. Recently, novel ICB targets beyond CTLA4 and PD-1 have been identified, such as LAG3, TIM3, HAVCR2, and TIGIT. Numerous clinical trials of these emerging ICB targets are underway. Overall, CD8+ T cells showed high expression level of LAG3 and TIM3 in our scRNA-seq data. Our analysis suggest that LAG3 and TIM3 might be potential ICB targets that are worth further investigation for the ICB therapy of CC patients. In conclusion, our study characterized the single-cell landscape of TME in CC. Then, we firstly classified all CESCs patients into four subtypes, which may present different response to immune checkpoint inhibitors. Although more datasets and experimental validation are needed, our results shed lights on T cell infiltration and response in CC, which might promote the development of more personalized diagnostic and therapeutic strategies in clinical practice. Human cervical samples were collected from three different patients in the Obstetrics and Gynecology Hospital of Fudan University, including three cancer samples and paired adjacent non-tumor (NAT) samples. All patients gave informed consent. The clinical information, including age, menstrual status, FIGO stage, histological type, HPV status, and treatment, are provided in Supplementary Table 1. The estrogen hormone of all included patients was at low levels. This study was approved and supervised by the ethics committee of the Obstetrics and Gynecology Hospital of Fudan University. Collected fresh cervical samples were washed with 1× PBS three times. Then tissue samples were cut into 1 mm3 pieces and incubated in the same dispase solution at 37 °C for half an hour. Pieced tissue was gently dissociated with a pipette and incubated in trypsin 0.05% solution diluted with PBS for 10 min. Single-cell samples were filtered out with a 70 mm filter after the trypsin was deactivated by RPMI 1640 medium (Gibco), supplemented with 10% FBS and 1% penicillin/streptomycin (Invitrogen). The trypan blue microscopy was used to determine the percentage of active cells, and only samples with no less than 85% of active cells were used for scRNA-seq. Single cells were then counted with a hemocytometer and live cells were sorted for the preparation of 10X Genomics scRNA-seq library. The single-cell suspension was loaded onto a 10X Chromium Single-Cell instrument to generate single-cell Gel Beads-in-emulsion (GEMs). The single-cell RNA library was then constructed and estimated by using 10X Genomics Chromium Single-cell 30 Library, Gel Bead & Multiplex Kit. The scRNA-seq was performed on the Illumina NextSeq500. All procedures were performed according to the standard manufacturer’s protocol. The raw scRNA-seq reads were first processed for sample demultiplexing, barcode processing, and genome mapping by using the Cell Ranger (version 3.0.1) software. The GRCh38 human reference genome was utilized in the read alignment process. The unique molecular identifiers (UMIs) were counted in each single cell. Low-quality cells were filtered as previously described. Specifically, cells with UMI number <200, gene number <200, or percentage of mitochondrion-derived UMI counts >10% were discarded as low-quality cells. The Seurat R package (version 4.0) was applied in the quality control procedure. In addition, the Scrublet software (version 0.2.2) was employed to identify and remove potential doublets. After removing low-quality and doublet cells, data of all samples was normalized and merged. The feature expression measurements for each cell were normalized by the total expression by using the “LogNormalize” method implemented in the NormalizeData function. Then normalized counts were then multiplied by a scale factor (10,000) and log-transformed. The normalized data was used to identify gene features with high cell-to-cell variations by utilizing the FindVariableFeatures function. The top 2000 highly variable genes were used to scale the data by using the ScaleData function. Then the principal component analysis (PCA) was adopted to reduce data dimensions. The FindNeighbors and FindClusters functions were consecutively used to perform a graph-based clustering and find the optimal cluster resolution. The RunTSNE function was applied for appropriate visualization. Differentially expressed gene markers in each cluster were identified by the FindAllMarkers function, which compares gene expression with those in all other cell clusters. The unbiased cell type recognition was performed by applying the SingleR package (version 1.4.1), which leverages reference transcriptomic datasets of pure cell types. Then the annotated cell clusters were checked by manually curated gene markers retrieved from the CellMarker database and published papers. The differential genes were then identified in each cell type with the following criteria: expressed in at least 20% of cells in either sample groups; |log2FoldChange| >0.585; adjusted p value < 0.01. The inferCNV software (https://github.com/broadinstitute/infercnv) was applied to infer copy number alterations (CNAs) in our scRNA-seq data. CNAs were computed according to a previous study. Briefly, genes were sorted by their chromosomal locations to evaluate initial CNAs from expression levels. A sliding window of 100 genes was used to calculate moving averages of relative expression values in each chromosome. In each epithelial cell, the relative CNAs were calculated from the inferCNV outputs. For each bin of 30 genes, an average value of CNA was estimated in nonoverlapping genomic regions. Average CNA values were rounded to the closest integers. The pseudo-time trajectory was inferred by utilizing the Monocle2 package (version 2.8.0) to reveal the cell-state transitions. The following parameters were adopted: average expression R0.125, num_cells_expressed R10, qval < 0.01 (differentialGeneTest function). The DDRTree function was applied to reduce the dimensions with default settings. The expression and variance levels were used to determine the ordering genes. Functional enrichment analyses in this study were conducted using the clusterProfiler R package (version 4.1). Differential genes in each cell type or cluster were used to compute enriched GO biological processes or KEGG pathways. The GSEA analysis was performed by using the gsea function. GO terms or KEGG pathways with adjusted p value < 0.05 were considered as significantly enriched by the gene sets of interest. The Single-Cell rEgulatory Network Inference and Clustering (SCENIC) method was employed to perform gene regulatory network analysis in different cell types or clusters. The SCENIC analysis was realized by the pySCENIC (version 0.10.2) software. Briefly, the processing consists of three major steps. First, co-expression modules of transcription factors (TFs) and targets were inferred by using the gradient boosting machine regression implemented in GRNBoost2. Second, these modules were optimized to remove indirect targets by using the i-cisTarget software. Third, enrichment scores for the regulons’ targets were calculated by the AUCell algorithm to estimate the activity of these regulons. The TFs and target motifs were collected by the SCENIC group. Statistical analysis and data visualization in the present study was performed by using the R software (version 4.0.2, R Foundation for Statistical Computing, Vienna, Austria; http://www.r-project.org). Unless specifically stated, p or FDR values < 0.05 were considered as statistically significant. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Reporting Summary

PMC9649756

Naor Granik,Noa Katz,Or Willinger,Sarah Goldberg,Roee Amit

Formation of synthetic RNA protein granules using engineered phage-coat-protein -RNA complexes

10-11-2022

Synthetic biology,Self-assembly,Protein delivery,RNA-binding proteins

Liquid-solid transition, also known as gelation, is a specific form of phase separation in which molecules cross-link to form a highly interconnected compartment with solid – like dynamical properties. Here, we utilize RNA hairpin coat-protein binding sites to form synthetic RNA based gel-like granules via liquid-solid phase transition. We show both in-vitro and in-vivo that hairpin containing synthetic long non-coding RNA (slncRNA) molecules granulate into bright localized puncta. We further demonstrate that upon introduction of the coat-proteins, less-condensed gel-like granules form with the RNA creating an outer shell with the proteins mostly present inside the granule. Moreover, by tracking puncta fluorescence signals over time, we detected addition or shedding events of slncRNA-CP nucleoprotein complexes. Consequently, our granules constitute a genetically encoded storage compartment for protein and RNA with a programmable controlled release profile that is determined by the number of hairpins encoded into the RNA. Our findings have important implications for the potential regulatory role of naturally occurring granules and for the broader biotechnology field.

Formation of synthetic RNA protein granules using engineered phage-coat-protein -RNA complexes Liquid-solid transition, also known as gelation, is a specific form of phase separation in which molecules cross-link to form a highly interconnected compartment with solid – like dynamical properties. Here, we utilize RNA hairpin coat-protein binding sites to form synthetic RNA based gel-like granules via liquid-solid phase transition. We show both in-vitro and in-vivo that hairpin containing synthetic long non-coding RNA (slncRNA) molecules granulate into bright localized puncta. We further demonstrate that upon introduction of the coat-proteins, less-condensed gel-like granules form with the RNA creating an outer shell with the proteins mostly present inside the granule. Moreover, by tracking puncta fluorescence signals over time, we detected addition or shedding events of slncRNA-CP nucleoprotein complexes. Consequently, our granules constitute a genetically encoded storage compartment for protein and RNA with a programmable controlled release profile that is determined by the number of hairpins encoded into the RNA. Our findings have important implications for the potential regulatory role of naturally occurring granules and for the broader biotechnology field. Phase separation, the process by which a homogeneous solution separates into multiple distinct phases, has been connected to a wide range of natural cellular processes in virtually all forms of life. In cells, phase separation results in the formation of membrane-less compartments containing a high-concentration mix of biomolecules (e.g., proteins, RNA, etc.), which are surrounded by a low-concentration solution. Generally, phase separations are classified by the different material states which can lead to multiple types of transitions (e.g., liquid-solid, gas-liquid, etc.). The forms commonly reported in cellular biology are broadly liquid-liquid and liquid-solid (e.g., gelation), however determining the exact mechanisms for phase separation in a living cellular environment is often challenging. Liquid-liquid phase transitions can be distinguished from liquid-solid by the dynamical properties of the resulting condensates. Liquid-based condensates show rapid internal rearrangement of molecules, fusions between different condensates upon contact, and dependency on the concentration of the molecules in the condensed phase. On the other hand, liquid-solid based condensates show none of the above qualities and are mainly dependent on the number of ‘cross-linkers’, which are points of contact between the molecules, rather than on the concentration of the molecules themselves. Recently, Jain & Vale reported on the formation of RNA granules both in vivo and in vitro, from highly repetitive RNA sequences associated with repeat expansion diseases. These RNA sequences, comprised of dozens of triplet-repeats of CAG or CUG nucleobases, form intramolecular hairpin structures, which facilitate multivalent intermolecular interactions. The RNA granules presented features associated with liquid-solid phase transition systems: a lack of internal mobility, virtually no fusion events, and dependence on the number of repeats in the RNA sequence (i.e., cross linkers) rather than the concentration of the RNA. These characteristics helped establish the granules as physical solids. Hairpin forming RNA sequences are widespread in the RNA world and are not strictly associated with disease phenotypes. Such sequences are commonly used in synthetic systems for biological research. Perhaps the most ubiquitous system is composed of RNA sequences that encode for multiple hairpin motifs that can bind the phage coat proteins (CPs) of PP7 or MS2. Using this system to label the 5’ or 3’ end of a transcript has become commonplace in the last two decades, and enables visualization of RNA transcripts when the CPs are co-expressed. This approach, originally introduced by Singer and others, was devised for the purpose of probing the dynamics of transcription and other RNA-related processes, irrespective of cell-type. When co-expressed, the coat-protein-bound RNA molecules yield bright puncta, which are similar in appearance to natural biomolecular condensates. Consequently, we hypothesized that co-suspension of synthetic RNA hairpin cassettes together with their binding CPs can lead to the formation of gel-like particles via liquid-solid phase separation in vitro. In addition, by utilizing the CP binding ability of the hairpins, we expect to be able to selectively incorporate proteins of our choosing into the solid-like granules, resulting in a selective platform for the stable concentration of proteins. In this paper, we rely on our previous works to design and synthesize a variety of PP7 coat-protein (PCP) binding synthetic long non-coding RNA molecules (slncRNAs). Using fluorescent RNA nucleotides, we show that these slncRNAs form isolated puncta in vitro in a manner dependent on the number of hairpins encoded into the RNA. We further show that addition of fluorescent PCP to the suspension results in almost complete co-localization between protein and slncRNA. By tracking puncta fluorescence signals over time, we demonstrate that for all slncRNAs used, the various puncta emitted similar signals characterized by bursts of increasing or decreasing fluorescent intensity. We further show that signal intensities and temporal characteristics are dependent on the number of hairpins present in the RNA. Using these observations, we conclude that these “fluorescence-bursts” corresponded to addition or shedding of slncRNA-PCP nucleoprotein complexes. These events occur at rates that are consistent with the puncta being phase-separated solid-like granules. Consequently, we present these slncRNA-protein granules as a genetically encoded platform for the selective storage of proteins as well as a model system for exploration of liquid-solid phase separation. To test whether hairpin containing RNA can phase separate in vitro we designed six synthetic long non-coding RNA (slncRNA) binding-site cassettes using our binding site resource. We divided our slncRNAs into two groups. For the first group (class I slncRNAs), we designed three cassettes consisting of three, four, or eight hairpins that encode for PCP binding sites (PCP-3x, PCP-4x, and PCP-8x, respectively). In this group, hairpins were spaced by a randomized sequence that did not encode for a particular structure. For the second group (class II slncRNAs), we encoded three cassettes that consisted of three, four, and fourteen PCP binding sites that were each spaced by hairpin structures that do not bind PCP (PCP-3x/MCP-3x, PCP-4x/MCP-4x, and PCP-14x/MCP-15x, respectively). In addition, we designed a negative control slncRNA which does not contain any hairpin binding sites. To ensure that the negative control sequence has a similar GC content as the other slncRNA molecules (45%), it was designed as a permutation of the PCP-8x sequence. The sequences encoding for the slncRNAs were cloned downstream to a pT7 promoter and transcribed in vitro to generate the corresponding RNA. To visualize the RNA, we incorporated fluorescent nucleotides in the transcription reaction such that an estimated 35% of uracil bases were tagged by Atto-488 fluorescent dye. Each slncRNA-type was separately mixed with granule forming buffer (see methods: In vitro granule preparation, and Fig. 1a) at equal concentration (8.5 nM final concentration) and incubated for 1 h at room temperature. 2–5 µl of the granule reaction were then deposited on a glass slide and imaged using an epi-fluorescent microscope. The images show that reactions with slncRNA molecules which contain hairpin binding sites result in the formation of a multitude of bright localized fluorescent condensates, with the exception of the PCP-3x case (Fig. 1b). Interestingly, when increasing the slncRNA concentration in the PCP-3x case, sporadic granules do begin to form at 20 nM, and more robustly at 40 nM (Supplementary Fig. 1), reminiscent of a concentration dependent, liquid-like phase separation rather than gel-like. In contrast to the above, the granule reaction containing the negative control slncRNA does not result in any discernible puncta (Supplementary Fig. 2). In addition to the localized, small condensates, we note the formation of larger structures in the reactions prepared with the longer slncRNA molecules (e.g., PCP-8x and PCP-14x/MCP-15x), consistent with a gel like solid network (Fig. 1c). We examined the median condensate fluorescence obtained per slncRNA sequence. To get a standardized measurement, we normalized the measured fluorescence values by the number of estimated labeled uracil bases in each sequence (assuming a 35% labeling efficacy as reported by the manufacturer). The standardized quantity is then dependent on the number of molecules in a condensate, as well as on the average fluorescence of a single uracil Atto-488 label. The results (Fig. 1d) reveal a dependence on slncRNA class, where class I slncRNA molecules yield condensates with weaker fluorescence when compared with class II. The exception to this is PCP-14x/MCP-15x which appears to be weakest on average. This can be due to reduced fluorescence of a single uracil label brought about by quenching (reduction in fluorescence is estimated to be 4.2x – see methods: Granule microscopy and Fig. 2d). We further examined the background fluorescence from each slncRNA granule reaction and found roughly similar background levels for all slncRNAs (350–450 [A.U]). Normalization of the background (under the assumption that the main contribution to the background is free floating slncRNAs), reveals a dependence on size with the shorter slncRNAs (PCP-4x and PCP-3x/MCP-3x) showing higher normalized fluorescence compared to the negative control (Supplementary Fig. 3), and longer slncRNAs showing weaker values. Quantification of the area of the localized condensates shows that granule size is generally dependent on slncRNA length rather than on number of hairpins, with shorter slncRNAs resulting in smaller granules (Fig. 1e). To further analyze the condensate structure, we fitted the normalized condensate fluorescence intensity distributions to a modified Poisson distribution (see Fig. 1f, Supplementary Fig. 4 and methods: Estimating the signal amount per slncRNA-RBP complex). The panels reveal three characteristic distributions. For PCP-4x, an exponential distribution is recorded (i.e., λ = 0). For PCP-3x/MCP-3x, and PCP-4x/MCP-4x, a Poisson distribution of λ~1 seems to be the best fit. Finally, for PCP-8x, and PCP-14x/MCP-15x, a Poisson distribution of λ~2-3 fit best. These results are consistent with the formation of condensates that are characterized by an increasing number of slncRNA molecules that are cross-linked to form a gel-like “granule”, where the number of hairpins encoded into the slncRNA determines the average number of molecules or cross-links within the observed field of granules. Moreover, the interpretation suggested by the shape of the distribution is contrasted by the counter-intuitive observation of decreasing value of the of the fitting parameter K0 as a function of an increasing number of hairpins (Fig. 1g). In this particular context, this observation is manifested by a significantly more gradual increase in mean or median granule fluorescence as compared to what would be naively expected by a simple rescale that takes into account the number of hairpins. Together, these observations suggest that slncRNA granules form via cross-linking interaction of multiple slncRNA molecules, and that an increasing number of hairpins and cross-links lead to a denser condensate. Denser granules, in turn, may result in fluorescence quenching of the labeled uracils leading only to a gradual and disproportionate increase in fluorescence observed (reduction in fluorescence ranging from 1.3x to 4.2x depending on cassette - see Fig. 2d for details). To test if the hairpins retain their ability to bind the PP7 phage coat protein while in the granule state, we added recombinant tandem dimer PP7 coat protein fused to mCherry (tdPCP-mCherry) to the granule formation reaction in large excess (reactions were set up with 10–20 nM slncRNA concentration and 800 nM protein concentration, see methods: In vitro granule preparation) to saturate the slncRNA molecules while accounting for the multiple binding sites present on one slncRNA molecule (Fig. 2a). The tdPCP-mCherry version used lacks the necessary moiety to form the wildtype viral capsid. The images (Fig. 2b and Supplementary Fig. 5) show colocalization between the 488 nm channel (Atto-488) and the 585 nm channel (mCherry) for all slncRNA designs used in the experiment implying that PP7 coat proteins are able to bind the RNA hairpins in the condensate state. Hence, the slncRNA and their protein partners form synthetic RNA-protein (SRNP) granules. Unexpectedly, PP7-3x granules were witnessed in the presence of the protein. This could indicate that the addition of a protein element adds a measure of multivalency, thus allowing RNA molecules to condensate at a lower concentration compared to the slncRNA-only case. Another possible explanation could be that three adjacent PP7 stem loops are able to co-localize a sufficient number of tdPCP-mCherry proteins to form a bright punctum which than attracts additional slncRNA molecules (i.e., the protein is the nucleating agent instead of the slncRNA). To check that this condensation was hairpin dependent, we tested whether the control RNA (of the same length and GC content as PCP-8x) containing no designed hairpins, condenses either on its own or in the presence of tdPCP-mCherry. In both cases, no condensates were detected in either the 488 nm or 585 nm channels (Supplementary Fig. 6). Finally, unlike for the slncRNA only case, SRNP granules (particularly for high number of hairpins) show an increased propensity to form large-scale extended structures, suggesting a more complex structure formation and condensation for the SRNP granules as compared with the slncRNA-only case. To check that the observed colocalization is due to affinity between the coat protein and the corresponding binding sites, we repeated the granule formation reactions with a tandem dimer MS2 coat protein fused to mCherry (tdMCP-mCherry) protein. In this case, granules formed with class I slncRNAs showed no colocalization with the protein, while those formed with class II slncRNAs (which contain MCP binding sites) were colocalized (Supplementary Fig. 7). Next, we measured the median fluorescence intensity of the tdPCP-mCherry protein in different SRNP granules. The distributions of median values (Fig. 2c) show a clear dependence on the number of binding sites available for protein binding. First, the PCP-3x/MCP-3x and PCP-4x granules appear to have a similar number of proteins in the granule and are both weaker than PCP-4x/MCP-4x granules, suggesting that PCP-4x slncRNAs inside the granules are not fully occupied by proteins. In addition, the PCP-14x/MCP-15x granules seems to be >2-fold brighter as compared with the PCP-8x granules, despite having <2-fold the number of hairpins. This stands in contrast to the observation that PCP-14x/MCP-15x granules appear to be ~3 times brighter than PCP-4x/MCP-4x granules, reflecting the difference in the number of binding sites available for binding. Finally, PCP-3x granules appear to be half as bright as PCP-14x/MCP-15x granules, providing more evidence that the former are not RNA-dependent entities. We also observe that when the spacing regions within the slncRNA encode for the MCP hairpins, the formed granules contain a larger protein cargo (Fig. 2c). To confirm this observation, we also observed the SRNP granules in the 488 nm channel. Here a slightly more complex image emerges, whereby the median normalized fluorescence values for the RNA granules decline as the number of hairpins increases, hinting once again at a quenching process due to the tightly packed nature of the condensates. The reduction in fluorescence is estimated to be from 1.3x to 4.2x as a function of slncRNA size due to the increasingly packed nature of the condensates. In contrast, measurements of RNA-protein granules reveal an increase in fluorescence which is proportional to the number of binding sites available for protein binding. This trend peaks at PCP-8x before the effects of quenching become more dominant for the PCP-14x/MCP-15x (estimated to be 2x and 4.2x for the SRNP and RNA-only granules respectively). This behavior indicates at the existence of an optimum point for slncRNA design in terms of number of binding sites and complexity of the design. (Fig. 2d). Together, the observations in both channels indicate that SRNP granules are less dense gel-like structures as compared with the slncRNA-only granules. To authenticate the granules as being solid-like RNA-protein structures, we imaged them using structured illumination microscopy (SIM) super resolution microscope with 120 nm resolution. Figure 2e-left shows a sample image of a PCP-14x/MCP-15x granule containing the tdPCP-mCherry protein. The image shows that the slncRNA is found mainly in the periphery of the granule, with filaments protruding into its core, where a high amount of protein is amassed in a network like configuration. The RNA seems to encase the protein cargo in a dense shell-like structure. Figure 2e right shows a sample image of PCP-4x granules, depicting cage-like structures with a solid protein core and slncRNA filaments protruding and connecting the different structures. RNA-only granules on the other hand appear to be more compact and uniform in nature, akin to solids (Fig. 2f). Finally, we explored the phase space of SRNP granule formation. To do so, we characterized formation of the PCP-14x/MCP-15x SRNP granules as a function of both slncRNA and protein concentration. For this we produced non-fluorescent RNA molecules (for higher concentrations) and mixed different titers of slncRNA and tdPCP-mCherry protein, each varied over two orders of magnitude. Puncta like structures were detected only for slncRNA and proteins concentrations of 10 nM and 100 nM respectively, or above (Fig. 2g). The images display bright puncta that are embedded within a filamentous structure. Quantification of the maximal intensity of the puncta both at time T = 0 (i.e., beginning of the reaction) and time T = 1 [hr] (Fig. 2h) reveals a fluorescent intensity distribution which declines by two orders of magnitude (i.e., from ~105 to ~103) in a step-like function as the RNA concentration is reduced from 100 to 1 nM, providing further indication that RNA is essential for granule formation. Likewise, the intensity distribution of the puncta declines in a more gradual fashion as the protein concentration is reduced, but overall, a similar disappearance of puncta is observed. A hallmark of liquid-liquid phase separation is the exchange of molecules between the dilute phase and the dense phase. This is also true for gels with non-permanent intermolecular interactions, wherein random breaks and rearrangement of the connections which form the inner network allow macromolecules (monomers and small polymers) to diffuse in and out of the gel phase, albeit at a significantly slower rate as compared with a high density liquid phase. These exchange events are predicted to occur independently of one another, at a rate which depends on multiple parameters: the probability of cross linking within the gel network (i.e., number of hairpins), the transient concentration of the molecules in the surrounding solution, and the average diffusion rate of the monomers. The movement of molecules (fluorescent CPs, slncRNA, and CP-bound slncRNA complexes) between the different phases should be reflected by changes in granule fluorescence intensity. To test whether the synthetic granules display this characteristic, we tracked the fluorescence intensity in both the 488 nm channel (for slncRNA), and the 585 nm channel (for protein), of each granule in a given field-of-view for 60 min. We analyzed the brightness of each granule at every time point using a customized analysis algorithm (see methods: Signal analysis and Identifying burst events). The resulting signals are either decreasing or increasing in overall intensity, and dispersed within them are sharp variations in brightness, that are also either increasing or decreasing. Next, we employed a statistical threshold which flagged these signal variation events, whose amplitude was determined to not be part of the underlying signal distribution (p-value<1e-3) (See methods: Identifying burst events, and Supplementary Notes 1–3). We term the statistically significant signal variation events as “signal bursts”. These were classified as either increasing bursts (green) or decreasing bursts (red). In addition, we mark the non-significant segments (blue), which are segments where molecular movement cannot be discerned from the noise (Fig. 3a). For each detected burst, we measure its amplitude (Δ intensity) and duration (Δ time), in addition to measuring the time between bursts and the order of their appearance. In Fig. 3b we plot the distributions of amplitudes for all three event types, obtained from ~156 signal traces, each gathered from a different granule composed of PCP-14x/MCP-15x and tdPCP-mCherry. We observe a bias towards negative burst or shedding events. Assuming an interpretation that fluorescent burst events correspond to entry and shedding events of slncRNA-CP complexes into or out of the synthetic granules, the amplitude bias towards negative events is consistent with RNA degradation and lack of transcription within the in vitro suspension, leading to a net shedding of slncRNA-protein complexes out of the granules over time. To confirm that we are observing entry and shedding events of what are likely single slncRNA molecules into and out of the fluorescent granules vis-a-vis the signal bursts, we tracked the intensity of PCP-14x/MCP-15x RNA-only granules with and without the presence of RNase A. We first found that at enzyme concentrations above 35 nM, no granules were observed whatsoever, indicating correct activity of the enzyme. At a concentration of 35 nM we were able to track identified granules for at least 60 min. Figure 3c depicts a typical signal of a granule in the presence of RNase, showing a steady decline over several minutes. In addition, while shedding events seem to maintain their amplitude, re-entry events seem to rapidly diminish in amplitude to a median level that is ~10–20% of the original entry burst level (Fig. 3d). In particular, the entry burst amplitude reduced in a statistically significant fashion in the latter part of the tracking (30–60’) as compared with the first part of the experiment (0–30’ – Wilcoxon rank test p-value<0.01). Together, this indicates that slncRNA degradation occurs outside of the granules, while inside the structure they seem to be protected from degradation, consistent with a gel-like phase. We then proceeded to gather statistical tracking data for granules produced from all previously described slncRNA designs (including the PP7-3x which does not phase separate on its own at our working concentration). Comparison of the amplitude distributions per design (class I vs. II), (Fig. 3e) reveals a dependence on the number of hairpins available for protein binding, where more protein binding sites translate directly into larger amplitudes. As before, PCP-3x is revealed to be an outlier in this case, presenting amplitudes akin to those observed in PCP-4x/MCP-4x granules, providing another indication of a different phase behavior. In addition, we measured the time duration between events for each granule type. The observed rate (~10 min or more) is two orders of magnitude above the typical rate observed in liquid phase separated condensates, but is in line with the measurements performed on RNA gels by Vale et. al., providing additional confirmation that the SRNP granules are gel-like particles (Fig. 3f). Examination of the median time between bursts reveals that shedding events (negative bursts) occur roughly every 10 min, regardless of slncRNA design and number of binding sites, indicating a global behavior of the formed granules (Fig. 3g-top). Entry events (positive bursts) on the other hand, appear to demonstrate some dependence on the number of binding sites available for protein binding, at least for the slncRNAs with four or more binding sites. For these, the average time between events rises, signifying a reduced ability of bound slncRNA molecules to enter the granules (Fig. 3g-bottom). Such a behavior could indicate saturation of the granule, or a high degree of entanglement in the internal granule structure, hindering entrance of new molecules, while also allowing the stochastic shedding of molecules from the periphery of the granule. We next carried out protein exchange or competition experiments on PCP-14x/MCP-15x granules and measured the dynamics and time-scales associated with protein replacement. To do so we initially prepared PCP-14x/MCP-15x granules with tdPCP-mcherry and allowed the system to equilibrate. Then, immediately prior to imaging, we added tdPCP-mCerulean and observed the granule dynamics in all three channels (i.e., 405 nm for tdPCP-mCerulean, 488 nm for slncRNA, and 585 nm for tdPCP-mCherry – see methods: In vitro granule preparation). Altogether, we tracked 39 co-labeled granules in multiple experiments. Tracking of the formed granules in both channels (405 nm for cerulean, and 585 nm for mCherry) reveals a variety of signals, indicating various types of mixing of the two labels. For instance, we observed events of apparent displacement where one signal increases as the other decreases (Fig. 4a-right) indicating a replacement of the mCherry by mCerulean within the granules. Other signals showed a relatively synchronized signal indicating equilibration of the two protein labels at least within the solution (Fig. 4a. – left). We next computed the Pearson correlation for each pair of signals and plotted the distribution of correlation coefficients in Fig. 4b. The correlation computation shows that while a synchronized signal can be detected in 24 of the 39 signals (R > 0.8), an anticorrelated or unsynchronized signal is still detected in a significant number of the pairs (15 of 39 signals R < 0.8) indicating that a significant percentage of the granules are not found in equilibrium. Comparing the total number of events in both channels during the first 20’ reveals that while the number of entry events is similar (Fig. 4c - N = 25 for mCherry and N = 36 for mCerulean), the number of shedding events is significantly lower in the mCerulean channel (Fig. 4d. N = 40 and N = 2 for the mCherry and mCerulean channels respectively). At longer tracking durations (20–60’), more mCerulean shedding events are observed (N = 23), but are nevertheless x3 lower than the total number of mCherry shedding events (N = 69) over the same duration. Together, the significantly smaller number of shedding events over the one-hour tracking period in the mCerulean channel (N = 109 for mCherry vs N = 25 for mCerulean) as compared with the relatively equal number of entry events (N = 69 for mCherry vs N = 63 for mCerulean) indicates that while the unbound protein equilibrated as expected, the contents of the granules are still not in equilibrium after 1 hr. This interpretation is supported by time interval measurements between successive burst events (Fig. 4e). This analysis reveals a similar rate of entry of tdPCP-cerulean and tdPCP-mCherry proteins into the granules (Fig. 4e-left), while a discrepancy between the two rates is observed for the shedding events (Fig. 4e-right) in both the duration and total amount of events (N = 73 for tdPCP-mCherry and N = 4 for tdPCP-mCerulean, respectively). We next examined the burst amplitudes as a function of tracking duration intervals (Fig. 4f–g) in the mCherry channel. The data reveals that burst amplitudes for both entry and shedding events decrease over time with a time scale of ~10 min. Specifically, burst amplitudes for the mCherry channel are higher in the first 20 min as compared with the 20–40 and 40–60 min time window in the latter part of the tracking. This indicates a transition from slncRNAs that are fully occupied by tdPCP-mCherry to ones that are increasingly dominated by the tdPCP-mCerulean consistent with rapid equilibration of the proteins within buffer. Consequently, the equilibration of the tdPCP-mCherry and tdPCP-mCerulean within the solution, and the apparent lack of equilibration within the granules at least over the 1 hr duration of the experiments indicated by the unbalanced number of shedding events in both channels, provides additional evidence for a liquid-gel phase transition associated with the SRNP granulation process. To provide a measure for the number of slncRNA molecules within the granules, we computed the ratio between the mean granule fluorescence and the mean burst amplitude, assuming the average burst corresponds to one slncRNA molecule bound by proteins. The results (Fig. 5a, b) show that type I gel-like granules (PCP-4x and PCP-8x) have a smaller median number of slncRNAs (~5), as compared with type II gel-like granules (~8–10), suggesting that type II granules form better crosslinked structures. For the PCP-3x/MCP-3x, PCP-4x/MCP-4x, and PCP-14x/MCP-15x the ratio in the red channel displays a dependence on the number of hairpins supporting a more robust solid-like behavior when compared with the type I SRNP granules. We next calculated the “net rate of slncRNA loss”, defined as the difference between the total number of observed shedding and entry events, divided by the number of tracked granules (wherein one granule constitutes one hour of tracking data). The rates (Fig. 5c) show a difference between class I and class II granules, as well. While the net loss rate for class I granules increases with number of binding sites, it decreases with the number of binding sites for class II granules. This observation is consistent with type I and type II structures that are characterized by a decreasing and increasing amount of cross-linking, respectively, as a function of the number of hairpins on the slncRNA. Together, this data and the super-resolution microscopy images (Fig. 2e) suggest that class I and II granules form different types of gel-like phases, with the former forming a structure that is permeable to proteins while the latter seem to form robust protein storage nano-particles. In particular, the class II granule characteristics are reminiscent of data and energy storage devices (e.g., capacitors), with the protein cargo replacing the electric charge in the biochemical analog. To further characterize the “capacitor-like” behavior of the type II granules, we performed a titration experiment with PCP-14x/MCP-15x slncRNAs. We formed granules with a constant slncRNA concentration (120 nM) and different protein concentrations, resulting in a 1:1, 10:1 and 100:1 protein to RNA ratio. We collected shedding burst data for each condition and calculated the previously reported observables. Comparison of the shedding burst amplitudes (Fig. 5d) reveals that granules formed with 10:1 and 100:1 ratios have almost identical burst amplitudes, indicating slncRNA binding saturation. A 1:1 protein to RNA ratio results in amplitudes one tenth the intensity, as expected. Interestingly, the increase in burst amplitude also apparently leads to an increase in the number of slncRNA molecules within the granules. This can be seen from the “duration-between-successive-events” distribution (Fig. 5e), which shows that the time intervals between entry events in the ratio “1” granules are significantly larger than the ratio “10” and “100” intervals (Wilcoxon p-value <0.005). This combined with a constant rate of shedding independent of ratio leads to granules that are composed of a smaller number of slncRNAs for the ratio “1” as compared with ratio “10” and “100”. Finally, the dependence of the stored protein load on the burst amplitude allows us to define a proportionality constant uniquely for every granule-type. This type of proportionality constant is analogous to electrostatic “capacitance”, and can be defined by a biochemical analog to the capacitance equation q=CV. Here, the biochemical “charge” is the total amount of protein stored within the granules, and the protein concentration within the solution corresponds to a biochemical “voltage”. Consequently, the protein to slncRNA titration measurement provides a more solid footing for the capacitor analogy, suggesting that a potential protein-storage set of applications may be facilitated both in vitro and in vivo. Given the capacitor analogy, we hypothesized that in vivo the granules can be used as devices that store granule-bound proteins indefinitely. This is due to the steady state production of slncRNAs and proteins via the cellular transcriptional and translational machinery, that ensures a constant flux of proteins into the granules. To show this, we first proceeded to test whether the granule material characteristics that are measured in vivo match the in vitro measurements. To do so, we decided to utilize two previously reported slncRNA designs which were shown to yield bright localized puncta in vivo in earlier work. The first slncRNA is of a class II design, PCP-4x/ QCP-5x, consisting of four native PCP binding sites and five native Qβ coat protein (QCP) hairpins used as spacers in an interlaced manner. The second slncRNA is the ubiquitous PCP-24x cassette, which from the perspective of this work can be regarded as a class I design slncRNA. To confirm the granules form condensates in vivo, we encoded the slncRNA component under the control of a T7 promoter, and the tdPCP-mCherry under the control of an inducible pRhlR promoter (Fig. 6a). We first wanted to test whether puncta develop in vivo and whether they are dependent on the existence of hairpins in the RNA. For this we co-transformed plasmids encoding either the negative control RNA or the PCP-4x/QCP-5x slncRNA, together with a plasmid encoding for the tdPCP-mCherry protein, into BL21-DE3 E. coli cells. Examination of cells expressing the slncRNA and protein following overnight induction of all components revealed the formation of bright puncta at the cell poles (Fig. 6b), which were absent in cells expressing the control RNA which lacks hairpins (Fig. 6c). In addition, the difference between the cultures was even visible to the naked eye (Supplementary Fig. 8), indicating copious amounts of protein which appear to be dependent on number of binding sites encoded in the slncRNA. We believe this phenomenon was missed in the past since such binding sites were exclusively used to track individual mRNA transcripts in vivo where both low concentrations and the effects of translation might hinder the formation of large macro-molecular structures. Next, to test whether cellular concentration of slncRNA influences the formation of the granules, we quantified the fraction of puncta per cell for cells expressing the PCP-4x/QCP-5x from a multicopy expression vector, and cells expressing the same slncRNA from a bacterial artificial chromosome (BAC) expression vector which is maintained at a single copy level in cells. We found that cells containing the multicopy plasmid frequently present puncta in at least one of the poles, while cells containing the single copy generally show between zero and one punctum (Fig. 6d). Given that cells expressing the slncRNAs from single copy vectors still present puncta, we decided to continue using this expression vector in follow-up experiments to reduce variability stemming from copy number differences. We compared cells expressing the PCP-4x/ QCP-5x or the PCP-24x (expressed from a BAC vector) in terms of the spot per cell fraction. Much like in the in vitro experiments, we found a dependence on the number of binding sites in accordance with the in vitro results and the cross linking model of gel phase formation (Fig. 6d). Finally, to test whether the polar localization of the granules is a consequence of nucleoid exclusion, we grew the cells in starvation conditions for several hours, triggering a transition to stationary phase. In stationary phase the nucleoid is known to condense, thus increasing the amount of cellular volume which is likely to be molecularly dilute. This, in turn, generates a larger accessible cellular volume for granule formation, which should lead to different presentation of the phase-separation phenomena as compared with exponentially growing cells. In Fig. 6e, we show an image of bacteria displaying ‘bridging’ (the formation of a high intensity streak between the spots) whereby granules seem to fill out the available dilute volume. This behavior is substantially different than the puncta appearing under normal conditions. Such behavior was observed in >40% of the fluorescent cells and was not detected in non-stationary growth conditions. Thus, SRNP granules with characteristics that are consistent with the in vitro observations form in vivo, in a semi-dilute bacterial cytosolic environment and independent of cell-state. To investigate the dynamic properties of granules formed in vivo, we utilized the same analysis approach as was used in the in vitro experiments, with minor differences. Normalizing the fluorescence of the granule by that of the cell (see methods: Image analysis and Signal analysis) for every time point results in a signal vs. time trace largely independent from the effects of photobleaching and cellular background noise, allowing us to search for and measure burst events, as was done previously. In Fig. 7a, we plot the distributions of amplitude of all three event types (positive, negative, and non-classified), obtained from 255 traces gathered from cells expressing the PCP-4x/QCP-5x slncRNA together with the tdPCP-mCherry protein. The symmetry in both shape and spread of the negative and positive distributions indicates that both are measurements of the same type of macromolecule, distinguished only by the direction in which it travels (into or out of the granule). Moreover, a similarly symmetric burst distribution is recorded for the PCP-24x slncRNA (Supplementary Fig. 9). This result contrasts with the in vitro amplitude distribution data (Fig. 3b), which presented a skewness towards negative bursts. This implies that in vivo, the transcriptional and translational processes in the cell balance the loss of granule components due to degradation. Next, we measured the amplitudes of the bursts for both slncRNAs and found that positive and negative amplitudes are proportional to the number of binding sites within the encoded cassette (Fig. 7b). In addition, a more quantitative analysis of these distributions (Supplementary Fig. 10) reveals that a single burst for the 24x cassette is ~2.5-3x more fluorescent as compared with the 4x cassette, indicating that the molecules transitioning in and out of the 24x granules are slncRNAs partially or fully bound rather than lone proteins. Moreover, estimations of the positive and negative amplitudes are practically equal per slncRNA, providing additional evidence that these are in fact representations of one physical process, with the difference being the directionality of the transitioning slncRNA-protein molecule. Finally, we measured the duration between burst events, revealing that slow shedding and absorption processes on the order of minutes are taking place for the in vivo granules as well (Fig. 7c). Altogether, the non-existence of puncta in cells expressing the negative control RNA, the slow shedding/entry rate of molecules, and the dependence on the number of binding sites, suggest that synthetic RNA protein granules are phase separated condensates in vivo and possess the same gel-like characteristics that were observed for the in vitro suspensions. Consequently, in vivo burst analysis is consistent with the capacitor model, where the amount of protein stored within the SRNP granule seems to be in steady state when there is a steady supply of protein and slncRNA. Next, to ascertain whether the granules facilitate increased protein titers in vivo in accordance with the capacitor model predictions, we measured for each bright granule the mean fluorescence intensity (Fig. 7d), and the mean intensity of the cell which contains it (Fig. 7e). We observed a dramatic increase in mean cellular fluorescence between cells which express only tdPCP-mCherry and cells which express it together with a slncRNA, suggesting that slncRNA molecules have some effect in the cytosol, regardless of the granules. To quantify this phenomenon more accurately, we measured the total fluorescence of the population using flow cytometry. For this, we grew cells expressing only the protein component (tdPCP-mCherry), and cells expressing both protein and a slncRNA (PCP-4x/QCP-5x or PCP-24x), with different combinations of induction: IPTG (induces the slncRNA) and C4HSL (induces the protein). The data (Fig. 7f and Supplementary Fig. 11) shows that cells expressing a slncRNA, regardless of induction (due to T7 leakiness), show higher fluorescence than cells expressing the protein only. In addition, induction of slncRNA expression with IPTG results in an increase in fluorescence, indicating that slncRNA is a deciding factor in this behavior. Finally, cells expressing the PCP-24x slncRNA show higher fluorescence than cells expressing PCP-4x/QCP-5x, displaying a dependence of the cellular protein titer on number of binding sites available for protein binding. In this study, we show that synthetic gel-like RNA—protein granules can be designed and assembled using phage coat proteins and RNA molecules that encode multiple CP hairpin binding sites, both in suspension and in vivo. Using fluorescently labeled RNA, we show that granule formation is nucleated by RNA-RNA interactions that are proportional to the number of hairpins encoded into the RNA. In addition, the binding of the proteins seems to further enhance and assist the granule formation process. Using fluorescent single molecule signal analysis, we reveal entry and shedding events of molecules into and out of the granules. By investigating their size and rate of occurrence, we show that these events correspond to entry and shedding of protein-bound slncRNA molecules, and that they are dependent on the number of hairpins available for protein binding. Transitioning of macro-molecules across a phase boundary is frequently observed in phase-separated condensates, particularly in liquld-liquid based system. In particular, the frequency of these transitions reflects the underlying order, internal interactions, and density of the condensed phase. While in liquid-liquid phase separation systems such transitions occur on the scale of seconds or less, here we observe shedding and insertion events on a much longer time scale of minutes or longer, that is more consistent with a solid or gel-like condensed phase. We provide additional evidence for the liquid-gel phase transition underlying the granulation process, by taking the system out of equilibrium and observing the different equilibration times of the dilute solution and dense granule phases. To do so, we used a binding competitor in one experiment and RNAse in the other. For those experiments, the liquid phase showed rapid mixing in the former and substantial catalysis in the latter, while the granule phase showed slow mixing and undetectable amount of catalysis, respectively. The out-of-equilibrium experiments allowed us to differentiate between the liquid-gel transition from the other options which include liquid-liquid, liquid-glass, and liquid-solid transition. As a result, we believe that equilibrium characterizations should be considered a standard tool in similar biomolecular phase separation studies. We further characterized two options for slncRNA design: a homogeneous design which is comprised of multiple CP hairpin binding site and non-structured spacing regions (class I), and a hybrid design which is comprised of hairpin binding sites and additional hairpins in the spacing regions (class II). We show that the design choice has implications for the granule’s protein-carrying capacity and dynamics. In particular, class II granules formed particles with increased cross-linking capability in the RNA-only granule, which in turn led to an increased ability to insulate the protein cargo in the SRNP granule phase. On the flip side, class I granules were characterized by decreased cross-linking in the RNA-only phase and increased permeability of the protein cargo in the SRNP-granule phase. In addition, class I granules displayed a faster shedding or dissolution rate, which in turn lead to a smaller protein cargo on average. The slow release and strong internal interactions which keep the granules intact for long durations within a gel-like phase, combined with the selectivity of our system due to the RNA binding component, could be utilized as a programmable controlled release mechanism in suitable biological settings. Hence, our granules can be thought as protein and RNA storage modules akin to a capacitor, with ‘capacitance’ that is dependent on protein concentration, and a monophasic release profile that can be tuned based on slncRNA design. This two-dimensional phase space of capacity vs. rigidity offers substantial flexibility and tunability when designing SRNP granules for a variety of applications. The capacitor- or storage-like behavior displayed by the SRNP granules implies that in vivo, the granules together with the gene-expression machinery form a biochemical analog of an RC-circuit. In a conventional RC-circuit, energy is stored within the capacitor for release at a later time. Such circuits are often used to protect electrical devices against sudden surges or stoppages of power. Here, the protein and slncRNA flux into the cytosol correspond to the current, which results in the formation of fully “charged” SRNP granules. This genetically encoded slncRNA and protein storage facility, which is constantly maintained, effectively increases the protein and slncRNA content of the cell beyond the steady-state levels facilitated by standard transcription, translation, RNA degradation, and proteolysis. This storage capacity is precisely the function that is carried out by capacitors in RC-circuits, allowing electrical devices to function even after “power” is cut-off. In the case here, the granules can be used not only to increase levels of a protein of choice by nearly an order of magnitude (as shown in Fig. 7 and Supplementary Fig. 8) without adversely affecting the cell but may also provide a mechanism to increase the cell’s ability to survive when a harsh or stressful environment is encountered. While the former may have important implications to the biotechnology sector, the latter may hint at an important function that natural granules (e.g., paraspeckles, p-bodies, etc.) may have evolved for in vivo. In particular, our finding of liquid-gel phase transitions may be also relevant to repeat expansion disorders, which are associated with RNAs containing multiple structural repeats. Further studies will be required to explore the biological relevance of SRNP granules to the survivability of cells and organisms under various forms of stress, and to potential underlying mechanisms for various diseases. E. coli BL21-DE3 cells which encode the gene for T7 RNAP downstream from an inducible pLac/Ara promoter were used for all reported experiments. E. coli TOP10 (Invitrogen, Life Technologies, Cergy-Pontoise) was used for cloning procedures. pCR4-24XPP7SL was a gift from Robert Singer (Addgene plasmid # 31864; http://n2t.net/addgene:31864; RRID: Addgene_31864). pBAC-lacZ was a gift from Keith Joung (Addgene plasmid # 13422; http://n2t.net/addgene:13422; RRID: Addgene_13422). All sequences encoding for the in vitro slncRNAs (i.e., PP7-3x, PP7-4x, PP7-3x/MS2-3x, PP7-4x/MS2-4x, PP7-8x and PP7-14x/MS2-15x. Sequences appear in Supplementary Data 1) were ordered from Integrated DNA Technologies (IDT), (Coralville, Iowa) as gBlock gene fragments downstream to a T7 promoter and flanked by EcoRI (New England Biolabs (NEB),Ipswich, MA, #R3101L) restriction sites on both sides. gBlocks were cloned into a high-copy plasmid containing an Ampicillin resistance gene and verified using Sanger sequencing. The 5Qβ/4PP7 slncRNA sequence was ordered from GenScript, Inc. (Piscataway, NJ), as part of a puc57 plasmid, flanked by EcoRI and HindIII (NEB, #R3104L) restriction sites. pBAC-lacZ backbone plasmid was obtained from Addgene (plasmid #13422). Both insert and vector were digested using the said restriction enzymes and ligated to form a circular plasmid using T4 DNA ligase (NEB, #M0202L). Sequence was verified by sanger sequencing. Fusion-RBP plasmids were constructed as previously reported. Briefly, RBP sequences lacking a stop codon were amplified via PCR off either Addgene or custom-ordered templates. Both RBPs presented (PCP and QCP) were cloned into the RBP plasmid between restriction sites KpnI and AgeI (NEB, catalog: #R3142L and #R3552L respectively), immediately upstream of an mCherry gene lacking a start codon, under the so-called RhlR promoter containing the rhlAB las box and induced by N-butyryl-L-homoserine lactone (C4-HSL) (Cayman Chemicals, Ann Arbor, Michigan. #10007898). The backbone contained either an Ampicillin (Amp) or Kanamycin (Kan) resistance gene, depending on experiment. A vector containing the slncRNA DNA sequence, flanked by two EcoRI restriction sites, was digested with EcoRI-HF per the manufacturer’s instructions to form a linear fragment encoding the slncRNA sequence. The enzyme was then heat-inactivated by incubating the restriction reaction at 65 °C for 20 min. For fluorescently labeled RNA, 1 µg of the restriction product was used as template for in vitro transcription using HighYield T7 Atto488 RNA labeling kit (Jena Bioscience, Jena, Germany, RNT-101-488-S), according to the manufacturer’s instructions. Non-fluorescent RNA was transcribed using the HiScribe™ T7 High Yield RNA Synthesis Kit (NEB, #E2040S). Following in vitro transcription by either kit, the reaction was diluted to 90 µl and was supplemented with 10 µl DNAse I buffer and 2 µl DNAse I enzyme (NEB #M0303S) and incubated for 15 min at 37 °C to degrade the DNA template. RNA products were purified using Monarch RNA Cleanup Kit (NEB, #T2040S) and stored in −80°. E. coli cells expressing fusion proteins (tdPCP-mCherry / tdPCP-cerulean / tdMCP-mCherry) were grown overnight in 10 ml LB with appropriate antibiotics at 37 °C with 250 rpm shaking. Following overnight growth, cultures were diluted 1/100 into two vials of 500 ml Terrific Broth (TB: 24 g yeast extract, 20 g tryptone, 4 ml glycerol in 1 L of water, autoclaved, and supplemented with 17 mM KH2PO4 and 72 mM K2HPO4), with appropriate antibiotics and induction (500 μl C4-HSL) and grown in 37 °C and 250 rpm shaking to optical density (OD) > 10. Cells were harvested, resuspended in 30 ml resuspension buffer (50 mM Tris-HCl pH 7.0, 100 mM NaCl and 0.02% NaN3), disrupted by four passages through an EmulsiFlex-C3 homogenizer (Avestin Inc., Ottawa, Canada), and centrifuged (10,000 g for 30 min) to obtain a soluble extract. Fusion protein was purified using HisLink Protein purification resin (Promega (Madison, WI) #V8821) according to the manufacturer’s instructions. Buffer was changed to 1xPBS (Biological Industries, Israel) using multiple washes on Amicon ultra-2 columns (Merck, Burlington, MA #UFC203024). In vitro experiments were performed in granule buffer (final concentrations: 750 mM NaCl, 1 mM MgCl2, 10% PEG4000). Reactions were set up at the appropriate concentrations and allowed to rest at room temperature for 1 h. 3–5 µl from the reaction was then deposited on a glass slide prior to microcopy. For the RNase experiment, granules were first formed as described and allowed to rest at room temperature for 1 h. Following this, RNase A enzyme (Thermofisher, Waltham, MA, #EN0531) was added at 35 nM final concentration. The reaction was then immediately deposited on a glass slide and proceeded to imaging. For the competition experiment, granules were first formed with tdPCP-mCherry at a final concentration of 40 nM and allowed to rest at room temperature for 1 h. Following this, tdPCP-mCerulean was added to the reaction at 80 nM final concentration prior to imaging. BL21-DE3 cells expressing the two plasmid system (single copy plasmid containing the binding sites array, and a multicopy plasmid containing the fluorescent protein fused to an RNA binding protein) were grown overnight in 5 ml Luria Broth (LB), at 37° with appropriate antibiotics (Cm, Amp), and in the presence of two inducers – 1.6 µl Isopropyl β-D-1-thiogalactopyranoside (IPTG) (final concentration 1 mM), and 2.5 µl C4-HSL (final concentration 60 μM) to induce expression of T7 RNA polymerase and the RBP-FP, respectively. Overnight culture was diluted 1:50 into 3 ml semi-poor medium consisting of 95% bioassay buffer (BA: for 1 L—0.5 g Tryptone [Bacto], 0.3 ml glycerol, 5.8 g NaCl, 50 ml 1 M MgSO4, 1 ml 10×PBS buffer pH 7.4, 950 ml DDW) and 5% LB with appropriate antibiotics and induced with 1 μl IPTG (final concentration 1 mM) and 1.5 μl C4-HSL (final concentration 60 μM). For stationary phase tests, cells were diluted into 3 ml Dulbecco’s phosphate-buffered saline (PBS) with similar concentrations of inducers and antibiotics. Culture was shaken for 3 h at 37° before being applied to a gel slide [3 ml PBSx1, mixed with 0.045 g SeaPlaque low melting Agarose (Lonza, Switzerland), heated for 20 seconds and allowed to cool for 25 min]. 1.5 μl cell culture was deposited on a gel slide and allowed to settle for an additional 30 min before imaging. Granules were imaged in a Nikon Eclipse Ti-E epifluorescent microscope (Nikon, Japan) with a 100×1.45 NA oil immersion objective. Excitation was performed by a CooLED (Andover, UK) PE excitation system at 488 nm (Atto 488) for experiments containing fluorescent RNA, 585 nm for mCherry protein, and 405 nm for cerulean protein. Images were captured using the Andor iXon Ultra EMCCD camera with a 250 msec exposure time for 488 nm, 250 msec exposure time for 585 nm, and 2 seconds for 405 nm. Microscopy control and data acquisition was performed using Nikon NIS-Elements version 4.20.02 (build 988) 64 bit. Fluorescence quenching for the granules was estimated by first placing each slncRNA sequence on a two-dimensional space, with the x-axis being the estimated number of fluorescent uracil nucleotides (assuming 35% labeling efficiency), and the y-axis being the normalized RNA-only granule fluorescence values as depicted in Fig. 2d top. Estimated non-quenched fluorescence values were extrapolated by fitting a linear line to the data from PCP-4x and PCP-3x/MCP-3x which are assumed to be roughly non-quenched. A numerical value per slncRNA was then calculated by dividing the estimated non-quenched fluorescence by the fluorescence value measured empirically. Gel slide was kept at 37° inside an Okolab microscope incubator (Okolab, Italy). A time lapse experiment was carried out by tracking a field of view for 60 min on Nikon Eclipse Ti-E epifluorescent microscope (Nikon, Japan) using an Andor iXon Ultra EMCCD camera at 6 frames-per-minute with a 250 msec exposure time per frame. Excitation was performed at 585 nm (mCherry) wavelength by a CooLED (Andover, UK) PE excitation system. Microscopy control and data acquisition was performed using Nikon NIS-Elements version 4.20.02 (build 988) 64 bit. Quantification of the fraction of cells presenting puncta was done by taking 10–15 snapshots of different fields of view (FOV) containing cells. The number of cells showing puncta and the total number of fluorescent cells in the FOV were counted manually. Super resolution images were captured using the Elyra 7 eLS lattice SIM super resolution microscope (Zeiss, Germany) with an sCMOS camera, a 63×1.46 NA water immersion objective, with a 1.6x further optical magnification. 405 nm, 488 nm and 585 nm lasers were used for excitation of the cerulean, Atto-488, and mCherry respectively. Microscope control and data acquisition was performed using ZEN Black version 3.3.89. 16-bit 2D image sets were collected with 13 phases and analyzed using the SIM^2 image processing tool by Zeiss. IPTG / C4HSL-induced or non-induced E. coli BL21 cells were grown overnight, diluted 1:100 the next day and grown until OD of 0.3. Cells were diluted 1:10 in 1xPBS in a 1.5 ml tube and vortexed. Samples and appropriate controls were loaded onto a 96-wells plate (Thermo Scientific, cat. 167008) in triplicates, each well containing 100 μl of the diluted bacterial cells. The cells were then measured using flowcytometry (MACSquant VYB, Miltenyi Biotec), with the 561 nm excitation laser and the Y2 detector channel (a 615/20 nm filter). The flowcytometer was calibrated using MacsQuant calibration beads (Miltenyi Biotec) before measurement. Running buffer, washing solution, and storage solution were all purchased from the manufacturer (cat. numbers 130092747, 130092749, and 130092748, Miltenyi Biotec, respectively). Voltages for the SSC, FSC, and mCherry (Y2) channel were 467, 313, and 333 volts, respectively. Events were defined using an FSC-height trigger of 3. An FSC-area over SSC-area gate (gate-1) was created around the densest population in the negative control (non-induced cells), and events falling inside this gate were considered live bacteria. From the gate-1-positive cells, an SSC-height over SSC-area gate (gate-2) was created along the main diagonal, and events falling inside this gate were considered single bacterium. Finally, from the gate-2-positive cells, a histogram of mCherry distribution was created, with the threshold for mCherry-positive cells set by leaving around 0.1% positive events in the negative control. The brightest spots (top 10%) in the field of view were tracked over time and space via the imageJ MosaicSuite plugin. A typical field of view usually contained dozens of granules (in-vitro) or cells containing puncta (in vivo) (Supplementary Fig. 12a, b). The tracking data, (x,y,t coordinates of the bright spots centroids), together with the raw microscopy images were fed to a custom built Matlab (The Mathworks, Natick, MA) script designed to normalize the relevant spot data. Normalization was carried out as follows: for each bright spot, a 14-pixel wide sub-frame was extracted from the field of view, with the spot at its center. Each pixel in the sub-frame was classified to one of three categories according to its intensity value. The brightest pixels were classified as ‘spot region’ and would usually appear in a cluster, corresponding to the spot itself. The dimmest pixels were classified as ‘dark background’, corresponding to an empty region in the field of view. Lastly, values in between were classified as ‘cell background’ (Supplementary Fig. 12c). We note that for the in vitro experiments the ‘dark background’ and ‘cell background’ pixel groups yield similar intensity values. This, however, does not affect the performance of the algorithm for in vitro experiments. Classification was done automatically using Otsu’s method. From each sub-frame, two values were extracted, the mean of the ‘spot region’ pixels and the mean of the ‘cell background’ pixels, corresponding to spot intensity value and cell intensity value. This was repeated for each spot from each frame in the data, resulting in sequences of intensity vs. time for the spot itself and for the cell background. (Supplementary Fig. 12d). We assume a noise model comprised of both additive and exponential components, corresponding to fluorescent proteins (bound or unbound) not relating to the spot itself, and photobleaching. This can be described as follows: Where is the observed spot signal, is the underlying spot signal which we try to extract, is the observed cell background signal, is the underlying background signal and is the photobleaching component. To find , we assume: This leads to: To get , we filter the measured spot signal with a moving average of span 13, in order to remove high frequency noise effects, and smooth out fluctuations (see section – Identifying burst events). To get , we fit the measured cell background signal to a 3rd degree polynomial (fitting to higher degree polynomials did not change the results). This is done to capture the general trend of the signal while completely eliminating fluctuations due to random noise. We assume the total fluorescence is comprised of three distinct signal processes: RNP granule fluorescence, background fluorescence and noise. We further assume that background fluorescence is slowly changing, as compared with granule fluorescence which depends on the dynamic and frequent insertion and shedding events occurring in the droplet. Finally, we consider noise to be a symmetric, memory-less process. Based on these assumptions, we define a “signal-burst” event as a change or shift in the level of signal intensity leading to either a higher or lower new sustainable signal intensity level. To identify such shifts in the base-line fluorescence intensity, we use a moving-average filter of 13 points (i.e., 2 min) to smooth the data. The effect of such an operation is to bias the fluctuations of the smoothed noisy signal in the immediate vicinity of the bursts towards either a gradual increase or decrease in the signal (Supplementary Fig. 13a). Random single fluctuations, which do not settle on a new baseline level are not expected to generate a gradual and continuous increase or decrease over multiple time-points in a smoothed signal. Following this, we search for contiguous segments of gradual increase or decrease and record only those whose probability for occurrence is 1 in 1000 or less given a Null hypothesis of randomly fluctuating noise. To translate this probability to a computational threshold, we first compute the intensity difference distribution for every trace separately. This distribution is computed by collecting all the instantaneous differences in signal and binning them (Supplementary Fig. 13b). Given a particular trace the likelihood for observing an instantaneous signal increase event in a time-point () can therefore be computed as follows:where N(>0) and Ntot correspond to the number of increasing instantaneous events and total number of events in a trace respectively. Likewise, the number of decreasing instantaneous events is defined as: This in turn allows us to compute the number of consecutive instantaneous signal increase events (m) to satisfy our 1 in 1000 threshold for a significant signal increase burst event m as follows: The threshold is calculated for each signal separately and is usually in the range of 7–13 time points. An analogous threshold is calculated for decrements in the signal and is typically in the range . To account for the presence of the occasional strong instantaneous noise fluctuations appearing in experimental signals, we allow isolated reversals in the signal directionality (e.g., an isolated one time point decrease in an otherwise continuous signal increase environment). Furthermore, since the moving average filter itself can induce correlations in the signal, we determined that the minimum allowed threshold is the moving average window span. This means that any calculated threshold lower than the moving average size is increased to this bare minimum. We mark each trace with the number of events whose duration exceeds the threshold and define those as bursts. Segments within the signal that are not classified as either a negative or positive burst event are considered unclassified. Unclassified segments are typically signal elements whose noise profile does not allow us to make a classification into one or the other event-type. For each identified segment we record the amplitude (), and duration (). In Supplementary Fig. 13c we mark the classifications on a sample trace with positive “burst”, negative “burst”, and non-classified events in green, red, and blue, respectively. We confine our segment analysis between the first and last significant segments identified in each signal, since we cannot correctly classify signal sections that extend beyond the observed trace. Given the fact that we cannot directly infer the fluorescence intensity associated with a single RNA-RBP complex, we fitted the distributions with a modified Poisson function of the form:where I is the experimental fluorescence amplitude, λ is the Poisson parameter (rate), and K0 is a fitting parameter whose value corresponds to the amplitude associated with a single RBP-bound slncRNA molecule within the burst. For each rate we chose the fit to K0 that minimizes the deviation (MSE) from the experimental data. Fits were validated by observing the resulting QQ-plots. Each granule formation reaction and subsequent microscopy experiment was successfully repeated in duplicate on multiple days as follows: PCP-3x – 4 days; PCP-4x – 3 days; PCP-8x – 3 days; PCP-3x/MCP-3x – 2 days; PCP-4x/MCP-4x – 2 days; PCP-14x/MCP-15x – 4 days. RNase A degradation experiments were repeated in duplicate on 2 separate days. Titration experiments were repeated in triplicate on 2 separate days. Super resolution microscopy was repeated on 3 separate days. In-vivo experiments were repeated in triplicate over 5 separate days. Flow cytometry measurements were performed in triplicate. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. supplementary information Peer Review File Description of additional Supplementary File supplementary data 1 Reporting Summary

PMC9649770

Alexander Vdovin,Tomas Jelinek,David Zihala,Tereza Sevcikova,Michal Durech,Hana Sahinbegovic,Renata Snaurova,Dhwani Radhakrishnan,Marcello Turi,Zuzana Chyra,Tereza Popkova,Ondrej Venglar,Matous Hrdinka,Roman Hajek,Michal Simicek

The deubiquitinase OTUD1 regulates immunoglobulin production and proteasome inhibitor sensitivity in multiple myeloma

10-11-2022

Myeloma,Ubiquitylation,Ubiquitylation

Serum monoclonal immunoglobulin (Ig) is the main diagnostic factor for patients with multiple myeloma (MM), however its prognostic potential remains unclear. On a large MM patient cohort (n = 4146), we observe no correlation between serum Ig levels and patient survival, while amount of intracellular Ig has a strong predictive effect. Focused CRISPR screen, transcriptional and proteomic analysis identify deubiquitinase OTUD1 as a critical mediator of Ig synthesis, proteasome inhibitor sensitivity and tumor burden in MM. Mechanistically, OTUD1 deubiquitinates peroxiredoxin 4 (PRDX4), protecting it from endoplasmic reticulum (ER)-associated degradation. In turn, PRDX4 facilitates Ig production which coincides with the accumulation of unfolded proteins and higher ER stress. The elevated load on proteasome ultimately potentiates myeloma response to proteasome inhibitors providing a window for a rational therapy. Collectively, our findings support the significance of the Ig production machinery as a biomarker and target in the combinatory treatment of MM patients.

The deubiquitinase OTUD1 regulates immunoglobulin production and proteasome inhibitor sensitivity in multiple myeloma Serum monoclonal immunoglobulin (Ig) is the main diagnostic factor for patients with multiple myeloma (MM), however its prognostic potential remains unclear. On a large MM patient cohort (n = 4146), we observe no correlation between serum Ig levels and patient survival, while amount of intracellular Ig has a strong predictive effect. Focused CRISPR screen, transcriptional and proteomic analysis identify deubiquitinase OTUD1 as a critical mediator of Ig synthesis, proteasome inhibitor sensitivity and tumor burden in MM. Mechanistically, OTUD1 deubiquitinates peroxiredoxin 4 (PRDX4), protecting it from endoplasmic reticulum (ER)-associated degradation. In turn, PRDX4 facilitates Ig production which coincides with the accumulation of unfolded proteins and higher ER stress. The elevated load on proteasome ultimately potentiates myeloma response to proteasome inhibitors providing a window for a rational therapy. Collectively, our findings support the significance of the Ig production machinery as a biomarker and target in the combinatory treatment of MM patients. Multiple myeloma (MM) is a malignancy of immunoglobulin (Ig)-producing plasma cells and the second most common hematological cancer. In the last decade, the life expectancy of MM patients has improved significantly, mainly due to the introduction of novel treatment options, with proteasome inhibitors (PIs) being the outstanding pioneers of rational anti-MM therapy. Three representatives of PIs (bortezomib, carfilzomib, ixazomib) have been approved and are commonly used in routine clinical practice. Even though PIs presented a tremendous success in MM therapy, they are currently not included as a backbone in all anti-myeloma regimens. In fact, PIs-free triplet - daratumumab, lenalidomide and dexamethasone - has become a new standard of care in newly diagnosed, transplant-ineligible MM patients with unprecedented outcomes. Indeed, the majority of newly diagnosed myeloma patients will not receive PIs in their first line therapy, thus a reliable biomarker that would predict the favorable response to PIs is eagerly needed. It is generally accepted that the tremendous capacity to produce Ig molecules is a prerequisite for the unique myeloma sensitivity to PIs. The secreted serum monoclonal Ig (M-protein) is considered as a diagnostic hallmark of all monoclonal gammopathies and the kinetic of serum M-protein levels is used to monitor the disease.. Classical hematological response criteria to the treatment are based on the decrease and/or disappearance of M-protein. A sudden rise in the serum M-protein is a characteristic feature of the upcoming relapse and is considered to reflect a higher tumor burden. To fully engage the clinical potential of this exclusive myeloma attribute, a deeper understanding of the molecular mechanisms behind Ig production is essential. During the process of plasma cell maturation, both transcription and synthesis of Ig markedly increase. This puts enormous pressure on endoplasmic reticulum (ER) folding machinery to generate correctly assembled Ig molecules. Overloading the ER folding capacity ultimately leads to the accumulation of misfolded proteins that need to be extracted from ER, ubiquitinated, and targeted for degradation in proteasome. An increase in ubiquitinated species creates an imbalance in proteasome load versus capacity in both normal and malignant plasma cells, making them exceptionally prone to apoptosis upon proteasome inhibition. Tight control of protein homeostasis is therefore critical for myeloma cells’ survival and any disruption of the proteosynthetic and proteolytic machinery is deleterious. This was proven by many mechanistic studies associating PI resistance with mutations and expression changes of proteasome and ribosome subunits, as well as the ER stress components. The prognostic potential of these parameters in the clinical setting is, however, not entirely accepted. Therefore, there is a constant need for robust assays to identify PI-insensitive MM patients who could profit from precision combination therapy that would eradicate the resistant clones in the early stage of the disease. Here, we identify levels of aberrant plasma cell intracellular Ig as an independent prognostic factor which could distinguish MM patients suitable for successful PI-based treatment. Additionally, we uncover regulatory mechanism driving Ig synthesis at the translation level. Tremendous proteosynthetic capacity of plasma cells is utilized in the treatment of MM patients with PIs. In addition to sensitivity to PIs, extreme amount of newly synthetized Ig molecules creates a burden of constantly elevated ER stress, high energy demands and nutrients consumption. Therefore, Ig synthesis can be one of the parameters of clonal selection in the development and progression of MM. Aberrant plasma cell clones which lost the ability to synthesize complete Ig molecule and produce only Ig light chain (IgL) form usually more aggressive tumors. However, those are relatively rare cases. Up to date, it has not been studied if variability in Ig production affects disease outcome in patients with secretory MM. A common diagnostic parameter measured in all patients with plasma cell dyscrasias is the level of serum M-protein (secreted monoclonal Ig). To thoroughly validate M-protein prognostic value, we analyzed progression-free survival (PFS) and overall survival (OS) in up to date the largest cohort of newly diagnosed MM patients (Supplementary Fig. 1a, b and Supplementary Table 1) and further, in a subgroup of patients who received bortezomib in the first line of therapy (Supplementary Fig. 1c, d and Supplementary Table 2). We did not observe any correlation between the levels of serum M-protein at the time of diagnosis and either PFS or OS, suggesting that amount of secreted M-protein does not reflect tumor aggressiveness or drug sensitivity. On the other hand, concentration of intracellular Ig (iIg) should not be affected by tumor size and thus, it could better reflect the proteosynthetic rate and response of MM patients to PI-based therapy. In this study we measured amount of iIgL in the aberrant plasma cells from a cohort of 86 newly diagnosed MM patients. We observed a positive correlation between iIgL and PFS (Fig. 1a; Supplementary Table 3). MM patients with iIgL concentration below 0.1 μg iIgL/μg of total plasma cell proteins (iIgL low) had a worse prognosis compared to patients with iIgL above 0.1 μg iIgL/μg of total plasma cell proteins (iIgL high) (Fig. 1b). The cubic spline analysis underlined the validity of the selected cut-off value for iIgL concentration (Fig. 1c). Multivariate analysis further supported significance of iIgL prognostic effect over other commonly used predicting factors (Supplementary Table 4). Additionally, we analyzed relative iIgL levels in 106 newly diagnosed MM patients by flow cytometry. This independent approach also confirmed the prognostic impact of iIgL (Supplementary Fig. 1e). It is well established that the production of Ig is regulated on a transcriptional level by plasma cell-specific factors including IRF4 and XBP1. However, the expression analysis of primary MM patient samples revealed that the differences in the iIgL protein content could not be explained by altered transcription from the IGL locus (Fig. 1d). Similarly, analysis of the public expression dataset showed no association of IGLC1 expression with the MM patients’ survival (Fig. 1e). Together, these data highlight the importance of post-transcriptional regulation of Ig production in regards to disease outcome. The enormous synthesis of Ig molecules in plasma cells sets high demands on the folding apparatus and is inherently connected with an extensive load of misfolded, highly ubiquitinated proteins. Therefore, we hypothesized that deubiquitinating enzymes (DUBs) might play a significant role in the Ig production pathway. To identify DUBs controlling Ig synthesis, we performed a CRISPR knockout screen in myeloma cells covering most of the currently annotated human DUBs and looked for those that alter Ig production and at the same time affect MM survival (Supplementary Fig. 1f). From the candidate list (Supplementary Table 5), we selected DUBs that are highly expressed in plasma cells (Supplementary Data 1). Ovarian Tumor Deubiquitinase 1 (OTUD1) was the only DUB that fulfilled all the applied criteria. Moreover, OTUD1 is the most highly expressed DUB in B cells compared to other hematopoietic cells (Supplementary Fig. 1g) and reaches its expression peak in the fully differentiated bone marrow plasma cells throughout B-cell development (Supplementary Fig. 1h). Altogether, the cumulative evidence indicates a crucial role of OTUD1 in plasma cell biology, particularly in Ig synthesis. To evaluate the effect of OTUD1 on Ig production in our cohort, we split the patients based on the median OTUD1 expression (OTUD1 low and OTUD1 high) and quantified the iIgL content. The aberrant plasma cells isolated from the OTUD1 low MM patients had less iIgL compared to the OTUD1 high group (Fig. 1f). Similarly, dividing the patients according to the median iIgL concentration (0.1 μg iIgL/μg of total plasma cell proteins) revealed differences in the amount of OTUD1 mRNA (Fig. 1g). Yet again, the various quantities of iIgL in the OTUD1 low and OTUD1 high MM patients could not be explained by changes in the transcription of the IGL genes (Supplementary Fig. 1i). Importantly, increased expression of OTUD1 correlated with better survival of MM patients both in our cohort (Supplementary Fig. 1j) and in publicly available datasets (Fig. 1h, i). Collectively, these results suggest that OTUD1 regulates production of IgL on the post-transcriptional level and OTUD1 expression has a similar prognostic potential as iIgL concentration in newly diagnosed MM patients. To further study the involvement of OTUD1 in Ig production, we generated a panel of MM cell lines with doxycycline (dox)-inducible OTUD1 overexpression (OTUD1 oe) and shRNA-mediated OTUD1 knock-down (sh OTUD1_1, sh OTUD1_2) (Fig. 2a–c). As controls, we used isogenic cells without dox induction (control) for OTUD1 oe and cells expressing non-mammalian shRNA (sh control) for sh OTUD1 cells. Similar to MM patients, we observed a significant increase of iIgL in cells with OTUD1 oe (Fig. 2d). On the contrary, iIgL dropped in cells expressing sh OTUD1 (Fig. 2e). In both cases, OTUD1 did not influence the expression of IgL mRNA (Supplementary Fig. 2a, b) nor the amount of secreted IgL (Supplementary Fig. 2c). Further, the knock-down of other DUBs did not affect IgL production in MM cells (Supplementary Fig. 2d), supporting the specific effect of OTUD1. Finally, the expression of the catalytically inactive OTUD1 C320R mutant did not cause any changes in iIgL levels (Supplementary Fig. 2e), therefore OTUD1 enzymatic activity is required to sustain high IgL production. Because both our and public data imply a potential inhibitory effect of OTUD1 expression on myeloma aggressiveness, we tested the proliferation capacity of our OTUD1 genetic models. As expected, OTUD1 oe suppressed myeloma cell growth while introduction of sh OTUD1 promoted proliferation (Fig. 2f, g). In the in vivo settings, the effect of differential OTUD1 expression on tumor growth was even more pronounced (Fig. 2h-k; Supplementary Fig. 2f, g). Additionally, the concentration of iIgL in myeloma cells extracted from the mouse xenografts correlated with OTUD1 levels (Fig. 2l, m). Further analysis revealed that OTUD1 oe induced partial arrest in the S phase of the cell cycle without affecting cell viability (Supplementary Fig. 2h, i). These results recapitulate the situation seen in MM patients and validate the use of cell line models to study the function of OTUD1 in myeloma pathogenesis. While exploring the mechanism behind the modulation of iIgL levels by OTUD1, we noticed dramatic changes in total ubiquitin pools in myeloma cells with altered OTUD1 expression. Surprisingly, OTUD1 oe led to a massive rise in total ubiquitination (Fig. 3a), while OTUD1 knock-down had the opposite effect (Fig. 3b). At the same time, the observed changes in protein ubiquitination directly correlated with the ER stress status (Fig. 3a–d). As we observed this phenomenon only in myeloma but no other cell lines (Supplementary Fig. 3a), and OTUD1 regulates IgL production in plasma cells (Fig. 2d, e), we hypothesized that the differences in global ubiquitination might be caused by alteration in the iIgL levels. Indeed, siRNA-mediated IgL knock-down significantly reduced the ubiquitination pattern (Fig. 3e). Additionally, immunoprecipitation of IgL from myeloma cells with OTUD1 oe revealed an increase in IgL ubiquitination (Fig. 3f) while ubiquitination of IgL dropped in sh OTUD1 cells (Fig. 3g). Because IgL is a secreted protein, it is not exposed to cytosolic ubiquitin ligases. Therefore, we speculated that the ubiquitinated forms of IgL represent the damaged or misfolded molecules retranslocated from ER and destined for proteasome degradation. Indeed, the application of the cell-permeable proteasome fluorescent substrate showed a significant occupation of proteasome in OTUD1 oe cells which was reverted by the introduction of siRNA-targeting IgL (Fig. 3h). Altogether, our results indicate that OTUD1 regulates synthesis of IgL and the vast amount of ubiquitinated proteins in myeloma cells belongs to misfolded Igs that are responsible for clogging proteasome. It is assumed that the unique sensitivity of MM to PIs is caused by a constant proteasome overload. Since our data indicate that OTUD1 activity correlates with the amount of iIgL, ubiquitinated products, and MM patient survival, we hypothesized that changes in the iIgL levels and IgL ubiquitination caused by altered OTUD1 expression might result in a different myeloma response to PIs. As expected, OTUD1 oe (high iIgL) increased the sensitivity of MM cells to all clinically used PIs, while the cells with OTUD1 knock-down (low iIgL) developed profound PI resistance (Fig. 3i, j). Analysis of proteasome expression and activity did not reveal any differences in myeloma cells with OTUD1 oe (Supplementary Fig. 3b–d). Together, these data suggest that OTUD1 modulates MM sensitivity to PIs by regulating a load of misfolded/ubiquitinated Ig on proteasome. To unveil the mechanism of how OTUD1 potentiates Ig production, we performed a proximity labeling assay in cells with dox-inducible expression of OTUD1 fused to the newest generation of the biotin ligase BirA (TurboID) (Supplementary Fig. 4a). The top potential OTUD1 interactors included two proteins involved in oxidative stress handling: the E3-ligase KEAP1 and the ER-resident peroxiredoxin 4 (PRDX4) (Fig. 4a; Supplementary Data 2). We explored both KEAP1 and PRDX4 because all Ig molecules are rich in disulfide bonds and their formation is likely accompanied by elevated oxidative stress. First, we validated the OTUD1-KEAP1 interaction by reciprocal co-immunoprecipitation using HA-tagged OTUD1 and Flag-tagged KEAP1 (Fig. 4b). We were able to immunoprecipitate both proteins from MM cells also on the endogenous level (Fig. 4c). A close inspection of the OTUD1 amino acid sequence identified a canonical KEAP1-interaction motif (ETGE) near its N-terminus (Fig. 4d). Deletion of this motif completely disrupted the association of OTUD1 with KEAP1 (Fig. 4e), providing additional evidence for the direct protein-protein binding. The presence of the KEAP1-substrate ETGE motif indicated that OTUD1 might be ubiquitinated by KEAP1. To test this possibility, we used the OTUD1 C320R construct to avoid self-deubiquitination or aberrant enrichment of ubiquitinated species during extraction. As expected, OTUD1 lacking the ETGE motif was less ubiquitinated compared to the wild type form (Fig. 4f) suggesting that OTUD1 might be a target of the KEAP1 ubiquitin ligase activity. Interestingly, removal of the ETGE sequence did not lead to any changes in OTUD1 protein levels (Fig. 4g). At the same time, none of the tested phenotypes altered upon OTUD1 expression in myeloma cells (Ig production, proliferation, PI sensitivity) was affected by the absence of the KEAP1-binding site in OTUD1 (Supplementary Fig. 4b–d), prompting us to further investigate the role of PRDX4. PRDX4 activity is critical for a continuous electron flow in the ER redox cycles, particularly during disulfide bond formation. The appearance of the S-S bridges is the initial and rate-limiting step in the synthesis and folding of Ig molecules. In accordance, elevated levels of PRDX4 were previously associated with the accumulation of Ig in both MM cell lines and primary patient samples. The importance of PRDX4 for Ig production is further highlighted by its unique expression profile that steadily rises during B-cell development and peaks similarly to OTUD1 in bone marrow plasma cells (Supplementary Fig. 5a). Initially, we validated the OTUD1-PRDX4 interaction in series of reciprocal co-immunoprecipitations with both tagged and endogenous proteins (Fig. 5a, b). Analysis of other members of the PRDX family confrimed the unique selectivity of OTUD1 for PRDX4 (Supplementary Fig. 5b). While OTUD1 is a cytosolic deubiquitinase, PRDX4 contains the N-terminal (amino acids 1-38) ER-localization sequence that restricts it to ER lumen. Thus, the direct protein binding is seemingly incompatible. To map and localize the OTUD1-PRDX4 interaction, we used the full length (FL) and the Δ1-38 PRDX4 constructs. Interestingly, OTUD1 failed to immunoprecipitate PRDX4 deletion mutant (Fig. 5c). Additionally, immunofluorescence analysis confirmed colocalisation of FL PRDX4 and OTUD1 at the site of the ER membrane (Fig. 5d, e). Moreover, extraction of subcellular organelles identified OTUD1 and FL PRDX4 but not Δ1-38 PRDX4 present in the ER fractions (Supplementary Fig. 5c, d). Together, these results suggest that localisation of PRDX4 to ER is a prerequisite for its interaction with OTUD1. Next, we tested whether OTUD1 acts upstream of PRDX4. In vitro, recombinant OTUD1 deubiquitinated PRDX4 indicating that PRDX4 might be an OTUD1 substrate (Fig. 5f). Immunoprecipitation of both Flag-tagged PRDX4 from OTUD1 oe cells (Fig. 5g) and endogenous PRDX4 from MM cells with sh OTUD1 (Fig. 5h) further supported this idea. Additionally, OTUD1 oe increased while OTUD1 knock-down decreased the amount of endogenous PRDX4 protein (Fig. 5i, j) without any effect on the levels of PRDX4 mRNA (Supplementary Fig. 5e). Finally, cycloheximide pulse-chase confirmed a positive effect of OTUD1 expression on PRDX4 protein levels (Fig. 5k). Thus, we conclude that OTUD1 deubiquitinates PRDX4 and regulates its abundance, possibly by protecting it from degradation. To date, no activity of ubiquitin ligases was detected inside ER. Therefore, we hypothesized that ubiquitination of PRDX4 might occur during its retrotranslocation to cytosol in the process of ER-associated degradation (ERAD). In accordance, we detected ubiquitination only of ER-localized FL PRDX4, but not cytosolic Δ1-38 PRDX4 (Fig. 5l). Because proteasome activity is crucial for successful completion of ERAD, we tested its effect on PRDX4. As expected, inhibition of proteasome rescued levels of FL PRDX4 to the same extent as OTUD1 oe (Fig. 5m). On the other hand, blocking the proteasome activity did not affect levels of Δ1-38 PRDX4 (Fig. 5n). Moreover, OTUD1 oe led to a higher amount of FL PRDX4 but not Δ1-38 PRDX4 in the ER fractions (Supplementary Fig. 5c, d), further suggesting that OTUD1 protects PRDX4 from ERAD. A previous study described OTUD2 (YOD1), a close homolog of OTUD1, as a non-selective regulator of protein retrotranslocation from ER. We analyzed levels of protein disulfide isomerase A6 (PDIA6) in cells with OTUD1 oe in order to examine whether OTUD1 acts in a similar manner. PDIA6 is localized in the ER lumen and, together with PRDX4, participates in the formation of disulfide bonds. Proteasome inhibition promoted PDIA6 levels indicating that PDIA6 is degraded by ERAD. Conversely, OTUD1 oe did not increase PDIA6 protein levels (Fig. 5m). These results advocate OTUD1 selectively regulates the amount of PRDX4 rather than affecting general ERAD mechanisms. Our data indicate that OTUD1 regulates Ig production, PI sensitivity, and proliferation of MM cells. At the same time, OTUD1 binds and elevates PRDX4 protein amount. Because PRDX4 is an important component of the ER folding machinery, we speculated that changes in the amount of iIgL in myeloma cells with altered OTUD1 expression might be caused by variations in the PRDX4 protein quantity. To test this hypothesis, we rescued the original levels of PRDX4 in OTUD1 oe and sh OTUD1 cells by introducing siRNA targeting PRDX4 (siPRDX4) or overexpressing PRDX4 (PRDX4 oe), respectively (Fig. 6a, b). As expected, the amount of iIgL restored almost to the basal state upon normalizing the PRDX4 protein amount in both OTUD1 oe and sh OTUD1 cells (Fig. 6c, d). Similarly, increased sensitivity (OTUD1 oe cells) and resistance (sh OTUD1 cells) to bortezomib were reverted in cells with rescued PRDX4 (Fig. 6e, f). Lastly, restoration of PRDX4 levels in cells with both OTUD1 oe and knock-down led to a complete rescue of the total pool of ubiquitinated proteins (Fig. 6g, h). On the other hand, differences in myeloma cell proliferation caused by differential OTUD1 expression were not changed by normalizing PRDX4 protein levels or by restoring the IgL levels (Supplementary Fig. 6a–c). Therefore, the effect of OTUD1 on MM cell growth seems to be not dependent on PRDX4 and IgL. PRDX4 is directly involved in the series of redox reactions required for disulfide bonds formation in the newly formed Ig molecules. Therefore, we hypothesized that increase in PRDX4 protein levels might alter the speed of IgL folding creating a disbalance and saturation of the ER folding machinery. Ultimately, amount of misfolded, highly ubiquitinated IgL molecules would rise leading to a proteasome overload. Indeed, PRDX4 oe significantly elevated the amount of ubiquitinated species which was diminished by the introduction of siRNA-targeting IgL (Fig. 6i). Additionally, knock-down of PRDX4 suppressed ubiquitination of IgL (Fig. 6j). Because only misfolded Ig is extracted from ER and ubiquitinated, our data indicate PRDX4 is involved in both Ig synthesis and folding. However, it remained unclear how OTUD1 and PRDX4 activity translates to the altered Ig synthetic rate. Recent study associated OTUD1 with a ribosome stalling. Moreover, a vast majority of OTUD1-associated proteins in our proteomic analysis belonged to ribosome or translation machinery (Supplementary Data 2) suggesting OTUD1 could mediate translation of Ig mRNA. Puromycin incorporation assay revealed that OTUD1 does not affect Ig translation initiation (Supplementary Fig. 6d, e). On the other hand, the overall rate of translation elongation was enhanced in cells with elevated levels of OTUD1 or PRDX4 (Fig. 6k, l). Together these results support the model where OTUD1 increases PRDX4 protein amount which promote the formation of new IgL molecules by enforcing translation elongation. Pronounced Ig synthesis further leads to oversaturation of the ER protein assembly capacity resulting in a raise of both folded and misfolded/ubiquitinated Ig molecules. Finally, to validate the suggested molecular mechanism, we analyzed total ubiquitin levels in aberrant plasma cells isolated from newly diagnosed MM patients who later received PI-based therapy. As in our cell line models, the amount of ubiquitinated proteins directly correlated with the iIgL and iIgH protein content (Fig. 7a top and middle), OTUD1 expression and PFS (Fig. 7a bottom). Moreover, when we treated several MM cell lines with bortezomib, iIgL levels were significantly lower in the surviving cells compared to the vehicle-treated controls (Fig. 7b). This shows that even within highly homogenous populations of cell lines variations in iIgL correlate with the sensitivity to PIs. In order to revert the poor prognosis of MM patients with low iIg, we hypothesized that increasing the amount of misfolded, highly ubiquitinated proteins might compensate for the lack of proteasome saturation and re-sensitize the iIg low myeloma cells to PIs. Because accumulation of incorrectly folded proteins is a typical hallmark of ER stress, we evaluated a set of the ER stress-inducing drugs that are approved for use in MM patients or are being tested in clinical studies. We applied this drug panel together with bortezomib to PI-resistant sh OTUD1 (iIgL low) myeloma cells (Supplementary Fig. 7a). The best-performing compound for bortezomib co-treatment was the Heat Shock Protein 90 (HSP90) inhibitor tanespimycin (17-N-allylamino-17-demethoxygeldanamycin, 17-AAG) that completely reverted PI resistance in sh OTUD1 cells (Fig. 7c). Analysis of ubiquitinated proteins confirmed the molecular mechanism of tanespimycin-mediated re-senitization of sh OTUD1 cells to bortezomib was mediated by elevating the amount of ubiquitinated proteins that re-saturated proteasome (Fig. 7d). Serum M-protein is one of the main diagnostic values measured for every MM patient. Recent reports indicated that the levels and early dynamics of M-protein in response to treatment could serve as an indicator of tumor size and potentially a predictor of PFS. However, these studies included a limited number of patients to draw a definitive conclusion on M-protein prognostic significance. Here, we retrospectively analyzed data from more than 4000 MM patients and provide an evidence that M-protein levels fail to predict patient outcome. In contrast to M-protein, iIg concentration is likely not affected by tumor burden and thus represents the myeloma Ig production capacity more precisely. Therefore, quantification of iIg from aberrant plasma cells could serve as a better estimate of the Ig synthesis rate. Indeed, we found the concentration of iIg as a robust, independent factor that can be used to anticipate the course of the disease upon induction of PI-based therapy. Previous studies suggested that the changes in the Ig expression might have a predictive potential. However, our results and publicly available data revealed no correlation between the expression of Ig mRNA and serum M-protein or iIg. These findings indicate that the control exerted by additional factors on Ig relevant to MM proteostasis occurs mainly at the post-transcriptional level. In a search for regulators of Ig production we identified deubiquitinase OTUD1 highly expressed in mature plasma cells. Decreased expression of OTUD1 correlated with poor survival in several unrelated MM patient cohorts. In myeloma cells, OTUD1 negatively affected progression through the cell cycle independently of the Ig expression. This observation is consistent with the previous findings describing OTUD1 as a tumor suppressor in other cancers. Additionally, our data support the positive role of OTUD1 in the regulation of Ig production and MM sensitivity to PIs. These phenotypes were fully dependent on ER-localized PRDX4. We found OTUD1 selectively deubiquitinates PRDX4 during retranslocation from ER, thus protecting it from ERAD. PRDX4 is a critical component of the ER folding machinery where it mediates the formation of intra- and intermolecular disulfide bonds during oxidative protein folding. This process is additionally catalyzed by protein disulfide isomerases (PDIs) and endoplasmic reticulum oxidoreductase 1 (ERO1) which pass the electrons to molecular oxygen generating H2O2. PRDX4 is responsible for the reduction of peroxide and, to some extent, also PDI enzymes. Therefore, PRDX4 determines the assembly speed of disulfide bond-rich Ig molecules. Accordingly, a previous study suggested that the expression of PRDX4 steadily rises during B-cell development and correlates with plasma cells’ capacity to produce Ig. However, a precise mechanism how OTUD1-PRDX4 axis mediates the Ig synthesis remained unclear. Currently emerging evidence indicates on an exciting phenomenon connecting protein folding and translation processivity. As for other multidomain proteins, folding of Ig molecules might require ribosome pausing and stalling. Recent study indicated that OTUD1 is able to revert stalled ribosomes. In agreement, our proteomic analysis revealed ribosomal proteins as the most enriched OTUD1 associating proteins. Biochemical assays shown that upregulation of OTUD1 and PRDX4 stimulates translation elongation and ribosome processivity. This mechanism might potentiate Ig synthesis without any apparent changes in Ig transcription. Elucidation of how OTUD1 and PRDX4 are linked to protein synthesis will require further research. Interestingly, OTUD1 binds and selectively promotes PRDX4 levels leaving other components of the ER folding machinery intact. Together with increased Ig production, it might create a disbalance and saturation of the ER redox system resulting in a rise in misfolded/ubiquitinated forms of Ig. Appearance of defected Ig ultimately augments the ER stress and congests available proteasomes which is reflected by potentiated sensitivity to PIs. In support, a recent study identified PRDX4 as one of the most highly expressed genes in the primary, PI-sensitive myeloma clones. The relevance of the Ig folding pathway to PI resistance in MM is further emphasized by the development of PDI inhibitors that are currently under investigation as promising agents to boost the efficacy of PIs. In addition to changes in the Ig levels, disturbance of the ER redox circuit might promote oxidative stress that further imbalances proteasome capacity. Interestingly, we identified the E3-ligase KEAP1, a protein sensor of oxidative stress, as an OTUD1 binding partner. Though KEAP1 directly interacts and ubiquitinates OTUD1, this modification did not alter OTUD1 protein content. Additionally, deletion of the canonical KEAP1-binding motif from OTUD1 did not affect all tested, OTUD1-mediated phenotypes in MM cells. Thus, the biological relevance of the KEAP1-OTUD1 complex remains to be elucidated. In summary, our work unveils the OTUD1-PRDX4 axis as a yet undescribed regulatory pathway driving Ig production and PI sensitivity in MM cells and highlights importance of OTUD1 in myeloma proliferation (Fig. 8). We suggest that expression of OTUD1 and, particularly, iIg concentration could be considered as promising prognostic and stratification parameters for newly diagnosed MM patients. Since the amount of iIg can be measured directly, future studies should identify precise and clinically relevant thresholds to recognize PI-responsive and non-responsive patients that would profit the most from PI-based combination therapy, especially in the era of PI-free standard of care regimens. Human RPMI8226, MM.1 S, HEK 293, and U2OS cell lines were obtained from American Type Culture Collection under respective cat. n. CCL-155, CRL-2974, CRL-1573, HTB-96. JJN-3, SK-MM-2 were obtained from Deutsche Sammlung von Mikroorganismen und Zellkulturen under respective cat. n. ACC 541, ACC 430. 293FT cells were obtained from Invitrogen under cat. n. R70007. Cell lines were maintained in RPMI1640 (for RPMI8226, MM.1S, JJN-3 and SK-MM-2) or DMEM (for HEK 293, U2OS, 293FT) medium containing 10% heat-inactivated fetal bovine serum. Analysis of clinical data was retrospective using data from the Czech Registry of Monoclonal Gammopathies (RMG: https://rmg.healthregistry.org/). All patients signed informed consent for data collection and the study protocol prior entering the RMG. The consent form and the study protocol have been approved by the ethical committee of the University Hospital of Ostrava. Parameters of interest in the Registry contain all demographic data, disease characteristics, and treatment intervals including overall survival (OS) and progression-free survival (PFS). Data were described by absolute and relative frequencies of categorical variables and median (min–max) of continuous variables. Survival analysis (PFS and OS) was computed by Kaplan-Meier method and statistical significance of differences in survival among subgroups was assessed using the log-rank test. All statistical tests were performed at a significance level of p = 0.05 (all tests two-sided). Analysis was performed in the SPSS software (IBM Corp. Released 2017. IBM SPSS Statistics for Windows, Version 25.0.0.1 Armonk, NY: IBM Corp.) and software R version 4.0.1. (www.r-project.org). Cubic spline analysis was performed using rms v6.2-0R package and R v4.0.5. All patients gave a written informed consent before sample collection. The collection and the study protocol were approved by the Ethical Committee of the University Hospital Ostrava. Fresh bone marrow aspirates from MM patients (n = 106) were routinely collected in the Haematooncology department of the University Hospital Ostrava between 2013 and 2019. Malignant plasma cells were isolated using CD138 MicroBeads (Miltenyi Biotec) by autoMACS Pro Separator. In total 16 SCID mice (CB17/Icr-Prkdc-scid/IcrIcoCrl, female, 7-12 weeks, The Jackson Laboratory) were used for in vivo experiments (4 mice per condition). All animal experiments were approved by the Animal Ethics Committee of the Faculty of Medicine, Ostrava University and the Animal Ethic Board of the Ministry of Education, Youth and Sport of the Czech Republic n. MSMT-4072/2021-3. 107 RPMI8226 cells expressing NanoLuc in 0.2 ml PBS (1:1 matrigel) were injected subcutaneously into an anesthetized mouse. Tumor size was analyzed by bioluminescence starting 14 days after inoculation. To measure bioluminescence D-luciferin was injected intraperitoneally 5 min prior to imaging once per week during 5 weeks of the experiment. Photon count (P/s) was measured analyzed using Bruker MI SE 7.2 software to estimate a tumor size. To induce OTUD1 expression, doxycycline was administered at a dose of 2.5 mg/kg per day intraperitoneally for 3 days every week. After sacrifice, tumor xenografts were mechanically dissociated into single-cell suspension and iIgL content and GFP expression were analyzed by flow cytometry. Maximal tumor size approved by the ethical committee was 18 mm in diameter and this size was not exceeded in any of the animal. Bone marrow aspirates were diluted in the Red Blood Cell lysis buffer (155 mM NH4Cl, 12 mM NaHCO3, 0.1 mM EDTA; 1 ml: 10 ml ratio) and incubated at room temperature for 15 min. Aberrant plasma cell phenotype was determined according to the EuroFlow protocol (stained for CD38, CD138, CD45, CD56, CD117, CD27). Cells were sorted using BD FACSAria II and processed immediately using. Flow cytometry data were acquired by BD FACSDiva™ Software v9.0 and CytExpert software v2.4; and analyzed by FlowJo v10. Aberrant plasma cells were lysed in RIPA buffer containing protease inhibitor cocktail (Roche), iodoacetamide (50 μM) and N-ethylmaleimide (10 mM). Protein concentration was measured using BCA Protein Assay Kit (Pierce) and cell lysates were diluted to 2 μg/ml of total protein. Concentration of iIgL was determined by Human Lambda or Human Kappa ELISA kit (Bethyl Laboratories). Total RNA was extracted from cells using the RNeasy Mini Kit (Qiagen). The RNA aliquots were stored at −80 °C. The quality (purity and integrity) of the RNA samples was assessed using the Agilent 2100 Bioanalyser with the RNA 600 NanoLabChip reagent set (Agilent Technologies). The RNA was quantified using a spectrophotometer. Complementary DNA (cDNA) synthesis was performed using the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific) according to the manufacturer’s instructions. Quantitative RT-PCR was conducted using PowerUpTMSYBRTM Green Master Mix (Applied Biosystems) on StrepOnePlus Real-Time PCR System (Applied Biosystems). Relative mRNA expression was calculated by 2−ΔΔCt method and normalized to GAPDH or GUSB gene. Oligonucleotide sequences used in the study can be found in Supplementary Data 3. Lentiviral constructs (1.64 pmol, pLKO.1 for shRNA and pCW57.1 for overexpression) were used for plasmid construction and transfected into 293FT cells (3.5 × 106 cells seeded in 10 cm dish overnight) together with the helper plasmids (0.72 pmol of pMD2.G and 1.3 pmol of psPAX2) using jetPRIME® transfection reagent (Polyplus transfection). Viral supernatants were collected 48 h post transfection, mixed with PEG-8000 solution (final concentration 10 % W/V) and sodium chloride (final concentration 0.3 M) and agitated overnight at 4 °C. Next day, viral particles were concentrated by spinning down at 1600 g for 60 min (4 °C). Resulted pellet was resuspended in PBS. 5 × 105 RPMI8226 or MM.1S cells were incubated with viral supernatant in presence of 10 μg/mL polybrene (Sigma-Aldrich) in a final volume of 2 mL and spin-infected for 1 h at 900 g (34 °C). Cells were then supplemented with 3 mL fresh medium, continued culture for at least 48 h and selected with puromycin (2 μg/ml) overnight. 5 × 105 RPMI8226 or MM.1S cells were stained with LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit (Invitrogen). Cells were fixed and permeabilized using FIX & PERM™ Cell Permeabilization Kit (Invitrogen) and labeled with Lambda-APC-C750™ (Cytognos) antibody. 104 events were collected in LIVE/DEAD negative gate, MFI was analyzed using geometric mean statistics. Flow cytometry data were acquired by BD FACSDiva™ Software v9.0 and CytExpert software v2.4; and analyzed by FlowJo v10. Gating strategy is shown in the Supplementary information file. The gating strategy is shown in Supplementary Fig. 8. 104 RPMI8226 or MM.1S cells were seeded per one well in 96-well plate in triplicates for each timepoint. Cell proliferation was measured by MTT assay at days 1–5. For viability assay 5 × 104 were seeded, treated with a drug or combination of drugs and analyzed in 16 h. In the MTT assay, cells were incubated with the MTT labeling reagent (final concentration 0.5 mg/ml) for 30 min at 37 °C, 5% CO2, formazan crystals were solubilized in DMSO and absorbance was measured at 540 nm. RPMI8226 cells were collected in proteasome activity assay buffer (50 mM TRIS, pH 7.5, 10 mM NaCl, 250 mM sucrose, 5 mM MgCl2, 1 mM EDTA, 1 mM DTT, and 2 mM ATP) and lysed by sonication. Lysates were centrifuged at 17,000 g for 10 min at 4 °C. Protein concentration was determined using the BCA protein assay (Thermo Fisher Scientific). 25 μg of total protein of cell lysates was transferred to a 96-well black plate (Corning Costar) and then the fluorogenic Suc-LLVY-AMC (100 μM, Enzo) substrate was added to lysates. Fluorescence intensity (340 nm excitation, 440 nm emission) was monitored using an Infinite F Plex (Tecan) every 20 min for 1.5 h at 37 °C and the data were analyzed by GraphPad Prism. Gating strategy is shown in the Supplementary information file. 2.5 × 107 HEK 293 cells with doxycycline-inducible expressing of TurboID N-terminally fused to OTUD1 and TurboID only were treated with biotin (50 μM) for 1 h before lysis with Urea buffer (8 M Urea, 50 mM TRIS, pH 7.6, 1 mM DTT) supplemented with 1% Triton X-100 and protease inhibitor cocktail (Thermo Fisher Scientific). After 30 min incubation on ice, the concentration of Triton X-100 was decreased to 0.5% by dilution with Urea buffer. The cells were further sonicated and lysates were clarified by centrifugation at 17,000 g for 10 min at 4 °C. Protein concentration was determined by BCA protein assay (Thermo Fisher Scientific). The lysates were mixed with 25 μL of Streptavidin sepharose high performance resin (Cytiva) and incubated on a rotator at 4 °C overnight. Beads were washed extensively with Urea buffer and transferred to new tubes before washing with ammonium bicarbonate buffer (1 mM biotin, 50 mM ammonium bicarbonate). The enriched biotinylated proteins were subjected to on-bead trypsin digestion (1 µg of trypsin at 37 °C overnight). The digested peptides were collected and desalted using in-house made StageTips packed with C18 disks (Empore) before mass spectrometry analysis. Peptides were separated and analyzed on an UltiMate 3000 RSLCnano system coupled to an Orbitrap Fusion Tribrid mass spectrometer (both from Thermo Fisher Scientific). Peptides were firstly loaded onto an Acclaim PepMap300 trap column (300 µm × 5 mm) packed with C18 (5 µm, 300 Å, Thermo Fisher Scientific) in loading buffer (0.1% trifluoroacetic acid in 2% acetonitrile) for 4 min at 15 μL/min and then separated in an EASY-Spray column (75 µm × 50 cm) packed with C18 (2 µm, 100 Å, Thermo Fisher Scientific) at a flow rate of 300 nL/min. Mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetonitrile) were used to establish a 60-min gradient from 4% to 35% B. Eluted peptides were ionized by electrospray. A full MS spectrum (350-1400 m/z range) was acquired at a resolution of 120,000 at m/z 200 and a maximum ion accumulation time of 100 ms. Dynamic exclusion was set to 60 s. Higher-energy collisional dissociation (HCD) MS/MS spectra were acquired in iontrap in rapid mode and normalized collision energy was set to 30% with maximum ion accumulation time of 35 ms. The automatic gain control for MS and MS2 was set at 1E6 and 5E4, respectively. Top speed mode with 2 s cycle time and lower intensity threshold 5E3 were selected. An isolation width of 1.6 m/z units was used for MS. All raw data were processed and searched using MaxQuant 1.6.3.4 with the UniProtKB reviewed human protein database (release 2020_07; 20,381 sequences). Trypsin specificity was set C-terminally to arginine and lysine residues, also allowing the cleavage of proline bonds. Two missed cleavage sites of trypsin were allowed. Carbamidomethylation of cysteine was selected as fixed modification and N-terminal protein acetylation and methionine oxidation as variable modifications. The false discovery rate of both peptide identification and protein identification was set as 1%. The options of “Second peptides” and “Match between runs” were enabled. Label-free quantification was used to quantify the difference in protein abundance between different samples. Data analysis was performed using Perseus 1.6.1.3 software and significance was compared using two-sided Student t-test (Supplementary Data 2). The obtained dataset was further correlated with the published OTUD1 interaction data. The full coding sequence of the human OTUD1 gene (gene ID 220213) was cloned into a pGEX-6P-1 vector containing N-terminal 6His-GST tag cleavable by tobacco etch virus (TEV) protease. The cloned gene was expressed in BL21(DE3) RIPL strain of E. coli. Cells were grown in LB medium at 37 °C until OD600 of 0.6 after which the expression was induced by 0.5 mM IPTG. Cells were grown for another 18 h at 18 °C, pelleted by centrifugation, and resuspended in GST-binding buffer (50 mM TRIS, pH 7.5, 500 mM NaCl). Cell suspensions were supplemented with lysozyme (1 mg/mL), PMSF (1 mM), MgCl2 (1 mM) and benzonase (10 mUnits/μL), and sonicated. Bacterial lysates were obtained by centrifugation for 30 min at 12,000 × g. 6His-GST tagged proteins were captured on a GSTrap 5 ml FF column (Cytiva) and eluted with 20 mM glutathione. 6His-GST-OTUD1 was further subjected to 6His-TEV protease cleavage at 4 °C overnight. 6His-GST tag with 6His-TEV protease was then captured using immobilized metal affinity chromatography on a HisTrap column (Cytiva), whereas purified recombinant OTUD1 protein was present in the flow-through fractions. The purified OTUD1 was concentrated and finally exchanged into assay buffers using 7 kDa molecular mass cut-off Zeba spin desalting columns (Thermo Fisher Scientific). TEV protease 6His-TEV(S219V)−5Arg was prepared in house following the modified method from ref. 56. Co-immunoprecipitation of OTUD1 and PRDX4 complexes was performed by lysing RPMI8226 cells in immunoprecipitation buffer (20 mM TRIS, pH 7.5, 150 mM NaCl, 0.5% NP40) containing phosphatase and protease inhibitor cocktail (Thermo Fisher Scientific), 20 mM N-ethylmaleimide, and 5 mM iodacetamide for 1 h on ice. Lysates were clarified by centrifugation at 17,000 g for 10 min at 4 °C and pre-cleared by incubation with Protein A/G PLUS-agarose resin (Santa Cruz Biotechnology) on a rotator for 1 h at 4 °C. Then, 4 mg of lysate was incubated with 4 μg of anti-PRDX4 rabbit polyclonal antibody (10703-1-AP, Proteintech) or with anti-OTUD1 rabbit polyclonal sera (Moravian Biotechnology) at 4 °C overnight. The antibody-bound protein complexes were incubated with Protein A/G PLUS-agarose beads (Santa Cruz Biotechnology) on a rotator for 4 h at 4 °C. After washing the beads with immunoprecipitation buffer, bound proteins were eluted by boiling in NuPAGE LDS sample buffer (Invitrogen) and analyzed by western blotting. HEK 293 cells expressing N-terminally HA-tagged (OTUD1) or Flag-tagged (KEAP1, PRDX4) proteins were harvested, washed with ice-cold PBS, and lysed in immunoprecipitation buffer (20 mM TRIS, pH 7.5, 150 mM NaCl, 0.5% NP40) containing protease inhibitor cocktail (Thermo Fisher Scientific), 20 mM N-ethylmaleimide, and 5 mM iodacetamide for 30 min on ice. The lysates were clarified by centrifugation at 17,000 × g for 10 min at 4 °C and mixed with anti-FLAG M2 or anti-HA agarose resin (both from Sigma Aldrich). After 2 h incubation on a rotator at 4 °C, beads were washed extensively with immunoprecipitation buffer. Bound proteins were eluted by excess of FLAG or HA peptide (0.4 mg/mL, both from Sigma Aldrich) and analyzed by western blotting or used for in vitro deubiquitination assay. 2 × 106 of RPMI8226 cells were treated with 50 μg/mL of cycloheximide for indicated time points. Afterwards, cells were washed with cold PBS and lysed in RIPA buffer. Cell lysates were analyzed by immunoblot. Densitometry analysis of immunoblots was performed using ImageJ v1.49 (National Institutes of Health). HEK 293 cells expressing N-terminally Flag-tagged PRDX4 and 6His-ubiquitin were lysed and Flag-PRDX4 was immunoprecipitated as described above. Next, Flag-PRDX4 protein was mixed with increasing concentration of purified recombinant OTUD1 protein (0.1, 1, and 2 μM) in the presence of DTT (0.5, 5, and 10 mM, respectively) in the in vitro deubiquitination assay buffer (50 mM TRIS, pH 7.5, 150 mM NaCl). The deubiquitination reactions were carried out at 37 °C for 30 min, quenched by boiling in NuPAGE LDS sample buffer (Invitrogen), and analyzed by western blotting. HEK 293 cells expressing N-terminally Streptavidin-Binding Peptide (SBP)-tagged OTUD1 C320R and C320R/ΔETGE mutants were lysed as for immunoprecipitation experiments. Next, ubiquitinated proteins were enriched using Ubiquitin pan Selector agarose resin (NanoTag Biotechnologies). After 2 h incubation on a rotator at 4 °C, beads were washed with immunoprecipitation buffer and bound proteins were eluted by boiling the beads in NuPAGE LDS sample buffer (Invitrogen) and analyzed by western blotting. The immunoprecipitates and cell lysates were resolved by SDS-PAGE and transferred to PVDF membrane. Membranes were blocked in 5% (w/v) non-fat milk (Roth) in PBS-T (phosphate buffer saline, 0.05% Tween-20) and incubated overnight at 4 °C in 1% (w/v) BSA/PBS-T with the appropriate primary antibodies. Primary antibodies used at indicated dilutions include: anti-β-Actin (3700S, clone 8H10D10, CST, 1:500), anti-FLAG (F1804, clone M2, Sigma Aldrich, 1:1000), anti-GAPDH (97166S, D4C6R, CST, 1:1,000), anti-HA (11867423001, clone 3F10 Roche, 1:1000), anti-human IgG (H + L) (SAB3701329, Sigma Aldrich, 1:500), anti-human Ig lambda (CYT-LAC750, Cytognos, 1:1000), anti-KEAP1 (8047S, D6B12, CST, 1:1,000), anti-OTUD1 (custom made rabbit polyclonal sera, Moravian Biotechnology), anti-polyHistidine (H1029, clone HIS1 Sigma Aldrich, 1:1,000), anti-PRDX4 (60286-1-Ig, Proteintech, 1:1000), anti-PRDX4 (19178-1-AP, Proteintech, 1:1000), anti-PSMB5 (19178-1-AP, Proteintech, 1:1000), anti-PSMC6 (A303-825A, Bethyl Laboratories, 1:1,000), anti-PSMA2 (sc-377148, clone B4, SantaCruz, 1:1000), anti-PSMB1 (sc-374405, clone D9, SantaCruz, 1:1000), anti-PSMB7 (sc-365725, clone H3, SantaCruz, 1:1000), anti-PCNA (13110S, clone D3H8P, CST, 1:1000), anti-PDIA6 (A304-519A, lot n. 1, Bethyl, 1:1000), anti-calnexin (sc-46669, clone E10, SantaCruz, 1:1000), anti-ATF6 (65880S, clone D4Z8V, CST, 1:1000), anti-puromycin (MABE343, clone 12D10, Sigma Aldrich), anti-phospho-eIF2α (Ser51) (3398, clone D9G8, CST, 1:1000), anti-ubiquitin (3936S, clone P4D1, CST, 1:1000). Membranes were subsequently washed with PBS-T and incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. HRP-coupled secondary antibodies used at indicated dilutions include: goat anti-rabbit-IgG (111-035-144, Jackson ImmunoResearch, 1:2000), goat anti-mouse-IgG (115-035-146, Jackson ImmunoResearch, 1:2000), goat anti-rat-IgG (NA935, Cytiva, 1:2000), streptavidin (016-030-084, Jackson ImmunoResearch, 1:10,000). After further washing, signal detection was performed using ECL (Thermo Fisher Scientific) and ChemiDoc MP System (Bio-Rad). ImageJ v1.49 (National Institutes of Health) was used to analyze protein bands by densitometry. Custom made rabbit polyclonal anti-OTUD1 antibody was validated for use in western blotting on lysates from human wild type and OTUD1 knockout cell lines and also using recombinant human full length OTUD1 protein. 3 × 107 HEK 293 cells were washed in PBS and resuspended in hypotonic buffer (10 mM HEPES, pH 7.8, 1 mM EGTA, 25 mM potassium chloride) in 3X volume of the cell pellet. After 20 min of incubation at 4 °C cells were centrifuged, resulted pellet was resuspended in isotonic buffer (10 mM HEPES,pH 7.8, 250 mM sucrose, 25 mM potassium chloride,1 mM EGTA) in a 4X volume of the new cell pellet. Cells were lysed by passing through 27-gauge needle 4 times. The lysate was centrifuged at 1000 × g for 10 min at 4 °C. Supernatant was transfered to a new tube and centrifuged ar12,000 × g for 15 min at 4 °C. Supernatant was transferred to 15 ml tube, 7.5 volumes of cold 8 mM CaCl were slowely added while constantly vortexed. The tube was rotated for 15 minutes at 4 °C and centrifuged at 8000 × g for 10 min at 4 °C. Supernatant containing cytoplasmic fraction was concentrated to 50 µl using Amicon® Ultra-4 10 K Centrifugal Filter Device. The pellet contained precipitated ER, it was washed in isotonic buffer at 8000 × g for 10 min at 4 °C and resuspended in 30 µl. RPMI8226 cells were incubated in growth media with 500 nm of Me4BodipyFL-Ahx3-L3-VS probe for 1 h at 37 °C. Cells were washed in PBS and analyzed by flow cytometry. As a negative control cells were treated with 1 µM of carfilozomib for 1 h before staining. Flow cytometry data were acquired by BD FACSDiva™ Software v9.0 and CytExpert software v2.4; and analyzed by FlowJo v10. Gating strategy is shown in the Supplementary information file. The gating strategy is shown in Supplementary Fig. 8. For microscope observations, the cells were maintained overnight in 24-well plate with the glass cover slip in and then transiently co-transfected with OTUD1-GFP and PRDX4-mApple or mApple-Sec61β constructs using PEI. After 48 h from transfection, the cells were washed in PBS, fixed with 4% PFA in PBS for 15 min at room temperature (RT) and washed in PBS. The residual aldehydic groups were reduced by 15 min incubation at RT in 0.1 M glycine. Subsequently, samples were washed in PBS and stained with (4′,6-diamidine-2′-phenylindole dihydrochloride) DAPI for 5 min at RT and washed 2 times in PBS. Washed slides were let to air dry and mounted in a mounting medium. All samples were examined with a Leica Dmi8 platform. Samples were documented using a 63×/1.40 NA (numerical aperture) oil immersion objective. The CHOPCHOP web tool (https://chopchop.cbu.uib.no/) was used to design sgRNAs. The sequences targeting the first or second exon were chosen. In case of multiple isoforms, sequences present in the common exon or the canonical isoform were selected. Sequences with the GC content higher than 45%, efficiency score higher than 0.5, low self-complementarity, and mismatch score were preferred. For generating gene knockout in RPMI8226, CRISPR-Cas9 technology based on two vector system was used. Plasmids for transfection were purchased from Addgene (Lenti-guide puro #52963 and Lenti-cas9 blast #52962). Three sgRNA per gene were simultaneously applied. For lentivirus generation, 1 × 106 293FT cells was seeded in 6 well plate and transfected with the respective plasmids the next day. Packaging plasmids (psPAX2, 2.58 µg; pMDG, 780 ng), 3 Lenti-guide puro vectors (1 µg each) containing sgRNA to specific gene and 11 µL of linear polyethylenimine (MW 25 000, 1 mg/ml) were mixed in 450 µl of Opti-MEM (Gibco) for 15 min and added to the cells. Media was changed the next day. After 72 h media was collected and mixed with PEG-8000 solution (final concentration 10 % W/V) and sodium chloride (final concentration 0.3 M) and agitated overnight at 4 °C. Next day, viral particles were concentrated by spinning down at 1,600 g for 60 min (4 °C) and resulted pellet was resuspended in PBS. 1.5 × 105 RPMI8226 cells stably expressing Cas9 were mixed with lentiviral particles and 10 µg/ml of Polybrene (Millipore), and spin-infected for 1 h at 900 × g (34 °C) in 96-well round bottom plate. The cells were then seeded into 24-well plate, and the media was changed the next day. After 72 h post infection, cells were selected with puromycin (2 μg/ml). Oligonucleotide sequences used in the study can be found in Supplementary Data 3. RPMI8226 cells were incubated with 10 μg/ml of puromycin for 10 min at 37 °C and washed in PBS and lysed in immunoprecipitation buffer (20 mM TRIS, pH 7.5, 150 mM NaCl, 0.5% NP40). Puromycin incorporation was analyzed by western blot. HEK 293 cells were treated with 2 μg/ml of harringtonine to block translation initiation and after 0, 3, 6 and 9 min were pulsed with 10 μg/ml of puromycin for 10 min at 37 °C, washed in PBS and lysed in immunoprecipitation buffer (20 mM TRIS, pH 7.5, 150 mM NaCl, 0.5% NP40). Puromycin incorporation was analyzed by western blot. Densitometry analysis of immunoblots was performed using ImageJ v1.49 (National Institutes of Health). The statistical significance of differences between various groups was calculated with the two-tailed paired t-test or ANOVA, and error bars represent standard deviation of the mean (SD). Statistical analyses, unless otherwise indicated, were performed using GraphPad Prism 5. Data are shown as mean ± SD. Multivariate Cox proportional hazard model was computed using R 4.0.3 and survival v3.2.11 package. Images of gels in the figures are representative experiments which have been repeated in form of independent biological replicates multiple times as indicated. Figures 2b, c, 3a, b, 6a, b, g, h, k, l (n = 3); 3e, 4b, c, e, g, 5a–c, f–h, k–n, 6i, j and Supplementary Figs. 3a, c, 5b–d, 6d, e (n = 2); 7a (n = 1). Source data are provided as a Source Data file. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary

PMC9649776

Louis M. P. Ter-Ovanessian,Jean-François Lambert,Marie-Christine Maurel

Building the uracil skeleton in primitive ponds at the origins of life: carbamoylation of aspartic acid

10-11-2022

Astrobiology,Origin of life

A large set of nucleobases and amino acids is found in meteorites, implying that several chemical reservoirs are present in the solar system. The “geochemical continuity” hypothesis explores how protometabolic paths developed from so-called “bricks” in an enzyme-free prebiotic world and how they affected the origins of life. In the living cell, the second step of synthesizing uridine and cytidine RNA monomers is a carbamoyl transfer from a carbamoyl donor to aspartic acid. Here we compare two enzyme-free scenarios: aqueous and mineral surface scenarios in a thermal range up to 250 °C. Both processes could have happened in ponds under open atmosphere on the primeval Earth. Carbamoylation of aspartic acid with cyanate in aqueous solutions at 25 °C gives high N-carbamoyl aspartic acid yields within 16 h. It is important to stress that, while various molecules could be efficient carbamoylating agents according to thermodynamics, kinetics plays a determining role in selecting prebiotically possible pathways.

Building the uracil skeleton in primitive ponds at the origins of life: carbamoylation of aspartic acid A large set of nucleobases and amino acids is found in meteorites, implying that several chemical reservoirs are present in the solar system. The “geochemical continuity” hypothesis explores how protometabolic paths developed from so-called “bricks” in an enzyme-free prebiotic world and how they affected the origins of life. In the living cell, the second step of synthesizing uridine and cytidine RNA monomers is a carbamoyl transfer from a carbamoyl donor to aspartic acid. Here we compare two enzyme-free scenarios: aqueous and mineral surface scenarios in a thermal range up to 250 °C. Both processes could have happened in ponds under open atmosphere on the primeval Earth. Carbamoylation of aspartic acid with cyanate in aqueous solutions at 25 °C gives high N-carbamoyl aspartic acid yields within 16 h. It is important to stress that, while various molecules could be efficient carbamoylating agents according to thermodynamics, kinetics plays a determining role in selecting prebiotically possible pathways. The question of the emergence of the first life forms, of which we know nothing and yet are the Darwinian descendants, can be tackled from the angle of the inert-to-living transition. The “geochemical continuity” hypothesis states that, at some stage during the evolution of life, key parts of metabolic pathways recapitulated reactions which previously occurred in a non-biological setting. It is both parsimonious and falsifiable, and also compatible with the idea that life developed in a continuous process rather than as a “freak accident”. Additionally, protometabolic paths from simple abundant precursors may continually resupply biochemical building blocks, avoiding the depletion problem encountered with exogenous delivery scenarios. In this hypothesis, life-defining structures (metabolism, information, compartments) may have been initiated along the same general paths, but with other alternatives for kinetic (inorganic, including heterogeneous, catalysis) and thermodynamic control (free energy issued from macroscopic environment fluctuations) than those we observe today in organisms. In this line of thought, we are exploring a typical metabolic sequence of nucleotide biosynthesis, the de novo synthesis of pyrimidines (orotate pathway), with the aim to transpose it to an abiotic environment. Non-enzymatic pyrimidine biosynthesis has been the object of much interest lately, either through attempts at transposition of the orotate pathway, or through alternative pathways involving different precursors. In a previous publication, we considered the prebiotic potential of carbamoyl phosphate, an activated carbamoylating agent used at the beginning of this biochemical pathway. In the present one, we concentrate on the formation of N-carbamoyl-aspartic acid (NCA), the 7-atom precursor of the uracil framework. NCA, also called ureidosuccinic acid, exists in all living species, ranging from bacteria to eukaryotes. NCA is present in cytoplasm as well as excreta (saliva) and organs (prostate). It is synthesized from carbamoyl phosphate and L-aspartic acid through the action of the aspartate carbamoyltransferase enzyme (ATCase). As it plays a key role in aspartate and pyrimidine metabolisms, it is involved in several dysfunctions such as Canavan disease and dihydropyrimidinase deficiency. After an additional cyclization step, NCA forms the core skeleton of orotic acid, the precursor of uracil (Fig. 1), hence it is a major prebiotic target to check whether the geochemical continuity hypothesis is valid. In previous work, we showed that carbamoyl phosphate (CP) is rather unstable under prebiotic conditions, but produces two other molecules still containing the energy-rich carbamoyl moiety: cyanate and urea. It is therefore unlikely that CP itself was involved in a prebiotic carbamoylation pathway. However, the potential of cyanate and urea-type compounds as alternate carbamoylating agents deserves to be explored. Cyanate and urea-like compounds can be produced by several pathways in prebiotic settings, contrary to carbamoyl phosphate. Therefore, we first investigated NCA synthesis through the reaction between cyanate and L-aspartic acid under alkaline aqueous conditions. Then, we also assessed mineral surface scenarios involving drying steps in order to test the predictions of the geochemical continuity hypothesis, including the idea that mineral catalysts can mimic the role of enzymes (Fig. 2). In our first experiment, the carbamoyl donor was sodium cyanate, reacting with aspartate under basic conditions (see Methods). The N-carbamoylation of aspartic acid has, in fact, been used long ago in Nyc’s protocol for orotic acid synthesis. The N-carbamoylation of generic amino acids by cyanate was further explored by the Commeyras team. The sodium hydroxide solution, initially containing aspartic acid and sodium cyanate in a 1:1 ratio, was left to react for 16 h at 25 °C, sampled and directly analysed in deuterated water by 1H NMR. Two sets of signals are identifiable (Fig. S1). The set of sodium aspartate (Fig. 3) is constituted by four signals corresponding to NH, Hα and the two Hβ protons, the latter overlapped with signals of the second set. Similarly, the second set is comprised of four signals corresponding to NH, Hα and the two Hβ protons, the latter overlapped with signals of the first set. This fits with NCA, a molecule that exhibits the same type of protons as aspartate, only slightly perturbed by carbamoylation. The identification of the two sets is fully consistent with COSY correlation (Fig. S2). Minor peaks correlate as one set (Asp) whereas all major peaks (NCA) correlate together. In each set, Hα is correlated to Hβ with a J3H-H. In addition, the two Hβ are correlated to each other with a J2H-H. 13C NMR confirms the assignments (Fig. S3). A DEPT 135 acquisition (Fig. S4) discriminated the signals amongst each set. An extra 13C signal at 162.0 ppm with no correlation with hydrogen atoms was identified as a carbonate resulting from cyanate partial decomposition. HSQC and HMBC correlations (Fig. S5 and S6) helped to link carbons and hydrogens without ambiguity. These results provide a clear spectroscopic signature of a successful NCA formation (Fig. 2). By integrating proton signals, the average NCA yield after 16 h is 92%. Upon further evolution, it is found to be 88% after 23 days (Figs. S7 and S8) and 91% after one year (Fig. S9). Thus, the NCA remained stable (or rather metastable) for a long period of time, although the pH value drifted from 6.96 to 8.66, then 9.41, respectively. For comparison, using carbamoyl phosphate (instead of cyanate) and aspartate at 8.3 mM concentration and pH 8, Yi et al. obtained NCA in a 37% yield. Two experiments were carried out under the same conditions (see Methods section) as the preceding one but cyanate was replaced by other potential carbamoyl donors. Biuret, issued from the dimerization of urea, is also relevant in prebiotic scenarios. Theoretically, each biuret molecule could carbamoylate two aspartates following the two successive reactions shown in Fig. 4: thus, although the “carbamoylating agent” concentration is the same as in the preceding experiment, the carbamoyl group/aspartate ratio is 2:1 instead of 1:1. After 16 h at 25 °C, the reaction products were analysed by 1H NMR (Fig. S10). Only one set of four signals was present and attributed to aspartate (Fig. 3). This lack of reactivity is confirmed by a COSY correlation (Fig. S11). It can be concluded that no detectable amount of NCA is formed in 16 h. Evolution was monitored for longer periods: after 23 days and even after 10 months, only unmodified aspartate was detected (Figs. S12 and S13). Since the expected yield of this reaction at equilibrium would theoretically be 99.98%, based on the reaction ΔrG°’ (standard transformed molar free enthalpy of reaction, − 44.1 kJ/mol), we can conclude that the biuret-aspartate reaction has extremely slow kinetics. We cannot exclude that for very long time scale (longer than several years) carbamoylation by biuret could occur but, at any rate, kinetic competition would favour cyanate as a carbamoylating reagent—not to mention the fact that the rate of NCA decomposition by amide bond hydrolysis would be comparable to its rate of formation. Urea was also checked as a potential carbamoylating agent but under the considered conditions (16 h, 25 °C), aspartate remained unmodified (Fig. S14). Carbamate itself (-O2C-NH2) cannot be an efficient carbamoylating agent because its free enthalpy content is too low: the calculated ΔrG°’ of NCA formation from carbamate and aspartate is largely positive at all pH (Fig. 5). One could have thought, however, that the addition of energy-rich reagents such as phosphorylating agents could allow aspartate carbamoylation by carbamate. Trimetaphosphate chemistry seems promising in this respect since this molecule and its degradation products have been widely used in prebiotic chemistry; it has even been suggested that the “RNA world” was characterized by the use of trimetaphosphate as an energy source. To check this idea, a solution of phosphorylating agents (generated from trimetaphosphate ammonolysis at 70 °C for 66 h) was reacted at 100 °C with L-aspartic acid and ammonium carbamate and analysed by 1H and 13C NMR. The targeted reaction—condensation of aspartic acid with carbamate—did not occur, but the initial signals of aspartate were nonetheless profoundly altered. The doubling of each 1H NMR signal from L-aspartic acid alone can be interpreted as evidence of dimerization. The split signals of L-aspartic acid present in 1H NMR could be unambiguously attributed to an Asp-Asp dimer (Figs. S15 and S16), that we assigned to the species β-Asp-Asp. Thermally obtained polyaspartic acid has a α/β ratio of 1/1.3. Thermal polymerization is therefore hardly selective; in contrast, our reaction seems to form a single compound. Thus, the free enthalpy contained in phosphoramidate was transferred to other high-energy bonds, namely the peptide bonds in the Asp-Asp dimer, affording another example of the versatility of phosphoramidates as coupling agents highlighted by Osumah and Krishnamurthy; but it did not result in NCA formation. Amorphous silica (SiO2) is well-known to be efficient for prebiotic condensation reactions, especially those of amino acids to peptides, and is a prebiotically realistic mineral. It is a moderately acidic catalyst, its catalytic properties relying on H-bonding and silanol acidity. Thus, we co-adsorbed aspartate on silica with cyanate or urea, but also with carbamate and with ammonium carbonate. Most of these potential carbamoylating agents were eliminated by drying at room temperature, i.e., before they could react with the aspartate molecules, as revealed by transmission IR spectra of the dried systems (Fig. S17). For the (aspartate + cyanate), (aspartate + carbamate) and (aspartate + carbonate) systems, only the bands of adsorbed aspartate were observed, indicating that the other partner was lost to the gas phase. For the (aspartate + urea) on SiO2 system, the characteristic bands of urea are still visible after drying, but a subsequent TG + MS study indicated that urea decomposes thermally at 140 °C without reaction with aspartate. Since carbamoylation failed on acidic silica, we turned to a mineral support with basic properties. Brucite (Mg(OH)2) would be a candidate, already studied for aspartate adsorption and reactivity. However, in the presence of a CO2-rich atmosphere, brucite is carbonated to magnesite (MgCO3), and so is periclase (MgO). Therefore, we chose magnesite as a support. Also, aspartic acid and Mg2+ ions from MgCO3 dissolution rapidly form the aspartate complex, which prompted us to select Mg(Asp)2 for Asp deposition. Despite being marketed as MgCO3, powder XRD studies showed the mineral phase after impregnation and drying under nitrogen is always hydromagnesite (Mg5(CO3)4(OH)2·4H2O) (Fig. S19). No crystallisation of organic molecules was observed by XRD in biomolecules-impregnated hydromagnesite (Figs. S20, S22, S24 and S26). After impregnation followed by drying at variable temperatures (Table S1), solid samples were directly desorbed with deuterated water. The desorption solutions were analysed by 1H (Figs. S25, S27, S28 and S29) and the spectra were compared with those of reference compounds (Figs. S18, S21 and S23). Hydromagnesite is partly soluble, so chances for organic molecules to stay adsorbed on the mineral and unanalyzed are negligible. According to 1H signal integration, a significant amount of NCA was already observable after drying at room temperature, and the NCA yield was improved by thermal treatment. An optimal yield seems to be reached at 150 °C (Fig. S27, Table S1). At 230 °C, new signals appeared, probably decomposition products of NCA and poly-Asp or poly-succinimide (Fig. S29). Two reaction attempts of biuret and magnesium di-aspartate on magnesite were tried, at 25 °C and 140 °C. No condensation reaction occurred, as indicated by 1H NMR (Figs. S30 and S31). The thermodynamic and kinetic parameters of biological compounds formation are considered with much care in many biochemical studies, especially in the two subdomains of bioenergetics and enzymatic catalysis. In contrast, they are often neglected in origins of life studies, although rigorous and insightful data have been published in some cases. A first rough indicator of whether a particular prebiotic reaction is thermodynamically possible is the standard (Gibbs) free energy of reaction, ΔrG°, or the transformed value at constant pH, ΔrG°’. Figure 5 displays the theoretically expected ΔrG°’ for NCA formation reactions from aspartate and several possible carbamoylating reagents in aqueous solution as a function of solution pH. These estimates permit differentiation between two sets of potential carbamoylating agents. With carbamate and urea, the reaction should be endergonic, although only slightly in the case of urea, and therefore these two molecules (especially carbamate) are not likely contenders for carbamoylation in solution. In contrast, aspartate carbamoylation by carbamoyl phosphate (the current biochemical agent), cyanate, and biuret should be exergonic. These values are pH-dependent, strongly so for biuret and cyanate; reaction with the latter becomes unfavourable at pH < 4, and conversely it is a more promising carbamoylating agent at high pH. While the ΔrG° (resp. ΔrG°’) allow to calculate the equilibrium constants K (resp. K’), the equilibrium reaction yields will depend, in addition to the pH, on the initial concentrations of all reagents. For instance, reaction c) with cyanate is a simple condensation and should proceed further in the forward direction if the total concentrations are increased. As an illustration, Fig. 6 shows the expected NCA yields (at equilibrium) as a function of pH for an initial 1:1 molar ratio of cyanate to aspartate, and concentrations of 50 × 10–3 mol.L-1 (our conditions) and 25 × 10–6 mol.L-1. In the conditions we used, the carbamoylation yield is expected to be at least 96% for all pH ≥ 5, and even in the micromolar range, it should be > 50% for pH ≥ 6. In the same way, excellent yields are expected using carbamoyl phosphate, and even more using biuret. Experimentally, aspartate carbamoylation with cyanate results in a 92% yield after 16 h, and this composition remains stable for over a year. While this is not as high as the theoretically expected yields in the experimental pH range (> 99.3%), the reaction is indeed highly favoured, and seems to reach equilibrium in a few hours. A very different situation is observed with biuret. While even higher carbamoylation yields would be expected, no reaction is detected even after 10 months. The kinetics of this reaction must therefore be extremely slow—no kinetic data were available in the literature to predict this conclusion. With urea, carbamoylation is expected to be only slightly endergonic, so that a > 30% yield would be predicted in our conditions. After 16 h, no reaction is observed, so this reaction is not very fast either. In previous studies, Fox obtained NCA from Asp and urea in “hot water” with solid mineral bases. High yields (46–80%) were obtained; fast kinetics are the consequence of high temperatures, but certainly also of specific catalytic effects of the solid phases. Yamagata et al. reacted aspartic acid with urea at pH 7 and 90 °C, in an open flask that allowed for constant elimination of the NH3 product (open system). In these conditions, about 80% of the aspartate transformed into NCA, and further products formed from it, in a matter of days. This would mean that the carbamoylation reaction by urea, while slow at 25 °C, is reasonably fast at 90 °C. In summary, aspartate carbamoylation is an illustration of the fact that in prebiotic chemistry, as in all branches of chemistry, kinetic and thermodynamic aspects must be considered in order to assess the feasibility of any given scenario. Thermodynamic data are available for many reactions; they tell us that carbamoylation by carbamate is not a feasible pathway, and by urea hardly so. Of course, an endergonic reaction can occur if it is coupled with an exergonic one, as happens often in biochemistry. However, such a coupling may only be considered if a convenient reaction pathway is available to induce coupling between the reagents: introducing the energy-rich trimetaphosphate in the system did allow an endergonic reaction, but it was aspartate dimerization instead of carbamoylation. In contrast, thermodynamics tells us that both cyanate and biuret are possible carbamoylating agents. Here, kinetics comes into play. Usually kinetic studies have to be completely implemented since few such studies are available, even for biochemically important molecules. Only the cyanate reaction is promising from both the thermodynamic and kinetic points of view. We have not studied the biochemical carbamoylating agent, carbamoyl phosphate (CP), because a previous study has led us to conclude that it is unlikely to be present in prebiotic conditions, or, if formed, would probably isomerize to cyanate before reacting. Carbamoylation of aspartate by cyanate has all the characteristics of a “good” prebiotic reaction. It is strongly favoured because of a negative ΔrG°; there exists a fast reaction path even in the absence of specific catalysis, and its product, NCA, while still energy-rich, is kinetically inert (not prone to decomposition even on the scale of several months), so that it will remain available for futher transformations for a long time after being formed. This is illustrated in Fig. 7 where fast reactions are represented by full arrows and slow ones by dashed arrows. This constitutes a simple example of how the elucidation of reaction thermodynamics and kinetics allows pruning a potentially complex network of prebiotic reaction pathways. As mentioned, silica has been known to constitute a good platform for condensation reactions between molecules. The reason for this is twofold. First, from the thermodynamic point of view, silica allows to work in conditions of low water activity, by simply applying a drying step. Because water is a product of the condensation reaction, continuously removing it drags the condensation to the right according to Le Châtelier’s principle, and this is also true for any volatile condensation product. Therefore, carbamoylation by all the agents we considered could in principle be made favourable on silica (as opposed to the solution where only some agents are allowed). All it takes is for the surface to be thoroughly dehydrated, which in the case of fumed silica occurs at about 100 °C. Second, from the kinetic point of view, silica catalyses the nucleophilic addition that constitutes the first step of the condensation. Although in-depth studies have been devoted to the complexity of this phenomenon, the basic mechanism probably implies weakly acidic surface silanols activating the leaving group (silica behaves as a weak solid Brönsted acid), as schematically illustrated in Fig. 8a for the aspartic acid/carbamic acid condensation. It is a basic principle of kinetics that everything that catalyses a reaction in the forward direction also catalyses it in the reverse direction. For instance, if silica catalyses a condensation that liberates one water molecule, it will also catalyse the reverse reaction of hydrolysis: which one will actually occur depends on the water activity in the system. In fact, in the present case additional pathways may be considered for the reverse reaction as illustrated in Fig. 8b for the hydrolysis of carbamic acid. Thus, we can understand the failure to observe carbamoylation reactions on silica. All carbamoylating agents already contain one C-N bond, which is usually quite labile. Silica is actually too good a catalyst: it will activate the C-N bond in the carbamoylating agent at low temperatures where water is still abundant on the surface, and thus cause its hydrolysis before it can functionalize the aspartic acid, a reaction that has a higher energy barrier. Carbamoylating scenarios involving mineral surfaces are not completely excluded since NCA was observed from (cyanate + aspartate) deposited on MgCO3. We can say therefore that whatever catalytic sites are present on the MgCO3 surface (it would be expected to exhibit at least general basic catalysis), they are not able to significantly activate amide bonds at room temperature. They might still activate condensation at higher temperatures, as the NCA yields increase somewhat after heating at 150 °C, although remaining lower than those obtained in a mineral-free reaction. On the other hand, excessive heating causes a degradation of the NCA product, in a matter of hours at 230 °C: this represents the upper limit of the temperature range available for the use of the prebiotic precursor NCA. In this work, we have shown that among the various carbamoyl donors that could have been present in prebiotic conditions, cyanate is probably the most efficient to perform aspartate carbamoylation. This reaction has all the features of a “good” prebiotic reaction, and only necessitates precursors that are known to be formed prebiotically. Our results underline the necessity of always considering both the thermodynamic and kinetic dimensions of potential prebiotic reactions. For instance, from a thermodynamic point of view, biuret should provide high carbamoylation yields, but the reaction rate at room temperature is so low that the corresponding scenario appears unlikely. In the same way, the input of chemical energy from a well-known inorganic precursor, trimetaphosphate, did result in energy transduction—but through the formation of an aspartate dimer, rather than the desired product, NCA. In this case, a different reaction pathway was opened. While thermodynamic data on many biomolecules and their precursors are often available, at least in standard conditions, kinetic data are scarce and reaction kinetics must systematically be assessed. The answer to the question asked at the beginning of this study—is aspartate carbamoylation a likely step in prebiotic pyrimidines synthesis?—is clearly positive, provided that cyanate is used as the carbamoylating agent. Thus, the geochemical continuity hypothesis is only partially vindicated. While the reaction could very well have happened in prebiotic settings, it is likely to have used cyanate rather than the carbamoylating agent of current biochemistry, carbamoyl phosphate, which was probably not easily available. An aqueous phase scenario is quite likely. Mineral surfaces can actually be detrimental if they are acidic as is the case for silica. On the other hand, the basic hydromagnesite did not prevent carbamoylation by cyanate, but neither did it bring clear benefits. However, once NCA is formed, deposition on a mineral surface could stabilize it against thermal degradation, with respect to the species in solution. This aqueous solution/mineral surface versatility of the carbamoylation reaction could be part of a primitive pond scenario. While we are trying to replicate the orotate pathway of extant biochemistry in a geochemical context, we do not want to rule out alternative sources of orotate. Indeed, Krishnamurthy et al. have investigated a different one-pot aqueous scenario, in which orotate is obtained from hydantoin and glyoxylate. Several pathways to orotate may have once coexisted, with the current one being selected at some stage of chemical evolution. In a forthcoming article, we will concentrate on replicating the next step in the orotate pathway, namely NCA cyclization, as it represents a key metabolic crossroads in current biochemistry. The following chemical compounds were purchased from commercial suppliers and used without further purification: ureidosuccinic acid (Sigma-Aldrich Co., cat. n°69037-500MG), biuret (Sigma-Aldrich Co., cat. n°15270-25G), L-aspartic hemimagnesium salt dihydrate (Sigma-Aldrich Co., cat. n°11260-100G), L-aspartic acid (Sigma-Aldrich Co., cat. n°A8949-100G), magnesium carbonate basic (Sigma-Aldrich Co., cat. n°13118-1 KG) with a Brunauer–Emmett–Teller surface area of 31.5 m2/g, sodium cyanate (Sigma-Aldrich Co., cat. n°185086-100G), sodium hydroxide anhydrous pellets (Carlo Erba reagents, cat. n°480507), deuterium oxide 99.90% D (Eurisotop, cat. n°D214FE), linear dimer H-Asp-Asp-OH (Bachem, cat. n°4010210.0250), ammonium carbamate (Sigma-Aldrich Co., cat. n°292834-100G), ammonia 28% analaR Normapur (VWR Chemicals, cat n°21190.292), trisodium trimetaphosphate (Sigma-Aldrich Co., cat. n°T5508-500G), urea ACS reagent (Sigma-Aldrich Co., cat. n°U5128-100G), ammonium carbonate ACS reagent (Aldrich chemical company, Inc., cat. n°20786-1), carbamoyl phosphate disodium salt (Sigma-Aldrich Co., cat. n°C4135-1G), fumed silica Aerosil 380 (Evonik Industries), with a Brunauer–Emmett–Teller surface area of 380 m2/g. The Gibbs free energies of reactions were calculated by using the eQuilibrator calculator at http://equilibrator.weizmann.ac.il/. NMR experiments were conducted on a Bruker Avance III 500 spectrometer (ωL = 500.07 MHz for 1H and 125.74 MHz for 13C) equipped with a 5 mm inverse double resonance broadband probe. Chemical shifts were calibrated as δ values (ppm) relative to the peak of TMS set at δ = 0.00 ppm (13C NMR), residual light water in D2O set at δ = 4.79 ppm (1H NMR). Coupling constants are given in Hertz. All spectra were processed with the Bruker TopSpin 4.0.6 and 4.0.8 softwares. The used 2D correlations were the following: 1H-1H COSY (COrrelated SpectroscopY), 1H-13C HSQC (Heteronuclear Single Quantum Coherence) and 1H-13C HMBC (Heteronuclear Multiple Bond Correlation). The used DEPT 135 (Distortionless Enhancement by Polarization Transfer) experiment gives inverted CH2 and C groups. pH measurements were carried out using a Fischer Scientific Accumet AE150 pH Benchtop Meter. A small portion of each sample was ground and mounted on a zero-background holder. The X-ray powder diffraction data were registered at room temperature with a D8 DISCOVER Bruker diffractometer at Sorbonne University. This instrument is equipped with a Cu anode (Kα1 and Kα2 copper doublet) source operated at 40 kV and 30 mA and a LynxEye XE-T 1D detector. The data were recorded from 5° to 80° in 1 h with steps of 0.02° L-Aspartic acid (6.50 g, 50 mmol, 1 eq.) and sodium cyanate (3.21 g, 49 mmol, 1 eq.) were dissolved into a 1 M sodium hydroxide solution (50 mL). The resulting mixture was stirred and allowed to stand for 16 h at room temperature, then sampled (400 µL into 200 µL D2O) for NMR analyses. L-Aspartic acid (6.50 g, 50 mmol, 1 eq.) and urea (3.00 g, 50 mmol, 1 eq.) were dissolved into a 1 M sodium hydroxide solution (50 mL). The resulting mixture was stirred and allowed to stand for 16 h at room temperature, then sampled (400 µL into 200 µL D2O) for NMR analyses. L-Aspartic acid (6.50 g, 50 mmol, 1 eq.) and biuret (5.08 g, 49 mmol, 1 eq.) were dissolved into a 1 M sodium hydroxide solution (50 mL). The resulting mixture was stirred and allowed to stand for 16 h at room temperature, then sampled (400 µL into 200 µL D2O) for NMR analyses. L-Aspartic acid (173 mg, 1.3 mmol, 1 eq.) and ammonium carbamate (132 mg, 1.7 mmol, 1.3 eq.) were dissolved into 2.5 mL of an ammonia solution containing phosphoramidates (sampled after 66 h at 70 °C, preparation described in). The solution was stored in a 25 mL airtight bottle in an oven at 100 °C for 16 h. 600 µL were collected for NMR analyses. To deposit organic molecules on minerals, we used a wetting-and drying procedure, also called “incipient wetness impregnation” or IWI in the heterogeneous catalysis literature. It implies putting a solution containing an Asp precursor and carbamoyl donor in contact with a mineral surface to obtain a paste or slurry. During this process, solubilised organic compounds are in prolongated contact with the dispersed mineral and the slurry is subsequently dried at room temperature. Similar events could happen in a geochemical scenario due to climatic fluctuations. L-Aspartic acid (20 mg, 0.13 mmol, 1 eq.) and a carbamoyl donor* (0.20 mmol, 1.3 eq.) were dissolved into 5 mL distilled water under stirring. 500 mg of silica Aerosil were impregnated by the solution. The resulting slurry was dried under N2, until a dry powder was obtained. Additional treatment in a desiccator under vacuum removed water traces. The powder was ground and sampled for IR analyses. *urea (12 mg), ammonium carbonate (19 mg), ammonium carbamate (15 mg), sodium cyanate (13 mg) or sodium carbamoyl phosphate (27 mg). Magnesium di-aspartate (24 mg, 0.074 mmol, 2 eq. Asp) was dissolved into 6 mL distilled water under stirring. 500 mg (5.9 mmol, 39 eq.) of magnesium carbonate (5.9 mmol) were impregnated by the solution. The resulting suspension was dried under N2, until a dry powder was obtained. The powder was ground and sampled for XRD and NMR analyses (30 mg suspended into 600 µL D2O). Asp-Asp dimer (24 mg, 0.15 mmol) was dissolved into 6 mL distilled water under stirring. 500 mg of magnesium carbonate were impregnated by the solution. The resulting suspension was dried under N2, until a dry powder was obtained. The powder was ground and sampled for XRD and NMR analyses (30 mg suspended into 600 µL D2O). Ureidosuccinic acid (22 mg, 0.12 mmol) was dissolved into 6 mL distilled water under stirring. 500 mg of magnesium carbonate (5.9 mmol) were impregnated by the solution. The resulting suspension was dried under N2, until a dry powder was obtained. The powder was ground and sampled for XRD and NMR analyses (30 mg suspended into 600 µL D2O). Magnesium di-aspartate (24 mg, 0.074 mmol, 2 eq. Asp) and sodium cyanate (13 mg, 0.20 mmol, 1.4 eq.) were dissolved into 6 mL distilled water under stirring. 500 mg (5.9 mmol, 39 eq.) of magnesium carbonate (5.9 mmol) were impregnated by the solution. The resulting suspension was dried under N2, until a dry powder was obtained. The powder was ground and sampled for thermal activations. 10 mg of the powder were thermally activated at four different temperatures (150 °C, 200 °C and 230 °C) for 30 min in an oven fitted with desiccants. Hygrometry measurement at 88 °C is 3.5% RH. After cooling down in a desiccator under reduced pressure, the samples were analysed by NMR (10 mg suspended into 600 µL D2O). Magnesium di-aspartate (24 mg, 0.074 mmol, 2 eq. Asp) and biuret (21 mg, 0.20 mmol, 1.4 eq.) were dissolved into 6 mL distilled water under stirring. 500 mg (5.9 mmol, 39 eq.) of magnesium carbonate (5.9 mmol) were impregnated by the solution. The resulting suspension was dried under N2, until a dry powder was obtained. The powder was grinded and sampled for thermal activations. 10 mg of the powder were thermally activated at 140 °C for 30 min into an oven packed with desiccants. Hygrometry measurement at 88 °C is 3.5% RH. After cooling down into a reduced pressure desiccator, samples were analysed by NMR (10 mg suspended into 600 µL D2O). Supplementary Information.

PMC9649778

Nienke Willemsen,Stefan Kotschi,Alexander Bartelt

Fire up the pyre: inosine thermogenic signaling for obesity therapy

11-11-2022

Endocrine system and metabolic diseases,Physiology,Target identification

Fire up the pyre: inosine thermogenic signaling for obesity therapy In a recent study published in Nature, Niemann et al. may have discovered a metabolite signaling pathway that could pave the way to new weight loss drugs (Fig. 1). Obesity and its comorbidities are a major threat to public health, but efficient therapeutics are still scarce. Brown adipocytes are fat cells that use large amounts of chemical energy in the form of fatty acids and glucose to generate heat, a process called non-shivering thermogenesis. It is well established that people with higher brown adipose tissue (BAT) activity are generally healthier but how this can be exploited therapeutically is to be determined. While norepinephrine is the primary stimulator, many other signals have been discovered to participate in the complex physiological regulation of thermogenesis. Niemann et al. set out to analyze the metabolome released by brown adipocytes undergoing apoptosis, a process linked to adipose tissue remodeling. They identified inosine, a purine-derived nucleoside, as one of the most prominent metabolites. The supernatant of apoptotic brown adipocytes and specifically inosine by itself, were sufficient to activate thermogenic signaling in brown adipocytes via the GPCR purinergic receptors A2A and A2B. In mice, injecting inosine increased energy expenditure and induced weight loss, which is strikingly similar to the natural thermogenic response mediated by norepinephrine. To decipher the mechanism of inosine-mediated thermogenesis, the authors investigated the inosine importer ENT1, encoded by Slc29a1. Lack of ENT1 or pharmacological inhibition of ENT1 with dipyridamole replicated the thermogenic effect of inosine. These mechanisms were also present in human adipocytes, and ENT1 was linked to human obesity on multiple levels. These results implicate inosine-derived therapeutics or compounds targeting ENT1 as new weight loss drugs. The work by Niemann et al. supports the concept that the involution of BAT induces cell death, which can be inhibited by adrenergic activation. It would follow that disuse of BAT in humans during weight gain was linked to higher rates of apoptosis. This would explain the observation that human obesity is associated with diminished BAT activity. While the authors showed an increase in apoptosis markers in BAT from mice exposed to thermoneutrality, the nature and cause of cell death under these conditions remain unclear. Inosine and potentially other signals from dying adipocytes could locally induce thermogenesis and stimulate the formation of more brown adipocytes. It is yet to be established if the same intercellular communication plays a role in the involution of BAT with warmer temperatures and high-fat diets, which is considered to be more similar to the human situation. Regardless, this work presents an elegant mechanism by which extracellular signals prevent complete loss of BAT during inactivity, thus preserving the plasticity of the tissue in response to cold. Interestingly, inosine increases intracellular cAMP levels and induces lipolysis in brown adipocytes. Lipolysis is also implicated in adipose tissue remodeling, as it was previously suggested that lipolysis-associated death of adipocytes leads to the recruitment of macrophages, which would then promote tissue remodeling and browning. Inosine could be a mediator of the interplay between adipocytes and macrophages, particularly in the context of obesity, in which adipose tissue is inflamed and displays substantial cell death. Thus, inosine could be driving paracrine communication to ensure tissue remodeling both in pathological (obesity) as well as physiological (cold) adaptations. Of note, extracellular ATP, along with other extracellular nucleotides, is a well-established “stench of decay” signal that is implicated in physiological and immune responses in a wide spectrum of cell types and tissues. The in vivo effects of inosine treatment on energy metabolism were mimicked by either global genetic deletion or inhibition of ENT1, which both increased extracellular inosine levels. However, ENT1 is rather promiscuous, and other metabolites could be involved in the ENT1 mechanism. The authors showed that either approach increased energy expenditure and Ucp1 levels in adipose tissues, indicating enhanced browning. ENT1 was recently identified as a brown adipocyte marker that is upregulated during cold, so it is somewhat surprising that ENT1 suppression increases browning. It should be taken into consideration that inosine could also come from and act on non-adipocytes. To address this, the authors analyzed adipocyte-specific ENT1 knockout mice, but obesity in these animals, unlike in the global deletion model, was independent of ENT1. This indicates that the role of ENT1 in obesity is complex. In line with the mouse data, the authors found that in human adipose tissue, ENT1 mRNA levels were inversely correlated with UCP1 gene expression. Furthermore, they identified the hypofunctional ENT1 missense gene variant p.Ile216Thr that was associated with lower BMI in humans. While these findings support an important role for ENT1 in energy metabolism, more work is needed to fully understand the cell type-specific roles of ENT1 and inosine in human obesity. The thermogenic effects of inosine place it into the spotlight as a potential weight loss drug. Inosine is already in clinical trials as a neuroprotective agent, as it is metabolized to uric acid, which has anti-oxidative effects. So far, no metabolic benefit has been observed during these oral supplementation studies, but obesity was not specifically investigated. As the metabolization of inosine prevents its function as an extracellular signal, a more direct application of inosine might be needed to induce its thermogenic effects. Likewise, dipyridamole, which is clinically used as a blood vessel dilator and antiplatelet drug, has not been reported to cause weight loss to our knowledge. It will be challenging to effectively increase extracellular inosine levels in desired tissues in humans to study the weight loss potential of the metabolite. In conclusion, Niemann et al. have identified inosine as a new metabolically active “death smell” messenger with important implications for our understanding of communication in the tissue microenvironment, adipose remodeling, and energy metabolism.

PMC9649783

Xiaodong Zou,Hongsheng Ouyang,Feng Lin,Huanyu Zhang,Yang Yang,Daxin Pang,Renzhi Han,Xiaochun Tang

MYBPC3 deficiency in cardiac fibroblasts drives their activation and contributes to fibrosis

10-11-2022

Cardiac hypertrophy,Mechanisms of disease

Genetic mutations in the MYBPC3 gene encoding cardiac myosin binding protein C (cMyBP-C) are the most common cause of hypertrophic cardiomyopathy (HCM). Myocardial fibrosis (MF) plays a critical role in the development of HCM. However, the mechanism for mutant MYBPC3-induced MF is not well defined. In this study, we developed a R495Q mutant pig model using cytosine base editing and observed an early-onset MF in these mutant pigs shortly after birth. Unexpectedly, we found that the “cardiac-specific” MYBPC3 gene was actually expressed in cardiac fibroblasts from different species as well as NIH3T3 fibroblasts at the transcription and protein levels. CRISPR-mediated disruption of Mybpc3 in NIH3T3 fibroblasts activated nuclear factor κB (NF-κB) signaling pathway, which increased the expression of transforming growth factor beta (TGF-β1) and other pro-inflammatory genes. The upregulation of TGF-β1 promoted the expression of hypoxia-inducible factor-1 subunit α (HIF-1α) and its downstream targets involved in glycolysis such as GLUT1, PFK, and LDHA. Consequently, the enhanced aerobic glycolysis with higher rate of ATP biosynthesis accelerated the activation of cardiac fibroblasts, contributing to the development of HCM. This work reveals an intrinsic role of MYBPC3 in maintaining cardiac fibroblast homeostasis and disruption of MYBPC3 in these cells contributes to the disease pathogenesis of HCM.

MYBPC3 deficiency in cardiac fibroblasts drives their activation and contributes to fibrosis Genetic mutations in the MYBPC3 gene encoding cardiac myosin binding protein C (cMyBP-C) are the most common cause of hypertrophic cardiomyopathy (HCM). Myocardial fibrosis (MF) plays a critical role in the development of HCM. However, the mechanism for mutant MYBPC3-induced MF is not well defined. In this study, we developed a R495Q mutant pig model using cytosine base editing and observed an early-onset MF in these mutant pigs shortly after birth. Unexpectedly, we found that the “cardiac-specific” MYBPC3 gene was actually expressed in cardiac fibroblasts from different species as well as NIH3T3 fibroblasts at the transcription and protein levels. CRISPR-mediated disruption of Mybpc3 in NIH3T3 fibroblasts activated nuclear factor κB (NF-κB) signaling pathway, which increased the expression of transforming growth factor beta (TGF-β1) and other pro-inflammatory genes. The upregulation of TGF-β1 promoted the expression of hypoxia-inducible factor-1 subunit α (HIF-1α) and its downstream targets involved in glycolysis such as GLUT1, PFK, and LDHA. Consequently, the enhanced aerobic glycolysis with higher rate of ATP biosynthesis accelerated the activation of cardiac fibroblasts, contributing to the development of HCM. This work reveals an intrinsic role of MYBPC3 in maintaining cardiac fibroblast homeostasis and disruption of MYBPC3 in these cells contributes to the disease pathogenesis of HCM. Hypertrophic cardiomyopathy (HCM) is a heterogeneous group of diseases affecting people of both genders and of various ethnic and racial origins [1]. HCM is mainly caused by autosomal dominant mutations in sarcomere genes and disseminated with a prevalence of about 1:500–1:200 in the general population [2]. The clinical manifestations of HCM include left ventricular (LV) hypertrophy, myocardial hypercontractility, reduced compliance, myofibrillar disarray, and fibrosis [3]. During the last two decades, a wealthy body of evidence revealed MYBPC3 as the most frequently mutated HCM gene, representing about 40–50% of all HCM mutations [4]. The MYBPC3 gene encodes cardiac myosin binding protein C (cMyBP-C), a flexible, rod-like protein that is a key component of the cardiac sarcomere. Among HCM patients with genetic defects in MYBPC3, 90% of mutations are heterozygous frameshift, nonsense, or splice site mutations that result in premature termination codons (PTCs) and truncated cMyBP-C protein [5]. Previous studies reported that the knockout of mybpc3 caused zebrafish displayed significant morphological heart alterations at systolic and diastolic states at the larval stages, and revealed that an impaired actin cytoskeleton organization as the main dysregulating factor associated with the early ventricular cardiac hypertrophy in the zebrafish mypbc3 HCM model [6]. Moreover, the loss of cMyBP-C protein results in left ventricular dilation, cardiac hypertrophy and impaired ventricular function in Mybpc3 null mice [7]. Microarray analysis on left ventricles of wild-type (WT) and cMyBP-C−/− mice at postnatal day (PND) 1 and 9 (before and after the appearance of an overt HCM phenotype) identified the importance of extracellular matrix pathways in hypertrophic growth and early dysregulation of potassium channels [8]. Cardiac fibrosis is a hallmark of most myocardial pathologies with limited treatment options [9]. Cardiac fibrosis-related diseases are associated with a high mortality rate and the morbidity increases with age [10]. Myocardial fibrosis (MF), as evidenced by the proliferation of cardiac fibroblasts (CFs) and excessive deposition of collagen in myocardial tissue [11], has proven to be an important marker and determinant in the pathogenesis of HCM [12]. Several signaling pathways have been implicated in the early activation of CFs. The TGF-β family of growth factors are the most extensively studied mediators of fibroblast activation, of which TGF-β1 plays a crucial role in pathological fibrosis [13]. The canonical pathway of TGF-β1 signaling involves the phosphorylation of Smad2/3, which then binds Smad4, translocates into the nucleus, and acts as a transcription factor, inducing the activation of numerous profibrotic genes [14]. In addition to the Smad-mediated pathways, TGF-β1 can also induce noncanonical signaling that involves the activation of TGF-β-activated kinase (TAK) 1, which is thought to contribute to pathological cardiac remodeling. Cardiac overexpression of constitutively active TAK1 induces cardiac hypertrophy and heart failure [15]. A growing body of evidence suggests that the noncanonical pathway may actually be the predominant driving force [16]. The TGF-β noncanonical signaling pathway is thought to propagate primarily through the type II TGF-β receptor, as supported by the evidence that cardiomyocyte-specific deletion of the type II TGF-β receptor resulted in reduced fibrosis and remodeling in the transverse aortic constriction model of heart failure [17]. In this study, we generated a pig model of HCM with R495Q mutation in MYBPC3 engineered with cytosine base editor (CBE) and observed severe MF in these mutant piglets soon after birth. Unexpectedly, we found that MYBPC3 and several other related myocardial genes were expressed in fibroblasts. We further showed that CRISPR-mediated genetic disruption of MYBPC3 in fibroblasts promoted their activation into myofibroblasts via the NF-κB/TGF-β1/HIF-1α aerobic glycolysis signal cascade. First, we established MYBPC3 R495Q mutant porcine fetal fibroblasts (PFFs) by CBE, cell monoclonal techniques and Sanger sequencing (Supplemental Fig. S1a). Electroporation of CBE and the MYBPC3-targeting sgRNA resulted in the conversion of CCG to CTG in the exon 16 of pig MYBPC3 (Supplemental Fig. S1b). The R495Q heterozygous PFFs were used to generate the mutant pigs via somatic cell nuclear transfer (SCNT). In total, three pregnant surrogates were carried to term, and six piglets were delivered. All piglets were genotyped by sequencing (Supplemental Fig. S1c), and three of newborns carried mutations at the target locus. Sequencing analysis of the six top predicted off-target sites showed no detectable editing (Supplemental Fig. S1d). The MYBPC3 R495Q mutation piglets began to die soon after birth, with all three dead within 10 days, whereas all wild-type piglets survived. Quantitative RT-PCR (qRT-PCR) analysis showed that the expression of MYBPC3 was significantly decreased in cardiac muscle of R495Q mutant pigs as compared with WT littermates (Fig. 1A). Consistently, Western blot analysis showed that the cMyBP-C protein was significantly reduced in the R495Q mutant pig hearts (Fig. 1B, C and Supplemental Fig. S7). Serological examination showed that the high sensitivity cardiac troponin T (cTNT) and procollagen type I carboxy-terminal propeptide (PICP), which are highly sensitive markers for myocardial injury and fibrosis, respectively [18, 19], were significantly elevated in the mutant pigs (Fig. 1D, E). In addition, the transcript and protein expression levels of the fibrosis-related genes such as type I collagens α1 (COL1A1), α-Smooth muscle actin (α-SMA) and profibrotic cytokines (TGF-β1) were significantly up-regulated in the mutant pig hearts (Fig. 1A–C). We also observed inflammatory infiltrates in heart tissue sections (Supplemental Fig. S1e) and the abnormally elevated pro-inflammatory cytokines in serum samples of the mutant pigs (Supplemental Fig. S1f–i). Remarkably, lactic acid concentrations in both serum (Fig. 1F) and myocardial muscles (Supplemental Fig. S1j) of the mutant pigs were highly elevated. Furthermore, Masson’s trichrome staining of paraffin sections showed severe myocardial fibrosis in the mutant pigs (Fig. 1G, H). Taken together, these results demonstrated that the MYBPC3 R495Q mutation led to the development of severe myocardial fibrosis and inflammation. The early onset MF phenotype in the mutant pigs suggests a potential intrinsic role of MYBPC3 deficiency in regulating fibroblast trans-differentiation. We first examined the expression of MYBPC3 in the major cell types within mouse heart, including mouse cardiac myocytes (CM), fibroblasts (CF), microvascular endothelial cells (MVET), aortic smooth muscle cells (ASM) and coronary artery smooth muscle cells (CASM). The CM and CF were isolated from neonatal C57BL/6J mice (day 1–3), while MVET, ASM, and CASM were isolated from 3–6 weeks C57BL/6J mice. Quantitative RT-PCR showed that Mybpc3 was expressed in CM and CF but not in other cell types (Supplemental Fig. S2A). Western blot analysis confirmed that the MYBPC3 protein was expressed in CF, at ~18% of that in CM (Supplemental Fig. S2B, C). MYBPC3 protein was not detectable in other cell types. We then examined the expression of MYBPC3 in primary porcine cardiac fibroblasts (PCF). Immunofluorescence staining showed that both MYBPC3 and myosin light chain 3 (MYL3) were readily detectable in WT PCFs, which were also positive for the fibroblast marker vimentin (VIM) (Fig. 2A). The fluorescence signals were specific as we did not detect their expression in PFFs. To examine whether the expression of MYBPC3 and MYL3 in cardiac fibroblasts are species-specific, we analyzed their expression in primary human cardiac fibroblasts (HCF) and mouse cardiac fibroblasts (MCF) and found that both MYBPC3 and MYL3 were expressed in these cells (Fig. 2A). Moreover, we found that MYBPC3 and MYL3 were expressed in NIH3T3 cells, a commonly used mouse embryonic fibroblast cell line (Fig. 2A). Consistently, Western blot analysis confirmed the expression of the cardiac proteins MYBPC3, MYL3 and TNNT2 in NIH3T3 fibroblasts and primary mouse, pig and human cardiac fibroblasts, but not in PFFs (Fig. 2B and Supplemental Fig. S8). To understand the role of MYBPC3 in fibroblasts, we utilized the CRISPR/Cas9 system to knockout Mybpc3 gene in NIH3T3 mouse fibroblasts (MYBPC3-KO). A guide RNA (gRNA) targeting the exon 2 of mouse Mybpc3 gene was transfected into NIH3T3 cells together with the Cas9-expressing plasmid (Supplemental Fig. S3a). Sequencing of the genomic DNA PCR amplicon showed efficient generation of indels at the gRNA target site (Supplemental Fig. S3b) but not at the top 5 predicted off-target sites (Supplemental Fig. S3c). The Mybpc3 transcript and protein expression was disrupted in MYBPC3-KO fibroblasts (Supplemental Fig. S3e, f and S15). Next, we examined whether Mybpc3 disruption could induce fibroblast activation. As shown in Fig. 3A–C and Supplemental Fig. S9, Mybpc3 disruption significantly increased the transcription and protein expression of COL1A1 and α-SMA, the markers of myofibroblasts. Moreover, Mybpc3 disruption led to accelerated proliferation and reduced apoptosis, as evidenced by EdU staining and apoptosis analysis, respectively (Fig. 3D–F and Supplemental Fig. S4a). In addition, Mybpc3 disruption increased cells in the G2/M phase while slightly reducing cells in the G0/G1 phase (Fig. 3G and Supplemental Fig. S4b). The wound scratch assay showed that Mybpc3 deficiency enhanced the migration of NIH3T3 fibroblasts (Fig. 3H and Supplemental Fig. S4c). Finally, the secretion of TGF-β1 into the culture medium and its expression in cell lysate were significantly increased (Fig. 3I, J). Together, these data suggest that Mybpc3 disruption induces fibroblast activation. Previous research demonstrated that glycolytic reprogramming facilitate progression to, and maintenance of, the transformed myofibroblast state and that enhances contractility and cellular migration. Similarly, we observed that the glucose concentrations in the culture medium and cell lysate of MYBPC3-KO fibroblasts were decreased while lactic acid levels were increased more rapidly than the control fibroblasts (Fig. 4A–D). ATP production was also elevated in MYBPC3-KO fibroblasts (Fig. 4E). Moreover, the expression of GCK, PFKM and LDHA, the key enzymes involved in glycolysis, was significantly increased at both the transcriptional and translational levels in MYBPC3-KO fibroblasts as compared to WT controls (Fig. 4F–H and Supplemental Fig. S10). Treatment with the glucose analog 2-deoxy-d-glycose (2-DG) to inhibit glycolysis dramatically reduced the expression of glycolysis genes and fibrosis marker genes (COL1A1 and α-SMA) but had no impact on the expression of TGF-β1 (Fig. 4F–H). These results suggest that the loss of Mybpc3 gene is a driving force for metabolic reprogramming in fibroblasts. A potential crosstalk between HIF-1α and TGF-β1 has been proposed in driving fibrosis [20]. The expression of HIF-1α was significantly increased in MYBPC3-KO fibroblasts as compared with WT control (Fig. 5A–C). Treatment with SB-431542, a selective inhibitor of TGF-β type I receptor, dramatically suppressed the expression of HIF-1α in MYBPC3-KO fibroblasts (Fig. 5A, C and Supplemental Fig. S11), while the selective inhibitor of HIF-1α, Oltipraz, did not affect the expression of TGF-β1 (Fig. 5B, D and Supplemental Fig. S12). Both SB-431542 and Oltipraz attenuated glycolysis and fibrosis genes (Fig. 5A–D). These data suggest that the disruption of Mybpc3 increased the expression of HIF-1α via TGF-β1 in fibroblasts. TGF-β1 and NF-κB were reported to be involved in liver fibrosis [21]. The cross-talk between TGF-β and NF-κB signaling pathways was mediated through TAK1 and SMAD7 in a subset of head and neck cancers [22]. First, we examined the NF-κB signaling pathway in WT and MYBPC3-KO fibroblasts. The phosphorylated p65, an indicator of NF-κB activity, was found to be increased in MYBPC3-KO fibroblasts and the TGF-β1 inhibitor SB-431542 did not impact the activation of NF-κB signaling pathway (Fig. 6A, B and Supplemental Fig. S13). NF-kB plays a critical role in controlling the expression of pro-inflammatory cytokines. We found that the levels of CCL2, IL-1β, IL-6, and TNF-α in both culture media and cell lysates of MYBPC3-KO fibroblasts were all significantly increased in MYBPC3-KO fibroblasts, which were significantly inhibited by the NF-κB signaling pathway inhibitor, PDTC (Fig. 6C–F and Supplemental Fig. 5). PDTC also significantly reduced the expression of TGF-β1 and HIF-1α (Fig. 6G, H and Supplemental S14), suggesting that they are the downstream signaling of NF-κB. Taken together, these results suggest that Mybpc3 deficiency triggers the NF-κB-mediated signaling cascade in fibroblasts to drive their activation. In this study, we generated a pig model of HCM with MYBPC3 mutation by CRISPR/Cas9 gene editing technology and demonstrated that the MYBPC3 mutant pigs developed premature MF. We unexpectedly found that MYBPC3 and other myofilament proteins were expressed in primary cardiac fibroblasts and NIH3T3 fibroblasts. We generated MYBPC3-KO NIH3T3 fibroblasts to dissect the signaling pathways involved in fibroblast activation. Our results revealed a novel fibroblast-intrinsic mechanism that links MYBPC3 deficiency to the activation of NF-κB-mediated signaling cascade through TGF-β1, HIF-1α, and aerobic glycolysis, contributing to the development of premature MF in HCM (Fig. 7). In this study, we demonstrated that MYBPC3 gene mutations activated the NF-κB signal pathway. NF-κB is a key molecule in regulating TGF-β1 levels, and a potential crosstalk mechanism between NF-κB and TGF-β/Smad has been proposed in hepatic macrophages [23]. It has been reported that TGF-β1 and NF-κB were both involved in liver fibrosis [24]. The expression of TGF-β1 and the nuclear translocation of NF-κB were required in liver cirrhosis [25]. In renal fibrosis, the loss of Chop gene represses Hmgb1/TLR4 signal pathway, leading to repressed NF-κB transcriptional activity along with suppressed IL-1β production, and reduced TGF-β1 production and PI3K/Akt activity to attenuate the development of fibrosis [26]. Our studies also revealed that mutations in MYBPC3 enhances HIF-1α expression and aerobic glycolysis. HIF-1α is a key transcription factor in response to chronic hypoxia and participates in fibrotic diseases, such as systemic sclerosis (SSc) [27]. Under hypoxic conditions, HIF-1α stably accumulates in the cytoplasm, and then transfers to the nucleus to form HIF-1α/ARNT dimer, and initiates the transcription of target genes, such as glycolytic genes (GLUT1, PFK, LDHA) [28]. However, hypoxia-independent mechanisms to regulate HIF-1α have also been proposed. In certain tumors, high levels of HIF-1α are observed in well-oxygenated environments [29]. Growth factor signal transduction has also been suggested to enhance HIF-1α expression [30]. Genetic mutations that result in hyperactivation of oncogenic signal transduction pathways also enhance HIF-1α expression [31, 32]. The related functions of HIF-1α in fibrosis include stimulating excessive ECM, vascular remodeling and ineffective angiogenesis, further aggravating chronic hypoxia and deteriorating pathological fibrosis [33]. Chronic hypoxia in skeletal muscle pathology is an important feature of fibrosis. HIF-1α and TGF-β1 co-driven CCN2 overexpression contributes to the establishment of fibrosis [34]. It has also been reported that TGF-β1 induces HIF-1α stabilization [35]. In this study, we demonstrated that TGF-β1 enhanced HIF-1α expression in MYBPC3-KO fibroblasts, which could be reversed by SB-431542. The switch of metabolism from oxidative phosphorylation to aerobic glycolysis (Warburg effect) and increased lactic acid production were observed not only in R495Q mutant pigs but also in MYBPC3-KO fibroblasts, which could be reversed by aerobic glycolysis inhibitor 2-DG. These observations are consistent with previous reports that aerobic glycolysis occurs under a wide range of fibrotic conditions [36]. For example, downregulation of fatty acid oxidation (FAO) and upregulation of glycolysis were found in fibrotic skin [37]. In addition, human keloid fibroblasts are characterized by higher rate of ATP biosynthesis, with glycolysis as their primary energy source, demonstrated by increased lactic acid production [38]. The Warburg switching in renal fibroblasts is the primary feature of fibroblast activation during renal fibrosis and that suppressing renal fibroblast aerobic glycolysis could significantly reduce renal fibrosis [39]. Previous studies also demonstrated that LPS promotes collagen synthesis in lung fibroblasts through aerobic glycolysis via the activation of PI3K-Akt-mTOR/PFKFB3 pathway [40]. Analysis of liver samples from patients with hepatocellular carcinoma (HCC) and patients with cirrhosis showed that the expression of glycolytic enzymes was up-regulated in precancerous cirrhotic livers and significantly associated with elevated risks for developing HCC [41]. These findings suggest that the metabolism switching is a fundamental feature of fibroblast activation in various disease conditions including HCM. In conclusion, our studies suggest that MYBPC3 deficiency in cardiac fibroblasts can promote their trans-differentiation into myofibroblasts via NF-κB/TGF-β1/HIF-1α/aerobic glycolysis signal cascade, contributing to premature myocardial fibrosis. These findings highlight the importance to consider cardiac fibroblasts as additional therapeutic target in future therapeutic development for HCM patients with MYBPC3 mutations. All animal studies were approved by the Animal Welfare and Research Ethics Committee at Jilin University (SY201903015), and all procedures were conducted strictly in accordance with the Guide for the Care and Use of Laboratory Animals. All surgeries were performed under anesthesia, and every effort was made to minimize animal suffering. The sgRNA oligonucleotides targeting mouse Mybpc3 gene (Comate Bioscience, China) were annealed and cloned into pX330-U6-Chimeric_BB-CBh-hSpCas9 (Addgene #42230) and the sgRNA oligonucleotides targeting pig MYBPC3 gene were annealed and inserted into pBluescriptSKII+ U6-sgRNA(F + E) empty expression vector (Addgene #74707). NIH3T3 (mouse embryonic fibroblast cells) and PFFs (porcine fetal fibroblasts) were cultured in DMEM (GIBCO) supplemented with 15% fetal bovine serum (FBS) at 37 °C and 5% CO2 in a humidified incubator. To generate R495Q mutant PFFs, approximately 3 × 106 PFFs were electro-transfected with pCMV_AncBE4max plasmid (Addgene #112094) and MYBPC3-sgRNA plasmid (15 μg each) in 200 μL of Opti-MEM (GIBCO) using 2-mm gap cuvettes and a BTX ECM 2001 electroporator. The parameters for electrotransfection were as follows: 340 V, 1 ms, 3 pulses for 1 repeat. At 36 h after electrotransfection, the cells were plated into ten 10-cm dishes at a density of 4 × 103 cells per dish. Single-cell colonies were picked and cultured in 48-well plates. When the plates reached 90% confluence, 50% of cells from each plate was lysed using 20 μL of lysis buffer (0.45% NP-40 plus 0.6% Proteinase K) for 60 min at 56 °C and 10 min at 95 °C to provide templates for genotyping with the following primers: pMYBPC3-JD-F, TCTTTGAGTCCATCGGCACC, and pMYBPC3-JD-R, CCCACAGTCAAGTCTGCGAT. The PCR conditions were 94 °C for 5 min; 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 1 min for 35 cycles; 72 °C for 5 min; and hold at 16 °C. The MYBPC3-KO NIH3T3 cells were similarly generated except that the pX330-U6-Chimeric_BB-CBh-hSpCas9 with Mybpc3-sgRNA plasmid (30 μg) was used for electrotransfection. The following primers were used for genotyping of MYBPC3-KO NIH3T3 cells: mMYBPC3-JD-F, GAACAGGCAAACGAAGGACAG, and mMYBPC3-JD-R, TCTTGGTGCAGAAGAGGGGAA. Potential off-target sites (OTSs) for each sgRNA were predicted by Cas-OFFinder (http://www.rgenome.net/cas-offinder/). OTS were analyzed via PCR and DNA sequencing to determine the target effects. The primer sequences used for analyzed the off-target activities are listed in Supplemental Tables S1 and S2. SCNT and embryo transfer were performed as described previously [42]. Briefly, recipient sows (100–120 kg, 9–10 months of age) were pre-anesthetized by subcutaneous or intramuscular injection of atropine (0.05 mg/kg). During operation, anesthesia was maintained by constant inhalation of 2% isoflurane. Subcutaneous injection of carprofen (2–4 mg/kg) was used for analgesia, once per day for three days starting from the day of operation. Euthanization of the animals was carried by intravenous injection of a lethal dosage of pentobarbital sodium (100–200 mg/kg), followed by chest opening and tissue collection. Serum samples were collected and measured by ELISA following the manufacturer’s instructions, in an infinite 2000 PRO Microplate Reader (Tecan, Switzerland). Samples were measured in triplicate, and the absorbance was monitored at 37 °C. Cell proliferation assay was detected by 5-ethynyl-2-deoxyuridine (EdU, RiboBio, China) following the manufacturer’s instructions. Briefly, cells were seeded into 96-well plates at a density of 1 × 105 cells per well and incubated with 10 μM EdU for 1 h. After that, cells were fixed and stained with Hoechst 33342. Images were captured using the fluorescence microscope (Olympus BX51). EdU-positive cells were quantified via ImageJ (NIH, Bethesda, MD, USA) on unmanipulated TIFF images. Cell wound scratch assay was performed as manufacturer’s instruction. Firstly, marker pen was used to draw a line at an interval of 1 cm behind the 6-well plate. Then, about 5 × 105 cells were plated into 6-well plate. Then, a scratch wound was generated using a pipette tip and washed three times with PBS to remove the scratched cells. Subsequently, serum-free medium was added and cells were further cultured in a humidified incubator at 37 °C. Cells were stained with Calcein-AM (Dojindo, China) and imaged at 0, 6, 12, 24, and 48 h. Numbers of fluorescent cell corresponding to wound closure rate were calculated by Image-J software (NIH, Bethesda, MD, USA). Cell cycle was detected by Cell Cycle and Apoptosis Analysis Kit (Beyotime, China) following the manufacturer’s instructions. First, cells were washed twice with PBS and fixed in 70% cold ethanol overnight at 4 °C. And then stained with propidium iodide (PI) for 30 minutes in the dark at 37 °C. Finally, the DNA content was measured by fluorescence-activated cell sorting (FACS) instrument (BD Biosciences). Cell apoptosis was detected by Annexin V-FITC apoptosis detection kit (Beyotime, China). A total of 1 × 105 cells were incubated with Annexin V-FITC and PI in the provided binding buffer for 30 min in the dark at 4 ˚C, and analyzed by fluorescence-activated cell sorting (FACS) instrument (BD Biosciences). Intracellular ATP level was detected by using Enhanced ATP Assay Kit (Beyotime, China). In accordance with the manufacturer’s instructions, cells were washed twice with ice-cold PBS, lysed using ice-cold ATP lysis buffer, and then centrifuged for 5 min to collect the supernatants. Then, ATP concentrations were measured by luminescence and ATP level was calculated according to standard curve with an infinite 2000 PRO Microplate Reader (Tecan, Switzerland). Samples were measured in triplicate, and the absorbance was monitored at 37 °C. All results were normalized to the total protein concentration. Glucose concentrations were detected by Tissue Cell Glucose Oxidase Assay Kit (Pplygen, China) following the manufacturer’s instructions. The glucose concentrations were measured and calculated according to standard curve with an infinite 2000 PRO Microplate Reader (Tecan, Switzerland). Samples were measured in triplicate, and the absorbance was monitored at 37 °C. All results were normalized to the total protein concentration. Lactic acid concentrations were determined by Lactic Acid assay kit (Nanjing Jiancheng Bioengineering Institute, China) according to the manufacturer’s instruction. The OD was measured by an infinite 2000 PRO Microplate Reader (Tecan, Switzerland). Samples were measured in triplicate, and the absorbance was monitored at 37 °C. All results were normalized to the total protein concentration. Equal numbers of viable cells were plated in 96-well plates. Cells were incubated with 200 μL drug-supplemented medium, treated with DMSO (vehicle) at 0.1% or the following drug concentrations standardized to 0.1% DMSO final concentration. For NIH-3T3 cells, the treatment regimens were: 2-DG (MCE, China), 1 mM, 5 mM, 10 mM, 15 mM and 20 mM; SB-431542 (MCE, China), 5 μM, 10 μM, 50 μM,100 μM and 200 μM; Oltipraz (MCE, China), 5 μM, 10 μM, 50 μM,100 μM and 200 μM; PDTC (MCE, China), 5 μM, 10 μM, 50 μM,100 μM and 200 μM; After incubation for 24 h, 48 h or 72 h, cell viability was measured using a Cell Counting Kit-8 (CCK8) assay (Dojindo, China) according to the manufacturer’s instructions. The optical density at 450 nm (OD450 nm) was measured using an infinite 2000 PRO Microplate Reader (Tecan, Switzerland). Samples were measured in triplicate, and the absorbance was monitored at 37 °C (Supplemental Fig. S6). Cell culture supernatant and cell lysates were measured by ELISA kits (Boster, China) following the manufacturer’s instructions. The OD was measured by an infinite 2000 PRO Microplate Reader (Tecan, Switzerland). Samples were measured in triplicate, and the absorbance was monitored at 37 °C. For detection of relative mRNA expression of genes, total RNA was isolated by TRNzol-A+ Reagent (TIANGEN, China) following the manufacturer’s recommendations. One μg RNA was reverse-transcribed (RT) to generate cDNA using a FastKing RT Kit (with gDNase) (TIANGEN, China) according to the manufacturer’s manual. The fluorescence intensity and amplification plots were analyzed by a BIO-RAD iCycler Thermal Cycler with iQ5 Optical Module for RT-PCR (Bio-Rad, ABI 7500, iQ5). The primers used in RT-PCR are listed in Supplemental Table S3. Western blotting was performed as described manufacturer’s instructions. Equal amounts of 40 μg proteins were separated through SDS-PAGE on a 10% separating gel, and the protein bands were electrophoretically transferred to polyvinylidene fluoride (PVDF) membranes. The protein bands were detected with the ECL-Plus Western blotting reagent. The primary and secondary antibodies involved in the process were shown in Supplemental Table S4. Fresh heart muscle tissues were fixed in 4% PFA, embedded in paraffin, and sectioned at 5 μm. HE staining and Masson staining were performed with standard techniques. For fibrosis analysis, nine non-overlapping pictures (400X) were randomly taken from each section and then calculated fibrotic area (blue) and total tissue area by ImageJ software (NIH, Bethesda, MD, USA). All data are expressed as the means ± standard error of the mean (SEM). Statistical differences were determined by unpaired Student’s t-test for two group comparisons, and one-way ANOVA with Bonferroni’s post-tests for multiple group comparisons. All statistical analyses were completed using GraphPad Prism 7.0 software. SUPPLEMENTAL MATERIAL

PMC9649788

Hsin-Jou Kao,Arun Balasubramaniam,Chun-Chuan Chen,Chun-Ming Huang

Extracellular electrons transferred from honey probiotic Bacillus circulans inhibits inflammatory acne vulgaris

10-11-2022

Microbiology,Microbiology techniques

Bacillus circulans (B. circulans) is widely used as an electrogenic bacterium in microbial fuel cell (MFC) technology. This study evaluated whether B. circulans can ferment glucose to generate electricity and mitigate the effects of human skin pathogens. The electricity production of B. circulans was examined by measuring the voltage difference and verified using a ferrozine assay in vitro. To investigate the fermentation effects of B. circulans on inhibition of human skin pathogens, Cutibacterium acnes (C. acnes) was injected intradermally into mice ears to induce an inflammatory response. The results revealed that the glucose–B. circulans co-culture enhanced electricity production and significantly supressed C. acnes growth. The addition of roseoflavin to inhibit flavin production considerably reduced the electrical energy generated by B. circulans through metabolism and, in vivo test, recovered C. acnes count and macrophage inflammatory protein 2 (MIP-2) levels. This suggests that B. circulans can generate electrons that affect the growth of C. acnes through flavin-mediated electron transfer and alleviate the resultant inflammatory response. Our findings demonstrate that probiotics separated from natural substances and antimicrobial methods of generating electrical energy through carbon source fermentation can help in the treatment of bacterial infections.

Extracellular electrons transferred from honey probiotic Bacillus circulans inhibits inflammatory acne vulgaris Bacillus circulans (B. circulans) is widely used as an electrogenic bacterium in microbial fuel cell (MFC) technology. This study evaluated whether B. circulans can ferment glucose to generate electricity and mitigate the effects of human skin pathogens. The electricity production of B. circulans was examined by measuring the voltage difference and verified using a ferrozine assay in vitro. To investigate the fermentation effects of B. circulans on inhibition of human skin pathogens, Cutibacterium acnes (C. acnes) was injected intradermally into mice ears to induce an inflammatory response. The results revealed that the glucose–B. circulans co-culture enhanced electricity production and significantly supressed C. acnes growth. The addition of roseoflavin to inhibit flavin production considerably reduced the electrical energy generated by B. circulans through metabolism and, in vivo test, recovered C. acnes count and macrophage inflammatory protein 2 (MIP-2) levels. This suggests that B. circulans can generate electrons that affect the growth of C. acnes through flavin-mediated electron transfer and alleviate the resultant inflammatory response. Our findings demonstrate that probiotics separated from natural substances and antimicrobial methods of generating electrical energy through carbon source fermentation can help in the treatment of bacterial infections. Honey is a naturally sweet substance, mainly consisting of glucose and fructose. Many studies have reported honey to have antibacterial activity and to be effective in wound management. However, some beneficial microorganisms have been found in honey in the form of spores to resist high concentrations of acids and sugar. Nonpathogenic bacterial strains in honey can also grow when honey is diluted with water. These nonpathogenic microorganisms include yeast (1%) and gram-positive bacteria (27%) such as Bacillus was one of the predominant genera. Previous study reported that B. circulans are present in the digestive tracts of honey bees and can inhibit the growth of Ascophaera apis, the causative agent of chalkbrood disease in honeybee larvae, possibly through bacteriocins and other antimicrobial molecules. Additionally, microorganisms such as Bacillus subtilis, and Clostridium butyricum can produce bioelectricity through extracellular electron transfer (EET). B. circulans is an electrogenic bacterium with potential for application in MFC technology. In MFCs, Bacillus cereus strain DIF1 and Rhodococcus ruber strain DIF2 actively secrete riboflavin and flavin mononucleotide (FMN), which contribute as electron mediators in EET, mediate electron transfer to extracellular acceptors, and enhance electric current production. Furthermore, through the addition of exogenetic flavins, Bacillus megaterium (B. megaterium) strain LLD-1 can increase the production of electricity by the fermentation of different carbon sources. In short, B. circulans can be isolated from honey and generate electricity. During metabolism and EET, NADH is oxidised to NAD+ through NADH dehydrogenase and delivers electrons to the extracellular space to reduce extracellular electron acceptors. EET includes direct and indirect modes, such as conductive protein filaments (microbial nanowires), electron-shuttling mediators (flavin or methyl viologen), and the extracellular polymeric substances of biofilms (redox proteins). Recently, flavins that mediate EET have received more attention. Flavins are common cofactors that are highly effective as redox enzymes in natural biological systems. In catalytic reactions, flavins oxidise electron donors, such as hydrogen and bacterial fermentation products, and release electrons. These electrons are then used to reduce extracellular electron acceptors such as Fe(III) or Mn(IV). Acne vulgaris is a skin disease in which the skin commensal C. acnes overcolonises the pilosebaceous unit and secretes lipase. The lipase breaks down triglycerides to release free fatty acids and stimulates the cells to produce proinflammatory cytokines, including interleukin (IL)-8, IL-12, IL-1β, and MIP-2, resulting in severe inflammation. Acne is most commonly treated through antibiotic application, which inhibits C. acnes overgrowth and lipase activity; however, this therapy has several side effects, such as promoting the emergence of antibiotic‐resistant C. acnes strains and nonspecific killing of other skin commensal bacteria. Alternatively, we reported that short-chain fatty acids (SCFAs), the fermentation metabolite from Staphylococcus epidermidis (S. epidermidis) could inhibit the growth of C. acnes. The injection of S. epidermidis with sucrose in an animal model led to decreases in MIP-2 levels and C. acnes-induced inflammation. In the present study, we further evaluated whether the honey probiotic B. circulans can ferment glucose to generate electricity, thereby reducing C. acnes lipase-induced MIP-2 levels, in addition to supressing C. acnes growth, through flavin-based EET. This study was carried out in strict with an approved Institutional Animal Care and Use Committee (IACUC) protocol at National Central University (NCU), Taiwan (NCU-106-016) and in compliance with the Arrive guidelines (https://arriveguidelines.org/). Institute Cancer Research (ICR) mice aged 8–9 weeks females (National Laboratory Animal Centre, Taipei, Taiwan) were sacrificed under CO2 anesthesia in a sealed chamber. All methods were performed in accordance with relevant guidelines and regulations. The wildflower honey (Neu Wang Feng Co., Ltd., Taoyuan, Taiwan) was diluted 1:10 with PBS and incubated at 37 °C on TSB agar plate. After 3 days, bacteria were collected and analyzed by 16S rRNA gene sequencing (Tri-I Biotech Inc., New Taipei, Taiwan). Three bacteria were identified as B. circulans, Lysinibacillus fusiformis (L. fusiformis), and Bacillus asahii (B. asahii). C. acnes (ATCC 6919) and B. circulans were cultured on Reinforced Clostridium Medium (RCM, BD, Sparks, MD, USA) under anaerobic conditions using a Gas-Pak (BD) and Tryptone Soy Broth (TSB, BD) medium. Bacteria were cultured at 37 °C until the logarithmic growth phase. Bacterial pellets were harvested by centrifugation at 5000 × g for 10 min, washed in phosphate-buffered saline (PBS), and then suspended in PBS for further experiments. B. circulans (107 colony-forming unit (CFU)/mL) was incubated in 5 mL TSB in the in the presence or absence of 2% glucose at 37 °C. Glucose alone in TSB was included as control. The 0.002% (w/v) phenol red (Sigma, Burlington, MA, USA) in TSB served as fermentation indicator. A colour changes from red–orange to yellow indicated the occurrence of bacterial fermentation, which was detected by optical density at 560 nm (OD560). MFC compartments was established for detection of bacterial electricity. The carbon cloth (9 × 9 cm) (Homy Tech, Taoyuan, Taiwan) as cathode, carbon filth (2.5 × 5 cm) (Homy Tech, Taoyuan, Taiwan) as anode and proton exchange membrane reated by Nafion membrane N117 (5 × 5 cm) (Homy Tech, Taoyuan, Taiwan). Anode and cathode were linked by copper wires, which in turn were connected to 1000 Ω external resistance. Bacteria (107 CFU) with and without 2% glucose or 0.1 μM roseoflavin (flavin inhibitor) in TSB media was loaded on the surface of anode. The cell voltage was recorded every 30 s by a digital multimeter (DM-9962SD, Lutron, Australia) for 20 min. Ferrozine assay was performed by suspending B. circulans (107 CFU) in TSB medium with and without 2% glucose and 0.1 μM roseoflavin total 50 μl, an equal volume of ferrozine (8 mM) (Sigma) and 100 μl ferric ammonium citrate (100 mM) (Sigma) were added into each well. The mixture was incubated at 37 °C for 1 h in 96-well. The colour change of media was detected from OD at 562 nm. In vitro, Co-Culture C. acnes (107 CFU) and B. circulans (107 CFU) in TSB with and without 2% glucose under anaerobic conditions for 3 days at 37 °C. B. circulans and glucose alone mix with C. acnes was included as a control. After 3 days, dilute with PBS for C. acnes bacterial counts. In vivo, the ears of ICR mice were injected intradermally with C. acnes (107 CFU) mix with B. circulans and 2% glucose in the presence or absence of 0.1 μM roseoflavin. B. circulans and glucose mix with C. acnes was included as a control. After 3 days, cut the mice ears and homogenized for C. acnes bacterial counts. The C. acnes loads in in vitro and in vivo sample were enumerated by plating serial dilutions (1:10–1:105) with PBS of selective agar plates containing RCM media and 10 μg/mL of furazolidone (Sigma). The plates were incubated for 5 days at 37 °C in an anaerobic chamber using Gas-Pak. Transformation of a plasmid encoding lipase (accession number: YP_056770.1) into Escherichia coli (E. coli) BL21 competent cells (Invitrogen, Carlsbad, CA, USA). The E. coli BL21 transformed with a plasmid encoding GFP was used as a control by following the same procedure. A transformant of E. coli BL21 was inoculated with Luria–Bertani (LB) (Biokar Diagnostics, Beauvais, France) medium containing ampicillin (Sigma) at 37 °C until the OD600 reached 0.6–0.8. 1 mM Isopropyl-B-D-thiogalactoside (IPTG) (Sigma, Burlington, MA, USA) was added into culture for 4 h at 30 °C to induce protein expression. Proteins were purified by ProBond™ Purification System (Invitrogen, Carlsbad, CA, USA). The ears of ICR mice were injected intradermally with 5/10 μl lipase or GFP (as a control) to induces the inflammation. After 24 h, cut the mice ears and the level of MIP-2 cytokine was measured by ELISA. In electrons treatment experiment, the ears of ICR mice were injected intradermally with 5/10 μl lipase to induces the inflammation. After 24 h, B. circulans, 2% glucose with 0.1 μM roseoflavin was injected. B. circulans and glucose without roseoflavin was included as a control. After 24 h, cut the mice ears and homogenized for MIP-2 quantified. The proinflammatory MIP-2 cytokines in the supernatants of ear homogenates was quantified by an ELISA kit, as directed by the manufacturer (R&D System. Inc., Minneapolis, MN, USA). Data analysis was performed by unpaired t-test using Prism software (https://www.graphpad.com/; Version 5.01, GraphPad Software, La Jolla, CA, USA). The levels of statistical significance were indicated as the following: *p < 0.05, **p < 0.01, ***p < 0.001 and ns = non-significant. The mean ± standard. deviation (SD) for at least three independent experiments was calculated. Animal experiments were performed with at least three animals per each treatment group. By using 16S rRNA gene analysis, we isolated three bacteria (Supplementary Fig. S1A), B. circulans, L. fusiformis, and B. asahii (Supplementary Table S1) from TSB agar plate with 10% honey. Particularly, one 16S rRNA gene sequence of the isolated three bacteria shares 99% similarity to that of B. circulans strain FDAARGOS_783 (GenBank accession no NZ_CP053989.1). To investigate whether B. circulans can ferment glucose, it was incubated with 2% glucose in TSB media with phenol red for 1 day. With B. circulans alone, phenol red colour changed to orange because of bacterial replication during incubation. However, when B. circulans was incubated along with glucose, phenol red colour changed to yellow with a decrease in pH value, indicating the use of glucose for fermentation (Fig. 1A; upper panel). Furthermore, the quantification of fermentation by measuring the optical density of phenol red at OD560 nm indicated a significant decrease of OD560 values in TSB media containing bacteria plus glucose medium compared with bacteria or glucose alone (Fig. 1A; lower panel). Next, electricity production in the B. circulans with glucose was identified by adding the fermented media to the anode of the MFC system. B. circulans and glucose alone were used as controls and exhibited a low voltage change at 20 min (Fig. 1B; blue and green lines). By contrast, the voltage significantly increased to approximately 4 mV in the B. circulans with 2% glucose group (Fig. 1B; red line). We further verified that B. circulans can produce electrons through glucose fermentation by using the ferrozine assay to identify ferric iron reductase activity. In Fig. 1C, it can be seen that the concentration of ferrozine-chelatable iron (dark brown) in the reaction solution containing a fermentation medium of B. circulans plus glucose was distinctly higher than the medium, glucose, or bacteria alone (Fig. 1C; left panel). Statistical tests further confirmed that the ferric iron reductase activity of B. circulans was significantly increased when with glucose than without glucose (Fig. 1C; right panel). When compared with glucose alone, B. circulans and glucose resulted in significant increase in OD value. Gram-positive bacteria used flavin-based EET to deliver electrons. Roseoflavin represses FMN riboswitch and subsequently mediates riboflavin and FMN gene expression. In Fig. 2A, 0.1 µM roseoflavin was added to the culture of bacteria and glucose, and phenol red colour changed to yellow with a decrease in pH, similar to that in the fermentation experiment illustrated in Fig. 1A, confirming that roseoflavin does not affect the fermentation of B. circulans. Next, to identify electricity production in the B. circulans-glucose-roseoflavin culture, the fermented medium was added to the anode of the MFC system. The voltage change induced by B. circulans in the presence of glucose was completely attenuated by the addition of roseoflavin (Fig. 2B). In Fig. 2C, the concentration of ferrozine-chelatable iron in the reaction solution containing the fermentation medium of B. circulans-glucose-roseoflavin was distinctly decreased compared with medium, glucose, and bacteria. In summary, the findings of similar acidities but significantly decreased voltage production when adding roseoflavin indicated that the voltage change was not due to the pH change during fermentation and the number of B. circulans because 0.1 μM roseoflavin did not influence bacterial growth (Supplementary Fig. S2 for details). Having established the electricity produced in B. circulans-glucose cultures, we cocultured C. acnes and B. circulans in vitro to test the impact of B. circulans fermentation on the growth of C. acnes. The number of C. acnes was significantly decreased when mixed with B. circulans and glucose (Fig. 3A). Furthermore, the B. circulans-C. acnes-glucose mixture was injected intradermally with or without roseoflavin in ICR mice ears to examine whether glucose fermentation of B. circulans with roseoflavin changed the bacterial growth by electron. The result shows that the C. acnes count was significantly decreased in the absence of roseoflavin but increased in the presence of roseoflavin (Fig. 3B). After 24 h induction of lipase injection (with GFP as the control), MIP-2 expression was elevated in the presence of lipase (Fig. 4A). In the electron treatment experiment, the use of roseoflavin dramatically decreased electricity production, attenuated the anti-inflammatory defence of B. circulans fermentation by glucose, and significantly increased the concentrations of lipase-induced MIP-2 (Fig. 4B). Through probiotics’ fermentation of proper carbon sources, pathogen-caused skin diseases can be reduced. For instance, SCFAs, when used as fermentation metabolites, can inhibit bacterial growth to achieve treatment effects. In addition, it has been shown that B. circulans can convert biomass into electrical energy. Its halophilic strain BBL03 can ferment 1% chitin and use degraded metabolites as electron donors to generate electricity in seawater; therefore, it can serve as electricity-producing bacteria in MFCs. In this study, after adding 2% glucose to the medium containing B. circulans, the electricity increased significantly when measured using changes in voltages, indicating that B. circulans can generate a substantial number of electrons through glucose fermentation. Notably, the low electricity was also detected in the medium containing B. circulans without the addition of glucose. The reason for this slight change of electricity production may be due to the presence of a small amount of glucose in TSB (Fig. 1B). In gram-positive bacteria, EET is involved in iron redox. Fe3+-reducing microorganisms belonging to the Geobacteraceae family can ferment sugars and other organic compounds to produce simple organic acids (such as acetate) that serve as electron donors to the electrodes. Similarly, by adding B. circulans and glucose to the solution containing ferric (Fe3+) ammonium citrate, we observed that the concentration of ferrozine-chelated Fe2+ was higher than that in the groups without glucose (Fig. 1C). The addition of glucose allowed more electrons to be generated, leading to increased ferric reduction, which rendered the culture medium-dark brown. The flavin-based EET mechanism has been confirmed in various gram-positive bacteria. Flavin in the suspension culture of the B. megaterium LLD-1 strain acted as an electron shuttle, enhancing electron transfer from LLD-1 to the electrode. Roseoflavin is a natural antibacterial compound. When combined with FMN riboswitch, it can inhibit Rli96 transcription, control the expression of downstream genes, and regulate the in vivo synthesis of flavin, thus impairing the bacterial metabolism and achieving bacteriostasis. Roseoflavin can also be converted into roseoflavin mononucleotide and roseoflavin adenine dinucleotide, both of which cause defects in cellular physiological functions. Therefore, FMN riboswitch may serve as a novel target for inhibiting pathogens. We used phenol red to monitor the degree of acid production by bacterial fermentation, and the addition of roseoflavin as flavin inhibitor to the culture medium did not alter the production of organic acids (Fig. 2A). This suggests that the reduction of electrons was not caused by organic acids but by the roseoflavin-induced inhibition of flavin generation. The ferrozine assay indicated that the decrease of iron concentration represented a decrease in electrons produced by fermentation (Fig. 2C). We previously reported that S. epidermidis can ferment glycerol and PEG-8 laurate to generate potential electron donors for electricity generation to combat ultraviolet damage or suppress acne vulgaris. In this study, we further demonstrated that adding glucose to the culture medium of B. circulans in vitro can inhibit the growth of C. acnes; specifically, this suppression was significantly reversed when roseoflavin was added to the mix. Taken together, B. circulan affected the growth of C. acnes through the electrons generated by glucose fermentation and flavin-mediated EET. These results also reveal an efficient EET mechanism to target pathogenic microorganisms by using electrons. Importantly, the intradermal application of B. circulans plus glucose to mice ears can suppress the C. acnes count while the use of nonselective anti-inflammatory drugs for treatment may lead to epidermal dysbiosis and the spread of resistant strains. In acne pathogenesis, the increased activity of the virulence factor lipase caused by C. acnes overcolonisation led to an inflammatory response that resulted in the release of proinflammatory cytokines and TNF-α, which modulated host immune response. In line with previous reports, lipase-induced immune responses were confirmed by an elevated MIP-2 content measured using ELISA in this study. Adding glucose to B. circulans can inhibit lipase-induced MIP-2 expression, but this inhibition was reversed when roseoflavin was also added. Therefore, B. circulans can generate electrons through glucose fermentation to affect the growth of C. acnes through flavin-mediated electron transfer, thereby reducing the resultant inflammatory response. In short, weak currents inhibit bacterial growth. The underlying mechanisms for the current-related lysis may be because of electron-induced electrolysis, the generation of free radicals, pH, and changes in biofilm structure. It was reported that exposure of gram-positive bacteria to pulsed electric fields can induce permeabilization of the plasma membrane, destabilising the cell wall and causing osmotic shock. In other words, electric current generated with conductivity electrodes can directly inhibit bacterial growth, but the transition of platinum complexes and metal ions generated during electrolysis can harm human cells. By contrast, the weak current produced by B. circulans through glucose fermentation, as demonstrated in this study, can efficiently and safely supress pathogenic bacterial growth. Overall, this study revealed the molecular mechanism by which the probiotic B. circulans in honey can generate electrical energy by using glucose as a prebiotic. B. circulans reduced the inflammatory response by disrupting C. acnes growth through FMN riboswitch and flavin-mediated electron transfer. Therefore, generating electrical energy from biomass through the metabolic activities of microorganisms may be a potential antimicrobial therapy. These results are beneficial for the future clinical treatment of acne-prone skin disorders and to development of skincare products. Supplementary Information.

PMC9649793

Lorenzo Vergani,Francesca Mapelli,Magdalena Folkmanova,Jakub Papik,Jan Jansa,Ondrej Uhlik,Sara Borin

DNA stable isotope probing on soil treated by plant biostimulation and flooding revealed the bacterial communities involved in PCB degradation

10-11-2022

Soil microbiology,Microbial ecology,Environmental biotechnology

Polychlorinated biphenyl (PCB)-contaminated soils represent a major treat for ecosystems health. Plant biostimulation of autochthonous microbial PCB degraders is a way to restore polluted sites where traditional remediation techniques are not sustainable, though its success requires the understanding of site-specific plant–microbe interactions. In an historical PCB contaminated soil, we applied DNA stable isotope probing (SIP) using 13C-labeled 4-chlorobiphenyl (4-CB) and 16S rRNA MiSeq amplicon sequencing to determine how the structure of total and PCB-degrading bacterial populations were affected by different treatments: biostimulation with Phalaris arundinacea subjected (PhalRed) or not (Phal) to a redox cycle and the non-planted controls (Bulk and BulkRed). Phal soils hosted the most diverse community and plant biostimulation induced an enrichment of Actinobacteria. Mineralization of 4-CB in SIP microcosms varied between 10% in Bulk and 39% in PhalRed soil. The most abundant taxa deriving carbon from PCB were Betaproteobacteria and Actinobacteria. Comamonadaceae was the family most represented in Phal soils, Rhodocyclaceae and Nocardiaceae in non-planted soils. Planted soils subjected to redox cycle enriched PCB degraders affiliated to Pseudonocardiaceae, Micromonosporaceae and Nocardioidaceae. Overall, we demonstrated different responses of soil bacterial taxa to specific rhizoremediation treatments and we provided new insights into the populations active in PCB biodegradation.

DNA stable isotope probing on soil treated by plant biostimulation and flooding revealed the bacterial communities involved in PCB degradation Polychlorinated biphenyl (PCB)-contaminated soils represent a major treat for ecosystems health. Plant biostimulation of autochthonous microbial PCB degraders is a way to restore polluted sites where traditional remediation techniques are not sustainable, though its success requires the understanding of site-specific plant–microbe interactions. In an historical PCB contaminated soil, we applied DNA stable isotope probing (SIP) using 13C-labeled 4-chlorobiphenyl (4-CB) and 16S rRNA MiSeq amplicon sequencing to determine how the structure of total and PCB-degrading bacterial populations were affected by different treatments: biostimulation with Phalaris arundinacea subjected (PhalRed) or not (Phal) to a redox cycle and the non-planted controls (Bulk and BulkRed). Phal soils hosted the most diverse community and plant biostimulation induced an enrichment of Actinobacteria. Mineralization of 4-CB in SIP microcosms varied between 10% in Bulk and 39% in PhalRed soil. The most abundant taxa deriving carbon from PCB were Betaproteobacteria and Actinobacteria. Comamonadaceae was the family most represented in Phal soils, Rhodocyclaceae and Nocardiaceae in non-planted soils. Planted soils subjected to redox cycle enriched PCB degraders affiliated to Pseudonocardiaceae, Micromonosporaceae and Nocardioidaceae. Overall, we demonstrated different responses of soil bacterial taxa to specific rhizoremediation treatments and we provided new insights into the populations active in PCB biodegradation. Polychlorinated biphenyl (PCB)-contaminated soils represent a major environmental source of persistent organic pollutants (POPs) worldwide, with important concerns for ecosystems and public health. Microbial communities providing PCB biodegradation abilities through reductive dichlorination and aerobic pathways are the main agents of natural attenuation in historically contaminated environments, hence, when properly biostimulated, constitute a natural resource for the reclamation of polluted soils. Several studies have reported that the “rhizosphere effect” exerted by plant root exudation and root deposition on the soil microbiome has a positive influence on the degradation of POPs, including PCBs, as a consequence of the induction of microbial degradation pathways by secondary plant metabolites. Therefore, plant-driven biostimulation of PCB-degrading soil microbial populations (i.e., rhizoremediation) has been identified as a way to accelerate the natural attenuation process in the attempt to restore ecosystem services in those sites where the implementation of traditional remediation techniques based on soil removal and incineration are not sustainable due to costs and environmental impact, such as extended agricultural areas. However, root growth and exudation profile, soil microbial community, and their interdependent relationships may undergo temporal and spatial variations depending on the plant species and phenological growth stage along with soil edaphic conditions and contamination profile. In a previous, 18-month rhizoremediation study we reported how ten plant species combined with soil treatments differentially affected the structure and activity of the soil microbial community and ultimately the degradation rate of different PCB congeners in a weathered contaminated soil. In the present work we focused on the soils that during this rhizoremediation trial had been i) biostimulated with reed canary grass (Phalaris arundinacea L.) and/or ii) subjected to periodic flooding with the aim to induce a redox cycle that would allow both reductive dichlorination and aerobic degradation of PCBs by autochthonous microorganisms. These treatments were chosen because they promoted the reduction of the concentration of tri-, tetra- and penta-chlorinated PCB in the original soil after 18 months. We hypothesized that the combined or independent application of plant biostimulation and cyclic oxic-anoxic conditions given by the periodical flooding would shape the soil bacterial communities and would enrich different populations among the autochthonous PCB degraders present in this historically polluted soil. Therefore, we applied DNA-based stable isotope probing (SIP) incubation combined with 16S rRNA amplicon sequencing to determine how the plant growth and the redox cycles affected the structure of bacterial communities and their diversity and, in particular, the metabolically active bacteria potentially involved in PCB degradation. SIP has been successfully employed in bioremediation-oriented studies to detect bacterial taxa able to incorporate carbon from diverse organic pollutants. Compared to other molecular techniques, this approach allows to directly link metabolic capability to phylogenetic information within a microbial community by tracking isotopically labeled carbon into the DNA, and therefore in phylogenetic biomarkers such as the 16S rRNA gene, only in actively duplicating cells. This technology enables the identification of only the organisms active in the utilization of a specific substrate, requiring less sequencing effort than a full metagenomic analysis of the total community. To the best of our knowledge, whereas the non-chlorinated biphenyl backbone is generally considered a model molecule for PCB mineralization, only one study compared the results obtained from SIP incubation of sediment samples with 13C-labeled biphenyl or 13C-4-chlorobiphenyl (4-CB). This implies a lack of knowledge on the metabolization of chlorinated congeners in environmental matrices and especially in soils differentially biostimulated. To overcome this limitation and target more specifically the PCB biodegradation, in this work we used 13C-labeled 4-chlorobiphenyl as a metabolic tracer. Samples of soil were collected at the end of an 18-month rhizoremediation trial performed in greenhouse conditions. The trial was conducted on former agricultural soils from the National Priority Site for remediation (SIN) Brescia-Caffaro which were contaminated by PCBs but also PCDDs, PCDFs, DDT and its isomers, metalloids, and metals (mainly As and Hg). The experimental set-up, conditions, and applied treatments are described in detail in our previous publication. In this work, we used soils collected from four different treatments: biostimulation with Phalaris arundinacea (Phal) and non-planted control (Bulk); biostimulation with Phalaris arundinacea subjected to periodic soil flooding obtaining an oxic-anoxic (redox) cycle (PhalRed) and non-planted control subjected to the same redox cycle (BulkRed). Each treatment was performed on the same homogeneous original batch of soil in separate pots in three biological replicates. We set up SIP microcosms for three incubation times: seven (D07), twenty-one (D21) and twenty-eight (D28) days. Triplicate microcosms were established for each time point and soil treatment, corresponding to each original biological replicate of the four rhizoremediation treatments, totaling 36 microcosms. The set-up procedure was as follows: 10 µl of 13C-labeled 4-chlorobiphenyl (4-CB, 99% 13C, Alsachim) in acetone solution (50 mg/ml) were spiked on 0.2 g of quartz sand placed at the bottom of 100-ml sterile serum bottles to have a final quantity of 0.5 mg of 4-CB per microcosm. After acetone evaporation, 2 g of soil were added to each microcosm, manually mixed with the sand, and supplemented with 200 µl of basal mineral salt solution. The microcosms were sealed and placed at 25 °C in the dark, to avoid the proliferation of photosynthetic microorganisms, until destructive harvesting when the content was stored at -80 °C. Parallel incubation with unlabeled 4-CB was carried out for each microcosm as a control. At each harvest time point, the % of O2 and CO2 in the headspace of each microcosm was measured using a CheckPoint II—Portable Headspace Analyzer (Dansensor). For the determination of CO2 isotopic composition, 500 μl of the headspace gas was sampled with a gas-tight syringe from labeled microcosms and transferred into helium-filled borosilicate vacutainers. The isotopic composition of headspace CO2 was analyzed by Gasbench II device equipped with a cryotrap coupled to a Delta V Advantage isotope ratio mass spectrometer (ThermoFisher Scientific, Bremen, Germany). The mineralization of 13C-labeled 4-chlorobiphenyl was calculated stoichiometrically based on the amount of 13CO2 evolved. Basing on the results of the isotopic CO2 analysis, the metagenomic DNA was extracted from D21 and D28 SIP microcosms and the initial, non-incubated, time zero (T0) soil samples using the FastDNA™ SPIN Kit for Soil (MP Biomedicals) and purified with the Genomic DNA Clean & Concentrator™ (Zymo Research) following the manufacturer’s instructions. D07 samples were excluded from DNA extraction because no 13C-labelled CO2 was retrieved in the microcosms’ headspace. DNA concentration was measured with a Qubit™ fluorometer (Thermo Scientific) and all samples were adjusted to a final concentration of 25 ng/µl. For isopycnic separation, 40 µl (1,000 ng) of DNA were placed into a Quick-Seal™ centrifuge tube (Beckman) filled with cesium trifluoroacetate (CsTFA) pre-diluted to a concentration of 1.62 g/ml. Ultracentrifugation was run for 72 h at 145,000 × g, 24 °C, using an Optima™ MAX-XP Ultracentrifuge with TLN 100 rotor (Beckman). After ultracentrifugation, each gradient was fractionated into fractions of 100 μL using a Beckman Fraction Recovery System (Beckman) and a Harvard Pump 11 Plus Single Syringe (Harvard Apparatus) by replacement with water at a flow rate of 400 μL/min. To ensure reproducibility and accuracy, each isopycnic centrifugation and gradient fractionation were run in parallel with two blanks (water instead of DNA), then the buoyant density (BD) of each gradient-recovered fraction was inferred by measuring the refractive index of fractionated blank samples with a Digital Hand-held Refractometer (Reichert Analytical Instruments). DNA was retrieved from CsTFA by isopropanol precipitation with glycogen and stored at -20 °C. The distribution of bacterial DNA across the density gradients was assessed by quantitative PCR (qPCR) targeting the 16S rRNA gene using primers 357F (3’-CCCTACGGGAGGCAGCAG-5’) and 907R (3’-CCGTCAATTCCTTTGAGTTT-5’) as previously described. In each gradient, fractions that were determined to contain 13C-labeled DNA based on 16S rRNA gene copies distribution as a function of buoyant density (BD) were combined into pools. Equivalent pools were prepared for controls incubated with unlabeled substrates and were subjected to the same downstream analyses as 13C-labeled DNA to identify any background contamination. Fractions corresponding to the main 16S rRNA peak containing 12C-DNA of each gradient from both labeled and unlabeled SIP incubations and from initial, non-incubated T0 soils were also pooled and analyzed. We applied Illumina MiSeq sequencing to the pooled-DNA samples after amplification using primers 515F (5’-GTGYCAGCMGCCGCGGTAA-3’) and 926R (5’-CCGYCAATTYMTTTRAGTTT-3’) targeting the V4-V5 hypervariable region of the 16S rRNA gene. Reactions were performed at the Core Facility for Nucleic Acid Analysis, University of Alaska Fairbanks (USA). The 16S rRNA gene amplicon sequence data were processed using the DADA2 pipeline in R software with a few modifications to the DADA2 SOP. In brief, the primer sequences were trimmed off or, if absent, the whole read was discarded. In order to manage the lower quality towards read ends, forward and reverse reads were truncated to the length of 248 and 176 bp, respectively. The values were calculated as the average positions where 75% of reads had a quality score > = 25 while ensuring a hypothetical minimum of 25 bp overlap between the paired reads. Filtering was based on the reads quality using the following parameters: maxN = 0, maxEE = 2, truncQ = 2. After dereplication and application of DADA2-based removal of sequencing errors, denoised forward and reverse reads were merged and chimeric sequences were removed using the method = "pooled". Based on the analysis of the mock community, which consisted of three bacterial strains (i.e., Rhodococcus sp. 3B12, ENA accession LT978383; Acinetobacter sp. P320 ENA accession LT838128; Bacillus sp. P28, ENA accession LT838114) that were amplified in parallel with the DNA samples, the sequences which differed by one base were clustered together and the most abundant sequence from each of the clusters was considered as the valid one. Taxonomy was assigned by using Silva v138 to create a database of amplicon sequence variants (ASVs). The sequence reads were deposited in the NCBI SRA database under the BioProject ID: PRJNA809320. Prior to further analysis, sequencing data were rarefied at 4,000 reads per sample to ensure the comparability between samples. A principal coordinate analysis (PCoA) was used to assess the phylogenetic β-diversity based on Bray–Curtis distance matrix on the normalized (log transformed) ASV table. Significant differences in bacterial community composition were investigated by canonical analysis of principal coordinates (CAP) and permutational multivariate analysis of variance (PERMANOVA), according to the factors ‘biostimulation’ (levels: ‘planted’, ‘no plant’), ‘redox’ (levels: ‘redox’, ‘no redox’), ‘treatment’ (levels: ‘Phal’, ‘Bulk’, ‘PhalRed’, ‘BulkRed’), ‘time’ (levels: T0, D21, D28) and ‘treatment’ and ‘time’ interaction. PERMDISP analysis was implemented prior to CAP and PERMANOVA to test the homogeneity of the data dispersion. Statistical analyses were conducted in PRIMER v. 6.1, PERMANOVA + for PRIMER routines. Alpha-diversity indices (i.e., Shannon diversity and dominance) were calculated using the PAST software and their statistical difference was evaluated using the R software version 4.0.2. To identify the bacterial taxa incorporating 13C from 13C-labeled 4-CB, we summed the number of sequences for each ASV present in each of the three 13C-labeled replicate samples and made a comparison with the respective unlabeled control. Only ASVs that were at least ten-fold more abundant in the 13C-DNA pools were considered to have incorporated 13C. Sequencing results of pools of fractions corresponding to the main 16S rRNA peak of each control DNA gradient from T0 and D21 / D28 unlabeled SIP incubations were analyzed to characterize the community dynamics among different soil treatments and over time. Results showed that after seven days (D07) of incubation there was no evolution of 13CO2 in the microcosms supplemented by 13C 4-CB, while it was detected after twenty-one days (D21) and increased at the end of SIP incubation (D28) (Fig. 1, Supplementary Table S1). For this reason, D07 samples were excluded from further processing and analysis. Mineralization of 4-CB was calculated from the concentration of 13CO2 and depletion of the labeled substrate occurred in all D21 and D28 microcosms and varied between 10% in Bulk soil and 22% in BulkRed soil microcosms at D21 and reached up to 39% for the PhalRed soil at D28. Even though planted and flooded soils seemed to support a higher 13C 4-CB mineralization rate (Fig. 1), no significant differences (p > 0.05, Student t-test) were observed among treatments, possibly because of the intrinsic variability of soil samples and the limited number of biological replicates. Soil samples from D21 and D28 were therefore selected for labeled DNA isolation and further analysis to identify 13C-labeled bacterial populations. Following density gradient ultracentrifugation and fractionation of DNA, we performed 16S rRNA gene qPCR analyses to assess DNA distribution in the CsTFA gradients (Supplementary Table S2 A-F). Gradients from all D28 samples presented a small 16S rRNA peak corresponding to heavy fractions containing labeled DNA in addition to samples Bulk 3 and BulkRed 2 from D21. Thus, we considered only D28 samples for further analyses. A main 16S rRNA peak containing unlabeled DNA was observed at buoyant density of 1.60 g/mL, while 13C-DNA-containing fractions were identified at BDs between 1.61 and 1.64 g/mL. Minor quantities of 16S rRNA gene (Supplementary Table S2 A-F) were detected at the same BDs in samples from parallel incubations with unlabeled 4-CB, that were analyzed as controls. A total of 10,921,966 high-quality paired-end reads with an average length of 352 bp were obtained from 16S rRNA amplicon sequence data spanning all samples and replicates. Taxonomy assignment resulted in a database of 21,185 unique ASVs. After rarefaction to 4,000 reads per sample, the resulting ASV table was composed of 21,167 ASVs. Samples that were below the rarefaction threshold were excluded from the dataset (Supplementary Table S3 A-B). ASVs representing less than 0.005% of the total bacterial community were cut off prior to further analyses. Following PERMDISP validation (p = 0.179), the beta-diversity analysis on the ASV dataset showed that bacterial communities of soils subjected to different treatments grouped separately as represented by the PCoA (Fig. 2A) and differed significantly from one another as demonstrated by CAP (delta_1^2: 0.98974, p = 0.0001) and PERMANOVA main and pairwise tests, with the exception of the two non-planted soils (Supplementary Table S4A-B). Beta-diversity based on ASV composition according to the time of SIP incubation cannot be discussed due to the difference in dispersion among soil samples of T0, D21, and D28 (PERMDISP: p = 0.021), however we could observe a diversity given by this factor when considered together with the soil treatment (PERMDIS p = 0.142; PERMANOVA: F2,6 = 2.46, p = 0.0001). Estimation of the factors’ contributions to the observed variations determined by PERMANOVA showed that “treatment” alone explained 28% of the variation and an additional 23% was explained by its interaction with the factor “time” (Supplementary Table S4C). Furthermore, PCoA and PERMANOVA (Supplementary Figure S1; Supplementary Table S5) separated the communities of planted soils from the bulk soils (PERMDISP: p = 0.072), and those of soils subjected to a redox cycle from those that were not (PERMDISP: p = 0.748). Bacterial alpha-diversity was affected by different soil treatments, with Phal soils hosting the most diverse community compared to the other treatments, as shown by the Shannon index and demonstrated by ANOVA (Fig. 2B). Soils planted with P. arundinacea and not subjected to redox cycle (Phal) were also characterized by lower dominance, even though the difference was statistically significant only compared to the non-planted Bulk control (Fig. 2C). The analysis of the alpha-diversity also showed a decrease in the Shannon index over time, while the dominance index followed an opposite trend. However, both the parameters differed significantly only between pre-incubation (T0) and SIP-incubated soils (D21, D28) according to two-way ANOVA (Supplementary Table S6). The main bacteria phyla, having relative abundance between 25 and 5%, observed in the total community of pre-incubation samples (T0) were Proteobacteria, Actinobacteria, Acidobacteria, Chloroflexi and Planctomycetes. The different soil treatments presented a different profile in terms of phyla/class relative abundance (Supplementary Figure S2). ASVs affiliated with Actinobacteria were dominant in soil planted with P. arundinacea (Phal) (25% on average between the replicates), while they declined in non-planted soil samples (Bulk, 15.5%) and in both planted and non-planted soils subjected to redox cycle (PhalRed, 13.2% and BulkRed, 9.2%). ASVs affiliated with Proteobacteria showed similar abundance in all the T0 soil samples but were mainly represented by Alphaproteobacteria in Phal (11.9%) and Bulk (16.7%) soils, and by Betaproteobacteria in the repeatedly flooded soils (Phalred and BulkRed, 10% on average). Chloroflexi were enriched in PhalRed and BulkRed soils, representing respectively 20.5% and 19.8% of the total community. Other less abundant taxa whose relative abundance changed with soil treatments were Firmicutes, which were reduced in plant-biostimulated soils, and Desulfobacterota, which were increased in soils subjected to redox conditions. After incubation with 4-chlorobiphenyl, ASVs affiliated with Proteobacteria became prevalent, representing the 30% of the communities of Phal and PhalRed soils at D21 and further increasing up to 40% in Phal soils at D28. Actinobacteria were more abundant in PhalRed soils at D28 compared to T0 while the abundance of Chloroflexi was decreased in all the D28 communities. Fifty-seven different ASVs corresponding to bacterial taxa incorporating 13C derived from the mineralization of 4-CB were identified in D28 soils after rarefaction (Supplementary table S7A). We observed the highest number of ASVs (n = 30) in Bulk soils, followed by PhalRed (n = 19), Phal (n = 14), and BulkRed (n = 6). Most of these ASVs (n = 46) were unique to the different soil treatments and none was shared among all of them (Fig. 3A). Only one ASV, affiliated with the genus Rhodococcus, was common in three of the considered soils (i.e., Bulk, PhalRed and BulkRed). Within the bacteria deriving carbon from 4-CB, the most abundant class/phyla in all four treated soils were Betaproteobacteria and Actinobacteria (Fig. 3B). In particular, the bacterial ASVs detected in Phal soils mostly belonged to the family Comamonadaceae (85.7% of the 13C-incorporating bacteria), which represented only 20% in Bulk and 24.3% in PhalRed soils and were absent in BulkRed soils (Table 1). Instead, we found Rhodocyclaceae among the 13C-incorporating bacteria only in non-planted soils (4.2% in Bulk and 30.8% in BulkRed). Among Actinobacteria, Nocardiaceae were represented only by ASV3, and were particularly abundant in Bulk and BulkRed soils (28.1% and 58.4% respectively), while in planted soils they represented only 2.6% in PhalRed and were absent in Phal. PhalRed soil bacteria that derived 13C from 4-CB were affiliated with other Actinobacteria families, such as Pseudonocardiaceae (18.3%), Micromonosporaceae (8.1%), and Nocardioidaceae (4.7%). Alphaproteobacteria deriving carbon from 4-CB were also retrieved in all the soils but were more abundant in planted (7.7% in Phal and 19.7% in PhalRed) than in non-planted soils (4.3% in Bulk and 5.7% in BulkRed). The most represented families within this clade included mainly Dongiaceae (7.8%) and other unidentified taxa (6.7%) in PhalRed soils and Sphingomonadaceae in BulkRed (5.7%). Other phyla that were identified to have derived 13C from 4-CB were Chloroflexi (13.6% in PhalRed, 8.7% in Bulk and 2.4% in Phal samples) and Acidobacteria (5.0% in BulkRed, 3.9% in PhalRed and 1.9% in both Phal and Bulk samples). Considering the relative abundance of ASVs corresponding to 13C-deriving bacteria in the total community of each sample (i.e., 4,000 reads obtained after rarefaction) and comparing it with the relative abundance of the same ASVs in T0 soil bacterial communities, we could infer that the most relevant families identified herein as deriving carbon from 4-CB were enriched during the SIP incubation period (Supplementary Table S8). Rhizoremediation through plant biostimulation of autochthonous microbial degraders is considered an environmentally sustainable and cost-effective promising technology for the clean-up of weathered PCB-polluted soils. However, plant species-specific rhizosphere effect together with soil treatment practices, differentially impact soil bacterial communities, which further affects pollutants’ depletion rates. Consequently, a lack of knowledge still exists about how biodegradation of PCB congeners occurs and which are the main microbial players under different soil rhizoremediation treatments. In this work, a historically contaminated soil containing an initial microbiome naturally selected during twelve years of natural attenuation was biostimulated for 18 months using reed canary grass and/or redox cycle, aiming at the enrichment of microbial degraders. Subsequent soil incubation with 13C-4-CB demonstrated the presence of bacteria that were able to assimilate 4-CB. The results of the evolution of 13CO2 in soil microcosms showed that after one week of SIP incubation there was no mineralization of the labeled substrate regardless of the original soil treatment, and, apart from two D21 samples, we were able to detect 13C labeled DNA only after twenty-eight days (D28). These results are in contrast with the only previous studies that used 4-CB as a metabolic tracer of PCB biodegradation in a stream sediment, in which the substrate mineralization and the identification of bacterial degraders occurred after one to seven days of incubation. Possible explanations could be that the different used matrices are characterized by diverse native bacterial communities together with environmental conditions and pollution profiles that can differentially impact the biodegradation rate, as shown for biphenyl by Chen and co-authors. These differences may reflect in the incorporation of 13C by bacterial cells and consequently in DNA-labelling: for instance, in other works labeled DNA was detected only after fourteen days of incubation with biphenyl. As 4-CB has a slower depletion rate compared to biphenyl due to the presence of chlorine, it is possible that the conditions of the soil samples used in this study, such as the presence of other classes of contaminants in addition to PCBs, determined a longer time for 4-CB mineralization and for the incorporation of derived carbon into bacterial DNA. The results of 16S rRNA amplicon sequencing from the DNA of the T0 soils before the 4-CB microcosm enrichment, confirmed that the applied treatments, i.e. biostimulation, periodic flooding, and their combination, induced the differentiation of the structure and taxonomic composition of the total bacterial community, corroborating previous information on beta-diversity obtained by DNA fingerprinting targeting the 16S rRNA- 23S rRNA intergenic spacers. Considering the dominant taxa, plant biostimulation with P. arundinacea induced an increase of Actinobacteria, though this effect was partially attenuated by the soil flooding. The phyla Chloroflexi and Desulfobacterota (formerly included in the class Deltaproteobacteria) were significantly enriched in soils subjected to the redox cycle, suggesting the occurrence of reductive dechlorination metabolisms, potentially able to reduce the concentration of highly chlorinated molecules. Members of dehalorespiring taxa, such as Dehalococcoides and Geobacter, have been in fact studied for their PCB dehalogenation activity. Besides the effect exerted by soil treatments on the original soil microbiota, our data also report a shift towards less diverse bacterial communities over SIP incubation time in terms of alpha-diversity, with the enrichment of Proteobacteria and Actinobacteria, putatively driven by the incubation condition. The combination of SIP and 16S rRNA gene sequencing allowed us to identify the ASVs deriving carbon from 4-CB at the end of SIP incubation (D28). The bacterial cells incorporating 13C may have acquired carbon both through direct degradation of 4-CB and via cross-feeding on catabolic intermediates or other metabolites. In a time-flow experiment with parallel incubation using biphenyl or benzoate as substrates it was suggested that different taxa are responsible for either the upper or lower PCB degradation pathways, even though a partial overlap in the two communities may include also bacteria performing the whole pathway. The dataset generated in this study does not allow for a clear discrimination between the two groups, however the results are consistent with previous DNA-SIP studies, where the predominant taxa detected also herein, such as Comamonadaceae, Rhodocyclaceae, Nocardiaceae and Pseudonocardiaceae, were identified as biphenyl and/or 4-CB degraders in PCB contaminated soils and sediments. The prevalent ASVs belonged to Comamonadaceae and were classified as Hydrogenophaga or Caenimonas in Phal and Bulk soil samples, and as Methylibium in PhalRed ones. In addition to the above-mentioned DNA-SIP works, Hydrogenophaga sp. has also been characterized as a PCB degrader by Lambo and Patel whereas Caenimonas and Methylibium were associated with the degradation of aromatic compounds in soil and groundwater and may therefore be cross feeders on PCB catabolic intermediates. Among Rhodocyclaceae, Azoarcus sp., a genus that we found abundant among 13C-labeled ASVs in Bulk soils, was also identified as capable of oxidizing biphenyl in upland soil by a recent study based on a protein-SIP approach. Considering the family Nocardiaceae, Rhodococcus strains are well known PCB degraders and have been shown to carry out different degradation pathways from biphenyl to benzoate in a metagenomic study conducted on an enrichment culture growing on biphenyl. Interestingly, in the present study, the only ASV representing this bacterial family was common to three of the considered soils (Bulk, BulkRed and PhalRed) and was classified as Rhodococcus. Three Rhodococcus sp. strains were indeed previously isolated from this soil after three months of rhizoremediation treatment with reed canary grass and a redox cycle (PhalRed), and their capacity to utilize biphenyl as a sole carbon source and to degrade different PCB congeners present in the Delor 103 commercial mixture was demonstrated in vitro through a resting-cell assay. Therefore, we can hypothesize that this genus was involved in the direct degradation of 4-CB in the present study as well as in the degradation of other PCB congeners in this historically polluted soil. Although most of the ASVs derived from D28 4-CB amended microcosms were unique to each of the four soil treatments, we could not observe the same pattern comparing their relative abundance in initial T0 soils. In fact, in pre-incubation samples, the same ASVs were mostly below or close to the detection threshold, indicating that the enrichment of the detected degraders occurred over time during the incubation in microcosms supplemented with 4-CB This may imply that, likewise the total bacterial community, the alpha-diversity of active PCB-degraders was higher in the initial soils subjected to the rhizoremediation treatments, with active taxa residing in the rare biosphere. Only ASV6 within Comamonadaceae, the prevalent taxon incorporating carbon from 4-CB in Phal soil at D28, even though representing less than 1% of the bacterial communities before SIP incubation, seemed to be enriched in previously planted treatments compared to bulk soils at T0. This taxon was already observed to be positively selected by plant species within the Poaceae family, both in uncontaminated and polluted soils and to be associated with an increase in organic pollutant biodegradation. It is possible that the predominance of ASV6 among degraders in Phal soils was favored by an efficient capacity to utilize the compounds derived from root deposition that may be implied in the co-metabolism of 4-CB or its catabolic intermediates. Instead, among Comamonadaceae in PhalRed soils we observed the enrichment of ASV4 classified as Methylibium, a facultative aerobe previously described for PAH anaerobic degradation, that could have been favored in this partially anoxic environment. In a former study, we showed that rhizoremediation with P. arundinacea alone or in combination with soil flooding resulted in a significant reduction in the concentration of low chlorinated PCBs (≤ pentaCBs) by stimulating microbial activity and inducing a shift in the overall soil bacterial community structure. In this study, we described at taxonomic level how these treatments shaped the structure of soil bacterial communities and we observed that Actinobacteria were enriched by the plant biostimulation. Using 4-CB as a metabolic tracer for DNA Stable Isotope Probing (SIP), we also provided new insights into the populations metabolically active in PCB degradation, identifying members of Actinobacteria and Betaproteobacteria as the main players involved. Altogether this information pointed out that Actinobacteria and Betaproteobacteria are the key bacterial targets to be enriched in a plant biostimulation-based approach aimed at the on-site bioremediation of the highly and historically PCB-polluted site considered in our study. Future research perspectives include in situ-SIP coupled with metagenomic analysis and will allow to deepen the investigation of plant-microbiome interactions and increase the knowledge of the metabolic reactions occurring during the PCB biodegradation in the field. Supplementary Information 1.Supplementary Information 2.Supplementary Information 3.Supplementary Information 4.

PMC9649799

Weiyin Wu,Xiangjun Kong,Yanhan Jia,Yihui Jia,Weimei Ou,Cuilian Dai,Gang Li,Rui Gao

An overview of PAX1: Expression, function and regulation in development and diseases 10.3389/fcell.2022.1051102

28-10-2022

Pax1,transcription factor,embryonic development,diseases,cancer

Transcription factors play multifaceted roles in embryonic development and diseases. PAX1, a paired-box transcription factor, has been elucidated to play key roles in multiple tissues during embryonic development by extensive studies. Recently, an emerging role of PAX1 in cancers was clarified. Herein, we summarize the expression and functions of PAX1 in skeletal system and thymus development, as well as cancer biology and outline its cellular and molecular modes of action and the association of PAX1 mutation or dysregulation with human diseases, thus providing insights for the molecular basis of congenital diseases and cancers.

An overview of PAX1: Expression, function and regulation in development and diseases 10.3389/fcell.2022.1051102 Transcription factors play multifaceted roles in embryonic development and diseases. PAX1, a paired-box transcription factor, has been elucidated to play key roles in multiple tissues during embryonic development by extensive studies. Recently, an emerging role of PAX1 in cancers was clarified. Herein, we summarize the expression and functions of PAX1 in skeletal system and thymus development, as well as cancer biology and outline its cellular and molecular modes of action and the association of PAX1 mutation or dysregulation with human diseases, thus providing insights for the molecular basis of congenital diseases and cancers. Embryonic development is under tight regulation of transcription factors, dysfunction of which could lead to congenital defects, or even early embryonic lethality. Thus, revealing the roles of key transcription factors not only helps us to better understand normal embryonic development, but also the etiology of birth defects. PAX1 is a transcription factor, playing crucial roles in diverse biological processes. It belongs to the paired box-containing (PAX) gene family which is highly conserved in both vertebrates and invertebrates (Thompson et al., 2021). Pax1 was initially identified from mouse in 1988 which showed highly homology to all the three paired-box genes isolated from Drosophila in 1986, called paired (prd), gooseberry proximal (gsb-p) and gooseberry distal (gsb-d) (Deutsch et al., 1988). PAX1 localizes to chromosome 2 in mouse and chromosome 20 in human (Chalepakis et al., 1991; Stapleton et al., 1993). PAX1 contains a 128-amino acid paired-box domain (PD) which is responsible for DNA-binding and an 8-amino acid octapeptide domain (OP), which exhibit extremely high levels of sequence identity among different species (Figure 1). Both PD and OP are involved in protein-protein interactions as reported in the PAX family. For example, PD in PAX3 is essential for its interaction with SOX10 (Lang and Epstein, 2003), while OP is required for the interactions between PAX5 and GRG4 (Eberhard et al., 2000). PAX1 is essential for the development of multiple tissues during embryogenesis, such as thymus (Wallin et al., 1996; Su and Manley, 2000; Su et al., 2001; Yamazaki et al., 2020), vertebral column (Wallin et al., 1994; Smith and Tuan, 1995; Furumoto et al., 1999; Peters et al., 1999; Sivakamasundari et al., 2017), chondrogenic differentiation and chondrocyte maturation (Rodrigo et al., 2003; Takimoto et al., 2013). In this review, we will summarize not only the roles of PAX1 in embryonic development and relevant diseases associated with PAX1 mutations, as well as its molecular modes of action during embryogenesis, but also significant advances in its emerging roles in various kinds of cancers, therefore providing insights for the molecular basis of PAX1 in congenital defects and cancers. During mouse embryonic development, Pax1 transcripts can be detected as early as E8.5 in the ventromedial part of newly formed somites when they are undergoing de-epithelialization (Wallin et al., 1994), as well as in the endoderm of the foregut region (Wallin et al., 1996). At E9.5, Pax1 starts to express in the sclerotome cells and becomes stronger at E10.5 in a subset of sclerotome cells which will migrate towards the notochord (Wallin et al., 1994). Meanwhile, Pax1 is also present in the limb buds (Koseki et al., 1993; Timmons et al., 1994), as well as in the first three pharyngeal pouches, which will fuse with ectodermal cells of the third branchial cleft to form the early thymic epithelial primordium and then from E11.0 onwards is populated by lymphoid progenitor cells (Wallin et al., 1996). At E12.5, when sclerotome cells migrated to locate close to the notochord, Pax1 can be detected in the condensation part which contributes to the intervertebral discs (IVDs), but not in the cells giving rise to the vertebral bodies (VBs) anlagen (Deutsch et al., 1988; Wallin et al., 1994). It also appears in the proximal part of the ribs, facial mesenchyme, sternum and pectoral girdle (Koseki et al., 1993; Timmons et al., 1994). In the meantime, Pax1 shows expression in a large number of epithelial cells in the thymus anlagen, that continues to be maintained up to the adult thymus, even though at which stage Pax1 can only be observed in a small number of cortical epithelial cells (Wallin et al., 1996). Till E14.5, Pax1 expression is confined to the anlagen of the IVDs and in a layer of cells in the perichondrium surrounding the VBs anlagen, but not in the ossifying vertebrae (Deutsch et al., 1988; Wallin et al., 1994). Similar observations were shown in quail embryos that Pax1 expression becomes restricted to the IVDs and the perichondrium of the VBs, as well as the connective tissue surrounding the spinal ganglia, after the Pax1-positive sclerotome cells migrated to surround the notochord (Ebensperger 1995). While in chick embryos, it is interesting to note that Pax1 not only shows expression in the IVDs, but also in the chondrocytes of immature vertebral bodies, so far unreported for mouse Pax1 (Peters et al., 1995). In situ hybridization and Northern blot hybridization showed that Pax1 is also expressed in the limb buds and pharyngeal pouches in developing chick embryos and continuously to be detected in the adult thymus (Peters et al., 1995). In Xenopus embryos, Pax1 transcripts starts to be detected in early somitogenesis (st. 17) and becomes more and more abundant (st. 20–45) in the sclerotome and endodermal pharyngeal pouches, as reported for other vertebrates (Sanchez and Sanchez, 2013). Zebrafish has two pax1 paralogues (pax1a and pax1b), both of which exhibit expression in the developing pharyngeal pouches, and sclerotomes from 18 to 96 hpf (Liu Y. H. et al., 2020). Altogether, the expression profiles of Pax1 in vertebrates show a highly restricted pattern and are mostly comparable to each other in different species, which might be due to the highly conserved sequences of Pax1 (Figure 1). In vertebrates, spontaneous mutants and genetic approaches help to reveal the biological functions of Pax1 in embryonic development. In accordance with its expression profile, Pax1 is essential for the development of axial skeleton, limb bud and pectoral girdle, as well as pharyngeal pouches-derived tissues, especially thymus (Chalepakis et al., 1991; Timmons et al., 1994; Wallin et al., 1994; Furumoto et al., 1999; Peters et al., 1999; Su and Manley, 2000; Su et al., 2001; Aubin et al., 2002; Sivakamasundari et al., 2017; Yamazaki et al., 2020). An allelic series of spontaneous mouse mutants have been reported and are important for understanding the critical roles of Pax1 in the development of skeleton system (Chalepakis et al., 1991; Timmons et al., 1994; Wallin et al., 1994; Dietrich and Gruss, 1995). Undulated (un) mice, carrying a point mutation with a Gly-Ser exchange at position 15 in the conserved part of the paired-box domain of Pax1, which dramatically decreases the DNA-binding affinity and alters the DNA-binding specificity of Pax1, showed kinked tails and skeletal deformity, suggesting a role of Pax1 in the skeleton formation (Wright, 1947; Chalepakis et al., 1991). Undulated extensive (un-ex) mice carry a deletion of at least 28.2 kb, removing the terminal Pax1 exon including the poly A signal, leading to a drastically reduced amount of Pax1 transcripts (Dietrich and Gruss, 1995). It showed severe skeletal malformations in homozygous animals, and only occasional mild skeletal abnormalities have been described in heterozygotes, similar with un mutant. Thus, un-ex and un are regarded to be recessive (Wallin et al., 1994; Wilm et al., 1998). Whereas, undulated short tail (un-s), is semidominant as heterozygotes exhibit clear skeletal abnormalities including a very short and strongly kinked tail (Wallin et al., 1994). Homozygous Pax1 un-s mice die perinatally displaying the most severe skeletal malformations among the undulated alleles (Wallin et al., 1994; Wilm et al., 1998). This can be explained by later findings of the molecular basis of un-s mutant mouse that harbors a Pax1 deletion interval of 125 kb, affecting four physically linked genes within or near the deletion, including Pax1, Nkx2-2, and their potential antisense genes (Kokubu et al., 2003). Moreover, another spontaneous mutant mouse, called scoliosis (sco), carries a new allele of Pax1 (un-i, undulated intermediate). The Pax1un-i allele is lacking the 5′-flanking region and exon 1 to 4 which is mapped to nt −2636 to 640 and −272 to 4271 of the Pax1 gene. Homozygous sco mice show a mild form of the known phenotypes of other Pax1 mutants and have a lumbar scoliosis and kinky tails in adults (Adham et al., 2005). In addition, the proximal part of the ribs are missing or severely malformed in Pax1 mutant mice (Wallin et al., 1994), suggesting an essential role of Pax1 for the development of the proximal part of the ribs. Moreover, Pax1 null mice generated by gene targeting showed strong skeletal abnormalities all along the vertebral column, in the sternum, and in the scapula, similar as those found in the mutant mice, with phenotypic differences in degrees of severity (Wilm et al., 1998), suggesting its roles in these tissues in accordance with its expression profile. In chick embryos, injection of an antisense oligodeoxynucleotide (ODN) against Pax1 resulting in decreased expression of Pax1 transcript leads to the loss of somite and/or disordered somite phenotype, implying the importance of Pax1 in proper segmentation of the somites and in sclerotomal differentiation (Smith and Tuan, 1995; Hol et al., 1996). Pax1 not only functions in the axial patterning, but also is involved in the formation of appendicular skeleton in chick embryos. Embryonic chick wings show morphological defects in the shoulder girdle when Pax1 expression is reduced by ectopic application of signaling molecules (BMP2 and BMP4) (Hofmann et al., 1998), which parallel the defects seen in Pax1 mutant mice (Timmons et al., 1994; Dietrich and Gruss, 1995). In zebrafish, microinjection of morpholino- (MO-) modified antisense oligonucleotides against pax1b induced pectoral fin bud defects which could not be phenocopied by pax9 MO and could not be rescued by either pax1a or pax9 overexpression, whereas could be partially rescued by mouse Pax1 mRNA injection (Liu et al., 2013; Chen et al., 2014). Loss-of-function of pax1b in zebrafish affects the expression of col2a1, uncx4.1, noggin3 and aggrecan, suggesting its roles in chondrocytes differentiation (Chen et al., 2014). While in chick embryos, Pax1 acts synergistically with Pax9, mediating Shh signaling from the notochord and the floor plate, and directly induces Bapx1 (also known as Nkx3.2) expression in chondrogenic differentiation of sclerotomal cells (Rodrigo et al., 2003) (Figure 2). Whereas, as cartilage formation proceeds, overexpression of Pax1 in chick embryos inhibits chondrocyte maturation, but does not affect mesenchymal condensation prior to cartilage formation and the subsequent early differentiation processes to give rise to proliferating chondrocytes (Takimoto et al., 2013). Cell culture studies showed forced expression of Pax1 in chondrocytes induced a morphological change from polygonal to fibroblastic, significant decrease in proteoglycans (PGs) accumulation, and downregulation of cartilage marker genes including Chm1, Col2a1, Aggrecan, Ihh, Nkx3.2 and Sox9, which can be partially rescued by Sox9 overexpression, suggesting Pax1 antagonizes the Sox9-driven chondrocyte maturation to act as a negative regulator during chondrogenic differentiation (Takimoto et al., 2013). This is in agreement with the study that PAX1/9 competes with SOX9 for occupancy of the binding site on the enhancer of Aggrecan, resulting in its reduced transactivation in the annulus fibrosus (AF) of IVDs (Takimoto et al., 2019). Consider that the expression of Pax1 becomes downregulated and restricted to the mesenchymal condensations that give rise to the anlagen of the IVDs, and to the perichondrium surrounding the cartilaginous VBs in physiological conditions (Wallin et al., 1994), supporting the notion that Pax1 plays multifaceted roles during the vertebral formation in a spatiotemporal and context-dependent manner. Using double mutant animal models, PAX1 was shown to act cooperatively with its cofactors PAX9 and MFH1 (also known as FOXC2) to regulate the proliferation of sclerotome cells during vertebral column formation (Furumoto et al., 1999; Peters et al., 1999), and cooperates with Hoxa5 in the development of the pectoral girdle (Aubin et al., 2002; Figure 2). Pax1/Pax9 double mutant mice completely lack the medial derivatives of the sclerotomes, the VBs, IVDs and the proximal parts of the ribs, which is much more severe than that of Pax1 single homozygous mutants (Peters et al., 1999). Whereas formation and anteroposterior polarity of sclerotomes, as well as induction of a chondrocyte-specific cell lineage, appear not affected (Peters et al., 1999). Reduced cell proliferation in the ventromedial regions of the sclerotomes was observed after sclerotome formation (Peters et al., 1999), implying that PAX1 works in concert with PAX9 in the migration, proliferation and mesenchymal condensations of the sclerotomes, but not in the earlier processes of somitogenesis. A similar effect was observed for MFH1 and PAX1. The differentiation of somites into sclerotome, myotome, and dermatome occurs in Mfh1/Pax1 double mutants, indicating that MFH1 and PAX1 are not essential for sclerotome formation (Furumoto et al., 1999). They function synergistically in the regulation of proliferation or migration of sclerotome cells toward the notochord, resulting in more severe malformation of the vertebral column in Mfh1/Pax1 double mutant animals (Furumoto et al., 1999). As a transcription factor, PAX1 is able to bind to the promoter region and regulate the expression of its downstream target genes. However, there are few direct target genes of PAX1 uncovered so far. There is no doubt that emerging molecular technologies will help in discovery of PAX1 target genes and revealing of the molecular mechanisms underlying its roles in skeleton development. Recently, more PAX1 downstream targets were found by a microarray in mouse vertebral column tissue, such as Wwp2, Col2a1 and Hip1, which are cooperatively regulated and occupied by PAX9 in mouse IVDs of the axial skeleton as shown by chromatin immunoprecipitation sequencing (ChIP-Seq) (Sivakamasundari et al., 2017). However, no evidence showed whether they are directly bound by PAX1 due to lacking of a good ChIP-grade antibody against PAX1 in the study (Sivakamasundari et al., 2017). Moreover, Shh signals from the notochord are necessary for the formation of the sclerotome from the epithelial somite which does not require PAX1, but the induction of PAX1 expression by Shh signals is essential for the specification of the sclerotome cells to obtain ventral or dorsal characteristics (Koseki et al., 1993; Furumoto et al., 1999). Noggin, which encodes a BMP antagonist, was required for Shh-mediated induction of sclerotomal development either by activating PAX1 alone or acting synergistically with Shh (Macmahon 1998), whereas BMP2 and BMP4 proteins supplied by the paraxial mesoderm itself, or by adjacent tissue completely abolished PAX1 induction in response to Noggin, Shh, or their combination (Macmahon 1998) (Figure 2). Overall, the signaling molecules cross talk with each other within the sclerotome cells and with tissues nearby, playing a role in PAX1 expression regulation and form a regulatory network together to control skeleton formation. In summary, PAX1 does not function in the formation of the sclerotome from epithelial somite, but is essential for the specification, proliferation, migration, as well as differentiation of sclerotome cells in the following stages after sclerotome formed, thus plays critical roles in the development of the skeleton system. We summarized the reported molecular modes of action of PAX1 in Figure 2. But the regulations of PAX1 are far from clear. For example, questions still remain that the regulation of PAX1 at transcriptional and post-translational levels are almost lacking. Answering these questions will not only help us to understand the biology of PAX1 better, but may also reveal novel diagnostic approaches that we can use in PAX1-associated diseases. The roles of Pax1 in thymus development have also been extensively investigated, since Pax1 is highly expressed in the third pharyngeal pouch, which gives rise to the thymus epithelium (Wallin et al., 1996). Mutations in Pax1 gene in un, un-ex, un-s mice all affected not only the total size of the thymus but also the maturation of thymocytes (Wallin et al., 1996). Further studies clarified that mutations in Hoxa3 and Pax1 act synergistically to cause defective thymic epithelial cell development, resulting in thymic ectopia and hypoplasia (Su and Manley, 2000), due to increased death and decreased proliferation of thymic epithelial cells caused by altered Hoxa3-Pax1 genetic pathway during thymus organogenesis (Su et al., 2001). Importantly, PAX1 mutants (Cys368*; Asn155del and Val147Leu) were identified in otofaciocervical syndrome type 2 (OTFCS) patients with severe combined immunodeficiency (SCID), which characterized by severe T cell lymphopenia, causing increased susceptibility to viral, bacterial, and fungal infections since early in life, failed to attain T cell reconstitution after allogeneic hematopoietic stem cell transplantation (Yamazaki et al., 2020). It showed an altered conformation and flexibility of the paired-box domain and reduced transcriptional activity of PAX1 (Yamazaki et al., 2020), which lead to an altered transcriptional profile in iPS (induced pluripotent stem cells) derived thymic epithelial progenitor cells, suggesting an essential role of PAX1 in human thymus development and function as well (Yamazaki et al., 2020). Consistent to its function in axial skeleton development, PAX1 coding variants were identified in the heterozygous state in exon 4 in two male patients with congenital vertebral malformations (Giampietro et al., 2005). One patient who showed thoracic vertebral segments T9 hypoplasia, T12 hemivertebrae and absent T10 pedicle, incomplete fusion of T7 posterior elements, ventricular septal defect, and polydactyly, had CCC (Pro) to CTC (Leu) substitution at amino acid 410, while the other patient had a T11 wedge vertebra and a missense mutation at amino acid 413 corresponding to CCA (Pro) to CTA (Leu) (Giampietro et al., 2005) (Figure 1). Since no molecular studies shown in the report, whether these two PAX1 mutations are responsible for the phenotypes shown in the patients need further investigations. PAX1 mutations were also detected in 8 out of 63 patients with Klippel-Feil syndrome (KFS), which is a human congenital condition, characterized by failed segmentation of the cervical vertebrae with the clinical sequelae of a short, immobile neck and a low posterior hairline (McGaughran et al., 2003). Among the 8 cases, three were missense mutations; two patients had intronic changes and three of them had silent mutations. The missense mutations of PAX1 (Pro61Ala; Ala283Pro; Gly289Ser) (Figure 1) could potentially have a pathogenic role, and the question remains whether PAX1 alone, or in conjunction with other genetic or environmental factors, plays a role in the pathogenesis of KFS, need to be confirmed by functional and molecular mechanism studies. These cases are reminiscent of the haploinsufficient roles of PAX1 in the development of skeleton system (Wilm et al., 1998). Moreover, two cases of Jarcho-Levin syndrome (JLS), with severe developmental alterations in the thoracic and vertebral skeleton, including “crab-like” thorax, typical clinical phenotypes of JLS, which have been revealed to stem from dysmorphogenetic reaction of the blastogenic axial skeleton developmental field, showed a significant reduction in protein levels of PAX1 and PAX9 in chondrocytes of the vertebral column (Bannykh et al., 2003). This report implies for the etiology and pathogenesis of JLS, in accordance with the synergistic roles of PAX1 and PAX9 in vertebral column development and chondrogenic differentiation of sclerotomal cells (Peters et al., 1999; Rodrigo et al., 2003). Though the report showed altered expression of PAX1 and PAX9 in JLS, the possibility remains that mutations or dysfunctions of the upstream factors of PAX1 and PAX9 might be involved in the pathogenesis of JLS. Moreover, a PAX1 missense mutation with changing of Gln to His at position 42 in the paired-box domain (Gln139His) was identified in one patient with spina bifida, a kind of congenital malformation due to the neural tube defect (NTD) (Hol et al., 1996). This reminiscent of the phenotypes with a high incidence of lumbar spina bifida in the mouse model crossed of un mutant and Patch (Ph) mutant mice (Helwig et al., 1995) which is a deletion of the gene encoding the platelet-derived growth factor receptor alpha subunit (PDGFRα) (Stephenson et al., 1991), suggesting the possibility of an involvement of PAX1 in the occurrence of NTD. Interestingly, PAX1 was shown to regulate the transcriptional activity of PDGFRα gene by a luciferase reporter assay, however, additional studies need to be done to indicate whether PAX1 actually binds to the PDGFRα promoter directly, or rather mediates its effect via interaction with other components of the transcription machinery (Joosten et al., 1998). Since mutation of PAX1 itself is not sufficient to cause NTD as shown in PAX1 mutant animals, the genetic interaction between PAX1 and PDGFRα, together with signaling cross talk between the sclerotome cells and the axial tissues that contribute to the neurulation, could increase the risk of NTD formation. In addition, whole-exome sequencing (WES) of a single pooled DNA sample of four affected individuals in a large consanguineous family with otofaciocervical syndrome (OTFCS) from Turkey, with cup-shaped ears, bilateral mixed hearing loss, bilateral preauricular fistulas, lacrimal duct abnormalities, protruding shoulders, and winged scapulae, identified a homozygous variant (c.497G>T), substituted the glycine at position 166 to valine (p.G166V) within the highly conserved paired-box domain (Figure 1), leading to a significantly reduced transactivation activity of PAX1 (Pohl et al., 2013). Later on, a nonsense homozygous mutation (c.1104C>A, p. Cys368*) and a homozygous small insertion (c.1173_1174insGCCCG, p. Pro392Alafs*19) in PAX1 gene were also identified in OTFCS patients (Paganini et al., 2017; Patil et al., 2018). The major clinical features of OTFCS include ear malformations, facial dysmorphism, shoulder girdle abnormalities, vertebral anomalies, and mild intellectual disability (Patil et al., 2018). Interestingly, some PAX1 mutations (Cys368*; Asn155del and Val147Leu) (Figure 1) found in OTFCS patients accompanied with severe combined immunodeficiency (SCID), who exhibited not only dysmorphic facial features, malformed vertebral bodies and appendicular structures, but also thymus aplasia, showed reduced transcriptional activity of PAX1 on its target gene (Paganini et al., 2017; Yamazaki et al., 2020). These reports confirmed the association of PAX1 gene with facial morphology (Adhikari et al., 2016; Qian et al., 2021), and agree with the roles of PAX1 in the extravertebral structures observed in mice, including the sternum, the scapula and the facial skull (Dietrich and Gruss, 1995), as well as its functions in thymus development (Wallin et al., 1996). Altogether, almost all the human syndromes associated with PAX1 mutations identified so far are associated with the functions and molecular roles of PAX1 in corresponding tissues reported in animal models, revealing the conserved roles of PAX1 among different species. Thus, the studies in animal models could help people to understand the pathogenesis of human diseases and aid to develop new therapeutic strategies. Epigenetic modifications have long been tightly connected with cancer (Dawson and Kouzarides, 2012). DNA methylation catalyzed by DNA methyltransferases (DNMTs) is one of the essential epigenetic modifications that control cell proliferation, apoptosis, differentiation, cell cycle, and transformation in eukaryotes (Pan et al., 2018). Typically, DNA methylation generates a stable epigenetic mark associated with silencing of gene expression (Moore et al., 2012). The genome-wide landscape of DNA methylation in cancer cells is largely altered compared with normal cells (Nishiyama and Nakanishi, 2021). In particular, the early findings have unraveled that aberrant DNA methylation promotes cellular oncogenesis through silencing of tumor suppressor genes (Nishiyama and Nakanishi, 2021). In the last decade, a growing body of evidence has implied an emerging tumor suppressing role of PAX1 in various human cancers, including cervical cancer (Huang et al., 2010; Kan et al., 2014; Lai et al., 2014; Liu H. et al., 2020; Li et al., 2021), ovarian cancer (Hassan et al., 2017), colorectal carcinoma (Huang et al., 2017), parathyroid tumor (Singh et al., 2022), as well as oral squamous cell carcinoma (Huang et al., 2014; Cheng et al., 2018; Sun et al., 2020), and so on. Higher DNA methylation levels of PAX1 were observed in most kinds of cancer cells (Lai et al., 2008; Huang et al., 2010; Cheng et al., 2016; Huang et al., 2016; Huang et al., 2017; Su et al., 2019; Tang et al., 2019; Zhao et al., 2020; Singh et al., 2022), significantly strengthened the observation that PAX1 often acts as a tumor suppressor. This is a little bit surprising, since PAX1 is only highly expressed in the early embryonic stages while its expression becomes undetectable or very low in the majority of adult human tissues according to the Genotype-Tissue Expression (GTEx) database (Thompson et al., 2021), which is consistent with the observation in mouse model (Deutsch et al., 1988). One possibility could be reactivation of PAX1 occurs under certain circumstances such as drug or stress-induced conditions. PAX1 reexpression was observed in cervical cancer cell lines after treatment with curcumin and resveratrol, may be due to their effect on histone deacetylase mediated through downregulation of UHRF1 which can regulate both DNA methylation and histone acetylation (Parashar and Capalash, 2016). Reactivation of PAX1 due to promoter hypomethylation has been achieved through silencing of DNMT1 in Hela and Siha cell lines (Zhang et al., 2011), suggesting DNMT1 might be the methytransferase responsible for the hypermethylation level of PAX1 in cervical cancer cells. So far, most reports of PAX1 in cancer cells focus in the methylation status of PAX1, limited articles showed further molecular mechanisms in these processes. Recently, a report showed that PAX1 plays a tumor suppressing role by forming a complex with WDR5 and SET1B, leading to increased trimethylation of histone H3, lysine 4 (H3K4me3), thus activates the expression of multiple phosphatases including PTPRR, DUSP1, DUSP5, and DUSP6 in cervical cancer cells (Figure 2), maintaining the homeostasis between kinases and phosphatases in cervical epithelium, revealed a functional relevance of PAX1 in cancer biology (Su et al., 2019). Epigenetic changes are recognized to occur in the early stage of tumor progress and in advanced of genetic alterations, thereby affording a reason for developing biomarkers for early identification and prevention of cancer (Reis et al., 2016). Therefore, PAX1 methylation level not only can provide biomarkers for early detection and diagnosis in cancer patients, but also could predict and monitor early therapeutic response (Pan et al., 2018; Li et al., 2021). Given its roles in the development of multiple tissues, PAX1 might regulate cell proliferation and differentiation of specific cancer cells and thus contributes to the activation or suppression of cancer development in specific tissue contexts when being dysregulated. As a transcription factor, PAX1 could also drive the suppression or activation of its downstream target genes and participate in the regulation of signaling pathways, or play a role in the epigenetic regulation on its target genes, which may further facilitate or inhibit tumorigenesis in a context-dependent manner. PAX1 is a highly conserved gene identified early in embryonic development. Its conserved roles can be demonstrated by similar observations in different species. For example, Pax1 is essential for fin bud development in zebrafish, wing bud development in chick and limb bud development in mouse embryos (Timmons et al., 1994; Hofmann et al., 1998; Chen et al., 2014). Interestingly, as a transcription factor in the same subfamily with similar protein domains, PAX9 was often investigated in parallel with PAX1. Undoubtedly, PAX1 and PAX9 play synergistic and redundant roles in some tissues, such as the axial skeleton (Balling et al., 1996; Peters et al., 1999; Sivakamasundari et al., 2017), while they also have their own specific expression and functional tissues as well (Chen et al., 2014; Sivakamasundari et al., 2018). It was found that PAX9 was unable to fully compensate for the loss of PAX1 but PAX1 could fully rescue PAX9 deficiency during the vertebral column formation probably by its own expression level via a positive auto-feedback mechanism, suggesting PAX1 is the more dominant player in the axial skeleton development (Peters et al., 1999; Sivakamasundari et al., 2017). Whereas, PAX9 plays essential roles in the palatogenesis and odontogenesis in which PAX1 is largely absent (Peters and Balling, 1998; Kist et al., 2005; Zhou et al., 2013; Jia et al., 2020a; Jia et al., 2020b). In addition, PAX9 expression is not affected when PAX1 is mutated suggesting the expression of PAX9 is not dependent on PAX1 (Neubüser et al., 1995; Sivakamasundari et al., 2017). More interestingly, PAX1 may act downstream of PAX9 in the expanding taste progenitor field of the mouse circumvallate papilla (Kist et al., 2014). Altogether, further studies will be needed to address the clearer relationships between PAX1 and PAX9 in a tissue-specific level. Even though the expression and functions of PAX1 have been elaborated in details, the cellular and molecular mechanisms that underlie its roles in multiple tissues are far from well-defined. One possibility may be no good commercial PAX1 antibody for biochemical and molecular studies; another reason could be knocking down or knocking out PAX1 in the molecular level is not efficient enough with common used methods such as siRNA mediated gene silencing techniques. With the development of the state-of-the-art techniques and the implement of emerging powerful tools of molecular and genetic studies such as high-throughput sequencing and CRISPR-Cas9 genome editing techniques, more and more PAX1 interacting partners and PAX1 direct targets in different tissues could be brought to light. Furthermore, the integrating of PAX1 activities to the regulatory network orchestrating embryonic development and diseases is a far-reaching task which will help us to see the whole picture and provide clinical implications in the future.

PMC9649800

Amal Bajaffer,Katsuhiko Mineta,Pierre Magistretti,Takashi Gojobori

Lactate-mediated neural plasticity genes emerged during the evolution of memory systems

10-11-2022

Computational biology and bioinformatics,Evolution,Neuroscience

The ability to record experiences and learning is present to different degrees in several species; however, the complexity and diversity of memory processes are cognitive function features that differentiate humans from other species. Lactate has recently been discovered to act as a signaling molecule for neuronal plasticity linked to long-term memory. Because lactate is not only an energy substrate for neurons but also a signaling molecule for plasticity (Magistretti and Allaman in Nat Rev Neurosci 19:235–249, 2018. https://doi.org/10.1038/nrn.2018.19), it is of particular interest to understand how and when memory-related genes and lactate-mediated neural plasticity (LMNP) genes emerged and evolved in humans. To understand the evolutionary origin and processes of memory and LMNP genes, we first collected information on genes related to memory and LMNP from the literature and then conducted a comparative analysis of these genes. We found that the memory and LMNP genes have different origins, suggesting that these genes may have become established gradually in evolutionarily and functional terms and not at the same time. We also found that memory and LMNP systems have a similar evolutionary history, having been formed with the gradual participation of newly emerging genes throughout their evolution. We propose that the function of LMNP as a signaling process may be evolutionarily associated with memory systems through an unidentified system that is linked by 13 common genes between memory and LMNP gene sets. This study provides evolutionary insight into the possible relationship between memory and the LMNP systems that deepens our understanding of the evolution of memory systems.

Lactate-mediated neural plasticity genes emerged during the evolution of memory systems The ability to record experiences and learning is present to different degrees in several species; however, the complexity and diversity of memory processes are cognitive function features that differentiate humans from other species. Lactate has recently been discovered to act as a signaling molecule for neuronal plasticity linked to long-term memory. Because lactate is not only an energy substrate for neurons but also a signaling molecule for plasticity (Magistretti and Allaman in Nat Rev Neurosci 19:235–249, 2018. https://doi.org/10.1038/nrn.2018.19), it is of particular interest to understand how and when memory-related genes and lactate-mediated neural plasticity (LMNP) genes emerged and evolved in humans. To understand the evolutionary origin and processes of memory and LMNP genes, we first collected information on genes related to memory and LMNP from the literature and then conducted a comparative analysis of these genes. We found that the memory and LMNP genes have different origins, suggesting that these genes may have become established gradually in evolutionarily and functional terms and not at the same time. We also found that memory and LMNP systems have a similar evolutionary history, having been formed with the gradual participation of newly emerging genes throughout their evolution. We propose that the function of LMNP as a signaling process may be evolutionarily associated with memory systems through an unidentified system that is linked by 13 common genes between memory and LMNP gene sets. This study provides evolutionary insight into the possible relationship between memory and the LMNP systems that deepens our understanding of the evolution of memory systems. Memory is the process that allows humans to retain information over time. The physiological properties of memory have been widely examined, revealing that the function of memory is conserved across species of vertebrates and invertebrates. Nevertheless, it is challenging to interpret the evolution of memory, focusing only on the physiological data. Besides, a number of studies that have concentrated on the evolution of a memory system is limited. Functionally, memory systems require the involvement of many genes; comprehensive research of memory-related genes is required due to the accumulation of genomics and relevant data to understand the origin and evolution of memory. Pellerin and Magistretti proposed that lactate released by astrocyte can be a functional energy source for neurons, through a mechanism known as the astrocyte-neuron lactate shuttle (ANLS); later Magistretti and colleagues showed that lactate is an intercellular signaling molecule by effectively stimulating the expression of some neural plasticity genes that are involved in the process of memory formation. Lactate is a metabolic intermediate that various species can aerobically and anaerobically produce in their tissues, and is a known energy source for humans, providing energy not only to muscles, but also to the heart and brain. Although the human brain constitutes only 2% of total body weight, the metabolic demands of the brain consume about 20% of the total oxygen consumption in the human body, and the energy required for appropriate brain function represents approximately 25% of the body’s entire glucose usage. As a signaling molecule, lactate changes the redox state of neurons once it is converted to pyruvate and also increases the intracellular NADH/NAD ratio. Consequently, it potentiates the activity of the N-methyl-D-aspartate receptor (NMDAR) to raise the calcium influx inducing the expression of neural plasticity-related genes. Besides, lactate has signaling features through the hydroxycarboxylic acid receptor 1 (HCAR1). Scavuzzo et al. showed the involvement of HCAR1 signaling in memory consolidation. Previous studies have revealed the evolutionary conservation of ANLS in mice and invertebrates. Lactate is produced from astrocytes by metabolizing glucose from cerebral blood vessels directly or from glycogen storage through glycolysis or glycogenolysis, respectively. Then, it is exported to neurons and can be used as a source of energy for neural activities through oxidative phosphorylation in the mitochondria. In Drosophila, for example, glial cells produce lactate and alanine from pyruvate to fuel mitochondria in neurons, indicating that metabolic coupling between glial cells and neurons is a global mechanism. In the adult brain of Drosophila, an increasing energy influx in the neural mushroom body has been described, which is the primary center of learning and memory in insects sufficient for LTM formation. Nonetheless, there is still a lack of fundamental understanding of the fluxes of energy-mediated molecules in the brain of Drosophila. The evolutionary origin of lactate dehydrogenase (LDH), an enzyme that produces lactate, has been intensively studied. Cristescu et al. suggested that L-Ldh evolved through multiple events of gene duplication, and its evolutionary history shared among vertebrates and invertebrates. Skory examined 26 LDH subunits, focusing on two subunits LDHA and LDHB from numerous hosts, including mammals, birds, fish, fungi, bacteria, and plants. The evolution of LDHA and LDHB have been previously studied, suggesting that LDHA and LDHB evolved from a single ancestral gene. It has been showed that the production of lactate by LDH is not limited in non-human animals, but it also observed in almost all organisms, from bacteria to humans. Duncan et al. discovered that some bacteria of the human colon produce lactate in low concentrations and can convert this lactate into butyrate, which is crucial for colonic health. To date, the function of lactate as a signaling molecule promoting neural plasticity and memory has only been detected in mice. It remains to be established when these memory and LMNP systems evolved. In this study, we aim to explore and understand the evolutionary origin and processes of memory-related genes, including LMNP genes. To this end, we collected orthologs in different taxonomic groups using memory and LMNP-related genes. Then, we conducted a comparative analysis of gene gain and loss of the orthologs of memory and LMNP genes to elucidate how LMNP-related genes were incorporated into the memory system. Lastly, we studied the evolutionary alteration of the functional groups of the orthologs of those genes to understand the evolutionary processes of the memory and LMNP-related genes. We conducted a literature survey by reviewing a total of 572 papers, 136 of which were not accessible and therefore not considered. We selected the papers that contained a list of more than ten genes, which resulted in 15 reviews (Supplementary Table S1). However, we excluded one review (Molecular basis of pharmacotherapies for cognition in Down Syndrome) because it was unclear if the genes were related to memory. Thus, we obtained 335 memory-related genes from the 14 published reviews. We excluded six of the memory-related genes (Dcx, Plec, Cit, Gdi1, ryr3, and Dlgap1) from the analysis because they had more than one euNOG orthologous cluster, as well as 27 memory-related genes because they did not have orthologs in humans. Therefore, the total number of memory-related genes used in this study is 302 genes. With regards to LMNP genes, we obtained 395 genes after omitting either one of 18 overlapping gene pairs between 1 and 6 h-data sets, 12 ncRNA (non-coding RNA), and one rRNA (ribosomal RNA). Seven LMNP genes (Kdm6b, Fras1, Agbl3, Met, Apaf1, Usp43, and Stk36) had more than one euNOG orthologous cluster, which we excluded from the analysis. We also removed from the analysis two LMNP proteins (Cygb and Hdac9) that were not detected by the eggnog mapper tool. Therefore, the total number of LMNP genes used in this study is 386 genes. For the reference, we collected 23,004 sequences from the mouse proteome. The eggnog mapper detected 22,574 reference sequences, 430 proteins were not found in the mapper, and no orthologous cluster or result was found. We excluded from the analysis 322 proteins that had more than one euNOG, 26 proteins that did not have an euNOG orthologous cluster, and 1505 proteins that did not have orthologs in humans. Therefore, we used a total number of 20,721 mouse amino acid sequences as references in this study. We collected 302 and 386 genes for memory and LMNP in mice, respectively. We compared these genes to detect any commonality between the memory and LMNP systems, giving 13 genes that are common between memory and LMNP systems, namely Arc, Junb, Bdnf, Kcna6, Egr1, Nr4a1, Atf4, Fos, Itga3, Bcl2l11, Hcn1, Adcyap1, and Cckbr. Some of the 13 genes have been previously detected to be required for memory formation in a lactate-dependent manner. Lactate induces the expression of several neural plasticity genes including (Arc, c-Fos, BDNF, and Egr1) that are involved in the process of LTM formation. The orthologous distribution of these genes is presented in Supplementary Table S2. The existence of overlapping genes may reflect the presence of an unidentified system that links the memory and LMNP systems. Interestingly, all these proteins have orthologs in humans and mice, suggesting that they must be essential in both systems. Further studies are needed to investigate what the unknown system might be that connects these two systems. We successfully detected memory-associated and LMNP genes based on our extensive literature review. We examined species from the following different taxonomic groups: vertebrates (Homo sapiens, Mus musclus, Gallus gallus and Danio rerio), invertebrates (Ciona intestinalis, Drosophila melanogaster, Nematostella vectensis, Trichoplax adhaerens, and Amphimedon queenslandica), Choanoflagellate (Monosiga brevicollis), and fungi (Saccharomyces cerevisiae). We based this on the variation in their taxonomic group to ensure an extensive coverage of the tree of life and because their complete proteomics are publically available. As seen in Fig. 1, vertebrates possess a substantial number of orthologous genes compared to other species in both systems. We detected 302 memory-related genes in mice, with the same number of orthologs of memory-related genes in humans. We detected 280 genes in chicken and 288 in zebrafish. The number of memory-related genes decreases significantly in invertebrates, choanoflagellate, and yeast compared to vertebrates (Fig. 1). This is consistent with our understanding that memory in vertebrates is more developed than in other species. One of the possible explanations could be that the memory system in vertebrates may have developed by increasing the number of newly participating genes to this system therefore enhancing its function. Another possible explanation is that different sets of genes could be involved in memory formation in invertebrates compared to mice. However, it is worth mentioning that memory formation in the mouse hippocampus and Aplysia, for example, has common cellular and molecular mechanisms that were highly conserved during evolution. Although the storage of short-term memory in mice and Aplysia requires different signaling, long-term storage of both species utilizes a core signaling pathway. Further components are likely recruited at least in a mouse. In this study, we collected genes related to memory systems without specifying the system considering that memory systems vary among different species. Although, the molecular mechanisms and genes involved in explicit and implicit memory have been studied in some vertebrates and invertebrates; however, due to a lack of information regarding the evolution of memory systems using genome-wide approaches, it is still unknown if there are specific genes associated with the emergence of certain memory systems. More experimental works should be done to determine whether the emergence of specific memory-related genes correlates with the formation of specific memory systems. Similarly, Fig. 1 shows the number of LMNP orthologous genes increases in vertebrates compared to other species. Remarkably, mice conserve 386 LMNP genes; however, the number of these orthologous genes is slightly reduced in humans. One explanation is that only some of these genes are conserved, and are specific for mice, such as Mex3a, Synj2, Nr1d2, Thbs1, Spred3, Ankrd33b, Chd2, Nup37, Mettl21a, Gdpd5, cml1, Rims1, Efr3a, Armc2, Nfil3, Igsf9, Peg10, Lppr5, Phf13, Lrp5, Ldlr, Stard10, Chrac1, and Stxbp5l. One further explanation could be that humans possess different and undiscovered network genes that may contribute to the LMNP system. The number of orthologous genes decreases slightly in chicken and fish by about 314 and 323 genes, respectively, and the number progressively decreases among invertebrates, choanoflagellates, and yeast, relative to vertebrates (Fig. 1). Thus, the number of orthologous genes of LMNP has consequently increased as a result of evolution. Our analysis shows that 22% of memory-related genes are conserved in yeast, indicating that these genes emerged before the memory function evolved. We identified 67 of 302 orthologous genes in yeast to be related to memory formation, suggesting that yeast is the origin of those genes (Fig. 2a). In contrast, we found that 35% (135 genes) of LMNP orthologous genes are conserved in yeast, which uses lactate as an energy source even before the emergence of neurons, as shown in Fig. 2b. Remarkably, yeast has neither neurons nor a memory system; thus, the original function of these proteins must have altered through evolution since they first emerged. For example, Gys1 gene is conserved in S. cerevisiae, and has function in the synthesis of glycogen. Lsm3 is another protein that is involved in the degradation and splicing of mRNA, the replication of telomere, and the formation of histone. Net1 is an essential protein for nucleolar structure in S. cerevisiae. Margineanu et al. experimentally examined the involvement of these genes in neural plasticity that is mediated by lactate. Further experimental works are necessary to investigate the exact role of other proteins in yeast and in species that do not possess memory. Further, we noticed that some genes have gained during the evolution of memory and LMNP systems. For example, the total number of genes conserved in the ancestral species that lack both systems such as choanoflagellate, A. queenslandica, and T. adhaerens is 150 (49%) and 113 (29%) genes, respectively (Fig. 2a,b). However, the total number of genes gained in the ancestral species of vertebrates, C. intestinalis, D. melanogaster, and N. vectensis is 85 (29%) memory genes and 114 (30%) LMNP genes (Fig. 2a,b). These novel genes must have an essential role in the development of both systems, suggested that improvement of memory in humans may have evolutionarily achieved due to the formation of very sophisticated neural networks in the brain based on those genes. The number of memory and LMNP-related genes might be affected by the Whole genome duplication (WGD) event. In fact, WGD plays an important role in evolution, particularly significant for the evolution of complex vertebrates, leading to an increase in biological complexity and facilitating the emergence of evolutionary novelties, a complex set of changes at the genetic and phenotypic level. Although WGD increases the genome complexity and supplies raw genetic materials, an extensive gene loss was reported after the WGD event. To clarify the evolutionary processes of both memory and LMNP genes, we used gene gain and loss analysis to detect any tendency in the genes during evolution. We compared the tendency of memory-related genes to the reference, i.e., the gain–loss pattern of the orthologous genes of all the proteins in mice (Fig. 2a,c). We calculated the correlation coefficient of the gene gain patten of the orthologous genes of the memory-related genes and the reference. Consequently, we detected a strong positive correlation (r = 0.814) between memory and the reference. Similarly, the examination of the tendency of gained LMNP genes helps to understand how these genes evolved. Therefore, we calculated the correlation coefficient of the gene gain pattern of the orthologous genes of LMNP genes and compared it to the gene gain pattern of the reference (Fig. 2b,c). We found a strong positive correlation between the LMNP genes and the reference genes (r = 0.937). Our analysis detected a similar tendency for the gene gain found during the development of the memory system and the LMNP system compared to the reference. We inferred functional features of memory-related proteins across species using Gene Ontology and the PANTHER classification system to examine whether evolutionary changes have occurred in the functional categories during evolution. We performed a statistical analysis [Student’s t-test with multiple test correction (p < 0.05)] to examine the differences in the functional categories between two groups: memory-possessing species and memory-lacking species. We found one significant category, which we define as “multicellular organismal processes” (Fig. 3a). To narrow down the functional categories in “multicellular organismal processes”, we analyzed two subcategories, “multicellular organismal development” and “system process”, to explore the significant differences between these two groups. We repeated the analysis on the subcategories up to the fourth layer. As a result, we obtained “Neurogenesis” as a category, which has a significant difference in the number of orthologous genes between memory-possessing and memory-lacking organisms. Fifteen proteins are related to the “Neurogenesis” category, as shown in Supplementary Table S3, as well as the species distribution of its orthologs. Nine out of 15 proteins are assigned to the “signaling molecule,” “nucleic acid binding,” “transcription factor” and “cytoskeletal protein” at the Molecular Function of Gene Ontology (Supplementary Fig. S1), indicating that the increase of “Neurogenesis” proteins in memory-possessing species may increase the neural networks required for the evolutionary development of the memory system. A growing body of evidence suggests that lactate participates in long-term potentiation (LTP) and memory formation. LTP is described as the increase in the strength of synapses following the activation of chemical synapses. O'Dowd et al. exhibited that the metabolism of glycogen from astrocytes provided energy for LTM in neonatal chicks, and the administration of glycogen inhibitor 1, 4-dideoxy-1, 4-imino-D-arabinitol (DAB), or glycolysis inhibitor 2-deoxyglucose (2DG) impaired memory performance in chickens. Suzuki et al. revealed that glycogen-derived lactate and monocarboxylate transporters (MCTs), i.e., lactate transporters MCT1, MCT2, and MCT4, mainly control LTM formation and LTP, and the blocking glycogenolysis or these transporters causes amnesia and weak synapses, leading to LTP impairment that can be rescued by lactate. Further, Netzahualcoyotzi and Pellerin have shown that decreased expression levels of MCT2 in mature neurons obstruct neurogenesis related to memory consolidation in the hippocampus, suggesting that neuronal MCT2 is required for LTM formation. To understand how LMNP proteins are involved in the LTM system, we conducted the t-test between LTM-possessing species and LTM-lacking species. LTM is a part of the memory system, and thus the species possessing LTM are the same as memory-possessing species. We performed statistical analysis for the categories and the significant subcategories between these two groups. We identified “biological adhesion," “biological regulation,” “cellular process,” “developmental process,” “immune system process,” “localization,” “metabolic process,” “multicellular organismal process,” “response to stimulus,” and “rhythmic process” as the categories that showed a significant difference between two groups (Fig. 3b). We repeated the analysis on the significant subcategories separately until the fourth level of layers, and we obtained "trans-synaptic signaling," "MAPK cascade," "apoptotic signaling pathway," "neuron development," "generation of neurons," "cellular protein modification process," "translation," "regulation of signal transduction," "regulation of cellular biosynthetic process," " regulation of phosphorus metabolic process," "cell morphogenesis involved in neuron differentiation," "neuron projection development," "embryo development ending in birth or egg hatching," "regionalization," "nervous system development," "neurogenesis," "glycoprotein biosynthetic process," "chordate embryonic development," "anterior/posterior pattern specification," "cellular response to growth factor stimulus," "response to fibroblast growth factor," and "cellular response to peptide” as categories, all of which have significant differences in the number of orthologous proteins between LTM-possessing and non-LTM-possessing organisms. A total of 86 proteins belong to these significant categories. Species distribution of those orthologous proteins is summarized in Supplementary Table S4. The majority of these proteins (more than ten) are classified as a "signaling molecule" and "transcription factor" (Supplementary Fig. S2). Therefore, the increase in the number of LMNP proteins in LTM-possessing organisms could be required to create complex pathways that enhanced the LMNP function during the evolution. Our evolutionary study showed that the memory- and LMNP-related systems seem to have similar evolutionary processes, formed gradually by increasing the number of participating genes, pointing out that these systems never took place at once during evolution. In both systems, a subset of genes had orthologs in yeast, which may be the ancestor of those genes. We found a leaping increase in memory and LMNP genes that took place just before the emergence of vertebrates during evolution. In a similar manner, increasing the number of proteins and their necessary functions among species that possess memory or LTM could be the main reason for the development of both systems. Further, a positive correlation between the gene gain of the memory and LMNP genes was detected once comparing them to the reference, suggesting a similar tendency in the evolutionary gained genes in both systems. Besides, the memory and LMNP systems commonly shared 13 genes; hence, we suggested that the function of LMNP as a signaling system could be evolutionarily linked with the memory system by an unknown system that connects them. Further experimental work should be conducted to examine the function of lactate as a signaling molecule and its association with the memory system among invertebrates to investigate its emergence during evolution. This will help to expand our knowledge to understand more about the involvement of the LMNP system in the evolutionary formation of memory. We used the PubMed database (https://www.ncbi.nlm.nih.gov/pubmed/) to collect memory-related genes, searching for the keywords Learning AND memory AND (gene OR protein) AND (Mus musculus). We sorted the search results by most recent and customized the selection to include all reviews conducted in the last 19 years (from 1/1/2000 to 19/03/2019). The number of 436 reviews were manually checked and filtered following the criteria that each must contain a list of more than ten genes. After selecting these genes, some were not subtype-specific genes. Those genes have several subtypes; it is not clear which subtype involved in memory formation. To obtain an accurate gene list, a gene that is written in general without determining its subtype was removed from the list. For example, dopamine receptor (Drd) is not specified to subtypes such as Drd1, Drd2, Drd3, Drd4, or Drd5; thus, in this case, we removed it from the list, and so on for all other genes in the list. LMNP genes were identified by a genome-wide RNA-sequencing study as described in Margineanu et al.. RNA sequencing was performed on RNA isolated from cultured cortical neurons 1 h/6 h after the addition of l-lactate to cell culture media. The expression level of each protein-coding gene was calculated via fragments per kilobase of exon per million mapped fragments (fpkm). In this study, we collected LMNP genes that were differently expressed (p < 0.05) after 1 h and 6 h of lactate treatmentand then combined these gene sets (1 h and 6 h) into one list following their definition. We converted all alias genes into official gene symbols using a python script to compile an accurate gene list and to avoid duplication. We obtained mouse genomic nucleotide accession versions and ID numbers for the all genes from NCBI-nucleotide database and NCBI-gene database, respectively. We extracted amino acid sequences of M. musculus for memory and LMNP-related genes from the nucleotide database. For the control, we downloaded the Mus musculus proteome from Refseq. We removed all isoforms by selecting the smallest accession number for a protein and we used an alias name for some LMNP proteins (cml1, Lppr5, Diap2, Mfsd4, Pvrl2, and Rp2h) as the sequences were not available using the symbol names. To detect orthologues, we used the eggNOG (evolutionary genealogy of groups: Non-supervised Orthologous Groups) database version 4.5.1 and the Eggnog mapper tool version 1. We downloaded all 335 memory and 386 LMNP amino acid sequences. The mapping mode was adjusted by default setting (Hmmer). We selected the eukaryotes taxonomic scope to search for mouse memory and LMNP related proteins. We selected all orthologs from the orthologous option, and selected the non-electronic term for the gene ontology evidence. For the control, the mapping mode adjusted by default setting (Diamond) as hmmer mode did not accept a large number of sequences. We selected and analyzed all proteins that contain one euNOG (eukaryotes) orthologous cluster and excluded those contain more than one orthologous cluster. We wrote a Perl script to detect orthologous proteins among species. Some files were essential for running the script; the first file (euNOG.members.txt.gz) contained eukaryotes non-supervised orthologous groups (euNOG) that were downloaded from eggnog version 3 (http://eggnog.embl.de/version_3.0/downloads.html), and the second file contained the species names and IDs (http://eggnogdb.embl.de/download/latest/eggnog4.core_ species_list.txt). We identified the gene gain and the gene loss by count. We used the default parameters for the analysis and analyzed the family history based on the Dollo parsimony. We conducted functional classification using the PANTHER functional classification system. We tested the functional classification of the orthologous proteins in all examined species using mouse proteins as reference. We studied the evolutionary changes of the functional categories of the orthologues of memory and LMNP proteins between species-possess memory (or LTM) such as (Homo sapiens, Mus musclus, Gallus gallus and Danio rerio, Ciona intestinalis, and Drosophila melanogaster) vs. species-lacking memory (or LTM), including (Trichoplax adhaerens, Amphimedon queenslandica, Monosiga brevicollis, and Saccharomyces cerevisiae). We excluded Nematostella from the analysis as it is unknown whether or not it possesses memory. We conducted a Welch’s t-test, applying two-tailed with unequal variance. We applied multiple testing corrections using the Benjamini–Hochberg procedure, and p values < 0.05 were considered significant in all analyses. We calculated the correlation coefficient of the gene gain pattens of the orthologous gene (i.e., a set of absolute numbers of gene gain at each node) of the memory-related genes vs the reference and LMNP-related genes vs the reference using Microsoft Office Excel software. Supplementary Information.

PMC9649808

Qiongling Peng,Yan Zhang,Binqiang Xian,Lianying Wu,Jianying Ding,Wuwu Ding,Xin Zhang,Bilan Ding,Ding Li,Jin Wu,Xiaowu Hu,Guanting Lu

A synonymous variant contributes to a rare Wiedemann-Rautenstrauch syndrome complicated with mild anemia via affecting pre-mRNA splicing

28-10-2022

Wiedemann-Rautenstrauch syndrome,Fanconi anemia,whole-exome sequencing,POLR3A,FANCA,pre-mRNA splicing

Wiedemann-Rautenstrauch syndrome (WDRTS) is an extremely rare autosomal recessive neonatal disorder. Currently, over 50 cases with variable phenotypes of WDRTS have been reported. In our cohort of prenatal and postnatal growth retardation, a female proband was found to have general growth retardation, neurocutaneous syndrome, and anemia. Karyotype test and array-CGH detected no obvious chromosomal aberrations. Trio-based whole-exome sequencing (Trio-WES) identified bi-allelic compound mutations in the coding sequence (CDS) of POLR3A gene (c.3342C > T, p.Ser1114 = and c.3718G > A, p.Gly1240Ser). For the mild anemia phenotype, the underlying causal genetic factors could be attributed to the compound heterozygous mutations in FANCA gene (c.2832dup, p.Ala945CysfsTer6 and c.1902 T > G, p.Asp634Glu). Mini-gene reporter assays revealed that the synonymous variant of POLR3A and the missense variant of FANCA could affect pre-mRNA splicing of each gene. For POLR3A, the synonymous mutation (c.3342C > T, p.Ser1114=) generated three types of aberrant isoforms. Therefore, the female patient was finally diagnosed as WDRTS caused by POLR3A. For FANCA, the missense variant (c.1902 T > G, p.Asp634Glu) disrupted the normal splicing between exon 21 and 22, and produced two types of abnormal isoforms, one carrying the 1902G and the other spliced between exon 21 and 23 to exclude exon 22. Network analysis showed that POLR3A and FANCA could be STRINGed, indicating both proteins might collaborate for some unknown functions. Current investigation would broaden the knowledge for clinicians and genetic counselors and remind them to interpret those synonymous or predicted “benign” variants more carefully.

A synonymous variant contributes to a rare Wiedemann-Rautenstrauch syndrome complicated with mild anemia via affecting pre-mRNA splicing Wiedemann-Rautenstrauch syndrome (WDRTS) is an extremely rare autosomal recessive neonatal disorder. Currently, over 50 cases with variable phenotypes of WDRTS have been reported. In our cohort of prenatal and postnatal growth retardation, a female proband was found to have general growth retardation, neurocutaneous syndrome, and anemia. Karyotype test and array-CGH detected no obvious chromosomal aberrations. Trio-based whole-exome sequencing (Trio-WES) identified bi-allelic compound mutations in the coding sequence (CDS) of POLR3A gene (c.3342C > T, p.Ser1114 = and c.3718G > A, p.Gly1240Ser). For the mild anemia phenotype, the underlying causal genetic factors could be attributed to the compound heterozygous mutations in FANCA gene (c.2832dup, p.Ala945CysfsTer6 and c.1902 T > G, p.Asp634Glu). Mini-gene reporter assays revealed that the synonymous variant of POLR3A and the missense variant of FANCA could affect pre-mRNA splicing of each gene. For POLR3A, the synonymous mutation (c.3342C > T, p.Ser1114=) generated three types of aberrant isoforms. Therefore, the female patient was finally diagnosed as WDRTS caused by POLR3A. For FANCA, the missense variant (c.1902 T > G, p.Asp634Glu) disrupted the normal splicing between exon 21 and 22, and produced two types of abnormal isoforms, one carrying the 1902G and the other spliced between exon 21 and 23 to exclude exon 22. Network analysis showed that POLR3A and FANCA could be STRINGed, indicating both proteins might collaborate for some unknown functions. Current investigation would broaden the knowledge for clinicians and genetic counselors and remind them to interpret those synonymous or predicted “benign” variants more carefully. Wiedemann-Rautenstrauch syndrome (WDRTS, OMIM#264090), an extremely rare progeroid disorder, was initially reported in two sisters by Rautenstrauch in 1977 (Rautenstrauch and Snigula, 1977). It was characterized by multiple distinct clinical features such as intrauterine growth retardation (IUGR), a progeroid appearance, lipodystrophy, failure to thrive, short stature, hypotonia, prominent scalp veins, teeth abnormalities and variable mental impairment (Toriello, 1990; Pivnick et al., 2000; Paolacci et al., 2017). In 1979, Wiedemann and Rautenstrauch considered this distinct neonatal progeroid syndrome to be transmitted under an autosomal recessive (AR) inheritance mode (Wiedemann, 1979). Till now, over 50 individuals with variable phenotypes of WDRTS have been reported (Paolacci et al., 2017). Homozygous or bi-allelic heterozygous mutations of RNA polymerase III subunit A (POLR3A, OMIM#614258) were proved to be the causal for WDRTS (Jay et al., 2016; Paolacci et al., 2018; Wambach et al., 2018; Temel et al., 2020). A few other cases with neonatal-onset progeria and lipodystrophy were identified to be caused by mutations in fibrillin 1 (FBN1, OMIM#134797; Graul-Neumann et al., 2010; Garg and Xing, 2014), caveolin 1 (CAV1,OMIM#601047; Garg et al., 2015; Schrauwen et al., 2015), catalytic subunit of DNA polymerase delta 1 (POLD1, OMIM#174761; Elouej et al., 2017; Sasaki et al., 2018) and solute carrier family 25 member 24 (SLC25A24, OMIM#608744; Ehmke et al., 2017; Rodríguez-García et al., 2018). Since neonatal-onset progeria and lipodystrophy were also core clinical phenotypes of WDRTS, it would pose a big challenge to discriminate WDRTS from other neonatal-onset progeria and lipodystrophy disorders in the early period, and to give appropriate and timely symptomatic treatments. Fanconi Anemia (FA, OMIM#227650) was a group of well-known clinically and genetically heterogeneous disorders (Bogliolo and Surralles, 2015), and characterized by distinct clinical features including developmental abnormalities in major organ systems, early-onset bone marrow failure, cellular sensitivity to DNA crosslinking agents, and a high predisposition to cancer (Nepal et al., 2017). The prevalence of FA was estimated at 1–5 in 1,000,000 live births (D’Andrea, 2010; Kottemann and Smogorzewska, 2013). It had been reported that FA could be caused by autosomal biallelic germline inactivation of any one of the 22 genes (FANCA-FANCW), except for the X-chromosomal FANCB gene (Niraj et al., 2019). Mutations in FANCA (OMIM#607139), FANCC (OMIM#613899) and FANCG (OMIM#602956) genes accounted for 60 ~ 65%, ~15% and ~10% of all the reported FA cases, respectively (D’Andrea and Grompe, 2003; Dimishkovska et al., 2018; Repczynska et al., 2022). In our clinic, a 3-years-old female patient was presumptively diagnosed as general growth retardation, neurocutaneous syndrome, left hip dysplasia and anemia. Later, she was diagnosed as WDRTS according to the clinical phenotypes and bi-allelic mutations in POLR3A gene detected by Trio-based whole exome sequencing (trio-WES). Besides, two heterozygous mutations were also detected in FANCA gene, resulted in a mild form of Fanconi Anemia. It is worth noting that one benign variant was identified in each gene and confirmed to affect proper pre-mRNA splicing to generate abnormal transcripts. To our knowledge, this was the first report for a WDRTS complicated with the occurrence of another recessive disorder, Fanconi Anemia (FA). It would broaden the molecular knowledge about WDRTS to clinicians and genetic counselors and reminded them to be more careful for analyzing genetic data and other relevant laboratory results. This study was conducted in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) for experiments involving humans. This study was approved by the Ethical Committee of the Shenzhen Bao’an Women’s and Children’s Hospital and Deyang People’s Hospital. Written informed consents were obtained from the female’s parent. Peripheral venous blood was collected from the proband and her parent. Genomic DNA was extracted using the TIANamp Blood DNA Kit (DP348, Tiangen Biotech, Beijing, China) according to the manufacturer’s instructions. Oligonucleotide Array-comparative genomic hybridization (array-CGH) was performed using the Fetal DNA Chip (Version 1.2) designed by the Chinese University of Hong Kong (CUHK) (Huang et al., 2014). The chip contains a total of 60,000 probes for more than 100 diseases caused by known microduplication/microdeletions. It does not include small fragment chromosomal abnormalities, copy number polymorphism, chimerism and chromosomal rearrangement (Iafrate et al., 2004). The experimental procedures were performed according to the standard Agilent protocol [Agilent Oligonucleotide Array-Based CGH (Array-CGH) for Genomic DNA Analysis, version 3.5]. Hybridized slides were scanned with SureScan High-Resolution Microarray Scanner (G2505B, Agilent Technologies), and the image data were extracted and converted to text files using Agilent Feature Extraction software. The data were graphed and analyzed using Agilent CGH Analytics software. To investigate the genetic cause of the disease, whole-exome sequencing (WES) was performed for the trio at MyGenostics. Briefly, the fragmented genomic DNAs were ligated with the 3′ end of the Illumina adapters and amplified by polymerase chain reaction (PCR). The amplified DNA was captured with Gencap Human whole Exon Kit (52 M) at MyGenostics. The capture procedure was performed in accordance with the manufacturer’s protocol. Finally, the generated libraries were sequenced on Illumina HiSeq 2,500 platform for paired-end sequencing. The sequencing depth was about 100x for each sample. The analysis of the WES data was carried out according to our previous reports (Lu et al., 2021, 2022). Briefly, clean reads were obtained after removal of adaptors and low-quality reads (multiple Ns and shorter than 40 bp) by Cutadapt (version 1.16) from raw data in fastq format (Kechin et al., 2017). The trimmed clean reads were aligned to the human reference genome (UCSC hg19) using BWA software (version 0.7.10) (Li and Durbin, 2010). The obtained files would be converted to bam format by SAMtools (version 1.2) (Li et al., 2009) and then filtered by BamTools (version 2.4.0) (Barnett et al., 2011). GATK (Genome Analysis Toolkit, version 4.0.8.1) was used to remove duplicated reads (by GATK/MarkDuplicates.jar), to recalibrate bases (by GATK/BaseRecalibrator.jar), and to obtain new bam files (by GATK/ApplyBQSR.jar) for subsequent variant calling by HaplotypeCaller (Auwera et al., 2013). Functional annotation for the GATK-called variants was performed by ANNOVAR (version 2018-04-16; Wang et al., 2010). Variants with a minor allele frequency (MAF) > 1% in the 1,000 Genome Project, or in-house data were removed. Synonymous single nucleotide variants (SNVs) were also removed. SNVs that caused splicing, frameshift, stopgain, or stoploss were retained for subsequent analysis. A position was called as heterozygous if 25% or more of the reads identified the minor allele. The location, type, conservation of the identified variants was obtained from several public databases, such as UCSC Genome Browser, NCBI dbSNP, NCBI ClinVar, 1000Genome, ExAC, TOPMED, gnomAD and gnomAD_exomes. Nonsynonymous SNVs were submitted to PolyPhen-2 (Polymorphism Phenotyping v2; Adzhubei et al., 2013) and PROVEAN (Protein Variation Effect Analyzer; Choi and Chan, 2015) for functional prediction. The pathogenicity of identified variants were also annotated according to the guidelines of American College of Medical Genetics (ACMG) (Riggs et al., 2020). The selected variants were confirmed by Sanger sequencing with an ABI3730xl sequencer (Applied Biosystems, Waltham, Massachusetts, United States). The possibility of identified variant for aberrant splicing was analyzed by SpliceAI (version 1.3.1) under default settings (Jaganathan et al., 2019). The protein sequences of POLR3A and FANCA were downloaded from NCBI GenBank, including 3 primates (Homo sapiens, Pan troglodytes, and Macaca mulatta), 1 cattle (Bos taurus), 2 rodents (Mus musculus, and Rattus norvegicus), 1 Chiroptera (Artibeus jamaicensis), 1 bird (Gallus gallus), 2 amphibians (Bufo bufo and Xenopus tropicalis), 2 fishes (Danio rerio and Nothobranchius furzeri). The protein sequences were aligned by the built-in ClustalW alignment algorithms of MEGA 11 (Gap opening penalty and Gap extension penalty for pairwise alignment and multiple alignment were set as 10.00, 0.10 and 10.00, 0.20, respectively; the Delay divergent cutoff was 30%). The effects of missense mutations on the structural changes were analyzed by the Missense3D and visualized using 3D View. The gene expressions were evaluated according to the normalized signal intensity of probe 227872_at for POLR3A and 236976_at for FANCA, which were extracted from a gene atlas of human protein-encoding transcriptomes for 79 human tissues (NCBI GEO#GSE1133; Su et al., 2004). The protein interaction network with POLR3A and FANCA was generated by STRING (version 11.5) under default settings. Gene Ontology (GO) analysis was performed for the 10 members of the network under default parameters in the Gene Ontology knowledgebase. The genomic regions containing the two mutations (c.3342C > T for POLR3A and c.1902 T > G for FANCA) were synthesized and cloned into the multiple cloning site (MCS) of pEGFP-N1 plasmid for minigene splicing reporter assays to test their effects on pre-mRNA splicing. As for c.3342C > T of POLR3A, the 1,646 bp genomic DNA spanning exon 25 to exon 27 (10:79,742,411-79,744,056, hg19) was cloned into the MCS of pEGFP-N1. For c.1902 T > G of FNACA, the 4,135 bp genomic DNA from exon 21 to exon 23 (16:89,838,089-89,842,223, hg19) was cloned into the MCS of pEGFP-N1. The two mutations were introduced by site-directed mutagenesis. The human embryonic kidney 293 cells (HEK293) or HeLa cells were cultured in high glucose DMEM medium (FI101-01, TransGen, Beijing, China) supplied with 5% fetal bovine sera (FBS) in 5% CO2. The constructs were transfected into HEK293 or HeLa cells by TransIntro EL/PL Transfection Reagent (FT231-02, TransGen) according to the manufacturer’s protocol. 24 h after transfection, cells were harvested, and lysed by adding 5 ml TransZol (ET101-01, TransGen). The wild-type (WT) and mutated (Mut) constructs were transfected into cells, respectively. Total RNAs were extracted and reversely transcribed into complementary DNAs (cDNAs) by TransScript Reverse Transcriptase (AT101-02, TransGen). The cDNAs were amplified by polymerase chain reaction (PCR) with paired primers (Supplementary Table S1), electrophoresed with agarose gel (1.5%, 120 V for 25 min), and then visualized by ChemiDoc XRS+ Gel Imaging System (Bio-Rad, Hercules, California, United States). DNAs of the bands were extracted and sequenced with an ABI3730xl sequencer (Applied Biosystems, United States). The single cell gel electrophoresis (SCGE) assay was performed as previously described with minor modifications (Li et al., 2014; Ji et al., 2018). After separated from 0.5 ml peripheral blood, lymphocytes were washed and resuspended at a density of 105 cells/mL in phosphate-buffered saline (PBS). 30 ml lymphocyte suspension were added in 70 μl of 0.75% low-melting-point agarose. The cell/agarose mixture was added onto the CometSlides which were precoated with 300 μl normal-melting-point agarose (0.75%) and was covered by a coverslip. After solidification, the coverslips were removed from the CometSlides. The CometSlides were submersed in cold fresh alkaline lysis solution for 1.5 h at 4°C. After lysis, the slides were electrophoresed at 30 V for 20 min in a horizontal tank which was filled with cold TBE buffer. Then, the slides were submerged in neutralization buffer for 20 min and stained with ethidium bromide (EB) in darkroom. The comets were observed using a digital fluorescence microscope (ECLIPSE 90i, Nikon, Tokyo, Japan), and images of 200 comets collected for each sample. The comets were analyzed by CASP (Comet Assay Software Project) software. The percentages of DNA in the comet head (HeadDNA%), DNA in the comet tail (TailDNA%), tail length (pix), tail moment (TM) and Olive tail moment (OTM) were calculated to evaluate the DNA damage of lymphocytes. The statistical analysis was conducted using the SPSS software (version 13) with Student’s t test for the mitomycin C-induced chromosome stress assay, and SCGE assay. p value less than 0.05 was considered as significance. The female proband (46, XX) was born naturally to a non-consanguineous couple in 2019. She has one unaffected healthy elder sister (Figure 1A). Her gestational period was 40+3 weeks. Her birth weight was 2.59 kg (P3). Her head occipitofrontal circumference (OFC) and body length at birth were 33 cm (P11) and 48 cm (P14), respectively. At the age of three, her weight, height and OFC were 11.5 kg (P6), 89.0 cm (P4), and 49.6 cm (P76), respectively (Figure 1D). Her mother accepted all regular inspections as required during her pregnancy. No abnormalities were found except for intrauterine growth retardation (IUGR) at 36 weeks of gestation. Her mother had no history of smoking or exposure to harmful hazards during pregnancy. She was breast fed in the first 6 months after birth. Mild feeding difficulty and sucking weakness were observed during that period. After 6 months of age, she gradually established a normal daily diet, but had persistent poor postnatal growth. The patient had presented progeroid appearance, with sparse scalp hair, poorly developed teeth, and thin subcutaneous fat (Figures 1B,C). Facial dysmorphic features were observed, such as triangular face, prominent forehead with frontal bossing, prominent scalp veins, sparse and broad eyebrows, deep set and long spaced eyes, pinched nose, small mouth with downturned corners, high-arched palate, malformed and low-set ears, and pointed chin. At birth, the patient had two natal teeth in the upper jaw and a gingival cyst of mucous gland in the lower jaw. At 6 months of age, the two natal teeth were removed by a dentist. No new teeth had grown at the same positions up to date. She was found mild neurodevelopmental delay and hypermyotonia at 3 months of age and received rehabilitation which lasted for 5 months till she could creep and sit without support. Neuropsychological development assessment was performed using the Children Neuropsychological and Behavioral Scale-Revision 2016 (CNBS-R2016) and the parent-rated Adaptive Behavior Assessment System II (ABAS-II) infant version at 3 years old. Her full-scale developmental quotient (DQ) of CNBS-R2016 was 115. The DQ in the five subscales involving gross motor, fine motor, adaptive behavior, language, personal-social of CNBS-R2016 were 120, 112, 112,120, and 112, respectively. The overall adaptive function score of ABAS-II was 106 (95% CI: 102–110, P66). The scores of social skills, conceptual skills, and practical skills in the three composite areas of adaptive function were 108 (95% CI 101–115, P70), 102 (95% CI 94–110, P55), and 106 (95% CI 99–113, P66), respectively. According to the neuropsychological development assessments, her intellectual development level was similar to that of children of the same age. Her brain magnetic resonance imaging (MRI) scan at 4 months old revealed no parenchymal abnormality except for a left arachnoid cyst (20.1 mm × 11.5 mm × 10.1 mm). Electroencephalogram (EEG) at the same time showed partial spikes at left occipital-parietal and anterior temporal region during sleep. Possible hereditary metabolic diseases were screened from blood and urine samples by liquid chromatography tandem mass spectrometry (LC–MS/MS), and no abnormalities were found. No structural abnormalities of urinary, cardiac and digestive systems were found by color Doppler ultrasound when she was 4 months old. Ultrasonic diagnosis revealed dysplasia of her left hip joint at 1 month after birth, but returned to normal at 6 months of age. Routine blood tests were performed seven times from 4 months to 3 years and 3 months after birth (Figure 2; Supplementary Table S2). No abnormality was found for the leukocyte and thrombocyte. Persistent mild anemia (HGB 88 ~ 98 g/L) with reduced mean corpuscular volume (MCV) and mean corpuscular hemoglobin (MCH) were observed, while the reticulocyte count was normal (36.3 × 109/L, 0.9%). Peripheral blood smear showed normal morphology of leukocytes and platelets, and smaller volume of mature red blood cells with enlarged central light stained area. Bone marrow puncture was refused by her parent. Karyotype analysis of G-banding chromosomes on peripheral blood mononuclear cells (PBMCs) detected no evident chromosomal abnormalities. Array-CGH analysis revealed no clinically significant microduplications or microdeletions. Trio-based whole exome sequencing (Trio-WES) was performed, and 53 rare variants were identified in the proband (Supplementary Table S3). All of the variants were inherited either from her father (parentally) or from her mother (maternally). Since the parent displayed no symptoms, genes with de novo, bi-allelic heterozygous or homozygous mutations were selected for subsequent analysis. Only one gene, FANCA, was remained in the list. Since homozygous or compound heterozygous mutations of FANCA gene resulted in the recessive Fanconi anemia of complementation group A (FANCA), this gene might be the unlying molecular factor for the anemia phenotype of the proband. However, no other genes were identified to be responsible for other clinical presentations. A Phenotype Profile Search was carried out at Human Phenotype Ontology (HPO) using eight key clinical features of the patient, such as progeroid facial appearance (HP:0005328), prominent scalp veins (HP:0001043), natal tooth (HP:0000695), intrauterine growth retardation (HP:0001511), and hypotonia (HP:0001252), sparse eyebrow (HP:0045075), sparse scalp hair (HP:0002209) and minimal (/thin) subcutaneous fat (HP:0003717). All eight input phenotypes were covered only in WDRTS, which was caused by mutations of POLR3A gene (Supplementary Table S4; Figure 3). Only one rare mutation (c.3718G > A, p.Gly1240Ser) of POLR3A gene (NM_007055.4) was identified in the proband (Supplementary Table S3). Since WDRTS was an autosomal recessive (AR) disorder, the removed variants of POLR3A were revisited and one synonymous variant (c.3342C > T, p.Ser1114=) was retrieved (Table 1). For POLR3A gene, c.3718G > A (rs1003620056) in the exon 28 (28/31) was transmitted maternally (Figures 4A,B) and generated a missense mutation from Gly1240 to Ser1240 (NP_008986.2, p.Gly1240Ser) in the RNA_pol_Rpb1_5 domain of POLR3A (Figures 4C,D). The Gly1240 residual was highly conserved in different vertebrate species. This mutation was extremely rare in TOPMED (n = 158,470, MAF = 0.000016), ExAC (n = 60,706, MAF = 0.000017), gnomAD (n = 76,156, MAF = 0.000007), and gnomAD Exomes (n = 125,748, MAF = 0.000028). Besides, this mutation had been reported in two patients (4H-42 and 4H-67) with 4H leukodystrophy (Wolf et al., 2014). Functional predictions by Polyphen-2 and PROVEAN showed this variant to be “Damaging” (score = 0.995) or “Deleterious” (score = −5.480) to the proper function of POLR3A, respectively. According to the ACMG guidelines, this mutation was classified as “Uncertain Significance” (PM2 + PP5). The structural changes introduced by Gly1240Ser were analyzed by Missense3D according to the cryo-EM structure of human POLR3A protein (7d58, chain A, 2.9 Å resolution). It’s revealed that Gly1240 was originally buried in a bend curvature (RAS 0.0%), which could be disrupted by the substitution with the Ser1240 residue (RSA 1.5%). The Serine could form new hydrogen bonds with Arg1104, Thr1238, and Tyr1097 (Figure 4E), which changed the surfaces of the local structure (Figure 4F). The splicing potential of this mutation on the pre-mRNA of POLR3A was evaluated by SpliceAI and obtained negative index (score = 0.00; Supplementary Table S5). The retrieved variant, c.3342C > T (rs183347762) in exon 26 (26/31) of POLR3A gene, was inherited paternally. It was synonymous without changing the amino acid Serine at position 1,114 (p.Ser1114=) in the RNA_pol_Rpb1_5 domain of POLR3A protein, and conserved in different species (Figures 4C,D). This variant was extremely rare in the international projects for large human cohorts, such as 1,000 Genomes (n = 2,504, MAF = 0.000200), TOPMED (n = 158,470, MAF = 0.000038), ExAC (n = 60,706, MAF = 0.000036), gnomAD (n = 76,156, MAF = 0.000043), and gnomAD_exomes (n = 125,748, MAF = 0.000036). It was assessed to be “Benign” (score = 0.0000) or “Neutral” (score = 0.0000) by Polyphen-2 and PROVEAN, respectively. According to ACMG guidelines, this mutation was annotated as “Likely Benign” (PM2 + BP4 + BP6 + BP7). Although as a synonymous mutation, c.3342C > T was only six nucleotides away from the canonical splicing acceptor site (c.3337-1G, 10:79743771) of the intron 25 (IVS25). SpliceAI displayed negative result (score = 0.00, Supplementary Table S5). The prediction by SPIDEX indicated that this mutation might affect the proper pre-mRNA splicing of POLR3A to generate abnormal transcripts. According to the analysis by ESEFinder (version 3.0), the mutated allele (3,342 T) might generate a novel binding site (CTGAGTAT) for serine and arginine rich splicing factor 1 (SRSF1), which might affect the splicing pattern or efficiency of POLR3A. For the anemia phenotype, two rare mutations in the CDS of FANCA gene (NM_000135.4) were detected by trio-WES (Table 1). A single-nucleotide insertion (c.2832dup, 16:89828377) in the exon 29 (29/43) was identified to cause a frameshift to the FANCA protein (p.Ala945CysfsTer6, NP_000126.2; Figures 5A,B). This mutation had been reported in an 8-year-old female patient (Li et al., 2018). However, this mutation had not been identified in any of the four public human genome projects, 1000Genomes, TOPMED, ExAC, gnomAD and gnomAD_exomes databases. It was poorly conserved in different animal species (Figure 5C). The mutant transcript might be the target of nonsense-mediated mRNA decay (NMD) or encode a putative shortened protein lacking the transmembrane (TM) and C-terminal Fanconi_A domain (Figure 5D). According to ACMG guidelines, this mutation was annotated as “Pathogenic” (PVS1 + PM2 + PP5). Another mutation in FANCA, c.1902 T > G (rs187300458) in exon 22 (22/43), was inherited maternally and changed the amino acid Aspartic acid (Asp., GAT) at position 634 to Glutamic acid (Glu, GAG; p.Asp634Glu; Figures 5A,B,D). This mutation was very rare in 1,000 Genome (n = 2,504, MAF = 0.000200), TOPMED (n = 158,470, MAF = 0.000008), gnomAD (n = 76,156, MAF = 0.000007), and GnomAD_exomes (n = 125,748, MAF = 0.000006). The conservation of this amino acid was very poor in different species (Figure 5C). This variant was predicted to be “Benign” (score = 0.255) or “Neutral” (score = −0.42) to the normal function of FANCA by Polyphen-2 or PROVEAN, respectively. According to ACMG guidelines, this mutation was annotated as “Likely Benign” (PM2 + BP3 + BP4). The structural changes introduced by Asp634Glu were analyzed by Missense3D according to the cryo-EM structure of human FANCA (7kzp, chain A, 3.1 Å resolution) and no structural damage was detected (Figures 5E,F). The c.1902 T > G was only two nucleotides away from the splicing acceptor site (c.1901-1G, 16:89839793, hg19) in intron 21 (IVS21) and might generate a novel putative splicing acceptor site (AT→AG). Negative impact (score = 0.03) of this missense on splicing of FANCA was identified by spliceAI (Supplementary Table S5). According to the prediction by ESEFinder (version 3.0), the mutant allele (1902G) might introduce novel binding sites for SRSF1 (CAGAGGC) and SRSF5 (ACAGAGG), and destroy a binding site for SRSF6 (TGCAGC). Although annotated to be benign, this mutation might affect the splicing pattern or efficiency of FANCA gene. Minigene reporter assay was carried out to verify the effects of c.3342C > T (POLR3A) and c.1902 T > G (FANCA) on pre-mRNA splicing. The genomic DNAs containing the selected mutations were cloned into the MCS of pEGFP-N1 and the mutations introduced by site-directed mutagenesis. As for c.3342C > T of POLR3A, the 1,646 bp genomic DNA spanning exon 25 to exon 27 (10:79,742,411-79,744,056) was cloned into the MCS (Figures 6A,B). Agarose electrophoresis revealed four different bands in the Mut samples, 435, 300, 243 bp and a short band (150 bp; Figure 6C). Sanger sequencing were carried out for the PCR products of these four bands and the band 300 bp was a non-specific product (pointed by the red arrow). The intensity of band 243 bp in Mut sample was about 70.60% of that in wild-type sample. The other two bands accounted for about 30% (Figure 6D). The band 243 bp was spliced with 3 exons (E25-E26-E27) and contained two types of isoforms, one wild type (Figure 6E) and one with 3,342 T (Figure 6F). The band 150 bp was produced by splicing between exon 25 and 27 to exclude the exon 26 (Figure 6G). In addition to the three consecutive exons, the band 435 bp retained the whole intron 25 (Figure 6H). After aligned all the Sanger-sequenced bands against POLR3A reference, three types of aberrant isoforms were revealed (Figure 6I). For c.1902 T > G of FANCA gene, the 4,135 bp genomic DNA from exon 21 to exon 23 (16:89,838,089-89,842,223) was cloned into the MCS of pEGFP-N1 (Figures 7A,B). Agarose electrophoresis revealed a novel short band (175 bp), in addition to the long band (289 bp) in the Mut sample (Figure 7C). However, the staining of the short band was rather weak. Sanger sequencing were carried out for PCR products of the two bands. It is verified that the long band was produced by the consecutive splicing of three exons (E21-E22-E23; Figures 7D,E). The long band in the Mut sample contained the 1902G allele. The short band was spliced between exon 21 and exon 23, excluding exon 22 (Figures 7F,G). Alimentary anemia due to deficiencies of iron, vitamin B12, vitamin D and folic acid was not considered according to relevant biochemical tests (Supplementary Table S6). Regular supplementation of iron for 3 months was ineffective. Trio-WES found no mutations for thalassemia-related genes such as HBA1 and HBA2 for α-thalassemia and HBB for β-thalassemia (Supplementary Table S1). It had been reported that thalassemia could also be caused by long-fragment deletions, recombinations and mutations in locus control regions (LCRs) involving α- or β-globin genes, which could not be detected by WES. Therefore, a third-generation single molecule real-time (SMRT) sequencing for long-molecules containing thalassemia-related genes were carried out and no mutations were identified (Supplementary Table S7). The mitomycin C (MMC)-induced chromosome stress assay was carried out for the peripheral blood samples from the patient and her mother, which was refused firmly by her father. After treated with different concentrations of MMC, 100 cells per sample were checked for chromosomal aberrations. However, no significant differences were observed between the two groups (Supplementary Figure S1). Genomic DNA damages were measured through cell-based alkaline comet assay, which was performed by the single cell gel electrophoresis (SCGE). As shown in Figure 8, after exposure to alkaline lysis solutions, the control lymphocytes (mother) failed to show any comet-like fashion (Figures 8A–C). About 17% of the patient’s lymphocytes showed the appearance of an obscure “halo” around the nucleus (Figures 8D–F), but no apoptotic cells were identified. The comet tail length of the patient sample was longer than that of the control (Supplementary Table S8). TailDNA%, TM, and OTM of the patient were much higher than those of her mother (p < 0.001), indicating that the level of DNA damage in the patient who carried FANCA mutations was higher than those in the control. RNA polymerase III was essential for the homologous recombination-dependent repair of DNA double-strand breaks (DSBs) (Liu et al., 2021) and FANCA was involved in inter-strand cross-link repair (Knipscheer et al., 2009). These two genes might act synergistically in this patient. According to the gene expression data of 79 human tissues, POLR3A and FANCA were co-expressed in many different tissues (Figure 9A). Ten proteins involving POLR3A and FANCA could form a stringed network (PPI enrichment value of p = 5.92E-10). The network showed that POLR3A could interact directly with POLR3B, POLR1A, POLR2F, and POLR2L to form a multi-subunit RNA polymerase complex possessing the DNA-directed 5′-3′ RNA polymerase activity (FDR = 4.63E-08; Figure 9B). FANCA could bind directly with BRCA1, which was an important component of the BRCA1-A complex (BRCA1, BARD1, BABAM1, and BRE; FDR = 1.92E-08). Interestingly, through the nodes of BRCA1 and POLR2F, FANCA could be stringed with POLR3A. Although all of the 10 proteins were involved in the nucleic acid metabolic process (FDR = 3.84E-06; Figure 9C), the synergistic function of FANCA on RNA polymerization III or vise verse was remained for further exploration. The gene POLR3A is located on chromosome 10q22.3, with 31 exons to encode a protein of 1,391 amino acid having a molecular mass of 154.7 kilodaltons. POLR3A is the largest catalytic subunit of the DNA-directed RNA polymerase III complex, which transcribes genes responsible for many small non-coding RNAs (ncRNAs), such as ribosomal 5S RNA, tRNAs, U6 small nuclear RNA, RNA components of mitochondrial RNA processing endoribonuclease (RMRP), ribonuclease P RNA component H1 (RPPH1), Ro60-associated RNA Y1 (RNY1), RNA component of signal recognition particle 7SL1 (RN7SL1) and RNA component of 7SK nuclear ribonucleoprotein (RN7SK). Some of these ncRNAs, such as RN7SL1 and RN7SK, regulate the activity of DNA-dependent RNA polymerase II, hence POLR3A mutations can also affect expression levels of polymerase II-transcribed genes (Azmanov et al., 2016; Flynn et al., 2016; Egloff et al., 2018). POLR3A also acts as a sensor to detect foreign viral DNAs and triggers an innate immune response (Ablasser et al., 2009). Recently, it has been reported that RNA polymerase III is an essential factor in the homologous recombination-dependent repair of DNA double-strand breaks (DSBs; Liu et al., 2021; Liu and Kong, 2021). Inhibition of POLR3A (also called Rpc1) could lead to the loss of genes in the DSB regions (Liu et al., 2021). Since POLR3A is ubiquitously expressed, the disability of this gene might be fatal to the prenatal and postnatal development of many systems. It had been reported that pathologic homozygous or bi-allelic heterozygous mutations in POLR3A could cause the occurrence of Wiedemann-Rautenstrauch syndrome (WDRTS; Paolacci et al., 2018) or Hypomyelinating leukodystrophy 7 (HLD7, OMIM# 607694) (Bernard et al., 2011) under an autosomal recessive (AR) mode of inheritance. WDRTS was one of the rare disorders having neonatal progeroid phenotype. The others included fontaine progeroid syndrome (FPS, OMIM#612289) (Writzl et al., 2017), autosomal recessive cutis laxa type IIIA (ARCL3A, OMIM#219150), and some forms of Marfan syndrome (MFS; Graul-Neumann et al., 2010; Takenouchi et al., 2013; Garg and Xing, 2014; Jacquinet et al., 2014). These syndromes had some characteristics similar to WDRTS. Besides, the clinical phenotypes of WDRTS were highly variable involving many systems. In addition, a variant of WDRTS was reported to have some atypical WDRST features (such as no lipodystrophy, no natal teeth and no sparse scalp hair), which was caused by a homozygous mutation in POLR3GL (c.358C > T, p.Arg120Ter) (Beauregard-Lacroix et al., 2020). These factors together made it difficult to accurately discriminate the WDRTS from other disorders having similar phenotypes. After checking clinical presentations of the above-mentioned syndromes, patients affected with WDRTS had neonatal tooth or teeth abnormalities, those with other neonatal progeroid phenotypes did not. It seemed that neonatal tooth might be an essential marker to discriminate WDRTS from other disorders having progeroid facial features. For our patient, she carried bi-allelic mutations in the CDS of POLR3A (c.3342C > T, p.Ser1114 = and c.3718G > A, p.Gly1240Ser). Except for c.3718G > A (p.Gly1240Ser) which could affect the structural conformation of POLR3A protein, the synonymous variant (c.3342C > T, p.Ser1114=) could lead to three types of abnormally spliced isoforms. The isoform 243 bp was consecutively spliced with three exons and carried the mutant allele. The isoform 435 bp was generated by the retention of intron 25, plus the three consecutive exons. After analyzed by Open Reading Frame Finder (ORF Finder), there was a premature stop codon in the intron 25 and might be translated into an aberrant protein (p.Glu1112GlufsTer7) or degraded by the nonsense-mediated mRNA decay (NMD; Lykke-Andersen and Jensen, 2015; Karousis and Muhlemann, 2019). As for the short isoform 150 bp, it only contained two exons (exon 25 and 27). Since the length of exon 26 was 93 base pairs (a multiple of three), the CDS of POLR3A should be left intact but missing 31 amino acids (aa1113-1,143) in the RNA_pol_Rpb1_5 domain (aa841-1,315). There were 4 missense mutations in the excluded exon 26 which were recruited in the NCBI ClinVar database, c.3350 T > C (p.Ile1117Thr), c.3388G > A (p.Val1130Ile), c.3392A > G (p.Lys1131Arg) and c.3407G > A (p.Arg1136Gln). These mutations were identified in patients with WDRTS or HLD7. In addition, c.3392A > G (p.Lys1131Arg) had been reported in a Caucasian WDRTS patient by targeted parallel sequencing (Paolacci et al., 2018). This indicated that the excluded region might be important for the function of POLR3A. Since also having core clinical phenotypes of WDRTS (Paolacci et al., 2017), the female proband was finally diagnosed as WDRTS caused by bi-allelic mutations in POLR3A. According to the mini-gene reporter assay, there were about 70% full-length wild-type and synonymous-containing transcripts. This indicated that a pathogenic hierarchy might be related to the two mutations. The missense mutation, c.3718G > A (p.Gly1240Ser), was the major contributor to the clinical presentations of our patient, with c.3342C > T (p.Ser1114=) as the minor one. Except for missense, nonsense, frameshifting, and mutations disrupting canonical splicing sites, there were 10 intronic mutations to affect pre-mRNA splicing of POLR3A (Hiraide et al., 2020), such as c.645 + 312C > T (Hiraide et al., 2020), c.1048 + 5G > T (Minnerop et al., 2017), c.1770 + 5G > C (Yan et al., 2021), c.1771-6C > G (Rydning et al., 2019; Wu et al., 2019), c.1771-7C > G (Minnerop et al., 2017), c.1909 + 22G > A (Minnerop et al., 2017; Morales-Rosado et al., 2020), c.1909 + 18G > A (Lessel et al., 2018), c.2003 + 18G > A (Bernard et al., 2011), c.3337-5 T > A (Lessel et al., 2018; Wambach et al., 2018), c.3337-11 T > C (Wambach et al., 2018). Among them, c.1909 + 22G > A was the most commonly reported mutations. However, synonymous variants affecting the pre-mRNA splicing of POLR3A were rarely reported. Till now, only one homozygous synonymous (c.3336G > A, p.Glu1112=) has been reported to generate two types of abnormal splicing isoforms, one with the retention of intron 25, and another with the exclusion of exon 25 (Lessel et al., 2022). Our patient was the second report of a synonymous variant to affect the pre-mRNA splicing of POLR3A. In order to have a comprehensive view of the phenotypes of WDRTS and HLD7, a literature review was made. Features of craniofacial dysmorphism and soft tissues were exclusively confined to WDRTS (Figures 10A,B). However, the majority of the abnormal phenotypes in central nervous system were mainly found in patients suffering from HLD7 (Figure 10B). The reported mutations of POLR3A in WDRTS and HLD7 were compiled and arranged according to their genomic position. To our surprise, there were a few different “hot spots” between WDRTS and HLD7. Mutations in intron 13, exon 19 and exon 28 were almost exclusively related to HLD7. For WDRTS, most of the mutations were distributed in exon 1, exon 6, intron 25 and 3’-UTR, with intron 25 as the highest (Figure 10C). For our patient, the splicing-altering mutation c.3342C > T was located at the junction between exon 25 and intron 25. It seemed that the aberrant pre-mRNA splicing at intron 25 might be correlated with the occurrence of WDRTS. Efforts to establish animal models with Polr3a mutation had been tried in mice, but had not been successful. It had been reported that double knockout (KO) of polr3a in mice was embryonically lethal (Choquet et al., 2019). Furthermore, no neurological or developmental abnormalities were identified in mice with whole-body homozygous knock-in (KI/KI) or heterozygous KI/KO of the pathogenic c.2015G > A (p.Gly672Glu) mutation of polr3a (Choquet et al., 2017). For RNA polymerase III (POLR3) in animals, it composed of 17 subunits to form a catalytic core, the stalk domain and Pol III-specific subcomplexes (Vannini and Cramer, 2012; Girbig et al., 2021). Till now, only six of 17 subunits (35.29%) were reported to be the causal for a spectrum of rarely inherited disorders. Mutations in POLR3A were responsible for WDRTS or HLD7 (Bernard et al., 2011; Wambach et al., 2018), POLR3B for HLD8 (OMIM#614381; Saitsu et al., 2011), POLR1C for HLD11 (OMIM#616494) (Thiffault et al., 2015), POLR3K for HLD21 (OMIM#619310) (Dorboz et al., 2018), POLR3GL for short stature, oligodontia, dysmorphic facies, and motor delay (SOFM, OMIM#619234; Terhal et al., 2020) and POLR3H for primary ovarian insufficiency (POI; Franca et al., 2019). Since only 35.29% of the members of POLR3 could be related to inheritable disorders, there might be a functional redundancy among other subunits. Inferred from the time-coursed routine blood testing, the patient had a moderate level of anemia. The anemia belonged to small cell hypochromic anemia, similar to iron deficiency anemia or thalassemia. However, the concentrations of serum ferritin, vitamin B12, folic acid and vitamin D were within a normal range, indicating that the anemia might be caused by other unknown reasons. Through trio-WES, two mutations were identified in the CDS of FANCA, a causal gene for Fanconi anemia of complementation group A (FANCA, OMIM#227650). No mutations were identified in genes responsible for thalassemia by trio-WES and the third-generation SMRT sequencing. For FANCA, the pathogenic insertion (c.2832dup) in exon 29 introduced a premature termination codon (PTC), which caused a frameshift of the FANCA protein (p.Ala945CysfsTer6) or rendered the resultant transcripts to be rapidly degraded by NMD. Another missense mutation (c.1902 T > G, p.Asp634Glu) was predicted to be benign. After carefully analyzing the genomic sequence containing c.1902 T > G, it was only two nucleotides away from the canonical splicing acceptor site (SA1, c.1901-1_1901–2, AG) in intron 21. The mutation might introduce a potential splicing acceptor site (SA2, c.1902_1903, AG) juxtaposed with SA1. Minigene reporter assay identified two types of aberrant isoforms. One carried the 1902G and translated into a full-length FANCA protein with Glu634. The other isoform was produced by splicing between exon 21 and 23 to exclude exon 22, but at a very low level. Since the length of exon 22 was 114 base pairs (a multiple of three), the CDS of FANCA should be left intact but missing 38 amino acids (aa634-672). However, the function of this region was not clear. There were seven pathogenic mutations in this excluded region recruited in the ClinVar database, namely, c.1912G > T (p.Gly638Ter), c.1944del (p.Glu648AspfsTer13; Levran et al., 1997), c.1951G > T (p.Gly651Ter), c.1979 T > C (p.Leu660Pro; Ameziane et al., 2008), c.1981A > T (p.Arg661Ter), c.2001dup (p.Ser668GlnfsTer4; Moghrabi et al., 2009), and c.2005C > T (p.Gln669Ter). This indicated that the excluded region was important for the function of FANCA protein. According to the mini-gene reporter assay, there were more than 80% full-length wild-type and missense-containing transcripts. A pathogenic hierarchy might be related to the two mutations. The frameshift mutation, c.2832dup (p.Ala945CysfsTer6), was the major contributor to the clinical presentations of our patient, with the missense c.1902 T > G (p.Asp634Glu) as the minor one. This indicated that there might be a delicate balance between wild-type and mutated transcripts to prevent the occurrence of macroscopic clinical phenotypes. In order to verify the phenotype of Fanconi’s anemia (FA), mitomycin C-induced chromosome stress (MMC) assay and single cell gel electrophoresis (SCGE) assay were performed for blood samples from the patient and her mother. MMC assay detected no significant chromosome aberrations. For SCGE assay, good-shaped comets were observed in the patient’s sample, indicating that the genomic DNA was seriously damaged. This implied that the patient might suffer from a mild type of FA without obvious classical phenotypes. To our knowledge, this was the first report of a patient with a rare Wiedemann-Rautenstrauch syndrome (WDRTS) complicated with another recessive disorder, Fanconi anemia of complementation group A (FANCA). It had been reported that both POLR3A and FANCA were involved in the homologous recombination-dependent repair of DNA double-strand breaks (DSBs) (Liu et al., 2021; Liu and Kong, 2021) and inter-strand DNA cross-link repair (Howard et al., 2015; Benitez et al., 2018) to maintain the chromosome stability. A network analysis showed that POLR3A could be STRINGed with FANCA via two nodes of BRCA1 and POLR2F (Krum et al., 2003; Lane, 2004). It is implied that both proteins might act synergistically to contribute to the complexity of clinical phenotypes. This should be verified by further cellular and model animal experiments. Generally, a WDRTS patient was identified to have rare bi-allelic compound mutations in POLR3A, one damaging missense and one synonymous. The synonymous mutation could affect the pre-mRNA splicing of POLR3A and should be pathogenic. It generated about 30% of aberrantly splicing transcripts. As for the anemia phenotype, the predicted benign missense mutation 1902 T > G could generate a small proportion of abnormally spliced isoform of FANCA. The expressed ratio between the aberrant and wild type isoforms might be correlated to the severity of the disease. Even patients carrying same splicing-altering mutations presented different phenotypes, other unidentified regulatory polymorphisms might be the modifying factors for the different penetration. Since the detrimental level of mutations vary greatly, different combinations of these mutations might be one of the underlying mechanisms for the varied clinical phenotype penetrance and prognosis. It might be very useful for clinical genetic consultors to have a comprehensive analysis for the relationship between genetic factors and clinical features. The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material. The studies involving human participants were reviewed and approved by Ethical Committee of the Shenzhen Baoan Women’s and Children’s Hospital. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. GL and QP designed the study. GL and YZ analyzed gnomic data. QP, BX, JD, ZX, and BD provided phenotype information. QP and BD performed followup inquiries. LW, WD, DL, and JW performed the minigene reporter assay, SCGE assay, and MMC assay. GL wrote the manuscript. GL, XH, and QP revised the manuscript. All authors contributed to the article and approved the submitted version. This work was supported by fundings from the Science and Technology Research and Development Foundation of Shenzhen (JCYJ20180305164359668), Natural Science Foundation of Sichuan (2022NSFSC0714), Key Research and Development Project of Deyang science and Technology Bureau (2021SZ003 and 2020SZZ085), and Special Fund for Incubation Projects of Deyang People’s Hospital (FHG202004). The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649811

Ying Zhao,Wusheng Zhu,Ting Wan,Xiaohao Zhang,Yunzi Li,Zhenqian Huang,Pengfei Xu,Kangmo Huang,Ruidong Ye,Yi Xie,Xinfeng Liu

Vascular endothelium deploys caveolin-1 to regulate oligodendrogenesis after chronic cerebral ischemia in mice

10-11-2022

Myelin biology and repair,Hypoxic-ischaemic encephalopathy,Oligodendrocyte

Oligovascular coupling contributes to white matter vascular homeostasis. However, little is known about the effects of oligovascular interaction on oligodendrocyte precursor cell (OPC) changes in chronic cerebral ischemia. Here, using a mouse of bilateral carotid artery stenosis, we show a gradual accumulation of OPCs on vasculature with impaired oligodendrogenesis. Mechanistically, chronic ischemia induces a substantial loss of endothelial caveolin-1 (Cav-1), leading to vascular secretion of heat shock protein 90α (HSP90α). Endothelial-specific over-expression of Cav-1 or genetic knockdown of vascular HSP90α restores normal vascular-OPC interaction, promotes oligodendrogenesis and attenuates ischemic myelin damage. miR-3074(−1)−3p is identified as a direct inducer of Cav-1 reduction in mice and humans. Endothelial uptake of nanoparticle-antagomir improves myelin damage and cognitive deficits dependent on Cav-1. In summary, our findings demonstrate that vascular abnormality may compromise oligodendrogenesis and myelin regeneration through endothelial Cav-1, which may provide an intercellular mechanism in ischemic demyelination.

Vascular endothelium deploys caveolin-1 to regulate oligodendrogenesis after chronic cerebral ischemia in mice Oligovascular coupling contributes to white matter vascular homeostasis. However, little is known about the effects of oligovascular interaction on oligodendrocyte precursor cell (OPC) changes in chronic cerebral ischemia. Here, using a mouse of bilateral carotid artery stenosis, we show a gradual accumulation of OPCs on vasculature with impaired oligodendrogenesis. Mechanistically, chronic ischemia induces a substantial loss of endothelial caveolin-1 (Cav-1), leading to vascular secretion of heat shock protein 90α (HSP90α). Endothelial-specific over-expression of Cav-1 or genetic knockdown of vascular HSP90α restores normal vascular-OPC interaction, promotes oligodendrogenesis and attenuates ischemic myelin damage. miR-3074(−1)−3p is identified as a direct inducer of Cav-1 reduction in mice and humans. Endothelial uptake of nanoparticle-antagomir improves myelin damage and cognitive deficits dependent on Cav-1. In summary, our findings demonstrate that vascular abnormality may compromise oligodendrogenesis and myelin regeneration through endothelial Cav-1, which may provide an intercellular mechanism in ischemic demyelination. With lower blood flow from distal parts of long deep arteries, cerebral white matter is susceptible to ischemia. In the hypoxia/ischemia, damages, such as oxidative stress, inflammation, and excitotoxicity, to oligodendrocyte precursor cells (OPCs) can trigger rapid and profound demyelination. Adult resident OPCs constitute approximately 6% of the total cell number in the CNS, providing a source for myelin sheath renewal. However, OPC maturation is often arrested in hypoxia/ischemia. Our previous study, consistently, found robust OPC proliferation after chronic cerebral ischemia, but it was not durable and failed to restore myelination. Thus, attempts to promote oligodendrogenesis may provide clues to ameliorate demyelination therapeutically. The coordination of oligovascular coupling is important for white matter maintenance. Oligodendroglial cells can regulate CNS endothelial cell proliferation and angiogenesis. Vascular endothelial cells, in turn, are documented to secrete factors to favor OPC survival and promote oligovascular remodeling. In cerebral ischemic injury, dysfunctional OPC-vascular interaction was detected as OPC might mediate early endothelial barrier opening. To the opposite, OPC in both human and animal models of hypoxic-ischemic encephalopathy was reported to promote white matter vascularization. However, whether this aberrant interaction would induce OPC changes is not fully understood. As endothelial dysfunction was detected even before demyelination occurred, it then prompts us to consider the contributing effects of vascular damage to the pathogenesis of ischemic demyelination based on OPC-endothelial association. Caveolin-1 (Cav-1), a component of caveolae, is highly expressed on endothelial cells, orchestrating signal transduction and vesicular trafficking. Decreased level of endothelial Cav-1 was implicated in blood-brain barrier (BBB) breakdown and neuroinflammation after ischemic stroke. Clinical data revealed that, in patients with ischemic stroke, a reduced level of serum Cav-1 was associated with cerebral microbleeds and symptomatic bleeding, suggesting an important role of Cav-1 in vessel stability. A recent paper has indicated that arteriolar Cav-1 has an active role in mediating neurovascular coupling, including neural activity and vascular dynamics. As such, we aim to investigate whether endothelial Cav-1 could participate in oligovascular coupling in chronic cerebral ischemia. In this study, we investigated the potential role of endothelial Cav-1 linking vasculature with ischemic demyelination in the mouse bilateral carotid artery stenosis (BCAS) model and leukoaraiosis patients. Endothelial Cav-1 was remarkably reduced, which was responsible for the aberrant relationship between the OPCs and the vasculature, following hypoxic-ischemic injury. We identified both an upstream cause of endothelial Cav-1 reduction and a resultant secreted downstream effector affecting OPC differentiation. The treatment to stabilize oligovascular interactions restored myelination in chronic cerebral ischemia, suggesting a new therapeutic strategy. The time-course changes in cerebral blood flow (CBF) and myelin damage were analyzed in the CC at 1, 2 and 4 weeks after BCAS surgery. At 1-week post BCAS, CBF significantly declined to nearly 49% of the control group (Fig. S1a, b, P < 0.001). Despite a slow recovery thereafter, the CBF remained at a level reduced by 23% 4 weeks after BCAS (Fig. S1a, b, P = 0.0002). Luxol fast blue (LFB) and black-gold II staining both confirmed myelin loss since BCAS_2w (Fig. S1c–e). Consistently, the fluorescent intensity of myelin-related proteins, myelin-associated glycoprotein (MAG), and myelin basic protein (MBP), was not reduced until 2 weeks after the insult (Fig. S1f, g). We next probed whether endothelial dysfunction was an initial change before demyelination under chronic ischemia. By Transmission Electron Microscopy (TEM), as early as 1 week after BCAS, the morphology of tight junctions (TJs) was significantly disrupted, characterized by an increased gap area between endothelial cells (Fig. S2a, b). To assess vascular permeability, the 3-kDa dextran was utilized as a fluorescent tracer. In the CC of BCAS mice, the extravasated dyes were observed around the CD31+ vessels at 1 and 2 weeks after BCAS, which was minimized at 4 weeks post-BCAS (Fig. S2c, d). The vessel regression, quantified by the number of collagen IV+ basement membranes without CD31+ endothelial cells (empty sleeves), increased after BCAS with a peak at 1 week (Fig. S2e, f). Moreover, the number of vessels with degraded collagen IV gradually increased from 1 week to 4 weeks post-BCAS (Fig. S2e, g), suggesting vessel instability during hypoperfusion. We then asked whether early endothelial dysfunction was associated with OPC changes in BCAS mice. By immunostaining of PDGFRα, oligodendrocyte transcription factor 2 (Olig2), and CD31, we found accumulative OPCs interacted with blood vessels 2 weeks after BCAS. Both frequency and size of clusters were increased on the vasculature (Fig. 1a–c; compared to controls, both P < 0.001). This association was more frequent at 4 weeks post-BCAS, suggesting a gradual strengthened association of OPCs with blood vessels (Fig. 1a–c). Immunoelectron microscopy also showed the process of Olig2 immunogold+ oligodendroglial cells contacting the vessels (Fig. 1d). Breast carcinoma amplified sequence 1 (BCAS1) and ectonucleotide pyrophosphatase/phosphodiesterase 6 (ENPP6) have been recognized as markers of newly-formed immature oligodendrocytes, representing a status of ongoing myelination and remyelination. Interestingly, these markers were greatly increased at 2 weeks followed by a reduction at 4 weeks of chronic hypoperfusion (Fig. 1e–g; compared with BCAS_2w, P = 0.0017 and 0.0062, respectively). Mature oligodendrocytes, which were stained for CC1 and Olig2, were gradually decreased at both 2 and 4 weeks after BCAS (Fig. 1h, i). These data suggest that OPCs tended to differentiate to generate new oligodendrocytes. However, this compensation, possibly with defective maturation and increased oligodendrocyte death, were not capable of supporting remyelination at 4 weeks of hypoperfusion. In vitro, we utilized CoCl2 (10 μM, 5 days) and oxygen-glucose deprivation (OGD, 4 h) to induce hypoxia. We performed PCR for the expression of classic hypoxia-related genes in endothelial cells. CoCl2 and OGD treatment both increased the mRNA level of hypoxic markers, including HIF-1α, HIF-2α, PDK-1, BNIP3, and VEGF, representing significant hypoxia (Fig. 2a). After exposure to CoCl2 or OGD, the proliferation of brain microvascular endothelial cells (BMECs) was increased with enhanced apoptosis by the quantification of EdU-positive and TUNEL-positive cells (Fig. 2b, c). ZO-1 and Claudin-5 in the immunostaining were both fractured at the boundary of CoCl2- and OGD-treated BMECs compared to controls (Fig. 2d). A consistent decrease was observed in the protein levels of TJs (Fig. 2e, f). To address the effects of dysfunctional BMECs on normal OPCs, we conducted conditioned media (CM) transfer assay. CM from either CoCl2- or OGD-treated BMECs was collected and then incubated with OPCs. When cultured in the CoCl2-CM or OGD-CM, the proliferation of OPCs was preferred, as Ki67+PDGFRα+Olig2+ cells were significantly increased compared to those in control CM (Fig. 2g, h; P < 0.001 and P = 0.004, respectively). The generation of MBP+ oligodendrocytes with highly developed ramified processes was inhibited by CoCl2-CM and OGD-CM (Fig. 2i–k; both P < 0.001). Immunoblotting displayed similar changes in that the expression of PDGFRα was increased while mature oligodendrocytes markers, MBP and 2′, 3′-cyclic nucleotide 3′-phosphodiesterase (CNPase), were reduced to less than half of control protein levels (Fig. 2l, m). These observations verified that BMECs, who suffered from hypoxia, could release detrimental factors to suppress OPC maturation. Given that Cav-1 plays a vital role in endothelial function, we explored the expression of Cav-1 in BMECs exposed to CoCl2 or OGD. Immunostaining for Cav-1 displayed that the intensity of Cav-1, which normalized to CD31 intensity, was decreased compared to BMECs under normoxic status (Fig. 3a, b; both P < 0.001). Besides, immunoblotting revealed that CoCl2 and OGD exposure reduced Cav-1 protein to 43.2% and 27.8% of control cells respectively (Fig. 3c, d; P = 0.0288 and 0.0068, respectively). To probe the extracellular engagement, we utilized an antibody array covering more than 1300 antibodies related to several signaling pathways to measure the changes in soluble factors secreted by BMECs. Compared to control CM, secreted HSP90α was the most significantly expressed functional protein in CoCl2-CM (Fig. 3e), which was further confirmed by ELISA (Fig. 3f). In vivo, reduced endothelial Cav-1, accompanied by elevated extravascular HSP90α, was detected in the CC after BCAS (Fig. 3g–i). Immunoblotting of extracted microvascular segments of the CC consistently demonstrated that the vascular Cav-1 was reduced while HSP90α was increased (Fig. 3j–m). As for the correlation between Cav-1 and HSP90α, we found the Cav-1 antibody specifically coprecipitated HSP90α in the co-immunoprecipitation assay (Fig. S3a). HSP90α was also be detected in a reverse co-immunoprecipitation analysis (Fig. S3b). Therefore, after hypoxic treatment, reduced Cav-1 could also be co-localized with HSP90α in BMECs. The Cav-1/HSP90α co-immunostaining further revealed that, under normoxic status, Cav-1 was co-localized with about 45.6% of total HSP90α at the endothelial membrane (Fig. S3c). However, during chronic hypoxia, the co-localization was reduced to 5.5% (Fig. S3c, d; P = 0.0028). Of particular interest, we evaluated Cav-1 expression among different neural cells, including endothelial cells, pericytes, oligodendroglial lineage cells, and astrocytes. The results showed that about 84% CD31+ endothelial cells expressed Cav-1, whose level was at least 2.5 times higher than other neural cells (Fig. S4). It thus indicated that the function of Cav-1 may be endothelial cell-autonomous. To determine whether Cav-1 reduction would affect the caveolae level after BCAS, we used TEM to directly assess caveolar change. We found capillary and arteriolar caveolae number did not change between control and BCAS mice (Fig. S5a–c). Co-immunostaining of Cavin-1, a caveolae marker, with vascular markers, including MFSD2A for capillary and α-SMA for arteriole, consistently showed no differences in vascular Cavin-1 expression between control and BCAS mice (Fig. S5d–f). In in vitro experiments, the immunostaining and the quantification demonstrated that chronic hypoxia did not affect the co-localization of Cav-1 and Cavin-1 in the membrane (Fig. S5g, h), further suggesting that hypoxia might not affect caveolar Cav-1. To identify the origin of HSP90α secretion, we performed combined fluorescent in situ hybridization (FISH) and immunofluorescence. We used the double-FAM-labeled probes to detect Hsp90α mRNA and different cell marker antibodies to locate different cells, including CD31 for endothelial cells, PDGFRβ for pericytes, GFAP for astrocytes, and Olig2 for oligodendroglial lineage cells. The results showed that Hsp90α mRNA was mainly expressed in endothelial cells and significantly increased under chronic ischemic conditions (Fig. S6). Thus, the secreted HSP90α close to or associated with the vessel might be mostly derived from vascular endothelial cells. We employed a genetic approach to increase endothelial Cav-1 using an Adeno-associated virus (AAV) carrying the Cav-1 gene under the promoter of Tie1. The viral vectors were stereotaxically injected into the CC 3 weeks before BCAS surgery (Fig. 4a). The GFP signal was found within endothelial cells in the CC. AAV-Tie1-Cav-1 successfully elevated endothelial Cav-1 expression (Fig. 4b, c; t = 18.41, P < 0.001). This 1.98-fold upregulation of endothelial Cav-1 led to 46.7% downregulation of endothelial HSP90α and 48.9% downregulation of secreted HSP90α in the perivascular region (Fig. 4d–f; P = 0.0111 and 0.0008, respectively), indicating HSP90α might be a downstream secretory factor of Cav-1. Besides, in the CC receiving AAV-Tie1-Cav-1, OPC clustering around the microvessels was significantly attenuated with decreased cluster numbers and occupied area (Fig. 4g–i). BCAS1+ cells as well as CC1+Olig2+ mature oligodendrocytes were remarkably increased in AAV-Tie1-Cav-1-transfected mice (Fig. 4j–l). As expected, chronic ischemia-induced myelin proteins reduction was rescued in the Cav-1 overexpression group (Fig. 4m, n), suggesting that endothelial Cav-1 may be an important intermediary in ischemic demyelination. To examine the role of HSP90α in myelin damage and its relation with Cav-1, we intravenously injected negative control (N.C.) or HSP90α siRNA packed with polyetherimide (PEI) into wild-type and Cav-1−/− mice after BCAS surgery at intervals of three days for 4 weeks (Fig. 5a). The capture of Cy5-labeled N.C. or HSP90α siRNA by microvessels was identified in the CC (Fig. 5b). HSP90α siRNA-induced inhibition on secretory HSP90α was significantly higher than the suppression of endothelial HSP90α in BCAS mice of two genotypes (Fig. 5c). As expected, the interference of HSP90α did not affect the endothelial Cav-1 level (Fig. S7). Compared to mice injected with N.C. siRNA, both wild-type and Cav-1−/− mice that received HSP90α siRNA exhibited less frequent OPC clusters attached to the vessels with decreased size (Fig. 5d–f). More BCAS1- and ENPP6-positive pre-myelinating oligodendrocytes were newly formed in HSP90α siRNA-treated BCAS mice (Fig. 5g–i), together with more mature oligodendrocytes (Fig. 5j, k). To explore the upstream mediator of Cav-1, high-throughput sequencing of miRNAs was carried out using the CC tissue of control and BCAS_4w mice. Among the miRNA sequences obtained, we filtered 30 significantly up-regulated miRNAs with fold change >= 2 (Fig. 6a). After cross-compared to the predicted miRNAs in the TargetScan, 13 potential miRNAs were screened out (Fig. 6a, b). We found that miR-3074-1-3p was one of the most significantly upregulated genes at BCAS_4w (Fig. 6c). Validated by real-time PCR, miR-3074-1-3p reached 10.2-fold higher expression in mice subjected to BCAS (Fig. 6d). Most importantly, to define the cell-autonomous nature of miR-3074-1-3p during chronic ischemia, real-time PCR was employed in the BMECs under chronic hypoxia and miR-3074-1-3p had the highest increase (Fig. 6e). miRNA fluorescent in situ hybridization further displayed the elevation of miR-3074-1-3p in CoCl2-treated BMECs (Fig. 6f, g). To determine whether miR-3074-(1)-3p could specifically recognize the 3′ untranslated regions (UTR) of Cav-1, we conducted the luciferase reporter assay in 293 T cells. Mouse miR-3074-1-3p and human homolog miR-3074-3p share conserved predicted binding sites in the 3′ UTR of mouse and human Cav-1 (Fig. 6h). miR-3074-(1)-3p mimics, instead of N.C., inhibited the luciferase activities of wild-type Cav-1 3′ UTR. However, when introduced to the mutated Cav-1 3′ UTR, the mimics could not induce any interference in the luciferase activities (Fig. 6i), implying that Cav-1 may be a direct target of miR-3074-(1)-3p in both mice and human. To examine whether miR-3074-1-3p could regulate Cav-1 at the protein level in BMECs, we analyzed endothelial Cav-1 expression under the transfection with miR-3074-1-3p agomir and antagomir. As depicted in Fig. 6j, k, miR-3074-1-3p agomir substantially suppressed the level of Cav-1 while antagomir enhanced the Cav-1 expression in BMECs. We included a total of 154 patients (54.5% male; mean age, 67.0 ± 9.5 years) in the present study. Eighty-eight subjects (57.1%) had imaging manifestation with leukoaraiosis. The comparison of baseline data was demonstrated in table S1 based on patients with or without leukoaraiosis. Patients with leukoaraiosis were more likely to have a higher level of hsa-miR-3074-3p and a lower level of Cav-1 (Table S1, both P = 0.001). After adjusting for potential confounders, increased hsa-miR-3074-3p [odds ratio (OR), 1.41; 95% confidence interval (CI), 1.17–1.69; P = 0.001] and reduced Cav-1 level (OR, 0.29; 95% CI, 0.16–0.53; P = 0.001) were also significantly associated with a higher risk of leukoaraiosis (Fig. 7a). Furthermore, linear regression analysis indicated that hsa-miR-3074-3p level was negatively associated with Cav-1 (Fig. 7b, β = −0.261, P = 0.027). Among the 88 patients with leukoaraiosis, 37 patients (42.0%) were diagnosed with severe leukoaraiosis. Compared to patients with mild leukoaraiosis, patients with severe leukoaraiosis tended to have a higher level of hsa-miR-3074-3p (Fig. 7c, P = 0.1309) but a lower level of Cav-1 (Fig. 7d, P = 0.0159). To ask whether the regulation of miR-3074-1-3p may exert effects on BMECs and OPCs and whether Cav-1 was involved, we treated wild-type and Cav-1−/− BMECs with miR-3074-1-3p antagomir respectively. Eight hours before exposure to hypoxia, BMECs were transfected with N.C. or antagomir. The Cy5-labeled antagomir was localized around the nucleus of BMECs, verifying the successful uptake of antagomir (Fig. 8a). By both immunostaining and immunoblotting, Cav-1 was remarkably increased in wild-type BMECs transfected with antagomir (Fig. 8a–c). Antagomir alleviated the loss of endothelial TJs almost to the basal conditions in CoCl2-treated wild-type BMECs (Fig. 8b, c). This restoration was counteracted by the ablation of Cav-1 (Fig. 8b, c). In wild-type BMECs, co-immunostaining revealed that antagomir could increase the percent of co-localized HSP90α with Cav-1 in the hypoxic environment (Fig. 8d, e). Correspondingly, the secreted HSP90α was strikingly decreased in the culture medium (Fig. 8f), which thereby rescued the defective maturation of OPC (Fig. 8g–j). On the contrary, under the deletion of Cav-1, antagomir could not suppress HSP90α secretion in response to CoCl2, which subsequently limited OPC differentiation, as illustrated by immunostaining and immunoblotting for myelin-related protein levels (Fig. 8f–j). Thus, our data suggested endothelial miR-3074-1-3p inhibition could rescue the hypoxia-induced endothelial dysfunction and subsequent defective OPC differentiation via Cav-1. We next examined whether antagomir could exert protection in vivo. The time-course expression of miR-3074-1-3p showed that it started to increase at 2 weeks after BCAS (Fig. 9a). Thus, continuous delivery of in vivo-jetPEITM-formulated Cy5-antagomir was started at BCAS_2w and lasted for another 2 weeks through the osmotic pumps into the CC (Fig. 9b). By live imaging and immunostaining, antagomir was located within the microvessels (Fig. 9c, d). It increased endothelial Cav-1 level and inhibited HSP90α secretion (Fig. 9d–g; Fig. S8a, b). The TJs level and structure in wild-type BCAS mice were significantly recovered by the antagomir (Fig. 9g; Fig. S8c–f). Nonetheless, antagomir could not affect the HSP90α increment and TJs loss in Cav-1−/− BCAS mice (Fig. 9e–g; Fig. S8). After treatment with antagomir, a lower frequency of OPC cluster was detected in wild-type mice of BCAS. Conversely, regardless of receiving N.C. or antagomir, the Cav-1 null mice manifested a parallel number and size of perivascular OPC clusters after BCAS (Fig. 10a–f). In addition, newly-formed oligodendrocytes were found in antagomir-treated wild-type CC, while equally treated Cav-1−/− BCAS mice displayed defective remyelination as fewer BCAS1+ and ENPP6+ cells were present in the CC (Fig. 10g–i). By the analysis of immunostaining, TEM, and immunoblotting, the antagomir significantly restored the white matter integrity in wild-type BCAS mice, while it was unable to induce detective attenuation without Cav-1 (Fig. S9). Cognitive function was evaluated by Morris water maze (MWM) and novel object recognition (NOR). From the 3rd day to the 5th day of the spatial trial in MWM, wild-type BCAS mice subjected to antagomir spent less time achieving platform (Fig. S10a, P = 0.0161, P = 0.0002 and P < 0.001 on each day) and swam shorter distances than the N.C.-treated group (Fig. S10b, c, P = 0.0276, P = 0.0184 and P = 0.014 on each day). In contrast, antagomir-administered Cav-1−/− BCAS mice did not acquire promotion in either escape latency or path length (Fig. S10a–c, both P < 0.01, on each day). For probe trial without the platform, antagomir treatment ameliorated chronic ischemia-induced impairment in the crossovers of platform location as well as the time spent in the target quadrant in wild-type mice (Fig. S10c–e), which was not seen within antagomir-administered Cav-1−/− BCAS mice. In the NOR with 1-hour interval, wild-type BCAS mice that received antagomir took a longer time to explore new objects (Fig. S10f, compared to the N.C. group of wild-type mice, P = 0.0005). Another NOR task with 24-hour interval consistently showed increased interactions with the novel object in the antagomir-administered wild-type BCAS mice (Fig. S10g, compared to the N.C. group of wild-type mice, P = 0.0028). Animals in other groups could not distinguish the new object from the old one. Therefore, Cav-1 was necessary for the protection of white matter and cognitive function induced by miR-3074-1-3p antagomir. Endothelial Cav-1 has been well-characterized in maintaining BBB homeostasis. Here, we extended insights into ischemic demyelination and identified a role of Cav-1 in regulating oligovascular coupling. Chronic hypoxia and ischemia led to endothelial dysfunction and diminished Cav-1 expression, triggering HSP90α secretion from endothelial cells. Extracellular HSP90α exerted remarkable paracrine function by inhibiting oligodendrogenesis and myelin regeneration. Endothelial miR-3074-1-3p was screened out and acted as an upstream regulator of Cav-1. In leukoaraiosis patients, human serum miR-3074-3p, which shared a conserved seed region with mouse miR-3074-1-3p, was increased while serum Cav-1 was decreased. This reciprocal change was in accord with the results from the animal study, indicating the translation of laboratory findings to human biology. Inhibition of miR-3074-1-3p by nanoparticle-antagomir, dependent on Cav-1, could ameliorate OPC clustering and promote oligodendrogenesis, which was ultimately beneficial for cognitive performance. However, as the hippocampus is important for both spatial memory and recognition memory, the results might be confounded by the chronic ischemia-induced changes in the hippocampal region. Therefore, the conclusion should be treated with caution. Structural and functional changes in the vasculature may promote hypoperfusion. The damage and chronic remodeling of microvessels, such as narrowing of the arteriolar lumen and thickening of the vessel wall, can impair CBF regulation and cause ischemia in distal territories. We found significant loss of endothelial junction proteins and early vascular hyper-permeability in the CC of BCAS mice, which coincided with the findings in patients and animal models with chronic cerebral ischemia. Despite angiogenesis reported in response to chronic hypoxia, we found vessel regression and BMECs apoptosis under hypoxic injury, reflecting endothelial instability in long-term ischemic injury. Vascular dysfunction has been identified as an early prelude to white matter abnormalities in various pathological settings. Clinical studies found increasingly destroyed BBB in normal-appearing white matter at the proximity of white matter lesions, a predilection site for the expansion of hyperintensity. A recent animal study confirmed that endothelial dysfunction was an upstream pathogenic factor in oligodendroglial pathologies. We also discovered that endothelial impairments occurred before ischemic demyelination, indicating that myelin loss may be secondary to vascular dysfunction in ischemic settings. Disrupted OPC-vascular interaction could promote early BBB opening in chronic cerebral ischemia and multiple sclerosis. Conversely, increased OPC could interact with endothelium and promote white matter vascularization. In contrast to these studies, we focused on the effects of altered vascular-OPC association on oligodendroglial injury under the context of chronic cerebral ischemia. We found increased OPCs accumulating on the vasculature at 2 weeks post-BCAS in the CC. Interestingly, newly-formed oligodendrocytes of BCAS1+ and ENPP6+ cells significantly increased at the same time. We attributed this phenomenon to an endogenous protective mechanism in chronic ischemia, whose reservation on OPC differentiation surpassed the inhibition from dysfunctional endothelial cells in the early phase. However, along with the hypoperfusion, proliferated OPCs were increasingly gathered on the blood vessels, more like in a “traffic jam”, with growing cluster frequency and size. The overwhelmed differentiation of endogenous OPCs may act in concert with increased oligodendrocyte death, leading to a significant decline in pre-myelinating and mature oligodendrocytes at 4 weeks after BCAS. The OPC clustering not only suggested a substantial defect in single-cell perivascular interaction but also indicated vascular detachment failure, which may disturb OPC dispersal in the area of demyelination. Hence, it led us to investigate the modulatory mechanism behind oligovascular crosstalk after BCAS. Cav-1 has been reported to express in different cells in the CNS, including endothelial cells, pericytes and OPCs themselves. Through co-immunostaining, we found that Cav-1 was mostly expressed in the endothelial cells. Thus, the changes observed in Cav-1−/− mice were mainly due to the knockout of endothelial Cav-1. However, limitations existed, as it was quite difficult to exclude the effects of other Cav-1-expressing cells in the global Cav-1−/− mice. To this end, the genetic delivery of AAV-Tie1-Cav-1 was introduced before BCAS to enhance Cav-1 expression exclusively in endothelial cells. The findings that aberrant vascular-OPC association and demyelination were remarkably attenuated by AAV-Tie1-Cav-1 suggest that the function of Cav-1 linking endothelial damage and ischemic demyelination may be cell-autonomous in endothelial cells. Cav-1 can maintain cellular structure and normal permeability through the regulation of junction proteins expression and assembly. Loss of Cav-1 could trigger BBB hyper-permeability and inflammatory injury, confirming our data that BCAS induced simultaneous reduction of endothelial Cav-1 and junction proteins. Cav-1 is an important regulator in cell signal transduction. Mounting evidence has shown that HSP90 can be accumulated in the Cav-1 microdomain. Competitive association with Cav-1 could facilitate HSP90 release. We found endothelial Cav-1 could co-localize with HSP90α in the membrane. However, this interaction was reduced by hypoxia/ischemia. The mechanism underlying HSP90α secretion by BMECs remains unclear. We learned that Cav-1 could bind to the protein kinase A (PKA), impeding the PKA signaling. PKA has been reported to phosphorylate HSP90α and promote the translocation and exocytosis of HSP90α. We are herewith proposing that hypoxia/ischemia promoted HSP90α release through the interference of Cav-1. Moreover, it revealed that Cav-1 overexpression by AAV could restrain intra- and extra-endothelial HSP90α, while HSP90α siRNA did not change Cav-1 level. Thus, we identified HSP90α as a downstream secretory factor of endothelial Cav-1. Our data suggest that hypoxia-treated BMECs released high levels of HSP90α, leading to the cessation of OPC differentiation. In line with this, the BMECs of stroke-prone spontaneously hypertensive rats could also secrete HSP90α, which caused a reduction in OPC maturation. Though we have validated that the Hsp90α mRNA was mainly enriched in endothelial cells after BCAS, the results should be interpreted with caution, as the expression or secretion of HSP90α from other cells cannot be excluded. The mechanism was unknown regarding the secreted HSP90α on OPCs. It is reported that secreted HSP90α is capable of binding to low-density lipoprotein receptor-related protein 1 (LRP1), which is highly expressed in OPCs and has been defined as a negative regulator of OPC maturation. Moreover, as a chaperone of Cxcr4, increased HSP90α may induce Cxcr4 activation and mediate OPC attraction to the endothelium, preventing OPC recruitment and differentiation. We identified hypoxia-responsive miR-3074-1-3p in the BMECs and validated that miR-3074-1-3p was an upstream regulator of Cav-1 by specifically binding with its 3′ UTR. Under the context of chronic ischemia, PEI-antagomir-induced inhibition of miR-3074-1-3p attenuated physical coupling of endothelial cells and OPCs and mitigated demyelination dependent on Cav-1, further emphasizing the core value of Cav-1 in therapeutics for ischemic white matter disease. Considering that the abnormal OPC-endothelial interactions were reversible later in the disease when vascular dysfunction was already present, miR-3074-1-3p inhibition may be of great clinical value in ischemic demyelination, or other white matter diseases sharing a similar mechanism. In summary, we show the mechanism engaging in the endothelial-oligodendroglial association in chronic ischemia, and regulation of this aberrant interaction could restore myelination. Chronic ischemia caused vascular dysfunction and OPC perivascular clustering, in which endothelial Cav-1 occupied a central place. A significant decline of endothelial Cav-1 was responsible for HSP90α secretion, blocking oligodendrogenesis. Specific overexpression of endothelial Cav-1 could reverse ischemic myelin damage. Intervention by antagomir preserved endothelial Cav-1 and suppressed the vascular HSP90α release, thereby attenuating OPC-vascular interactions and demyelination. Therefore, our findings may shed light on the therapeutic implications of endothelial Cav-1 in diseases involving white matter pathologies. All experiments were performed following the National Institutes of Health Guide and approved by Jinling Hospital Animal Care Committee. The participants or legal representatives who acknowledged the use of blood samples signed an informed consent form before being included in the study. The protocol was approved by the Ethics Committee of Jinling Hospital. Antibodies against CNPase (ab6319), GFAP (ab53554), Ki67 (ab15580), MAG (ab89780), MBP (ab40390), PLP (ab28486), Cav-1 (ab2910), Collagen IV (ab6586), HSP90α (ab79849), NeuN (ab104224) were purchased from Abcam, UK; antibody against BCAS1 (bs-11462R) was purchased from Bioss, China; antibody against CD31 (550274) was purchased from BD Biosciences, USA; antibodies against β-actin (#8457), Cav-1 (#3267) were purchased from Cell Signaling Technology, USA; antibody against HSP90α (ADI-SPS-771-D) was purchased from Enzo Life Sciences, USA;antibody against Occludin (R1510-33) was purchased from Huabio, China; antibodies against ENPP6 (PA5-25140), Claudin-5 (35-2500), Occludin (71-1500) were purchased from Invitrogen, USA; antibodies against CC1 (OP80), Olig2 (AB9610) were purchased from Millipore, USA; antibody against O4 (MAB1326-SP) was purchased from R&D Systems, USA; antibodies against PDGFR-α (sc-398206), ZO-1 (sc-33725), MBP (sc271524), Iba-1 (sc-32725), Cav-1(sc-53564) were purchased from Santa Cruz Biotechnology, USA; All antibodies were used at a dilution of 1:50–1:1000 for immunofluorescence, 1:500–1:1000 for immunoblotting analysis unless otherwise specified. Secondary antibodies were donkey-anti-mouse or anti-rabbit or anti-rat conjugated with either Alexa 488 or Alexa 594 or Alexa 647 (Jackson, USA; 1:400), goat anti-mouse or rabbit or rat IgG HRP (Cell Signaling Technology, USA; 1:5000). Mouse anti-goat IgG-HRP (sc-2354, Santa Cruz Biotechnology) was used for co-immunoprecipitation. Adult male (24–29 g; for BCAS surgery) and C57BL/6J mice (6–8 weeks old; for BMECs primary culture) were purchased from Gempharmatech co., Ltd (Nanjing, Jiangsu, China). Cav-1 knock-out mice (Cav-1tm1Mls/J, C57BL/6 background, 007083) were purchased from the Jackson Laboratory (Bar Harbor, Maine, USA). All animals were housed in a 12 h light/dark cycle at approximately 25 °C and provided with free access to food and water. For BCAS surgery, mice were anesthetized with 5% isoflurane and maintained with 2% isoflurane in oxygen (RWD Life Science Co., LTD). Both common carotid arteries (CCAs) were carefully dissected and exposed through the midline incision. A microcoil with an internal diameter of 0.18 mm (Sawane Spring Co. Japan) was twined around the CCA on one side, followed by another twinning on the other side 30 minutes later. Rectal temperature was maintained between 36.5 °C and 37.5 °C during surgery. After the operation, the mice were taken care of and provided with ad libitum access to water and food. To study the disease progression, mice were divided into four time-point groups: Control, BCAS_1 week, BCAS_2 weeks and BCAS_ 4 weeks. Mice with sham operation served as controls. Control mice were given a skin incision and their CCAs were exposed without inserting the microcoil. To confirm the BCAS modeling, CBF was measured before, 1 week, 2 weeks, and 4 weeks after surgery by laser speckle contrast imaging (RWD Life Science Co., LTD). Anesthetized by 2% isoflurane in oxygen, mice were placed in the prone position. The skull was exposed by a midline scalp incision and cleaned with sterile normal saline. Color-coded blood flow images obtained in high-resolution mode (2048 ×2048 pixels; 1 image/sec) were captured by a CMOS camera positioned above the head and transferred to a computer for analysis. By the color image program incorporated in the flowmetry system, images were analyzed to obtain the average value of blood flow. The mean CBF of 5 mice in each group was determined. The value of blood flow was expressed as a percentage of the baseline blood flow. Spatial learning and memory were evaluated by MWM. Briefly, mice were trained for five consecutive days to find a hidden platform. On the sixth day, the platform was removed and mice were placed to swim for 1 min in probe trial. The escape latency to find the platform, the swim path length, and the percentage time in the platform quadrant and platform crossovers were recorded and analyzed by the ANY-maze video tracking software (Stoelting, USA). The NOR test was conducted to assess novelty preference. In brief, mice were habituated in a box for testing and got familiarization with 2 identical objects in the box. Then memory function was evaluated either 1 h or 24 h later by replacing one of the familiar objects with a novel one that was different in shape and appearance. The time spent on exploring the familiar object (F) and the new object (N) was respectively recorded in a 5 minute-trial. The index of discrimination defined as (N-F)/(N + F) was calculated. As modified from a published method, brains were obtained from wild-type or Cav-1−/− mice at the age of 6–8 weeks with cerebellum and meninges removed. After homogenization, the homogenate of brains was centrifugated at 150 × g for 5 min. The pallet was then resuspended in a 15% dextran solution (mol wt 60,000–76,000, Sigma-Aldrich) and centrifugated at 400 × g to isolate blood vessel fragments. The pallet of fragments was digested in collagenase/dispase (1 mg/ml, Roche) and DNase I (10 μg/ml, Roche) at 37 °C. The dissociated BMECs were washed and seeded onto coated plates in DMEM/F12 media with 20% FBS, 1% PS, 1% endothelial cell growth supplement (SclenCell), 1% l-glutamine, 1% heparin, and 2 ng ml−1 bFGF (Biolegend). In vitro hypoxia was induced by either chronic CoCl2 (10 μM, Sigma-Aldrich) for 5 consecutive days or OGD treatment for 4 h. Primary mouse OPCs were isolated from pups of postnatal days 0–2 as reported. In brief, mixed glial cells were cultured in DMEM/F12 with 10% FBS and 1% PS for 7 days. Then the flasks were shaken on the orbital shaker at 200 × g for 1 h. After washed, the flasks continued to shake for another 18–20 h at 4 × g. OPCs were obtained and plated onto coated plates in DMEM/F12 media containing 1% BSA (Gibco), 20 ng ml−1 ITSS (Roche), 20 ng ml−1 PDGF-BB (Biolegend) and 20 ng ml−1 bFGF (Biolegend). After exposure to CoCl2/OGD or antagomir/agomir, the medium was removed and BMECs were washed by PBS. Then BMECs were cultured in serum-free medium, including DMEM/F12 media with 1% PS, 1% endothelial cell growth supplement (SclenCell), 1% l-glutamine, 1% heparin, and 2 ng ml-1 bFGF, for another 2 days. CM was collected and stored at −80 °C until use. Before applying to OPCs, the CM was centrifugated at 230 × g for 5 min to remove the cell debris. The normal culture media was replaced by CM 1 day after OPCs were seeded onto the plates. OPCs were cultured under CM for 2 days before being analyzed. For HSP90α experiments, recombinant HSP90α (1 μg/ml, #ADI-SPP-776-D, Enzo Life Sciences) or HSP90α blocking antibody (1.7 μg/ml, #ADI-SPS-771-D, Enzo Life Sciences) was added into the CM before transferring to OPCs. From January 2021 to April 2021, 154 patients who were more than 45 years old and referred to Jinling Hospital for further assessment due to dizziness and nonspecific headaches without migraines were recruited. All subjects performed an MRI examination for assessing the presence and severity of leukoaraiosis. Baseline data including age, gender, body mass index, and vascular risk factors (including hypertension, diabetes, hyperlipemia, and smoking) were recorded. The exclusion criteria of this study were as follows: (1) leukoaraiosis which was not due to vascular origins, such as multiple sclerosis and leukodystrophy; (2) signs of neurological deficit; (3) history of stroke, brain trauma, intracranial tumor, central nervous system infection, active malignancy, thyroid diseases, autoimmune diseases, and active or chronic inflammatory diseases; (4) contraindication for MRI examination. The patients gave written informed consent before entering the study. The patients were not financially compensated. MRI scan of whole brain was performed with a 3.0 T system (TIM Trio; Siemens Healthineers, Erlangen, Germany) in all patients within 7 days after admission. Leukoaraiosis was defined on T2-fluid-attenuated inversion recovery (FLAIR) imaging (T2-FLAIR parameters: number of slices = 25, slice matrix = 512 × 512, field of view (FOV) = 220 × 220 mm2, repetition time (TR) = 8000 ms, echo time (TE) = 93 ms, echo train length = 24, slice thickness = 5 mm, spacing between slices = 7 mm, flip angle = 130°, inversion time = 2500 ms.) and graded using the Fazekas scale. According to previous studies, we dichotomized leukoaraiosis according to its severity as mild (Fazekas scores of 0 or 1) and severe (Fazekas scores ≥ 2) in periventricular and deep subcortical region. All images were analyzed by two experienced MRI-specialized neuroradiologists who were blinded to the patients’ clinical information. In case of disagreement, lesions were ascertained by consensus. An intra-rater reliability test was performed on 50 subjects, the kappa values for the presence and severity of leukoaraiosis were 0.91 and 0.87, respectively. BMECs in the control and CoCl2 groups were washed and cultured in the serum-free media for 2 days. CM was then collected and concentrated for the array test. The antibody array covering more than 1300 antibodies (#SET100, Full Moon BioSystems) was performed according to the manufacturer’s manual. ELISA was conducted as specified in the protocol of the HSP90AA1 ELISA Kit (AVIVA SYSTEMS BIOLOGY, USA) and Human Cav-1 ELISA Kit (Sabbiotech, USA). Briefly, 100 μl of CM or patient serum was added into each well of the microplate and incubated at 37 °C for 2 h. The liquid was then removed and a 100 μl Detector Antibody was added to the well for 1 h of incubation. After washing, 100 μl avidin-HRP conjugate was added and incubated for another 1 h. With wells washed, 90 μl TMB substrate was added, followed by the addition of a stop solution 30 minutes later. O.D. absorbance was obtained at the wavelength of 450 nm. The RNA was extracted from the CC tissue of control and BCAS_4w mice (n = 4 in each group). miRNA sequencing was performed by Illumina HiSeqTM 2500 platform and data analysis was conducted by Ribo Bio (Co., Ltd, Guangdong, China). The quantity and integrity of RNA yield were evaluated by the K5500 (Beijing Kaiao, China) and Agilent 2200 TapeStation (Agilent Technologies, USA). Total RNA (1 µg) of each sample was used to prepare small RNA libraries by NEBNext® Multiplex Small RNA Library Prep Set for Illumina (NEB, USA) according to manufacturer’s instructions. The libraries were sequenced by HiSeq 2500 (Illumina,USA) with single-end 50 bp at Ribobio Co. Ltd. The raw reads were processed by filtering out containing adapter, poly ‘N’, low quality, smaller than 17nt reads by FASTQC to get clean reads. Mapping reads were obtained by mapping clean reads to reference genome of by Burrows-Wheeler-Alignment Tool. The miRNA expression was calculated by Reads Per Million (RPM) values (RPM = (number of reads mapping to miRNA/ number of reads in clean data) × 106). The expression levels were normalized by RPM, which equals to (number of reads mapping to miRNA/number of reads in clean data) × 106. Differentially expressed miRNAs were then screened by an adjusted P value of <0.05 and at least a two-fold change of expression. Serum miRNAs were extracted from patients’ samples according to the manufacturer’s manual of miRNeasy Serum/Plasma kit (QIAGEN, Germany). Total RNA was extracted from cells and tissues using TRIzol (Invitrogen, USA), followed by reverse transcription using RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific, USA). Real-time PCR based on SYBR Green was performed by Stratagene Mx3000P QPCR system (Agilent Technologies, USA). Primers for reverse transcription were listed in Table S2. The primer pairs used for real-time PCR were listed in Table S3. With Lipofectamine 3000 Reagent (Thermo Fisher Scientific, USA), BMECs were transfected with 50 nM miR-3074-1-3p N.C./agomir/antagomir (Ribo Bio, Co., Ltd, China). The 293 T cells were seeded in the 96-well plates and co-transfected with a 1 μg vector containing 3′ UTR of mouse/human Cav-1 and 100 nM mimic N.C./mmu-miR-3074-1-3p/hsa-miR-3074-3p mimic (Ribo Bio, Co., Ltd, China) using Lipofectamine 3000. After 24 h of transfection, the luciferase activity was measured by the Dual-Luciferase® Reporter Assay system (Cat# E1910, Promega, USA). HSP90α-siRNA or miR-3074-1-3p antagomir was diluted with 10% glucose solution and RNAse/DNAse free water. Separately, in vivo-PEI™ was diluted to the same volume by 10% glucose. The solutions were then mixed and incubated at room temperature for 15 min to form complexes with an N/P ratio of 6. For Cav-1−/− Mice, 20 μg PEI-N.C. siRNA or PEI-HSP90α siRNA was injected into mice via tail vein every 3 days for a month. To over-express endothelial Cav-1 in BCAS mice, a cDNA encoding Cav-1 sequence, the enhanced GFP reporter gene, and endothelial-specific Tie1 promoter were produced and inserted into AAV packaging vectors. AAV containing control and Cav-1 vectors was purchased from GeneChem co., Ltd (Shanghai, China) and injected in the CC (0.5 mm anterior-posterior, 1.0 mm medial-lateral, −2.1 mm dorsal-ventral relative to bregma) using stereotaxic injection 3 weeks before BCAS surgery. All the animals were injected bilaterally with 1 μl of AAV-Tie1-C (1.13 × E13 v.g/ml) or AAV-Tie1-Cav-1 (1.95 × E13 v.g/ ml). For continuous in vivo delivery of antagomir, osmotic infusion pumps (model 1002, Alzet) loaded with PEI-antagomir N.C. or PEI-antagomir and cannulas (Plastics One, USA) were connected and stereotactically inserted in the CC on the 15th-day post BCAS. Approximately a total of 2 μg PEI-antagomir was infused daily for a consecutive 14 days. For antagomir tracking in mice, fluorescence whole-animal imaging was performed on the IVIS imaging system 1 day and 2 weeks after the implantation of pumps loaded with Cy5-labeled antagomir. For TEM, the CC tissue was fixed in 2.5% glutaraldehyde. After dehydration and embedding, samples were cut to 60–80 nm slices and scanned by an H7500 Transmission Electron Microscope (Hitachi, Japan). The TJ length, gap area, the percentage of myelinated fibers, and myelin thickness were measured and analyzed. For immuno-TEM, samples were cut into 70–80 nm slices after fixed and embedding. By immunogold labeling of Olig2, samples were observed and images were captured with H7500 Transmission Electron Microscope (Hitachi, Japan). Mice were anesthetized and transcardially perfused with PBS and 4% PFA. Brains were dissected, postfixed in 4% PFA, and dehydrated in gradient sucrose. Embedded in Tissue-Tek O.C.T compound (Sakura Finetek, USA), frozen brains were cut into 20 μm-thick sections. For immunostaining, sections or cell coverslips were fixed in PFA, followed by blocking with 5% goat serum, 1% BSA, and 0.1% Triton X-100 for 1 h. After incubation with primary antibodies at 4 °C overnight, the sections were incubated with secondary antibodies. Finally, images were captured using a BX51 microscope (Olympus, Japan) or a laser-scanning confocal microscope (FV3000, Olympus) and were prepared using Adobe Photoshop (version 21.0.2). The 3D reconstruction was analyzed by IMARIS software (Bitplane). White matter LFB staining was carried out with the degree of white matter damage scored and Black-gold II staining (Sigma-Aldrich, USA) was performed with the calculation of myelinated area. For histochemical evaluation of BBB integrity, mice were injected with FITC-dextran (3 kDa, Sigma-Aldrich) via CCA under deep anesthesia. After circulation for 90 min, brains were isolated and immediately fixed in 4% PFA, followed by dehydration in gradient sucrose. Dextran was visualized with a laser-scanning confocal microscope (FV3000, Olympus). TUNEL staining (Beyotime, China) and EdU proliferation assay (Ribo Bio, Co., Ltd, China) were performed on cell coverslips following the manufacturer’s protocol. For TUNEL staining, cell coverslips were fixed in 4% PFA and stained with TUNEL reagent. For EdU proliferation assay, cells were incubated with EdU buffer and fixed with 4% PFA. EdU solution was then added to coverslips followed by the staining of DAPI. Finally, the slides were visualized by fluorescence microscopy. The positively labeled cells were calculated by Image J. Combined FISH for miR-3074-1−3p and HSP90α with immunostaining was conducted. The FAM-labeled miR-3074-1-3p probe was used to detect miR-3074-1-3p in Cav-1-stained endothelial cells. The FAM-labeled HSP90α probe was used to detect HSP90α in CD31-labeled endothelial cells, PDGFRβ-labeled pericytes, GFAP-labeled astrocytes, and Olig2-labeled oligodendroglial lineage cells. Briefly, the frozen sections were fixed in 4% PFA and incubated with probe hybridization solution overnight. The slides were then washed and blocked for the immunostaining. Nuclei were stained with DAPI and the slides were mounted in an anti-fade reagent with DAPI (Vector Laboratories). Lysates were extracted from the CC brain tissue, microvascular segments, and cultured BMECs. Briefly, as for microvascular segment extraction, a pool of two mouse brains was defined as one group. Collected brains were homogenized and centrifuged at 150 × g. The pellet was then resuspended in 15% dextran solution for twice layered centrifugation at 400 × g for 10 min. The supernatant and upper myelin debris were discarded and the microvascular segment at the bottom was resuspended in RIPA lysis buffer (Cell Signaling Technology, USA) with 1% PMSF. Also, brain and cell lysates were harvested using RIPA lysis buffer with 1% PMSF. The concentration of protein was quantified by BCA Protein Assay Kit (Beyotime, PR China). For co-immunoprecipitation, protein extracted from endothelial cells was incubated with Cav-1, HSP90α or negative control IgG overnight in a 4 °C shaker. Protein A/G agarose beads were then added to the complex and incubated for 4 h at 4 °C. After centrifugation, the agarose beads were collected and protein was eluted. For immunoblotting, an equal amount of protein was loaded, separated by SDS-PAGE, and incubated with primary antibodies. After incubation of the secondary antibodies for 1 h, the special protein signals were detected by Immobilon Western Chemiluminescent HRP substrate (Millipore, USA). CD31/PDGFRα/Olig2 triple staining was used to detect the association of OPCs with blood vessels in the corpus callosum. An OPC cluster is defined as a perivascular aggregation of OPCs with more than 4 cell bodies in direct contact with blood vessels. The total area of the cluster by PDGFRα+ or Olig2+ signals was outlined and measured by Image J, which was further normalized to the area of the image. The vessel density was quantified by dividing the CD31+ vessel number by the area of the selected region. CD31+ vessel length was measured by Image J, which was normalized to the area of the image. The extravasated dextran was outlined and its fluorescence intensity was measured by Image J, which was then normalized to the fluorescence intensity of the vessels. The extra- and intra-endothelial HSP90α were chosen respectively. And the HSP90α level changes were measured by dividing the HSP90α fluorescence intensity by the CD31 fluorescence intensity. The severity of demyelinated lesions was graded by LFB staining as follows: grade 0 (normal), grade 1 (disarrangement of the nerve fibers), grade 2 (marked vacuoles), and grade 3 (the disappearance of myelinated fibers). (1) We chose 4 capillaries and 4 arterioles for each sample. The area of tight junction gap between endothelial cells was measured and caveolae number was counted by Image J. (2) The percentage of myelinated axons and myelin thickness were calculated and analyzed. Cell counting and fluorescence intensity analyses were conducted on five randomly chosen fields for each sample using Image J. The results were normalized to the areas of interest or the total cell population in the selected region. Results were analyzed by GraphPad Prism software, SPSS version 24.0 (SPSS Inc., Chicago, IL, USA), and Stata version 16.0 (StataCorp, College Station, TX). All data were expressed as mean ± S.D. Continuous variables were assessed by paired t-test for two groups and one-way ANOVA followed by Tukey post hoc test for multiple comparisons. Escape latency and swimming path length in the MWM test were analyzed by two-way repeated-measures ANOVA, followed by Tukey’s post hoc test. We compared patients with and without leukoaraiosis using unpaired t-test or the Mann-Whitney U tests for continuous variables and Pearson χ2 and Fisher exact tests for categorical variables. We constructed a logistic regression analysis to evaluate the risk factors of leukoaraiosis. A multivariable regression model was adjusted for variables with a P-value <0.1 in univariate analysis. Results were presented as odds ratio with a 95% CI. We also performed the linear regression to assess the association between levels of hsa-miR-3074-3p and Cav-1 in patients with leukoaraiosis. Two-sided P value < 0.05 was considered statistically significant. Further information on research design is available in the Nature Research Reporting Summary linked to this article. Peer Review File Supplementary Information Reporting Summary

PMC9649812

Wei Chen,Pingting Liu,Dong Liu,Haoliang Huang,Xue Feng,Fang Fang,Liang Li,Jian Wu,Liang Liu,David E. Solow-Cordero,Yang Hu

Maprotiline restores ER homeostasis and rescues neurodegeneration via Histamine Receptor H1 inhibition in retinal ganglion cells

10-11-2022

Optic nerve diseases,High-throughput screening

When the protein or calcium homeostasis of the endoplasmic reticulum (ER) is adversely altered, cells experience ER stress that leads to various diseases including neurodegeneration. Genetic deletion of an ER stress downstream effector, CHOP, significantly protects neuron somata and axons. Here we report that three tricyclic compounds identified through a small-scale high throughput screening using a CHOP promoter-driven luciferase cell-based assay, effectively inhibit ER stress by antagonizing their common target, histamine receptor H1 (HRH1). We further demonstrated that systemic administration of one of these compounds, maprotiline, or CRISPR-mediated retinal ganglion cell (RGC)-specific HRH1 inhibition, delivers considerable neuroprotection of both RGC somata and axons and preservation of visual function in two mouse optic neuropathy models. Finally, we determine that maprotiline restores ER homeostasis by inhibiting HRH1-mediated Ca2+ release from ER. In this work we establish maprotiline as a candidate neuroprotectant and HRH1 as a potential therapeutic target for glaucoma.

Maprotiline restores ER homeostasis and rescues neurodegeneration via Histamine Receptor H1 inhibition in retinal ganglion cells When the protein or calcium homeostasis of the endoplasmic reticulum (ER) is adversely altered, cells experience ER stress that leads to various diseases including neurodegeneration. Genetic deletion of an ER stress downstream effector, CHOP, significantly protects neuron somata and axons. Here we report that three tricyclic compounds identified through a small-scale high throughput screening using a CHOP promoter-driven luciferase cell-based assay, effectively inhibit ER stress by antagonizing their common target, histamine receptor H1 (HRH1). We further demonstrated that systemic administration of one of these compounds, maprotiline, or CRISPR-mediated retinal ganglion cell (RGC)-specific HRH1 inhibition, delivers considerable neuroprotection of both RGC somata and axons and preservation of visual function in two mouse optic neuropathy models. Finally, we determine that maprotiline restores ER homeostasis by inhibiting HRH1-mediated Ca2+ release from ER. In this work we establish maprotiline as a candidate neuroprotectant and HRH1 as a potential therapeutic target for glaucoma. The most common cause of irreversible blindness, glaucoma, will affect an estimated 3% of the world population over 40 years old by 2040 (more than 100 million people), which will impose a multi-billion dollar economic burden on society. Glaucoma is characterized by optic neuropathy with optic nerve (ON) degeneration followed by progressive retinal ganglion cell (RGC) death. The only available treatments act by reducing intraocular pressure (IOP), a risk factor associated with glaucoma. However, IOP reduction fails to completely prevent the progression of glaucomatous neurodegeneration, indicating the urgent need for innovative neuroprotection therapies. We previously found that ON injury induces neuronal endoplasmic reticulum (ER) stress in RGCs, suggesting a detrimental role of RGC-specific ER stress in glaucoma. When the protein or calcium homeostasis of the ER is adversely altered, cells experience ER stress and activate three signaling pathways initiated by three ER-resident stress-sensing proteins: inositol-requiring protein-1 (IRE1α), activating transcription factor-6 (ATF6) and protein kinase RNA-like ER kinase (PERK), together called the unfolded protein response (UPR). IRE1α, a bi-functional enzyme that contains both a Ser/Thr kinase domain and an endoribonuclease (RNase) domain, mediates the splicing of X-box binding protein 1 (XBP-1) mRNA to generate an active (spliced) form of the transcription factor, XBP-1s. The IRE1α-XBP-1s pathway targets genes that increase ER protein-folding capacity and facilitate degradation of misfolded proteins. On the other hand, IRE1α kinase activity also activates pro-apoptotic c-Jun kinase (JNK), which contributes to Bax-dependent IRE1α-induced apoptosis. ATF6 is a transcription factor that is truncated and thereby activated by ER stress to control the expression of a group of UPR target genes. PERK phosphorylates and inactivates eukaryotic translation initiation factor 2α (eIF2α) to attenuate global cap-dependent mRNA translation and thereby reduce protein load on the ER. Activating transcription factor 4 (ATF4) downstream of PERK-eIF2α induces expression of ER stress-specific transcription factor C/EBP homologous protein (CHOP). CHOP is a well-known pro-apoptotic transcription factor that mediates ER stress-induced cell death by downregulating anti-apoptotic Bcl2, upregulating pro-apoptotic BH-3 only molecules Bim and PUMA, increasing expression of death receptor 5 (DR5) and caspase 8 cleavage. ATF4 can also form heterodimers with CHOP to cause cell death by upregulating protein synthesis and inducing oxidative stress. Chronic ER stress with prolonged PERK-eIF2α-ATF4-CHOP signaling has been associated with many acute and chronic neurodegenerative diseases; genetic manipulation and small molecular modulators of this pathway have proven to be beneficial in animal models of various neurodegenerative diseases. We also previously found that genetic inhibition of CHOP or its upstream regulator eIF2α significantly protects RGCs’ somata and axons and preserves visual functions in mouse models of traumatic ON injury, glaucoma, and optic neuritis. Therefore, identification of small modulators of ER stress to block CHOP is an important step toward developing effective neuroprotectants for glaucoma. We reasoned that identifying CHOP inhibitors from FDA-approved drugs would significantly shorten the drug development process. Therefore, we used a reporter cell line expressing CHOP promoter-driven luciferase to perform a high throughput screening (HTS) of five compound libraries with a total of 4846 compounds that have known bioactivities and have been approved for clinical application. In this work, we identify three FDA-approved drugs, amoxapine, desloratadine, and maprotiline, as potent ER stress inhibitors. They share similar tricyclic chemical structures and a common antagonistic target, histamine receptor H1 (HRH1), through which we find that they restore ER homeostasis and achieve significant neuroprotection of RGCs and ONs in vivo in two mouse optic neuropathy models. Finally, we determine that maprotiline inhibits HRH1-mediated ER Ca2+ release and thereby inhibits the damaging intracellular Ca2+ influx induced by axon injury. This readily testable small molecule drug is a promising neuroprotectant, and HRH1 is a potential therapeutic target for glaucoma and other neurodegenerative diseases associated with ER stress. We generated a stable cell line to express CHOP promoter-driven luciferase (CHOP-Luc) in HEK293T cells. This promoter shows dose-dependent responses to the ER stress inducers thapsigargin (Tg) and tunicamycin (Tm) (Supplementary Fig. 1a–c). We used this reporter cell line in a multiple dose-response assay to screen 4846 compounds from five FDA-approved drug libraries. All compounds were run in 7-point dose response based on a previous publication, except the NIH Clinical Collection (446 compounds), which were tested in duplicate due to the small quantity of compounds on hand. We identified 89 “hits” (Fig. 1a–c) based on the criteria: (1) >30% inhibition of Tm/Tg induced CHOP-Luc signal; (2) no cell toxicity or luciferase inhibition activity; (3) dose-dependent effect. Among the 89 hit compounds, five (a–e) with good dose-dependent inhibition of CHOP-Luc share a similar tricyclic chemical structure (Supplementary Fig. 1d). We then focused on this series of compounds and added nine more tricyclic or tetracyclic FDA-approved small molecule drugs (not in the libraries) for retesting with the CHOP-Luc reporter line; we excluded compounds “c” and “d” because they are not commercially available (Supplementary Fig. 1d). Among the 12 retested compounds, only three (amoxapine, desloratadine, and maprotiline) significantly inhibited CHOP-luciferase activity induced by ER stress (Tm/Tg) at 10 µM (Fig. 1d). Two known inhibitors of the PERK-CHOP pathway, GSK2606414 and ISRIB, were used as reference compounds. However, only GSK2606414 but not ISRIB significantly inhibited CHOP-luciferase activity induced by Tm/Tg (Fig. 1d and Supplementary Fig. 1e), indicating the low sensitivity of the reporter line. Another five compounds (desipramine, trifluoperazine, clomipramine, amitriptyline, olanzapine) inhibited CHOP-luciferase activity at 20 µM (Supplementary Fig. 1e). Interestingly, a previous study also identified trifluoperazine as a CHOP inhibitor using a different CHO luciferase reporter cell line driven by murine CHOP promoter. Maprotiline appeared to be the most potent of the three as it has the lowest IC50 (Fig. 1e, f). Importantly, these three drugs had no obvious cell toxicity within the concentration ranges that we tested (Fig. 1g) and they did not induce ER stress by themselves (Supplementary Fig. 1f). To determine whether these compounds inhibit the CHOP branch of UPR preferentially or act as general ER stress modulators that also affect the other two branches, we examined the signaling cascades downstream of the three UPR pathways induced by ER stress after treatment of HEK293T cells with each of these compounds. We again used Tm/Tg (1 µM) to induce ER stress and compared the three hit compounds (10 µM) to DMSO control: (1) The three compounds significantly inhibited PERK phosphorylation and expression of ATF4 and CHOP, indicating upstream modulation of the PERK-CHOP pathway (Fig. 2a). (2) The three compounds also significantly inhibited ATF6 expression (Fig. 2b). (3) Amoxapine and maprotiline likewise significantly inhibited the third UPR pathway downstream of IRE1α activation, as indicated by XBP-1 mRNA splicing (Fig. 2c). All three compounds also downregulated XBP-1s protein level (Fig. 2d). JNK phosphorylation is another downstream effector of IRE1α and contributes to cell death, especially in glaucoma. Desloratadine and maprotiline significantly inhibited JNK phosphorylation and maprotiline inhibited IRE1α phosphorylation (Fig. 2d). Lastly, we used qPCR as additional confirmation that these three compounds inhibited various downstream genes of UPR at the mRNA levels; maprotiline showed the most striking modulation of all the UPR pathways (Fig. 2e). We previously demonstrated that traumatic ON injury induces ER stress in RGCs at 3 days post crush (3dpc). To determine the in vivo effects of these three ER stress modulators, we delivered each of the three compounds into mouse eyes by intravitreal injection on the same day after ON crush (ONC) and examined the expression of the key ER stress molecules in RGCs at 3dpc and the survival of RGC somata and axons at 14dpc (Fig. 3a). Consistent with our cell-based in vitro assays, local administration of the three compounds significantly inhibited ONC-induced CHOP and ATF4 expression, and the phosphorylation of eIF2α and JNK in RGCs examined in both retinal sections (Fig. 3b–i) and retinal wholemounts (Supplementary Fig. 2a, b). Interestingly, western Blot assays of ON lysates demonstrated that the three compounds also inhibited ER stress molecules elevated by ONC in the ONs (Supplementary Fig. 2c, d). ONC is extensively used as a traumatic optic neuropathy model that injures all RGC axons and causes universal RGC and ON degeneration. Therefore, we next used this model to examine the effects of the three ER stress inhibitors on RGC soma and axon survival. We again delivered the three compounds into one of the mouse eyes by intravitreal injection on the same day as ONC, left the contralateral eye as internal control, and then maintained the compound exposure by daily intraperitoneal (i.p.) injection to avoid repeated intravitreal injection. The control group was treated with vehicle (DMSO). We have used optical coherence tomography (OCT) before to image and measure retina thickness in living animals as an accurate in vivo morphological readout for RGC degeneration. OCT images at 14dpc showed significant thinning of the ganglion cell complex (GCC), in crushed eyes treated with DMSO compared to contralateral naïve eyes, whereas crushed eyes treated with amoxapine, desloratadine, or maprotiline showed significantly thicker GCC than DMSO-treated eyes (Fig. 4a, b), suggesting significant RGC neuroprotection by these compounds. Histological analysis of post-mortem retina wholemounts and ON semi-thin sections consistently demonstrated significant loss of RGC somata and axons at 14dpc in the DMSO control group, whereas the survival of RGCs and axons was much higher in the compound-treated eyes (Fig. 4c–f). We confirmed the axon protection effect of these compounds by TEM analysis of ON cross sections (Supplementary Fig. 3). ISRIB is a small molecule inhibitor of the PERK pathway identified through cell-based screening. It showed no effect with the reporter cell line (Fig. 1d), but ISRIB provided neuroprotection in the ONC mouse model, to a lesser degree than maprotiline (Fig. 4a–f). Taken together, these results show that in vivo application of each of the three compounds (amoxapine, desloratadine, and maprotiline) effectively inhibits RGC ER stress and significantly protects RGCs and ONs after traumatic ON injury. Because maprotiline shows the most potent and consistent effects on ER stress modulation and neuroprotection both in vitro and in vivo, we focused on characterization of maprotiline in the subsequent experiments. We previously developed the silicone oil-induced ocular hypertension (SOHU) mouse glaucoma model, which faithfully replicates human secondary glaucoma with persistent elevation of IOP and severe degeneration of RGCs and ON. To test the effect of maprotiline on glaucomatous neurodegeneration, we generated the SOHU glaucoma model in one eye, used the contralateral eye as sham control, and treated the animals both systemically by i.p. injection + by local retrobulbar injection of compounds or vehicle (DMSO). We performed i.p. injection daily based on the presence of maprotiline in the retina 6 and 24 h after injection (Supplementary Fig. 4a). We did not use intravitreal injection to deliver the drug directly into the eye because intravitreal injection itself can lower IOP and therefore compromise the ocular hypertension glaucoma model. CHOP and ATF4 expression were elevated in glaucomatous RGCs at one week post SO injection (1wpi) in the SOHU model; their expression was significantly inhibited by systematic administration of the three compounds (Supplementary Fig. 4b, c). Next, we focused on maprotiline. Maprotiline treatment did not affect normal IOP in naïve mice, nor elevated IOP in SOHU mice (Fig. 5a). In vivo OCT retinal imaging showed significant thinning of the GCC at 3 weeks post SO injection (3wpi) in the DMSO group, whereas maprotiline treatment significantly increased GCC thickness (Fig. 5b, c). Histological analysis of post-mortem retina wholemounts and semi-thin ON sections consistently demonstrated significantly greater RGC soma and axon survival in the maprotiline group than in the DMSO group (Fig. 5d, e). We confirmed the axon protection effect of maprotiline in the SOHU glaucoma model by TEM analysis of ON cross sections and CTB tracing in wholemount ONs (Supplementary Fig. 5). The clinical significance of neuroprotection depends on preserving neuronal function. Therefore, we also investigated whether maprotiline preserves visual function in the glaucomatous mice. The optokinetic tracking response (OKR) is a natural reflex that objectively assesses mouse visual acuity. Another important electrophysiological assessment of RGC function is the pattern electroretinogram (PERG), in which the ERG responses are stimulated with contrast-reversing horizontal bars alternating at constant mean luminance. We used both techniques, which are well established in our lab, to evaluate maprotiline’s effect on glaucomatous eyes. Consistent with our morphological and histological results, maprotiline significantly preserved visual function in glaucomatous eyes, as demonstrated by improved amplitude of PERG (Fig. 5f) and visual acuity (Fig. 5g) compared to the DMSO control group. Taken together, these results show that maprotiline treatment achieves significant RGC and ON neuroprotection and preserves visual functions in a mouse glaucoma model, confirming its potential as a neuroprotectant. We next explored potential downstream effectors of the hit compounds for ER stress modulation and neuroprotection. We reasoned that since the three ER stress modulators, amoxapine, desloratadine, and maprotiline, have similar chemical structures, they may act on a common downstream target to restore ER homeostasis. Because, intriguingly, all three agents are potent antagonists of HRH1 with high binding affinities, we hypothesized that HRH1 inhibition may mediate the effects of these compounds on ER stress modulation. To test this hypothesis, we first overexpressed (OE) human HRH1 (Supplementary Fig. 6a) in the CHOP-Luc reporter cell line by transient transfection. The overexpression of HRH1 significantly but not completely reversed the three compounds’ inhibitory effect on Tm/Tg induced CHOP expression (Fig. 6a), indicating that other mechanisms in addition to HRH1 inhibition may also contribute to the compounds’ effects on ER stress. To test whether other UPR pathways are also modulated by HRH1, we generated a stable XBP-1-Luc HEK293T reporter cell line expressing the human XBP-1 fragment-fused luciferase construct (Supplementary Fig. 6b). This construct contains a 26 nt intron sequence that will be removed from the mRNA by IRE1α upon ER stress. Splicing of the 26 nt intron will allow a shift of the open reading frame in the mRNA to express luciferase, which will serve as a reporter for the activation of IRE1α pathway. This cell line shows consistent dose response to Tg and Tm (Supplementary Fig. 6c–e). Maprotiline also shows a dose-dependent inhibition of XBP-1 splicing induced by Tm/Tg. Overexpression of HRH1 reversed the inhibitory effects of maprotiline (Supplementary Fig. 6f) and amoxapine and desloratadine (Supplementary Fig. 6g) on XBP-1 splicing. Therefore, overexpression of HRH1 significantly blocked the activities of the three ER stress modulators, consistent with their antagonistic effect on this receptor. We next tested whether blocking HRH1 itself has a similar ER stress modulation effect as the three compounds. First, we generated a pair of gRNAs targeting human HRH1 (Supplementary Fig. 6a) and confirmed the HRH1 knockdown (KD) effect of CRISPR in HEK293T cells (Fig. 6b). We then transfected Cas9 and HRH1 gRNAs into the two reporter cell lines and compared Tm/Tg-induced ER stress with or without HRH1 KD. HRH1 inhibition consistently downregulated Tm/Tg-induced CHOP expression (Fig. 6c) and XBP-1 splicing (Supplementary Fig. 6h). To further confirm the HRH1 KD effect on ER stress, we also examined the protein levels of various UPR molecules and again found that HRH1-KD significantly inhibited Tm/Tg-induced ATF4 and CHOP expression (Fig. 6d); JNK phosphorylation and ATF6 expression (Supplementary Fig. 6i); and XBP-1 mRNA splicing (Supplementary Fig. 6j). We previously demonstrated AAV-mSncg promoter mediated Cas9 expression and CRISPR-mediated gene KD in RGCs in vivo. Using the same strategy, we designed gRNAs targeting mouse HRH1 and injected the mixture of AAV-mSncg-Cas9 + AAV-mouse HRH1-gRNAs (Supplementary Fig. 7a) or AAV-control gRNAs intravitreally into mouse eyes (Supplementary Fig. 7b). The endogenous HRH1 mRNA level was detected in some mouse RGCs (Supplementary Fig. 7c). Immunostaining showed HRH1 protein levels to be more extensive in RGCs, and crush injury, but not glaucoma, decreased protein expression in RGCs (Supplementary Fig. 7d, e). AAV-mediated CRISPR KD of HRH1 significantly blocked CHOP and ATF4 expression induced by ONC injury (Fig. 6e, f). Taken together, our studies demonstrated that the three hit compounds inhibit ER stress through their antagonistic effects on their common target, HRH1, suggesting that HRH1 inhibition may provide a potential neuroprotection strategy. We next investigated whether HRH1 inhibition also furnishes neuroprotection in two mouse optic neuropathy models. We injected the mixture of AAV-mSncg-Cas9 + AAV-mouse HRH1-gRNAs or AAV-control gRNAs intravitreally into one of a mouse’s eyes five weeks before ONC (for traumatic ON injury model) or SO intracameral injection (SOHU glaucoma model) and used the contralateral eye as sham control (Fig. 7a). In the ONC model, in vivo OCT imaging showed that the GCC was significantly thicker in HRH1 KD mice than control mice (Fig. 7b, c). Histological analysis of post-mortem retina wholemounts and ON sections consistently demonstrated significant protection of RGC somata and axons by HRH1 KD (Fig. 7d, e). In the SOHU glaucoma model, HRH1 KD also showed a significantly thicker GCC and greater survival of RGC somata and axons than controls (Fig. 7f–i). Importantly, we also confirmed visual function preservation by HRH1 KD, measured by OKR and PERG (Fig. 7j, k). Therefore, like maprotiline treatment, blocking HRH1 significantly protects RGCs and ONs and preserves visual functions in two mouse optic neuropathy models, indicating the promising therapeutic potential of HRH1 inhibition in traumatic and glaucomatous neurodegeneration. Investigation of the long-term safety of maprotiline and HRH1 KD on naïve mouse retinas revealed no RGC or ON degeneration one month after systemic maprotiline administration or three months after local retina AAV-mediated CRISPR HRH1 KD (Supplementary Fig. 8a, b). We also found no immune cell infiltration in retina and ON after either of these treatments (Supplementary Fig. 8c–f). HRH1 is a Gq protein-coupled receptor that can activate phospholipase C (PLC)-IP3 pathway; it leads to ER Ca2+ release through IP3 receptors and cytosol Ca2+ influx (Supplementary Fig. 9a). ON injury is well-known to induce rapid intra-axonal Ca2+ influx that leads to axon degeneration. Ca2+ release from the ER, the major intracellular Ca2+ storage site, contributes to the deleterious intra-axonal Ca2+ influx, and at the same time, the disturbance of ER Ca2+ homeostasis is also an important initiator of the ER stress. We reasoned that maprotiline may block ER Ca2+ release by inhibiting HRH1, and therefore restore ER Ca2+ homeostasis and prevent ER stress. Using pharmacologic small molecule inhibitors of signaling downstream of HRH1, we confirmed that blocking PLC or IP3, but not DAG-PKC, decreased CHOP and XBP-1 activation induced by Tm/Tg (Supplementary Fig. 9b, c), suggesting that HRH1-mediated ER Ca2+ release contributes to ER stress. Moreover, PLC inhibitor U-73122 protected RGC somata and axons in vivo after ONC injury (Supplementary Fig. 9d). Therefore, we investigated the effect of maprotiline on the intracellular and ER Ca2+ levels. First, we transfected genetically encoded Ca2+ sensor jGCaMP7s into HEK293T cells (Fig. 8a), and confirmed that maprotiline significantly blocked Tm/Tg-induced intracellular Ca2+ influx (Fig. 8b, c). Next, we assessed Ca2+ influx in RGCs after ONC in vivo: we confirmed efficient AAV-mediated jGCaMP7s expression in RGCs (Supplementary Fig. 10a); and then recorded in vivo RGC Ca2+ imaging in living animals by scanning laser ophthalmoscope (SLO) at different time points after ONC injury. Within minutes after ONC, intra-RGC Ca2+ levels were significantly elevated, indicating rapid Ca2+ influx induced by axon injury, whereas maprotiline significantly decreased intra-RGC Ca2+ levels at both early time points (Fig. 8d, e) and later time points (Supplementary Fig. 10b). This significant decrease indicates efficient blocking of Ca2+ influx by maprotiline, presumably through inhibition of HRH1-mediated Ca2+ release from the ER. To definitively prove this mechanism, we measured intra-ER Ca2+ levels by expressing a FRET-based ER Ca2+ sensor, D4ER, driven by the mSncg promoter, in mouse RGCs specifically (Supplementary Fig. 10c). We confirmed that ONC significantly depleted ER Ca2+ of RGCs, whereas, in dramatic contrast, maprotiline maintained ER Ca2+ concentration at much higher levels (Fig. 8f, g). Taken together, our data demonstrated that maprotiline blocks ER Ca2+ release through HRH1 inhibition, by which it restores ER homeostasis, prevents deleterious intracellular Ca2+ influx and ultimately protects injured/diseased RGCs and ONs. The present experiments first identified three FDA approved medicines, two antidepressants (amoxapine and maprotiline) and one antihistamine/anti-allergy drug (desloratadine), as potent blockers of ER stress-induced CHOP expression, and then as general modulators of all three UPR pathways and as effective neuroprotectants. Although it is formally possible that these compounds inhibit the PERK-CHOP pathway directly and indirectly affect the other two pathways through cross talk, we favor a model in which they have a global effect inhibiting the UPR by modulating upstream signaling of ER stress. Indeed, this notion receives support from our finding that inhibition of HRH1, a common antagonistic target of all three drugs, also achieved comparable ER stress inhibition and in vivo neuroprotection in both traumatic ON injury and ocular hypertension glaucoma models. It is known that HRH1 activation leads to Ca2+ release from ER and that depletion of ER Ca2+ worsens ER function and induces ER stress. Using both cytosol and ER-targeted Ca2+ biosensors, we found that maprotiline inhibits ER inducer (Tm/Tg)-induced and axon injury-induced ER Ca2+ release and cytosol Ca2+ influx both in cultured cells and mouse RGCs in vivo. The restoration of Ca2+ homeostasis then attenuates the global UPR signaling. We found maprotiline to be the most potent of the three drugs in modulating ER stress based on in vitro cell-based assays and that its systemic administration caused no detectable toxicity on the normal retina, but significantly protected RGCs and ONs and visual functions in mouse disease models of glaucoma and traumatic injury. The potent in vivo neuroprotection of maprotiline correlates with its potent ER stress modulation, further evidence for its on-target mechanism of action. Our studies not only identified potent ER stress modulators and effective neuroprotectants, but also revealed a molecular mechanism that regulates ER stress: maprotiline (possibly amoxapine and desloratadine as well) restores ER homeostasis and keeps the three UPR pathways in check by blocking HRH1-mediated Ca2+ release from the ER. Many small molecule modulators of ER stress that target the signaling molecules in the three main UPR pathways have been developed for different purposes, some of them are neuroprotective but none target intracellular Ca2+ signaling, which is critical for many neurodegenerative diseases associated with axon degeneration. Other HRH1 antagonists may also be able to modulate ER stress and therefore merit further exploration as neuroprotectants. However, some of the other HRH1 antagonists that we tested inhibited ER stress only at high concentration (Fig. 1d, Supplementary Fig. 1e), and overexpression of HRH1 only partially blocked the three hit compounds’ effect (Fig. 6a). These results suggest that other mechanisms in addition to HRH1 may also be responsible for amoxapine/desloratadine/maprotiline-mediated ER stress inhibition, possibly through other receptors that are also modulated by these compounds. Interestingly, maprotiline has previously been shown to be neuroprotective in Huntington’s disease, potentially through mitochondrial protection and anti-apoptotic mechanisms. Our in vivo findings establish maprotiline as a candidate neuroprotectant and HRH1 as a potential therapeutic target for glaucoma, and possibly for neurodegenerative diseases more generally. Because their safety profiles, pharmacokinetics, and pharmacodynamics, including penetration of blood–brain barrier, are well-known, and because of their extensive clinical usage, maprotiline and/or HRH1 inhibition is a promising pharmacological approach for neuroprotection that can be readily translated to pre-clinical studies in large animals and evaluation in human patients. To this end these compounds may be even more attractive if sustained, local delivery to the eye is pursued, thereby minimizing the potential for systemic side effects. In this study, we constructed two ER stress reporter HEK293T cell lines that stably express human CHOP promoter-driven luciferase or fused human XBP-1 fragment-luciferase containing a 26 nt intron sequence that will be removed upon ER stress and IREα activation to allow luciferase expression. Through a small scale HTS with the CHOP-Luc reporter line and further validation with both CHOP-Luc and XBP-1-Luc reporter lines, we demonstrated that this powerful strategy efficiently identifies ER stress modulators. We previously found that the two ER stress molecules, CHOP and XBP-1, play opposing roles in glaucomatous degeneration: deletion of CHOP and activation of XBP-1 protect diseased RGCs and ON synergistically. In this study, the three hit compounds inhibited all three UPR branches, which may not be desirable for neuroprotection because of the inhibition of the IRE1α-XBP1 pathway. A more selective blocker of the PERK-eIF2α-ATF4 pathway, such as ISRIB, may, therefore, be more valuable. Using a murine CHOP promoter-luciferase CHO cell line, another group confirmed that ISRIB inhibits CHOP expression. Unfortunately, ISRIB did not have a significant inhibition effect with our luciferase reporter cell line driven by a human CHOP promoter (Fig. 1d), although it showed significant neuroprotection in the ONC model (Fig. 4), indicating low sensitivity of the reporter line. The much larger murine CHOP promoter (8.5 kb) in the CHO cell line may be more sensitive to compounds or have more cis-regulatory components than the human CHOP promoter (~1 kb) in our HEK cell line, but they do share similar “hits”, including GSK2606414 and trifluoperazine. Cross-checking hit compounds with these two reporter lines will be worthwhile to further confirm CHOP inhibitory effects. Small molecular modulators that inhibit CHOP but activate XBP-1 may furnish even better neuroprotection than those acting by a single mechanism. Encouraged by recent success in identifying preferential activators of IRE1α/XBP-1s and ATF6 with a counter screening strategy, we are currently pursuing complementary cell-based HTS using these two reporter lines with much larger chemical libraries to identify advanced ER stress modulators that inhibit the CHOP pathway but activate the XBP-1 pathway. In summary, we identified three FDA-approved drugs as potent ER stress modulators and effective neuroprotectants through a small scale of HTS, a strategy that warrants further application in identifying additional ER stress modulators. We found that both systemic administration of maprotiline and locally applied CRISPR-mediated RGC-specific HRH1 inhibition achieve significant neuroprotection and visual function recovery in in vivo mouse models of glaucoma and traumatic ON injury. Based on the demonstration of their molecular target and mechanism that we provide in this report and their well-established safety, pharmacological and clinical usage profiles, maprotiline, its structural analogs, and HRH1 antagonists appear to be highly promising candidates for thorough pre-clinical and clinical evaluation as neuroprotectants. C57BL/6J WT (#000664) mice (7–9 weeks old, male) were purchased from Jackson Laboratories (Bar Harbor, Maine) and housed in standard cages on a 12-h light–dark cycle with room temperature at 25 ± 2 °C and humidity between 40 and 60%. All experimental procedures were performed in compliance with animal protocol (#32093) approved by the IACUC at Stanford University School of Medicine. The phCHOP-Luciferase (−954) construct containing human CHOP promoter driven-luciferase was originally made by Dr. Pierre Fafournoux and given by Dr. Shigeru Takahashi. The XBP1-Luciferase construct was from Dr. Albert Koong, containing the luciferase gene fused downstream of an XBP1 fragment containing the 26 nt intron, splicing of the intron by IRE1 under ER stress results in a frameshift and luciferase translation. The coding regions of D4ER (a gift from Dr. Paola Pizzo, Department of Biomedical Sciences, University of Padua, Italy) and jGCaMP7s (Addgene, #104487) were cloned into our pAM-AAV-mSncg-WPRE backbone containing the RGC-specific mSncg promoter. The human HRH1 coding sequence was cloned from HEK293T cell genomic DNA and inserted into a backbone containing the CMV promoter to create a CMV-HRH1 vector. The AAV2-mSncg-Cas9 and the AAV-U6-sgRNAs-Syn-EGFP have been described before. The mouse HRH1 gRNA sequences are: gRNA1 (5′-GCTCCACAACCCTTCCGAGTA-3′) and gRNA2 (5′-GTCCGTCTTCTCCACAACCCT-3′). The human HRH1 gRNA sequences are: gRNA1 (5′-GTCTCCGTCCTCCTTAACCCC-3′) and gRNA2 (5′-GATTCTCCGTCCTCCTTAAC-3′). AAV2 vector was co-transfected with the pHelper plasmid (Stratagene) and pAAV2 (pACG2)-RC triple mutant into HEK293T cells for 72 h before purification with polyethylene glycol and cesium chloride density gradient centrifugation. The AAV titers were determined by real-time PCR and diluted to 1.5 × 1012 vector genome (vg)/ml for mouse intravitreal injection. For intravitreal injection, mice were anesthetized by xylazine and ketamine based on their body weight (0.01 mg xylazine/g + 0.08 mg ketamine/g). A pulled and polished microcapillary needle was inserted into the peripheral retina just behind the ora serrata. Approximately 2 µl of the vitreous was removed to allow injection of 2 µl AAV into the vitreous chamber to achieve 3 × 109 vg/retina. HEK293T cells were transiently co-transfected with phCHOP-Luciferase or XBP-1-Luciferase with pEGFP-puro using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) at ratio 5:1 to ensure EGFP positive cells are also luciferase construct positive and puromycin resistant. After a serial selection with puromycin and EGFP expression, multiple stably expressing clones (CHOP-Luc/puro or XBP-1-Luc/puro) were isolated by a serial dilution. After expansion of individual clones, one CHOP-luciferase stable line and one XBP-1-luciferase line were selected based on their responses to Tm/Tg treatment, and maintained by puromycin as stable reporter cell lines used in this study. HEK293T cells were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) (Invitrogen, 11995081) supplemented with 100 μg/mL streptomycin and 100 units/mL penicillin (Gibco, 15140122), and 10% fetal bovine serum (FBS) (Invitrogen, 10437-028). Cells were maintained under standard tissue culture conditions (5% CO2, 37 °C). PolyJet™ (SignaGen Laboratories, SL100688) transfection reagent was used for transient cell transfection. For detection of relative cellular viability levels, cells (10000/well) were seeded into poly-d-lysine coated 96-well plates (Falcon, 353072) and treated as described. Then, MTT (5 mg/mL, 10 μL/well) (MedChemExpress, HY-15924) was added. After incubation at 37 °C for 3–4 h, DMSO was added to dissolve the precipitate, and the absorbances were determined at 570 nm by a Tecan Infinite M1000 Pro plate reader. The HTS to identify CHOP expression inhibitors was performed at the Stanford High-Throughput Bioscience Center with small molecule libraries containing 4846 total known bioactive, FDA approved drugs and clinical trial compounds; libraries included Biomol FDA, Biomol ICCB, Microsource (MS) Spectrum, Sigma LOPAC, and NIH Clinical Collection (NIH-CC). All the compounds were run in 7-point dose response based on a previous publication; except the NIH Clinical collection (446 compounds) which were tested in duplicate due to the small quantity of compounds on hand. Most of the NIH-CC was screened at 10 µM, but this varied per compound. The Z’ of the assay was 0.5, details on instrumentation can be found here: https://med.stanford.edu/htbc/equipment/liquid.html. We used the Pin Tool to transfer 100 nL of the compounds into 50 µL final volume. Therefore, we did a 500-fold dilution of the stock plates, which were at 10, 5, 2.5, 1.25, 0.625, 0.3125, and 0.156 mM resulting in final concentrations of 20, 10, 5, 2.5, 1.25, 0.625, and 0.3125 µM (for most of the compounds but not all). The final DMSO concentration was 0.2% for all wells. Briefly, CHOP-Luc cells were seeded in 384-well plates (Greiner Bio-One CELLSTAR™) at a density of 2 × 105 cells per well in 40 µL medium and cultured for 24 h before treatment with tunicamycin (Tm) + thapsigargin (Tg) at 1 µM in 10 µL medium followed by adding 100 nL of one of the testing compounds using a Staccato SciClone ALH3000 small molecule liquid handling system (Caliper Life Sciences) and V&P Scientific 384 pin tools. AeraSeal sterile adhesive microplate seals (Excel Scientific; Victorville, CA) were used to seal plates that were incubated at 5% CO2, 37 °C for 24 h, when luciferase activity was assayed by adding 10 μL of BrightTM-Glo Luciferase reagent (Promega) to each well and detected by a Tecan Infinite M1000 Pro plate reader. Percentage of CHOP-Luc inhibition = 100%−(Compound value−DMSO control value)/(Tm/Tg value−DMSO control value) × 100%. The compounds achieving >30% of CHOP-Luc inhibition were considered as “hits”. Luciferase activity of all the tested compounds was determined by meta-analysis of luciferase activity in different cell lines under the control of a generic or other promoters. Any compound that appeared in more than 3 non-related luciferase screens was likely a toxic compound or luciferase inhibitor and therefore eliminated. There was no true cutoff, other than an IC50 < 20 µM (or the highest concentration tested). Most of these compounds were toxic, but we did not specifically examine toxicity because our goal here was to eliminate the non-specific hits regardless of whether they were luciferase inhibitors or toxic compounds. Thapsigargin (Sigma, T9033), Tunicamycin (Sigma, T7765), Maprotiline (Sigma, M9651), Amoxapine (MedChemExpress, HY-B0991), Desloratadine (MedChemExpress, HY-B0539), Desipramine (Sigma, D3900), Trifluoperazine (Sigma, T8516), Clomipramine (Sigma, C7291), Amitriptyline (Sigma, A8404), Quetiapine (Sigma, Q3638), Olanzapine (MedChemExpress, HY-14541), Doxepin (MedChemExpress, B078), Loxapine (Sigma, L106), Norquetiapine (Sigma, 07849), dimethyl sulfoxide (DMSO) (Sigma, D8418), 2-APB (MedChemExpress, HYW009724), U-73122 (MedChemExpress, HY13419), Go 6983 (MedChemExpress, HY13689), GSK2606414 (Sigma, 516535), ISRIB (MedChemExpress, HY-12495). Cell and tissue lysates were prepared in RIPA buffer (Themo Fisher Scientific, 89901) and supplemented with Halt Protease inhibitor cocktail (Themo Fisher Scientific, PI78437). The total protein concentration of lysates was measured by the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 23227). 20 µg protein of each sample were denatured at 95 °C for 15 min in 100 mM DTT + 1× Laemmli buffer before being separated by SDS-PAGE. We then transferred the protein samples to 0.2 µm nitrocellulose membranes (Bio-Rad, 1610097) and blocked with 5% BSA for 2 h. Subsequently, the membranes were incubated with primary antibodies (1:1000) overnight at 4 °C. After washing in TBST, these membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling, 7074S and 7076S, 1:1000) and visualized using a GE-AI600 imaging system. Images were quantified with ImageJ software. The primary antibodies used were PERK (Cell Signaling, 3192S), p-PERK (Cell Signaling, 3179 S), eIF2a (Cell Signaling, 5324S), p-eIF2a (Cell Signaling, 3597L), ATF4 (Cell Signaling, 11815S), CHOP (Cell Signaling, 2895S), ATF6 (Cell Signaling, 65880T), IRE1a (Cell Signaling, 3294T), IRE1 (phosphor-S724) (Thermo Fisher, PA116927), phospho-SAPK/JNK (Thr183/Tyr185) (Cell Signaling, 9251S), phospho-p38 MAPK (Thr180/Tyr182) (Cell Signaling, 9215S), β-Actin (sigma, A5441), XBP-1s (Biolegend, 647502), XBP-1 (Santa Cruz, sc-8501), anti-RBPMS (Custom made at ProSci Inc). Cells were plated into poly-d-lysine coated 6-well plates (Fisher, 353046) and treated as indicated at 37 °C with 5% CO2. The total RNA was isolated using Trizol (Thermo Fisher Scientific, 10296010) and 500 ng RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, 4374966) to acquire total cDNA. The XBP-1 mRNA splicing primers (forward primer 5′-GGGGCTTGGTATATATGTGG-3′, reverse primer 5′-CCTTGTAGTTGAGAACCAGG-3′) were utilized to amplify the XBP-1 amplicon containing the 26 nt intron that will be released by IRE1α upon ER stress, with Q5 High-Fidelity DNA Polymerase (NCB, M0491L). PCR products (2 µg) were resolved on 2.5% agarose gels, visualized with GelRed (Biotium), and quantified by ImageJ (NIH). The relative mRNA expression levels of target genes were detected by PowerUP SYBR Green Master Mix (Thermo Fisher, A25776) and C1000 TOUCH CYCLER w/48 W FS RM PCR System (Bio-rad, 1851148). Thermal cycles were 95 °C for an initial 5 min followed by 40 cycles of denaturation at 95 °C for 15 s, annealing at 60 °C for 30 s and extension at 72 °C for 60 s. Transcripts were normalized to GAPDH and all measurements were performed in triplicate. Primers used were: GAPDH, forward 5′-GTCTCCTCTGACTTCAACAGCG-3′ and reverse 5′-ACCACCCTGTTGCTGTAGCCAA-3′; ATF6, forward 5′ TGGAAGCAGCAAATGAGACG-3′ and reverse 5′-TGAGGAGGCTGGAGAAAGTG-3′; ERN1, forward 5′-CGAACGTGATCCGCTACTTC-3′ and reverse 5′-ATGTTGAGGGAGTGGAGGTG-3′; EIF2AK3, forward 5′-GTCCCAAGGCTTTGGAATCTGTC-3′ and reverse 5′-CCTACCAAGACAGGAGTTCTGG-3′; CHOP, forward 5′-ACCAAGGGAGAACCAGGAAACG-3′ and reverse 5′-TCACCATTCGGTCAATCAGAGC-3′; ATF4, forward 5′-GTCCCTCCAACAACAGCAAG-3′ and reverse 5′-TGTCATCCAACGTGGTCAGA-3′; DNAJC3, forward 5′-GCCTGCATCTGCTTTATGCT-3′ and reverse 5′-TCTGCAAGGCTGTGAAGAGA-3′; GADD45a, forward 5′-GGAGGAAGTGCTCAGCAAAG-3′ and reverse 5′-ACATCTCTGTCGTCGTCCTC-3′; DNAJB9, forward 5′-GGAAGGAGGAGCGCTAGGTC-3′ and reverse 5′-ATCCTGCACCCTCCGACTAC-3′; Calreticulin, forward 5′-CGATGATCCCACAGACTCCA-3′ and reverse 5′-CCGTCCATCTCTTCATCCCA-3′; Bip, forward 5′-GCCTGTATTTCTAGACCTGCC-3′ and reverse 5′-TTCATCTTGCCAGCCAGTTG-3′; XBP-1s, forward 5′-CTCCAGAGACGGAGTCCAAG-3′ and reverse 5′-CACCTGCTGCGGACTC-3′. The ON was exposed intraorbitally while care was taken not to damage the underlying ophthalmic artery, and crushed with a jeweler’s forceps (Dumont #5; Fine Science Tools, Foster City, California) for 5 s approximately 0.5 mm behind the eyeball. Eye ointment containing neomycin (Akorn, Somerset, New Jersey) was applied to protect the cornea after surgery. For compound treatment, each eye received intravitreal injection with 2 µl of 2 mM test compounds once and intraperitoneal (i.p.) injection daily (15 mg/kg) for 14 days after ONC. Control groups received the same volume of DMSO as vehicle control. Mice were anesthetized by an intraperitoneal injection of Avertin (0.3 mg/g) and received the SO (Alcon Laboratories, 1000 mPa.s) injection at 9–10 weeks of age. Prior to injection, one drop of 0.5% proparacaine hydrochloride (Akorn, Somerset, New Jersey) was applied to the cornea to reduce its sensitivity during the procedure. A 32 G needle was tunneled through the layers of the cornea at the superotemporal side close to the limbus to reach the anterior chamber without injuring lens or iris. Following this entry, ~2 µl silicone oil (1000 mPa.s, Silikon, Alcon Laboratories, Fort Worth, Texas) was injected slowly into the anterior chamber using a homemade sterile glass micropipette, until the oil droplet expanded to cover most areas of the iris (diameter ~1.8–2.2 mm). After the injection, veterinary antibiotic ointment (BNP ophthalmic ointment, Vetropolycin, Dechra, Overland Park, Kansas) was applied to the surface of the injected eye. The contralateral control eyes received mock injection with 2 µl normal saline to the anterior chamber. Throughout the procedure, artificial tears (Systane Ultra Lubricant Eye Drops, Alcon Laboratories, Fort Worth, Texas) were applied to keep the cornea moist. For compound treatment, each eye received retrobulbar injection twice on day 0 and day 10 with 50 µl of 2 mM maprotiline, and daily i.p. injection with maprotiline (15 mg/kg) for 3 weeks after SO injection. Control groups received the same volume of DMSO as vehicle control. The IOP of both eyes was measured by the TonoLab tonometer (Colonial Medical Supply, Espoo, Finland) according to product instructions under a sustained flow of isoflurane (3% isoflurane at 2 L/min mixed with oxygen) delivered to the nose by a special rodent nose cone (Xenotec, Inc., Rolla, Missouri). 1% Tropicamide Sterile Ophthalmic Solution (Akorn, Somerset, New Jersey) was applied three times at 3-min intervals to fully dilate the pupils (about 10 min) before taking measurements. During this procedure, artificial tears were applied to keep the cornea moist. Since IOP measurement requires pupil dilation, which essentially relieves the ocular hypertension during the period of pupil dilation, we only measure IOP 3 weeks after SO injection immediately before sacrificing the animals in the ND (no dilation) SOHU model that we described before. Retinal tissues were homogenized with 100 µL of pre-chilled 20% acetonitrile and then diluted 2-fold with blank mouse plasma. An aliquot of 20 µL of diluted retina homogenate was extracted with 100 µL of methanol:acetonitrile (5:95, v-v) containing the internal standard (Verapamil). The mixture was shaken on a shaker for 15 min and then centrifuged at 3220 × g for 15 min. An aliquot of 70 µL of the supernatant was mixed with 70 µL of water for the injection to the LC-MS. Calibration standards and quality control samples were prepared by spiking 2 µL of the test compound into 18 µL of blank mouse plasma, and the resulting plasma was processed with the unknown samples in the same batch. The extracts were analyzed by a Shimadzu LC-30AD interfaced to a Sciex API 5000 system. The extracts were injected onto an ACE 3 C18 column (50 × 2.1 mm, 3.0 µm) and separated by the gradient elution using water with 10 mM ammonium acetate (A) and acetonitrile with 0.1% formic acid (B) as mobile phases. The gradient program started at 10% B, held for 0.2 min, ramped to 95% B at 1.5 min, remained at 95% at 2.4 min, dropped to 10% B at 2.45 min, and stayed at 10% B till 3.2 min. The mass spectrometer was operated in positive electrospray ionization under the multiple reaction monitoring (MRM) mode for the detection of the maprotiline (278.277-to-250.2 m/z) and the internal standard (455.346-to-165 m/z). The calibration curve fitted by linear regression was used to quantify the analytes in the matrix using Analyst software 1.6.2 (Sciex). After perfusion fixation with 4% PFA in PBS, mice eyeballs and ONs were dissected out and post-fixed with 4% PFA for 2 h at room temperature. Retinas were dissected out for whole-mount retina immunostaining. For cryo-section with Leica cryostat, the eyeballs and ONs were embedded in tissue-tek OCT (Sakura) on dry ice for subsequent cryo-section. The sections were blocked with 10% goat serum (Sigma, G9023) for 2 h before incubating with primary antibodies: RBPMS 1:4000, others 1:200 overnight at 4 °C. After washing 3 times with PBS, samples were incubated with secondary antibodies (1:400; Jackson ImmunoResearch, West Grove, Pennsylvania) at room temperature for 2 h. Tissues were washed with PBS 3 times before mounting with Fluoromount-G (SouthernBiotech, Alabama). Confocal images were obtained by a Zeiss LSM 800 microscope (Carl Zeiss Microscopy). HEK293T cells were seeded at 10% confluence the day before transfection in poly-D-lysin-coated glass bottom 35 mm dishes (MatTek, P35GC1.510C). Two hours prior to transfection, media was removed and 1 mL of FBS-free DMEM media was added to each well. We then mixed 2 µg of the jGCaMP7s plasmid in 100 μl DMEM. In a separate tube, we diluted 4 μl PolyJet™ (SignaGen Laboratories, SL100688) reagent in 100 μl with DMEM, and incubate for 10 min. The complete mixture of DNA and PolyJet was incubated for another 15 min before being added to the HEK293T cells. Six hours later, the media was changed to 1 ml/well culture media. The next day, cytosolic Ca2+ was imaged in cells transiently transfected with jGCaMP7s through a Zeiss LSM 800 confocal microscope. Dynamic intracellular Ca2+ influx in response to Tm/Tg-induced ER stress was measured every 10 s for 100 s before Tm/Tg treatment to acquire baseline fluorescence (F0) and for 400 s after Tm/Tg administration to acquire F1. ΔF = F1−F0. Data were analyzed with ImageJ. The mice were intravitreally injected with AAV2-mSncg-jGCaMP7s (9 × 109 vg/retina) 4 weeks before imaging. The mice were anesthetized by xylazine and ketamine after dark adaptation for 30 min. Mydriasis was achieved by applying a drop of 1% tropicamide solution and a drop of 2.5% phenylephrine hydrochloride solution, which prevents pupillary contraction during recording. Right after ON crush, the mice were placed on a 3D-printed mouse holder with a 37 °C heater, and a custom-made +10D mouse contact lens (3.0 mm diameter, 1.6 mm BC, PMMA clear, Advanced Vision Technologies) attached to keep the cornea from drying. The retinal fundus was imaged by the Heidelberg Spectralis SLO/OCT system (Heidelberg Engineering, Germany) with a 55° lens using the fluorescein angiography scanning mode under the same sensitivity (sensitivity 75–85) and high-resolution (1536 × 1536 pixels). ER calcium levels were measured using the Förster resonance energy transfer (FRET)-based ER targeted calcium sensor, D4ER. Mouse received intravitreal injection of AAV2-mSncg-D4ER to express D4ER in RGCs in vivo 4 weeks before ON crush injury, as well as intravitreal injection of DMSO (vehicle) or maprotiline compound. For imaging, retinas were dissected out at 3, 7 or 14dpc and plated onto laminin (Sigma, L2020) and poly-D-lysin-coated glass bottom dishes (MatTek, P35GC1.510C) and maintained in Neurobasal-A medium (ThermoFisher Scientific, 10888022) supplemented with L-glutamine (Gibco, 25030-081), penicillin/streptomycin (Gibco, 15140122) and B-27 (ThermoFisher Scientific, 0080085SA). Retina explants were imaged using a Zeiss LSM 800 microscope. ER[Ca2+] levels were determined by exciting D4ER at 440 nm to record the emitted light at 465–485 nm and 530–550 nm, and analyzed with ImageJ. Whole-mount retinas were immunostained with the RBPMS antibody, 6–8 fields randomly sampled from peripheral regions of each retina using a 40X lens with a Zeiss M2 epifluorescence microscope, and RBPMS + RGCs counted by Volocity software (Quorum Technologies). The percentage of RGC survival was calculated as the ratio of surviving RGC numbers in injured eyes compared to contralateral uninjured eyes. The investigators who counted the cells were masked to the treatment of the samples. ONs were post-fixed in situ with 2% glutaraldehyde and 2% PFA in 0.1 M PBS. Semi-thin (1 µm) cross sections of the ON 2 mm distal to the eye (globe) were collected. The sections were stained with 1% PPD for 0.5 h before washing with methanol: isopropanol (1:1) 3 times × 10 min and then mounted with Fluoromount-G. Four sections of each ON were imaged through a 100× lens of a Zeiss M2 epifluorescence microscope to cover the entire area of the ON without overlap. Two areas of 21.4 µm × 29.1 µm were cropped from the center of each image, and the surviving axons within the designated areas counted manually using ImageJ. After counting all the images taken from a single nerve, the mean of the surviving axon number was calculated for each ON. The mean of the surviving axon number in the injured ON was compared to that in the contralateral control ON to yield a percentage of axon survival value. 70 nm ultrathin sections were collected onto formvar-coated copper grids and dried overnight. Sections were then stained with uranyl acetate for 30 min, washed in PBS, and then stained with lead citrate for 7 min. Sections were again washed and dried before observing under TEM. The cross-sections of the entire ON were examined and imaged randomly without overlap at 4000× with 11.6 μm × 11.6 μm frames on a JEOL JEM-1400 TEM microscope (JEOL USA, Inc., Peabody, MA). For each ON, 25–45 images were taken to cover the whole area of the ON. Axons were counted manually with ImageJ’s Cell Counter plugin. Intravitreal injection of CTB was performed 48 h before perfusion of the animals with 4% PFA in PBS. The ONs were carefully dissected with fine forceps and scissors and cleared with a modified iDISCO method: wash with PBS for 4 × 30 min; then immersed in a series of 20%, 40%, 60%, 80%, and 100% methanol in PBS for 30 min at each concentration; dichloromethane (DCM)/methanol (2:1) for 30 min; 100% DCM for 30 min and dibenzyl ether (DBE) for 10 min before mounting on slides. Tiled images of the wholemount ON were captured and stitched by a Zeiss LSM 880 confocal laser scanning microscope with 40x/1.0 Oil DIC (Carl Zeiss Microscopy, Thornwood, NY, USA). Positive CTB areas were identified based on a fluorescence intensity greater than the baseline intensity threshold. The percentage of the CTB positive area in the optic nerve was measured by NIH ImageJ. Fluorescent in situ hybridization (FISH) was performed by using the RNAscope Multiplex Fluorescent Detection Reagents V2 (Advanced Cell Diagnostics, ACD, Hayward, CA, USA) according to the manufacturer’s instructions. RNAscope probe Mm-Hrh1 (491141) was purchased from ACD. Adult mice were perfused with ice-cold 4% PFA/PBS, and eyes were dissected out and fixed in 4% PFA/PBS at 4 °C overnight. The eyes were dehydrated with increasing concentrations of sucrose solution (10%, 20 and 30%) overnight before embedding in OCT on dry ice. Serial cross sections (12 µm) were cut with a Leica cryostat and collected on Superfrost Plus Slides. The sections were pretreated with protease and then subjected to in situ hybridization with RNAscope Multiplex Fluorescent Detection Reagents V2 according to the manufacturer’s instruction (Advanced Cell Diagnostics, Hayward, CA). Briefly, sections were hybridized with the probe solution, followed by amplification and probe detection using TSA plus fluorophores (AKOYA, Marlborough, MA, USA). The sections were mounted with Fluoromount-G (SouthernBiotech, Birmingham, AL, USA). Images were captured by a Zeiss LSM 880 confocal laser scanning microscope with 40×/1.0 Oil DIC (Carl Zeiss Microscopy, Thornwood, NY, USA). The mouse retina was scanned by the Heidelberg Spectralis SLO/OCT system (Heidelberg Engineering, Germany) with the ring scan mode centered by the ON head under high-resolution mode (each B-scan consisted of 1536A scans). The ganglion cell complex (GCC) includes retinal nerve fiber layer (RNFL), ganglion cell layer (GCL) and inner plexiform layer (IPL). The average thickness of GCC around the ON head was measured manually with the aid of Heidelberg software. After anesthetization and pupil dilation, PERG of both eyes was recorded simultaneously with the Miami PERG system (Intelligent Hearing Systems, Miami, Florida) according to manufacturer’s instructions. Two consecutive recordings of 200 traces were averaged to achieve one readout; each trace recorded up to 1020 ms. The first positive peak in the waveform was designated as P1 and the second negative peak as N2. The amplitude was measured from P1 to N2. Mice were placed on a platform in the center of four 17-inch LCD computer monitors (Dell, Phoenix, AZ), with a video camera above the platform to capture the movement of the mouse. A rotating cylinder with vertical sine wave grating was computed and projected to the four monitors by OptoMotry software (Cere- bralMechanics Inc, Lethbridge, Alberta, Canada). The sine wave grating, settled at 100% contrast and speed of 12 degrees per second, provides a virtual-reality environment to measure the spatial acuity (cycle/degree) of the left eye when rotated clockwise and the right eye when rotated counterclockwise. The maximum frequency (cycle/degree) that the mouse could track was identified and recorded by investigators masked to treatment. The relative percentages of visual acuity were calculated as the ratio of maximum frequency in disease eye compared to contralateral control eye. GraphPad Prism 7 was used to generate graphs and for statistical analyses. Data are presented as means ± s.e.m. Student’s t-test was used for two groups comparison and One-way ANOVA with post hoc test was used for multiple comparisons. Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary Information Reporting Summary

PMC9649816

Ji Zhang,Yujing Feng,Shengqiang Han,Xueting Guan,Ziliang He,Chao Song,Lingyun Lv,Qiaoyu Luo

Incarvillea compacta Maxim ameliorates inflammatory response via inhibiting PI3K/AKT pathway and NLRP3 activation 10.3389/fphar.2022.1058012

28-10-2022

Incarvillea compacta Maxim,anti-inflammatory,RNA sequencing,PI3K/Akt signaling pathway,acute gastritis

Incarvillea compacta Maxim is a traditional Tibetan medicine used to treat inflammation-related diseases, such as pneumonia, fever, jaundice, and otitis media. However, no studies have examined its anti-inflammatory mechanism. To validate the anti-inflammatory activity of I. compacta extract (ICE) and its protective effect on acute alcoholic gastritis, Phytochemicals of I. compacta were identified using Ultra-performance liquid chromatography quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS). Lipopolysaccharide (LPS)-induced RAW 264.7 macrophages were used in vitro along with an in vivo a mouse acute gastritis model. Pro-inflammatory mediators and cytokines were measured using the Griess reagent and Cytometric bead array (CBA) assay. Furthermore, inflammation-related molecules were analysed by Western blotting, RNA-Seq, and real-time quantitative PCR (RT-qPCR). The experimental results revealed that ICE decreased the nitric oxide (NO), IL-6, MCP-1, and TNF-α levels in LPS-stimulated RAW 264.7 cells, and downregulated the expression and phosphorylation of PDK1, AKT, and GSK3β. Moreover, ICE also downregulated the activation of NLRP3. The RNA-Seq analysis revealed that 340 differentially expressed genes (DEGs) response to ICE treatment was enriched in several inflammation-related biological processes. The results of the in vivo mouse acute gastritis model showed that ICE significantly reduced inflammatory lesions in the gastric mucosa and remarkably downregulated the expression of iNOS, TNF-α, IL-1β, and IL-6 mRNA in gastric tissue. Therefore, the results of this study obtained scientific evidence supporting the use of I. compacta.

Incarvillea compacta Maxim ameliorates inflammatory response via inhibiting PI3K/AKT pathway and NLRP3 activation 10.3389/fphar.2022.1058012 Incarvillea compacta Maxim is a traditional Tibetan medicine used to treat inflammation-related diseases, such as pneumonia, fever, jaundice, and otitis media. However, no studies have examined its anti-inflammatory mechanism. To validate the anti-inflammatory activity of I. compacta extract (ICE) and its protective effect on acute alcoholic gastritis, Phytochemicals of I. compacta were identified using Ultra-performance liquid chromatography quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS). Lipopolysaccharide (LPS)-induced RAW 264.7 macrophages were used in vitro along with an in vivo a mouse acute gastritis model. Pro-inflammatory mediators and cytokines were measured using the Griess reagent and Cytometric bead array (CBA) assay. Furthermore, inflammation-related molecules were analysed by Western blotting, RNA-Seq, and real-time quantitative PCR (RT-qPCR). The experimental results revealed that ICE decreased the nitric oxide (NO), IL-6, MCP-1, and TNF-α levels in LPS-stimulated RAW 264.7 cells, and downregulated the expression and phosphorylation of PDK1, AKT, and GSK3β. Moreover, ICE also downregulated the activation of NLRP3. The RNA-Seq analysis revealed that 340 differentially expressed genes (DEGs) response to ICE treatment was enriched in several inflammation-related biological processes. The results of the in vivo mouse acute gastritis model showed that ICE significantly reduced inflammatory lesions in the gastric mucosa and remarkably downregulated the expression of iNOS, TNF-α, IL-1β, and IL-6 mRNA in gastric tissue. Therefore, the results of this study obtained scientific evidence supporting the use of I. compacta. Inflammation is a defensive response that protects the body against infection and injury. The inflammatory response can be localised to eliminate the injurious agent and remove damaged tissues to restore health (Chen et al., 2018). Infectious and non-infectious factors that cause injury can result in inflammation (Molteni et al., 2016). In acute inflammation, the inflammatory response may last only for a few days, although it can last much longer and develop into chronic inflammation (Varela et al., 2018). Inflammation is not always beneficial, it often causes discomfort, such as pain or itching, and in some cases can even cause severe disease, including hypersensitivity and autoimmune disease (Straub and Schradin, 2016; Montero-Melendez, 2018). Many cytokines are involved in the inflammatory process that begins when macrophages are triggered by bacterial Lipopolysaccharide (LPS) (Abdulkhaleq et al., 2018). When cytokines bind to their receptors in signal transduction pathways, they alter the receptor conformation, resulting in the phosphorylation of receptors or receptor-related kinases, and then activate a variety of phosphorylation transcription factors (Hughes and Nibbs, 2018). The intracellular signal transduction cascade pathway is also activated. This alters the amounts of substances involved in mediation of the inflammatory response. Intracellular signal transduction allows cells to respond to external stimuli via cell membrane or intracellular receptors, and triggers specific biological effects (Romani et al., 2021). Toll-like receptors (TLRs) are important transmembrane proteins in mammals that transmit extracellular antigen recognition information to cells and trigger inflammatory responses (Sameer and Nissar, 2021). TLRs are the main factors mediating immune and inflammatory responses. TLR4 is activated by agonists and initiates downstream signal transduction to prompt a series of reactions via MyD88-dependent and -independent TRIF pathways (Yanagibashi et al., 2015). The LPS receptor is a protein complex composed of LPS-binding protein, CD14, TLR, and MD-2. These receptors can form a high-affinity complex, tightly bind with LPS, activate cells, and release proinflammatory cytokines, growth factors, and enzymes (Latz et al., 2003). LPS mainly activates NF-κB, TBK1-IRF3, MAPKs (including ERK, JNK and p38), JAK-STAT1, and various other signal transduction pathways to trigger the expression of inducible nitric oxide synthase (iNOS) (Pålsson-McDermott and O’Neill, 2004). Therefore, identifying natural products that inhibit nitric oxide (NO) production has become a major goal to allow the development of anti-inflammatory agents. Incarvillea compacta Maxim is a perennial herb in the family Bignoniaceae that mainly occurs in Gansu, Qinghai, Yunnan, and Tibet of China (Zhao et al., 2017). As a traditional Tibetan medicine, it is widely used to treat jaundice, stomach ache, and otitis media (Zhang et al., 2016). Two of the earliest Tibetan books, the “Tibetan Medicine Record” and “Jing Zhu Ben Cao”, mentioned the ethnopharmacological uses of this plant (Pengcuo, 1986; Yang, 1991). Among the many pharmacological activities of I. compacta, its anti-inflammatory effects have attracted much attention, although the underlying mechanism is unclear. Zhao et al. investigated the chemical constituents of I. compacta, identifying 23 compounds, including phenylpropanoid glycosides, flavonoids, iridoid glycosides, triterpenes, and steroids, among others (Zhao et al., 2017). Wang et al. isolated and identified 10 compounds from the ethyl acetate fraction of a 95% ethanol extract of I. compacta: methyl linoleate, tricin, trihydroxy-7-megastigmen-9-one, syringaresinol, salcolin A, rhamnazin, dibutyl-phthalate, trimethyl ellagic acid, kaempferol rhamnopyranoside, and tricin glucopyranoside (Wang et al., 2019). In our laboratory, 12 phenylethanoid glycosides were isolated from the roots of I. compacta, among which Z-3‴-O-methylisocrenatoside and 7-O-metylleucoseceptoside were novel (Shen et al., 2015; Wu et al., 2016). This work examined the anti-inflammatory effects of an I. compacta extract (ICE) on RAW264.7 cells and in a mouse acute gastritis model. The differentially expressed genes (DEGs) of RAW264.7 cells in response to ICE treatment were also analysed at the transcriptome scale using RNA-Seq. The results improve our understanding of the mechanisms underlying the anti-inflammatory activity of ICE. I. compacta maxim were purchased from Xining medicinal materials market (Qinghai China) and authenticated by Professor Haifeng Wu, Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences. Voucher specimens (ICM-2019) was deposited at School of Life Sciences, Huaiyin Normal University. 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was purchased from BioFroxx (Einhausen, Germany). LPS was obtained from Sigma (St. Louis, MO, United States). Roswell Park Memorial Institute (RPMI) 1,640 was bought from Invitrogen-Gibco (Beijing China). Fetal bovine serum (FBS) was purchased from Corning (New Zealand). Penicillin/streptomycin was obtained from Invitrogen-Gibco (Carlsbad, CA, United States). L-NG-monomethyl-arginine (L-NMMA) was obtained from Beyotime Biotechnology (Nantong, China). Cytometric bead array (CBA) kit was purchased from BD Biosciences (San Diego, CA, United States). Trizol reagent was obtained from Ambion (Waltham, MA, United States). Rabbit monoclonal antibodies against p-PDK1, PDK1, p-AKT, AKT, p-GSK3β, GSK3β and NLRP3 were obtained from Cell Signaling Technology (Danvers, MA, United States). The goat anti-rabbit IgG H&L (HRP) and β-actin were purchased from Abcam (Cambridge, UK). The dried plant powder (50 g) was extracted three times with 70% ethanol (500 ml) at 90°C for 2 h. The crude extract (ICE) was obtained by removal of the solvent under a rotary evaporator (RV 10, IKA, Germany). The phytochemical profile of ICE was carried out using a Waters Acquity UPLC BEH C18 column (2.1 mm × 100 mm, 1.7 µm). The mobile phase consists of 0.1% formic acid (A) and acetonitrile (B) with a linear gradient elution as follows: 2–40% B from 0 to 5 min; 40–100% B from 5 to 15 min. The flow rate of mobile phase was 0.3 ml/min and detection wavelength was set at 260 nm. The sample inject volume was 2.0 μl. The Ultra-performance liquid chromatography quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS) included a SYNAPT G2-Si HRMS instrument (Waters, Milford, MA, United States) equipped with an electrospray ionization (ESI) interface. The data were collected in both positive and negative ion modes and the full scan range was set between m/z 50 and 1,500. The source temperature was 120°C, and the desolvation gas temperature was 350°C. The cone and solvent removal gas was nitrogen gas and the flow rates of cone and desolvation gas were set at 50 and 600 L/h, respectively. The capillary voltage was set at 2500 V, while the cone voltage was set at 50 V. Leu-enkephalin was used as a reference mass. All data collected were acquired using MassLynx 4.2 software. The RAW264.7 cells were obtained from American Type Culture Collection (Manassas, VA, United States) and maintained in RPMI 1640 medium supplemented with 10% FBS and 1% antibiotics solution at 37°C under 5% CO2. In briefly, RAW264.7 cells were seeded on 96 well plates at a density of 1 × 105/well and incubated overnight. Cells were pre-treated with different concentration of ICE (0, 12.5, 25, 50, 100 μg/ml) or 100 μg/ml of L-NMMA for 30 min and stimulated by LPS for 24 h. NO content was measured using Griess reagent (Li F. et al., 2019). Cytokines of TNF-α, IL-6, and MCP-1 were detected by CBA according to the manufactures protocol. RAW264.7 cells were seeded on 6-well plates at a density of 5×106/well and incubated overnight. Cells were pre-treated with different concentration of ICE for 30 min and stimulated by LPS for 24 h. RNA were extracted with Trizol reagent kit according to the manufacture’s protocol, and were sequenced on a HiSeq 2,500 sequencing platform (Illumina, San Diego, CA, United States). The raw sequencing data were filtered by remove the raw reads that containing adapters or low quality bases, then the obtained clean reads were further mapped and assembled. Expression abundance and variations of each transcription region were quantified by calculate the Fragment per kilobase of transcript per million mapped reads (FPKM) value. DEGs were defined based on the mRNAs differential expression between control and ICE treated groups. False Discovery Rate (FDR) ≤ 0.05 and absolute fold change ≥2 were the significance threshold for DEGs selecting. To further understand the biological functions of selected DEGs, bioinformatic analyses were performed on Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. Significantly enriched GO terms and KEGG pathways were defined when the calculated p-value gone through FDR correction (FDR ≤0.05). The RNA sequencing data were validated by real time quantitative RT-PCR analyses using Luna universal qPCR kit according to the manufactures protocol on a CFX-96™ Real-Time instrument (Bio-Rad, Hercules, CA, United States). 16 DEGs were selected for real-time quantitative PCR (RT-qPCR) analysis with the GAPDH gene as a house keeping gene to normalize the expression levels of test genes. Primers used in this work were synthesized by Sangon Biotech (Shanghai, China) and were listed in Table 1. The relative gene expression levels were calculated using 2−ΔΔCT method. All of the samples were triplicated. The data generated in this study was uploaded to the China National GeneBank DataBase (CNGBdb) with accession number CNP0003554. RAW264.7 cells were seeded on 6-well plates at a density of 5×106/well and incubated overnight. Cells were pre-treated with different concentration of ICE for 30 min and stimulated by LPS for different time points. Cells were then rinsed with PBS and the total protein prepared using RIPA kit (CoWin Biosciences, Beijing, China). Western blot analysis was carried out as described previously (Zhou et al., 2021). The 6 week old ICR mice were purchased from SPF Biotechnology Co. Ltd (Beijing, China) and were housed in standard mouse cages at 20°C with a 12 h light/dark cycle. Food and water were freely provided. 40 ICR mice were randomly divided into five groups with eight mice in each group and were named as Normal Control (NC) group, Model Control (MC) group, Low-Dose (LD) group, High-Dose (HD) group and Positive Control (PC) group, respectively. Mice in NC group and MC group were intragastrical administered with 0.5% CMC-Na, LD and HD group were intragastrical administered with ICE at 35 and 100 mg/kg of body weight. 35 mg/kg of ranitidine were intragastrical administered as the positive control. All the mice were administered 6 times in 3 days and an 18 h starvation were carried out after the last administration. To build the acute gastritis model, all the administered mice were treated with 400 μl HCL/EtOH solution (150 mM HCL in 60% ethanol) for 1 h, then the mice were sacrificed and the stomach were taken out. Total RNA of the stomach were prepared after grounded the tissues to powder with liquid nitrogen. The mRNA expression levels of iNOS, TNFα, IL-1β and IL-6 in stomach were analyzed by RT-qPCR analysis. All animal experiments in accordance with those approved by the Institutional Animal Care and Use Committee at the Affiliated Huaian No. 1 People’s Hospital (approval number: DW-P-2022-001-01). Results in the present study were represented as the mean ± SD (standard deviation). The significance was analyzed between the two groups using Student’s t-test and p-values less than 5% were considered to be statistically different. UPLC-QTOF-MS was used to identify the major compounds in ICE. Nine chromatographic peaks were identified in the ICE profile (Figure 1). Comparing the retention times and ultraviolet and MS spectra of the samples with those of the standards (Table 2), nine main compounds were identified: peaks 1-9 were protopine, 8-epideoxyloganic acid, eriodictyol, coumaric acid, naringenin, quercetin-3-O-glucoside, rutin, nicotiflorin, and phellopterin, respectively. The cytotoxicity of ICE on RAW264.7 cells was estimated using the MTT assay. A high ICE concentration did not significantly affect cell viability (Figure 2). RAW264.7 cell viability was decreased slightly on treatment with 200 or 400 μg/ml ICE. When the ICE concentration was 100 μg/ml or less, there was almost no effect on cell viability. Therefore, the ICE concentrations used in subsequent studies were 100 μg/ml or lower. To validate the anti-inflammatory effect of ICE, the NO generated was measured (Figure 3A). The NO content in LPS-stimulated RAW 264.7 cells was increased markedly compared with controls. However, in the presence of 12.5, 25, 50, and 100 μg/ml ICE, NO production significantly reduced to 98.62%, 74.63%, 60.31%, and 45.29% of control, respectively. L-NMMA (100 μg/ml) inhibited NO production to 24.88% in LPS-stimulated RAW264.7 cells. The proinflammatory factors IL-6, MCP-1 and TNF-α were assessed with the CBA assay, and ICE reversed the increased pro-inflammatory cytokine production in a dose-dependent manner (Figures 3B–D). To validate whether ICE exerts its anti-inflammatory activity via the PI3K/AKT signalling pathway, PDK1, AKT, and GSK3β proteins, and the phosphorylation thereof, as well as NLRP3 were measured in LPS-stimulated RAW 264.7 cells with or without ICE treatment by Western blotting (Figures 4A,B). This revealed that ICE downregulated PDK1, AKT and GSK3β phosphorylation (Figure 4A), meanwhile, ICE also downregulated the expression level of NLRP3 (Figure 4B). Therefore, we speculate that the PI3K/AKT signalling pathway and NLRP3 are involved as the targets of ICE anti-inflammatory activity. To understand the mechanisms underlying ICE anti-inflammatory activity at a broader scale, the transcriptomes of LPS-stimulated and ICE-treated RAW 264.7 cells were analysed using RNA-Seq. This revealed 340 DEGs, including 205 upregulated and 135 downregulated genes (Figure 5A). The DEGs are displayed in a volcano plot (Figure 5B) and heatmap (Figure 5C). DEGs that clustered in the heatmap showed repeatable gene responses to ICE treatment among experimental replicates. To gain insight into the functions of selected DEGs, GO and KEGG enrichment analyses were conducted. In the GO enrichment analysis (Figure 6), DEGs were enriched in biological process (BP), cellular component (CC), and molecular function (MF). For BP, DEGs were mainly enriched in cellular process, single-organism process, metabolic process, biological regulation, regulation of BP, and response to stimulus. Note that in BP, several DEGs were enriched in signalling, positive and negative regulation of BP, and immune system process. For CC, DEGs were mainly enriched in cell, cell part, organelle, membrane, membrane part, organelle part, membrane-enclosed lumen, and macromolecular complex. For MF, DEGs were mainly enriched in binding, catalytic activity, molecular transducer activity, signal transducer activity, transporter activity, MF regulator, transcription factor activity, and protein binding. In the KEGG enrichment analysis (Figure 7), DEGs were significantly enriched in cytokine–cytokine receptor interaction, IL-17 signalling pathway, TNF signalling pathway, and inflammatory bowel disease. To verify the reliability of the RNA-Seq data, the mRNA expression of 16 representative DEGs was assessed by RT-qPCR. The mRNA expression of these 16 DEGs showed the same trend as in the RNA-Seq analysis (Figure 8). Therefore, our RNA-Seq analysis was accurate and reliable, and reflects the expression changes of genes in response to ICE treatment. To study the anti-inflammatory activity of ICE in vivo, a mouse HCL/EtOH-induced acute gastritis model was established, and ICE or the positive control ranitidine was administered intragastrically. Compared with the control, the gastric tissue in the MC group showed severe inflammatory lesions; the lesions were less obvious in the LD, HD, and PC groups (Figures 9A,B). The inflammatory lesions in the HD group were markedly less severe than in the LD group, indicating that ICE has dose-dependent anti-inflammatory activity. To validate the anti-inflammatory activity of ICE in the acute gastritis model at the molecular level, the mRNA expression of iNOS and the proinflammatory cytokines TNF-α, IL-1β, and IL-6 in gastric tissue in each group was detected by RT-qPCR. The iNOS, TNF-α, IL-1β, and IL-6 mRNA expression decreased significantly with ICE administration in a dose-dependent manner (Figures 9C–F), validating the in vivo anti-inflammatory activity of ICE. As a widely used traditional Tibetan medicine, I. compacta reduces inflammation and is used to treat otitis media (The State Pharmacopoeia Commission of the People’s Republic of China, 1995). According to the Tibetan Medicine Record, I. compacta can be used as red “Ou-qu” to prevent Qi-stagnation and turgidity, and to treat otopathy and cough; it can also been used to reduce abdominal distention (Yang, 1991; Dge-bśes, 1986). In the last decade, a few papers have reported the pharmacological activity of I. compacta based on modern pharmacological techniques. A phytochemical analysis revealed that I. compacta contains several bioactive compounds, including polyphenols, flavonoids, alkaloids, and phenylethanoid glycosides (Guo et al., 2019). In this study, the compounds in ICE were analysed using UPLC-QTOF-MS, which identified nine compounds: protopine, 8-epideoxyloganic acid, eriodictyol, coumaric acid, naringenin, quercetin-3-O-glucoside, rutin, nicotiflorin, and phellopterin (Figure 1; Table 2). Several of them were reported with anti-inflammatory properties, such as protopine (Saeed et al., 1997), eriodictyol (Lee, 2011), and naringenin (Ahmed et al., 2018). These compounds may act as the phytochemical basis of ICE to exert its anti-inflammatory property. Previously, we found that phenylethanoid glycosides extracted from I. compacta root had hepatoprotective effects in a carbon tetrachloride-induced liver HepG2 cell injury model, imparted via antioxidation and NF-κB downregulation (Shen et al., 2015; Wu et al., 2016). The trichloromethane fraction of I. compacta root, designated R2 in a previous report, exerted anti-proliferation effects on AGS-EBV cancer cells by regulating related protein expression and inducing EBV lytic replication, apoptosis, and G0/G1 arrest (Zhang et al., 2016). Besides antioxidant effects, an aqueous extract of I compacta showed dose-dependent analgesic effects in the formalin test (Guo et al., 2019). This work investigated the anti-inflammatory activity of I. compacta in LPS-stimulated RAW 264.7 cells. Compared with controls, RAW264.7 cells treated with 1 μg/ml LPS generated large amounts of NO, while LPS-stimulated RAW264.7 cells pre-treated with different ICE concentrations generated significantly less NO, in a dose-dependent manner (Figure 1). The intercellular messenger NO has many roles in the immune system. Activated macrophages release NO in response to infection (Bogdan, 2001). As an immune system modulator, NO indicates inflammation (Tripathi et al., 2007). Therefore, the inhibitory activity of ICE on NO generation in LPS-stimulated RAW 264.7 cells suggests anti-inflammation activity. The levels of the pro-inflammatory cytokines IL-6, MCP-1, and TNF-α in LPS-stimulated RAW 264.7 cells also decreased after ICE treatment (Figure 3), verifying the anti-inflammatory activity of ICE. PI3K/AKT is an important signalling pathway that controls many cellular processes, including cell division, autophagy, survival, and differentiation; it also modulates the inflammatory response (Fruman et al., 2017). The PI3K/AKT signalling pathway involves many anti-inflammatory compounds. The compounds that we identified in ICE, including protopine, eriodictyol, coumaric acid, naringenin, rutin, and phellopterin, regulate the PI3K/AKT signalling pathway to exert their pharmacological activities (Zhang et al., 2012; Lim and Song, 2016; Lim et al., 2017; Lim et al., 2017; Bao et al., 2018; Zhou et al., 2019; Li et al., 2020; Liu and Li, 2021; Nie et al., 2021). Hence, we speculate that the PI3K/AKT signalling pathway is involved in the mechanism underlying the anti-inflammatory activity of ICE. To test this hypothesis, the expression and phosphorylation of PDK1, AKT, and GSK3β in LPS-stimulated RAW 264.7 cells was assessed by Western blotting; the ICE pre-treated cells showed downregulated PI3K expression and p-PDK1, p-AKT, and p-GSK3β phosphorylation (Figure 4A). NLRP3 inflammasome was assembled of NLRP3, ASC and pro-caspase-1, the activation of NLRP3 inflammasome can leads to the activation of caspase-1 and further to release the proinflammatory cytokines IL-1β and IL-18. The activation of NLRP3 inflammasome were reported to relating with a wide range of diseases, such as type 2 diabetes, Alzheimer’s disease, obesity, cerebral and myocardial ischemic diseases, and a variety of auto-immune and auto-inflammatory diseases (Fusco et al., 2020). Therefore, pharmacological research on NLRP3 inhibitors gained widely concern in the resent years. In the past decades, several inhibitors targeting on NLRP3 inflammasome have been reported (Shao et al., 2015), including Oridonin, OLT1177, Tranilast and many other agents (Zahid et al., 2019). In the present study, the expression level of NLRP3 in RAW264.7 cells was revealed been downregulated by ICE treatment (Figure 4B). These results indicate that the PI3K/AKT signalling pathway and inhibition of NLRP3 inflammasome are involved in the anti-inflammatory effects of ICE. To better understand the anti-inflammatory mechanism of ICE, RAW 264.7 cell genes that responded to ICE treatment were analysed at the transcriptome level using RNA-Seq, this revealed 340 DEGs (205 upregulated, 135 downregulated; Figure 5A) enriched in several inflammation-related GO terms and KEGG pathways (Figures 6, 7). In the GO enrichment analysis, DEGs were enriched in inflammation-related biological processes, such as cellular processes, biological regulation, BP regulation, response to stimulus, signalling, positive/negative regulation of BP, and immune system process. In the MF sub-ontology, DEGs were mainly enriched in binding, catalytic activity, molecular/signal transducers activity, MF regulator, and transcription factor activity. Therefore, ICE exerts its anti-inflammatory activity via these biological processes and molecular functions. IL-17 is induced during bacterial infection, and in turn induces the production of inflammatory cytokines including IL-1, GM-CSF, TNF-α, and IL-6, and chemokines such as MCP-1, MCP-3, and MIP-3A (Qian et al., 2010). Therefore, IL-17 plays critical roles in host immunity and inflammation. IL-17 activates several inflammation-related signalling pathways, including NF-κB, MAPK, C/EBPs, PI3K, and STAT (Li X. et al., 2019). In this study, the top pathways in the KEGG enrichment analysis were the IL-17 signalling pathway and other inflammation-related pathways, including the TNF, PPAR, and Jak-STAT signalling pathways. Therefore, we speculate that ICE modulates the inflammatory response via the IL-17 signalling pathway and other related pathways. Alcohol abuse can cause severe gastric mucosa inflammatory lesions (Repetto and Boveris, 2010). Therefore, to study the anti-inflammatory activity of ICE in vivo, a mouse ethanol-induced acute gastritis model was used to assess the protective effects of ICE against inflammatory lesions. Oral ICE strongly ameliorated gastric inflammation by downregulating pro-inflammatory expression, similar to the anti-gastritis effects of plants such as Alisma canaliculatum, and Geranium koreanum (Kim et al., 2018; Nam and Choo, 2021). This study demonstrated that ICE exerts anti-inflammatory activity. The PI3K/AKT signalling pathway may be involved in the mechanisms by which ICE exerts its activity. At the transcriptome level, ICE treatment regulated hundreds of genes, many of which are involved in inflammation-related biological processes and functions, and in signalling pathways via which ICE exerts its anti-inflammatory effects.

PMC9649822

Yun-cheng Li,Kang-shuai Li,Zeng-li Liu,Yong-chang Tang,Xiao-Qiang Hu,Xing-yong Li,An-da Shi,Li-ming Zhao,Li-Zhuang Shu,Shuo Lian,Zhang-di Yan,Shao-hui Huang,Guo-li Sheng,Yan Song,Yun-jia Liu,Fan Huan,Ming-hui Zhang,Zong-li Zhang

Research progress of bile biomarkers and their immunoregulatory role in biliary tract cancers

28-10-2022

biliary tract cancers,bile components,biomarkers,immune effect,liquid biopsy

Biliary tract cancers (BTCs), including cholangiocarcinoma and gallbladder carcinoma, originate from the biliary epithelium and have a poor prognosis. Surgery is the only choice for cure in the early stage of disease. However, most patients are diagnosed in the advanced stage and lose the chance for surgery. Early diagnosis could significantly improve the prognosis of patients. Bile has complex components and is in direct contact with biliary tract tumors. Bile components are closely related to the occurrence and development of biliary tract tumors and may be applied as biomarkers for BTCs. Meanwhile, arising evidence has confirmed the immunoregulatory role of bile components. In this review, we aim to summarize and discuss the relationship between bile components and biliary tract cancers and their ability as biomarkers for BTCs, highlighting the role of bile components in regulating immune response, and their promising application prospects.

Research progress of bile biomarkers and their immunoregulatory role in biliary tract cancers Biliary tract cancers (BTCs), including cholangiocarcinoma and gallbladder carcinoma, originate from the biliary epithelium and have a poor prognosis. Surgery is the only choice for cure in the early stage of disease. However, most patients are diagnosed in the advanced stage and lose the chance for surgery. Early diagnosis could significantly improve the prognosis of patients. Bile has complex components and is in direct contact with biliary tract tumors. Bile components are closely related to the occurrence and development of biliary tract tumors and may be applied as biomarkers for BTCs. Meanwhile, arising evidence has confirmed the immunoregulatory role of bile components. In this review, we aim to summarize and discuss the relationship between bile components and biliary tract cancers and their ability as biomarkers for BTCs, highlighting the role of bile components in regulating immune response, and their promising application prospects. Biliary tract cancers (BTCs) constitute a diverse group of malignancies emerging from the biliary epithelium, including cholangiocarcinoma and gallbladder carcinoma (1). According to its anatomical site, cholangiocarcinoma can be divided into intrahepatic cholangiocarcinoma(iCCA), perihilar cholangiocarcinoma(pCCA) and distal cholangiocarcinoma(dCCA) (2–4). iCCA, pCCA, and dCCA have distinct clinicopathologic characteristics in terms of epidemiology, molecular characteristics, clinical features, management, and outcomes (5). BTCs account for about 3% of all gastrointestinal tumors, and their incidence has increased in recent years (6). BTCs have an abysmal prognosis, with a five-year survival rate of about 10% (4). Radical resection is the only cure chance for BTCs. However, since patients are usually asymptomatic or only exhibit non-specific symptoms in the early stage, most patients are diagnosed at the advanced stage, missing the opportunity for radical surgery (7). Therefore, early diagnosis of BTCs is vital to improve patient prognosis. The diagnosis of BTCs is based on a combination of imaging techniques, including ultrasound, CT, MRI, and tumor biopsy (5, 8).CT scanning is convenient and economical and is usually the first choice for diagnosing BTCs. MRI can provide high-resolution anatomical information about the biliary system, while MRCP is audio-visual (1). ERCP can directly display the biliary system and improve the accuracy of diagnosis. ERCP can also be used to confirm the diagnosis of cholangiocarcinoma by brushing cytology or biopsy (9). However, widely accepted biomarkers for diagnosing and dynamically monitoring the disease are still lacking. Currently, widely applied tumor markers, such as serum CA199 and CEA, have limited diagnostic value for BTCs (10). Tumor markers can be used for tumor screening, early detection, differential diagnosis and staging, prognosis judgment, efficacy monitoring, recurrence and metastasis monitoring, and guiding individualized treatment. The ideal tumor markers should have high sensitivity and specificity and reflect the tumor’s dynamic changes. The detection of tumor markers should be simple, fast, and accurate. Tumor markers have great application prospects in cancer. Thus, many researchers are working on searching for biomarkers used in the diagnosis, treatment response, and prognosis of BTCs (11–13). Tumor immunotherapy is a new and hot field. In addition, understanding the interaction between molecules and immune system is the premise of implementing immunotherapy. Bile is produced by hepatocytes and secreted into the intestine through the biliary system. The compositions of bile are complex. Bile is composed mainly of bile acids, phospholipids, cholesterol, bilirubin, proteins, inorganic salts, etc. Proteins account for about 7% of the total bile compositions (14). Under the condition of disease, the ingredients of bile can be changed, especially in terms of BTCs, which is directly in touch with bile. Therefore, compared with blood or other body fluids, bile is an essential source for searching for tumor markers in the biliary system. Current advances in omics techniques have enabled researchers to discover new and valuable biomarkers in biological fluids (5). This review focuses on changes in bile composition and their interaction with the immune system. Changes in small molecules, proteins, ctDNA, miRNA, and extracellular vesicles have also been discussed. When tumors of the biliary system occur, small molecules in the bile are altered. Existing studies have focused on differences in bile acid molecules and lipids in bile. The production of bile acids in the liver is the body’s primary way of cholesterol metabolism. The primary bile acids, such as cholic and chenodeoxycholic acid, are synthesized from cholesterol in the liver, then secreted into the intestine to produce secondary bile acids through the action of intestinal bacteria. Primary and secondary bile acids are unconjugated and can be combined with glycine or taurine to form conjugated bile acids (15). The intestine can reabsorb more than 95% of all bile acids; the rest is excreted with the stool (16). Bile acids can be divided into hydrophobic and hydrophilic bile acids according to their affinity for water. Hydrophobic and hydrophilic bile acids play opposite roles in tumor carcinogenesis. Hydrophobic bile acids, such as deoxycholic acids (DCA) and lithocholic acid (LCA), can promote tumorigenesis, whereas hydrophilic bile acids, such as ursodeoxycholic acid (UDCA), can prevent tumorigenesis (17–19). However, the relationship between the composition of bile acid in bile and the malignant transformation of biliary tract tumors has not been fully elucidated. Compared with patients with biliary stones or controls (liver transplant donors without biliary disease), patients with biliary system tumors had lower concentrations of total bile acids and lower ratios of DCA and LCA. For patients with bilirubin ≤ 2.0mg/dL, biliary tract tumor patients also showed lower total bile acid concentration and a lower percentage of DCA and LCA compared with the control group, suggesting its value in the early diagnosis of biliary tract tumors (20). A later study further confirmed the reduction in total bile acid concentrations. They also found an elevated ratio of primary bile acids and conjugated bile acids in CCA patients compared with BBD or PC, which is consistent with previous results (21, 22), indicating that conjugated bile acids can promote the growth of bile tract cancers. They further identified GCA as a positive biomarker and TUCDA as a negative biomarker for cholangiocarcinoma (23). Researchers have also found that compared with benign stenosis, the conjugated bile acids(C)/unconjugated bile acids(U) ratio of malignant stenosis patients is relatively low. As the C/U ratio decreases, the diagnosis rate of inflammation and tumors increases (24). The synthesis of bile acids is dynamically regulated and bile acid concentration can be a real-time and reliable indicator of liver function (25).However, the regulatory mechanisms of bile acid synthesis is complex and have not yet been fully elucidated. The results observed in those studies may reflect a partial imbalance in the process of bile acid during tumorigenesis. In addition to digestive function, bile acids also participate in the regulation of immune function. Primary and secondary bile acids can inhibit the expression of pro-inflammatory cytokines from monocytes, macrophages, dendritic cells, and Kupffer cells (26). Bile acid metabolites affect the host immune response by directly regulating the balance of TH17 and Tregs (27). One LCA derivative, 3-oxolCa, can inhibit TH17 cell differentiation. Another LCA derivative, IsoalloLCA, can enhance Treg differentiation. Biliary primary and secondary bile acids can regulate the expression of RORγ+ regulatory T cells via bile acid nuclear receptor (28). Primary bile acids increase the expression of CXCL6, thereby causing aggregation of hepatic CXCR6+ NKT cells and producing anti-tumor effects through IFN-γ, while secondary bile acids have opposite effects (29). SIRT5 is a metabolic regulator involved in the occurrence and development of a variety of tumors and plays context-specific roles. A recent study found that the decreased expression of SITR5 in hepatocellular carcinoma patients leads to increased bile acid synthesis. Bile acids act as signaling molecules to stimulate their nuclear receptors and promote the polarization of M2-type macrophages, thereby creating an immunosuppressive tumor microenvironment and promoting tumor growth (30). It is reasonable to infer that bile acids play a role in the occurrence and development of biliary tract tumors by interacting with the immune system. In conclusion, bile acids play a complex role in immune regulation, and understanding the role of bile acids in immunity can promote the implementation of immunotherapy. Lipids include phospholipids, glycolipids, and cholesterol and their esters. Oxidized phospholipids(oxPLs) are essential in tumor cell apoptosis (31). Ten oxPLs were detected in bile from patients with diseases including CCA, PSC, PC, and BBD. Among them, ON-PC(1-palmitoyl-2(9-oxononanoyl)-sn-glycero-3-phosphatidylcholine) and S-PC(1-palmitoyl-2-succinoyl-sn-glycero-3-phosphatidylcholine) were significantly increased in CCA compared with other biliary stenosis patients. The sensitivity and specificity of ON-PC for distinguishing CCA from other biliary strictures were 85.7% and 80.3%, respectively. The sensitivity and specificity were increased to 100% and 83.3% by combining ON-PC and S-PC (32). The lipid peroxidation product 4- hydroxynonenal (4-HNE) was raised in gallbladder cancer patients compared with stones or controls. These findings indicate that lipids and their metabolites are also important bile markers for BTCs (33). Metabolomics was applied to analyze low molecular metabolites in the body (10). The techniques used in metabolomics studies typically include Fourier transform infrared spectroscopy, gas chromatography/mass spectrometry(GC/MS), or lipid chromatography/mass spectrometry(LC/MS) (34). A new nuclear magnetic resonance spectra (NMR)-based metabolomics approach was also established to compare differences in the bile of patients with biliary tract cancer and benign biliary tract disease. A specific NMR spectroscopy can distinguish the two groups with 86% and 81% sensitivity and specificity, respectively (35). Lysophosphatidylcholine, phenylalanine, 2-octenoylcarnitine, and tryptophan levels in biliary tract cancer were decreased in patients of BTCs compared with benign diseases (36, 37).1H-MRS method was also applied to analyze the composition of bile. The levels of phosphatidylcholine and Taurine in bile were decreased in CCA patients compared to benign groups. The four regions in the final classifier (which represents phosphatidylcholine, bile acids, lipid, and cholesterol) can distinguish the two groups with a sensitivity and specificity of 88.9% and 87.1%, respectively (38). Another study examined volatile organic compounds in bile headspaces (gas above the sample) and found that several compounds (ethanol, acrylonitrile, acetonitrile, acetaldehyde, benzene, carbon disulfide, dimethyl sulfide, 2-propanolol) were decreased in patients with CCA compared with PSCs. A model was established based on the bile concentration of acrylonitrile, 3-methyl hexane, and benzene to distinguish patients with or without CCA. The sensitivity and specificity of the model are 90.5% and 72.7%, respectively (39). Studies have focused on changes in other small molecules in the bile of patients with biliary tract cancer. The levels of glutathione (GSH), peroxide, ferrous iron (Fe2+), glutathione peroxidase (GPX), and farnesyl transferase/geranylgeranyltransferase type-1 subunit alpha (FNTA) in bile were investigated to illustrate the ferroptosis level of eCCA. The ability of these substances to distinguish between eCCA and common bile duct stones was further investigated. Compared with patients with common bile duct stones, the levels of these substances in bile are significantly reduced. Bile can be used as a biomarker source for diagnosing eCCA, especially for differentiating eCCA from benign bile duct stenosis (40). The lactate level in bile differs between the malignant disease group and non-malignant diseases or healthy people and may serve as a biomarker to distinguish benign people and malignancies (41). The content of ions and heavy metals in bile also changes in the diseased state. Compared with gallstones, patients with gallbladder cancer have decreased zinc levels and elevated copper and copper/zinc ratio (42). They tend to have higher concentrations of heavy metals such as chromium and cadmium (43). Bile is believed to be the most relevant body fluid, which can closely indicate the pathological conditions of the biliary tract. Proteins are involved in tumor development (44–47). More and more data on protein biomarkers in bile is emerging. Serum CA199 and CEA are widely used markers for diagnosing biliary duct cancer, although the sensitivity and specificity are relatively low (48). Consequently, researchers have also studied the possibility of CA199 and CEA as bile biomarkers of CCA. The results indicate that bile CA199 and CEA are not valuable in distinguishing benign and malignant obstructive biliary tract disease (49). Thus, more potential bile markers with high sensitivity and specificity are urgently needed. Proteins in bile that may serve as tumor markers are summarized in Table 1. The mucin family is a group of glycosylated macromolecules. Some mucins are abnormally expressed in cancer cells and are involved in tumorigenesis and progression (70). Several mucin family members also exist in the bile and can serve as bile markers. MUC2 and MUC16 were the first identified mucin family members which were differentially expressed in bile between CCA and controls using proteomics methods (71), although the sensitivity and specificity of bile MUC2 and MUC16 in diagnosing CCA still need further validation. Previous studies have shown that intratumoral MUC4 and MUC5AC are associated with the diagnosis and prognosis of biliary tract tumors (72, 73). Therefore, whether bile MUC4 and MUC5AC are indicators of biliary tract tumors has attracted attention. Bile MUC4 was identified as a particular marker of BTCs (74). However, whether bile MUC5AC expression could distinguish benign and malignant biliary tract diseases is still controversial. A study reported that bile MUC5AC couldn’t distinguish benign from malignant biliary tract diseases (74). However, another study showed that bile MUC5AC could distinguish benign and malignant patients with higher diagnostic accuracy than serum MUC5AC (AUC=0.85 in bile vs. AUC=0.82 in serum). Evaluating MUC5AC in bile and serum simultaneously and based on the serum-to-bile ratio can achieve better diagnostic performance (AUC=0.97) (50). As glycosylated macromolecules, studies have also focused on the performance of glyco-alteration of glycoproteins as a diagnostic marker of CCA. In a study, researchers have established a new sandwich ELISA system to detect the glyco-alteration of MUC1. The results revealed that bile Wisteria floribunda Agglutinin (WFA)-Positive Mucin 1 is a promising CCA biomarker, which can effectively distinguish CCA from benign biliary tract diseases with sensitivity and specificity of 90.0% and 76.3%, respectively and the AUC was 0.86 (51). Neutrophil gelatinase-associated lipocalin (NGAL), also called lipocalin-2 (LCN2), is a 25-kDa multifunctional protein released by activated neutrophils. NGAL plays a role in the onset and development of tumors (75). Zabron et al. compared the bile protein profiles of benign and malignant diseases. Biliary NGAL was the most valuable biomarker in distinguishing between benign and malignant groups. The area under the receiver-operating characteristic curve was 0.76 in differentiating malignant from benign pancreaticobiliary disease with sensitivity and specificity of 94% and 55%, respectively (52). Its diagnostic effect was validated in an independent cohort, and this capability is independent of CA199. The combination of bile NGAL and serum CA199 can improve the diagnostic accuracy for malignancy. Another study showed that bile NGAL levels could distinguish pancreaticobiliary cancer and benign biliary diseases with sensitivity and specificity of 73.3% and 72.2%, respectively. The area under the curve was 0.74 (53). In addition, Chiang et al. reported that biliary NGAL could differentiate CCA patients from gallstone patients with sensitivity and specificity of 87% and 75%, respectively. The AUC was 0.81. CCA patients with higher NGAL expression were also found to have poor overall survival (OS) of patients with CCA (54). Bile samples from patients having biliary stenosis caused by CCA or other malignant tumors were analyzed by the proteomics method (76). A total of 127 differentially expressed proteins were identified. Several of them have been reported to be associated with pancreatic cancer. Immunoblotting assays were performed to validate these markers’ expression, and significant CEACAM6 elevation in CCA has been identified. A subsequent study explored the role of bile CEACAM6 in diagnosing CCA. Bile CEACAM6 could discriminate cholangiocarcinoma from benign disease (AUC=0.738) (55). CEACAM6 was further screened and validated as a biomarker to distinguish malignant and benign bile duct stenosis with sensitivity and specificity of 93% and 83%, respectively. The AUC was 0.92. The combination of bile CEACAM6 and serum CA199 can improve the diagnostic accuracy for malignancy (AUC 0.96) (56). CCA cells can secrete several cytokines and promote tumor progression by paracrine. Some of the cytokines secreted by CCA can also enter bile and may serve as biomarkers. Among them, IGF-1 and VEGF were the most studied. IGF-1 is a bioactive protein polypeptide, and it has been previously reported that serum IGF-1 levels are associated with prostate, breast, pancreatic, lung, and colorectal cancer (77, 78). VEGF promotes tumor angiogenesis, and the relationship between serum VEGF levels and various tumors has also been reported (79, 80). Expression levels of IGF-1 and VEGF in the bile and serum of extrahepatic cholangiocarcinoma, pancreatic cancer, and benign biliary abnormalities were measured to evaluate their role as a tumor marker (57). Biliary IGF-1 perfectly differentiated extrahepatic cholangiocarcinoma from benign biliary abnormalities or pancreatic cancer. The AUC was 1. Bile VEGF levels have no differentiating effect on biliary obstruction. Meanwhile, another study shows the level of bile VEGF could distinguish pancreatic patients from other etiologies of biliary stricture (58). Biliary VEGF-1 may help rule out distal common bile duct cancer (81). Some cytokines related to chronic inflammation, such as IL-6 and TNF-a, play a role in the development of tumors by interacting with the immune system. Chronic inflammation-driven cytokines released from cholangiocytes, fibroblasts, or immune cells can promote the development of CCA (82). BTCs are always complicated with an anomalous arrangement of pancreaticobiliary ducts and affect bile digestive enzymes, especially for pCCA and dCCA. Abnormal digestive enzymes and their relative product expression in bile may serve as biomarkers. One study used the quantitative proteomics method to compare bile proteins in six CCA patients with three different gross-appearance tumor types to identify possible biomarkers. They selected α-1- antitrypsin (AAT) for further validation. Fecal AAT, which can indirectly reflect bile AAT, can differentiate CCA from controls with a sensitivity and specificity of 80% and 75%, respectively. The AUC was 0.83 (83). Elevated amylase activity in bile is also associated with pancreaticobiliary reflux, a significant risk factor for biliary system tumors (84, 85). Therefore, high biliary amylase activity is associated with BTCs tumorigenesis (86). However, pancreaticobiliary reflux is affected by the location and extent of biliary tract obstruction. The amylase level in bile is not specific for diagnosing biliary tract cancers (87, 88). Compared with patients with gallbladder stones, the pancreatic elastase (PE) level in CCA patients is elevated. Biliary PE can distinguish patients with CCA from that with gallstones. The combined measurement of PE and amylase and the PE to amylase ratio can improve the accuracy of CCA diagnosis (AUC=0.0877) (59). Dhar et al. (89) discussed the role of pyruvate kinase M2(PKM2) in the development of cholangiocarcinoma. They proposed that PKM2 can be used as a diagnostic marker of BTCs and is associated with prognosis. A subsequent study (60) showed that biliary PKM2 had low sensitivity (52.9%), but the specificity was high (94.1%) for the diagnosis of malignant and benign bile duct stenosis. The AUC was 0.77. In the first published bile proteomics, Mac-2-binding protein (Mac2BP) was identified as the potential biomarker for CCA. The value of Mac2BP as a diagnostic marker for cholangiocarcinoma was further validated in the following study (61). Bile Mac2BP can distinguish between malignant and benign biliary diseases with sensitivity and specificity of 69% and 67%, and the AUC was 0.70. The diagnostic effect is similar to biliary CA199 (AUC 0.69), and the combination of bile Mac2BP and CA19-9 can improve the accuracy of diagnosis (AUC 0.75). Our group (62) applied ELISA to detect survivin and CA199 in bile. Biliary survivin could differentiate CCA and benign biliary obstruction with sensitivity and specificity of 67.27% and 80.85%, and the AUC was 0.78. Combining biliary survivin and CA199 could improve diagnostic accuracy (AUC=0.838). Wang et al. (90) analyzed the proteins of gallbladder bile samples from gallbladder cancer, gallbladder adenoma, and chronic calculous cholecystitis patients using 2D LC-MS/MS. They proposed S100A8 as a potential biomarker for GBC and used immunohistochemical analysis to confirm it. Another study showed that S100P levels in bile were significantly elevated in patients with CCA, and bile S100P could distinguish between patients with CCA and choledocholithiasis with sensitivity and specificity of 92.9% and 70.0%, respectively (63). Some other bile proteins including total fibronectin(tFN) and cellular fibronectin(cFN) (64), sB7-HB3 (65), sCD97 (66), LR11 (67), minichromosome maintenance protein 5(Mcm5) (68), HSP27, and HSP70 (69) was also reported to be bile biomarkers to distinguish benign and malignant biliary tract diseases with different sensitivity and specificity. However, these biomarkers still need further validation. The term liquid biopsy was first used to describe how the same diagnostic information can be obtained from blood samples as tissue samples. In oncology, the term broadly refers to the sampling and analysis of various biological fluids (91). Analytes include cell-free DNA(cfDNA), cell-free RNA(cfRNA), circulating tumor cells (CTCs), and extracellular vesicles (EVs) (91). The interaction between liquid biopsy components and anti-cancer immunity is a new area of research (92). CfDNA refers to fragmented DNA found in non-cellular components of blood, which are usually double-stranded fragments of about 150-200 base pairs in length (93). In cancer patients, the cfDNA released from tumor cells is called circulating tumor DNA (ctDNA) and constitutes part of the total cfDNA (94, 95). Assessing the ctDNA in bile is a promising liquid biopsy method, especially for patients with insufficient tumor tissue to biopsy. However, whether the mutation profiles detected by ctDNA reliably reproduce the mutation profiles obtained by tumor biopsy remains a question (96). Bile contact with tumor cells directly, Therefore, detecting genetic alterations in bile may be more valuable than in plasma for biliary tract cancers. Several studies (97, 98) have demonstrated the feasibility of detecting genetic mutations of ctDNA in tumors of the biliary system, with a high compatibility rate with tissue biopsies. A recent study shows genetic mutations of ctDNA in bile are similar to those in tissues and are more effective than in plasma (97). Similar conclusions were reached in another study (98). cfDNA in bile samples is mainly long DNA fragments, significantly different from plasma samples. Targeted deep sequencing can reliably detect mutant variants in bile cfDNA. Tumor stage and location affect the sensitivity of mutation detection in bile cfDNA samples. In patients with GBC, the positive rate of mutation detected in bile ctDNA was 58.3%, which is higher than cytology. The mutation compliance rate between tissue DNA and bile ctDNA in GBC patients was 85.7%. KRAS and TP53 are frequent mutation genes in tumors of the biliary system. Several previous studies have evaluated the diagnostic effects of KRAS and P53 mutations in bile on tumors of the biliary system (99–102). Overall, the frequency of KRAS and P53 mutations in bile is too low to rely on these molecular markers to establish a reliable diagnosis. It was initially thought that higher cfDNA might indicate tumor growth. However, many other diseases contribute to a similar increase. Thus, recent research has focused on epigenetic features of cfDNA, such as methylation (103). One study showed that hypermethylation of UCHL1 and RUNX3 in pancreaticobiliary fluid might be useful for diagnosing biliary duct cancers (104). INK4a/ARF’s promoter methylation status in bile might be useful for diagnosing benign and malignant biliary tract diseases (105). Shen et al. (106) established and validated a methylation panel in the bile, which can improve the sensitivity of detecting BTCs. ctDNA can be used to diagnose biliary duct cancers at an early stage. Recently, a prospective study using ddPCR to detect the methylation status of four genes (CDO1, CNRIP1, SEPT9, VIM) in bile showed that detecting DNA methylation markers in bile allowed for more accurate and early detection of CCA patients in PSCs (107). Another prospective study suggests that next-generation sequencing (NGS) mutational analysis of bile cell-free DNA (cfDNA) in the initial stages of biliary stenosis can significantly improve the detection of malignant tumors. In addition, it can also test for the presence of mutations suitable for targeted therapy (108). miRNAs are small non-coding RNA molecules of 18-25 nucleotides in length, which can regulate gene expression by binding to corresponding mRNA sites (109). Extracellular miRNAs are stable and detectable in other body fluids (110). There are significant differences in miRNA concentration and composition in different diseases. Therefore, miRNA can be used as a biomarker to evaluate the body’s condition (111). Extracellular miRNAs are not homogenous groups and exist in various body fluids in different forms. Its origin and function are not fully understood. It is currently believed that some extracellular miRNAs are packaged in apoptotic bodies, shedding microvesicles, exosomes, or high-density lipoprotein (HDL) particles. Most extracellular miRNAs (90%-95%) bind to AGO proteins (110).Therefore, attention should be paid to the origin of the miRNAs when evaluating their roles as biomarkers. Extracellular vesicles are all kinds of vesicles with membrane structures released by cells, and their contents include proteins, lipids, nucleic acids, and so on. According to the diameter and occurrence mode, it can be divided into exosomes, microvesicles, and apoptotic bodies (112, 113). Tumor-derived exosomes are likely to serve as potential cancer markers. Shigehara et al. (114) compared the difference of miRNA in the whole bile between patients with malignant and benign biliary tract diseases for the first time and pointed out that miR-145 and miR-9 could be used as diagnostic markers of BTCs ( Table 2 ), especially miR-9, which has higher diagnostic accuracy. A subsequent study showed that RNA from free-floating cells degrades rapidly. The bile processing process can produce unpredictable deviations. Therefore, they focused on extracellular vesicles in bile instead of the whole bile and established a miRNA panel based on extracellular vesicles in bile to diagnose CCA with a sensitivity and specificity of 67% and 96%, respectively (115). However, miRNA in extracellular vesicles only accounts for a small part of extracellular miRNA. Thus, one study suggested that extracellular vesicles were not a reliable miRNA source for CCA diagnosis. They recommend cell-free bile rather than whole bile or extracellular vesicles in bile to find potential miRNA markers (118). miRNA profiles in the bile of PSC, PSC with CCA, and CCA patients were different, among which miR-142, miR-640, miR-1537, and miR-3187 were significantly different in PSC and CCA patients (116) ( Table 2 ). In another study, extracellular miR-30d-5p and miR-92a-3p were increased dramatically in the bile of CCA patients compared with benign biliary disease (BBD) patients (117) ( Table 2 ). miR-30d-5p showed the best performance in distinguishing CCA and BBD patients. In addition, it has been demonstrated that high methylation of miR-1247, -200a, and -200b in bile may help determine benign and malignant biliary tract diseases (119). The extracellular vesicles can also be used to distinguish between patients with benign and malignant diseases. As aforementioned, an established miRNA panel ( Table 3 ) in the extracellular vesicles can be used to diagnose CCA (115). Ge X et al. (120)compared the expression of lncRNAs in bile exosomes from CCA patients and biliary obstruction patients. They identified that two lncRNAs were significantly elevated in CCA, and the combined use of two lncRNAs showed better diagnostic value ( Table 3 ). Besides, both of them were associated with CCA prognosis. Bile exosome circCCAC1 can distinguish CCA from benign hepatobiliary diseases (AUC=0.857) while tissue circCCAC1 is associated with the prognosis of patients (122). In addition to non-coding RNA, proteins in extracellular vesicles can also be used as markers for patients with CCA. Ikeda, C et al. (121)used EDEG to separate human bile EVs and analyzed proteins in extracellular vesicles for the first time. They found that Claudin-3 in bile EVs can distinguish between CCA and patients with bile duct stones with a sensitivity and specificity of 87.5% and 87.5%, respectively ( Table 3 ). The concentration of EVs in bile may be relevant to diagnosing malignant biliary tract disease. One study showed that the concentration of EVs in bile can distinguish malignant from non-malignant CBD stenosis with 100% accuracy (123). In addition to being a marker of disease, the potential therapeutic role of EVs in disease is gradually being discovered (124). EVs can target receptor cells, increase drug concentration in target tissues, and reduce drug toxicity and side effects. The outer membrane of the vesicle increases drug stability and prevents drug degradation. The study of vesicle-related functions has become a research hotspot and is expected to play a role in the early diagnosis and treatment of many diseases. BTCs are a group of highly heterogeneous diseases with abysmal prognosis. Finding a method for early diagnosis of BTCs in a non-invasive or minimally invasive way and evaluating the treatment effect to improve the prognosis of patients with BTCs is promising. The search for tumor markers is a feasible approach. Many previous studies have focused on finding a substance in the blood, evaluating its relationship with BTCs, and determining whether it can be used as a tumor marker. Blood samples are readily available, but blood often reflects the body’s overall conditions and cannot be accurately localized to an organ or tissue. The components in the blood may interfere with each other and mask the characteristic information of the disease, especially in the early stages (125). Therefore, many studies have focused on finding tumor markers in other body fluids. Bile is directly exposed to BTCs, and tumors will release certain substances into bile. Thus, bile may contain tumor-specific information, an essential source for searching for tumor markers. The acquisition of bile is invasive, and it is difficult to obtain information about a healthy person’s bile (14). While in the disease state, some treatments such as ERCP, PTCD, and surgery become essential sources of bile without adding additional trauma to patients. Therefore, bile biomarkers research is more valuable. This study focused on bile and summarized the research progress of the relationship between bile components and BTCs. Metabolomics is used to analyze low molecular metabolites in bile to determine the pathophysiology of the human. Bile acids and lipids are two kinds of small molecules that are most concerned at present. However, the results of different studies are inconsistent or even opposite. Many factors are involved, including pretreatment before bile testing, tumor anatomical site and methods of bile collection, individual differences in disease severity, and various other factors affecting the results. For the early diagnosis of BTCs, more attention was paid to whether bile composition changes in the absence of biliary obstruction (20). Bile acids and lipids are involved in the occurrence and development of BTCs, and the mechanism needs to be further studied. Proteomics studies the composition and variation of proteins in cells, tissues, or living organisms. Clinical detection of disease is mainly based on protein analysis. Bile protein analysis is challenging to carry out because of the interference of many other substances (14). Thus, pretreatment is required before proteomic analysis. The proteome is a fluctuating description of the host’s response to disease. Many minor but significant changes may occur in the early stage of the disease (126).Therefore, combining multiple markers is more effective than single markers (127). Protein changes arise during the malignant transformation of healthy cells (128–130). Researchers currently focus on finding overexpressed proteins that flow into body fluids due to disease as disease biomarkers (131–133). Since bile is directly in contact with BTCs, tumor markers of BTC are more likely to be identified in bile. The use of mass spectrometry to analyze proteomic patterns in tissues and body fluids, combined with bioinformatics tools to distinguish patterns in normal, benign, or malignant disease states, namely, proteomic pattern diagnosis (134). Thus, artificial intelligence (AI) is required for processing high-throughput omics data. Liquid biopsy is a hot field now, which can be used to detect cfDNA, cfRNA, CTCs, EVs, etc. (91). Compared with the traditional methods, it can comprehensively reflect the overall state of the tumor in a minimally invasive and fast way, and it applies to a wide range of people. Liquid biopsies are good supplements for patients who have difficulty obtaining biopsy tissues or have insufficient tissues for genetic testing. Liquid biopsies can be used to diagnose and guide the treatment of patients (135). The low ctDNA content of the samples used for liquid biopsies, especially in the early stages, poses a great challenge to the detection of ctDNA (135). ctDNA is currently the most widely used in clinical practice. ctDNA can be used as a tumor marker of BTCs by detecting the related mutation status of ctDNA (125). ctDNA in bile provides a comprehensive view of the tumor genome, reflecting DNA released from multiple tumor regions. It is dynamic. Thus, ctDNA in bile can monitor treatment response, drug-resistant, tumor heterogeneity, and detect residual disease after surgery. In conclusion, detecting ctDNA in bile is an auspicious method. Multiple miRNAs have been found in the blood as markers of BTCs. However, miRNAs are rarely reported in bile. Extracellular miRNAs are heterogeneous and may play an essential role in the occurrence and development of BTCs (110). Extracellular miRNAs may act as tumor markers for BTCs and are associated with prognosis. EVs are a recently emerging class of molecules that play an essential role in cell-to-cell communication, with exosomes being the most mature (113). There are also growing studies of EVs in bile. EVs contain various substances such as proteins, nucleic acids, etc. These substances within EVs or the concentration of EVs in bile may play an essential role in the occurrence and development of BTCs and may act as biomarkers for BTCs. EVs can also be used in the treatment of diseases. Bile-specific exosomes may lead to better therapeutic outcomes. Bile enters the intestine through the biliary tract and may be helpful in the treatment of intestinal diseases (124). In addition to traditional chemotherapy and targeted therapy, tumor immunotherapy is a revolutionary breakthrough. Instead of acting on tumor cells, it activates the body’s immune system to produce an anti-tumor effect. Understanding the mechanism of interaction between molecules and the immune system is immunotherapy’s premise. Anti-tumor immunotherapy is promising theoretically and ideologically. However, it still seems a long way to go. Bile biomarkers are promising and meaningful in diagnosis and prognosis judgement of CCA. However, bile collection is invasive, and it is difficult to obtain normal bile from a healthy donor (14). Setting the control group to patients with benign diseases may lead to unpredictable biases. Instead, finding a non-invasive way, such as blood, feces, etc., to get an accurate diagnosis is preferred. However, this requires building a relationship between bile and the clinically readily available samples such as blood, urine, stool, etc. Besides, single bile biomarker is hard to reach a high specificity and a high sensitivity simultaneously. An accurate BTCs diagnosis involves combining data from multiple omics, including genomics, epigenomics, transcriptomics, proteomics, and metabolomics. The high-throughput multi-dimensional information, combined with AI technology, makes it possible to diagnose BTCs more accurately and improve the judgment of patient prognosis. Y-CL, K-SL, Z-LL, Y-CT, and X-QH collected the references. Y-CL and K-SL analyzed the references and wrote the paper. X-YL, A-DS, L-MZ, L-ZS, SL, Z-DY, S-HH, G-LS, YS, Y-JL, FH, and M-HZ read the paper and provided revising advice. Z-LZ contributed to study supervision and revised the manuscript. All authors contributed to the article and approved the submitted version. This work was supported by the National Natural Science Foundation of China (Grant No. 81900728, 82072676,82172791,82203766), Shandong Province Natural Science Foundation (Grant No. ZR2019MH008, ZR2020MH238, ZR2021QH079), Shandong Province Key R&D Program(Major Scientific Innovation Projects,2021CXGC011105), Shandong Medical and Health Technology Development Project (Grant No. 2018WSB20002), Clinical Research Foundation of Shandong University (Grant No. 2020SDUCRCA018), Key Research and Development Program of Shandong Province (Grant No. 2019GSF108254). The funders had no role in study design, data collection, analysis, interpretation, and manuscript writing. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649824

Yipeng Dong,Chuwei Zhang,Qingrong Zhang,Zihan Li,Yixiao Wang,Jun Yan,Gujie Wu,Ling Qiu,Zhihan Zhu,Bolin Wang,Haiying Gu,Yi Zhang

Identification of nanoparticle-mediated siRNA-ASPN as a key gene target in the treatment of keloids 10.3389/fbioe.2022.1025546

28-10-2022

keloid,asporin gene,nanoparticle,keloid fibroblast,nude mice allograft,ASPN

Background: Keloid, also known as connective tissue hyperplasia, is a benign proliferative disorder with a global distribution. The available therapeutic interventions are steroid injections, surgical removal of keloids, radiotherapy, compression therapy, the application of cryosurgery, and many other methods. Objectives: Existing treatments or approaches for keloids may lead to similar or even larger lesions at the site of keloid excision, leading to a high recurrence rate. Therefore, this study aims at identifying a new gene-based therapy for the treatment of keloids. Methods: An ASPN-siRNA/nanoparticle combination (si-ASPN) and a negative siRNA/nanoparticle complex (NC) was developed on the basis of bioinformatics studies and used in vitro and in vivo experiments. Results: The results showed a strong correlation between the development of keloids and high expression of ASPN protein. With the expression of ASPN protein greatly reduced in keloid fibroblasts and nude mice allografts after treatment with si-ASPN, the collagen and fibroblasts were also uniform, thinner, parallel and regular. Conclusion: All the above experimental results suggest that keloid and ASPN are closely related and both fibroblast growth and metabolism of keloid are inhibited after silencing ASPN. Therefore, ASPN-siRNA delivered via nanoparticles can serve as a novel intervention therapy for the treatment of keloids.

Identification of nanoparticle-mediated siRNA-ASPN as a key gene target in the treatment of keloids 10.3389/fbioe.2022.1025546 Background: Keloid, also known as connective tissue hyperplasia, is a benign proliferative disorder with a global distribution. The available therapeutic interventions are steroid injections, surgical removal of keloids, radiotherapy, compression therapy, the application of cryosurgery, and many other methods. Objectives: Existing treatments or approaches for keloids may lead to similar or even larger lesions at the site of keloid excision, leading to a high recurrence rate. Therefore, this study aims at identifying a new gene-based therapy for the treatment of keloids. Methods: An ASPN-siRNA/nanoparticle combination (si-ASPN) and a negative siRNA/nanoparticle complex (NC) was developed on the basis of bioinformatics studies and used in vitro and in vivo experiments. Results: The results showed a strong correlation between the development of keloids and high expression of ASPN protein. With the expression of ASPN protein greatly reduced in keloid fibroblasts and nude mice allografts after treatment with si-ASPN, the collagen and fibroblasts were also uniform, thinner, parallel and regular. Conclusion: All the above experimental results suggest that keloid and ASPN are closely related and both fibroblast growth and metabolism of keloid are inhibited after silencing ASPN. Therefore, ASPN-siRNA delivered via nanoparticles can serve as a novel intervention therapy for the treatment of keloids. Keloid, also known as connective tissue hyperplasia, is a benign proliferative disorder with a global distribution. It is characterized by the loss of normal restraint and control of collagen anabolism during skin wound healing, excessive deposition of extracellular matrix collagen and excessive growth of fibroblasts, resulting in excessive proliferation of collagen fibers invasively (Tan S et al., 2019). The lesion often interferes with normal surrounding skin tissue, exceeds the original skin damage, raising a nodular, striated or flaky mass on the skin surface with different shapes, hardness, toughness and red color. Keloids are the mostly common in people aged 10 to 30, usually located on the chest, earlobes, shoulders and back (A. Onoufriadis et al., 2018). Keloids can cause both physical and psychological burden to the patient as it may result in cosmetic impairment, itching, pain,etc. and limited joint activity in severe cases (B. Han et al., 2018). The precise etiology of keloid formation is unknown, but recent studies have pointed out a number of factors that may contribute to the keloid formation. Skin damage is one of the main causes of keloid (Finnerty CC et al., 2016; Ogawa, 2017). Clinically, a keloid forms after an inflammation or skin injury, typically months or even years after the initial event. Another significant factor related to keloid formation is genetics. There’s an increased risk of keloid after trauma of about 15%–20% in dark-skinned individuals, such as Blacks, Hispanics and Asians, which is uncommon in Caucasians, and no cases have been reported in albinos (Gauglitz, G.G. et al., 2011). Asporin (ASPN), an extracellular matrix protein, is a member of a family of small leucine-rich proteoglycans that has been found to play an important role in collagenogenic fibril production, signal transduction and tumour growth. Recent studies have reported its role in the microenvironment of various types of cancers, including breast, pancreatic, prostate, and colorectal cancers. It has also been reported that ASPN secreted by cancer-associated fibroblasts can promote cancer cell invasion and metastasis (Simkova D.et al., 2016; Castellana B.et al., 2012; Maris P.et al., 2015; Hurley PJ.et al., 2016; Wang L.et al., 2017; Huo W.et al., 2016). Asporin has also been considered to be a beneficial regulator in cardiac remodeling (Chengqun Huang.et al., 2022). Previous studies have also shown an association between ASPN, keloid gene expression and increased protein expression (Liu L et al., 2021). Therefore, the ASPN may play an important role in the formation of keloids. Particles with a diameter of roughly 1–1000 nm are known as nanoparticles (Yan-xiong Wu et al., 2017; Sun et al., 2015). When transported to the target cell, nanoparticles can protect payloads such as vaccine antigens, proteins, nucleic acids, and medicines from destruction. (Grabowska et al., 2021; Padilha et al., 2021; Yousefi et al., 2021). Compared to viral vectors, nanoparticles are non-viral delivery vehicles, easier to process and modify, biocompatible, biodegradable, and more favourable safety profiles (ShenJ et al., 2013; S. Arany et al., 2013; Y.C. Chung et al., 2010; M.E. Davis et al., 2010). The slow release properties of nanospheres facilitate sustainability of keloid treatment (DelgadoD et al., 2012; Wong SP et al., 2012). Therefore, it was hypothesized that the ASPN-siRNA/nanoparticle combination would be effective in inhibiting the growth of keloid. Hence, in this study, we applied bioinformatics analysis to first identify genes differentially expressed between diseased and normal skin tissues and ASPN was identified as a key gene. The ideal ASPN-siRNA was developed, tested, and loaded onto PLGA nanoparticles for an in vitro investigation to evaluate its characteristics and transfection efficacy. For in vivo trials, we transplanted patient-based keloid samples into nude mice to assess the therapeutic impact after transplantation, we administered si-ASPN/nanoparticle complexes into the keloid tissue (Figure 1). A retrospective analysis of the case data accumulated during our clinical work was performed to guide the study. Human keloid tissue was obtained from 10 keloid patients who underwent surgery at Affiliated Hospital of Nantong University (Jiangsu province, China) from January 2022 to May 2022, while 10 cases were obtained from normal skin tissue of other plastic surgery patients. Inclusion criteria: ①Selection of patients with pathologically confirmed or diagnosed with keloids to obtain pathological tissue samples; ② Selection of patients who choose other plastic surgery procedures to obtain normal skin tissue; ③ Patients with complete clinical information and signed informed consent were included. Exclusion criteria: ① Patients who have been treated for keloid in the past; ②Patients with fibroproliferative diseases, such as Dupuytren disease; ③ Patients with skin infections and subcutaneous nodules in the area of the keloid; ④ Patients diagnosed with autoimmune diseases, malignancies, haematological diseases, serious infectious diseases, serious respiratory diseases and circulatory diseases or severe kidney and liver dysfunction. The Affiliated Hospital of Nantong University’s Ethics Committee gave its approval to the current study. All participants affirmed their approval to the publication of the photos by signing an informed consent form and a statement. A pathologist with experience in the area analyzed each specimen for keloid growth. (NO.2022-K142-01) The NCBI-Gene Expression Omnibus (GEO) database, a global public gene data repository run by the National Center for Biotechnology Information (NCBI) of the National Library of Medicine, provided the source data for this study (https://www.ncbi.nlm.nih.gov/geo/). The GEO data information includes genes, proteomic analyses, non-coding RNA analysis, etc. The majority of worldwide research organizations maintain these data with their original study findings. For this study, the GEO database’s two keloid-related datasets, GSE92566 and GSE7890 mRNA expression profiles (Homo sapiens), were retrieved. The authors of GSE92566 are Fuentes-Duculan J et al. It was created using the [HG-U133 Plus 2] Affymetrix Human Genome U133Plus2.0 array on the GPL570 platform. For this experiment, six samples from GSE92566 were selected, including two groups (three keloid lesions and three adjacent non-lesioned samples; in addition, we excluded one newly formed keloid sample from GSE92566 for better data consistency). The GSE7890 microarray dataset [(HG-U133 Plus 2) Affymetrix Human Genome U133 + 2.0 array], which contains 10 samples, was also annotated on the GPL570 platform (five keloid, five normal skin). The platform files and sequence matrices were downloaded, and R (V4.1.2) (https://www.bi-oconductor.org/packages/release/bioc/html/limma.html) was used to convert the gene IDs, merge datasets, analyze potential batch effects, normalize data, and calculate gene expression. The probes averaged as the final expression of the genes matched by multiple probes were downloaded. The Limma package was then used to identify probes that were differentially expressed in keloid compared to normal tissue samples. Adjusted p-values were accessed after performing multiple testing corrections using the Benjamini-Hochberg (BH) method. To distinguish between up-regulated and down-regulated genes, absolute log2 fold increases larger than 1.5 and adjusted p-values less than 0.05 were used as thresholds.We selected the differentially expressed genes (DEGs) depicted on the volcano plot. Using the ggplot2 package deg, prominent DEGs were simultaneously plotted as a heat map in R (V4.1.2) (Gu et al., 2016). R package clusterProfiler (V3.14.3) and ggplot2 were used to analyze the DEGs for gene ontology (GO) functional enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis to better understand the biological activities of genes (V3.3.3) (Yu et al., 2012). Further Gene Set Enrichment Analysis (GSEA) (Subramanian A et al., 2005) was performed to determine the underlying molecular mechanisms or potential functional pathways for comparing keloid and skin tissue based on a p value <0.05. Lasso Cox regression and Support Vector Machine (SVM) dimensionality reduction algorithm were used to screen for significant genes among the DEGs, and “glmnet” package (V2.1.1) and “kernlab” package (0.9–31) were used respectively. Simultaneously, the “Venn"packages were used to screen the intersection genes and Venn diagram was drawn, in which the overlapping part of the two circles represented the intersecting genes. Boxplots were generated using the “ggpubr"package to analyze the expression of each intersection gene in keloid (test) group and normal (control) group. The ROC curve was plotted using the R package “pROC” (v1.8) to analyze the sensitivity and specificity of each related gene (Robin X et al., 2011). Total RNA was extracted from normal and keloid samples using Invitrogen TRIzol reagent according to the manufacturer’s instructions (ThermoFisher Scientific, China). Next, total RNA was reverse cleaved using the Human-ASPN test kit according to the manufacturer’s protocol (Shanghai Rui Mian Biological Technology Co. Ltd.) into cDNA followed by qPCR. Relative expression levels were normalized to the final control GAPDH. (Human-ASPN-F Seuence (5’→3′): TGG​GAG​TCT​TGC​TAA​CAT​ACC; Human-ASPN-R Seuence (5’→3′): CAT​CTT​TGG​CAC​TGT​TGG​AC; Human-GAPDH-F Seuence (5’→3′): CTG​GGC​TAC​ACT​GAG​CAC​C; Human-GAPDH-R Seuence (5’→3′): AAG​TGG​TCG​TTG​AGG​GCA​ATG). To assess the total protein content, tissues or cells were treated with protein lysate (Servicebio, Wuhan, China) containing Phenylmethanesulfonyl fluoride (PMSF) followed by Western blotting analysis.Antibodies used included: rabbit anti-ASPN antibody (DF13642,Affinity, China), rabbit anti-GAPDH antibody (Huabio, Hangzhou, China), Anti-rabbit IgG (H + L) (Cell Signaling Technology,United States).The results of relative quantification were analyzed using ImageJ greyscale scanning software. The acquired sample segments were submerged in 10% paraformaldehyde for 24 h. A paraffin microtome (RM2235, Leica, Wetazlar, GER) was used to slice samples embedded in paraffin blocks into 5-m thick sections.The slides were rehydrated dried and given a PBS washing. The slides were subsequently stained with hematoxylin and eosin (HE) (Servicebio, Wuhan, China) for histological analysis to gauge the percentage of inflammatory cells. Masson’s trichrome staining (Servicebio) was used to examine the density of collagen fibers and variations in collagen under a microscope.Subsequently, the slides were then treated with rabbit anti-ASPN antibody, rabbit anti-SMA antibody, and rabbit anti-collagen I antibody overnight at 4°C each and incubated with the secondary antibody at room temperature for 2 h. The slides were next stained with Meyer’s hematoxylin and 3,3-diaminobenzidine (DAB; Sigma, St. Louis, Missouri, United States) for 1–2 min (Sorabio, Beijing, P.R.C.). After blocking with neutral gum (Invitrogen, San Diego, CA, United States), the slides were examined under a microscope (Leica DMR 3000; Leica, Bensheim, Germany). ImageJ was used to quantitatively analyse the immunohistochemistry image data. Three double-stranded siRNAs for in vitro testing in accordance with the genetic sequences of Human ASPN to identify the gene sequence of a siRNA that greatly reduced the production of the target gene. The sequences (5′to 3′) of the three ASPN (human)siRNA are ASPN (human) siRNA1:/rG//rG//rA//rG//rU//rA//rU//rG//rU//rG//rC//rU//rC//rC//rU//rA//rU//rU//rA/TT,/rU//rA//rA//rU//rA//rG//rG//rA//rG//rC//rA//rC//rA//rU//rA//rC//rU//rC//rC/TT; ASPN (human) siRNA2:/rG//rC//rC//rA//rU//rU//rU//rU//rU//rU//rU//rC//rC//rA//rU//rU//rU//rG//rA/TT,/rU//rC//rA//rA//rA//rU//rG//rG//rA//rA//rA//rA//rA//rA//rA//rU//rG//rG//rC/TT, and ASPN (human) siRNA3:/rC//rC//rU//rU//rU//rC//rU//rA//rA//rC//rC//rA//rC//rA//rA//rA//rG//rA//rA/TT,/rU//rU//rC//rU//rU//rU//rG//rU//rG//rG//rU//rU//rA//rG//rA//rA//rA//rG//rG/TT. Both strands of each siRNA, containing the negative control (negative siRNA), were commercially synthesized (GenePharma, Shanghai, P.R.C.) and then annealed to form a double-stranded oligonucleotides. The actual molecular weight differs from the theoretical molecular weight by less than 0.05%. Nanoparticles have been reported to be obtained using a double emulsion method (Yang et al., 2018; Zhou YL et al., 2019). The main component of the nanoparticles is polyis poly (D,L-lactide-co-glycolide) [PLGA, lactide:glycolide (65:35), Mw = 40–75 kDa] from Sigma-Aldrich (St. Louis, MO, United States). The nanoparticles were modified with polydisperse branched polyethyleneimine (PEI, Mw = ∼25 kDa, Sigma, St. Louis, Mo., United States) and loaded with siRNA to attract negatively charged siRNA with a positive charge. The final concentration of nanoparticles obtained was approximately 1 μg/μl. To develop ASPN siRNA/nanoparticles (si-ASPN) and negative siRNA/nanoparticles (NC), the nanoparticle solution was combined with PEI in deionized water. The combination was then gently centrifuged at a N/P ratio of 6:1 (polyethyleneimine molar to DNA phosphate molar) (Figure 5A). Previous studies have demonstrated that keloid formation could be mainly caused by abnormal keloid fibroblasts associated with the microenvironments of keloid lesions (Xianglin Dong ei al., 2013). Our pathological keloid tissue was taken from keloid patients with complete excision of subcutaneous tissue. The samples were then sliced into 5 mm × 5 mm size and digested with 0.25% Dispase II (D6430, Solrabio, China) for 4 h to completely separate the epidermis from the dermis. The epidermal tissue was removed and dermal tissue retained. The retained dermal tissue were minced and digested with 1 mg/ml collagenase type I (C917425, Macklin, China) for 3 h, followed by filtration through 50-µm strainers to obtain fibroblasts. The keloid fibroblasts were subcultured in 10% FBS in DMEM supplemented with 1% penicillin and streptomycin. Culture dishes were placed in a moist atmosphere of 5% CO2 at 37°C.All primary fibroblasts were used before reaching the fifith passage. The 24-well plates were filled with keloid fibroblasts until each well’s cell density reached 70%–80%. Three different types of ASPN siRNA/nanoparticles were diluted with 1 mL of culture medium (1 ml of each contained 2.5 µl of nanoparticles and 2.5 µl of ASPN siRNA). Keloid fibroblasts in well plates were divided into four groups. Keloid fibroblasts that had not been interfered with served as the control group, whereas keloid fibroblasts that had been transfected with one of three different ASPN siRNA/nanoparticle complexes—siRNA1, siRNA2, or siRNA3—served as the experimental groups. They underwent a 24-h CO2 incubation at 37°C and Western blot was used to detect ASPN protein expression. To investigate the transfection effectiveness of the of the ASPN siRNA/nanoparticle (si-ASPN) complex in keloid fibroblasts in vitro. Cells were seeded in 24-well plates until roughly 70%–80% coverage was reached in each well. Using a fluorescent microscope (Leica DMR 3000; Leica Microsystem, Bensheim, GER) equipped with GFP channels, fibroblasts were grown in non-transfected (Control) or ASPN siRNA/nanoparticle (si-ASPN) mixture for 24 h (excitation 488 nm, emission 507 nm). Then, the cells were collected in plates, suspended in PBS, and a FACSCalibur flow cytometer was used to assess the transfection efficiency (BD FACSCalibur, BD Bioscience, San Jose, CA, United States). The EdU (5-ethynyl-2′-deoxyuridine) test was used to detect the keloid fibroblast growth. The total number of cells and the number of EdU-positive cells were counted under a microscope. Positive cells are calculated by dividing the total number of positive cells by the percentage of all cells using ImageJ. To assess keloid fibroblasts’ capacity for cell migration, a scratch test was employed. The state of the wound was assessed every 24 h, and photographs are obtained under a light microscope at 48 h. ImageJ was used to calculate wound healing rate = (0 h scratch area—48 h scratch area)/0 h scratch area x 100%. In accordance with The Affiliated Hospital of Nantong University’s Guidelines for Care and Use of Laboratory Animals, and the Nantong University Animal Ethics Committee Approval to all animal treatment operations (Animal ethics: S20220515-902), animal models for this study included six-week-old Foxn1nu BALB/c male nude mice (Laboratory Animal Center of Nantong University, Nantong, China). Six nude mice were randomly divided into two groups of three each. They were a part of the negative control group (NC) group treated with negative siRNA/nanoparticles and the si-ASPN group treated with ASPN-siRNA/nanoparticles, respectively. In a sterile setting, the diseased keloid tissue was divided into twelve pieces, each weighing roughly 105.2 ± 6.6 mg. A 5-mm transverse incision was made, and a pair of segments were put into the incisions on each side of the nude mice’s spine and the incisions were stitched up with non-absorbable sutures. These nude mice were raised for 7 days for the keloid tissue samples could integrate into the animals’ subcutaneous tissue, and they were freely fed and drink throughout a typical light/dark cycle. After 7 days, keloid tissues including the si-ASPN group were injected separately with a 100 µl mixture containing 24 µl ASPN siRNA/nanoparticles (containing 12 µl nanoparticles and 12 µl ASPN siRNA) and 76 µl normal saline. Keloid tissues in the NC group were injected separately with a 100 µl combination containing 24 µl negative siRNA/nanoparticles (12-L nanoparticles and 12-L negative siRNA), and 76 µl normal saline. The mice were put to death 2 weeks following the injection, and the keloid grafts were collected for examination (Figure 7A). The results are presented as the mean ± standard deviation (M±SD). The threshold for statistical significance was P 0.05. The Student’s t-test was used to analyze differences in protein volume, EdU-positive cells, wound closure, and graft quality. p-value < 0.01 in statistical analysis was considered to be statistically significant by definition. Version 8.0.2 of GraphPad Prism was used to visualize the data (GraphPad Software, San Diego, CA, United States). As shown in Figure 2A, the keloid lesion exceeds the extent of the original skin damage and has a raised, tough surface with a reddish coloration (Figure 2A). According to histological evaluation, the keloid tissue was beyond the boundary of the original wound, and had a claw-like leading edge, which is an invasive edge. The positive percentage of collagen I in keloid was clearly increased compared to normal skin tissue and the keloid tissue was clearly invading inwards (Figure 2B). The positive percentage of inflammatory cells in keloid was significantly higher than in normal skin tissue and the cells were dense and irregularly arranged compared to the normal skin ( Figure 2C ). The collagen fraction volume was also higher than in normal skin tissue. The collagen fibers were large and thick and wavy, with an abnormal and unstable shape in keloid. Normal skin tissue has a limited margin, slender and parallel fibrocytes with regular, thin and uniform linear arrangement of collagen fibers ( Figure 2D ). Therefore, compared with normal skin tissue, keloid has unique characteristics. Sixty-four (64) DEGs were screened between keloid and normal skin through bioinformatics analysis, including 24 up-regulated and 40 down-regulated genes. The results are shown as a volcano map and a star cluster heat map ( Figures 3A,B). Functional enrichment analysis was performed to elucidate their functions and related pathways in keloidogenesis and development. Cellular component (CC), molecular function (MF), and biological process (BP) context was supplied in the GO word analysis of DEGs. It was discovered that the above genes were related to biological processes like “kidney development,” “renal system development,” and “kidney epithelium development” in the BP entry. The DEGs for “extracellular matrix structural constituent giving compression resistance,” “Wnt protein binding,” and “extracellular matrix structural constituent” for GO MF were considerably enriched. The “apical plasma membrane,” “apical portion of the cell,” and “collagen-containing extracellular matrix” were all high in GO CC ( Figures 3C,D ). Wnt signaling pathway, rheumatoid arthritis, and PPAR signaling pathway are the other three ways included in KEGG. A gene set enrichment analysis (GSEA) was further performed to elucidate the key pathways involved in the DEGs ( Figure 3E ). Samples classified as ‘fatty acid metabolism','metabolism of exogenous substances by cytochrome P450’, ‘PPAR signaling pathway','steroid hormone biosynthesis’ and ‘tyrosine metabolism’were the closest in the degree of enrichment. Therefore, it was hypothesized that these genes may affect the development of keloid by participating in the above biological processes and pathways. Combining Lasso Cox regression (Figure 3F), SVM dimensionality reduction algorithm (Figure 4A) and Venn diagrams (Figure 4B), six specific genes were shown, which are respectively ASPN, MEST, F2RL2, ACTC1, KRT19, and IL7. The expression of ASPN gene in keloid group and normal control group were analyzed. The boxplot showed significantly higher expression of ASPN in the keloid group than in normal skin.In analyzing the sensitivity and specificity of ASPN gene, the area under the ROC curve was 0.984. Therefore, ASPN was effective in the detection of keloid (Figures 4C,D). In previous studies, there is a very important connection between ASPN and fibrosis, but there are few studies between keloid and ASPN. Therefore, for the above reasons, we opted for ASPN as the target gene (Huang S et al., 2022; Liu L et al., 2021; Tanaka, 2022). The specific expression of ASPN was confirmed in keloid by RT-qPCR, immunohistochemistry and Western blot analysis; ASPN was strongly expressed in keloids, but the expression signal was negative or weakly positive in normal skin (Figures 4E–G). This was also observed in other studies that examined protein or mRNA levels in keloids (Ong CT et al., 2010; Shih B et al., 2010b). The PLGA siRNA/nanoparticles under scanning electron microscopy (SEM) were nearly spherical in all cases (Figures 5A,B). Western blot analysis was performed to verify whether keloid fibroblasts transfected with si-ASPN/nanoparticles or not expressed ASPN protein. The expression of ASPN was successfully downregulated by all three ASPN siRNAs. The three ASPN siRNAs inhibited 33.9% (p0.01), 51.6% (p0.01), and 64.9% (p0.05) of the protein level in comparison to the negative control group, respectively (Figure 5C). As a result, ASPN siRNA3 was the most successful of the three ASPN siRNAs. As a result, for further trials, we chose ASPN siRNA3. In Figure 5D, cells tagged with ASPN siRNA/nanoparticles (si-ASPN) or untransfected treatment (Control) were depicted in typical fluorescence and bright field views (Figure 5D). The transfection effectiveness of the ASPN siRNA/nanoparticle complexes was evaluated using flow cytometry. The si-ASPN group displayed fluorescence with a strong expression. These outcomes demonstrated the viability of transfecting keloid fibroblasts with ASPN siRNA/nanoparticle complexes. EdU staining was used to assess the proliferation of keloid fibroblasts. The percentage of EdU-positive cells in the si-ASPN group was significantly lower than that in the NC group (p < 0.05). The ratio of EdU-positive cells were 13.7% ± 1.8% and 3.35% ± 1.36% in the si-ASPN group and NC group respectively. The EdU assay results showed that modulation of ASPN inhibited the proliferation of keloid fibroblasts (Figures 6A,B). The migration ability of keloid fibroblasts was evaluated by scratch assay (Figure 6C). After forty-eight (48) h of scratching, the wound healing rate of the keloid fibroblasts treated with si-ASPN (58.76% ± 1.77%) was markedly lower than that of cells in NC group (75.60% ± 2.99%) (p < 0.05) (Figure 6D). The results of the scratch assay indicated that silencing ASPN inhibited the migration of keloid fibroblasts. External morphological changes of the keloid before and after transplantation was observed and quantified (Figures 7B,E). When keloid tissues from keloid patients were separated into almost identical portions and weighed prior to surgery, it was discovered that the si-ASPN group had an initial mass of 83.82 mg while the NC group had an initial mass of 87.37 mg. The mice were executed 2 weeks after injection, the grafts were removed and weighed again. The weight of the grafts in the si-ASPN group was 34.32 ± 25.90 mg, a reduction of 49.50 ± 23.46 mg. The weight of the grafts in the NC group was 68.75 ± 22.64 mg, a reduction of 18.62 ± 19.01 mg. However, due to the vulnerability of the nude mice, two died in the NC group due to infection and other reasons, so ended up with eight samples from the si-ASPN group and four samples from the NC group.Graft quality decreased both in the si-ASPN and NC groups compared to pre-inoculation quality, but the decrease in graft mass was significantly greater in the si-ASPN group than in the NC group (p < 0.01) (Figures 7B,E). To detect the role of ASPN siRNA/nanoparticles, the expression of ASPN protein in tissues was examined from the si-ASPN and NC groups, respectively, using Western blot assays (Figures 7C,D). After isolating and blocking the protein, it was incubated with anti-ASPN antibody. As shown in Figures 7C,D, ASPN expression was significantly downregulated after 2 weeks of treatment (p < 0.01). To ascertain the expression of ASPN protein, samples were immune-stained with ASPN antibody 2 weeks after injection of ASPN siRNA/nanoparticles. Tissue sections were then observed under a microscope and images were recorded. The tissues in the si-ASPN group were all negative or weakly positive for ASPN expression signal, whereas the tissues in the NC groups had a more pronounced ASPN expression signal (Figure 8D ). Hematoxylin and eosin (HE) staining and Masson staining showed that collagen cells and fibroblasts in the si-ASPN group were more regular, slender, parallel and evenly spaced compared to the NC group (Figures 8A,B ). Activated fibroblasts can excrete massive amounts of α-SMA, which promotes the secretion and sedimentation of extracellular matrix (LeeWon et al., 2013; G.Carpino et al., 2005). To a large extent, activated fibroblasts are the exclusive source of collagen I (Michelle Tallquist and Molkentin, 2017; Rana et al., 2020; Qiyao Li et al., 2016). Therefore, 2 weeks after injection, immunohistochemical analysis of α- SMA and collagen I were performed to investigate the effect of ASPN-siRNA/nanoparticle for further study (Figures 8C,E). By contrast, the use of the ASPN siRNA/nanoparticle complexes was effective in reducing fibroblast levels (Figure 8F). These results generally showed that silencing of ASPN resulted in modest antiproliferative of fibroblasts and promoting the more orderly arrangement of collagen fibers. In this study, the transcriptome sequencing results of keloid and normal skin were compared. Bioinformatics analysis was used to screen a total of 64 DEGs, of which 24 showed up-regulation and 40 showed down-regulation, and the “Wnt signaling pathway” was obtained, among others. Combining Lasso Cox regression, SVM dimensionality reduction algorithm and Venn diagrams, six specific genes were screened out.We then selected ASPN, one of the six genes, for a follow-up study. Research proves that keloid tissues had considerably higher levels of ASPN protein expression and that the Wnt signaling pathway maybe play an important role in keloid development. This suggests that the ASPN gene may control the developmental rate of keloids by modulating the Wnt signaling pathway. There has been a very important association between ASPN and fibrosis in previous studies, but the relationship between keloids and ASPN has been sparsely studied (Huang S et al., 2022; Liu L et al., 2021; Tanaka, 2022). Due to the rapid development of mechanical biology and the research on the formation and progress of keloids promoted by mechanical forces, the ECM molecular ASPN can also become an effective therapeutic target (Huang C et al., 2012a; Huang C et al., 2012b; Huang C et al., 2013; Deng Z et al., 2021; Syed F et al., 2012). Keloid are also strongly associated with fibroplasia (Huang C et al., 2014; Huang and Ogawa, 2020; Jonathan et al., 2016; DorothySupp et al., 2012). Based on these conclusions, we constructed siRNA/nanoparticle complexes containing ASPN siRNA to investigate the effects of ASPN gene regulation on the keloid fibroblasts. In fibroproliferative diseases, fibroblasts are responsible for increased ECM stiffness (Nelesen et al., 2019). α-SMA is a well-known signature protein used to assess activated fibroblasts in a variety of tissues and organs (LeeWon et al., 2013; G. Carpino et al., 2005). Collagen is the most plentiful structural protein in the human body. While type I collagen (COLI) is the predominant collagen type in connective tissue (Li W et al., 2021). The expression of COLI was significantly higher in keloids compared to the normal group. Therefore, COLI can precisely reflect the growth rate and status of keloid fibroblasts, and its down-regulation can suppress the proliferation of keloid fibroblasts. According to our findings, the si-ASPN group’s -SMA and COLI levels were significantly lower than those of the NC group. Additionally, the EdU assay demonstrated that inhibiting ASPN caused keloid fibroblasts to stop proliferating. This demonstrates that keloid fibroblast proliferation and metabolic activity can be inhibited by a reduction in ASPN protein expression. Our study developed a xenograft mouse model, and after 14 days, it was discovered that the expression level of ASPN protein in the si-ASPN group had been significantly lower than in the NC group. In comparison to the NC group, the collagen in the si-ASPN group was more homogeneous, slender, parallel, and regularly arranged. This shows that si-ASPN can be delivered from the nanoparticles into the keloid grafts continuously and steadily. The si-ASPN/nanoparticle complexes may therefore be a useful kind of gene therapy for the treatment of keloid. Our investigations’ uniqueness is the utilization of nanoparticles as vector or vehicle to transport si-ASPN. Numerous studies are being done on nanoparticles as non-viral gene delivery vehicles because of their biocompatibility and better regulatory release. Because PLGA nanoparticles can deliver their contents for a long time, they could be used to release genes into keloid fibroblasts (Delgado D et al., 2012; Wong SP et al., 2012; XingGuo et al., 2020). Thus, si-ASPN can be slowly released into keloid tissues and exert therapeutic effects. In fact, both the NC and si-ASPN groups had a decline in graft quality after transplantation, although this decline in graft quality was noticeably larger in the si-ASPN group than in the NC group, according to our findings. We hypothesize that the aforementioned may be related to the immunological health and dietary needs of nude mice. According to the outcomes of our investigation, xenografts were inhibited in the same way by both groups of age-matched nude mice, while the si-ASPN group had a greater decrease in xenograft quality, indicating that si-ASPN has a certain effect. Additionally, earlier research has demonstrated that the ASPN gene is present in both humans and animals, and that si-ASPN functions in xenografts made from nude mice. As a result, we speculate that si-ASPN may have some sort of function in human keloid tissue (Maccarana M et al., 2017). There are, of course, several limitations to this experiment. One of these limitations is that nude mice only exhibit T lymphocyte immunodeficiency, B cells and NK cells are still functional, and their immune systems are still capable of blocking the growth of external tumors (Wu et al., 2017). Furthermore, for xenografts implanted subcutaneously in nude mice, the blood supply, lymphatic reflux, and delivery of numerous nutrients necessary for xenograft growth are insufficient, which has an impact on the growth of the transplanted xenograft. Moreover, keloid is a progressive proliferative disease, which recurs repeatedly and has a long course of disease. Inadequately, due to the limitations of this experiment, the long proliferation cycle of tissues was not fully expressed in the nude mice experiment. Thus, we assumed that the proliferation cycle of nude mice tissues is very long. Second, we only investigated the effect of ASPN proteins on the regulation of keloid. Whereas keloid genesis may be the result of a combination of factors (Macarak et al., 2021; Jennings et al., 2022; Kumar and Kamalasanan, 2021). Also, as keloids may be linked to genetic factors (Shih and Bayat, 2010a; Glass, 2017) and our numbers are small, it is not possible to strongly demonstrated the significance of ASPN. Based on the results of the trial outside of our animal experiments, further experimental validation remains to be done on whether the results are valid for humans. Besides, some studies have shown that the Wnt signaling pathway is thought to be the main cause of fibrosis in different organs (Bowley et al., 2007). Therefore, the Wnt signaling pathway may play an important role in keloid. In the upcoming experiments, the role of the Wnt signaling pathway in keloid may be another novel direction of research. In this study, keloid fibroblasts and tissues were transfected with ASPN-siRNA to decrease the expression of ASPN protein. The results of the study showed that the growth and metabolic activity of keloid fibroblasts were declined significantly. This suggests that ASPN siRNA/nanoparticles may provide a novel therapeutic strategy for the management of keloid.

PMC9649828

36386611

Tiantian Ma,Xinjun Zhang,Ruihong Wang,Rui Liu,Xiaoming Shao,Ji Li,Yuquan Wei

Linkages and key factors between soil bacterial and fungal communities along an altitudinal gradient of different slopes on mount Segrila, Tibet, China

28-10-2022

soil microbial communities,Tibetan Plateau,elevational gradient,slope,microbial interaction,environmental influencing factor

Soil microbes are of great significance to many energy flow and material circulation processes in alpine forest ecosystems. The distribution pattern of soil microbial community along altitudinal gradients is an essential research topic for the Tibetan Plateau. Yet our understanding of linkages between soil microbial communities and key factors along an altitudinal gradient of different slopes remains limited. Here, the diversity, composition and interaction of bacterial and fungal communities and in response to environmental factors were compared across five elevation sites (3,500 m, 3,700 m, 3,900 m, 4,100 m, 4,300 m) on the eastern and western slopes of Mount Segrila, by using Illumina MiSeq sequencing. Our results showed that microbial community composition and diversity were distinct at different elevations, being mainly influenced by soil total nitrogen and carbonate. Structural equation models indicated that elevation had a greater influence than slope upon the soil microbial community. Co-occurrence network analysis suggested that fungi were stable but bacteria contributed more to among interactions of bacterial and fungal communities. Ascomycota was identified as a key hub for the internal interactions of microbial community, which might affect the soil microbial co-occurrence network resilience of alpine forest ecosystems on the Tibetan Plateau.

Linkages and key factors between soil bacterial and fungal communities along an altitudinal gradient of different slopes on mount Segrila, Tibet, China Soil microbes are of great significance to many energy flow and material circulation processes in alpine forest ecosystems. The distribution pattern of soil microbial community along altitudinal gradients is an essential research topic for the Tibetan Plateau. Yet our understanding of linkages between soil microbial communities and key factors along an altitudinal gradient of different slopes remains limited. Here, the diversity, composition and interaction of bacterial and fungal communities and in response to environmental factors were compared across five elevation sites (3,500 m, 3,700 m, 3,900 m, 4,100 m, 4,300 m) on the eastern and western slopes of Mount Segrila, by using Illumina MiSeq sequencing. Our results showed that microbial community composition and diversity were distinct at different elevations, being mainly influenced by soil total nitrogen and carbonate. Structural equation models indicated that elevation had a greater influence than slope upon the soil microbial community. Co-occurrence network analysis suggested that fungi were stable but bacteria contributed more to among interactions of bacterial and fungal communities. Ascomycota was identified as a key hub for the internal interactions of microbial community, which might affect the soil microbial co-occurrence network resilience of alpine forest ecosystems on the Tibetan Plateau. Soil microbial communities have an important role in energy flow and material circulation processes in terrestrial ecosystem, whose diversity can be used to characterize many aspects of the soil environment (Zhang et al., 2018). Bacterial and fungal communities has received much interest over recent decades with a view to identifying key biotic drivers of soil ecosystem services (Rashid et al., 2016). For example, bacteria and fungi as decomposers contribute to the regulation of organic matter and nutrient cycling, but have diverse substrate preferences and metabolic strategies, leading to distinctive organic carbon accumulation and stabilization outcomes in soils (Soares and Rousk, 2019). Bacteria and fungi usually share same microhabitats in soil but may have divergent ecological functions. Both bacterial and fungal communities are sensitive to the impacts from environmental disturbances entailing both biotic and abiotic factors, such as temperature, water content, pH, etc., which may in turn influence ecosystem stability (de Menezes et al., 2017; Wei et al., 2018). Therefore, focusing on co-occurring bacteria and fungi in soil as well as their interactions in response to varied environmental conditions could contribute to a more complete understanding of soil microbes’ respective roles. The Tibetan Plateau, as the “third pole of world,” is crucial for global biodiversity and climate change due to the complex interactions there among environmental, cryospheric, and geographic processes (Wang et al., 2020). Abundant forestry resources can be found in the southeastern part of the Tibetan Plateau where alpine forest ecosystems fulfill an important role in maintaining the water balance, carbon sequestration, and even ecological security (Kai et al., 2016). Given the high average altitude, large elevation span, and seasonal freezing and thawing events of alpine forests in Tibetan Plateau, pronounced fluctuation patterns along elevational gradients and their associated environmental conditions may contribute differentially to microbial diversity and activity (Yue et al., 2016). Hence, obtaining a better understanding of the complex interactions among bacterial, fungal, and abiotic components in alpine forest soils is one of most challenging issues due to harsh geographic and climatic conditions, but is likely crucial to predict the stability of microbial networks’ response to future climate change (de Vries et al., 2018). Distribution patterns of microbial communities along elevational gradients are an essential topic in biogeography, especially for mountains in the Tibetan Plateau, which have steeply compressed gradients in terms of abiotic and biotic conditions across short geographic distances (Bragazza et al., 2015; Yao et al., 2017). It is widely reported that the diversity and composition of soil microorganisms will vary across differing elevations (Bahram et al., 2012; Siles and Margesin, 2016; Ren et al., 2018). For example, work by Yang et al. (2014) indicated that the diversity of soil microbial genes in grasslands changes along an elevation gradient on the Tibetan Plateau. Klimek et al. (2015) found that soil bacterial catabolic activity in Beskidy Mountains was mainly affected by organic matter and dissolved organic nitrogen. Furthermore, autotrophic microbial composition in grassland soils on the Tibetan Plateau can markedly change along an elevational gradient and is jointly affected by temperature, nutrients, moisture, and plant types (Guo et al., 2015). A part of studies found that there is no general altitudinal pattern in bacterial diversity, and fungal biodiversity changed by ecological processes such as temperature along forest altitude gradients (Tian et al., 2017; Wang et al., 2017). Some studies showed that aboveground community composition and soil moisture played determining roles in restoring microbial community (Xu et al., 2014; Duan et al., 2021). Considering that altitude-driven environmental conditions on differing slopes of mountains may represent an analogue to ecological changes in different latitudinal-driven climatic zones (Diaz et al., 2003), that different variables act as key factors determining microbial community variation in mountains may be due to differing elevation ranges and neglecting the orientation of mountain slopes. In temperate regions of Asia, east- and south-facing mountain slopes may favor plant or microbial richness due to less cold temperatures, in comparison with western and eastern slopes of same mountains, especially in high altitude areas (Huang et al., 2015; Kumar et al., 2019). However, research on elevational patterns of microbial community structure and microbial interactions in alpine forest ecosystems on different slopes of mountains above 3,500 m a.s.l. remains rather limited. In this study, Mount Segrila, located in the center area of a pristine forest tract in southeastern Tibet (Li et al., 2014), was selected to investigate the spatial patterning of bacterial and fungal community composition and diversity in soils at high elevation, ranging from 3,500 to 4,300 m a.s.l. Mount Segrila has a discernible and relatively geologically uniform altitude gradient, harboring vegetation representative of typical alpine forest on the Tibetan Plateau (Xu et al., 2014). Here, we used the Illumina MiSeq platform to compare the patterns of microbial communities at five elevations (3,500 m, 3,700 m, 3,900 m, 4,100 m, 4,300 m) on the eastern and western slopes of Mount Segrila. This study had three objectives: (1) to compare the diversity and composition of bacterial and fungal communities along an elevational gradient on different slopes; (2) to understand relationships between biotic and abiotic factors across elevations and between slopes; (3) to detect linkages of bacterial and fungal communities to key factors affecting the modes of microbial interactions. This study was carried out at Mount Segrila (29°10′ N–30°15′ N, 93°12′ E–95°35′ E), situated in Linzhi County, Tibet, China. Harboring typical montane frigid-temperate forest of southeastern Tibet, Mount Segrila has a peak elevation is >5,200 m a.s.l. and stark differences in climate and terrain between its east and west slopes. The climate here is humid and cool, with a mean annual temperature of −0.73°C, mean annual precipitation of 866 mm (mainly falling from April to October), and an annual evaporation of 544 mm (Xu et al., 2014). Soil and vegetation types in the studied elevational range are as follows: dark-brown forest soil (luvisols) from 3,500 to 4,200 m, with frigid, dark coniferous forests; black mattic soil (cambisols) from 4,200 to 4,500 m, with alpine sub-frigid shrub meadows; mainly forest between 2,700 and 4,300 m consisting of the conifer Abies georgei var. smithii. Shrubs such as Sabina saltuaria are dominant at higher elevations of 4,300–4,500 m. Soil samples were collected from the five elevations (3,500 m, 3,700 m, 3,900 m, 4,100 m, and 4,300 m) of Mount Segrila, on both its east slope and west slope, in August 2020 (Figure 1). Soil samples were collected from nearby forest (within a 30-m distance). Fresh surface soil samples (each 20 cm × 20 cm, 0–15 cm in depth) were obtained after carefully removing the litter layer from each quadrat (n = 3 replicates per elevation site). Each quadrat’s soil sample consisted of mix of five subsamples randomly taken within a 10-m radius. All samples were passed through a 2.0-mm sieve to remove visible stones, invertebrate animals, large root fragments, and plant materials. One part of every sample was stored in a refrigerator at −80°C for DNA extractions, and the other part air-dried at room temperature to analyze the soil physical and chemical properties. Various soil physicochemical characteristics were determined, namely total organic carbon (TOC), total nitrogen (TN), C/N ratio, available phosphorus (AP), available potassium (AK), carbonate content (CO32−), and pH. Soil pH was measured in a water-to-soil mass ratio of 5:1, using a pH meter. TOC and TN were measured with an elemental analyzer (Model EA1108, Carlo Erba, Torino, Italy; Wei et al., 2019). Soil AP was extracted with 0.5 M NaHCO3 and assayed by applying the ascorbic acid/molybdate reagent blue color method (Wei et al., 2018). Soil water content was quantified as the mass lost after drying at 105°C for 24 h. To determine the AK content, the ammonium acetate extraction-flame photometric method was used (Li et al., 2014); for carbonate content, it was determined by measuring the volume of CO2 gas released after applying an HCl treatment to the soil (Luo et al., 2015). Soil mechanical composition was assessed according to methodology of the Chinese Agricultural Industry Standard (NY/T 1121.3–2006). Soil DNA was extracted using a FastDNA Spin Kit for soil (MP, Biomedicals, Santa Ana, Carlsbad, CA, United States), by following the manufacturer’s instructions. Purified DNA from each sample was PCR-amplified with the primer pair 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) for the bacterial 16S rRNA gene, and likewise by the primer pair ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2R (5′-GCTGCGTTCTTCATCGATGC-3′) for the fungal ITS1 region, on an ABI GeneAmp® 9,700 PCR thermocycler (ABI, CA, United States). The ensuing PCR products were collected using AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States). Purified amplicons were pooled in equimolar amounts and then paired-end sequenced on an Illumina MiSeq PE300 platform/NovaSeq PE250 platform (Illumina, San Diego, United States) by the Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China), this done according to standard protocols. All obtained raw reads were deposited into NCBI Sequence Read Archive (SRA) database (accession number: SRP331508). After their sequencing, the operational taxonomic unit (OTU) assignment of the reads was carried out using >97% as the shared identity threshold, by the USEARCH (version 8.0.1623) from which representative sequences of OTUs were assigned for taxonomic analysis. Descriptive statistics including the mean, standard deviation (SD), minimum and maximum values, and the coefficient of variation (CV) were calculated in MS Excel software. Two-way analyses of variance (ANOVA) and multiple comparison tests (LSD [least significant difference]; all-pairwise comparisons test) were used to compare differences for different variables and test significance of differences (p < 0.05). For these statistical analyses, we used SPSS v 25.0 software (SPSS Inc., Chicago, IL). Data for physical and chemical indicators were using Origin 2021 software. To express taxonomic alpha diversity, the Chao1, Shannon, and heip indexes were calculated in QIIME 2. Principal coordinate analysis (PCoA) was employed to graphically explore differences in microbial community structure between different elevation sites, based on Bray–Curtis distances using OTUs dataset, using Canoco v5.0. Redundancy analysis (RDA) with forward selection was also carried out in Canoco v5.0 software, to analyze the relationships between microorganisms and environmental factors. A “robust” maximum likelihood evaluation program was used to analyze a fitted structural equation model (SEM), in AMOS 24.0 software, based on several criteria: Chi-square (p > 0.05), high goodness-of-fit index (GFI > 0.90), and root mean square errors of approximation (RMSEA <0.05). Microbial co-occurrence networks of mixed fungal and bacterial communities were constructed based on the relative abundances of OTUs. For this, the most abundant species (i.e., top 10%) in the bacterial and fungal communities were selected, that is 860 spp. of bacteria, 323 spp. of fungi. We used the MENA pipeline, which implements the random matrix theory (RMT) approach, to construct the correlation-based relevance network of microbes (Zhou et al., 2010; Deng et al., 2012). A minimum threshold of 0.08 was used for this co-occurrence construction; our inspection of p-values for the calculated r-values of each built network found them all to be <0.05. The co-occurrence network analysis was then conducted in Gephi v0.9.2 according to Pearson correlation coefficients (p < 0.05, |r| > 0.8). Relationship maps of key environmental factors and key microbial species were also analyzed in Gephi 0.9.2 (p < 0.05, |r| > 0.5). Results revealed that the physicochemical properties of soils along the elevational gradient differed significantly between western and eastern slopes (Table 1). The pH level at each sampling site was acidic, ranging from 4.7 to 6.1, but soil pH was slightly lower on the western than eastern slope, on average. The lowest pH for both slopes occurred at an elevation of 4,100 m. In most sample sites, pH decreased as the elevation increase except for 4,300 m on the eastern slope. TOC and TN in soil showed no significant trends along the elevation gradient of either slope. TOC was similar among elevation sites, ranging from 42 g/kg to 82 g/kg, while the TN varied between 3.2 g/kg and 4.6 g/kg. For both slopes, their highest TOC and C/N values were observed at 4100 m. The content of AK as well as that of CO32− fluctuated with rising elevation, peaking at 3900 m on the western slope. The highest AP content was observed in soil at the lowest elevation on the western slope; AP ranged from 1.9 to 8.2 mg/kg across all sites. There was a significant correlation between pH, TOC, C/N, and CO32− (p < 0.05), but these indices were not significantly associated with elevation or slope. The proportion of sand (0.02–2 mm in size), including coarse sand and fine sand particles, was highest in all soil samples, accounting for ~80% of the mechanical fractions (Supplementary Figure S1; Supporting Information). Illumina sequencing of bacterial 16S rRNA gene amplicons resulted in 465,348 high-quality paired-end reads. After removing the short low-quality reads, singletons, replicates, and chimeras, a total of 3,741 OTUs were obtained from all soil samples based on at least 97% similarity. Sequencing of fungal ITS amplicons led to an initial total of 497,916 paired-end reads; after applying similar data treatments used for bacterial 16S rRNA, a final total of 39,676 sequences remained that clustered into 1,500 OTUs. Together, the bacterial and fungal reads accounted for more than 98% of all obtained soil sequences (Table 2). A rarefaction analysis based on the number of observed species suggested that sequencing depth of bacterial and fungal communities was sufficient for carrying out robust downstream microbial community diversity analyses. According to the alpha diversity indices, microbial diversity along elevational gradient of the two slopes was distinct. With respect to the bacterial community, Chao1, Shannon, and heip indices at 3500 m exceeded those of other elevation sites on the western slope. However, soil bacterial community richness, diversity, and evenness at lower elevations, especially 3,700 m, were less than those at higher elevations (i.e., 4,100–4,300 m; Table 2). Whether on the east- or west-facing slope, fungal diversity and richness were greatest at 3500 m, being relatively reduced at middle elevations (i.e., 3,700–4,100 m). PCoA was also used to visualize the dissimilarity matrix of bacterial OTUs among all sampling sites (Supplementary Figure S1; Supporting Information). The first two ordination axes together explained 74% of variability found in bacterial composition. Distinctive β-diversity of microbial communities was generally apparent across the elevation gradient, but those on same slope clustered more, especially for bacteria. In total, we detected 37 bacterial phyla and 13 fungal phyla in all soil samples. As Figure 2 shows, Proteobacteria was the most abundant phylum in soil of both eastern and western slopes. The relative abundance (expressed as proportion) of Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi was above 60%, and they were dominant members of bacterial communities in all soil samples. Nevertheless, bacterial composition varied greatly among elevation sites. The relative abundance of Actinobacteria increased significantly at 3500 m and 4,300 m, whereas that of Proteobacteria or Acidobacteria was obviously enhanced at middle elevations. The prevalent fungal phyla were Basidiomycota, Ascomycota, Mortierellamycota, and Rozellomycota. Ascomycota was most abundant at 3500 m and 4,300 m, especially on the western slope. The greatest relative abundance of Mortierellamycota was found in soil at 4100 m on the eastern slope and at 4300 m on the western slope, while Basidiomycota were more abundant at middle elevations (3700–4,100 m). SEM as shown in Figure 3 indicated that elevation and slope directly affected bacterial community composition (p < 0.01) as well as soil physiochemical properties, which indirectly influenced bacterial composition and diversity. The standardized effect of slope (>70%) on the soil bacterial community was positive and surpassed that of elevation on Mount Segrila. By contrast, elevation had a significantly positive effect on fungal diversity but a negative one upon fungal composition (p < 0.01), with no significant correlations between slope and fungal community detected. Standardized total effects of elevation and slope on the fungal community were negative, suggesting different modes by which environmental factors influenced bacterial and fungal communities. In summary, it was found that both elevation and slope had an obvious effect on soil microbial communities but the former exerted a greater influence than the latter upon microbial community composition and diversity, while the bacterial community was more sensitive than fungal community in responding to environmental changes along the elevational gradient of different slopes on Mount Segrila. According to the PCoA results (Supplementary Figure S1; Supporting Information), bacterial communities of soil samples at 3500 m and 4,300 m clustered together, while those from 3,700, 3,900, and 4,100 m were more similar. According to clustering analysis, two fungal–bacterial co-occurrence networks were constructed to investigate differential biotic interactions. The microbial co-occurrence network of soil taken from middle elevations (3,700 m, 3,900 m, and 4,100 m) incorporated 162 nodes and 276 edges (Figure 4A), while the other network had 115 nodes and 172 edges (Figure 4B), exhibiting a U-shape across the elevational gradient. In these fungal–bacterial co-occurrence networks, bigger nodes had stronger correlations and more edges with other nodes, implying they sustained more complex microbial interactions as key microbes. Comparing the networks among different elevation sites, seven core fungal species were ubiquitous: Sebacina sp., Pleotrichocladium opacum, Mortierella humilis, Clavaria falcata, Clavaria sp., Saitozyma sp., and Goffeauzyma gastrica (Supplementary Table S2, S3; Supporting Information), but the key bacteria linked with these fungal species differed among the networks (Supporting Information). This result suggested these hubs, especially core fungi, might be stable and responsive to environmental changes associated with varying elevation and slope orientation as keystone taxa, which may continually reconstruct the fungal–bacterial co-occurrence network based on transmission effects. The RDA let us explore the relationships between soil properties and microbial communities, and to distinguish key environmental factors influencing bacteria and fungi. Monte Carlo tests for all canonical axes were significant (p < 0.05), and the first two axes accounted for 30 and 28% of variance in the bacterial community and fungal community, respectively. The forward selection indicated that TOC, CO32−, and TN were main drivers of changes to bacterial community composition (p < 0.05; Figure 5A), while TN and CO32− contributed significantly to fungal community composition (p < 0.05; Figure 5B). Soil nutrient features (e.g., TN, TOC) were significantly correlated with bacterial diversity as well as the relative abundances of Acidobacteriota, Planctomycetota, and Proteobacteria, while those of Desulfobacterota and Firmicutes were strongly correlated with the CO32− content of soil. Concerning the fungi, Ascomycota and Rozellomycota were associated with variation in soil CO32−. Next, the interactions between key soil physiochemical properties and individual species were visualized by building a co-occurrence network. This was based on only strong (|r| > 0.6) and significant (p < 0.05) correlations; that is, prominent relationships between key environmental factors (TN, TOC and CO32−) and bacteria at the species level (Figure 5C) as well as correlations between TN, CO32−, and fungal community members (Figure 5D). The bacterial network contained 297 nodes and 320 edges, in which CO32− connected more with bacteria (172 nodes), followed by TOC (97 nodes) and TN (50 nodes). Key bacterial species in the bacterial–environmental network changed significantly vis-à-vis the fungal–bacterial interaction network (Supplementary Table S4; Supporting Information), indicating that soil nutrient features of differing slopes and elevations may influence the bacterial community more and thereby lead to a reorganization of the core fungal–bacterial network. The fungal network consisted of 57 nodes and 55 edges, having more nodes linked to TN than CO32−. Intriguingly, one species (Clavaria sp.) that had significant correlations with CO32− was also a core species in fungal–bacterial interaction networks (Supplementary Table S4; Supporting Information). Generally, there were more bacteria (> 60%) than fungi in each microbial–environmental network or fungal–bacterial network, suggesting that bacteria contribute more to reorganizing the co-occurrence network along an elevation gradient (3500–4,300 m) on different slopes of Mount Segrila. Accumulating studies show that the diversity and composition of soil microbial communities varies with elevational distribution characteristics (Xu et al., 2014; Wang et al., 2017; Cui et al., 2019). Climatic conditions can change considerably across different elevations, resulting in differences in vegetation and soil physicochemical properties, which further affects the growth, activity, and composition of soil microbial communities. The climate characterizing high-elevation mountain regions usually confers more environmental stress to affect soil properties, which often has much more of an impact on a soil microbial community than either abiotic or biotic factors such as vegetation (Wang et al., 2020). In this study, soil physicochemical properties as well as microbial communities differed greatly along the sampled elevational gradient on east- vs. west-facing slopes of Mount Segrila, lacking any significant linear relationships (Table 1). SEM results revealed that the elevational gradient significantly shaped both the composition and diversity of bacterial and fungal communities yet slope only affected bacterial community composition (Figure 3). suggesting complexity of environment when considering slope of mountain and variability of bacterial and fungal communities together. Accordingly, this study emphasizes the effects of high elevation on differently oriented slopes upon soil physiochemical properties and microbial community structure, especially the interaction of bacterial and fungal communities. Soil bacterial community along the elevational gradient of two different slopes in Mount Segrila was significantly affected by TN, TOC, and CO32−, while TN and CO32− had distinct effects on fungal community composition. Although different elevations of Mount Segrila can all support alpine forest ecosystem of the Tibetan Plateau, they can differ in the vegetation communities formed there (Li et al., 2014). This could generate disparate nutrient inputs to soil from various litter types of dark coniferous forests versus broadleaved and coniferous mixed forests that may function as microbial habitat filters over a short span of elevation. Meanwhile, local environmental conditions (e.g., light and moisture) can directly affect the activity and growth of vegetation, for example the dynamics of root growth (Huang et al., 2015), which could alter the soil nutrient cycle and influence the microbial community at same elevation due to shady or sunny slopes in the sampling season (summer) of our study. Still, most studies to date have concluded that pH significantly affects bacterial and fungal diversity (Yao et al., 2017; Shen et al., 2020). It is worth noting that CO32− is also considered as another index of soils’ acid–base condition. A reason for why CO32− had an impact upon microbial community instead of pH per se could be that CO32− is associated with nutrient ions in soil, confirming nutrients’ content could greatly influence the soil microbial community of Mount Segrila. And carbonate is an important tool for soil ecosystems to cope with climate change due to its essential role in regulating soil pH (Luo et al., 2015), so microorganisms may respond to climate change by adapting to carbonate content in alpine forests with extreme climate variation. Many studies have reported bacteria and fungi having distinct network patterns in response to differences in environmental factors along elevational gradients but few have concentrated on biotic interactions within soil bacterial and fungal communities (Allison et al., 2010; Yao et al., 2017). For example, Ho et al. (2016) suggested that microbial community composition and nitrogen cycle are closely connected. In our study, a co-occurrence network was constructed to visualize biotic interactions within alpine forest soil along an elevational gradient on Mount Segrila. Interestingly, seven fungal species were present and stable in all sites and had strong correlations with other core microbes, such as Sebacina sp., Pleotrichocladium opacum, Mortierella humilis, to name a few, indicating those species play critical roles in the biotic network of soils on Mount Segrila, also being capable of strong environmental adaptability at 3500–4300 m elevation. Most of them belonged to Ascomycota (42.86%), which were mainly influenced by CO32− and the decomposition of soil organic matter (Figure 5). The Clavaria sp. (phylum Ascomycota) was designated a core microbe in Mount Segrila, given its high proportion in the derived fungal–bacterial interaction network, implicating a large influence on organic matter transformation and nitrogen cycling. Core bacteria connected to seven core fungal species in fungal–bacterial interaction network was dissimilar at different elevations, which suggests these bacterial species might serve as a reservoir of genetic and functional diversity and/or buffer ecosystems against species loss or environmental change (Brown et al., 2009). In the bacterial–soil network, we found evidence for some species acting as a “reservoir,” these being correlated with TOC and CO32−. The bacteria linked to TOC all belonged to Acidobacteriota or Actinobacteriota, illustrated that these two phyla contributed more to soil organic carbon cycling in Mount Segrila by influencing biotic interactions. Meanwhile, organic carbon compounds in soil had a significant effect upon its bacterial community, contributing chiefly to the soil core microbial interaction network yet not responsible for sustaining the stability of the microbial co-occurrence network. Although soil microbial community composition and diversity were affected by similar factors (i.e., TN and CO32−) driven by the elevational gradient in this study, the responses of bacterial and fungal community were clearly distinct (Figure 3). We found a stronger effect of elevation characteristics on the composition of the bacterial community than that of the fungal community, perhaps due to the greater sensitivity of bacterial communities (Ren et al., 2018). Soil nutrients’ content can be negatively correlated with bacterial diversity, especially heterotrophic bacteria, at higher altitudes (Kumar et al., 2019). Earlier research showed that diazotrophs have selective advantages under low-nutrient environment (Chan, 1994). Most bacteria that are heterotrophic microbes could mineralize organic N compounds and produce various forms of available N for uptake by roots, leading to tight connections between bacteria and forms of soil nutrients (Kuypers et al., 2018). Conversely, N nutrients tend to exert negative effects on fungal community diversity and composition (Cox, 2010; Pellissier et al., 2014; Wang et al., 2015; Siles and Margesin, 2016; Yao et al., 2017) because this is linked with nutrient deposition. Enhanced TN in soil along an elevational gradient due to differing quality litter inputs and nitrogen deposition levels could drive more bacterial growth and diversity given that bacteria are thought to require more N per unit biomass C accumulation in comparison with fungi (Deng et al., 2016). In this study, we found that Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi were dominant in soil, accounting for more than 60% of total bacterial community, similar to survey results reported for the Taibai Mountain and some temperate forests (Ren et al., 2018; Ma et al., 2019; Bayranvand et al., 2021). Microbial community structure at the phylum level at different elevations was distinct but the distribution of bacterial phylum was similar to results from the Changbai Mountain (Heděnec et al., 2019), perhaps because both areas have the same low pH in soil. On Mount Segrila, there was pronounced variation in the relative abundances of Actinobacteria and Acidobacteria along the elevational gradient; this could be due to the slow-growing characteristics, large genomes, and preference for oligotrophic environments of these taxa (Johnston et al., 2019; Lladó et al., 2019). Acidobacteria reportedly prefer a dry and warm environment and has a lower relative abundance under moister and colder soils (Placella et al., 2012; Lauber et al., 2013). Interestingly, Acidobacteria was more common at high altitudes and western slope, unlike Ren et al. (2018) who found the abundance of Acidobacteria was greater at low altitudes in warm and low-humidity environments. In our study, the Actinobacteria was more abundant at lowest elevations and relatively less so at higher elevations, likely because Actinobacteria spp. are oligotrophic and common in environments with limited soil nutrients, such those typically found at high elevations in mountains regions (Yao et al., 2017; Johnston et al., 2019). Compared with its bacterial community, the fungal community of Mount Segrila’s soils was more stably distributed along the elevational gradient, with Basidiomycota and Ascomycota comprising more than 80% of sampled fungi. Many researchers believe the distribution of fungal community composition is influenced principally by climatic conditions and soil physical properties (e.g., Vetrovsky et al., 2019). For example, fungal communities changed substantially under experimental warming conditions, indicating that climate change could significantly impact soil fungi (Allison et al., 2010; Ihsan et al., 2021). In the present study, however, the abundance of those two phyla did not show a clear geographical signature, for which a plausible explanation could be the dense needle-leaf litter in alpine forest of Mount Segrila. Furthermore, it is widely reported that the main ecosystem role of Basidiomycota and Ascomycota is to decompose organic matter as well as inter-root sediments, a process influenced by organic carbon content of soil (Hannula et al., 2012; Clemmensen et al., 2013). Because of the low temperature and slow decomposition of litter, the surface of the soil in Mount Segrila is covered with a thick layer of litter after years of accumulation. As fungi’s main function is to decompose organic C compounds occurring in recalcitrant forms (Treseder et al., 2016) and it can be stable in extreme climates, so fungi such as Basidiomycota and Ascomycota play an important role at high elevations. The results here demonstrated that soil physiochemical characteristics and microbial communities vary considerably along an elevational gradient on Mount Segrila. The Proteobacteria, Actinobacteria, Acidobacteria, and Chloroflexi all occur in high proportions in the soil bacterial community, while the Basidiomycota and Ascomycota are major phyla in the soil fungal community. Fungal diversity and composition are both correlated with TN and CO32−, while the bacterial community is jointly influenced by TN, TOC, and CO32−. Elevation rather than slope orientation is the main driver of soil microbial community properties of Mount Segrila. Biotic interactions between core bacteria and fungi as affected by edaphic factors could be deconstructed and reorganized along the elevational gradient. Fungi, especially Ascomycota, are important for shaping and sustaining the interrelationship of bacterial and fungal co-occurrence networks. Fungi are evidently stable, but bacteria contribute more to reorganizing the soil microbial network across the elevation span of 3,500–4,300 m on Mount Segrila. The original contributions presented in the study are publicly available. This data can be found at: NCBI Sequence Read Archive (SRA) database (Accession Number: SRP331508). TM: methodology, data curation, conceptualization, and funding acquisition. XZ: data curation, visualization, and writing—original draft. RW: conceptualization, formal analysis, validation, and writing—review and editing. RL: methodology and visualization. XS: validation and funding acquisition. JL: conceptualization and resources. YW: conceptualization, funding acquisition, resources, and project administration. All authors contributed to the article and approved the submitted version. This work was financially supported by the Open Research Fund from the Key Laboratory of Forest Ecology in Tibet Plateau (Tibet Agriculture and Animal Husbandry University), Ministry of Education, China [grant XZAJYBSYS-2020-02], National Natural Science Foundation of China (42007031 and 31960013), Joint project of China Agricultural University and Tibet Agricultural & Animal Husbandry University, and Central government guides local projects (XZ202101YD0013C), The Independent Research Project of Science and Technology Innovation Base in Tibet Autonomous (XZ2022JR0007G). The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649842

Hassan Rasouli,Fatemeh Dehghan Nayeri,Reza Khodarahmi

May phytophenolics alleviate aflatoxins-induced health challenges? A holistic insight on current landscape and future prospects

28-10-2022

Climate change,Aflatoxin,Phytophenolics,Cancer,Diabetes,Alzheimer's disease,Oxidative stress,Inflammation

The future GCC-connected environmental risk factors expedited the progression of nCDs. Indeed, the emergence of AFs is becoming a global food security concern. AFs are lethal carcinogenic mycotoxins, causing damage to the liver, kidney, and gastrointestinal organs. Long-term exposure to AFs leads to liver cancer. Almost a variety of food commodities, crops, spices, herbaceous materials, nuts, and processed foods can be contaminated with AFs. In this regard, the primary sections of this review aim to cover influencing factors in the occurrence of AFs, the role of AFs in progression of nCDs, links between GCC/nCDs and exposure to AFs, frequency of AFs-based academic investigations, and world distribution of AFs. Next, the current trends in the application of PPs to alleviate AFs toxicity are discussed. Nearly, more than 20,000 published records indexed in scientific databases have been screened to find recent trends on AFs and application of PPs in AFs therapy. Accordingly, shifts in world climate, improper infrastructures for production/storage of food commodities, inconsistency of global polices on AFs permissible concentration in food/feed, and lack of the public awareness are accounting for a considerable proportion of AFs damages. AFs exhibited their toxic effects by triggering the progression of inflammation and oxidative/nitrosative stress, in turn, leading to the onset of nCDs. PPs could decrease AFs-associated oxidative stress, genotoxic, mutagenic, and carcinogenic effects by improving cellular antioxidant balance, regulation of signaling pathways, alleviating inflammatory responses, and modification of gene expression profile in a dose/time-reliant fashion. The administration of PPs alone displayed lower biological properties compared to co-treatment of these metabolites with AFs. This issue might highlight the therapeutic application of PPs than their preventative content. Flavonoids such as quercetin and oxidized tea phenolics, curcumin and resveratrol were the most studied anti-AFs PPs. Our literature review clearly disclosed that considering PPs in antioxidant therapies to alleviate complications of AFs requires improvement in their bioavailability, pharmacokinetics, tissue clearance, and off-target mode of action. Due to the emergencies in the elimination of AFs in food/feedstuffs, further large-scale clinical assessment of PPs to decrease the consequences of AFs is highly required.

May phytophenolics alleviate aflatoxins-induced health challenges? A holistic insight on current landscape and future prospects The future GCC-connected environmental risk factors expedited the progression of nCDs. Indeed, the emergence of AFs is becoming a global food security concern. AFs are lethal carcinogenic mycotoxins, causing damage to the liver, kidney, and gastrointestinal organs. Long-term exposure to AFs leads to liver cancer. Almost a variety of food commodities, crops, spices, herbaceous materials, nuts, and processed foods can be contaminated with AFs. In this regard, the primary sections of this review aim to cover influencing factors in the occurrence of AFs, the role of AFs in progression of nCDs, links between GCC/nCDs and exposure to AFs, frequency of AFs-based academic investigations, and world distribution of AFs. Next, the current trends in the application of PPs to alleviate AFs toxicity are discussed. Nearly, more than 20,000 published records indexed in scientific databases have been screened to find recent trends on AFs and application of PPs in AFs therapy. Accordingly, shifts in world climate, improper infrastructures for production/storage of food commodities, inconsistency of global polices on AFs permissible concentration in food/feed, and lack of the public awareness are accounting for a considerable proportion of AFs damages. AFs exhibited their toxic effects by triggering the progression of inflammation and oxidative/nitrosative stress, in turn, leading to the onset of nCDs. PPs could decrease AFs-associated oxidative stress, genotoxic, mutagenic, and carcinogenic effects by improving cellular antioxidant balance, regulation of signaling pathways, alleviating inflammatory responses, and modification of gene expression profile in a dose/time-reliant fashion. The administration of PPs alone displayed lower biological properties compared to co-treatment of these metabolites with AFs. This issue might highlight the therapeutic application of PPs than their preventative content. Flavonoids such as quercetin and oxidized tea phenolics, curcumin and resveratrol were the most studied anti-AFs PPs. Our literature review clearly disclosed that considering PPs in antioxidant therapies to alleviate complications of AFs requires improvement in their bioavailability, pharmacokinetics, tissue clearance, and off-target mode of action. Due to the emergencies in the elimination of AFs in food/feedstuffs, further large-scale clinical assessment of PPs to decrease the consequences of AFs is highly required. Nowadays, the estimations predict that the global demand for intensified food production has been increased, and the statistics are expected to be doubled by 2050 (1, 2). Providing safe foods to nurture the world population requires a significant improvement in crop cultivation systems, plant breeding techniques, and the development of climate-smart crops (3). GCC and environmental forces are two determinant factors to influence crops' sustainable growth. GCC influences crop production practices in sophisticated modes (3, 4). Numerous studies have been addressed the direct and indirect effects of GCC on world food demand and agricultural systems (5, 6). Accordingly, modification of cultivation systems and increasing the susceptibility of crops to future climate are typical direct effects of climate change. From a large-scale perspective, affecting the world economy, food demand, and distributions of incomes are indirect effects of climate change on world societies (4). Studies have shown that changes in climate humidity, temperature, and precipitation patterns are associated with the outbreak of some invasive fungal pathogens (7). The uncontrolled growth of these fungi affects the quality and quantity of crop yield, stored foodstuffs, grains, processed foods, and herbaceous products (8). Studies highlighted the possible health risks linked to fungal toxins to the human body (8). The APF including Aspergillus, Fusarium, and Penicillium, are well-known mycotoxigenic fungal species with potentially fatal effects on human and animal health (9). The occurrence of AFs in food commodities depends on environmental factors such as air temperature, humidity, CO2 levels, pH, susceptibility of foods to contaminations, and improper harvest and storage of food products (10, 11). According to scientific reports, nearly 25% of global food supplies are contaminated with AFs, making them a serious issue of concern for world nations' health (12). AFs are health hazardous environmental risk factors in the onset of liver cancer, kidney failure, and gastrointestinal problems (13), in turn, their health consequences depend on duration of exposure and enzymatic/genetic alterations in target organs (14). Poverty, hunger and contaminated foods are the foremost health risk factors to threaten human life. Numerous studies suggested that mycotoxins, including aflatoxins, ergot alkaloids, ochratoxins, trichothecenes, zearalenone, and fumonisins, are among the most lethal naturally occurring toxins (15, 16). Characterization and elimination of mycotoxins in food/feed items require specific technical and analytical methods (17). Developed countries implemented strict regulatory gates to control and monitor the occurrence of AFs in import/export sites to reduce the health complications of these toxins (18). In contrast, many people in low-income countries are at risk of long-term exposure to AFs, resultantly this might increase the progression of different cancers in these locations (19). Studies have shown that natural products (e.g., PPs, berberine, plant extracts, polysaccharides, etc.) can reduce the toxicity and production of AFs (20, 21). PPs are a large heterogeneous group of secondary plant metabolites, display a wide range of biological activities, and are substantially studied for their anticancer activities (22–24). The regular consumption of PPs is associated with lower risks for developing cardiovascular diseases, obesity, DM, cancer, stroke, and AD (23–26), though this finding requires further well-designed clinical validations. Global interest in PPs studies to alleviate AFs health consequences has been increased during the past decades. Preliminary studies showed that PPs can directly and/or indirectly alter the possible toxicity effects of AFs in complicated ways (27, 28). However, the molecular mechanisms underlying PPs effects on AFs toxicity in the human body are still not understood comprehensively, and scientific efforts are ongoing to find out the health-promoting content of PPs against AFs. In this regard, this review aims to summarize recent findings on AFs and the application of PPs in alleviating AFs end effects. Due to the strategic roles of antioxidant phytochemicals in preventing human chronic diseases (29), and beneficial effects of PPs in the cornerstone of therapeutic programs (30, 31), we have followed five goals herein: (1) providing a comprehensive insight on the association of GCC and the occurrence of AFs; (2) characterizing the most potent anti-AFs phenolics; (3) role of AFs in the onset of nCDs; (4) understanding anti-AFs mechanism of actions of PPs; and (5) addressing the current gaps regarding the large-scale application of PPs for clinical applications. Scientific databases including Scopus, PubMed, Google Patents, and Scholar were separately searched to find relevant papers using keywords such as “aflatoxin or human diseases,” “aflatoxin B1/B2/G1/G2/M1,” “aflatoxin and GMO,” “aflatoxin and diabetes,” “aflatoxin and cancer,” “aflatoxin B1 or chronic diseases,” “aflatoxin or transcriptome,” “aflatoxin and/or Alzheimer's,” “aflatoxin and flavonoids,” “aflatoxin or polyphenols,” “aflatoxin and/or stilbenes,” “aflatoxin and curcumin,” “aflatoxin and/or epigenetic,” “aflatoxin and/or plant extract,” “aflatoxin and/or crops,” “aflatoxin M1 and/or foodstuff,” aflatoxin and/or spices,” “aflatoxin and climate change,” “aflatoxin and/or temperature,” “aflatoxin and/or CO2 levels.” The outputs of searches were used to generate bibliometric network using VOSviewer software (32). More than 20,000 papers (published from 1990 to 2021) were appeared in search outputs. To unify the results, we compared the outcomes together, thus the redundant papers with similar title or content were deleted from search outputs. To interpret the statistical output of literature searches, we used Scopus data to illustrate the relevant graphs. We also used Cytoscape (ClueGo module) software (33) to construct simplified protein-protein interaction networks where it was needed. Because the exact binding modes of all AFB1 metabolites into human serum albumin (HSA) were not available in the literature, we used AutoDock Vina tools (34) to generate the expected interactions. The Protein-Imager software (35) was recruited to inspire some of graphical illustrations. Fungal AFs including B1(AFB1), B2(AFB2), M1(AFM1), G1(AFG1), and G2(AFG2) are well-known AFs in contaminated crops, foods, dairy products, herbal materials, spices, and processed foods (19, 36). AFs showed mutagenic, carcinogenic, hepatogenic, teratogenic, and immunosuppressive toxicological properties. The toxic properties of AFs depend on alteration of enzymatic activity, modification of gene expression patterns, epigenetics changes, and dysregulation of signaling pathways (11–13, 37, 38). Chemically, AFs can be classified into two main groups including difuran-coumarin lactones (AFG1/2) and difuran-cyclo-pentanones (AFB1/2 and AFM1/2) (39, 40). These fungal toxins displayed a “CHO” molecular formula with a different number of H/O atoms. The molecular weight of AFs ranges from 312 to 330 g·mol−1. AFs displayed a colorful fluorescent pattern under the UV light. In this respect, AFB1/2 showed a blue color while AFGs displayed green color (39, 40) (Figure 1). To date, more than 20 different AFs have been identified with moderate to high toxicity effects on the human body (40). The evidence suggests that AFs are potent carcinogenic toxins with potential side effects on human liver organs and tissues (37). The laboratory analysis of AFs identified several metabolized AFs derivatives that rarely found in the human body (40). The most abundant AFs derivatives found in human and animal bodies are AFM1/2 metabolites-the derivatives of AFB1 metabolism in the liver (41). These AFs are widely found in milk and milk-based products, and recently received much attention from the literature due to their potential health problems for consumers of dairy products (41). Studies have shown that AFs are thermostable substances, therefore elevated temperature might not destroy these mycotoxins (40). The toxicity content of AFs is in the order of AFB1 > AFG1 > AFB2 > AFG2 (42). These mycotoxins have been considered as group I highly carcinogenic substances by IARC (40, 43). No only structural features of well-known AFs, but also the chemistry of emerging toxins mainly produced by Fusarium species should be taken into account because of their potential to outbreak and influence the future food safety and security (44). For more details, the chemistry and distribution of emerging mycotoxins have been reviewed by Gruber-Dorninger and colleagues (45). The GCC is becoming a serious concern, and being the future driver of food safety and security (46). The expanded worldwide industrial activities and the increasing rate of world population are the two of the most crucial components of countries' climate modification (47–49). The outbreak of invasive plant pests/pathogens, soil erosion, drought, erratic rainfall patterns, GW, salinity, crop cultivation failure, shortage of irrigation water, and reduction in the fertility of arable soils are the foremost consequences of the GCC (47, 48, 50, 51). Studies disclosed that GCC induces the occurrence of plant fungal pathogens by providing appropriate environmental requirements for their growth and development (52, 53). While elevated temperatures and irregular precipitation patterns can significantly reduce crop yield, the growing evidence suggests that these climatic consequences also triggered the production of AFs at the beginning of crops growing seasons (53). The emerging evidence postulates that under at least +2 to +5°C climate temperature increase, maize and wheat crops might prone to higher levels of AFs contamination (46). Studies also demonstrated that APF tolerated a wide range of temperatures, resultantly enabled these fungi to grow easily in the production/storage sites of crops (54). The GCC might support the growth and development of APF in crop production sites (7, 55). The overgrowth of these fungi leads to a rapid expansion of spores in the environment, increases AFs contaminants level, ultimately leading to health impacts on consumers (7, 54). Due to the increased universal demand for food, the GCC intensifies crop cultivation and production, in turn, might lead to the establishment of single-crop cultivation systems (56). Additionally, underestimating of crop rotation and plantation of susceptible crops may also increase the risk of future AFs occurrence (57). Different strategies have been applied to address the GCC in advanced and developing countries (3, 54). These policies mainly focused on improving breeding strategies (3) and recruiting integrated and novel methods to control plant pathogens (58). In contrast, farming systems in undeveloped countries are almost entirely influenced by GCC (1). On such occasions, traditional methods of crop cultivation, harvest, and storage have mainly been followed by local farmers, these may in turn be leading to the exacerbation of APF outbreak (59). Although a variety of sophisticated methods have been employed in developed countries for integrated control of mycotoxin-producing fungi in food and feed (54, 60), the outcomes, however, suggested that these methods were rarely successful to completely reduce the occurrence of AFs (18, 61). The emerging reports on the GCC demonstrated that GW has begun to occur but the empirical data on how GW might affect crops yield is markedly overlooked (62). Strategically, GW and GCC might alter the distribution of plant pests and diseases, resulted in a significant damage to crop production. The literature suggests that pest-infested crops are prone to AFs contaminants (63). Therefore, predicting the exact roles of GCC on the prevalence of AFs relies the development of accurate models to estimate the future damages of AFs under GCC (64). Presently, mechanistic (65), empirical, and hybrid models in predicting future economic costs of AFs occurrence have been developed in Australia, the USA and some European countries (64) while the lack of comprehensive predictive models in less developed countries (e.g., Africa, Middle-East, Latin America) might decrease the effectiveness of such estimations in preventing future AFs production (64, 66). Therefore, recruiting predictive models to simulate the occurrence of AFs requires an in-depth knowledge of future AFP-GCC interactions (67) to know where and how these mycotoxins will be emerged in the target production/storage sites (64, 66), which future crops are more susceptible to AFs, and ultimately which future country-specific regulations must be taken into consideration in order to better elimination of AFs production. Indeed, in addressing modeling of future APF-GCC interactions, it is also meritorious to highlight this note that the contemporaneous predictive studies have been validated in limited geographical regions, therefore it is imperative to conduct large-scale multinational investigations for better understanding of GCC impacts on future APF mycoflora and global pattern of AFs distribution (68). In this regard, Yu et al. reported that global temperature modification has an impact on the prevalence of AFs contaminants (62). According to the given model for simulation of AFs occurrence based on corn phenology, it is believed that some corn grown US states will experience an increased level of AFs occurrence by 2031–2040 (62). On the contrary, this estimation also postulated that under elevated temperatures AFs might be inactivated, therefore, some the US counties might experience lower level of AFs occurrence (62). Correspondingly, other outcomes also suggested that water stress and elevated temperatures are two determinant factors in changing the relative expression pattern of structural genes (aflD, aflR) involved in the production of AFB1, leading to higher occurrence of this carcinogenic mycotoxin (67). Scientists recruited high-throughput multi-omics technologies to assess the impact of GCC on the production of AFB1 (69). The outcomes unraveled that the elevated CO2 levels, as a consequence of GW, might alter aflR gene expression in AFB1 biosynthetic pathway (69). In another study, researchers simulated climate change condition to investigate how it might influence the growth of A. carbonarius and OTA production under elevated temperature/CO2 levels (70). The results surprisingly displayed that the interaction between elevated CO2 levels and temperature lead to the up-regulation of velvet complex regulatory elements and OTA biosynthetic genes in A. carbonarius. This finding suggests that elevated CO2/temperature levels are two quintessential factors in increasing the risk of OTA contaminants in grape-based products (70). The outcomes from a similar study also demonstrated that changes in temperature/CO2 levels in stored coffee beans and coffee-based media attributed to OTA production in A. westerdijkiae compared to A. carbonarius (71). In another study, it was also reported that the GCC-associated factors have differential impacts on AFB1 production in pistachio nuts (7). In a case study conducted on maize grown in Eastern Europe using different climatological models, the estimations predicted that climate change can lead to a probable increase in the occurrence of maize AFB1 and cow's milk AFM1 (72). Other studies on the effects of GW on the occurrence of AFB1 and trichothecenes mycotoxins in wheat and maize crops also suggested that the prevalence of these hazardous mycotoxins is expected to increase as consequences of the future GCC (46, 68). Although little is understood on the effects of future GCC on mycotoxigenic fungi growth and mycotoxins production, developing accurate predictive models to characterize mechanistic interactions of APF with climatological factors will provide a ground for better controlling of these fungi (68). It is estimated that economic losses due to the occurrence of AFs are between $500 million to $1.6 billion for maize, peanuts and other crops in the USA (73, 74). These obvious economic costs, however, are associated with GCC and its impact on AFs production (73). Another pivotal issue, that is crops grown in low and northern latitudes might negatively or positively deal with future GCC (44). In this regard, the evidence suggests that low latitude regions will be suffered from consistent and negative consequences of GCC compared to the northern regions where its effects may be positive or negative (44). Considering the relationship between GCC and increased global demand for food, this might threaten the production of certain crops such as maize using the available infrastructures. In this respect, the occurrence of AFs and other mycotoxins is expected to increase on such occasions (75). In spite of the fact that the universal temperature may be rising above the optimum condition for APF, it is pivotal to consider the threat posed by emerging thermotolerant fungi that can produce novel health hazardous toxins (76). Therefore, smart crop breeding for developing resistance against both GCC/GW and APF might be considered as an alternative scenario in managing the reduction of AFs occurrence in food and feed (75). In addition to financial losses to animal and agricultural commodities, AFs-contaminated foods are leading to significant clinical costs due to the side effects of long-term exposure to these toxins for the human body (18). However, decreasing milk production, reducing of crops quality, weakening animal immune system, and many other demerits are few examples of AFs financial burden for animal and crop products (18). Various chemical and biological methods have been suggested in controlling AFs-producing molds (60, 77, 78), nevertheless, the evidence suggests that there has been no efficient method to completely eliminate these molds (60, 78, 79). It should be expected that the universal quantity/distribution of AFs will probably be grown under GCC/GW. Therefore, due to the upcoming future GCC, systematic modification of crop cultivation practices, applying crop rotation, developing climate-smart crops, set limits to APF growth in production sites, modernizing crop storage facilities, utilizing modern crop irrigation systems, and increasing public awareness about the association of climate change and AFs risks are influential scenarios to reduce the occurrence of AFs under modified future climate (Figure 2). Studies have also highlighted the role of GCC in the prevalence of nCDs such as cancer, MetSys, stroke, chronic respiratory disorders, and cardiovascular diseases (80, 81). In this regard, several investigations purported that the future GW/GCC will highly increase exposure levels to GCC-associated health hazardous risk factors (49, 66, 68, 72), consequently leading to higher rates of global deaths (49, 81). As discussed, AFs exposure will probably be growing in the upcoming years owing to a significant modification in countries' climate patterns (82), nevertheless, the current estimations require further validations to address all gaps and challenges in preparing world communities for future changes (68). The co-occurrence of GCC-associated risk factors offering synergistic effects on human health and the onset of nCDs (83, 84). Therefore, due to the complexity of interactions between GCC/exposure to AFs (and other emerging mycotoxins) (85), and the onset of nCDs (86), it is obligatory to establish country-specific regulations to deal with the upcoming challenges (64) and decrease the global burden of incurable human diseases. While only 2% of global health funds allocated to treat these diseases, the estimations predicted that the number of people affected by nCDs have dramatically been increased over the past decades (81, 87). Therefore, due to the lack of specific international leadership to combat nCDs (81), the increased exposure to AFs and other emerging mycotoxins under countries' climate change will worsen global health status, particularly in low and middle-income countries (88, 89). In this regard, to decrease the economic costs of AFs exposure and the progression of nCDs, the possible threats of future GCC should be taken into account in alleviating the health consequences of AFs risks. According to the Scopus statistics (https://www.scopus.com/), the global studies on AFs have markedly increased during the past few decades. The literature mining of scientific databases showed that more than 20,000 papers have been published on different classes of AFs from 1990 to 2021. As depicted in Figure 3, the frequencies of studies on different types of AFs metabolites displayed that most of these studies had targeted AFs metabolites such as AFB1/2, AFG1/2, and AFM1. Interestingly, the number of studies on AFM2 was lower than other AFs metabolites during the investigated timeline, and these investigations have increased from 2008 to 2021. Accordingly, a large proportion of scientific studies on AFs were conducted in field of agriculture and biological sciences, followed by biochemistry, medicine, and chemistry, respectively. These frequencies of studies on AFs displayed that these mycotoxins received much attention from academia and clinical sectors due to their health hazardous risks. Evaluating the searched papers to highlight the top countries for research on AFs showed that the USA, China, and India were occupied the top ranks for studies on AFB1. In the case of AFM1, the USA, China and Iran were the three top countries to publish academic investigations on this aflatoxin. Indeed, China, the USA, and Brazil published many articles on AFG1 and ranked first to third countries to conduct research in this area. Interestingly, these results are in consistent with previous outcomes on the frequency of global studies on AFs in which the USA, China, and India were the top publishing sources from 1998 to 2017 (90). These outcomes together suggest that the global frequency of scientific studies on AFs has markedly extended over the past few years. Evaluating and monitoring academic publications on AFs can support researchers to identify the critical gaps in these studies for improving current regulatory policies, local and international awareness programs and making political decisions to protect target consumers from complications of AFs. Figure 4 shows a detailed representation on the proportion of the top 10 countries to conduct studies on different classes of AFs metabolites. The review of literature unraveled that the conducted studies on AFs can be specified in different clusters. The majority of these studies targeted technical procedures for detecting and monitoring AFs in food sources, crops, spices, dairy products, nuts, and herbaceous products. Additionally, research interest in characterization of AFs in vegetable oils has been increased over the past 5 years. For example, a recent meta-analysis disclosed that vegetable oils such as sesame oil showed a differential AFs contamination (91). In another study, vegetable oil samples such as coconut oil was contaminated with different AFB1 concentrations (92). This indicates that consumption of such oil samples may pose serious health risks. Analytical methods such as HPLC, ELISA, TLC, HPTLC, UHPLC, mass spectroscopy, and immunoassays were the most frequent procedures used for diagnosing the AFs contaminants in foods/feeds. Although the basic concept of these methods was previously reviewed in several studies (93–95), however, a fast and accurate method to characterize AFs in suspected sources has not been reported. Miklós et al. reviewed recent trends in developing accurate analytical/immunological measurements to identify AFs in different food and feed items (94). According to their finding, ELISA and LFIA are two promising methods to quantify the minimum concentrations of AFs in food or feed items. Indeed, IAC-clean up followed by HPLC-FLD is another accurate system for AFs measurements (94). Due to the co-occurrence of mycotoxins in food/feed items, the currently applied diagnostic methods might not appropriately detect different types of mycotoxins in evaluated items, therefore, the application of LC-MS/MS technique has been received much attention from academia owing to its ability in multiplex identification of AFs (94). As shown in Figure 5, a considerable number of studies also highlighted the carcinogenic effects of AFs and their role in the development of human cancers such as liver and gastrointestinal tumors. Studies on biological and chemical control of APF were also frequent among screened investigations. We also found that studies on the effects of AFs on human and animal reproduction systems were increased during the past two decades. This is an interesting topic because the evidence suggests that AFs have negative impacts on reproductive organs (96). Studies on the toxicity of AFs for animals and chickens were also increased up to 5–10 folds over the past few years. More interestingly, the frequency of intervention studies using synthetic and/or natural compounds to alleviate the complications of AFs were also grown during the past decade. According to these statistical data, AFB1 and AFM1 were the most studied AFs during the past decades, though studies on AFB2, AFG1/2, and AFM2 were remarkably increased in the same time. Figure 5 highlights the major clusters for studies on AFs using searched keywords in scientific databases. Literature mining using available scientific resources has also manifested that the risk assessment investigations for characterization of AFs contaminants were dramatically increased. As depicted in Figure 6, the detailed bibliometric networks of studies on AFs unraveled that the carcinogenic effects of AFB1 in inducing hepatocellular carcinoma were significantly investigated. Interestingly, the outcomes showed that the application of PPs (e.g., curcumin and resveratrol) to alleviate health-related complications of AFs has increased during the last 10 years due to the health-promoting effects of these natural metabolites. In this regard, studies suggested that antioxidant therapy might ameliorate the hepatotoxicity of AFs metabolites, leading to lower health risks for cancer development (97). As it is discussed in the next sections, antioxidants are critical chemical agents provided the human body with ability to protect both lipid and protein elements from free radicals and oxidative agents (98). Therefore, considering antioxidants as promising agents in the prevention of AFs complications might decrease the deterioration of hepatic cells and prevent the development of liver cancer, though this claim requires further clinical assessment. Based on the data discussed in this section, it can be concluded that in which fields AFs have considerably studied and where they have been ruled out in scientific investigations. Because AFs are emerging health hazardous threats to the human society, studies on the complications of these fungal toxins, developing reliable medicines to decrease the toxicity of AFs and integrated management of AFs in production/storage sites should be increased to mitigate the quantity of AFs in food/feed items. As discussed in the previous sections, the worldwide occurrence of AFs depends on several factors such as climate, the availability of standard crop/food storage facilities, the awareness of local farmers, food processing methods, post-harvest contaminations, and temperature and moisture of post-harvest storage sites. Of all, air humidity and temperature are two determinant factors to enhance the emergence of APF (99, 100). In an interesting study, the worldwide occurrence of several mycotoxins in feed has been summarized (101). Accordingly, in some geographic regions, the average concentrations of mycotoxin contaminants in feed are considerably higher than other ones owing to the imbalance distribution of these toxins. For instance, the highest median concentration of AFB1 was reported for Sub-Saharan Africa (23 μg/kg), South Asia (20 μg/kg), Southeast Asia (10 μg/kg), and East Asia (10 μg/k) (101). Fumonisins (B1, B2, and B3) are other mycotoxins that pose high health risks to the human societies. The highest median concentration of fumonisins was reported for South America (1390 μg/kg), Central America (929 μg/kg), and Sub-Saharan Africa (789 μg/kg) (101). Compared to other mycotoxins, fumonisins showed a broader geographic occurrence. In contrast, OTA displayed a domineering abundancy ratio in Central Asia (22 μg/kg) and South America (17 μg/kg), respectively (101). Figure 7 represents the worldwide median concentration of five mycotoxins found in feed in different geographical regions based on information adopted from Gruber-Dorninger et al. (101). Other studies have also highlighted the observed level of AFs in different geographical regions (102, 103). The results showed that the occurrence of AFs in various foods, cereals, nuts, oilseeds and processed foods is inevitable and these sources showed a differential level of contaminations with AFs derivatives (102). According to these outcomes, the prevalence of AFs in each region might be affected by local climate and the abundancy of crops found in these regions (103, 104). Therefore, it can be concluded that AFs are almost not equally distributed in production/storage sites. Therefore, the construction of predictive models for the occurrence of AFs during specific seasons is helpful in identifying the contaminated food/feed items to prevent the circulation of AFs in local and international food chains (105, 106). Over the past decades, studies on various food/feed items to trace the fingerprint of AFs contaminants have been increased (13). Cereals are most commonly cultivated in the world, supporting human societies in reaching essential nutrients in their diet (10, 107). Studies reported that AFs occurrence in cereals is becoming a serious worldwide concern (11). The formation of AFs in cereals and cereals-based processed products depends on several factors such as fungal genotype, processing methods (drying, milling, blending, chemical additives), and environmental factors such as oxygen level, environmental pH, field temperature and humidity content (11). A comprehensive analysis of different cereals samples using published records in GEMS/Food database showed that around 12.7% of all samples were positive for contamination with at least one of AFs (10). Correspondingly, rice, sorghum, and maize samples possessed a higher level of AFs (10). Other investigations also suggested that the highest level of AFs was detected in maize in the concentration of 3,760 μg/kg, which extremely exceeded from the USA and EU permissible standards (108). Among cereals, rice is surprisingly a susceptible crop prone to AFs pollutants (103, 104, 109). After rice, corn and sorghum are prone to AFs contaminants (109). The evidence suggests that AFB1 is the main AFs found in cereals (11). Another study on 108 Brazilian wheat and wheat by-products samples disclosed that 30.6% of studied samples were positive for at least contamination with one of the AFs in which AFB1 was the most dominant fungal mycotoxin in these samples (110). The highest contamination levels were observed for wheat grains, followed by the barn, whole and refined flour (110). Presently in EU and other countries, only limited concentrations of AFs are allowed to be found in food products (36). In this regard, the allowed concentration of AFB1 and total AFs for nuts and cereals in the EU is 2 and 4 μg/kg, respectively (109). In cereals, applying inappropriate drying methods allowed for maintaining higher humidity content in these crops, leading to a higher ratio of AFs-contaminated crops (13). Recent studies on contaminated cereals demonstrated that differential concentrations of AFs are found in cereals grown in different countries (109). However, nuts, groundnuts, and cereals are prone to AFB1 contaminations under field and non-standard storage whenever temperature, humidity, and field soil are suitable for APF growth (13, 111). A recent study on the prevalence of AFs in nuts from different origins demonstrated that peanuts from Argentina, Congo, Nigeria, and South-western Uganda were differentially contaminated with AFB1 (112). Accordingly, the average concentrations of AFB1 in peanuts from these countries were 530, 163.22, 110.95, and 103.10 μg/kg, respectively (112). Countries such as Taiwan, Morocco, and Iran ranked as first to third countries for contamination of pistachio with AFs (112). The maximum average concentration of AFB1 in almond, hazelnut, walnut, and Brazilian nut samples was observed for countries such as Cyprus, Italy, Morocco, and Brazil (112). Other interesting studies have also reviewed the contamination of Iranian pistachio with AFB1 using different procedures (113). Accordingly, the outcomes suggested that there has been differential AFB1 concentrations in this nut. Based on these results, around 37% of studies reported AFB1 contaminations in the concentration of ≥10 μg/kg, 35% of studies reported ≤ 10 μg/kg, and 28% of studies reported ≤ 5 μg/kg, respectively (113). In another interesting study, Bui-Klimke et al. also analyzed the global regulations on prevalence of AFs in pistachio samples and reported that pistachio nuts are accounting for substantial quantity of dietary AFs (114). Accordingly, estimations showed that contaminated pistachio nuts remarkably affected the global market of this valuable nut and ignoring the presence of AFs in these samples will increase the health risks of these mycotoxins for target exported/imported locations (114, 115). Presently, global markets between Asia and other world countries have been spectacularly increased (116). Due to the higher rates of AFs occurrence in pistachio samples (114), top producers of this nut should develop modernized infrastructures for drying and processing pistachio to eliminate the expected level of AFs. Increasing monitoring gates in import/export gates of target consumers of pistachio can help to early detection of AFs sources and prevention of APF growth. Walnut kernel and oil are two important products worldwide. Studies have shown that walnut kernel and oil have an interesting metabolic profile which in turn can be considered as a source of antioxidants and mineral elements (117, 118). Recent findings on the contamination of Iranian walnuts with AFs showed that nearly half of these samples were contaminated with AFB1 in concentrations of 0.8–14.5 μg/kg, respectively (119). Nuts such as walnut maintain remarkable moisture in their structure, leading to providing a favorable environment for APF growth (119). The grown fungi inside walnuts significantly affect these tree nuts quality and destroy their taste and flavor. Applying modern drying technologies to process, drying and shipping walnut and other nuts can extraordinarily decline the occurrence of AFs in such nuts. The prevalence of AFs in salt-roasted nuts is also becoming an emerging concern. In this regard, Ostadrahimi et al. (120) reported that salt-roasted pistachio and peanuts possessed a differential concentration of AFs. Correspondingly, the observed average AFs concentration in these samples was about 19.88 μg/kg in comparison to pure nuts (6.51 μg/kg), respectively (120). These outcomes indicate that processing of economic nuts should be carefully conducted because the occurrence of AFs after nut preparing steps for increasing their taste and flavor is inevitable; therefore, it can pose health risks to target consumers. Considering the world climate zones map (Supplementary Figure S1), it can be said that the higher prevalence of AFB1 in peanuts harvested from different geographic regions may be related to the type and dominancy of countries' climate. The literature suggested that the moisture content above 17% and warmer temperatures (above 24°C) are effective in inducing the formation of AFs in corn and feed (121, 122). Indeed, the review of literature manifested that due to the higher moisture content of nuts, these products are the main susceptible foodstuffs for AFs contaminations (123). Therefore, in addition to local climatological factors, standard storage of nuts and decreasing the moisture content of these products before entering storage sites and local/global markets can dominantly affect the prevalence of AFB1 in such foodstuffs. A majority of AFB1/2 concentrations in contaminated crops have entered the animal feed network and metabolized to AFM1 (109). In this regard, the evidence suggests that nearly 1% of AFB1 metabolized into AFM1 in dairy cows (124). About 1–3% ingested AFB1 and metabolized AFM1 are excreted by feces and urine (109). However, the remained AFM1 level in the animal body will later enter the human food chain through dairy products. Due to the considerable affinity of AFM1 to dairy proteins (125), it seems that in dairy products with higher protein content (e.g., cheese), the occurrence of AFM1 is more probable compared to other AFs (109, 125). Generally, the occurrence of AFM1 in milk, cheeses, butter, and yogurt is surprisingly high; however, the final concentration of AFM1 in these products depends on processing methods and the quality of animal feed (109, 126). Today, the presence of AFM1 is becoming a global concern (127), as discussed in previous sections, the frequency of studies on AFM1 has been dramatically increased over the past decades. A recent meta-analysis indicated that the prevalence of this mycotoxin among dairy products is averagely between 40 and 60% which has seriously been considered a biological threat to public health (128). Although the toxicity of AFM1, a hydroxylated derivative of AFB1, is relatively lower than other types of AFs (129), the available evidence indicates that the long-term exposure to AFM1 might be effective in the onset of liver cancer (39, 128). Animals supplemented with AFs-contaminated feed are major sources of AFM1. Because dairy products are unique sources of proteins, vitamins, and calcium, they are becoming the principal part of the human diet (127). Therefore, AFM1 contaminants in these products pose a threat to public health (128), and regular monitoring measurements should be conducted to diagnose target AFs in dairy products. However, to eliminate the AFM1 in animal products, increasing public awareness and regular monitoring of dairy products can be helpful (130). Scientific studies also reported that the occurrence of AFs contaminants in livestock meat products is probable (131–133). Accordingly, AFs such as AFB1/B2 and AFG1/G2 with different concentrations occurred in meat-based foods (132–134). In an interesting study on meat products collected in Riyadh, Saudi Arabia, incredibly 37.5% of gathered samples were contaminated with AFs, and 4% of samples have exceeded from permissible standards (the acceptable Saudi limit: 20 μg/kg). Correspondingly, AFB1 and AFG1 were the most commonly identified AFs, followed by AFB2, respectively (133). In another study, the occurrence of AFs in meat products such as basterma, sausage, kofta, and luncheon was investigated, and the results unraveled that AFB1/2 occurred in higher concentrations compared to AFG1/G2 (132). Investigations on domestic fowls feeding diets containing AFs also displayed that the accumulation of AFs in their liver is higher than other organs (135). Indeed, the outcomes showed that the highest abundancy of AFB1 was observed for tissues of quails compared to other birds (135). In another study conducted on meat, milk, and eggs samples, collected in Jordan, the outcomes indicated that the samples were contaminated with AFB1/2, AFG1/G2, and AFM1/AFM2 (136). In milk samples, however, the highest concentration of AFM1 was exceeded from EU standard for liquid milk (50 ng/L) (136). These outcomes suggested that the proportion of AFs contaminants in meat samples depends on several factors, including meat processing methods, post-storage contaminations, and non-standard transportation and shipment facilities of meat products to local and international markets. Presently, the maximum permissible AFs concentration in animal feeds is 5 μg/kg based on EU limits (137). Therefore, preparing animal feeds from credited sources, improving of the storage condition of livestock inputs (137), improving the quality and accuracy of AFs detection systems (138), improving of animal feed manufacturing procedures (139), and regular monitoring of animal feeds to identify the source of AFs contaminants are possible strategies to reduce the concentration of AF pollutants in dairy products. Governments must seriously deal with providers of livestock feeds that might supply contaminated animal inputs to dairy farms by imposing strict limitations on their business to prevent further consequences of AFs. It is now well-documented that AFB1 was widely found in contaminated livestock products fed on contaminated forage and grain (140). Therefore, due to health threats of fungal toxins to public health, dairy products should be repeatedly monitored to decrease the quantity of AFs contaminants. Such strategies will later decrease the health consequences of AFs and help consumers to reach safe and AFs-free dairy items. In practice, however, characterization of AFs-contaminated animal feeds is difficult because the available techniques require allocating sufficient time and financial support for early detection of different types of AFs metabolites (137–139). Governments should support researchers in developing accurate, fast and low-cost measuring systems to abate AFs metabolites from suspected feeding resources. As discussed in the previous sections, large quantities of scientific studies have been conducted on measuring processes of AFs diagnosing systems. Expanding academic studies without practical innovations cannot alone help the elimination of AFs in animal feeds; therefore, efforts to convert the results of academic investigations into touchable outcomes should be highly followed to recruit the power of science in controlling health hazardous mycotoxins. As we discussed later, AFs are major environmental risks factors in developing nCDs. Therefore, elimination of these fungal toxins in animal feeds can primarily decrease their occurrence in human food chain. Spices are interesting food additives that constructed a valuable financial global market for spice-producing countries (141, 142). Presently, countries such as India, China, Nigeria, Indonesia, Thailand, Vietnam, Bangladesh, Nepal, Ethiopia, and Turkey occupy the first to tenth ranks of the top spice-producing countries (143). Estimations indicated that the demand for fresh, powdered, and processed spices have been increased over the past decades due to multi-functional applicability of these products for various purposes such as traditional medicine, cooking, etc. (142, 144). Interestingly, AFs contaminants also occur in spices with higher content of moisture. Since 2002, the EU has implemented rigorous regulatory policies to identify AFs in spices (36, 144). Accordingly, the permissible concentration of AFB1 and total AFs in spices has been reported up to 5 and 10 μg/kg, respectively (36). Studies proclaimed that well-distinguished spices such as pepper, ground red pepper, paprika, curcumin (or turmeric), chili, nutmeg, and ginger are the susceptible natural food additives prone to AFs pollutants (36, 109). The evidence also suggests that the highest permissible AFs limits for all foods in the USA are 20 μg/kg (145). In Croatia, the permissible AFB1 and total AFs levels for spices are 30 and 15 μg/kg, while in Bulgaria, the accepted limits are in the concentrations of 2 and 5 μg/kg (144). Iran also follows the EU regulation on spices and the permitted maximum levels of AFB1, and total AFs range in the concentrations of 5 and 10 μg/kg (144). In this respect, several studies comprehensively reviewed the occurrence of AFs in spices (144, 146, 147). The outcomes of these scientific investigations demonstrated that the occurrence of AFs in spices depend on the type and processing methods of spices (144, 146, 147). Compared to cereals, and edible nuts, spices and plant-based food additives are highly prone to maintain higher humidity levels in their structure (148). On such occasions, opportunist APF can easily grow among stored spices to decrease the quality, taste, fragrance, color and marketability of these popular plant-based food additives (149). In Asian countries, in particular Iran and India, food spices (or additives) are indispensable parts of daily cooking and different forms of spices including raw and processed stuffs can be purchased from local providers. Due to bulk production of spices, these products short immediately after harvest would send for local and international markets. Recent studies on herbal products and spices of different locations of Iran indicated that AFB1 is the most prevalent AFs among these spices and nearly 100% of analyzed red pepper samples were contaminated with AFs (148). In another interesting study, different samples of commercial spices in Iran has been analyzed using HPLC method to identify the quantity and abundance of culprit AFs (149). According to these results, spices such as cinnamon, turmeric, black and red pepper diagnosed with different concentrations of AFs (149). Similar to the previous results (148), AFB1 was the most domineering AFs among analyzed spices, though different concentrations of AFB2, AFG1/2 were also observed among the evaluated samples (149). These results are in agreement with previous studies that confirmed herbal spices are remarkably prone to AFs contaminants (150, 151). Monitoring of spices marketed in Africa also showed that the Ethiopian ground red pepper was extremely contaminated with AFB1 in a dose of 250–525 μg/kg (152). Simultaneous investigations on Iranian and Indian spices to detect AFs contaminants purported that spices from these origins were differentially contaminated with AFB1 in a concentration of 63.16–626.81 ng/kg (Iranian samples) and 31.15–245.94 ng/kg (Indian samples), respectively (153). More interestingly, the outcome of this investigation demonstrated that contamination of studied samples was not exceeded from EU standard concentration of AFB1 in spices (5 μg/kg) (153), though AFs metabolites were characterized in the monitored samples. In another study in Turkey, 93 spices and 37 medicinal herbs were evaluated to identify hazardous AFs derivatives. The results manifested that AFB1 was domineering fungal metabolite in nearly 32 herbs and 58 spice samples (154). Resultantly, the maximum concentration of AFB1 was found in cinnamon at the concentration of 53 μg/kg so that the measured concentrations in these samples were obviously exceeded from EU permissible values (154). Evaluation of marketed spices in Doha, Qatar, showed that Aspergillus and Penicillium spp. were the most prevalent fungi in these samples (155). Interestingly, this investigation demonstrated that five spices, including turmeric, black paper, chili, tandoori and garam masala, were contaminated with AFB1 (155). For the first four samples, the detected AFB1 concentrations have deviated from EU standards (155). Indeed, according to the outcomes of a recent meta-analysis on the occurrence of AFB1 in red pepper, the prevalence of AFB1 among studied samples was 50.8%, respectively (156). This study reported that the minimum and maximum concentrations of AFB1 were detected in Korean and Turkey samples in the concentration of 0.14 and 31.13 mg/kg, respectively (156). The growing body of evidence suggests health promoting medicinal plants are also prone to AFs contaminants because of their ability to support AFs-fungi growth (12, 157). In an interesting investigation the results displayed that AFB1 metabolite was found among medicinal plants with a significant prevalence percentage (12). The results also confirmed that other AFs metabolites, including AFB2 and AFG1/2, were also characterized in herbal supplies (12). These outcomes are in agreement with other studies that confirmed that AFs are commonly found among medicinal plants in different concentrations (157, 158). More interestingly, herbal products not only hosted AFs metabolites, but also various scientific reports confirmed that OTA is another hazardous mycotoxin found in these herbal stuffs (159). In some studies that revolved around the occurrence of AFs in medicinal plants, spices and herb-teas incongruous results come to view in which spices were contaminated with AFs while medicinal plant samples from tropical countries were free of these hazardous fungal metabolites (36). Evaluation of medicinal plants fungal flora also showed that Aspergillus species were the most superior fungal strains isolated from target medicinal herbs (160). Monitoring of medicinal herbs in Thailand for AFs contaminants showed that these fungal toxins in the range of 1.7–14.3 ng/g have been occurred in these samples and AFB1 was abundantly detected in the evaluated herbal products (161). Although medicinal herbs are at risk of AFs pollutants, however, it is trustworthy to note that these herbal supplies have the ability to produce specific metabolites to detoxify AFs metabolites. In this regard, in vitro studies have shown that aqueous extracts of medicinal herbs such as Centella asiatica, Hybanthus enneaspermus and Eclipta prostrata displayed nearly 70% degradation of AFB1 (162). Meanwhile, it can be said that the concentration of characterized AFs in medicinal plants and herbal supplies depends on the type of plants, herbal processing methods, storage condition, variation of grown mycotoxigenic fungal strains, temperature and humidity of storage sites, secondary contaminations during herbal supplies storage, and the infestation of pests and plant pathogens to stored herbal products (163). Contaminated herbal products in local markets are potential health risks to consumers because there are no accurate scrutinizing systems to monitor local herbal suppliers and identify contaminated commodities (164). The local and international markets of medicinal herbs are flourishing yearly, and it is now valued at more than 100 billion dollars (165). Therefore, by regular risk assessment of spices and herbal supplies as well as increasing the number of monitoring gates for local and international markets and also by increasing the quality of processed spices and herbal products and suitable storage and packaging of these supplies, official health and agricultural organizations can significantly mitigate the prevalence of AFs contaminations in this industry (144). As shown in Figure 8, many herbal products and spices are marketed outdoors in local markets, which in turn might lead to post-harvest contaminations. Because herbal and spice providers might be not cautious enough to identify the source of AFs contaminants; sequentially, AFs-contaminated products may be intentionally or inadvertently passed to consumers, eventually leading to the progression of nCDs. Therefore, increasing the awareness of herbal farmers, manufacturers, and consumers is an effective strategy to eliminate AFs in such herbaceous products (166). As discussed in previous sections, like many other foodstuffs, spices and herbal products were prone to AFs contaminants (148). In some cases, however, the occurrence of AFs is not in detectable concentrations. Additionally, due to the increasing demand for herbal spices and plant-based food additives (142), many national and international suppliers adhered to spice markets, leading to intensified cultivation and extensive processing of these products. The lack of fundamental infrastructures and suitability of climate factors enhanced the growth of APF, consequently leading to an increased level of detectable AFs contaminants. Therefore, to mitigate the total level of AFs in spices, medicinal herbs, and other popular plant-based food additives, local and international herbal markets, however, should be constantly monitored for detection of AFs metabolites (167, 168). Keeping in mind the most popular quote perhaps assigned to Hippocrates (400 BC) “let food be thy medicine and medicine be thy food” (169), healthy foods are indispensable parts of our dietetic regime and are complement to modern pharmacology (169). Therefore, preparing foodstuffs from mycotoxin-free sources not only improve our lifestyle, but also can decrease the progression of nCDs and improve the quality of daily diet. In this regard, to bring concentration of AFs in spices and herbal products down, taking precautionary actions, such as raising public awareness, might support consumers to purchase healthy and AFs-free products (170, 171). According to scientific data, AFs contaminants in herbal products used in traditional medicine mainly occur in two stages during drying/processing and storage of target herbs/spices (172–174). Therefore, the lack of strict regulations on herbal ingredients used in traditional medicine might increase the occurrence of AFs in these health-promoting products (174). It is pivotal to implement rigorous regulations on the production, processing, packaging, manufacturing and exporting/importing of herbal products that are prone to AFs contaminations. Improving packaging systems (108) and developing standard infrastructures for the distribution, storage, and transportation of medicinal herbs and spices can also help consumers to use more safer products (175, 176). Providing safety guidelines for preparing herbal products and spices helps the public to broaden their knowledge about the consequences of AFs and associated health complications; therefore, leading to increasing demands for mycotoxin-free commodities and a healthy lifestyle. In the next sections, we will discuss the circulation of AFs in the human body and major AFs health consequences reported in the literature. This helps readers to a better understanding of AFs biological properties and their role in developing human chronic diseases. Studies have shown that AFs metabolites are prone to bind human serum albumin (HSA) (177). HSA is one of the most prevalent proteins in human blood plasma (178). HSA is a globular protein produced in the liver and constructed from a monomeric structure with several subdomains (179, 180). HSA functions as a carrier in the human body to transport fatty acids, drugs, hormones, and other biomolecules (178). This multifunctional, negatively charged, and non-glycosylated protein also participates in the regulation of plasma osmotic pressure. HSA is formed from 585 amino acids, and its 3D crystallographic structure is well-documented (179). Structurally, HSA spatial conformation is formed by a heart-shape molecule, possesses three helical domains (I, II, and III), and is divided into A and B subdomains (IA, IIA, IIIA and IB, IIB, and IIIB) (178, 179, 181) (Figure 10). In the structure of HSA, there are two distinct binding sites, including Sudlow's I and II, each prone to bind different types of chemical agents (178, 182). Generally, negatively charged large heterocycles bind to site I, whereas small molecules prefer to interact with site II (182). The decreased concentration of HSA in blood plasma is associated with AD (183), cancer, obesity, diabetes, heart failure, stroke, and venous thromboembolism (178, 184). HSA plays a critical role in the tissue distribution of AFs metabolites (185). To date, only few studies have been conducted on the possible interaction of AFs and HSA binding sites. The available evidence suggests that AFs could non-covalently bind to HSA binding cavities (185). Evaluating the binding mode of chemical ligands to target receptors helps researchers to characterize the molecular behavior of these molecules in vivo (186). To understand how AFs metabolites might interact with HSA, we computationally investigated the binding affinity of AFs and AF-derived metabolites and HSA binding sites. As explained, only few studies are available to show the exact binding mode of well-known AFs to HAS (185). Therefore, to broaden the literature consistency on this topic, as part of this review 14 AFs metabolites, including AFB1/2, AFG1/2, AFM1/2, AFB2a/G2a, AFP1, AFH1, aflatoxicol, AFB1exo-8,9-Epoxide-GSH, AFB1exo-8,9-Epoxide, and aflatoxin-N7-guanine have been structurally prepared and docked into HSA using supervised and blind docking protocols (187, 188). The docking results for the interaction of AFs metabolites and HSA binding site I showed that docked AFs possessed different binding affinities to interact with HSA site I residues. The calculated binding energies for these metabolites ranged from −6.2 to −9.5 kcal·mol−1, respectively (Figure 9). AFB1exo-8,9-Epoxide, AFB1, aflatoxicol, AFG1/2, and AFM1 significantly formed H-bonds and Van der Waals forces to interact with HSA binding site I. Interestingly, docking results demonstrated that AFs metabolites such as aflatoxicol, AFB1, and AFB1-exo-8,9-epoxide, might also interact with HSA subdomain IB. Previous studies suggested that AFB1/2, AFG1, and AFM1 mainly interacted with HSA binding site I (185), but the binding affinities of the remaining AFs metabolites are not comprehensively investigated in the literature. In the case of other mycotoxins such as zearalenone, the binding mode inhibitory assays showed that this fungal toxin could strongly bind to a non-conventional binding cavity between Sudlow's site I and II (189). More interestingly, OTA has two binding sites in the structure of HSA with different binding constants so that the highest binding affinity for this toxin was observed for subdomain IIA HSA protein (190). Other interesting experimental studies also confirmed that AFB1 is mainly bound to HSA in binding site I located in subdomain IIA with a binding affinity around 104 M−1 (177, 191). Similarly, the results of spectroscopic and computational assays also determined that AFB1 and AFG1 also interacted with subdomain IB residues in HSA (192). AFB1 also displayed a similar binding affinity to interact with bovine serum albumin (BSA) binding site I with a binding constant of nearly 4.20 × 104 M−1 (193). These outcomes together demonstrated the precise interaction of AFs and HSA, leading to a better understanding of toxicokinetic properties of these mycotoxins. Therefore, displacement of HSA-AFs complexes has been suggested as a therapeutic strategy to diminish the affinity of these mycotoxins to HSA and decrease the tissue delivery and uptake of AFs (177). Decreasing the affinity of HSA to AFs with chemical compounds sharing similar binding patterns with higher affinities in comparison to AFB1 might bring down the toxicity of this mycotoxin for the human body (193). Studies have shown that natural products such as PPs interfere with the interaction between AFB1 and HSA and reduce the transportation of AFs to delivery locations (177, 194). More interestingly, scientific outcomes reported that administration of vitamins A and E could reduce carcinogenic properties of AFs in studied animals (195–197), though controversial results on the protective roles of vitamin E in cancer therapies have been reported (198). These findings are in agreement with previous outcomes demonstrated that exposure to AFs metabolites is associated with plasma micronutrient deficiencies (199). However, mycotoxin metabolites could bind to HSA (189, 193); therefore, these toxins are easily transported to different parts of the human body and causing chronic health consequences (189). In the next section, we explain how metabolized AFs derivatives are accounting for prevalence of nCDs. The discussed nCDs have been selected based on the frequency of conducted studies on each field of interest. DM is a chronic metabolic disease mainly characterized by elevated blood glucose level and insulin deficiency (200, 201). More than 422 million people are expected to suffer from DM in which the number of DM affected people in low-income countries has steadily grown during the past decades (202). To date, various subtypes of DM have been identified by which scientists can treat affected individuals through observed symptoms. Generally, type 1 DM (T1D) and type 2 DM (T2D) are the two the most prevalent subtypes of DM, leading to thousands of deaths yearly. T2D is responsible for more than 95% of all diabetic cases, while T1D only represents 5% of diabetic individuals. T1D is more prevalent among juvenile people and is significantly dependent on insulin deficiency (188, 203). In contrast, T2D is insulin-free DM, by which affected people suffer from elevated blood glucose and associated complications (203). The emergence of modern drug design technologies leads to the development of potent anti-diabetic drugs. Different anti-diabetic medicines with specific molecular targets have presently been introduced into global markets (200). However, these drugs could not entirely suppress the complications of DM (200, 201), in turn, leading to an increased economic healthcare cost that allocated on caring for DM-affected people. Recently, the role of exposome measurements has been highlighted in the progression of DM (204, 205). Exposome-associated factors can be divided into external and internal factors. External factors are features that directly linked to nearby environment such as pollutants, chemical materials, lifestyle and dietary regimes (205). Instead, internal factors are accounting for epigenetics alterations, gut microbiota and relevant molecular processes (204). This ongoing paradigm helps to understand how and where exposure to environmental factors lead to the progression of MetSys and other human diseases (204). As a complicated metabolic disorder, DM progression depends on various factors (206–208). By considering exposome-associated factors in the development of DM, it is worthy to note that the adopting of a healthy lifestyle can decrease the incidence of this metabolic disorder (206). Environmental factors such as exposure to hazardous chemical agents (209) and toxins might increase the onset of DM (210). Indeed, the complex interaction between environmental and genetic risk factors might worsen the health complications of DM (206, 211). Biological toxins might act as health hazardous diabetogenic agents to disrupt normal function of the human body in controlling blood sugar levels and associated signaling pathways (212). In this regard, evidence-based studies imparted that long-term exposure to particular types of AFs, such as AFM1, might increase health risk factors for developing T2D and other metabolic disorders (213). Interestingly, long-term exposure to AFB1 increased liver injuries in mice, disrupting blood glucose levels, insulin sensitivity, and a high chance of inducing liver cancer (39). Recent studies have shown that type 1 diabetic mice exposed to AFB1 showed a significant reduction in MUP1 levels, in turn, indicated an elevated blood glucose level and decreased insulin sensitivity (214). Molecular mechanisms underlying the diabetogenic effects of AFs are not completely understood, however recent investigations reported that AFs metabolites, in particular AFB1, might influence the regulatory switches of specific signaling pathways, genes, transcription factors, and receptors such as IGF2 and IGF1 receptor IGF-IR (215). In this regard, the evidence suggests that the increased level of IGF2 expression in pancreatic islets is associated with the onset of DM and dysfunction of β-cells (216). The overexpression of IGF2 affects the functionality of β-cells, leading to chronic endoplasmic reticulum stress and dysfunction of pancreatic islets (216). Indeed, the evidence also suggests that IGF1 plays a critical role in DM by lowering blood glucose levels and insulin secretion (217). Therefore, the interaction of AFs with such molecular targets might negatively cause molecular abnormalities, which later lead to the development of DM. Additionally, hepatorenal injuries, lipid peroxidation, DNA damage, oxidative stress, and inflammation are other symptoms of animal models exposed to AFB1 metabolite (218). In an interesting study (219), it has shown that the long-term exposure to mycotoxins was significantly associated with DM development in affected rats (219). In this finding, OTA could remarkably increase blood glucose levels, cause damage to pancreatic islets, and decrease insulin secretion (219). The cross-talks between the progression of MetSys and prevalence of HCC have been widely investigated (220, 221). The evidence suggests that MetSys might be connected to the progression of cancer (221). Yesheng et al. meta-analysis (221) reported a possible link between MetSys and pathogenesis of HCC among Euro-US societies, though there has not been association between HCC and MetSys clinicopathological feature (221). In another study, Marchioro et al. (222) reported that in broilers chickens supplemented with a mixture of AFs (B1/2-G1/2) in the concentrations of 0.7–2.8 mg/kg for 42 days, chickens' performance features and enzymatic activity of pancreas have notably been altered (222). The outcomes imparted that long-term chronic exposure to AFs mixture increased the activity of pancreatic α-amylase and lipase while trypsin levels has been affected by the maximum concentration of AFs mixture (2.8 mg/kg) (222). The literature has disclosed that AFs altered the accumulation of lipids droplets and lipoproteins in addition to the dysregulation of lipid metabolism-related genes (CHO, TAG, PHOL, MDA, Lipc, Lcat, Scarb1, etc.) (223, 224). The evidence imparted that the dysregulation of fatty acids, cholesterol, and other health affecting lipids biosynthesis and metabolism is accounted for the progression of DM (225). Therefore, exposure to AFs in the dose-dependent fashion might contribute to the development of DM and cardiovascular diseases via alteration in the body lipids metabolism pathways (223, 226), though this claim requires future confirmation. In rats exposed to penitrem A, a highly toxic mycotoxin from Aspergillus genus, a considerable diabetogenic properties has been observed (227). In this regard, chronic exposure to OTA (45 μg/daily diet) for 6–24 weeks caused a significant decrease in insulin levels and increase in blood glucose and glucagon levels (227). The observed diabetogenic activity of OTA is attributed to its impact on degeneration of pancreatic Langerhans islets (227). The elevated diabetogenic effect of mycotoxins in combination with chemical agents such as insecticides has also been investigated (228). Correspondingly, the outcomes displayed a remarkable synergistic interaction between mycotoxins and chemical agents in the onset of DM by increasing blood glucose and dysregulation of liver enzymes (228). In a cross-sectional study conducted on Guatemalan participants, the outcomes manifested a significant association between AFB1-albumin adduct levels and pathogenesis of DM (229). Additionally, there was no significant association between AFB1-adducts and the progression of other metabolic diseases such as central obesity, obesity, non-alcoholic fatty liver diseases (229). This result, however, was aligning with previous animal-based studies that confirmed the association between exposure to mycotoxins and the onset/progression of DM (219). The evidence imparts that fungal toxins may increase the susceptibility to the onset of MetSys; therefore, well-designed human-based studies are needed to show how mycotoxins and AFs may contribute to the progression of MetSys (213, 230). According to WHO statistics (231), there are more than 50 million AD-affected people worldwide such the statistics that have projected to increase by 2050 (187, 232). This prevailing neurodegenerative disorder is chiefly characterized by a remarkable decline in thinking, memorial dysfunction, unpredictable behaviors, language problems, and cognitive impairments, in turn, sequentially causes significant damage to the brain cells (231–233). Accordingly, the lesion of brain cells, the accumulation of amyloid plaques, neurofibrillary tangles, oxidative stress, NIF, and synaptic dysfunction are typical clinical symptoms of AD (187, 232, 234). Different hypotheses have been postulated for the progression of AD; however, it is not completely clear which of molecular switches drives the inception of AD to cause obvious damages to the brain (234, 235). Scientists suggested that environmental and genetic risk factors, exposure to chemical pollutants, heavy metals, mycotoxins, lifestyle, age, infections, cardiovascular dysfunctions, T2D, cellular senescence, and head injuries may play a critical role in pathogenesis of AD (232, 236–238). Studies have shown that the AFs metabolites can alter various brain enzymatic actions, leading to AD development. For instance, in rodent models, exposure to AFB1 could significantly decrease the activity of brain protein kinases (239). The SH-SY5Y human neuroblastoma cell lines exposed to 100 and 50 μM AFB1 and FB1 mycotoxins for 24 h, manifested a significant increase in ROS formation, though the trace of endoplasmic reticulum stress was not observed (240). On the contrary, in adult male rats treated with 25 μg/kg/week AFB1 for 8 weeks, AFB1 could trigger obvious neurotoxicity, inflammatory responses, oxidative stress and, anxiety and depression-like behaviors (241). The finding showed that AFB1 supplementation was linked to a reduction in the activity of GSH, GST, SOD, and GSH-Px enzymes and increased MDA, IL-1 and TNF-α levels in right region of cerebral tissues (241). The AFB1 also negatively influenced the distribution of astrocytes in rats' cerebral cortex and hippocampus (242). The effects of AFs on the BBB were also investigated such that the outcomes showed that AFs (in particular AFB1) could alter mitochondrial gene expression profile in the human BBB cells model (243). More interestingly, AFB1 could inhibit the electron transport chain function, affect ATP synthesis and dysregulate key genes in mitochondria (243), leading to genetic mutations and DNA damage (244). The AFB1-NIF is attributed to the interaction of AFB1 metabolized derivatives with neuroinflammatory signaling pathways (245). It is now well-established that neuroinflammation promotes the pathogenesis of AD and other neurodegenerative diseases (246). The molecular mechanisms underlying neuroinflammation have partially been investigated, however, little is known on how exposure to mycotoxins may have impact on the incidence of AFB1-NIF (247). To elucidate the AFB1-NIF mechanism of action, it is important to take this question into consideration how AFB1 metabolites may alter the NIF signaling pathways? Briefly, the activation of microglial cells elevates glial neuroimmune responses (248, 249). Next, CNS-related genes might be up- and/or down-regulated, sequentially resulted in the reactivity of astrocytes and expression of pro-inflammatory molecules such as IL-1/1β/6, INF-γ, and TNF-α (248). Activation of these neuroinflammatory signaling mediators will increase the ROS/RNS levels in the brain, leading to a significant oxidative/nitrosative stress and neuronal damage (246, 250). Studies have proven that TLRs, MAPK, MyD88, CxCR4, PI3K/AKT, mTOR, COX-2, iNOS, Nrf2, HO-1, γ-enolase, STAT, AMPK, JAK, and NF-κB signaling pathways are major components of NIF (248, 251, 252). The alteration of kynurenine/tryptophan ratio (253), dysregulation of intracellular protein kinases (PKs) (254), the loss of neuronal integrity (255), and dysregulation of neurotransmitters signaling circuits are other pivotal components of NIF in response to brain abnormalities (256). The evidence introduced thus far supports the scenario that AFB1 enhances the secretion of pro-inflammatory cytokines such as TNF-α and IL-6 in CNS-derived cells, leading to the promotion of immune responses and significant oxidative stress in the CNS (247). These elevated level of neuroimmune reactions and activated signaling pathways in astrocytes and glial cells have been reported as consequences of AFs (257). Interestingly, low and high-dose exposure to AFs might alter the activity of brain signaling cascades based on exposure time duration and toxicodynamic properties of culprit AFs (258). The acute exposure to AFs could notably affect the expression of genes and enzymatic activation in the brain. In rats, acute treatment with AFB1 influenced the activation of protein kinase C by phosphorylation of Ser957 position in the cerebral cortex (259). CCK is another critical protein kinase in the brain accounted for pathogenesis of AD (260). Studies have revealed that the functionality of brain cells depends on ATP molecules produced by CCK (245). Blocking CCK activity is associated with energy depletion in the brain, which can lead to significant oxidative stress and brain abnormalities. AFB1 inhibited the CCK enzyme to decrease ATP metabolism and trigger oxidative stress in the brain (245, 261). Park et al. reported that AFB1 decreased human astrocyte cell proliferation by arresting cell cycle, sequentially induced the mitochondrial dysfunction and apoptosis of astrocytes (262). Accordingly, the interesting part of this finding that is AFB1 dysregulated calcium hemostasis and increased ROS formation, leading to neurotoxic effects on astrocytes cells in vitro and in vivo (262). In female Wistar rats given 15.75 μg/kg/orally for 8 weeks, the outcomes suggested that AFB1 decreased the distribution of astrocytes in frontal cortex without effect on neuronal numbers (242). In contrast, AFB1 increased neuronal number and decreased astrocyte distribution percentage in the hippocampal CA1 subfield. Importantly, the withdrawal of AFB1 restored the observed changes in rat brain (242). In the support of these outcomes, Alsayyah et al. reported that the severity of chronic neurodegenerative effects of exposure to AFB1 is associated with astrocyte immune responses and alteration of brain enzymes (239). Accordingly, chronic exposure to AFB1 altered the activity of antioxidant enzymes (GPX, CAT, SOD, GSH), increased the activity of AP and LDH, and decreased CK activity (239). The observed AP and LDH increased activity are attributed to neuronal death, astrocytes damage, and necrosis (239). This finding also indicated that the chronic side effect of long-term exposure to AFB1 depends on the passed quantity of these toxin from the BBB and duration of exposure (239). In another study, animals feed with 5 ml AFB1 for 8 weeks also showed noxious neuronal degenerative changes in cerebral cortex (263). Another animal-based investigation also manifested that AFB1 increased the activity of AChE and ADA enzymes (264). The up-regulation of these enzymes might be responsible for elevated level of inflammatory responses due to tissue damage. Additionally, AChE and ADA may be contributed to clinical signs of apathy because of their participation in neuromodulation and neurotransmission (264). By the way of illustration, the evidence suggests that AFB1 triggered acute neurodegenerative consequences in the CNS, induced encephalopathy by influencing glutamate neurotransmitters, increased ATP depletion, modified brain catalase, SOD, MDA levels, and GST concentrations (245). What is outstanding in these outcomes is that differential exposure to AFB1 induces oxidative stress (265, 266) and neuronal damage in the brain (245). Additionally, exposure to AFB1 is associated with reduction in CNS phagocytic ability, increased levels nitrosative stress, increased expression of cytokines (TNF-α and IL-1β/6/8/10) (267), induced microglia cell apoptosis, dysregulation of p-NF-κB signaling pathway (268), significant alteration in brain integrity, substantial DNA damage, S-phase cell cycle arrest (269), and other neuroimmunotoxic complications (245). AFM1 also could degrade the BBB structure by influencing astrocytes, vascular endothelial and microglia cells to trigger remarkable neurotoxicity in the brain (262, 270). These data indicated that AFB1-NIF affected brain enzymatic and none-enzymatic reactions as well as other CNS molecular components, ultimately leading to the onset of AD. As an epidemiological standpoint, the clinical framework of AD pathogenesis and exposure to AFs has still not transparent, consequently further well-supervised trials should be conducted to know how do exactly these mycotoxins contributed to molecular dysfunctionalities in the brain (245). As explained earlier, the neurotoxicity and side effects of AFs in the brain have been documented through in vitro and animal studies, therefore, for cautionary reasons the elimination of these mycotoxins in food/feed should be repeatedly followed to reduce their clinical end effects. Cancer is one of the most lethal chronic diseases, leading to millions of deaths yearly (271, 272). According to global statistics (271, 273), the prevalence of cancers is remarkably increased during the past decades (271, 273). Still, an efficient and safe medicine for the treatment of cancer has not been introduced (274). Among all influencing environmental and genetic risk factors, the carcinogenic effects of fungal toxins in contaminated foods raised global concerns about the prevalence of cancer (271). As discussed, studies have reported that AFs are toxic, carcinogenic substances (154). These fungal metabolites disrupt the normal activity of signaling pathways, gene expression, and enzymatic activities in the human body. The evidence suggests that long-term exposure to high concentrations of AFs remarkably influences liver and kidney function (19). Briefly, AFs affect the expression level of many genes involved in phase I and II metabolism in the human body (275). Compared to other xenobiotics, the highly liposoluble AFs are rapidly absorbed at the site of exposure (276). Studies have shown that respiratory tracts and gastrointestinal organs are two major sites for the absorption of AFs into the body (276, 277). AFs affinity to carrier proteins in the body helps these carcinogens to enter the blood and circulate around tissues and organs (191, 194). Highly toxic and reactive AFs metabolites bind to DNA to form AFs-DNA adducts (278, 279). Studies have reported that the interaction of AFs and DNA causes significant damage to DNA and associated biological processes such as transcription and chromatin packaging (280). Binding AFs to DNA and key enzymes in the liver and other organs can induce cancer in different ways. Studies suggest that occurring mutation in specific sites of DNA and proteins is linked to the pathogenesis of cancer (281). The formation of AFs-DNA adducts can affect the topology of packaged DNA and DNA conformation (276, 278). The exo-8, 9-epoxide metabolites of AFB1 are surprisingly prone to construct DNA adducts. This AFB1 metabolite is highly reactive, and its genotoxic effects have been well-documented (276, 278). Aflatoxin-N7-guanine metabolite also binds to the DNA, induces transversion mutations (pyrimidine ⇆ purine), affects the expression of tumor suppressor proteins and transcription factors, and ultimately dysregulates cell cycle events (276, 278). Acute exposure to AFs pollutants disrupts the fundamental function of genes involved in the glutathione pathway (282, 283). Studies reported that AFs could dysregulate the cellular levels of PKC, PKA, p53, CDK, NF-κB, Bcl2, CKI, and cyclins (276). Disruption of mitochondrial function, ATP synthesis, and mitochondrial gene expression profile have been observed as side effects of AFs in the animal and human bodies (276). The growing evidence suggests that AFs also influence the expression of xenobiotics metabolism genes, leading to upregulation of CYP3A4 and pregnane X receptor (PXR) (275). In this regard, scientific outcomes have proven that the activation of PXR upregulates phase I and II metabolism genes and proteins such as CYP2B6, CYP2C9, CYP3A4, CYP3A7, UGT, GST, and SULT enzymes (275, 284). The dysregulation of MDR1 and OATP2 genes after activation of PXR has also been reported in several studies (275, 285). Studies reported that in primary liver cancer and cirrhosis, these genes are significantly overexpressed, leading to highlighting their critical role in the pathogenesis of cancer (286, 287). Considerable number of scientific investigations also demonstrated that exposure to AFB1 could disrupt the expression of ERK, PKC-β, COX-2, caspases3/7/9, ASK1, SAPK, STAT3, E2FA, MYC, Bax/Bak, PUMA, CDKN1A, p21, DNA/RNA polymerases, PLK, MAPK, and TRPs signaling pathways (288–291). Additionally, continuous exposure to AFB1 promotes epigenetic modifications in liver cells (292). As detailed in the literature, AFB1 triggered various epigenetic alterations such as increased levels of aberrant DNA methylation, histone post-translational modifications, and up/down-regulation of non-coding RNAs (ncRNAs) and transcription factors (TFs) (293). AFB1-induced epigenetic drivers in the liver cells are associated with the development of hepatocellular carcinoma (14). Both AFB1-Lysine-protein and AFB1-DNA adducts inhibited the fundamental molecular process of infected cells by preventing transcription/translation of target genes. AFB1-based epigenetic alterations potentially increased the level of genomic mutations, inhibited the interaction TFs-gene-promoter complexes, modified the normal pattern of ncRNAs expression (e.g., miR34a/21/221 and lncRNA-H19), and altered signaling pathways (14, 293). The evidence suggests that AFs strengthen the consequences of hepatocellular carcinoma risk factors such as DM, obesity, over-drinking alcohol, and viral infections (HBV, HCV) to influence the onset and progression of this catastrophic diseases (294). On the other hands, studies proven that AFB1/M1 end effects are not limited to liver cells so that it was shown that these AFs could also influence the metabolic profile of kidney (295). For example, in CD-1 mice co-treated with AFB1+AFM1 (0.5 mg/kg + 3.5 mg/kg) for 35 days, the results manifested that AFB1/M1 promoted the onset of oxidative stress in mice kidney, altered proline dehydrogenase and L-proline levels, sequentially induced upstream apoptosis, in turn, leading to kidney damage (295). Evaluation of PPI networks of genes/proteins involved in AFs and xenobiotics metabolism unraveled that these genes (or proteins) constructed a network of interactions with various key proteins in cancer-linked signaling pathways. Up or down-regulation of these genes provides a ground for toxicity of AFs. In the human body, genes including CYP1A2, CYP3A4, CYP2A13, GSTT1, GSTM1, EPHX1, AKR7A2, and AKR7A3 are driving AFs metabolism (296). The upregulation of these genes affects the expression of upstream/downstream genes, leading to a remarkable disturbance of cancer-associated signaling pathways (296) (Supplementary Figure S2). Studies have also shown that genetic polymorphisms in the structure of dominant genes involved in AFs metabolism might increase the risk of developing cancer (297). Due to the presence of complex interactions between genes, proteins, and transcription factors in the human body, modification of gene expression levels after exposure to AFs can surprisingly affect the genotoxic, immunosuppressive, and mutagenic properties of these fungal metabolites (278). However, our literature searches showed that AFs interfere with DNA integrity, increase the secretion of pro-inflammatory factors, cytokines, and chemokines, inhibit DNA repair mechanisms, induce genomic instability, increase lipid peroxidation, induce DNA damage, cause tissue necrosis and organ failure, and dysregulate innate and adaptive immunity (290, 298–300). Reactive metabolites generated from AFs metabolism are highly toxic for body tissues, leading to obvious oxidative stress and the accumulation of ROS/RNS radicals (301). Indeed, DNA-AFs structures can inhibit protein synthesis by disrupting the interaction of transcription factors and polymerase enzymes with DAN or increasing the occurrence of mutation in promoters and coding sequences (290). Considering the PPI networks, it is trustworthy to note that the dysregulation of the AFB1 metabolism network might be associated with disruption of gene expression networks in hepatocarcinoma, colorectal, pancreatic, melanoma, thyroid, bladder, and other types of cancers (Supplementary Figure S3). It seems that chronic exposure to AFs expedites a wide range of molecular irregularities in cells, ultimately leading to the pathogenesis of cancer, and more probably other nCDs. According to WHO statistics (302), nCDs are accounted for more than 41 million deaths yearly, in turn, 77% all deaths occurred in low or middle-income countries (302). Environmental and genetic risk factors, such as exposure to biological toxins, pesticides, air pollutants, smoking, alcohol, unhealthy diet, cholesterol, obesity, physical inactivity, mental stress, work tension, elevated blood glucose, and blood pressure are common factors in triggering nCDs (303–306). Studies manifested that withdrawal of environmental risks factors and decreasing the exposure rate to health hazardous substances are effective strategies in preventing nCDs (306). In the case of AFs, however, it is worthy to point out this fact that the elimination of these mycotoxins in food/feed is impossible, leading to the continuous existence of these risk factors in the environment. As evidenced in the previous sections, exposure to AFs is associated with development of liver cancer, though its role in the progression of MetSys and neurodegenerative diseases has not extensively been investigated. Therefore, this part of our review, by focusing on data obtained from high-tech omics-assisted outputs, describes how AFs may alter biological networks and/or gene expression profile being effective in the onset and development of nCDs. Integrated transcriptomics and metabolomics analyses conducted on male Fischer rats given 0.25–1.5 mg/kg/b.w. AFB1 for 7 days reported that exposure to low-high concentrations of AFB1 is associated with dysregulation of tumor suppressor genes (at least 27 critical genes were up/down-regulated), antioxidant enzymes, cyclins, cyclin-dependent kinases, cytokine receptors, and inflammatory signaling pathways (307). According to this finding, acute exposure to AFB1 resulted in p53-induced oxidative damage, dysregulation of gluconeogenesis, and lipid metabolism, in turn, leading to hepatotoxicity of AFB1 (307). In ducklings exposed to 0–40 μg/kg/b.w. AFB1 for 2 weeks, the RNAseq data disclosed that at least 749 transcripts responded to chronic exposure to AFB1 (308). Interestingly, these genes were critical components of phase I/II metabolism (CYP1A5, CYP2H1, CYP2K1, CYP2F3, etc), antioxidant enzymes (GST1/3/K1), fatty acid metabolism (ACAA1, ACOX1, ACAT1, FADS1, FASN, HADH, etc), apoptotic genes (CASP3, CBR1, CCBL1, PPIF, KRT18, etc), protein kinases (PLK2), oxidative responsive genes (AKR1A1, AR, AO, FMO3, GPX4, NQO1, TXN, TDO2, etc), cell cycles and cancer-associated genes (PRELID1, PLK2, UGT1A8, MDM2) (308). This study manifested that phase II detoxification enzymes such as GST1, GSTK1, GST3 were up-regulated under chronic exposure to AFB1 (308), though previous studies found that these genes were down-regulated or not significantly influenced due to the difference between animal models selected for omics-based investigations (308, 309). In Wistar male rats received 100–200 μg/kg intraperitoneal AFB1 for 4 weeks, the high-throughput gene expression analyses showed that exposure to AFB1 altered gene and lncRNAs expression (310). According to this finding, the identified differentially expressed lncRNAs were associated with upregulation of genes involved in cancer, apoptosis, DNA repair, and cell cycle arrest (310). Because several genes such as Bcl2, MAPK8, and NFKB1 were up-regulated after exposure to AFB1, therefore, it has suggested that apoptotic-associated responses to AFB1 exposure played a critical role in hepatotoxicity of this mycotoxin (310). Similarly, in another omics-assisted study on chickens, the outcomes showed that exposure to AFB1 dysregulated genes involved in apoptosis and lipid metabolism in liver (311). This finding highlighted that Bcl-6 gene was down-regulated whereas PPARG was up-regulated, in turn, might be leading to hepatic fat deposition and hepatocellular apoptosis (311). The whole transcriptome of BFH12 (bovine fetal hepatocyte cell line) exposed to 0.9–3.6 μM AFB1 for 48 h displayed that AFB1 significantly dysregulated the expression of genes involved in inflammatory responses, apoptosis, oxidative stress, cancer and xenobiotics metabolism (312). Indeed, this investigation disclosed that exposure to low-high concentrations of AFB1 markedly influenced the activation of TLR2, p38β MAPK, AP-1 and NF-κB signaling pathway and pro-inflammatory cytokines (312). This outcome is important because a large quantity of studies published on toxicity of AFB1 for the animal and human bodies addressed oxidative stress and inflammatory responses as health consequences of this carcinogenic mycotoxin. Bao et al. (313) reported that in Caco-2 cell lines exposed to 0.0005–4 μg/ml AFM1 for 48 h, totally 165 genes were down-regulated after exposure to AFM1. This finding demonstrated that exposure to AFM1 is associated with dysregulation of CDK1, AMPK, SOS1/Akt signaling pathways that are involved in cell cycle arrest (313). The metabolomics and transcriptomics analyses of mice liver and serum showed that co-exposure to AFM1 + OTA (3.5 mg/kg/b.w. for 35 days) significantly affected the phase I metabolism enzymes (ALT, AST, glutamyltransferase) by increasing their levels (314). Additionally, due to synergistic side effects of AFM1 + OTA, the accumulation of lipid droplets and liver steatosis were observed in co-treated groups (314). The metabolome profile of groups co-exposed to AFM1 + OTA disclosed that in liver and serum lysophosphatidylcholines levels were significantly increased (314). In another transcriptomics study conducted on wild and domesticated turkey exposed to 0.015–0.32 mg/kg AFB1, a significant dysregulation in phase I and II metabolism, inflammatory responses, and apoptotic genes was observed, in turn, provided another evidence for side effects of AFs on cellular and molecular targets (315). As evidenced in these findings, chronic and acute exposure to AFs lead to remarkable alterations in cellular and molecular pathways. Based upon the given standpoints in the previous sections, it is important to note that continuous exposure to low-high doses of AFs certainly promotes the onset and progression of nCDs in societies with higher risks to encounter AFs in food/feed. These studies also addressed the role of inflammation, apoptosis and oxidative stress in progress of AFs-induced consequences, which in turn, can be considered as a scenario to characterize the susceptible signaling pathways and enzymes to develop future clinical management strategies. On the other hand, the growing body of evidence postulates that the public health impact of climate change is negative, might leading to expedite the onset of nCDs (316). Climate change alone is one of the most important drivers of AFs production under future conditions (72). Other climate change associated environmental risk factors such as improper storage of crops, infestation of secondary pests/pathogens, humidity and temperature can strengthen the production of AFs and other carcinogenic mycotoxins (53, 62, 72). Therefore, more contaminated foodstuffs will be entered local and international food chains, in turn, probably leading to increased risk of nCDs (305). The high-throughput outcomes of multi-omics-based data disclosed that AFs mainly have impact on the inflammatory responses and antioxidant defensive system. In HCC, however, dysregulation of inflammatory responses and antioxidant enzymes promoted the onset of hepatocarcinoma cancer (317–319), though these factors function in corporation with a complex network of protein-gene interactions. However, for other nCDs such as DM and AD, the mechanism underlying inflammatory responses, antioxidant enzymes, and cellular signaling cascades is markedly a sophisticated process and dysregulation of these pathways has a significant negative effect on the progression of diseases. Therefore, elimination of AFs risk factors in food/feed might, at least, provide a ground to decline the onset of nCDs. Figure 10 represents a cross-link between exposure to AFs and the onset of nCDs. An exciting point should be highlighted that is nCDs such as cancer, AD, and DM shared a commonality in the occurrence of oxidative/nitrosative stress and inflammatory responses. As shown in Figure 10, the onset of inflammation is the core component of these nCDs, in turn, leading to cellular apoptosis, dysregulation of enzymatic activity and ultimately the severity of nCDs pathogenesis. According to FDA guidelines (320, 321), there has been no medicine to prevent the poisoning of AFs, though several protective interventional substances (e.g., NovaSil clay, chlorophyllin, oltipraz, sulforaphane, and tea PPs) have been addressed (322). Safe withdrawal of AFs, using modern agricultural practices, improving food processing methods, and preventing the consumption of contaminated foods have been reported as scenarios to reduce AFs quantity in food commodities (320–322). Recent findings suggested that the supportive and symptomatic care is a reliable health management strategy to reduce the poisonous effects of AFs (323). Correspondingly, using specific carbohydrate-rich and protein-restricted diets followed by administration of vitamins (e.g., B and K) can be helpful in the suitable prevention of AFs end effects (323). However, preventative methods should be cost-effective, and available for all individuals. By considering the link between climate change, onset of nCDs, and AFs health complications, it can be said that phytochemicals and plant-based products are contemporary solution in preventing the poisonous effects of AFs. It is now well-documented that plant secondary metabolites are health promising compounds in the cornerstone of human diseases prevention (324). Medicinal herbaceous metabolites are ubiquitously found in fruits, vegetable, and even inedible plant materials (324). Efforts to reduce the complications of AFs demonstrated that plant-based products such as extracts, teas, essential oils, pure metabolites can decrease the toxicity of AFs by maintaining cellular normal function (324, 325). These herbaceous materials not only display health promoting properties in detoxification of AFs, but also prevent the formation of AFs in producing molds (326). Considering the link between exposure to chronic/acute doses of AFs and progression of nCDs, and also paying attention to this fact that there is no antidote to treat AFs poisonous effects (323), screening natural products pools to identify anti-AFs substances can alternatively help scientists in developing practical medicines against these biological carcinogens (20). Recently, the potential benefits of plant extracts and naturally occurring phytochemicals in mitigating mycotoxins consequences has been partially reviewed in the literature (324). Accordingly, herbaceous products could show rigorous antifungal activities to improve enzymatic function in degrading AFs (324) and enhancing their metabolism. Despite the effectiveness of phytochemicals in reducing the severity of AFs side effects, there have been no comprehensive clinical studies to use these herbaceous metabolites for detoxifying of AFs. However, among natural products PPs are presently taken into consideration to detoxify AFs, and their health benefits expedited scientific investigations in this respect (191, 327–336). The popularity of PPs can be attributed to their spectacular antioxidant and biological properties (337, 338). According to our literature searches, hundreds of studies have been conducted on anti-AFs properties of PPs to show how these phytochemicals might ameliorate the toxicity of AFs in the animal body (339, 340). Previous studies substantially confirmed the health promoting impacts of PPs in alleviating MetSys (341), neurodegenerative (342) and chronic diseases (343). PPs are a diverse category of plant secondary metabolites, displayed health-promoting properties, and are marketed as dietary supplements (338, 344, 345). Originally, PPs are defensive metabolites in plants secreted in response to abiotic and biotic tensions (346). It is now well established that the regular consumption of PPs and/or PPs-rich foods are linked to a healthy lifestyle (347–349). To date, the chemical structure of more than 8,000 PPs has been characterized, and the emerging evidence suggests that global demands for PPs-rich foods (or supplements) have increased over the past decades (344, 347). PPs could scavenge free toxic radicals and display potential antioxidant activity (350). Evidence-based data reported that PPs showed no toxicity effects on the human body (338, 350). These metabolites mainly function as potent anti-bacterial, antiviral, anti-diabetic, anti-cancer agents, and are modulators of molecular signaling pathways (351, 352). A plethora of studies has been conducted on PPs to modify their structure in developing reliable medicines for treatment of human diseases (353). After ingestion of PPs-rich foods, these metabolites are metabolized in liver, enter the blood circulation system and are transported to different parts of the human body (345). The chemical backbone of PPs consists of at least one aromatic ring and several OH groups. In this regard, PPs can be classified into two main groups, including flavonoids and non-flavonoid metabolites (346). Simple PPs mainly shared C1–C6 and C3–C6 backbones. These PPs have a low molecular weight and are widely biosynthesized in flowering plants. Instead, flavonoids share a C6–C3–C6 backbone, and two benzene rings are existed in their structure (354). Flavonoids are among the most abundant and well-studied PPs, displayed potential antioxidant and health-promising effects. This class of PPs divided into several subgroups with distinct biological and chemical features (355). Studies have shown that the biological activity of flavonoids is associated with the substitution of functional chemical moieties on their backbone (356). In nature, flavonoids and other PPs mainly occur in the form of O-glycosylated metabolites. These highly hydroxylated PPs are prone to other chemical modifications such as methylation and acetylation (347). The review of the literature showed that the majority of studies on PPs targeted O-glycosylated PPs, while the biological properties of C-glycosylated PPs have remained relatively unclear (357, 358). Stilbenes, lignans, curcuminoids, coumarins, and xanthones are other classes of non-flavonoid metabolites with potential biological activities (345, 347, 359). Figure 11 summarizes relevant information on PPs. The advancement of biotechnological methods supports researchers to integrate new genes, and other interested molecular components into target organisms for improving their qualitative and quantitative traits (360, 361). Genetically modified organisms have a particular gene expression profile in which genes might be silenced or over-expressed to obtain the desired traits (362–365). In this regard, ME of PBPs received much attention from academic and industrial sectors due to health benefits of PPs (366). Today, sufficient pieces of information about PBPs are available to introduce their biosynthetic genes into new hosts for overproduction of these highly valuable metabolites (360, 366). The ME of PBPs can be performed in different ways, including overexpression of PPs biosynthetic structural genes and transcription factors, up and down-regulation of genes/enzymes involved in the biosynthesis of certain PPs, silencing of specific metabolic routes in phenylpropanoid pathway, down-regulation of PPs biosynthesis structural genes using interference miRNAs, and enhancing the production of intermediary substrates of phenylpropanoid enzymes (365–369). In this regard, several lines of evidence suggest that the metabolic engineering of PBPs could enhance crop resistance against plant pathogens (369, 370). In an interesting study, ME of PBPs in soybean hairy roots influenced root resistance to fungal pathogens (370). Additionally, scientific outcomes also reported that the presence of gallic acid and hydrolysable tannins in the pellicle tissues of walnuts were accounting for prevention of AFs biosynthesis (371). Biofortification lignin biosynthesis in plants has been considered as a defensive strategy in the retardation of pathogen entrance and growth (372). For example, the overexpression of OsWRKY89 gene in GM rice plants has shown to influence resistance to rice blast by modulating the biosynthesis of PPs and increasing the lignification process of GM lines (373). The overexpression of PAL gene in GM tobacco plants could enhance their resistance against Phytophthora parasitica and Cercospora nicotinae fungal pathogens (372, 374). In this respect, studies manifested that transformation of other key regulators of phenylpropanoid pathway along with PAL gene might increase phenolic content (in particular rutin and chlorogenic acid) of GM plants compared to wild-type lines (375). These outcomes, as documented in the literature, indicated that production of PPs-rich GM plants not only improved plant resistance to fungal pathogens, but also enhanced their antioxidant properties (372). Correspondingly, engineering of flavonoids pathways in flax plants has shown to increase the accumulation of fatty acids in GM flax seeds and oil (376). This outcome has shown that the overproduction of flavonoids resulted in the prevention of lipid oxidation during seed development and maturation and increased its antioxidant properties for biomedical applications (376). Multi-level engineering of tomato plants by targeting specific transcription factors (e.g., AtMYB12) has been increased the fruit dry weight, carbon metabolisms, and improved the functionality of shikimic acid and phenylalanine pathways (377). This outcome indicated that the accumulation of phenylpropanoid bioactive metabolites not only improved the quality of engineered tomato plants, but also provided a ground for biofortification of healthy foods enriched in health promoting secondary metabolites (377). The overexpression of maize Lc gene in transgenic apples spectacularly increased the production and accumulation of anthocyanins and falvan-3-ols resulted in a higher resistance to Erwinia amylovora bacterium and Venturia inaequalis fungus (378). In another line of research, the cloning of Solanum sogarandinum 7-O-glycomethyltransferase enzyme in transgenic flax plants significantly increased the accumulation and stability of flavonoid glycosides (anthocyanidins and flavonols), leading to a higher resistance to the fungus Fusarium and improved antioxidant and oil content of GM flax (379). Further information regarding scientific investigations on ME of plants PBPs can be obtained in the available literature reviews (366, 367). Indeed, Microbial production of PPs using GM bacterial strains have also resulted in significant over-production of these health promising metabolites (360, 380) which can be used to expand clinical and experimental studies on PPs. The available evidence suggests that such biotechnological methods not only increased the quantity and availability of PPs for industrial, food and medicinal purposes (360, 381, 382), but also expedited the number of scientific investigations conducted on PPs to find the most potent metabolites for large-scale applications (361). The available literature suggests that there have been no comprehensive studies in the literature to engineer crops PBPs against APF. Therefore, developing highly PPs-rich GM plants among susceptible crops to AFs using modern biotechnological and breeding procedures might be considered as alternative mycotoxin management strategy to decrease the global spread of AFs. Although presently GM crops are not universally popular due to the rumors revolved around (383, 384), nevertheless, GMOs are part of hundreds of suggestions to eliminate AFs in food/feed. Additionally, the ME of PPs biosynthetic pathways might increase the long-term resistance of recombinant plants to invasive fungal pathogens and decrease the application of chemical fungicides, though this benefit will require further large-scale investigations to confirm its efficacy against crop fungal pathogen damages. As anticancer agents, PPs could modulate signaling pathways involved in cancer by up and down-regulation of gene expression and suppression of the release of inflammatory factors (385, 386). Experimental assays also unraveled that PPs could inhibit cancerous tumor growth by triggering apoptosis and blocking signaling pathways involved in tumor cell angiogenesis (387). The antioxidant content of PPs also support these metabolites in decreasing ROS/RNS levels in cancerous cells (388). Indeed, PPs affect the cancerous cells gene expression profile, leading to suppressing cancerous cell development and metastasis (389, 390). The growing body of evidence suggests that PPs have impacts on DNA methylation and epigenetic modifications associated with progress of cancer (390). PPs also inhibited the activity of DNA methyltransferases, leading to significant changes in the methylation pattern of specific genes involved in various types of cancers (391). PPs could also modify the expression of microRNAs involved in the regulation of cancer metastasis pathways (392). Proteins such as G protein-coupled receptors, PI3K, AKT, MMPS, EGFR, VEGF, ERK, STAT3, p53, FOX, JNK, caspases, JAK, PKC, FGF, Nrf2, ALK, ROS1, mTOR, and MAPK are pivotal derivers in the pathogenesis of human cancers (386, 392). In a dose-dependent fashion, PPs could modulate the activity of these proteins, leading to inhibition of primary phases of tumor development (386, 393). Blocking and/or modulating the secretion of cancer-induced pro-inflammatory mediators is another health-promising effect of PPs (394). Inhibition of cancerous cell proliferation (395), inducing apoptosis (387), and suppressing cell cycle events are promising strategies in the cornerstone of cancer therapies. PPs have been shown to regulate these critical processes in cancerous cells through a range of molecular mechanisms (396). Anticancer activity of PPs in clinical trials and animal studies has also substantially studied (397–400). As evidenced in the literature, as anti-inflammatory phytochemicals certain PPs (e.g., flavonoids, stilbenes, curcuminoids) could ameliorate the consequences of neuroinflammation (248). These phytochemicals principally interacted with pivotal neuro-inflammatory signaling waterfalls, improved brain enzymatic activity, decreased nitrosative stress and RNS formation, improved brain antioxidant defense, regulated of pro-inflammatory-related gene expression, restored the activity of astrocytes and microglial cells, modulated brain transcription factors expression, alleviated COXs expression, enhanced expression of anti-inflammatory genes, and protected brain neuronal cells (248, 401, 402). As anti-diabetic phytochemicals, PPs showed a broad-spectrum of biological activities to alleviate the complications of DM (403, 404). In this regard, several studies generally attributed the anti-diabetic potential of PPs to their capability to reduce blood glucose level, improve insulin sensitivity and secretion, alleviate oxidative/nitrosative stress, inhibit carbohydrate digestive enzymes, alleviate β-cells apoptosis, ameliorated lipogenesis, alleviate glucogenolysis and gluconeogenesis, up- and down-regulate of DM-associated genes, and modulate signaling pathways (NF-κB, ERK, PPAR, AMPK, cytokines, protein tyrosine phosphatases, glucose transporter receptors, hepatic enzymes, tyrosine kinases, insulin receptors) (403–406). Biochemical and metabolic factors, including concentrations of PPs, duration of PPs administration, PPs metabolism and post-metabolism modifications, interaction with intestine metabolites/enzymes, interaction with gut microbiota, gastrointestinal uptake, and their bioavailability the body tissues affect the molecular effects of PPs. The following sections summarized the recent trends on the application of PPs in prevention of AFs consequences. As documented in the literature, PPs exhibited anti-fungal activity against various types fungal pathogens (407). Respectively, Ahmed and colleagues comprehensively reviewed the inhibitory profile of PPs in preventing AFs production (407). This study, however, focused on PPs mechanism of actions to inhibit AFs formation, and highlighted the associated food safety issues (407). Indeed, another review by Fan et al. summarized recent updates on the application of phytochemicals in detoxifying AFB1-induced hepatotoxicity (21). This study also included PPs as possible candidates to ameliorate AFB1 side effects, however, the study mainly discussed anti-AFB1 mechanisms of different phytochemicals (21). There have not been similar comprehensive review studies on anti-AFs activity of PPs in the literature until the time we prepared this review. Therefore, to increase our current knowledge of anti-AFB1 properties of PPs, in this section we provided an in-depth insight into PPs/PPs-rich extracts mechanism of actions in alleviating health hazardous effects of AFs, in particular AFB1, by summarizing recent trends obtained from animal-based and in vitro studies. Due to multiscale biological activity of PPs, scientific investigations promoted these naturally occurring metabolites in preventing AFs complications (324, 408). Studies suggested that PPs can directly or indirectly affect the metabolism of AFs, leading to significant reduction of AFs toxicity (409–414). These interesting results purported that PPs substantially interfered with the formation of AFs-HSA complex (191), reduced the construction of AFs-DNA adducts (415), regulated AFs-induced inflammation (416), and also improved detoxification of AFs in liver (417). Bearing this fabulous quote “All things are poison, and nothing is without poison; the dosage alone makes it so a thing is not a poison,” credited to Paracelsus (418) in mind, it can be said that only particular doses of PPs might suppress the onset of AFs-induced metabolic and chronic disorders. Therefore, understanding the biological properties of PPs after their metabolism and tissue intake in the body can supposedly help to identify the exact behavior of these metabolites in AFs therapy scenarios. In this regard, there are hundreds of studies confirmed that PPs could alleviate the end effects of AFs-induced complications, as detailed in the following paragraphs. In rats fed with 72 μg/kg/b.w. AFB1 and 100 mg/kg/b.w. PPs-rich leaf extract of artichoke (Cynara scolymus L) (PLEA) for 42 days, the outcome manifested that the PLEA promoted partial neuroprotective properties. Accordingly, PLEA down-regulated total plasma lipids/LDL/VLDL, and simultaneously increased HDL levels (419). Although supplementation of PLEA alone had no effects on TNF-α, TIMP3 and IDO concentrations in the brain of rats, however, the results displayed that when rats supplemented with PLEA + AFB1 the total concentrations of these biomarkers relatively decreased but still was not meaningful compared to control group (419). Co-treatment of rats with PLEA +AFB1 was also ameliorated the oxidative stress in the brain of rats, and improved antioxidant enzymes. It seems that the joint administration of PLEA + AFB1 could alleviate oxidative stress in the brain and improve the histological effects of AFB1 on the brain (419). Similarly, quercetin (30 mg/kg) showed significant neuroprotective effects in Balb/c mice co-administrated with AFB1 (0.75 mg/kg/b.w.) (328). Quercetin reduced TNF-α and IL-1β levels, increased GSH, CAT, and SOD levels, and prevented memory impairment in mice exposed to chronic levels of AFB1 (328). In rats co-supplemented with 80 μg/kg/b.w. AFB1 and 300 m/kg/b.w. PPs-rich ethanolic extract of Chelidonium majus (PEEC), the outcome showed that PEEC alleviated the neurochemical biomarkers (420). In AFB1-treated group, rats showed a significant increase in TNF-α, IL-1β and CD4, AChE, dopamine, and caspase 3 levels whereas co-treatment of PEEC (not PEEC alone) significantly alleviated the increased levels of studied neurochemical markers in rats' cortex and hippocampus areas, and improved the activity of antioxidant enzymes (GSH, SOD, CAT, GPx) (420). As evidenced in the literature, gallic acid (GAc) abated the health consequences of exposure to AFs. In a study conducted by Owumi et al. (421), the evidence suggested that gallic acid exhibited potential anti-inflammatory and antioxidant activity in rats co-treated with 75 μg/kg AFB1 and 20–40 mg/kg GAc for 28 days. GAc exhibited its antioxidant activity by increasing GSH, SOD, CAT, GPx levels (421). This compound improved antioxidant balance in testes, hypothalamus, and epididymis of rats. The outcome also unraveled that GAc decreased lipid peroxidation, and oxidative/nitrosative stress (421). Correspondingly, GAc alleviated apoptosis mediators in rats by reducing IL-1β, TNF-α, nitric oxide levels, and suppressing myeloperoxidase (421). This observation proven that GAc could ameliorate AFB1-induced oxido-inflammatory responses in reproduction system (421). In Swiss male albino rats co-supplemented with 750–1,000 μg/kg/b.w./day AFB1 and 2 mg/0.2 ml olive oil/day, the outcomes showed that curcumin could ameliorate the AFB1-toxic effects in reproduction system of rats by improving caput and cauda epididymis weight and enzymatic activity (422). In a similar study, co-treatment of curcumin and AFB1 displayed that curcumin alleviated AFB1-induced toxic effects in rats' reproduction system by improving semen parameters such as sperm quality and quantity, mobility, viability and other sperm relevant features (423). In another study, the whole transcriptome analysis of BFH12 cell lines co-treated with curcumin and AFB1, the finding purported that curcumin abated the inflammatory responses and improved antioxidant enzymes in AFB1-treated cells (424). In a dose-dependent manner, phenolic metabolites such as ellagic acid improved the activity of endogenous antioxidant enzymes, prevented DNA damage and exhibited antimutagenic properties in animal models exposed to AFB1 (324). Caffeic acid in the concentration of 40 mg/kg exhibited protective effects in reproduction system of male rats exposed to 50 μg/kg/b.w. AFB1 through modulation of antioxidant enzymes, apoptosis, and inflammatory factors (425). Apigeninidin-rich extracts of Sorghum bicolor L. Moench (ASBEs) including ASBE-05/06/07 modulated inflammation and apoptosis mechanisms in kidney and liver of rats exposed to 50 μg/kg doses of AFB1 (426). ASBE-06 with IC50 = 6.5 μg/ml suppressed lung cancer cell lines. Correspondingly, ASBEs were also modulated the expression of STAT3 and caspase 3 proteins and displayed protective role against oxidative and nitrosative stress (426). In an interesting study, total flavonoids extract of Rhizoma Drynariae in the concentration of 125 mg/kg inhibited AFB1-induced apoptosis and regulated the expression of PI3K, AKT, Bax and Bcl2 in broilers chickens (427). In rats fed with 400 mg/kg/b.w. AFB1, oxidized tea phenolic compounds in the concentration of 100 μg/kg directly bound to AFB1, lowered its plasma level, and increased AFB1 fecal excretion (27). Quercetin in the concentration of 50 and 100 mg/kg/b.w. displayed protective hepatocellular effects in liver of rats received 1.4 mg/kg/b.w. AFs-containing diet (428). This finding is in agreement with previous outcomes that examined the efficacy of different doses of quercetin (15–45 mg/kg/b.w.) in prevention hepatic damage of AFs in mice (327). Quercetin also showed protective role against DNA damage when HepG2 cells treated with 5 μg/ml quercetin and 1 μM AFB1 for 2 h, leading to a significant decrease in DNA damage from nearly 60–32% (21, 333). In an interesting investigation, chicks exposed to 5 mg/kg AFB1 displayed an alteration in the activity of AST, ALT, nitric oxide synthase, COX-2, caspase1/3/11, antioxidant enzymes, and pro-inflammatory factors such as TNF-α, IL-1β/6 (429). Morin, a flavonol derivative, in a dose-dependent manner (20–80 mg/kg) ameliorated inflammatory responses, restored AFB1-induced liver and kidney damages by modulation of gene expression and prevention of hepatocyte disruption in AFB1 fed chicks (429). Kolaviron, a bioflavonoid extracted from Garcinia kola in mice administrated with 100 and 200 mg/kg of this compound and 2 mg/kg AFB1 for 4 weeks significantly reduced the AFB1-induced genotoxicity and oxidative stress (430). This bioflavonoid not only abated the total level of AFB1-DNA adducts, but also decreased the AFB1-induced GGT, AST, and ALT activity by 72, 62 and 56% (430). In adult rats treated with either 10 mg/kg/b.w. quercetin nanoparticles (QNPs) or quercetin and 80 μg/kg/b.w. AFB1, the outcomes have shown that QNPs showed significant anti-aflatoxigenic properties compared to pure quercetin (330). In this regard, QNPs (52.70 nm size) significantly reduced ROS formation, AST/ALT and alkaline phosphatase levels, improved cell viability, glutathione level, and mitochondrial function, and decreased lipid peroxidation (330). This outcomes suggested that Nano-formulation of PPs such as quercetin strengthened their hepatoprotective properties to alleviate the compilations of AFs (330). In rats administrated with subcutaneous 25 mg/kg/b.w. ternatin, a tetramethoxyflavone extracted form Egletes viscosa, and 1 mg/kg/b.w. for 72 h, the finding suggested that this bioflavonoid significantly reduced AFB1-induced AST/ALT levels, modulated MDA level, and displayed chemoprotective effects against liver injury (431). Ternatin markedly inhibited lipid peroxidation, bile-duct proliferation and hepatic necrosis as vitamin E did (431). In another study, the outcomes showed that ternatin decreased plasma liver GSH quantity, alleviated liver oxidative stress and glycemic state, but had no effects on liver regeneration (432). As discussed formerly, AFs displayed a potent binding affinity to human HSA (191) as well as BSA (193). PPs such as resveratrol could compete with AFs to bind to critical active sites of HSA by which showed an influential effect to decrease the bioavailability of AFB1 and displace the stability of AFB1-HSA complex (177). Studies demonstrated that flavonoids generally displayed a moderate binding potency to HSA, however, flavones and flavonols disclosed a higher tendency to interact with HSA (433). Chemical modification (e.g., sulfation and glycosylation) of flavonoids backbone might also be effective in lowering or increasing their binding affinity to HSA (433). In a dose-time dependent manner, green tea PPs (GTPs) modulated the AFB1 metabolism (by inhibiting of phase I and inducing of phase II metabolism), and decreased the formation of AFB1-HSA adducts (408). Additionally, GTPs in the concentration of 500 and 1,000 mg significantly increased the excretion of AFB1–mercapturic acid, the metabolized by-product of AFB1-8,9-epoxide, into urine which indicated a significant modification in the activity of GSTs (408). The outcome of this clinical trial was also displayed that there was no significant change in urine AFM1 levels, however, this finding strongly supports the protective roles of GTPs in regulating AFB1 metabolism and detoxification (408). Notably, AFB1 induced hepatocellular pyroptosis (434), and caused critical impairment of liver KCs (435). In mice treated with 1 mg/kg/b.w. AFB1 for 4 weeks, the outcomes showed that AFB1 activated NLRP3 inflammasome and inflammatory infiltration, up-regulated COX-2, enhanced the secretion of IL-1β, and activated KCs, leading to inflammatory-induced liver injury (434). The flavonoid silibinin displayed hepatoprotective activity via selective modulation of certain pathways in activated KCs isolated form rat liver (436). Accordingly, silibinin inhibited nitrosative and oxidative stress in a dose-dependent fashion (IC50 = 80 μM/L). While it has not inhibited prostaglandin E2, silibinin was effective in inhibiting of leukotriene B4 (IC50 = 15 μM/L) and 5-lipooxygenase pathway (436). Curcumin was also found to be functional in preventing hepatic pyroptosis and oxidative stress (416). In this respect, in mice given oral curcumin (100–200 mg/kg) and AFB1 (0.75 mg/kg) for 30 days, the outcome showed that this metabolite mitigated AFB1-induced inflammatory liver injury by inhibiting of NLRP3 inflammasome activation, enhancing phase II AFs detoxifying metabolism, up-regulating of Nrf2 signaling pathway, and preventing the release of pro-inflammatory IL-1β/18 (416). An increasing volume of evidence suggests that flavonoid subclasses mainly attenuated toxin-induced liver injury by regulating MAPK/NF-κB, CYPs, TLRs, c-JNK/ERK, cytokines/chemokines, Nrf2/CYP2E1, Bcl2/AKT/caspases signaling cascades, preventing oxidative stress, and enhancing antioxidant enzymes (437). AFs has shown to induce the expression of CYPs genes (438). Studies also proven that PPs prone to interaction with CYPs isoforms, which in turn can decrease the biotransformation of AFs after ingestion (439). Generally, mechanism of AFs-induced liver injury is a sophisticated process. The accumulating body of evidence suggests that the liver toxicity of AFs is mainly associated with oxidative/nitrosative stress, cellular apoptosis, mitochondrial dysfunction, lipid/protein peroxidation, construction of AFs-DNA adducts, DNA damage, induction of genomic mutation, inhibition of tumor suppressor proteins, up-regulation of gene expression, induction of inflammatory signaling pathways (21, 440). In this regard, robinetin and other polyhxdroxy flavonols inhibited microsome-assisted formation of AFB1-DNA adducts (334). Mutually, in Wistar male albino rats exposed to 2 mg/kg/b.w. AFB1 for 6 weeks, administration of 25 mg/kg/b.w. silymarin (or silibinin) could decrease lipid peroxidation and improve the activity of antioxidant enzymes of liver (up to 44–100%) and kidney (up to 82–100%) (441). It could also protect liver from the altered levels of DNA, RNA, glycogen and cholesterol by 70–100%, and hepatic GSH up to 25–37%, respectively (441). Grape seed proanthocyanidins (GSPAs) also showed protective role against AFB1-induced DNA damage. In male Swiss albino rats received 0.5–1 mg/kg AFB1 for 2 days and 100–200 mg/kg/day GSPAs, the outcome suggested GSPAs modulated the expression of Ogg1, Parp1, and p53 genes involved in DNA repair (442). PPs-rich cocoa extract has also been tested to investigate its anti-AFs properties (443). The outcomes suggested that it was not effective against AFB1 but it significantly reduced the ROS formation and increased cell viability in cells treated with AFB1 alone or mixture of AFB1 + OTA (443). On the contrary, flavonoids-rich fractions prepared from Rhus verniciflua Stokes (FRVs) displayed both in vitro and in vivo chemoprotective against AFB1-induced liver injury. FRVs remarkably decreased ROS formation and MDA level and improved cell viability in HepG2 cells (444). Correspondingly, oral administration of FRVs suppressed AFB1-increased serum level of ALT, lactate dehydrogenase and alkaline phosphatase. FRVs improved glutathione balance and superoxide dismutase activity in AFB1-adminstarted mice liver (444). Indeed, FRVs increased IgA and IgG titers in mice serum. Form this outcome, it can be concluded that FRVs increased the formation of AFB1-GSH complex and restored antioxidant defense (444). In male Wistar rats treated with 100 and 250 mg/kg ginger PPs-rich extract (GPE) and 200 μg/kg AFB1 on the basis of daily, the results showed that GPE could significantly reduce liver damage, AFB1-induced toxicity, and showed remarkable hepatoprotective properties (445). Additionally, it was observed that GPE could improve the activity of endogenous antioxidant enzymes, up-regulate Nrf2/HO-1 pathway, and reduce lipid peroxidation (445). In piglets fed with 8% PPs-rich grape seed extract (PGSE) and 320 μg/kg AFB1 for 30 days, the outcome showed that lower concentration of PGSE has a low to moderate impact on oxidative stress and inflammation in piglet spleen, suggesting that greater concentration of PGSE is required for better alleviation of AFB1 toxicity (446). The flavone aglycone diosmetin displayed anti-OTA activity in MDCK kidney cells by regulation of cellular ATP levels (447). By-products of palm oil industry such as PPs-rich palm kernel cake (PPKC) also alleviated AFB1-induced cell damage and showed hepatoprotective effects in chicken hepatocytes (448). It is believed that the molecular mechanisms underlying anti-AFB1 properties of PPKC are associated with prevention of lipid peroxidation, modulation of pro-inflammatory and apoptosis genes, and improving the activity of antioxidant enzymes (448). A study has displayed that in rats administrated with 250 mg/kg Korean red ginseng (Panax ginseng) extract (KRGE) and 150 μg/kg AFB1, KRGE could alleviate the adverse effects of AFB1-induced inflammation and hepatotoxic effects (449). Accordingly, KRGE improved serum biomarkers and antioxidant enzymes and prevented apoptosis in hepatocytes and liver inflammation (449). The recent studies have shown that P. ginseng comprised various types of PPs in which ferulic acid, rutin, chlorogenic acid, gentisic acid, p-/m-coumaric acid, catechin, and kaempferol were the foremost domineering phenolic metabolites in different tissues of this plant (450, 451). Grapefruit juice extract in the concentration of 100 mg/kg has displayed protective effects against AFB1-induced liver DNA damage in F344 rats treated with 5 mg/kg AFB1 by gavage (452). Correspondingly, the administrated extract remarkably reduced hepatic CYP3A content but had no effects on CYP1A and CYP2C quantities (452). Studies confirmed that flavonoids (in particular naringin a flavanone-7-O-glycoside), vitamin C and other organic acids are major metabolic components of this extract (453), accounting for its antioxidant and biological properties (454). In rats treated with different concentration of olive cake PPs-rich extracts (OCPEs) (0.2–0.5 ml), its nanoparticle-based formulation and 22 μg/kg AFB1 for 4 weeks, OCPEs and its nano-formulation improved the neurotoxicity of AFB1 in rats brain (455). In male Wistar rats given 20 g/kg basal diet bee pollen (BP) and 3 mg/kg basal diet AFs for 30 days, the BP could ameliorate the toxicity of AFs by increasing the proliferation of lymphocytes (456). Although this observed benefit of BP was attributed to minerals, vitamins, and amino acids, however, PPs are also accounting for 1.6% total BP metabolites (456), which in turn might affect the anti-AFs properties of BP. The evidence also suggests that honeybee propolis, a resinous mixture of phytochemicals such as flavonoids and non-flavonoid metabolites, vitamins B/C/E, amino acids and other aromatic metabolites (457) could alleviate the toxicity of AFB1 by improving the activity of cytochrome P450 in honeybees (458). In male rats received 50 mg/kg/b.w. Iraqi propolis and 0.025 mg/kg/b.w. AFB1, the propolis has found to be effective in the restoring AFB1-induced gastrointestinal damages (459). In mice orally administrated with different concentration of AFB1 and 2% aqueous black tea extract (ABTE) (instead of drinking water) for 30 days, the outcomes showed that ABTE ameliorated the AFB1-induced lipid peroxidation in mice liver by increasing the activity of enzymatic and non-enzymatic antioxidant contents (460). The observed benefit is accounted for the fact that ABTE is a PPs-rich fraction which exhibited significant antioxidant activities (460). Similarly, in mice co-administrated with low/high doses of ABTE and AFB1, supplementation of AFB1 resulted in significant reduction of DNA, RNA, glycogen and protein contents, and increased phospholipase activity and cholesterol content (461). In this regard, the ABTE co-administration displayed a protective role in mice liver against AFB1-induced biochemical changes (461). The co-supplementation of 2% ABTE, 200 μg/kg/b.w. curcumin in rats given 750 μg/kg/b.w. for 90 days, the outcomes confirmed that the co-administration of ABTE-curcumin displayed synergistic effects in alleviating AFB1-induced liver damages in rats (462). Correspondingly, ABTE-curcumin could improve liver architecture, activity of antioxidant enzymes, lipid profile (in particular lowering cholesterol content) and liver biomarkers (462). In rabbits treated with 5 g/kg/b.w. coumarin and 0.25 mg/kg/b.w. AFB1, coumarin improved body weight and carcass gains, and reduced the toxicity of AFB1-induced complications (463). As detailed, PPs could exhibit their anti-AFs activity in a concentration-dependent manner, thus optimization of PPs concentration for treatment of AFs therapies is the most pivotal step in experimental assays regarding this field. These outcomes together confirmed that these plant metabolites are promising substances to reduce the health consequences of AFs. Table 1 provides detailed information on the mechanism of action of anti-AFs properties of studied PPs. As detailed in Table 1, PPs ameliorate AFs toxicity in different ways. Accordingly, the anti-AFs activity of PPs mainly contributed to preventing oxidative stress and inflammatory responses, inhibiting mutations in DNA, regulating signaling cascades, modulating phase I and II metabolism enzymes, improving cellular antioxidant balance, and interfering interaction of AFs and HSA. These data showed that flavonoids, in particular oxidized tea PPs, were the most studied PPs in the prevention of AFs toxicity. Indeed, the anti-AFs activities of resveratrol and curcumin were also highly investigated. Interestingly, different classes of PPs exhibited a wide range of heterogenous biological properties against toxicity of AFs. Our review clearly disclosed that PPs targeted the core pathways (inflammation-based responses) in the pathogenesis of AFs. This interaction is important because not only in the onset of cancer, but also inflammation (in particular NIF) play a critical role in in the progression of neurodegenerative disorders and MetSys (248, 250, 256). Therefore, as modulator secondary metabolites, PPs have the potential to maintain the normal status of cell by regulating the onset of inflammation-assisted signaling pathways and preventing the development of nCDs (338). PPs also prevent the formation of AFs in target fungi (A. flavus and A. parasiticus strains) by modulating fungal transcription factors activity (513, 514). For example, water-soluble and methanol extract of peanut tannins were also inhibited A. parasiticus growth and impaired the formation of AFs in a dose-dependent manner (515). Studies have also shown that the type of PPs extraction methods might affect antifungal activity of these metabolites. In this regard, solid-phase extraction of PPs-rich citrus peel extract displayed up to 40% antifungal properties against A. flavus compared to crude extracts (516). Correspondingly, 300–400 mg/ml mandarin PPs-rich extract is enough to completely inhibit A. flavus growth depending on extraction method and applied solvents (516). PPs-rich methanolic extract of Zanthoxylum bungeanum (a traditional Chinese food additive) with the IC50 2–4 μg/ml significantly repressed the AFB1 biosynthetic pathway (517). The omics-based analysis of this extract unraveled that it could show the anti-aflatoxigenic properties by down-regulating of the global regulators of AFB1 biosynthesis such as velvet complex proteins, Medusa and brlA genes, and GPCR/oxylipin-based signaling cascade (517). Intriguingly, PPs-rich olive processing wastes (POPWs) also showed differential anti-aflatoxigenic properties in a dose-dependent manner to inhibit the growth of APF (518). Of all POPWs, olive pomace extract displayed a higher anti-aflatoxigenic properties in comparison to olive leave and pomace olive oil extracts, respectively (518). 5′-hydroxy-auraptene, a coumarin derivative isolated from Lotus lalambensis, in the concentration of 40 μg/ml prevented the AFB1 production and exhibited potential antifungal activity against A. flavus by down-regulation of genes involved in different phases of AFB1 biosynthetic pathway and inhibition of conidial germination of this fungus by 60% (519). The outcomes also suggested that 5′-hydroxy-auraptene disrupted the structure of mycelia sugar units and up-regulated stress mediated transcription factors (atfA and atfB) up to 2 and 2.5 folds (519). Buckwheat hull PPs-rich extracts (BHPEs) displayed a dose-dependent inhibitory profile against A. flavus growth and AFB1 biosynthesis (520). According to this outcome the higher concentration of BHPE was accounting for longer inhibition of AFB1 formation (520). In another study, the pau ferro (Libidibia ferrea), a Brazilian medicinal plant, ethanolic fruit extract has found to be effective in the inhibition of A. parasiticus growth (521). PPs-rich Cistus incanus L. methanolic extract in the concentration of 0.2 g/ml significantly decreased the formation of AFB1 production from 72.5 to 90.1% and inhibited A. parasiticus growth (522). The essential oil of C. ladanifer, another species of Cistus genus, has also found to be effective in the prevention of A. flavus growth by suppressing AFB1 production, and inhibiting the fungus ergosterol biosynthetic pathway with MIC value of 0.6 μl/ml, respectively (523). An increasing trend of evidence purported that glycosidic and aglycone derivatives of flavonoids and non-flavonoids in a dose-dependent manner suppressed the production of AFB1 and other mycotoxins by targeting the critical routes in biosynthetic pathways of toxicogenic fungi (407). For instance, compound of interest, quercetin also disrupted AFs-producing fungal proliferation in addition to blocking the formation of AFs (28). Similarly, Green tea PPs in the concentration of 70 mg/ml suppressed the formation of AFs without side effects on the mycelial growth of APF (524). Phenolics also showed potential inhibitory profile against the production of other mycotoxins. In an interesting study, Boonmee and colleagues reported that simple hydroxycinnamic acid derivative, ferulic acid inhibited the production of OTA in A. westerdijkiae and P. verrucosum by 35 and 75% (525). Synthetic derivatives of flavonoids such as 5,6-dihydroxy-flavone and 5,6-dihydroxy-7-methoxy-flavone, in the concentration of 25 and 50 μg/ml, have significantly decreased the production of OTA in A. carbonarius after 8-day incubation (526). Similar to these outcomes, Romero et al. reported that caffeic acid, rutin and quercetin in the concentration of 250 mg/L remarkably decreased the production of OTA in A. carbonarius (527). Indeed, higher concentrations of these PPs (500 mg/L) completely inhibited the growth of OTA-producing fugus (527). The aqueous seed extract of Trachyspermum ammi also showed beneficial effects in degrading AFB1 mycotoxin. The phytochemical analyses revealed that T. ammi has different types of metabolites such as PPs, alkaloids, tannins, and other well-known natural substances (528). Olive mill wastewater (OMWW) pure PPs such as caffeic acid, hydroxytyrosol, tyrosol and verbascoside also showed a decrease of nearly 99% in AFB1 production but had not influenced A. flavus growth (512). Accordingly, OMWW extract in the concentration of 15% was also decreased the formation of AFB1 ranged from 88 to 100%, respectively (512). In a dose-dependent manner, flavonoids-rich spent coffee grounds extract (PSCGE) has also found to be effective in degrading AFs (B1/2-G1/2) and OTA in vitro (529). The PSCGE was remarkably decreased the growth of toxicogenic fungi such as A. flavus and A. ochraceus as well as Fusarium species (529). These outcomes demonstrated that food wastes/residues have considerable level of health promoting metabolites; alternatively, can be used as potent inhibitors of APF and the production of AFB1 and other mycotoxins (529). Therefore, it should be noted that these waste by-products are trustworthy candidates to develop and formulate modified extracts with added values as anti-fungal agents. Compared to PPs-rich extracts, alkaloids-rich extracts (ALEs) showed potential inhibitory properties in detoxifying of AFB1. In this respect, vasaka (Adhatoda vasica Nees) leaf ALEs displayed functionality to degrade AFB1 up to ≥98% after 24 h incubation at 37°C (530). The amide alkaloid piperlongumine isolated from Piper longum L. in the concentration of 0.2%w/v inhibited the biosynthesis of AFB1 in A. flavus up to 96% (531). Another piperidine alkaloid, piperoctadecalidine isolated from P. longum displayed 100% inhibitory profile against biosynthesis of AFB1 in the concentration of 0.7%, respectively (531). Other relevant studies also reported that ALEs are promising anti-fungal agents to prevent the formation of AFB1 and APF growth (532). It seems that the anti-aflatoxigenic property of alkaloids is dose-dependent, and the chemical variation of alkaloids might determine their inhibitory profile. In this regard, it can be said that both alkaloids/PPs-rich extracts and pure metabolites are influential compounds in the prevention of AFs; however, their chemistry and concentration are two determinant factors in defining their inhibitory/biological profile. In modern food industries, however, these metabolites are promising candidates to develop antifungal agents. These results together permit to harness phytochemicals for dealing with health hazardous mycotoxins. Chemical fungicides (ChFs) are presently recruited for large-scale inhibition of AFs production and APF growth (533). However, as detailed in the literature, long-term application of ChFs may lead to ChFs-resistant APF (533) and the onset of health threatening symptoms (534). In this regard, there are only few registered patents publicly available to use PPs as candidate inhibitors of AFs biosynthesis. These innovations developed specific products or GM plants to prevent the spread of AFs-contaminated foods/crops. In an interesting patent, engineered transgenic plants with elevated levels polyphenol oxidase/gallic acid content showed resistance to A. flavus and AFs production (535). In another patent issued by Xiang et al., tea PPs in the concentrations of 0.2–1% were effective in the reduction of AFs biosynthesis by 21–81% (536). Dacheng et al. developed an anti-mycotoxigenic feed additive for cattle in which tea PPs (1–5 parts of whole patent formulation) have been used as main ingredients of this product (537). In another patent, a preparation method for constructing catechin nanoparticles has been suggested that could decrease the bioavailability of AFB1 and prevent hepatic injury (538). Despite the scarcity of patents on PPs for inhibition of AFs production, however, in the light of discussed materials herein further formulation of PPs can be crafted to develop potent anti-AFs products. Increasing knowledge on the antioxidant content of natural products and vitamins leads to developing “antioxidant therapy” to relieve human diseases (539), though there have been doubts in this field (540, 541). Studies have shown that the combination of PPs and vitamins enhances their antioxidant activity (539). It is now proven that cancer, MetSys and neurodegenerative diseases are associated with a higher level of oxidative/nitrosative stress and ROS/RNS production (30, 542–544). In this regard, natural antioxidants are primary defensive agents in preventing early phases of cancer, AD and DM progression (30, 388, 545). Accordingly, PPs received a huge volume of attention due to their antioxidant properties and becoming the relevant metabolites in antioxidant therapy programs (546). Presently a considerable number of PPs in the form of antioxidant supplements, cosmetic and food additive products, sanitary agents, antibacterial products, and pain relievers are available in global markets for non-clinical applications (338). Regarding PPs applications in alleviating AFs-induced health challenges, there are several key points should be highlighted before consideration of these metabolites for large-scale studies. First, despite considerable studies conducted on PPs, the optimum doses of phenolic compounds for amelioration of AFs side effects have not yet determined. This case causes a significant variability in observed biological effects assigned to PPs. For example, studies have shown that the anticancer activity of PPs is dose-dependent, in which PPs shared different IC50 and Ki values for inhibition of target enzymes (547). On the other hand, the intake of high concentration of PPs might show adverse effects on kidney and thyroid hormones level, as reported in animal models (548, 549) and argued in the literature (540). Second, standardization of PPs mode of action toward cellular receptors requires sufficient data generated from clinical trials and large-scale studies (546). For instance, the anticancer activity of PPs depends on their chemical structure, doses and subtypes of cancers (399). Third, as detailed in Table 1, anti-AFs properties of PPs are not generally focused on specific pathways, though the majority of studies confirmed that these metabolites function through suppressing of AFs-induced oxidative stress. This demonstrated that PPs might show off-target effects to interact with several different receptors in the body. Fourth, the pro-oxidant activity of PPs is another concern (547, 550) might delay the recruitment of these metabolites against AFs. On the other hand, some phenolic compounds such as daidzein and genistein exhibited controversial effects on the pathogenesis of hormone-associated cancers (399). Such incongruous results have still not been reported for other phenolic categories, and it is now believed that PPs displayed their health-promising effects by interacting with various receptors and signaling pathways (551). In this respect, the emerging scientific investigations reported that supplementation of antioxidants (vitamin E and N-acetylcysteine) in mice increased the progression of lung cancer by inducing P53-assisted oxidative stress (198). This finding suggests that excessive supplementation of antioxidants might increase the progression of nCDs, therefore, to avoid further complications, and for cautionary reasons, the consumption of antioxidants should be followed by considering optimum doses under strictly controlled condition. Therefore, recruiting PPs or PPs-rich extracts in anti-AFs therapies requires a deep insight into their bioavailability, on-target mode of action, pharmacokinetics, drug interventions, and chemical stability in the human body (188, 547). Another worthy point that should be addressed that is PPs-rich waste products (e.g., OMWW) disclosed potential inhibitory activity against AFB1 formation (512). Such products have a certain group of PPs, called non-extractable PPs (NEPPs) (552, 553), in turn, their biological activities have not extensively been investigated due to limitations in extraction methods or insolubility and polymeric essence of these metabolites (554). Studies have shown that the optimization of PPs extraction methods using innovative technologies reduced required time and energy to elucidate PPs, and improved the quantity of achievable phenolics metabolites (555). These innovations help researcher to access the whole PPs profile of herbaceous materials to determine their biological activities. In this regard, we previously reviewed the valorization methods of OMWW PPs extraction to highlight the biological benefits of these phenolics that widely released into the nearby environment (556). Similar studies also suggested that such NEPPs from waste products have a remarkable antioxidant content, resultantly may exhibit health promoting effects in the cornerstone of the human diseases prevention (557). On the other hand, the current knowledge of PPs is mainly associated with extractable O-glycosidic derivatives. The evidence purported that C-glycosidic PPs such as schaftoside derivatives exhibited potential biological activity to enhance crop resistance against pests (558). Therefore, such metabolites can also be used to inhibit the growth of APF or prevent the production of AFs. Additionally, the current findings on anti-AFs activity of PPs obtained from animal models (rodents and chicken) and experimental assays, in turn, requiring further validations. Future studies on how PPs might interact with nCDs-associated molecular receptors and signaling pathways (in particular inflammation) and their exact mode of action in relieving AFs end effects should be conducted to confirm their efficacy and safety for medical management of mycotoxins. Figure 12 summarized the current findings of this study. Our literature review clearly manifested that shifts in world climate can influence the distribution/quantity of AFs and prevalence of nCDs. Additionally, the triangle of GCC/exposure to AFs/progression of nCDs has significant complications for human health, in turn, can increase the economic costs to countries health care system. It is important to conduct more large-scale and long-term investigations in susceptible countries to GCC (in particular low and middle-income regions) to predict future threats for designing effective preventative policies and health risk assessments. Various types of food/feed items have been identified with exceeded level of AFs contaminants, resultantly this phenomenon might bring countries more deaths during the upcoming years owing to the onset of nCDs complications. The co-occurrence of other GCC-associated risk factors may expedite the progression of nCDs, though this claim requires further approvement to know how synergistic effects of environmental health hazardous risks will threaten human health and lifestyle. Countries implemented rigorous regulatory measurements to monitor and diagnosis of AFs in contaminated foods/feeds. However, the literature indicates that the current regulatory settings should immediately be revisited due to the advent of AFs and other emerging mycotoxins in various food commodities (559, 560). AFs decreased the quality and marketability of food/feed products, therefore causing damages to countries food safety and economy (561). In this regard, increasing public awareness, in particular among farmers and local food/feed producers, plays a critical role in elimination of AFs and preparing countries local society to deal with complications of GCC and health threatening environmental risk factors. Owing to the role of AFs in triggering immunosuppression (37), interfering with protein metabolism and micronutrient deficiency (276), and reducing antibody production (562), specific diagnostic/predictive biomarkers should be characterized in early detection of AFs-induced health challenges to a better management of clinical symptoms. On the contrary, PPs promoted initiate immune responses by activating certain signaling pathways (563). This indicates that dietetic intervention using PPs is an effective way to modulate immune responses to alleviate the toxicity of AFs, though current PPs gaps for clinical applications should be addressed in detail. In neurodegenerative diseases, the accumulation of aggregated proteins leads to the progression of these disorders (30, 564). PPs have also shown regulatory effects in the modulation of protein metabolism and activation of protein degrading systems to prevent the accumulation of misfolded proteins (564). Having such biological properties enabled PPs to combat AFs health problems. As discussed, no antidote has been introduced to alleviate the toxicity of AFs and current management scenarios of AFs complications are based on the removal of these mycotoxins in foods. Indeed, the frequency of international studies on AFs has spectacularly surged up in the past years in which the USA, China, India occupied the top ranks of studies in this field. Bibliometric analysis of the literature displayed that the majority of studies conducted on AFs were focused on carcinogenic, properties and detection methods of these toxins, though association of exposure to AFs and the onset of DM and AD has been remarkably taken into consideration. This indicates that long-term exposure to AFs evoked multidimensional health challenges in addition to their potential in inducing HCC. The literature reviewed herein suggested that the quest for characterization of natural inhibitors of AFs is becoming a global trend in food safety field. According to our literature review, phenolics and PPs-rich extracts are promising AFs detoxifying products. Green and black tea, turmeric, anthocyanins and flavonols were the most studied phenolics metabolites to detoxify AFs complications in animal and in vitro studies. Despite the lack of enough clinical data on the effectiveness of PPs in preventing AFs consequences, the available data displayed that PPs showed a heterogenous biological activities in preventing the side effects of AFs by targeting several different molecular receptors. PPs reviewed in this paper can be used for decontamination of AFs (and possibly other emerging mycotoxins) either in the human/animal body or production/storage sites after considering safety cautions. As discussed, due to the negative effects of nCDs and GCC on world economy, future studies should seek to develop strategies that may improve the bioavailability, mode of action, and pharmacokinetics properties of PPs in the cornerstone of AFs-induced nCDs treatment. Coupling antioxidant-assisted interventions with conventional physio-chemical removal procedures of AFs in food/feed items and developing GM PPs-rich crops are highly recommended to decrease the quantity of AFs to lowest concentrations and improve lifestyle and longevity of affected individuals. It is also treasured to address this point that regular consumption of PPs or PPs-rich functional foods might help in early preventing of AFs-induced nCDs, though this finding requires further clinical assessments. HR designed the study, wrote the first draft, and prepared the graphical illustrations. HR and FN conducted a literature search. RK revised the draft. HR, FN, and RK revised the final draft. All authors contributed to the article and approved the submitted version. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649867

María Belén Fernández,Lucas Latorre,Natalia Correa-Aragunde,Raúl Cassia

A putative bifunctional CPD/ (6-4) photolyase from the cyanobacteria Synechococcus sp. PCC 7335 is encoded by a UV-B inducible operon: New insights into the evolution of photolyases

28-10-2022

cryptochrome/photolyase family,Synechococcus sp. PCC 7335,UV-B,photolyase operon,cyanobacteria,bifunctional CPD/(6-4) photolyase- like

Photosynthetic organisms are continuously exposed to solar ultraviolet radiation-B (UV-B) because of their autotrophic lifestyle. UV-B provokes DNA damage, such as cyclobutane pyrimidine dimers (CPD) or pyrimidine (6-4) pyrimidone photoproducts (6-4 PPs). The cryptochrome/photolyase family (CPF) comprises flavoproteins that can bind damaged or undamaged DNA. Photolyases (PHRs) are enzymes that repair either CPDs or 6-4 PPs. A natural bifunctional CPD/(6-4)- PHR (PhrSph98) was recently isolated from the UV-resistant bacteria Sphingomonas sp. UV9. In this work, phylogenetic studies of bifunctional CPD/(6-4)- photolyases and their evolutionary relationship with other CPF members were performed. Amino acids involved in electron transfer and binding to FAD cofactor and DNA lesions were conserved in proteins from proteobacteria, planctomycete, bacteroidete, acidobacteria and cyanobacteria clades. Genome analysis revealed that the cyanobacteria Synechococcus sp. PCC 7335 encodes a two-gene assembly operon coding for a PHR and a bifunctional CPD/(6-4) PHR- like. Operon structure was validated by RT-qPCR analysis and the polycistronic transcript accumulated after 15 min of UV-B irradiation. Conservation of structure and evolution is discussed. This study provides evidence for a UV-B inducible PHR operon that encodes a CPD/(6-4)- photolyase homolog with a putative bifunctional role in the repair of CPDs and 6-4 PPs damages in oxygenic photosynthetic organisms.

A putative bifunctional CPD/ (6-4) photolyase from the cyanobacteria Synechococcus sp. PCC 7335 is encoded by a UV-B inducible operon: New insights into the evolution of photolyases Photosynthetic organisms are continuously exposed to solar ultraviolet radiation-B (UV-B) because of their autotrophic lifestyle. UV-B provokes DNA damage, such as cyclobutane pyrimidine dimers (CPD) or pyrimidine (6-4) pyrimidone photoproducts (6-4 PPs). The cryptochrome/photolyase family (CPF) comprises flavoproteins that can bind damaged or undamaged DNA. Photolyases (PHRs) are enzymes that repair either CPDs or 6-4 PPs. A natural bifunctional CPD/(6-4)- PHR (PhrSph98) was recently isolated from the UV-resistant bacteria Sphingomonas sp. UV9. In this work, phylogenetic studies of bifunctional CPD/(6-4)- photolyases and their evolutionary relationship with other CPF members were performed. Amino acids involved in electron transfer and binding to FAD cofactor and DNA lesions were conserved in proteins from proteobacteria, planctomycete, bacteroidete, acidobacteria and cyanobacteria clades. Genome analysis revealed that the cyanobacteria Synechococcus sp. PCC 7335 encodes a two-gene assembly operon coding for a PHR and a bifunctional CPD/(6-4) PHR- like. Operon structure was validated by RT-qPCR analysis and the polycistronic transcript accumulated after 15 min of UV-B irradiation. Conservation of structure and evolution is discussed. This study provides evidence for a UV-B inducible PHR operon that encodes a CPD/(6-4)- photolyase homolog with a putative bifunctional role in the repair of CPDs and 6-4 PPs damages in oxygenic photosynthetic organisms. Ultraviolet- B (UV-B) is the solar electromagnetic radiation with wavelengths between 280–315 nm. Although most of this radiation is absorbed by the stratospheric ozone layer it affects all living organisms. Around 0.3% of sunlight energy at sea level corresponds to UV-B and it can damage aquatic organisms triggering a decrease of ecosystem productivity (Häder et al., 2011; Banaś et al., 2020). UV-B is perceived by the UV-B response locus 8 (UVR8) photoreceptor present in photosynthetic organisms ranging from green algae to higher plants (Fernández et al., 2016). Although no defined photoreception systems have been described for other photosynthetic microorganisms (red and brown algae and cyanobacteria), they have developed several mechanisms of protection against UV-B damage (Montgomery, 2007; Sinha and Häder, 2008; Rastogi et al., 2014). Cyanobacteria are an ancient group of Gram-negative bacteria, and the first prokaryotes that perform oxygenic photosynthesis. Cyanobacteria are ubiquitous and occupy diverse ecological niches, adapting to various extreme environments, such as high or low temperatures, highly acidic or basic pH, high salt concentrations, desiccation and UV-B (Wada et al., 2013). As photosynthetic organisms, they depend on solar energy and have to cope with harmful UV-B. Solar UV-B affects the DNA and protein structures, photosynthesis, ribulose 1,5-bisphosphate carboxylase/oxygenase (RuBisCO) activity, N2 fixation, cellular morphology, growth, survival, pigmentation and buoyancy (Tyagi et al., 1992; Rastogi et al., 2014; Kumar et al., 2020; Vega et al., 2020). Thus, cyanobacteria developed several UV protection mechanisms that include UV absorbing and screening compounds such as scytonemin and mycosporine-like amino acids (MAAs), antioxidant protection, apoptosis, migration, mat formation and DNA reparation to recover UV induced DNA damages (Sinha and Häder, 2008; Moon et al., 2012; Rastogi et al., 2014). Solar UV radiation can induce two types of pyrimidine dimers in the double helix DNA, being the predominant cyclobutane pyrimidine dimers (75%; CPDs) and to a lesser extent pyrimidine (6-4) pyrimidone photoproducts (6-4 PPs) (Banaś et al., 2020; Vechtomova et al., 2020). Photoreactivation is a blue/ UV-A light- dependent mechanism used to specifically repair CPD or 6-4 PPs damages by photolyases (PHRs) (Pathak et al., 2019; Soumila Mondal et al., 2022). These enzymes are a class of flavoproteins found in all organisms excluding placental mammals, who lost all genes encoding functional photolyases in the course of evolution (Banaś et al., 2020; Xu et al., 2021). During repair, flavin adenine dinucleotide (FAD) cofactor is fully reduced via Trp or Tyr surface transfer of electrons in a process named photoreduction (Graf et al., 2015). Then, a rapid electron transfer from the excited fully reduced FAD chromophore to the DNA lesion triggers both kinds of repair. The 6-4 PPs repair also requires proton transfer which is a limiting step resulting in much lower reaction efficiency (Banaś et al., 2020). A bifunctional photolyase called as PhrSph98 was recently characterized in the Antarctic bacterium Sphingomonas sp. UV9. This enzyme is able to repair both types of damage since it displays a larger catalytic pocket compared to CPDs or 6-4 PPs repairing enzymes (Marizcurrena et al., 2020). The cryptochrome/photolyase family (CPF) comprises proteins that have conserved the FAD binding site and switch between basal and excited states. CPF proteins bind either damaged or undamaged DNA and are classified in two groups according to their function: (1) CPD and 6-4 PPs photolyases, which repair CPD or 6-4 PPs damages, respectively, and (2) Cryptochromes (CRY) that regulate growth and development in plants and the circadian clock both in plants and animals (Mei and Dvornyk, 2015; Vechtomova et al., 2020). CPD PHRs are classified, based on sequence similarity in (i) class 0, which repair CPD damages in single-stranded DNA (ssDNA PHR, previously classified as CRY with Cry-DASH designation), (ii) class I, present mostly in unicellular organisms; (iii) class II, in unicellular and multicellular organisms; and (iv) class III found only in some eubacteria (Ozturk, 2017; Zhang et al., 2017). Recently, an exhaustive phylogenetic analysis identifies a new class of CPD repair enzymes, called short photolyase-like (SPL), because they lack the N-terminal α/β domain of normal photolyases. They are similar to class I/III CPD PHRs and authors speculated that this class constitutes the real ancestor of the CPF (Xu et al., 2021). CPD PHRs III may be considered as an intermediate form between CPD PHR I and plant CRYs, as well as a protein that has retained the traits of their common ancestor (Vechtomova et al., 2020). The iron–sulfur cluster containing bacterial cryptochromes and 6-4 PPs repair photolyases (FeS –BCPs) are newly characterized CPF members containing the amino acids necessary to bind cofactors, and four conserved cysteine residues for the coordination of an iron -sulfur cluster, an ancient feature (Scheerer et al., 2015). Some CPFs have dual roles, like ssDNA PHR proteins. They are restricted to UV lesions and repair CPD damages in single-stranded DNA, having also blue light photoreceptor activity. These proteins may represent the link between photolyases and cryptochromes (Kiontke et al., 2020). It has been proposed that the bifunctional CPD/(6-4)- PHR from Sphingomonas sp. UV9, PhrSph98, is a sister group of CPD class II PHRs and that may represent a missing link in the transition from 6-4 PP to CPD PHRs (Marizcurrena et al., 2020). It is widely accepted that photolyases are ancient DNA repair enzymes, which have evolved far before the oxygen accumulation and the ozone layer establishment in the atmosphere (Vechtomova et al., 2020). However, the evolutionary scenario was not fully elucidated yet. Prokaryotic 6-4 photolyases were suggested as the first common ancestor of photolyases. However, as CPDs are the major UV-induced DNA damage it is not convincing that the 6-4 photo repair occurred earlier than the CPD one during evolution (Xu et al., 2021). Exhaustive phylogenetic trees do not include the recently described bifunctional CPD/(6-4)- photolyase PhrSph98, and its occurrence among other species was not explored. Thus, the aim of this work was (1) to analyze the distribution of CPD/(6-4) photolyase- like proteins and its evolutionary relationship with other CPF family members; and (2) characterize the operon structure that encodes a bifunctional CPD/(6-4) PHR- like gene from the cyanobacteria Synechococcus sp. PCC 7335. We show that PhrSph98 homologs are present both in heterotrophic and phototrophic bacteria and that cyanobacteria Synechococcus sp. PCC 7335 encodes this homolog, in a two-gene assembly operon, which is induced by UV-B light. Evolutionary and structural aspects of this operon are discussed. A protein–protein BLAST (BlastP) analysis was performed against the non- redundant protein sequence (nr) database using as template the amino acid sequence of PhrSph98 from Sphingomonas sp. UV9 (accession number: NCBI ANW48627), and restricting the maximum target sequences to 250 to increase taxon- specific occurrence. All the retrieved sequences match the criteria of E- values lower than 0.001 (being lower than 1×10−63) and percentage identity higher than 30% according to Pearson (2013). Subsequently, partial sequences and those sharing >85% of identity (redundant proteins) were identified and discarded using CD-HIT software (Huang et al., 2010). Multiple sequence alignment and curation was performed using MAFFT and BMGE 1.12_1 (Criscuolo and Gribaldo, 2010) software, respectively, using default settings. Subsequently, the phylogenetic tree was inferred using a total length of the alignment of 259 aminoacids with PhyML 3.0 using the automatic model selection AIC, which defined LG5 + G + I + F as the best model of evolution, and 1,000 bootstrap. The tree was visualized using iTOL (Letunic and Bork, 2021). A sequence alignment logo for the proteins described as homologs of PhrSph98 was created using the online available WebLogo tool (Crooks et al., 2004). Conserved domains were identified using the sequences from each target protein employing the NCBI conserved domain database (CDD)-NIH (Lu et al., 2020). Operon prediction was performed using the webservers Operon mapper (Taboada et al., 2018) and SoftBerry restricting the search to bacterial genomes. Operon promoter and potential transcription factors binding sequences were predicted using BPROM software from SoftBerry platform. Transcriptional terminator sequences were predicted using iTerm-PseKNC/predictor (Feng et al., 2019). The linear amino acid sequence of PhrSph98 was used as template in BlastP against the nr protein database restricting the search to Synechococcus sp. PCC 7335. Tertiary structure of PhrSph98 and Synechococcus sp. PCC 7335 homolog were predicted using Modeller in the HHPred server (Söding et al., 2005). The structure of class II CPD-photolyase from Methanosarcina mazei (PDB ID: 2XRY) was the most similar to both queries, and thus selected as template. The 3D model structures were validated using ProSA web software (Wiederstein and Sippl, 2007). Visualization, structure overlaps and distance calculations were done using the Molecular Graphics System PyMOL. The cyanobacterial strain Synechococcus sp. PCC 7335 used in this study was acquired from the Pasteur Culture Collection of Cyanobacteria. Synechococcus sp. PCC 7335 was originally isolated from a snail shell in an intertidal zone near Puerto Peñasco, Mexico (Rippka et al., 1979). PHR genes were identified in the genome of Synechococcus sp. PCC 7335 (NCBI accession number NZ_DS989904.1). Cultures were grown in Erlenmeyer flasks in volumes of 150 ml containing ASNIII marine medium at 25°C and photon flux density of 5 μmol m−2 s−1 under a photoperiod of 12 h light (4,500 K LED tubes): 12 h dark. The optical density at 750 nm (OD750 nm) was measured with GeneQuant 1,300 spectrophotometer to monitor cell growth. Synechococcus sp. PCC 7335 cells were grown in at least three replicate cultures to an OD750 nm ~0.2–0.4. Aliquots of cells (~35 ml) were collected from each replicate culture, placed in Petri dishes and exposed to 3.34 μmol m−2 s−1 of UV-B supplemented with ~5 μmol m−2 s−1 of PAR for 15 and 30 min in a controlled environment chamber. The UV-B irradiance intensities were chosen according to He et al. (2021). Control samples were covered with a polycarbonate filter of 1 mm to screen UV-B. UV-B was provided by narrowband UV-B Philips TL 100 W/01 lamps. The spectral irradiance was determined with an UV- B photo-radiometer (Delta ohm HD2102.1). After treatment, cells were immediately pelleted by centrifugation at 11,000× g and 4°C and stored at −80°C until being processed for total RNA extraction. RNA from Synechococcus sp. PCC 7335 was extracted with the RNeasy mini Kit (Qiagen) following the manufacturer’s protocol. The concentration and purity of the isolated RNA were measured with a UV spectrophotometer (NanoDrop™ One, Thermo Scientific). Total RNA was used for cDNA synthesis in a reaction containing 3 μM random primer, 10 mM dNTP, 0.1 M DTT, and 200 U of M-MLV reverse transcriptase (Invitrogen). RT-qPCR was performed at a 10 μl total reaction volume including 1 μl of a dilution of cDNA template, 250 nM of forward and reverse primers, and 5 μl of 2× Power SYBR Green PCR Mix (Applied Biosystem). Amplified signals were monitored continuously with a Step One Plus Real-Time Thermal Cycler (Applied Biosystem). Thermocycling was performed using the following conditions: 10 min of denaturation and enzyme activation at 95°C, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min. Melting curve analysis was performed from 60 to 95°C at 0.3°C increments to verify primer specificity. Negative controls (no template and minus RT transcriptase) were included in every PCR run to verify no genomic DNA contamination. A default threshold of 1 was used. The size of qPCR products for each primer pair was verified on 2% (w/v) agarose gel electrophoresis using 1× TAE buffer. Ladder 100 pb was used as molecular marker (PB-L productos Bio- lógicos). Real- time data was analyzed using the StepOne™ Software v2.3 Tool (Applied Biosystem). The primers used are listed in Supplementary Table S1. RNase P RNA (RNPB) and Phosphoenolpyruvate carboxylase (PPC) were used as reference genes for gene expression normalization. Results were expressed as 2(−∆Ct) being ∆Ct the ratio between the target gene and the geometric mean of PPC and RNPB (Livak and Schmittgen, 2001). In order to analyze phylogenetic positioning of the CPD/(6-4)- PHR from Sphingomonas sp. UV9 (PhrSph98) we performed a BLASTp search of the NCBI non-redundant protein database. Results retrieved 250 sequences (Supplementary Table S2) with similarities ranging from 90 to 31% and E-values from 0 to 1.10−63 matching the criteria of homology reported previously (Pearson, 2013). MSA and phylogenetic analysis showed that the closest homologs to PhrSph98 were grouped in a clade (numbered as 7, 8, 9 and 10, Figure 1) containing proteins with E-values of 0 and percentages of identity ranging from 61.32 to 82.04 (Supplementary Table S3). These proteins correspond to Gram- negative bacteria, including Sphingomonas as the most abundant (31/37), and others like Belnapia (1/37), Dankookia (1/37), Sphingosinicella (1/37), Methylobacterium (1/37), Polymorphobacter (1/37) and Croceibacterium (1/37) (Supplementary Table S3). According to the domain profile search using the NCBI conserved domain database (CDD)- NIH, the majority of these proteins contain the PHR2 domain, except for two proteins from node 9 that have a PHRB domain (WP_157177235.1 from Sphingomonas prati and WP_093400154.1 from Sphingomonas jinjuensis, Supplementary Figure S1). In silico proposed 3D model from PhrSph98 suggests that electron transfer to FAD involves Y389-W369-W390-F376-W381-FAD pathway (Marizcurrena et al., 2020). We analyzed the conservation of amino acids involved in electron transfer using a MSA. Figures 1 and 2 show that W369 and 390 were conserved in the 149 sequences analyzed. Y389 is conserved from nodes 1 to 12, being changed for a Thr or Leu (which could not participate in the electron transfer) in the species from node 13 (Figure 1; Supplementary Figure S1). F376 was conserved in nodes 7, 9, 10 and 11 and in WP_115381109 from Sphingomonas sp. strain FARSPH in node 8 (Figure 1). This amino acid was replaced for a Trp in nodes 1 to 6, and for a Tyr in nodes 8 and 12 (Figure 1). In the other predicted proteins, there was a change to non- aromatic amino acids in this position (Supplementary Figure S1). Finally, W381 was conserved from nodes 1 to 12 (Figure 1). Thus, aromatic amino acids important for electron transfer and DNA repair are conserved in species from nodes 1 to 12, suggesting that they constitute PhrSph98 homologs. In class II CPD PHRs, N403 is highly conserved and involved in FAD cofactor binding (Kiontke et al., 2011). This residue from M. mazei class II photolyase (template for PhrSph98 tertiary structure modeling) corresponds to N412 in PhrSph98 (Marizcurrena et al., 2020). MSA analysis showed that this residue was conserved in all the sequences predicted as homologous by phylogeny (Figure 2). A charge compensation during FADH- photoreduction is performed by H- bonding of N403 and the surface exposed D404 in M. mazei CPD PHR (Kiontke et al., 2011). This residue (D413 in PhrSph98) was also conserved in PhrSph98 homologs (Figure 2) indicating that FAD binding is favored in these enzymes. CPD and 6-4 PPs DNA lesions are predicted to be positioned in the binding pocket of PhrSph98 through interaction with the hydrophobic residues W314, A430 and M388 (Marizcurrena et al., 2020). W314 was conserved among all the proteins identified as PhrSph98 homologs (Figure 2). In nodes 1 to 8 and 11 to 13 the position of A430 was replaced for a Trp, as occurs in M. mazei CPD photolyase II (W421) (except for WP_095511993.1, WP_095515698.1 from node 11 and TWT53083.1 from node 13). In nodes 9 and 10, A430 was conserved except for WP_157177235.1, WP_107959005.1, WP_133188229.1, WP_010215059.1, WP_056426589.1. M388 was conserved among all the protein sequences analyzed, except for protein TWT53083.1 (Figure 2). Amino acids E310 and N384 from PhrSph98 are hypothesized to be involved in the stabilization of (i) the CPD radical after electron transfer (by the transfer of a proton from a neutral Glu at the bottom of the active site) and (ii) the anionic thymine radical after bond breakage (through hydrogen bond formation with the N3 amide and C4 carbonyl from CPD lesion) respectively (Marizcurrena et al., 2020). Both amino acids were conserved in the sequences predicted to have bifunctional CPD/(6-4)- activities (Figure 2). Considering these results, over a total of 149 sequences we propose that 55 are true homologs of PhrSph98 (proteins from nodes 1 to 12, Figure 1), as they conserve all the important amino acids described to be involved in electron transfer, cofactor and DNA lesion binding and stabilization (Figure 2). A widespread distribution of these enzymes was observed among bacteria, including different genera from proteobacteria (43/55), planctomycete (1/55), bacteroidetes (7/55), acidobacteria (1/55) and cyanobacteria (3/55) (Figure 1) being the latter the only oxygenic photosynthetic organisms encoding this enzyme. The phylogenetic tree obtained from the CPF proteins was unrooted, with an apparent root on the branch near the group of FeS-BCPs. We manually placed the branches from FeS-BCPs, SPL and bifunctional photolyases as roots in independent trees (Supplementary Figure S3). It was found that the simplest tree which has minimum changes across evolution correspond to SPL as root. The maximum likelihood tree reconstruction features 10 main clades. It comprises SPL, CPD PHRs class I, II and III, plant CRY, plant PHR2, ssDNA PHRs, animal CRY and eukaryotic (6-4) photolyases, FeS-BCPs and bifunctional CPD/(6-4) PHR- like proteins (Figure 3, see Supplementary Figure S4 for full phylogenetic tree). All bifunctional CPD/(6-4) PHR- like proteins constituted one monophyletic group, which was located as a sister group of class II CPD PHRs. This clade was located as a sister group of FeS-BCPs enzymes and all together as sister group of all the rest of the members from the CPF (Figure 3). According to tree inference, FeS-BCPs and eukaryotic 6-4 photolyases lost the CPD activity and gained 6-4 PP photorepair activity through two independent events, whereas bifunctional CPD/(6-4) PHR -like proteins gained 6-4 PP activity but retained CPD repair activity (Figure 3). Results suggest that bifunctional CPD/(6-4) PHR- like proteins are related to class II CPD PHRs, both sharing a common ancestor with FeS-BCPs, in a clade that includes only prokaryotic organisms. Phylogenetic analysis revealed that canonical CPD/(6-4) PHR- like proteins are also present in a few strains of oxygenic photosynthetic organisms. Blast analysis using PhrSph98 as query toward the Viridiplantae group gave no results, indicating that the only oxygenic photosynthetic organisms encoding a bifunctional photolyase were the cyanobacteria Phormidesmis sp. LEGE 1147, Chondrocystis sp. NIES-4102 and Synechococcus sp. PCC 7335 (Figure 1; Supplementary Figure S1). Amino acids involved in DNA repair, lesion and cofactor binding were conserved among PhrSph98 and the cyanobacterial homologs (Figure 4A). Synechococcus sp. PCC 7335 has several adaptations to different ambient conditions as chromatic acclimation (CA), far- red light photoacclimation (FARLIP) and the presence of a non- canonical nitric oxide synthase enzyme (SyNOS) (Ho et al., 2017; Correa-Aragunde et al., 2018; Herrera-Salgado et al., 2018). In this work, we were particularly interested in the strategies of Synechococcus sp. PCC 7335 to cope with UV-B radiation. To analyze that, tertiary structure of PhrSph98 homolog from Synechococcus sp. PCC 7335 (herein named as SyCPD/(6-4) PHR- like protein) was predicted using as template a class II CPD-photolyase from the archaea M. mazei (PDB ID: 2XRY) using HHPred online available server. Results obtained show that PhrSph98 and SyCPD/(6-4) PHR- like protein (identity of 40.71%) adopted similar tertiary structure (Figure 4B). Molecular modeling indicate that the positioning of the amino acids Y389-W369-W390-F376-W381 were conserved in SyCPD/(6-4) PHR- like protein (Figure 4B) suggesting it may be a true bifunctional CPD/(6-4)- PHR. Genomic context analysis revealed that SyCPD/(6-4) PHR- like gene constitutes a two gene operon assembly with another PHR (herein refer as PHRs operon). This operon is only conserved in Phormidesmis sp. LEGE 11477 (accession JADEXO010000055.1, 81.68% of identity and 95% of query coverage) not in Chondrocystis sp. NIES-4102. This study provides evidence of a bifunctional CPD/(6-4) PHR- like protein, as well as a putative PHRs operon in an oxygenic photosynthetic organism. PHRs operon is encoded in the DNA negative strand (coding sequence from position 38,587 to 35,587). Two putative −10 (TGCTATACA) and −35 (TTGAAG) boxes were found in the promoter region using the bacterial promoter prediction software BPROM. The analysis of potential transcription factor binding sites revealed the sequences TCACAATT from cyclic AMP receptor protein (CRP) and ACAGACAA from integration host factor (IHF) (Figure 5A). The operon consists of two unidirectional structural genes: an upstream deoxyribodipyrimidine photo-lyase (PHR, WP_038015784.1) and a downstream hypothetical protein (SyCPD/(6-4) PHR- like, homolog to PhrSph98, WP_038015781.1) with different translation frames (Figure 5A). Both ORFs overlap in 11 nt within the sequence ATGTCAGTTGA where TGA is the stop for the upstream gene, and ATG is the start of the second (Figure 5A). We also found a Shine- Dalgarno (SD) intragenic sequence (AGGAG) preceding in seven nucleotides the ATG from the downstream gene. This sequence is absent in the leading gene (Figure 5A). Details of PHRs operon nucleotide sequence are shown in Supplementary Figure S5. Photolyases transcript accumulation is induced upon UV-B exposition in several organisms, protecting them from cell death (An et al., 2021; Fernández et al., 2021, 2022; He et al., 2021). First, we analyzed the effect of UV-B irradiation (1.6 W m−2) on Synechococcus sp. PCC 7335 cell culture measuring the OD at 750 nm. As shown in Figure 5B, UV-B does not affect Synechococcus sp. PCC 7335 optical density at 750 nm up to 3 h- treatment. To investigate whether the two PHRs ORFs form a polycistronic mRNA, the intergenic region between PHR and SyCPD/(6-4) PHR- like was amplified by qPCR from non- irradiated culture cDNA. Figure 5C shows the amplification of the expected DNA fragment (233 nt), confirming the expression of both PHRs as a polycistronic mRNA. The expression of PHR, SyCPD/(6-4) PHR- like and the intergenic region were also evaluated during UV-B treatment. Results show that PHR, bifunctional CPD/(6-4) PHR- like and the intergenic region expression increased 3.5, 2 and 3- fold, respectively, after 15 min of UV-B exposition (Figure 5D). PHR transcript was statistically significantly higher compared to bifunctional photolyase. After 30 min of UV-B exposure, abundance of all amplicons decreased to control levels (Figure 5D). Photolyases repair the UV-induced DNA damages of CPDs and 6-4 PPs using blue/ UV-A light. The FAD catalytic cofactor, conserved in the whole protein superfamily of photolyase/cryptochromes, adopts a unique folded configuration at the active site and plays a critical role in DNA repair. FADH- functions as the active state to allow efficient electron injection into DNA damage in a process that involves intramolecular electron transfer (Zhang et al., 2017). Based on the in silico modeling of PhrSph98 from Sphingomonas sp. UV9, Marizcurrena et al. (2020) suggested that electron transfer to FAD involves Y389-W369-W390-F376-W381 amino acids pathway. In class II CPD photolyases, an Asn residue (N403) is highly conserved and involved in FAD cofactor binding. Kiontke et al. (2011) showed that the class II CPD photolyase from M. mazei owing N403A and N403L mutations has impaired FAD binding. In contrast, the N403D mutant has FAD incorporation of at least 70% (Kiontke et al., 2011). Also, in class II CPD PHRs, a charge compensation during FADH- photoreduction is performed by H- bonding of the side chain from N403 with D404, whereas in class I this role is taken up by a Glu residue (Kiontke et al., 2011). We showed that all these residues are conserved in PhrSph98 homologs, indicating that bifunctional CPD/(6-4) PHR- like proteins are present in heterotrophic bacteria like proteobacteria, planctomycete, bacteroidete, acidobacteria and in oxygenic photosynthetic bacteria. Conservation of class II CPD PHRs residues suggests that these enzymes belong to this class, which was also supported by phylogenetic analysis, positioning bifunctional photolyases as a sister group of CPD class II photolyases. Additionally, tertiary structure model comparison between PhrSph98 and SyCPD/(6-4) PHR- like protein shows that the structural arrangement of residues involved in electron transfer to FAD, FAD binding, and DNA damages binding are conserved. The tertiary structure conservation of PhrSph98 hallmarks, allows us to suggest that SyCPD/(6-4) PHR- like is a true homolog of this enzyme. Several ancestors for the CPF family have been proposed (Mei and Dvornyk, 2015; Miles et al., 2020; Vechtomova et al., 2020). After analyzing different phylogenetic tree topologies, we found that positioning of SPL proteins as root retrieves the tree with lesser changes during evolution. Accordingly, Xu et al. (2021) suggested that the first common ancestor of the CPF might be a SPL protein, with CPD photolyase activity and a FeS cluster, characteristic of ancient proteins. The phylogenetic tree shows that bifunctional CPD/(6-4) PHR- like proteins are a sister group of CPD class II photolyases and both constitute a clade with FeS-BCP proteins, represented only by prokaryotic organisms. 6-4 PP photorepair appear independently in FeS-BCP and bifunctional proteins, probably by mutational events, and conservation of CPD activity is only retained in bifunctional enzymes. A 6-4 PP repair enzyme can be converted into a CPD repair one with only three mutations. However, eleven mutations are needed for the vice versa conversion (Yamada et al., 2016). This asymmetric functional conversion was suggested as a more complex repair mechanism for 6-4 PPs repair (Yamada et al., 2016). Although it seems that conversion of CPD to 6-4 PP activity is more difficult in terms of the number of mutations, the phylogenetic tree supports the origin of the CPF with a CPD repair family. An operon is a cluster of neighboring genes that are transcribed together and therefore encodes several proteins. Its evolution is hypothesized to be adaptive and toward coordinated optimization of functions (Memon et al., 2013). Results presented here indicate that only the cyanobacteria Synechococcus sp. PCC 7335 and Phormidesmis sp. LEGE 11477 encode a PHRs operon. It has recently been proposed a new classification of the genus Synechococcus, based on habitat distribution patterns (seawater, freshwater, brackish and thermal environments) that reflect the ecological and evolutionary relationships of its members. According to Salazar et al. (2020), Synechococcus sp. PCC 7335 genome did not cluster with any other Synechococcus genome, and matched with the Phormidesmiales order, being named Phormidesmis mexicanus PCC 7335. This new genus assignment is supported by the conservation of PHRs operon both in Phormidesmis sp. LEGE 11477 and Synechococcus sp. PCC 7335. Thus, the Phormidesmis genus is apparently the unique containing a PHRs operon including a bifunctional CPD/(6-4) PHR- like protein. Currently, there is a low number of publicly available cyanobacteria genomic sequences (0.6% compared to the total number of genomes available for bacteria and archaea; Alvarenga et al., 2017). Future genome sequencing of this phylum will allow us to determine whether this operon occurs in other cyanobacteria. Operon evolution analysis in cyanobacteria shows that genes in highly and moderately conserved operons code for key cellular processes, such as photosynthesis. Conversely, genes in poorly conserved operons may code for functions possibly linked to niche adaptation. Also, newly acquired operons are greater in number, smaller in size, with wider intergenic spacing, and weakly coregulated compared to ancient operons. However, a sub-clade comprising the genera Synechocystis, Microcystis, Cyanothece, and Synechococcus sp. PCC 7002 forms an exceptional case with small intergenic spaces (Memon et al., 2013). Synechococcus sp. PCC 7335 PHRs operon matched the criteria of poorly conserved and no intergenic space, indicating that may be a newly formed operon with an adaptative role to protect this organism from high environmental UV-B levels. Here, we observed that UV-B induces the expression of the PHRs operon. This arrangement guarantees that the production of the proteins related to DNA repair under UV-B induced damage is simultaneously switched on and off and provides the ability to fast acclimate to new growth conditions. Moreover, there is a potential binding site of the transcription factor CRP upstream of the PHRs operon promoter −35 and −10 boxes. Class I CRP-dependent promoters have this sequence upstream of the promoter boxes. This transcriptional activation involves the interaction between CRP and the carboxy-terminal domain of the RNA-polymerase (RNA-P) α-subunit, facilitating the binding of RNA-P to the promoter (Soberón-Chávez et al., 2017). Additionally, a CRP knockout strain of Deinococcus radiodurans is sensitive to UV radiation, suggesting an important role of CRP in UV protection (Yang et al., 2016). Also, in D. radiodurans, a CRP homolog regulates the expression of different DNA repair proteins as RecN (in response to double-stranded DNA breaks), PprA (RecA-independent, DNA repair-related protein) and UvsE (UV damage endonuclease that is involved in nucleotide excision repair). Based on this, CRP transcription factor is an interesting candidate to be explored for PHRs operon UV-B inducible expression regulation. Until now, only two reports describe photolyases encoded in operons. Osburne et al. (2010) characterized an operon of two DNA repair genes (nudix hydrolase- photolyase operon) in the UV hyper-resistant strain marine cyanobacterium Prochlorococcus MED4 which has constitutively upregulated expression but was induced by UV-B in the WT Prochlorococcus strain. Also, Sancar et al. (1984), reported an operon of a photolyase with a protein of unknown function in E. coli. To our knowledge, our study constitutes the first report that describes a UV-B inducible operon encoding two photolyases, being one of them with a putative bifunctional role in the repair of CPD and 6-4 PP damages. We found that upon 15 min of UV-B exposition, PHR transcript accumulates faster than bifunctional CPD/(6-4) PHR- like. The increase in transcripts level during UV-B irradiation, suggest a role in DNA repairing. The distinct expression of both intra-operonic genes constitutes a deviation from the generally expected co-expression behavior. A possible explanation may be the occurrence of translational coupling where the juxtaposition of the translational stop codon of an upstream gene with the translational start codon of the downstream one reduces the expression levels of the latter and affects mRNA stability (Dogra et al., 2015; Wright et al., 2021). In this sense, the upstream translating ribosome destabilizes RNA secondary structure which would prevent optimal expression of the downstream gene (Schümperli et al., 1982). We also found an intragenic SD sequence preceding the bifunctional CPD/(6-4) PHR- like gene, that may be involved in termination- reinitiation of translation. In overlapping genes, the start codon of the downstream gene is typically preceded by an SD motif in most archaea and bacteria (Huber et al., 2019). Cyanobacteria make ample use of the SD motif for reinitiation at overlapping gene pairs while less than 20% of SD sequences are found in the leading gene from operon (Huber et al., 2019). In agreement, no SD sequence was found in the leading gene of PHRs operon. Synechococcus sp. PCC 7335 was isolated from a snail shell, from the intertidal zone in Puerto Peñasco, Mexico. As previously mentioned, they have several specific adaptations to prevailing light. For example, they contain chlorophyll f that absorbs far-red light and go under FaRLiP response and also perform complementary chromatic acclimation. They also contain an operon of PHRs that may contribute to UV-B tolerance, and one of these PHRs may encode for a bifunctional CPD/(6-4) PHR- like protein. Genome analysis of Synechococcus sp. PCC 7335 denotes the presence of two other CPF: a cryptochrome/photolyase family and a deoxyribodipyrimidine photo- lyase (WP_006457129.1, locus 27315-28829 and WP_006453873.1, locus 88140-86491 respectively). BlastP analysis of these two proteins shows similarity with cyanobacteria proteins. However, PHRs operon shares more similarities with proteobacteria photolyases. In this sense, it is possible that Synechococcus sp. PCC 7335 acquires the PHRs operon gene through horizontal gene transfer from proteobacteria. Whether these photolyases have redundant roles or contribute to classifying Synechococcus sp. PCC 7335 as an UV-B resistant bacteria needs to be explored. A metagenomic analysis combined with solar radiation measurements described a positive correlation between higher UV-B levels in the aquatic microbiome environments and the abundance of CPF genes (Alonso-Reyes et al., 2020). The finding of an operon of PHRs including a bifunctional CPD/(6-4)- photolyase- like protein may explain its ability to survive in an environment exposed to high doses of UV-B irradiation as are the intertidal zones (Hanelt et al., 1997). Additionally if bifunctional CPD/(6-4) PHR- like gene is under horizontal gene transfer is also an open question. Furthermore, it will be interesting to analyze whether the repair of two DNA damages by a single protein may confer on Synechococcus sp. PCC 7335 an advantage over expressing a CPD and 6-4 PP photolyases. Exposure of humans to UV-B provokes DNA damage in keratinocytes that contributes to photoaging and implies a risk of progression to squamous cell carcinoma. Photolyases are not encoded by placental mammals. Photoprotection by the application of conventional sunscreens is not sufficient once DNA damage occurred. Diverse studies suggested that the addition of liposome-encapsulated CPD photolyases to sunscreens helps to diminish CPDs in human skin (Luze et al., 2020). Indeed, the biochemical characterization of cyanobacteria photolyases with potential bifunctional activity in the repair of both CPDs and 6-4 PPs may provide useful biotechnological information for the cosmetic and healthcare industries, as these organisms are emerging as sources of bioactive compounds. The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material. MBF conceived the original idea, carried out the phylogenies and experiments, and wrote the manuscript. LL collaborated in qPCR experiments. MBF, LL, NC-A, and RC contributed to the interpretation of the results. NC-A and RC contributed to the design of the experiments and critically revised the manuscript. RC supervised the project. All authors contributed to the article and approved the submitted version. This work was supported by the Agencia Nacional de Promoción Científica y Tecnológica (ANPCYT) grant number 2019-1577 to MBF, 2018-2524 to NC-A and 2019-3436 to RC, and Universidad Nacional de Mar del Plata grant number EXA 1030/21 to RC. MBF, NC-A, and RC are permanent researchers of CONICET Argentina. LL is PhD student fellow of CONICET. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649892

36323312

Zsolt Czimmerer,Laszlo Halasz,Bence Daniel,Zsofia Varga,Krisztian Bene,Apolka Domokos,Marten Hoeksema,Zeyang Shen,Wilhelm K. Berger,Timea Cseh,Karoly Jambrovics,Zsuzsanna Kolostyak,Ferenc Fenyvesi,Judit Varadi,Szilard Poliska,Gyorgy Hajas,Istvan Szatmari,Christopher K. Glass,Attila Bacsi,Laszlo Nagy

The epigenetic state of IL-4-polarized macrophages enables inflammatory cistromic expansion and extended synergistic response to TLR ligands

01-11-2022

SUMMARY Prior exposure to microenvironmental signals could fundamentally change the response of macrophages to subsequent stimuli. It is believed that T helper-2 (Th2)-cell-type cytokine interleukin-4 (IL-4) and Toll-like receptor (TLR) ligand-activated transcriptional programs mutually antagonize each other, and no remarkable convergence has been identified between them. In contrast, here, we show that IL-4-polarized macrophages established a hyperinflammatory gene expression program upon lipopolysaccharide (LPS) exposure. This phenomenon, which we termed extended synergy, was supported by IL-4-directed epigenomic remodeling, LPS-activated NF-κB-p65 cistrome expansion, and increased enhancer activity. The EGR2 transcription factor contributed to the extended synergy in a macrophage-subtype-specific manner. Consequently, the previously alternatively polarized macrophages produced increased amounts of immune-modulatory factors both in vitro and in vivo in a murine Th2 cell-type airway inflammation model upon LPS exposure. Our findings establish that IL-4-induced epigenetic reprogramming is responsible for the development of inflammatory hyperresponsiveness to TLR activation and contributes to lung pathologies.

The epigenetic state of IL-4-polarized macrophages enables inflammatory cistromic expansion and extended synergistic response to TLR ligands Prior exposure to microenvironmental signals could fundamentally change the response of macrophages to subsequent stimuli. It is believed that T helper-2 (Th2)-cell-type cytokine interleukin-4 (IL-4) and Toll-like receptor (TLR) ligand-activated transcriptional programs mutually antagonize each other, and no remarkable convergence has been identified between them. In contrast, here, we show that IL-4-polarized macrophages established a hyperinflammatory gene expression program upon lipopolysaccharide (LPS) exposure. This phenomenon, which we termed extended synergy, was supported by IL-4-directed epigenomic remodeling, LPS-activated NF-κB-p65 cistrome expansion, and increased enhancer activity. The EGR2 transcription factor contributed to the extended synergy in a macrophage-subtype-specific manner. Consequently, the previously alternatively polarized macrophages produced increased amounts of immune-modulatory factors both in vitro and in vivo in a murine Th2 cell-type airway inflammation model upon LPS exposure. Our findings establish that IL-4-induced epigenetic reprogramming is responsible for the development of inflammatory hyperresponsiveness to TLR activation and contributes to lung pathologies. Features of macrophages (Macs) as well as their contribution to disease processes are determined by the tissue microenvironment, pathogen-derived molecules, and cytokines. Conventionally, the endpoints of cytokine-induced Mac activation states are T helper-1 (Th1) cell-type cytokine interferon-gamma (IFNγ)-induced classical (M(IFNγ)) and T helper-2 (Th2) cell-type cytokine interleukin 4 (IL-4)-mediated alternative (M(IL-4)) Mac polarization. M(IFNγ) Mac polarization programs are associated with inflammatory response, while M(IL-4)-type Macs protect against helminth infection and promote tissue regeneration. The classical and alternative Mac polarizing cytokines could be present simultaneously or sequentially in the microenvironment along with various pathogens or normal microbiota components and can lead to more nuanced, specialized Mac phenotypes and functions often associated with disease progression. Thus, the complex and changing milieu inevitably leads to heterogeneous polarization states in vivo, which are not well characterized (Gordon and Martinez, 2010; Murray et al., 2014; Murray and Wynn, 2011). There are examples for enhanced inflammatory gene expression upon two distinct pro-inflammatory signaling events and blunting by anti-inflammatory cytokines. Activation of IFNγ--polarized Macs by Toll-like receptor (TLR) ligands leads to the so-called super-induction of many canonical inflammatory genes (Qiao et al. 2013). In addition, IFNγ prevents and reverses TLR ligands-induced Mac tolerance resulting in an exaggerated inflammatory phenotype in autoimmune diseases (Hu et al., 2008; Chen and Ivashkiv, 2010). Helminth infection-induced IL-4-dependent alternative Mac polarization markedly modifies the response against bacterial pathogens (Mylonas et al., 2009; Potian et al., 2011; Weng et al., 2007; Du Plessis et al., 2013). In addition, environmental lipopolysaccharide (LPS) contamination or expansion of particular Gram-negative bacteria in the bronchial airway microbiota exacerbates Th2 cell-type asthma (Eisenbarth et al., 2002; Goleva et al., 2013; Huang et al., 2011; Wang et al., 2021). The epigenomic basis of such interactions is not well understood. The epigenomic and transcriptional programs of Mac polarization and inflammatory signal responses are tightly and dynamically regulated. First, lineage-determining transcription factors (LDTFs), including ETS domain, transcription factor PU.1, CCAAT-enhancer-binding proteins (C/EBPs), activator protein 1 (AP-1), or Runt-related transcription factor 1 (RUNX1), determine the Mac-specific-enhancer repertoire. Second, signal-dependent transcription factors (SDTFs), including IL-4- and IL-13-activated signal transducer and activator of transcription-6 (STAT6), IFNγ-activated STAT1, and LPS-activated nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), or AP-1, are responsible for immediate-early transcriptional responses for Mac polarization and inflammatory signals (Glass and Natoli, 2016). Third, additional Mac-polarization-signal-induced or activated transcription factors such as IL-4-induced EGR2 or IFNγ-activated IRF1 and IRF8 also contribute to determining the late, stable epigenomic program of the differentially polarized Macs along with other transcription factors further downstream (Langlais et al., 2016; Daniel et al., 2020). STAT1, STAT3, and IRF1 transcription factors contribute to the inflammatory hyper-responsiveness of IFNγ-primed Macs (Qiao et al. 2013; Kang et al., 2019). We recently found that the anti-inflammatory features of IL-4-primed Macs are based on the direct repressor activities of STAT6, affecting many TLR target genes (Czimmerer et al., 2018). Moreover, IL-4-enhanced inflammatory responsiveness has been also observed for certain genes, indicating that the effect of IL-4 on the TLR response is much more complex (Czimmerer et al., 2018; Major et al., 2002; Varin et al., 2010). This is clinically relevant because there are niches in the body, such as the lung, where a Th2 cell environment and pathogenic TLR activation may occur simultaneously and can lead to asthma exacerbation with glucocorticoid resistance (Eisenbarth et al., 2002; Goleva et al., 2013; Huang et al., 2011; Wang et al., 2021). Here, we uncovered how IL-4 priming influences the epigenomic and transcriptomic outcomes of inflammatory responses in Macs. We identified a specific gene set displaying enhanced or de novo LPS responsiveness in IL-4-primed murine bone marrow-derived Macs (BMDMs). Enhanced LPS responsiveness was associated with IL-4-priming-dependent epigenomic reprogramming mediated by BRD4, as well as increased NF-κB-p65 binding and enhancer activity. Finally, IL-4-facilitated LPS responsiveness (termed extended synergy) was found to be conserved between human and mouse Macs and regulated by transcription factors STAT6, NF-κB-p65, and EGR2 ex vivo in distinct murine tissue-resident and monocyte-derived murine Mac subsets and in vivo in alveolar Macs following the induction of Th2-type airway inflammation. We set out to study the LPS responsiveness of alternatively polarized Macs focusing first on the LPS-induced chemokine and cytokine signature. Using RNA-seq analysis, we found that 72 out of 229 genes from the “cytokine activity” gene ontology (GO) category (GO:0005125) showed LPS-induced mRNA expression in either non-polarized or alternatively polarized murine BMDMs or both. The LPS-dependent induction of many (14) chemokines and cytokines was significantly attenuated by IL-4 priming. However, another 29 chemokines and cytokines showed significantly enhanced LPS inducibility in IL-4 primed BMDMs (Figures 1A and S1A). In order to investigate whether the modulating effects of IL-4 priming on LPS-induced cytokine and chemokine expression are restricted to TLR4 activation, we activated IL-4-primed and non-polarized BMDMs with ten different TLR agonists (Figure 1B). We measured the mRNA expression of five augmented and three attenuated LPS-inducible genes using RT-qPCR. Among the synergistically activated genes, Ccl17 and Ccl22 mRNA expression showed significantly elevated induction in IL-4-primed Macs upon each TLR agonist activation (Figure 1C). Although the increased responsiveness of the additional three selected genes in alternatively polarized Macs showed greater TLR ligand specificity, it was limited to the activation by TLR4 ligand on Il23a (Figure 1C). Similarly, IL-4-mediated repression was not restricted to TLR4 activation (Figure S1B). NF-κB transcription factor complex mediates the response to TLR ligands; thus, we examined the activation of the NF-κB signaling in non-polarized and IL-4-primed BMDMs following LPS activation. Immunoblot data showed that IL-4 pre-treatment did not influence the LPS-mediated inhibitory IκBα protein degradation in wild-type (WT) and Stat6−/− BMDMs (Figure 1D). Immunoblot and fluorescence microscopy revealed that neither the expression nor the LPS-induced nuclear translocation of NF-κB-p65 was affected by IL-4 priming (Figures 1D–1F). We asked whether NF-κB signaling was necessary for the enhanced gene-specific LPS responsiveness in IL-4-primed BMDMs. We studied the selected five genes in the presence of two NF-κB pathway inhibitors (BAY11–708 and Bot64). We found that the inhibitors attenuated the LPS-induced expression of the selected genes in both non-polarized and IL-4-primed murine BMDMs (Figure 1G). Finally, we investigated whether IL-4 removal after 24 h of priming influences the ability of Macs to respond more robustly to LPS; thus, transcriptional memory developed. Therefore, we measured LPS-induced mRNA expression of the selected five cytokines and chemokines without IL-4 washout and at different time points after washout (Figure S1C), As expected, IL-4-induced expression of alternative Mac polarization markers Arg1 and Chil3l3 returned to baseline within 24 h after the IL-4 washout (Figure S1D). In contrast, the elevated LPS responsiveness of Ccl2, Il12a, and Il23a was partially retained for a minimum of 24 h following IL-4 washout (Figure S1E), while the augmented LPS-mediated induction of Ccl17, and Ccl22, was not observed immediately after IL-4 washout or 24 h later (Figure S1E). However, the enhanced LPS responsiveness of Ccl17 and Ccl22 returned 48 and 72 h after the IL-4 washout (Figure S1D). Therefore, IL-4 priming synergizes with TLR ligand activation that requires an intact NF-κB signaling but does not affect the nuclear translocation of NF-κB-p65. Moreover, there is gene-specific and continuous or intermittent time-limited transcriptional memory after IL4 exposure and thus the signals do not need to co-occur to induce extended synergy. Next, we investigated whether the enhanced LPS responsiveness of the previously identified cytokine and chemokine gene set was a consequence of IL-4 priming-modulated NF-κB-p65 binding and enhancer activation. Therefore, we first performed NF-κB-p65-specific chromatin immunoprecipitation followed by sequencing (ChIP-seq) in WT and Stat6−/− BMDMs (Figure S2A). Several differences were identified in the NF-κB-p65 cistrome between non-polarized and IL-4-primed WT Macs following 1-h LPS exposure (Figure S2B). IL-4 priming increased the number of NF-κB-p65 peaks from 25,428 to 48,507 (Figure 2A) and the occupancy of LPS-activated NF-κB-p65 at 26,440 genomic regions (Figure 2B; Table S1). In contrast, there was a significant reduction of NF-κB-p65 binding at only 1,545 genomic sites (Figure 2B; Table S1). Based on the presence or absence of NF-κB-p65 peaks in LPS-activated non-polarized Macs, the IL-4-facilitated NF-κB-p65-bound genomic regions could be divided into two groups. “De novo” NF-κB-p65 peaks could be detected at 15,236 genomic regions, while “enhanced” NF-κB-p65 binding could be found at an additional 11,204 regulatory regions (Figures 2C, 2E, and 2F). The IL-4 priming-induced NF-κB-p65 peak number and occupancy in LPS-activated alternatively polarized BMDMs proved to be strictly STAT6 dependent (Figures 2A, 2E, and 2F). The majority of genomic sites exhibiting IL-4-facilitated NF-κB-p65 binding were located in distal regulatory regions, including intergenic and intronic elements, with marginal differences between the different categories (Figure 2D). To gain insight into the enhancer activity at the IL-4 priming-facilitated NF-κB-p65 binding-associated genomic sites, we examined the elongation-specific RNAPII-pS2 binding by ChIP-seq in LPS-exposed non-polarized and alternatively polarized Macs. 73% (11,166) of de novo and 94% (10,564) of enhanced NF-κB-p65 binding-associated genomic regions proved to be RNAPII-pS2 positive in at least one experimental condition (Figure S2C). K-means clustering of RNAPII-pS2-positive sites identified five clusters among both the de novo and the enhanced NF-κB-p65 peaks (Figure S2D). 42.9% of de novo and 49.9% of enhanced NF-κB-p65 binding-associated genomic regions showed maximal RNAPII-pS2 occupancy in the LPS-activated alternatively polarized BMDMs (Figure S2D). Among these genomic sites, 2,034 (18.2%) de novo and 1,831 (17.3%) enhanced NF-κB-p65 peaks (cluster I) were associated with RNAPII-pS2 enrichment in IL-4-primed Macs, which was further increased by LPS activation (Figures 2G and 2H). An additional 2,752 (24.6%) de novo and 3,444 (32.6%) enhanced NF-κB-p65-bound genomic regions (cluster II) showed LPS-induced RNAPII occupancy in non-polarized BMDMs. However, the LPS-induced RNAPII binding was significantly elevated in the IL-4-primed Macs (Figures 2G and 2H). The other three clusters were associated with IL-4 priming-attenuated (cluster III) or LPS-repressed (clusters IV and V) RNAPII-pS2 binding in the LPS-activated alternatively polarized Macs (Figures S2D–S2F). These findings indicate that IL-4 priming induces the expansion of LPS-activated NF-κB-p65 cistrome resulting in a distinct active enhancer repertoire. To identify the gene sets affected by the LPS-induced NF-κB-p65 cistrome expansion and increased enhancer activation in alternatively polarized Macs, we performed RNA-seq in IL-4-primed and LPS-activated WT and Stat6−/− BMDMs. Our global transcriptome analysis demonstrated that the alternatively polarized WT BMDMs have a distinct LPS response compared with the WT non-polarized and IL-4-primed Stat6−/− Macs (Figure S3A). This distinct LPS response included extended synergy impacting the mRNA expression of 1,318 genes upon IL-4 priming and LPS activation (Figures 3A and S3B; Table S2). The elevated LPS response was strictly STAT6 dependent in IL-4-stimulated BMDMs (Figures 3A, S3A, and S3B). Based on the individual IL-4 and LPS responsiveness, the synergistically activated genes could be divided into nine clusters, of which four contained more than 90% of the genes (Figure 3B). 663 genes (50.30%) were induced by both IL-4 and LPS stimulation. 328 and 87 genes (24.88% and 6.60%) were activated by LPS or IL-4 alone, respectively. Finally, LPS-dependent activation of 141 genes (10.69%) was restricted to the IL-4-primed BMDMs, showing so-called de novo LPS responsiveness (Figure 3B). Next, we investigated RNAPII-pS2 binding at the gene bodies of genes induced by extended synergy using ChIP-seq. RNAPII-pS2 binding showed a similar pattern to the steady-state mRNA expression suggesting that prior IL-4 exposure enhances LPS-induced gene expression primarily at the transcriptional level (Figure S3C). To determine the link between enhanced LPS responsiveness and de novo and enhanced NF-κB-p65-bound genomic regions, we assigned the regulatory elements from CI and CII clusters of de novo and enhanced NF-κB-p65 peak sets to the synergistically activated genes. We annotated 592 (cluster I) plus 910 (cluster II) de novo and 455 (CI) plus 933 (CII) enhanced NF-κB-p65 peaks falling into a ±100-kb genomic window around the transcription start site of the synergistically activated genes (Figures 3C–3E and S3D). 871 genes showed extended synergism and were associated with a minimum of one regulatory element exhibiting a de novo or enhanced pattern of NF-κB-p65 binding, while we could not assign such regulatory regions to the remaining 447 genes (Figure 3F). Based on the very high degree of enrichment of de novo or enhanced NF-κB-p65 peak in the examined gene loci, we divided the de novo and/or enhanced NF-κB-p65 bound enhancer-associated genes into three categories. We identified 281 genes, among which more than 62.5% of the annotated synergistically activated and NF-κB-p65-positive enhancers belonged to the enhanced NF-κB-p65 binding-associated enhancer subset (Figure 3G; “dominantly enhanced” category). In contrast, 377 genes were associated with de novo NF-κB-p65 binding-linked enhancers in more than 62.5% (Figure 3G; “dominantly de novo” category). The remaining 213 genes were associated with de novo and enhanced NF-κB-p65 peaks of nearly 50–50 (Figure 3G, “both” category). During the characterization of the NF-κB-p65-mediated direct synergistic activation, we found that the number of regulatory regions annotated to highly synergistically activated genes showed that a remarkable portion of genes possessed only 1–2 enhanced and de novo NF-κB-p65-binding-associated regulatory elements. Additionally, the genes with a minimum of one de novo and enhanced NF-κB-p65 peak were associated with slightly more genomic regions with de novo NF-κB-p65 binding (Figure 3H). To evaluate the identified three de novo and/or enhanced NF-κB-p65 peak-associated gene clusters, we performed GO biological process analysis using the Enrichr algorithm (Xie et al., 2021). We found that several immunologically relevant biological-process-associated gene sets were significantly enriched in each gene cluster, indicating that synergistic gene activation affects many aspects of the inflammatory and immunomodulatory functions of Macs. For instance, “monocyte chemotaxis” (GO:0002548) and “cytokine-mediated signaling pathway” (GO:0019221) GO biological process categories were enriched in the dominantlyenhancedNF-κB-p65-bound enhancers-associated gene cluster. Among others, the “neutrophil-mediated immunity” (GO:0002446) and “NIK-NF-κB signaling” (GO0038061) categories-linked genes were enriched in the dominantly de novo NF-κB-p65-bound enhancers-associated gene cluster, while the enrichment of the “inflammatory response” (GO:0006954) and “positive regulation of cytokine production” (GO:0001819) GO-terms-linked genes was observed in both de novo and enhanced NF-κB-p65 binding-associated gene clusters (Table S2), indicating that several aspects of inflammatory response and immunoregulation are affected by the extended synergy. Next, we chose six representative genes associated with de novo and enhanced NF-κB-p65 peaks for further examination (Figures 3G, 3I, 3J, and S3E). In these cases, elevated RNAPII-pS2 binding was observed both on the gene bodies and at the NF-κB-p65-bound regulatory elements in alternatively polarized and LPS-activated BMDMs (Figures 3I and S3E), suggesting that the given enhancer(s) is(are) regulating the gene. To test this assumption, we measured the mRNA expression of these genes with RT-qPCR and confirmed their IL-4 priming-mediated and STAT6-dependent high-activation following LPS exposure (Figures 3J and S3F). In the case of three selected chemokines (CCL2, CCL17, and CCL22), we confirmed the significantly elevated protein secretion in IL-4-primed and LPS-activated WT BMDMs. In contrast, IL-4 priming failed to increase the LPS-induced secretion in the absence of STAT6 (Figure S3G), establishing a requirement of this factor. Enhancer RNA (eRNA) expression is a proven and well-accepted surrogate for detecting and characterizing enhancer activity (Daniel et al., 2014; Natoli and Andrau, 2012). Therefore, we measured eRNA expression on three selected enhancers with elevated NF-κB-p65 and RNAPII-pS2 bindings in LPS-exposed alternatively polarized WT BMDMs. As expected, IL-4 priming facilitated LPS-induced eRNA expression in a STAT6-dependent manner (Figure 3K). Therefore, the distinct LPS responsiveness in alternatively polarized Macs is mediated, in large part, by elevated NF-κB-p65 binding and enhancer activity at their annotated regulatory regions of highly activated genes. In order to investigate the epigenomic and chromatin level changes at the elevated NF-κB-p65 and RNAPII-pS2 co-bound regulatory elements of the synergistically activated genes, we assessed chromatin openness using the assay for transposase-accessible chromatin sequencing (ATAC-seq). We found that both de novo and enhanced NF-κB-p65-bound genomic regions showed weak ATAC-seq signals in the non-polarized state without LPS activation (Figures 4A, 4B, 4E, and S4A). IL-4 priming and LPS activation alone could induce chromatin opening, but maximal chromatin accessibility was observed in the alternatively polarized MACs following LPS activation (Figures 4A, 4B, 4E, and S4A). Thus, the elevated LPS-activated NF-κB-p65 binding and enhancer activity is linked to IL-4 priming-induced chromatin remodeling in alternatively polarized Macs, and most likely, it is required for that. It has been demonstrated that bromodomain and extraterminal (BET) proteins, including BRD2 and BRD4, play an essential role in the LPS-induced gene and enhancer activation in Macs (Belkina et al., 2013; Hah et al., 2015). Therefore, we decided to examine whether the elevated NF-κB-p65 and RNAPII-pS2 co-binding are accompanied by enhanced BRD4 occupancy during the inflammatory response of alternatively polarized Macs. To do this, we assessed the genome-wide localization of BRD4. Cistrome analysis (ChIP-seq) unequivocally supported our hypothesis and showed that IL-4 priming and LPS activation-dependent BRD4 binding perfectly aligned with our RNAPII-pS2-specific ChIP-seq results at de novo and enhanced NF-κB-p65-bound genomic regions (Figures 4C–4E and S4A). These results raised the possibility that BRD4 is required for elevated LPS responsiveness in alternatively polarized Macs. To test the potential role of BET proteins, we aimed to examine the LPS-induced extended synergistic activation of the six previously selected genes in alternatively polarized BMDMs in the presence and the absence of BRD bromodomain inhibitor JQ1 (Figure S4B). The LPS-dependent induction was completely or partially inhibited in both non-polarized and alternatively polarized BMDMs except for Ccl2 (Figures 4F and S4C). Similarly to the JQ1-dependent attenuation of extended synergistic activation of the mRNAs mentioned above, JQ1 could also inhibit the LPS-induced eRNA expression in IL-4-primed and control Macs at the previously selected distal regulatory regions except for Ccl2_−19Kb (Figures 4G and S4D). These findings show that the IL-4-facilitated LPS responsiveness is accompanied by chromatin remodeling and requires BRD4 in most cases. In order to identify transcription factors participating in the elevated LPS responsiveness of alternatively polarized Macs, we performed transcription factor motif enrichment analysis at elevated NF-κB-p65 and RNAPII-pS2 co-bound genomic elements in the loci of synergistically activated genes. As expected, we could detect significant enrichment of Mac-specific various LDTF such as PU.1 and AP1, and the LPS-activated NF-κB-p65 binding motifs at both de novo and enhanced NF-κB-p65 binding-associated regulatory elements (Figure S5A). Additionally, the EGR binding motif was also significantly enriched at the examined NF-κB-p65 peak clusters except for cluster II from the enhanced NF-κB-p65 peak set, indicating that EGR transcription factors can play a pivotal role in the elevated inflammatory responsiveness of alternatively polarized Macs. EGR2 transcription factor is an IL-4 inducible member of the EGR transcription factor family in Macs and is essential to establish and maintain the late, stable epigenomic program in alternatively polarized Macs (Daniel et al., 2020; Hoeksema et al., 2021). Therefore, we decided to study the contribution of EGR2 to the distinct LPS responsiveness of alternatively polarized Macs. First, we performed NF-κB-p65-specific ChIP-seq experiments in EGR2 deficient (Egr2fl/fl) and control (Egr2+/+) BMDMs to investigate whether EGR2 is necessary for the de novo and enhanced LPS-activated NF-κB-p65 binding in the alternatively polarized Macs. Both IL-4 priming-dependent de novo and enhanced NF-κB-p65 bindings were observed in Egr2+/+ BMDMs following LPS activation, but IL-4 failed to facilitate NF-κB-p65 binding in the absence of EGR2 (Figures 5A and 5B). Next, we examined the LPS-induced high-activation at the gene expression level in IL-4-primed and non-polarized Egr2+/+ and Egr2fl/fl BMDMs using RNA-seq. Our analysis revealed that the LPS response proved to be EGR2-independent in non-polarized Macs, while the LPS-induced high-activation was modified in the EGR2 deficient alternatively polarized Macs (Figure S5B). Specifically, the elevated LPS responsiveness of 469 genes was partially or entirely abolished in the IL-4-exposed EGR2 deficient Macs (Figure 5C; Table S3). In contrast, the mRNA expression of 16 genes was further increased in alternatively polarized Egr2fl/fl BMDMs following LPS activation (Figure S5C; Table S3). Four of the six previously characterized genes showed EGR2-dependent elevated LPS responsiveness in alternatively polarized Macs (Figure S5D). The LPS-induced secretion of the selected three chemokines showed a similar pattern to the mRNA expression. Specifically, EGR2 deficient alternatively polarized Macs secreted a significantly lower amount of CCL22, while their CCL2 and CCL17 secretion remained unchanged following LPS activation (Figure S5E). We have recently described that EGR2 controls the late epigenetic program of alternative Mac polarization by direct and indirect mechanisms (Daniel et al., 2020). Therefore, we examined EGR2 binding at the distal regulatory regions of 469 highly activated genes, which showed attenuated LPS responsiveness in the EGR2 deficient IL-4 primed Macs. We classified these genes into three subgroups based on the binding of EGR2 to either de novo or enhanced NF-κB-p65-peak-associated regulatory regions. We found 226 genes, including Il12a and Edn1, with a minimum of one de novo or enhanced NF-κB-p65 peaks associated with IL-4-induced EGR2 binding (Figures 5D–5F). Most of these genes were also annotated with de novo and enhanced NF-κB-p65 peaks, which did not overlap with EGR2 binding (Figures 5E and 5F). Additionally, 152 genes were associated with de novo or enhanced NF-κB-p65 binding at their regulatory regions without direct IL-4-induced EGR2 binding. The remaining 91 genes were not linked to de novo or enhanced NF-κB-p65 peaks using our criteria (Figure 5D). Finally, we investigated the regulatory role of EGR2 in the elevated enhancer activation at two selected EGR2-bound enhancers, including Il12a_−57Kb, and Edn1_−9Kb (Figure 5F). RT-qPCR-based measurement of eRNA expression demonstrated that the IL-4 priming and LPS stimulation-dependent synergistic activation of both enhancers requires the presence of EGR2 in BMDMs (Figure 5G). These results suggest that EGR2 contributes to the increased inflammatory responsiveness of alternatively polarized Macs and is required for many highly induced genes. To investigate the direct contribution of the EGR2 transcription factor to the elevated LPS responsiveness of IL-4-primed Macs, we further studied the EGR2-bound de novo and enhanced NF-κB-p65 binding-associated regulatory regions. First, we analyzed the de novo EGR motif enrichment in these regions. According to the IL-4-induced EGR2 binding, the specific EGR binding motif was found at 94% of EGR2-bound regulatory regions, indicating the direct DNA binding of EGR2 at these genomic sites (Figures S6A and S6B). Next, we investigated the effects of IL-4 priming on chromatin accessibility (by ATAC-seq), the enrichment of BRM (chromatin remodeling factor), PU.1 (LDTF), CEBPβ (LDTF), BRD4, RNAPII-pS2, and H3K27Ac at the EGR2-bound regulatory regions in WT and EGR2 deficient Macs. 24 h of IL-4 exposure led to elevated chromatin openness and BRM binding in an EGR2-dependent manner at the examined genomic regions (Figures 6A and 6B). Among the LDTFs, PU.1 binding was slightly induced, while CEBPβ binding was strongly induced by IL-4 priming in an EGR2-dependent fashion. Similarly, H3K27Ac, BRD4, and RNAPII-pS2 binding were induced in an IL-4 and EGR2-dependent manner (Figures 6A and 6B). To provide further direct evidence that IL-4-induced EGR2 binding to specific DNA sequences is necessary for extreme synergism, we investigated the single nucleotide polymorphisms (SNPs) at the EGR2-bound enhancers in three different mouse lines, including C57BL/6J (C57), SPRET/EiJ (SPRET), and BALB/cJ (BALBc). We first trained a deep neural network model based on the de novo and enhanced NF-κB-p65 peaks within cluster I and II using a strategy as described previously (Hoeksema et al., 2021) and then used DeepLIFT (Shrikumar et al., 2017) to interpret the importance of single nucleotides. Top k-mers based on DeepLIFT scores correspond to previously identified motifs, including NF-κB, EGR, AP-1, etc. (Figure S5A; Table S4). As this deep learning model classifies de novo and enhanced peaks without specifically considering known transcription factor recognition motifs, the finding of these motifs within the top-ranked nucleotides provides an independent line of evidence for their functional importance. SNPs overlaying with the top-ranked nucleotides are predicted to be functional. We detected 1–1 SNP at Tmco3_+6Kb, and Btg1_−10Kb enhancers in the SPRET mouse line, which were both predicted to be functional by DeepLIFT and decreased the affinity of the EGR binding motif. As measured by EGR2 ChIP-seq, the basal and IL-4-induced EGR2 binding was diminished at both enhancers in SPRET mice-derived BMDMs compared with the C57 and BALBc mice-derived Macs (Figure 6C). According to these findings, we observed the LPS-dependent de novo induction of Btg1 and Tmco3 expression in IL-4-primed WT C57 and BALBc mice-derived BMDMs, which was completely abolished in both EGR2 deficient C57 and SPRET mice-derived Macs (Figures 6D and 6E). These findings support the conclusion that EGR2 contributes to the IL-4 priming-induced epigenomic reprogramming, mediating the enhanced LPS responsiveness of a set of genes in alternatively polarized Macs. To broaden the scope of our studies we examined the IL-4 priming and LPS activation-induced gene expression changes in alveolar and large peritoneal Macs derived from embryonic precursors and thioglycolate-elicited small peritoneal Macs derived from blood monocytes (Ghosn et al., 2010; Epelman et al., 2014). Alternative Mac polarization markers Chil3 and Arg1 were induced in each Mac type by IL-4 (Figure S7A). By determining the mRNA expression of the six synergistically activated genes, we could observe elevated LPS responsiveness following the IL-4 priming in the studied Mac populations with minor Mac subtype-specific differences (Figure 7A). We concluded that extended synergy is not restricted to bone marrow and blood monocyte-derived Macs but can also be observed in distinct populations of embryonic precursor-derived tissue-resident Macs. Next, we assessed human CD14+ monocyte-derived differentiating Macs. We investigated the mRNA expression of the selected synergistically activated genes in human alternatively polarized Macs following LPS activation. As shown in Figure S7B, two genes, including IL12A and CCL22, showed elevated LPS responsiveness in the IL-4-primed human differentiating Macs indicating that the extended synergy is evolutionarily conserved. To determine whether the synergy between the alternative Mac polarizing and inflammatory signals is detectable in vivo, we studied a mouse model of alternative Mac polarization. Allergen-induced airway inflammation and asthma are associated with elevated Th2-type cytokine production, including IL-4 and alternative Mac polarization (Deng et al., 2019; Saradna et al., 2018; Robbe et al., 2015). We induced Th2-cell-type airway inflammation in sensitized animals using the clinically relevant ragweed pollen extract (RWE) challenge and characterized the dynamics of inflammation and alternative Mac polarization. As shown in Figure S7C, first, we sensitized the mice with two intraperitoneal RWE injections on days 0 and 4. Next, we induced Th2-cell-type airway inflammation with intranasal RWE treatment on day 11. We have characterized the developing Th2-cell-type inflammation, including Il4 and Th2 cell-type airway inflammation marker Muc5ac (Evans et al., 2015) RNA expression in total lung RNA extract and the immune cell composition of bronchoalveolar lavage fluids (BALFs), on day 11 before the RWE challenge (0 h) and 24 and 48 h after the RWE treatment. Il4 and Muc5ac expression did not show any differences between RWE and PBS-sensitized mice-derived lung mRNA extracts at 0 h. Still, the intranasal RWE stimulation significantly induced them at both examined time points (Figures S7D and S7E). The immune cell composition of the BALFs has also dynamically changed following the intranasal RWE stimulation. Before the intranasal RWE challenge (at 0 h), we could not observe any differences in the immune cell composition of the PBS and RWE-exposed mice-derived BALFs, and the alveolar Mac was the dominant immune cell type in both experimental groups (Figure S7E). However, the intranasal RWE treatment induced the recruitment of neutrophil and eosinophil granulocytes with different dynamics (Figure S7F) (Hosoki et al., 2016). Finally, we examined the mRNA expression of the alternative Mac polarization markers Arg1 and Chil3 in F4/80 and CD11c double-positive alveolar Macs derived from our model system. Each alternative polarization marker was induced following the intranasal RWE stimulation and showed the highest expression at 48 h (Figure S7G). Since alternative polarization of alveolar Macs is the most pronounced 48 h after intranasal RWE sensitization, we performed an LPS challenge intranasally at this time point, and isolated F4/80+ and CD11c+ alveolar Macs 6 h following treatment and studied synergistic gene activation in vivo (Figure 7B). We selected four synergistically activated genes for in vivo study of the identified phenomenon and measured their mRNA expression. In contrast to the in vitro results, the Ccl17 mRNA expression was already induced in alveolar Macs from PBS-treated and LPS-exposed mice. However, the RWE challenge further enhanced the LPS-dependent induction of Ccl17 (Figure 7C). In addition, Ccl22, Ccl2, and Edn1 mRNA expression were slightly enhanced following PBS treatment by LPS, and their LPS responsiveness was markedly elevated in RWE-pretreated mice-derived alveolar Macs (Figure 7C). After that, we determined the secreted CCL17, CCL22, CCL2, and EDN1 protein amounts in BALFs derived from PBS or RWE-challenged mice following intranasal LPS stimulation. We could detect the same highly synergistic expression patterns in the secreted protein levels in each case (Figure 7D). Elevated CCL2 expression in the lung leads to increased number of CD11b+ exudative Macs in BALFs, contributing to the immune pathology in different infections and injuries (Lin et al., 2008; Winter et al., 2007; Liang et al., 2012). Thus, we decided to investigate whether RWE challenge-induced Th2-cell-type airway inflammation can influence the LPS-dependent increase of CD11b+ exudative Mac content in BALFs. The exudative Mac content of the BALFs was negligible in the PBS and RWE-exposed mice, and its induction was observed 24 h after the intranasal LPS stimulation (Figure 7E). However, similarly to CCL2 production, the LPS-induced CD11b+ exudative Mac content was significantly elevated in the BALF-derived RWE-challenged mice (Figure 7E), indicating that synergistic gene activation in alveolar Macs leads to LPS-induced immunopathology in the lung. Looking at the physiological consequences of the interactions between the LPS-activated inflammatory and Th2-cell-type inflammatory pathways at the whole-body level, we found that LPS-induced transient body weight loss and hypothermia were significantly enhanced in the RWE-challenged mice, indicating that Th2-cell-type airway inflammation exacerbated the LPS-induced inflammatory response (Figures 7F and 7G). These findings indicate that the gene-specific enhanced LPS responsiveness of alternatively polarized Macs has pathophysiological consequences in vivo in a murine Th2 allergic airway inflammation model associated with elevated exudative Mac content in BALFs and exacerbated LPS-mediated inflammatory disease symptoms. Finally, clarified the role of EGR2 in regulating synergistic gene activation in alveolar Macs. We examined the IL-4-induced mRNA expression of canonical alternative Mac polarization markers Chil3 and Arg1 ex vivo in alveolar Macs isolated from Egr2+/+ Lyz2-cre and Egr2fl/fl Lyz2-cre mice. We found that the IL-4-dependent induction of Chil3 mRNA expression was partially diminished in EGR2 deficient alveolar Macs (Figure S7H). In contrast, IL-4-induced Arg1 expression was further enhanced in the Egr2fl/fl Lyz2-cre mice-derived alveolar Macs (Figure S7H), indicating the modified alternative polarization state in alveolar Macs in the absence of EGR2. Next, we wanted to investigate whether the IL-4 priming and LPS activation-mediated synergistic induction of Ccl17, Ccl22, End1, and Ccl2 expression is EGR2 dependent in the alveolar Macs ex vivo. Synergistic induction of Ccl2 was significantly increased in EGR2-deficient alveolar Macs, while the synergistic activation of the other three selected genes was not affected by the absence of EGR2 (Figure 7H). Finally, we examined the synergistic activation of Ccl2 mRNA expression in sorted alveolar Macs and the secreted protein content in the bronchoalveolar lavage fluid samples from Egr2fl/fl Lyz2-cre mice (Figure 7B). Similarly to our ex vivo alveolar Mac-derived data, in vivo synergistic activation of Ccl2 expression by RWE challenge and LPS treatment was significantly elevated at both mRNA expression and secreted protein levels in Egr2fl/fl Lyz2-cre mice compared with the control Egr2+/+ Lyz2-cre animals (Figures 7I and 7J). Overall, we find that EGR2 is a major modulator of synergistic gene activation in alveolar Macs ex vivo and in vivo, but distinct from its observed role in murine BMDMs. Understanding the nature of interactions between Mac polarization signals and pathogen-derived molecules is essential and will ultimately lead to exploitable insights into the development of differential immune responses and foster the development of novel therapeutic strategies. It is becoming clear that the innate immune responses of each individual are determined by (1) natural genetic variations (Fairfax et al., 2014; Lee et al., 2014) and (2) prior experience, disease state, and pathological or physiological exposures, including earlier infections or aging (DiNardo et al., 2021; Schultze and Aschenbrenner, 2021). Therefore, understanding the molecular underpinning of such interactions is critical not only at the organismal but also at the individual level. However, due to the complexity of the microenvironment, the molecular bases of these interactions are difficult, if not impossible to study in vivo. Thus, we investigated the interactions between isolated signaling events at the epigenomic and transcriptomic levels ex vivo by utilizing murine BMDMs, and then, tissue-resident Macs. We unraveled a highly synergistic interaction between alternative Mac polarizing signal IL-4 and inflammatory signal LPS at the epigenomic and transcriptional levels. We have termed this phenomenon extended synergism for the following reasons: (1) the transcriptional output of LPS activation is 5 to 100 times larger for a distinct gene set in IL-4-primed Macs vs. non-polarized cells, (2) LPS-activated NF-κB-p65 cistrome is expanded with more than 15,000 new NF-κB-p65-bound genomic regions in IL-4-exposed Macs, (3) the enhanced gene activation is linked to the significant increment in enhancer activation, chromatin accessibility, and cofactor binding, and (4) some of the resulting secreted cytokine amounts are orders of magnitude higher than the ones produced by either signal alone. Importantly, extended synergism is present in all Mac subtypes we examined including human primary Macs and is induced by at least nine different TLR activators, in a gene-specific manner. Here, we provided mechanistic insight into the IL-4 priming-enhanced gene subset-specific TLR4 response and the underlying IL-4-STAT6 signaling-directed epigenomic reprogramming. Integrative analysis of global transcriptional changes, chromatin structure, and NF-κB-p65 and RNAPII-pS2 cistromes showed that elevated inflammatory responsiveness is accompanied by the expansion of NF-κB-p65 cistrome, more accessible chromatin, and increased enhancer activation in alternatively polarized Macs. The identified highly activated genes and enhancers could be divided into two groups: (1) LPS activation was observed in non-polarized Macs, but IL-4 priming further enhanced it and (2) LPS responsiveness was only detected in the IL-4-exposed Macs, leading to de novo gene induction. The latter represented a qualitatively distinct biological response, similarly to our findings of repeated IL-4 exposure leading to the induction of a gene set not induced by the first stimulus (Daniel et al., 2018), illustrating the likely universality of the concept that repeated activating signals lead to the formation of an epigenomic memory and/or highly synergistic responses. Despite some Mac subtype-, or TLR-ligand-specific differences, our observations held and revealed that this phenomenon is universal and is not restricted to the bone marrow and/or monocyte-derived Macs or TLR4 activation, and evolutionarily conserved. However, the transcriptional regulatory mechanisms leading to the extended synergism show distinct gene and/or Mac subtype-specific differences. It appears that both STAT6 and NF-κB-p65 or to a lesser degree AP1 appear to be required components of extreme synergy at the enhancer selection and activation level. The role of EGR2 is more complex; (1) in BMDMs, IL-4-STAT6-induced EGR2 establishes and maintains the late, stable epigenomic and transcriptomic program of alternative Mac polarization (Daniel et al., 2020; Hoeksema et al., 2021); (2) however, in vivo, EGR2 is also required for the differentiation and specification of various tissue-resident Macs, including alveolar and MHCII+ serous cavity Macs (Bain et al., 2022; McCowan et al., 2021). Here, we showed that EGR2 is an important but not exclusive regulator of extended synergism in different Mac subsets. EGR2 regulates different aspects of synergistic gene activation in alveolar Macs—both ex vivo and in vivo—if compared with BMDMs; synergistic induction of Ccl2 expression is EGR2-independent in BMDMs, while it restrains it in alveolar Macs in the context of Th2 cell-type inflammation superimposed with LPS exposure. These differences might be explained by the reported, different expression patterns and/or distinct roles of EGR2 in BMDMs versus alveolar Macs. While EGR2 is expressed only in IL-4 polarized cells and contributes to alternative polarization of BMDMs but not to their differentiation per se (Daniel et al., 2020), it is constitutively expressed in mature alveolar Macs and essential for their proper maturation and the maintenance of their homeostatic functions (McCowan et al., 2021). Therefore, we propose that this difference might account for the distinct expression patterns of regulated genes and regulatory roles of EGR2 in the extended synergy that we observed. Specifically, the altered basal state and pre-mature or incomplete phenotype in the EGR2 deficient alveolar Macs may result in enhanced synergistic activation of the inflammatory genes such as Ccl2, while the failure of the IL-4 induced epigenetic reprogramming in the absence of EGR2 leads to the abolished synergistic activation of a specific gene subset in BMDMs. In addition, other transcription factors may compensate for the lack of EGR2 in alveolar Macs in synergistic activation of the EGR2-dependent genes such as Ccl22 and Edn1. Regarding the requirement for co-stimulation and the existence of epigenetic memory in extended synergism, we investigated the LPS-induced elevated responsiveness in BMDMs at different time points after IL-4 washout. Similarly, as reported by van den Bossche and colleagues (van den Bossche et al., 2016), we showed that the 24 h resting period between the IL-4 priming and LPS activation does reduce the observed respective regulatory interaction but does not completely eliminate it for some genes. Additionally, two gene-specific phenomena could be observed regarding transcriptional memory: (1) for many genes, including Ccl2 and Il12a, there is partially retained responsiveness; (2) for some genes, such as Ccl22 and Ccl17, the responsiveness to synergistic gene activation returns at 72 h following IL-4 washout. In principle, both patterns prove that there is no absolute requirement for simultaneous or co-stimulation, broadening the biological and clinical relevance of the discovered phenomenon. The transient, refractory period is quite unusual, and likely to be linked to some yet unknown, intermittent signaling or epigenomic event. However, recent work on allergen-induced inflammation-driven TNF-dependent innate memory supports our finding and the likelihood of transcriptional memory for regulating Ccl17 expression in Macs (Lechner et al., 2022). The allergic asthma is accompanied by the elevated production of Th2 cell-type cytokines IL-4 and IL-13 and alternative Mac polarization although the pathological relevance of the alternatively polarized Macs is not completely understood (Lambrecht and Hammad, 2015; Abdelaziz et al., 2020). Th2 cell-type airway inflammation and asthma severity are influenced by the lung microbiome and pathogen infections. Several viral and bacterial infections or the expansion of specific Gram-negative bacteria in the airway microbiome can cause asthma exacerbation often combining with glucocorticoid resistance (Goleva et al., 2013; Huang et al., 2011; Busse et al., 2010; Johnston et al., 2006; Lieberman et al., 2003; Talbot et al., 2005), while it has also been described that asthma can sensitize to pneumonia (Zaidi and Blakey, 2019). Although the molecular basis of the complex interactions between the bacterial pathogens and the allergens-activated Th2-cell-type immune response is not fully understood, there is a growing recognition that bacterial infections in pre-existing Th2-type inflammation can further potentiate Th2 cell-type inflammation and/or activate Th1 cell-type pathways resulting in a mixed immune activation and increased disease severity (Maltby et al., 2017). In line with these, we observed the development of complex inflammation as the consequence of the extended synergy; first, a wide range of immunomodulatory factors can be identified among the synergistically activated genes from the Th2-cell-mediated immune response-linked chemokines Ccl17, Ccl22, and Ccl24 through the potent vasoconstrictor Edn1 and the L-arginine transporter Slc7a2 to the classical inflammation-associated genes including Ccl2, Il6, and Nos2 in BMDMs; second, alveolar Macs also exhibit synergistic induction of Ccl17, Ccl22, Ccl2, and Edn1 both ex vivo and in vivo. Finally, RWE-induced Th2-cell-type airway inflammation leads to exacerbated TLR4 activation-mediated inflammation, resulting in well-defined symptoms (e.g., loss of body weight and hypothermia) at the organismal level. Nevertheless, exploring the precise pathogenic role of the extended synergism in the development of the exacerbated inflammatory response in asthmatic mice requires further investigation. Our work here provides insights into how IL-4 priming-induced epigenetic reprogramming leads to NF-κB-p65 cistrome expansion and gene and enhancer set-specific synergistic transcriptional activation in Macs following TLR4 activation. The described extreme synergy is inducible in Mac subtypes regardless of origin and by all TLR activation tested. Furthermore, the presented data raise the possibility that the alternatively polarized Macs have specific responsiveness to various pathogen-derived signals in vitro and in vivo that is not limited to the well-characterized transcriptional repression. The synergistically activated genes and resulting proteins but also the enhancers driving these processes can potentially be targeted to modulate alternative Mac polarization and TLR activation-associated complex pathological processes such as Th2 cell-mediated inflammation, severe asthma, or pneumonia. In this study, we used murine BMDMs and alveolar Macs to study the extended synergism. The inherent difference between the epigenome of these two cell types, in particular, the distinct role EGR2 plays, necessitates future work to investigate the basis of the Mac subtype-specific action of EGR2 and also to characterize the pathological consequences of extended synergy in vivo in other models. Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Laszlo Nagy (lnagy@jhmi.edu). This study did not generate new unique reagents. Sequencing data sets performed in this study are available at the NCBI GEO under accession numbers: GSE181223. Publicly available, published ChIP-seq data sets can be accessed on the following GEO accession numbers: GSE159630 and GSE151015. Female and male breeder mice for C57BL/6, BALBc and SPRET mice were used and bred under specific-pathogen-free (SPF) conditions. Egr2fl/fl Lyz2-cre and Stat6−/− animals were kept on the C57BL/6 genetic background. The Egr2fl/fl animals were a generous gift from Patrick Charnays laboratory. We crossed these animals with lysozyme-Cre (Lyz2-cre)+ animals to establish the conditional EGR2 deficient strain (Egr2fl/fl Lyz2-cre). These mice were backcrossed to the C57BL/6J strain for eight generations. As controls we used Egr2+/+ Lyz2-cre littermates. Full-body Stat6−/− (Jackson Laboratory) animals were maintained by breeding STAT6 deficient male and female mice. WT C57BL/6 animals were used as controls. Animals were handled according to the regulatory standards of the animal facilities of the University of Debrecen, Johns Hopkins All Children’s Hospital, and the University of California San Diego. Eight- to 12-week-old healthy female mice were used for all our experiments. Isolation and differentiation were completed as described earlier (Daniel et al., 2014). Isolated bone marrow-derived cells were differentiated for 6 days in the presence of L929 supernatant. Differentiated BMDMs were treated with IL-4 (20 ng/ml), LPS (100 ng/ml), PAM3CSK4 (100 ng/ml), HKLM (107/ml), polyI:C HMW (10 ug/ml), polyI:C LMW (10 ug/ml), FLA-ST (100 ng/ml), FSL-1 (100 ng/ml), ssRNA40 (1 ug/ml) and ODN1826 (5 uM) for the indicated period of time. Six- to 8-week-old male C57/B6 WT mice were used for these studies. Allergic airway inflammation was induced with endotoxin-free ragweed pollen extract (RWE, Greer Laboratories) as we previously described (Boldogh et al., 2005) with some modification. Briefly, animals were sensitized with two intraperitoneal (i.p.) administrations (on days 0 and 4) of 300 μg RWE in calcium and magnesium free Dulbecco’s phosphate-buffered saline (PBS, Sigma Chemical) injection combined in a 3 : 1 (75 μl : 25 μl) ratio with alum adjuvant (Thermo Fischer Scientific) or injected with the same volumes of PBS, as a vehicle control. On day 11, parallel groups of mice were challenged intranasally under ketamine and xylazine sedation with 240 μg RWE dissolved in 60 μl of phosphate-buffered saline or same volumes of PBS. On day 13, mice were challenged intranasally by 20 μg LPS (Salmonella enterica serotype minnesota Re 595-derived, Sigma-Aldrich) dissolved in 60 μl PBS or mice were challenged intranasally with same volumes of PBS. Animals were handled according to the regulatory standards of the animal facilities of the University of Debrecen. Animal studies were approved by the Animal Care and Protection Committee at the University of Debrecen (16/2019/DE MAB). RNA was isolated with Trizol reagent (Ambion). RNA was reverse transcribed with High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to manufacturer’s protocol. Transcript quantification was performed by qPCR reactions using SYBR green master mix (Roche). Transcript expressions were normalized to Ppia. Primer sequences are available in Table S5. The quality of total RNA samples was checked on Agilent BioAnalyzer using Eukaryotic Total RNA Nano Kit according to manufacturer’s protocol. Samples with RNA integrity number (RIN) value >7 were accepted for library preparation process. RNA-Seq libraries were prepared from total RNA using Ultra II RNA Sample Prep kit (New England BioLabs) according to the manufacturer’s protocol. Briefly, poly-A RNAs were captured by oligo-dT conjugated magnetic beads then the mRNAs were eluted and fragmented at 94-Celsius degree. First strand cDNA was generated by random priming reverse transcription and after second strand synthesis step double stranded cDNA was generated. After repairing ends, A-tailing and adapter ligation steps adapter ligated fragments were amplified in enrichment PCR and sequencing libraries were generated. Sequencing runs were executed on Illumina NextSeq 500 instrument using single-end 75 cycles sequencing. ChIP was performed with minor modifications of the previously described protocol (Daniel et al., 2014). We lowered the sonication strength to low and shearing was performed in two consecutive rounds of 5 min (total 10 min). ChIP-Seq libraries were prepared by using Ovation Ultralow System V2 (Nugen Technologies) reagent kit according to manufacturer’s protocol. Briefly, 1–5 ng of IP DNA sample was used for library preparation. End repair step was followed by A-tailing, adapter ligation and amplification steps. ChIP-Seq libraries were sequenced on Illumina NextSeq 500 instrument using single-end 75 cycles sequencing. The following antibodies were used: NFκB-p65 (sc-372), BRD4 (A301–985A100), RNAPII-pS2 (ab5095). Macrophages were scraped and counted to achieve 60,000 cells per condition. Cells were washed in ice-cold PBS and nuclear isolation was performed in the following lysis buffer: 1% Triton-X-100, 0.1% SDS, 150 mM NaCl, 1 mM EDTA pH 8.0, and 20 mM Tris-HCl pH 8.0. Nuclei were subjected to tagmentation using Tn5 transposase at 37C for 30 minutes in the following reaction: 12.5 μl TD buffer, 10.5 μl H2O, and 2 μl Tn5 transposase (Illumina). Tagmentation was stopped by the addition of 75ul TE buffer and DNA was isolated with the Qiagen Minelute columns according to the manufacturer’s protocol. Tagmented DNA was PCR amplified using Nextera primers, and the reaction clean-up was performed with AMPure XP beads by adding 90 μl beads to the 50 μl PCR reaction. Samples were subjected to Bioanalyzer fragment analysis and sequenced on the Illumina NextSeq 500 instrument. The collected cells were centrifuged at max RCF at 4°C for 15 minutes followed by the lysis of pellet in lysis buffer (50 mM TRIS, 1mM EDTA, 0.1 % MEA, 0,5% Triton X-100, 1 mM PMSF) containing a protease inhibitor cocktail (Sigma-Aldrich) with a 1:100x dilution ratio and sonicated with 4–5 strokes with 40% cycle intensity (Branson Sonifer, 450). The lysed cells were centrifuged at maximum rcf at 4°C for 15 min. The supernatant was collected for measurement of the protein concentration. The protein concentration was measured with the Pierce BCA Protein Assay Kit (Thermo Scientific) at a wavelength of 570 nm (Thermo Multiskan Ascent). Every sample was measured with 3 technical parallels, normalized using BSA standard (Thermo Scientific, Stock: 2 mg/mL). The samples were diluted up to 2 mg/mL concentration, mixed with equal volumes of 23SDS denaturation-buffer (0.125 M Tris-HCl, pH 6.8, containing 4% SDS, 20% glycerol, 10% MEA, 0.02% bromophenol blue dye), and incubated at 99°C for 10 min. Proteins were separated on 10 % SDS-polyacrylamide gels and blotted onto a PVDF membrane (MERCK-Millipore) using the semi-dry blotting method. Membranes were blocked with 5% non-fat dry milk or with 5% BSA in Tris-buffered saline and Tween 20 (TTBS) for 1 hour at room temperature. Primary antibodies were diluted in 0.5% milk or 5% BSA in 1xTTBS, with a dilution ratio of 1:1000–1:5000, incubated overnight at 4°C. The membranes were washed three times with 1xTTBS for 15 min at RT, incubated with horseradish peroxidase labelled secondary antibodies (Advansta) at a 1:10000–1:20000 dilution ratios for 1 hour at RT, followed by three times wash with 1xTTBS for 15 min at RT. The targeted protein bands were visualized using ECL-Kit (Advansta). For the investigation of nuclear translocation of NFκB-p65 800,000 BMDMs/well were seeded on round glass coverslips placed into 12 well plates. After the cells attached onto the coverslips, BMDMs were primed with IL-4 for 24 hours and activated with LPS for 1 1hour. Samples were washed twice with HBSS and fixed with methanol: acetone 1:1 for 10 min at room temperature, then cells were washed once with HBSS and incubated with FBS (Fetal Bovine Serum, Sigma) for 15 min at 37 °C to block the nonspecific binding sites. Cells were then rinsed with HBSS and incubated with 2 μg/ml primary anti-NFκB-p65 antibody (polyclonal rabbit IgG antibody, Santa Cruz Biotechnology Inc., Dallas, USA) for 1 h at 37 °C. After this incubation, cells were washed four times with HBSS and incubated with 5 μg/ml secondary antibody (Alexa Fluor 488 goat anti-rabbit IgG antibody, Invitrogen) for 1 h at 37 °C in the dark. Cells were washed four times again with HBSS and cell nuclei were stained with DAPI (283 nM) for 15 min at 37 °C. DAPI was removed from the samples by washing the specimens with HBSS and the round glass coverslips were glued onto microscope slides. Fluorescence microscopy measurements and analyses were carried out by a Zeiss Axioscope A1 (Jena, Germany) fluorescent microscope. The following filters were used to examine the samples: DAPI: excitation G 365 nm, emission BP 445/50 nm; fluorescein: excitation BP 470/40 nm, emission BP525/50 nm. The green fluorescence intensity of cell nuclei and cytosol were analysed by ZEN 2011 software (Zeiss, Jena, Germany) and the nuclear/cytosol intensity was calculated to express the extent of NFκB-p65 nuclear translocation. 24 hours after the last intranasal injection, measurements of the body temperature and the body weight of mice were started once a day for 6 or 7 days, respectively. Bronchoalveolar lavage (BAL) was performed at 6 hours after the last intranasal challenge. To collect BAL samples, animals were euthanized, and their tracheas were cannulated. Lavage was performed with 2 aliquots of 0.7 ml of ice-cold PBS (pH 7.3). The BAL cells were centrifuged at 400 g for 10 minutes at 4 °C and the supernatants were removed and stored aliquoted at −80 °C for further analysis. BAL cells were suspended in ACK buffer to lyse red blood cells for 2 minutes at room temperature and then adding 1 mL of MACS buffer. After washing step (at 800 rcf, 10 minutes, 4 °C), collected BAL cells were used for flow cytometry and cell sorting. In respective experiments, 2.5% thioglycolate was injected 3 days prior to cell isolation. Mice were euthanized and peritoneal cells were collected by washing the peritoneum with 7 mL PBS. Collected cells were filtered using a 100 mm filter cap. After centrifugation (350×g, 5 minutes at 4oC) and red blood cell lyses was performed with 2 mL ACK buffer for 2 minutes at room temperature. After washing step with 10 mL PBS buffer, single cell suspensions in MACS buffer were prepared for flow cytometry and fluorescence-activated cell sorting (FACS). BAL cells were labeled for anti-mouse CD11c-phycoerythrin ((PE), clone HL3, BD Biosciences) and anti-mouse F4/80-allophycocyanin ((APC), clone BM8, BioLegend) antibodies. Peritoneal cells were labelled for anti-mouse F4/80-APC and anti-mouse CD11b-PE-Cy7 (clone M1/70, eBioscience). The FcR Blocking Reagent (Miltenyi Biotec) was used to prevent non-specific bonding of antibody conjugates. To discriminate live and dead cells, the eBioscience™ Fixable Viability Dye eFluor™ 506 (Thermo Fischer Scientific) was used based on the manufacturer’s recommendation. The CD11c-F4/80 double positive alveolar macrophages, F4/80loCD11blo small peritoneal macrophages, and F4/80hiCD11bhi large peritoneal macrophages were sorted by FACSAria™ III (BD Biosciences). Approximately 15,000–25,000 cells were separated for transcript analysis. The flow cytometry analysis and cell sorting were performed by BD FACSAria III (BD Biosciences) using BD FACSDiva Software 6.0 (BD Biosciences). To determine total immune cell numbers of BAL samples, the CountBright™ Absolute Counting Beads (Thermo Fischer Scientific), Fc Receptor blocker (Miltenyi Biotec), anti-mouse F4/80-APC, anti-mouse CD11c-PE, and anti-mouse CD24-fluorescein-5-isothiocyanate (FITC) (clone M1/69, eBioscience) were used. The acquired flow cytometry data were analyzed with FlowJo v10.8 (BD Biosciences). Total cell counts in BALF were determined from an aliquot of the cell suspension. Differential cell counts were performed on cyto-centrifuge preparations stained with eosin and thiazine (ELITech Biomedical Systems: Red Stain Reagent, Blue Stain, Rinse, Aerofix Additive). The mRNA expressions of Muc5ac and Il-4 were analyzed by RT-qPCR measurements. RNA was isolated from a frozen mouse lung tissue with Trizol reagent. We cultured isolated cells at 24-well plates in a 0.5 mL RPMI medium containing 5% penicillin, 5% streptomycin, and 10% FBS. After attachment, we added 20 ng/ml IL-4 for 24 hours and then 100 ng/ml LPS for further 3 hours. total RNA was isolated by TRIZOL for mRNA transcripts detection. Human monocytes were isolated from peripheral blood mononuclear cells (PBMC) of healthy volunteers. Buffy coats were obtained from the Regional Blood Bank.Monocyte separation was carried out using CD14 MicroBeads (Miltenyi Biotec) according to the manufacturer’s instructions. Monocytes were cultured and differentiated to macrophages by their attachment to cell culture plate in RPMI 1640 supplemented with 10% FBS, 2 mM glutamine, penicillin and streptomycin for the indicated time (Daniel et al., 2020). Cells were pretreated with IL-4 (20 ng/ml) for 24 hours and activated by LPS (100 ng/ml) for 3 hours. Culture supernatants from cells were collected at the indicated times. Samples were centrifuged at 1000g for 10 min at 4°C, and the supernatants were separated and stored at −20°C until analysis. These cell culture and BAL supernatants were then probed for the presence of the following cytokine using ELISA kits according to the manufacturer’s instructions: EDN1 (R&D Systems), CCL2/MCP-1 (BioLegend), CCL17/TARC (R&D Systems) and CCL22/MDC (R&D Systems). Plates were read using the BIO-TEK Synergy HT Multi-Detection Microplate Readermicroplate reader. The single-ended mRNA sequence reads were analyzed using the nf-core/rnaseq v3.0 pipeline (Ewels et al., 2020). Sequencing quality was evaluated by FastQC software and reads were aligned to the mm10 (GRCm38) genome assembly with STAR (Dobin et al., 2013) default parameters. Genes were quantified using Salmon (Patro et al., 2017). Genes with at least 10 Counts Per Million mapped read (CPM) were considered expressed. Multidimensional scaling (MDS) analysis was used to identify broad library-wise trends using the cmdscale function in R. Statistically significant difference was considered FDR < 0.1 from GLM test using R package edgeR (Robinson et al., 2010). Annotation of enhancers to differentially expressed genes was based on linear proximity (+/− 100 kilobase) to the transcription start sites (TSSs). Coverage profiles represent Reads Per Kilobase Million (RPKM) values, calculated using deeptools2 bamCoverage (Ramirez et al., 2016) and visualized in IGV (Robinson et al., 2011). Tables S2 and S3 contain the synergistically expressed genes with additional statistics. Sequencing quality was evaluated by FastQC software. The primary analysis of ATAC-seq was carried out using nf-core/atacseq pipeline. ChIP-seq read alignment and filtering were carried out using our ChIP-seq command line pipeline (Barta, 2011). Reads were mapped to the mouse reference genome (mm10) using the default parameters of BWA MEM aligner (Li and Durbin, 2009). Low mapping quality reads (MAPQ < 10), reads mapping to ENCODE mouse blacklisted regions (Amemiya et al., 2019) and duplicated reads were discarded from the downstream analyses, using bedtools intersectBed (Quinlan and Hall, 2010) and samtools rmdup (Li et al., 2009). MACS2 (Zhang et al., 2008) was used to call peaks at 5% false discovery rate (FDR). Coverage profiles represent Reads Per Kilobase Million (RPKM) values, calculated using deeptools2 bamCoverage (Ramirez et al., 2016) and visualized in IGV (Robinson et al., 2011). ATAC-seq and ChIP-seq read density profiles for region set summits were calculated using deeptools2 computeMatrix (-a 1500 -b 1500 -binSize 50) and plotted with plotHeatmap function. Violin plots represents the mean RPKM of replicates. Genomic annotation of region sets was calculated using HOMER (Heinz et al., 2010) and plotted in R. To determine differentially bound NFκB-p65 genomic regions, we first used DiffBind (Ross-Innes et al., 2012) to generate a consensus peak set with minOverlap=2 for all samples. DESeq2 (Love et al., 2014) was used to determine differentially bound sites at 10% false discovery rate (FDR). Library-wise similarities were determined using hierarchical clustering of the normalized values using the hclust function in R. ″De novo″ and ″enhanced″ NFκB-p65 sites were classified based on the presence of a called peak in the LPS-treated samples. IL-4 pretreatment facilitated NFκB-p65 sites (n=26440) were further classified based on their RNAPII activity patterns. For this, RNAPII reads were counted, low occupancy sites were removed and k-means clustering (k=5) was applied using pheatmap in R. Table S1 contains the detailed information for the NFkB-p65 consensus with additional statistics and annotation. We used HOMER’s findMotifsGenome.pl script to search for de novo motif enrichments with mm10 -size given -dumpFasta -bits -homer2 parameters. The resulting EGR position weight matrices (PWMs) were used as an input for annotatePeaks.pl (mm10 -size 1200 -hist 20) to calculate motif enrichment scores and histograms. EGR motif occurrences in regions were calculated using FIMO (Grant et al., 2011) in R using the MEME Suite wrapper memes runFimo function with p < 0.001 cutoff. Data were visualized in R with ggplot2. To predict potential SNPs that alter enhancer functions, we trained a deep neural network model as described previously (Hoeksema et al., 2021). In brief, we adapted a strategy of AgentBind (Zheng et al., 2021) by fine-tuning a pre-trained DeepSEA model (Zhou and Troyanskaya, 2015) on all the ″de novo″ and ″enhanced″ NFκB-p65 peaks within Cluster I and II. Next, we used DeepLIFT (Shrikumar et al., 2017) to interpret the importance of single nucleotides. For each input sequence, we generated two sets of importance scores, one for the original sequence and the other for its reverse complement. The final DeepLIFT scores were displayed as the absolute maximum importance score at each aligned position. We defined top-scoring variants by their overlay with the top 20% (i.e., top 60) nucleotides of each 300-bp region. To interpret the important sequence patterns learned by neural networks, we computed the odds ratio of each 5-mer within the top 10% of all 5-mers (Zheng et al., 2021) and used Fisher’s exact test to determine significant enrichment. For enriched 5-mers, we used TOMTOM (Gupta et al., 2007) to match with known transcription factor binding motifs from the JASPAR database (Fornes et al., 2020). The position weight matrix (PWM) of EGR2 motif was downloaded from JASPAR database with motif ID MA0472.1 (Fornes et al., 2020) and was used to compute motif score or PWM score and motif score difference between strain genomes. Motif matches were identified with a motif score cutoff at false positive rate < 0.1%. Statistical analysis for qRT-PCR and ELISA methods: the error bars represent standard deviation (SD). The two-tailed Student’s t test was used to evaluate the significance of differences between two groups. Quantification and alignments of NGS analysis for RNA-seq, ChIP-seq, and ATAC-seq are also described in more detail in the methods section above.

PMC9649902

Shumin Yang,Ji Luo,Yingying Chen,Rui Wu,Huazhen Liu,Zutao Zhou,Muhammad Akhtar,Yuncai Xiao,Deshi Shi

A buffalo rumen-derived probiotic (SN-6) could effectively increase simmental growth performance by regulating fecal microbiota and metabolism

28-10-2022

Bacillus,fiber decomposition,production performance,fecal microbiota,metabolomic

Microorganisms play a key role in ruminal digestion, some of which can be used as probiotics to promote growth in ruminants. However, which potential bacteria are responsible for ruminant growth and how they potentiate the basic mechanism is unclear. In this study, three bacterial strains, Bacillus pumilus (SN-3), Bacillus paralicheniformis (SN-6), and Bacillus altitudinis (SN-20) with multiple digestive enzymes were isolated from the rumen of healthy buffaloes. Among these strains, SN-6 secreted cellulase, laccase, and amylase, and significantly inhibited Staphylococcus aureus ATCC25923 and Escherichia coli K99 in vitro. In addition, SN-6 exhibited strong tolerance to artificial gastric juice, intestinal juice, and high temperature. Antibiotic resistance test, virulence gene test, and mouse toxicity test confirmed the safety of SN-6. Further, SN-6 significantly increased the body weight (p < 0.01), affects the intestinal microbiota structure, and alters the metabolomic patterns of Simmental. There was a remarkable difference in the β diversity of fecal microflora between SN-6 and control groups (p < 0.05). Furthermore, SN-6 significantly increased the abundance of Clostridium_sensu_stricto_1, Bifidobacterium, Blautia, and Cellulolyticum, decreased the relative abundance of Monoglobus and norank_f_Ruminococcacea. Moreover, SN-6 feeding significantly enriched intestinal metabolites (i.e., 3-indoleacrylic acid, kynurenic acid) to maintain intestinal homeostasis. Finally, the microbial and metabolic functional analysis indicated that SN-6 could enhance amino acid metabolism (mainly tryptophan metabolism) and lipid metabolism pathways. Overall, these findings indicated that SN-6 could be used as a probiotic in ruminants.

A buffalo rumen-derived probiotic (SN-6) could effectively increase simmental growth performance by regulating fecal microbiota and metabolism Microorganisms play a key role in ruminal digestion, some of which can be used as probiotics to promote growth in ruminants. However, which potential bacteria are responsible for ruminant growth and how they potentiate the basic mechanism is unclear. In this study, three bacterial strains, Bacillus pumilus (SN-3), Bacillus paralicheniformis (SN-6), and Bacillus altitudinis (SN-20) with multiple digestive enzymes were isolated from the rumen of healthy buffaloes. Among these strains, SN-6 secreted cellulase, laccase, and amylase, and significantly inhibited Staphylococcus aureus ATCC25923 and Escherichia coli K99 in vitro. In addition, SN-6 exhibited strong tolerance to artificial gastric juice, intestinal juice, and high temperature. Antibiotic resistance test, virulence gene test, and mouse toxicity test confirmed the safety of SN-6. Further, SN-6 significantly increased the body weight (p < 0.01), affects the intestinal microbiota structure, and alters the metabolomic patterns of Simmental. There was a remarkable difference in the β diversity of fecal microflora between SN-6 and control groups (p < 0.05). Furthermore, SN-6 significantly increased the abundance of Clostridium_sensu_stricto_1, Bifidobacterium, Blautia, and Cellulolyticum, decreased the relative abundance of Monoglobus and norank_f_Ruminococcacea. Moreover, SN-6 feeding significantly enriched intestinal metabolites (i.e., 3-indoleacrylic acid, kynurenic acid) to maintain intestinal homeostasis. Finally, the microbial and metabolic functional analysis indicated that SN-6 could enhance amino acid metabolism (mainly tryptophan metabolism) and lipid metabolism pathways. Overall, these findings indicated that SN-6 could be used as a probiotic in ruminants. Ruminants play a significant role in our society because they were domesticated more than 10,000 years ago. They uniquely use a variety of digestive enzymes and digest the most complex polysaccharides, which are undigestible by the human digestive tract (Mizrahi et al., 2021). Such enzymes are predominantly produced by microbes of ruminants and are responsible for breaking down plant fibers (Russell and Rychlik, 2001; Jose et al., 2017). It is interesting to note that the microbiota in the gut of ruminants is diverse and abundant. That is why, the rumen, an important foregut fermenter with a strong capacity to digest plant feed, is strictly dependent on a complex array of gut microbes for its physiological and biochemical responses. These complex rumen microorganisms degrade plant fibers in ruminant roughage by expressing and secreting various digestive enzymes, i.e., cellulase, protease, amylase, etc. (Weimer, 1996). To promote host growth, these bacteria release energy stored in complex plant carbohydrates (Flint and Bayer, 2008), by converting them into short-chain fatty acids, vitamins, and other compounds. Meanwhile, the bacterial-released protein is also an important protein source for ruminants (Flint, 1997; Mizrahi et al., 2013). In addition, microorganisms secrete a large number of antagonistic factors (e.g., hydrogen peroxide, bacteriocins, diacetyl, etc.) that have a significant inhibitory effect on a wide range of bacteria (Cox and Dalloul, 2015). These antagonistic factors protect the organisms from various pathogenic bacteria and also reduce the colonization of potentially pathogenic bacteria. Subsequently, they help to maintain host health and normal physiological functions throughout the life cycle. Therefore, the development of beneficial rumen-derived microorganisms is of great significance to promote the development of the ruminant industry. The ruminant gut microbiota is rich in microorganisms and the interactions between these microorganisms are complex and crucial to the host’s health. A growing body of research has highlighted that the gut microbiota and its metabolic activities with the host are essential in understanding nutrition and metabolism (Del Chierico et al., 2018; Valdes et al., 2018), of which the role of probiotics has, indeed, been emphasized. Additionally, it has been discovered that gut microbiota’s metabolic reactions help the body in nutrient absorption from the diet and transform them into a range of secondary metabolites to maintain gut health. The beneficial microorganisms ultimately cause widespread changes in intestinal metabolites, which in turn maintain the equilibrium in the intestinal metabolic microenvironment by carrying out a variety of metabolic functions in the gut. Additionally, these metabolites could also be utilized by the microorganisms for their proliferation (Foroutan et al., 2020). Host physiological activity supports intestinal homeostasis by lowering the amount of tryptophan and promoting indole derivatives that activate aryl hydrocarbon receptors (Williams et al., 2014). Conversely, toxins produced by intestinal microbes could potentially impact intestinal epithelial cells and result in intestinal injury (Kim et al., 2014). Numerous studies found that microorganisms have been increasingly used as feed additives in ruminants. They achieve this by stimulating a shift in the harmful gut microbiota toward a healthier microbiota, improving feed utilization and daily weight gain (Marsalková et al., 2004; Sun et al., 2010). They promote in vitro fermentation and fiber degradation microbiota (Izuddin et al., 2018), affect body metabolism (Liu et al., 2017; Kim et al., 2018), and help to build better immune status. Moreover, they improve intestinal health (Izuddin et al., 2019) and even prevent diseases (Larsen et al., 2014). Hence, we hypothesized that ruminant probiotics achieve weight gain via strengthening intestinal microbiota/metabolites and based on intestinal microecology. Until now, many swine and poultry studies have demonstrated that probiotics help to develop healthy microflora (mainly probiotics), which prevent pathogen adhesion and invasion of intestinal epithelial cells, induce the production of antibacterial compounds, maintain epithelial barrier integrity, and regulate metabolism and immune system (Wang et al., 2019; Šikić Pogačar et al., 2020; Chance et al., 2021). However, there is little evidence that ruminant-derived probiotics regulate gut microbial composition and thus affect metabolism to promote beef cattle body weight. Furthermore, most previous studies used a combination of strains (some not ruminant-derived) and primarily focused on the synergy of strains rather than the mechanism of a specific strain. This study aimed to isolate strains with fiber degradation potential and antibacterial ability from the buffalo rumen, evaluate their growth-promoting effects, investigate their influence on fecal flora and host metabolism, and analyze the impact of changes in intestinal flora on host metabolism. The gut microbiota, microbial metabolism, and potential probiotic effects were also investigated with possible mechanisms. In this study, the addition of a single probiotic seems a precise intervention, providing a meaningful reference for probiotic development in ruminants. The modified cellulase identification medium (CMC-Na medium) was used as an isolation medium (Supplementary Table 1). The rumen fluid was obtained from the rumen of healthy buffaloes with rumen fistulas and filtered by four layers of sterilized gauze. The rumen fluid was diluted with sterilized double-distilled water, evenly coated on a CMC-Na medium, and then cultured in an anaerobic container at 39°C for 3 days. After that, the single colonies were selected and cultured on the CMC-Na medium. The culture mediums were dyed with 0.1% Congo red (Solarbio, China) staining solution to observe whether there were light yellow hydrolysis circles around the coating (Teather and Wood, 1982). The strains producing hydrolytic circles were selected for purification and subculture. The isolated strains were confirmed and identified by genetic analysis using PCR and 16S rRNA sequencing for further verification. The genomic DNA was extracted with the bacterial genome DNA fast extraction kit (Aidlab Biotech Co., Ltd., China) according to the manufacturer’s protocol. Universal PCR primers 27F (5′-AGAGTTTTGATCCTGGCTCAG-3′) and 1492R (5′-GGTTACCTTGTTACGCACTT-3′) were used to amplify the 16S rRNA gene. PCR products were sequenced by Sangon Biotech Co., Ltd. (Shanghai, China). The sequencing results were analyzed using Basic Local Alignment Search Tool (BLAST) on the NCBI website. The phylogenetic tree of bacteria was constructed by the neighbor-joining method using MEGA7.0 software. The phylogenetic tree was statistically evaluated using 1,000 bootstrap replicates. Potato dextrose agar (PDA) medium-guaiacol (0.04% guaiacol), PDA-aniline blue medium (0.1 g/L aniline blue), Luria–Bertani (LB) plate (1% soluble starch), and an LB plate (1% skimmed milk) were, respectively, used to detect the laccase (Lac), manganese peroxidase (Mnp), lignin peroxidase (Lip), amylase, and protease in the strains. The antibacterial activity of the isolates was determined by the Oxford cup method (Bian et al., 2016). Escherichia coli O157, O139, K88, K99, Salmonella C78-1, and Staphylococcus aureus ATCC25923 were used as an indicator at 1.0 × 107 CFU/ml. These indicator bacteria were obtained from the State Key Laboratory of Agriculture Microbiology of Huazhong Agricultural University. In a super clean bench, the bacterial solution of the indicator bacteria (Escherichia coli O157, O139, K88, K99, Salmonella C78-1, and Staphylococcus aureus ATCC25923) was evenly coated onto solid LB plates, respectively. Sterilized Oxford cups (small round tubes with an inner diameter of 6 nm, an outer diameter of 8 nm, and a height of 10 nm) were then placed in the LB plates so that they were in contact with the LB plates without gaps. A 200 μL of SN-6 bacterial solution was added to the Oxford cup. The size of the inhibition zone was measured with a vernier caliper after overnight incubation. The bacterial liquid (2.4 × 109 CFU/ml) in the logarithmic growth phase was placed in a water bath at 70 and 90°C, respectively. Samples were taken at 3 and 10 min time points to count the viable bacteria in the samples. Artificial gastric juice and intestinal juice (Yuanye Biotechnology Co., Ltd., Shanghai, China) were prepared according to the Chinese Pharmacopeia (D’Aldebert et al., 2009). The bacterial liquid (2.4 × 109 CFU/ml) in the logarithmic growth phase was inoculated into artificial gastric juice (pH = 3.0) and artificial intestinal fluid (pH = 7.0) with 1% inoculation amount. Samples were taken at 3 and 4 h, respectively. Finally, the viable bacteria in the samples were counted. The survival rate was calculated as follows: survival rate = [C/C0] × 100%. Here, C and C0 represented the number of colonies in the experimental and control groups, respectively. The drug sensitivity of isolated strains was tested with the disk diffusion method (Ghosh et al., 2015). Fifteen drug tablets (Hangzhou microbial Reagent Co., Ltd., China) were selected. The drug sensitivity detection was performed according to the latest version of the CLSI standard (Institute and Laboratory) (CLSI, 2018). Bacillus cereus, which contains nheA, nheB, nheC, and entFM genes was used as the positive control strain. The specific synthesized primers of virulence genes were obtained from Sangon Biotech Co., Ltd. (Shanghai, China). The amplification program was as follows: pre-denaturation at 94°C for 3 min; 35 cycles (95°C 3 min, 58°C 30 s, 72°C 33 s); and extended for 10 min at 72°C (Rowan et al., 2003). All animal experiments were approved and reviewed by the animal welfare and research department, ethics committee, Huazhong Agricultural University, Wuhan, China (Approval number: HZAUMO-2019-047). Twenty-three-week-old KM (Kun Ming) mice (an equal number of male and female subjects) were randomly divided into the experimental group and control group (n = 10). The mice in the experimental group were given SN-6 by gavage at 2.0 × 108 CFU/day for 2 weeks, while mice in the control group were given the same volume of saline. Behaviors, hair loss, mental state, and general health of the reared mice were observed throughout 2 weeks. After 2 weeks, the mice were sacrificed using chloral hydrate as anesthesia, and the heart, spleen, liver, lung, and kidney were collected to detect organ index. T-test was used to analyze the data. p < 0.05 was considered statistically significant. Five-six-month-old healthy Simmental beef cattle (female) with the same genetic background and similar initial weight from Hubei Liangyou Jinniu animal husbandry technology Co., Ltd. (China, Hubei) were selected. The initial weights of Simmental are shown in Supplementary Table 2. Cattle (n = 66) were randomly divided into control group (n = 33) and experimental group (n = 33). There was no significant difference in the initial weight of Simmental between the two groups (p > 0.05). The feeding lasted for 33 days. To assess the long-term sustained effects of SN-6, we fed Simmental beef cattle with SN-6 for an additional 28 days. Both the control cattle and experimental cattle were fed with the basic diet (Supplementary Table 3) ad libitum during the experiment. The control cattle were given normal water, whereas, the experimental cattle were given water that contained SN-6 at 1.0 × 1010 CFU/day/individual. Before and at the end of the experiment, the cows were weighed at fasting. Data are expressed with mean ± SD, analyzed by one-way analysis of variance using SPSS 21.0 software, and p < 0.05 was considered statistically significant. Fresh fecal samples were collected from the rectum with sterile gloves at the end of the experiment and immediately stored in sterile centrifuge tubes. All samples were immediately frozen on dry ice and stored at –80°C for further analysis. Total DNA was extracted from fecal samples using an E.Z.N.A.® soil Kit (Omega Bio-Tek, Norcross, GA, United States). The extracted DNA was qualitatively and quantitatively detected by 1% agarose gel electrophoresis and NanoDrop 2000 UV-vis spectrophotometer (Thermo Scientific, Wilmington, NC, USA). The V3–V4 region of 16S rRNA was amplified by PCR with specific primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACHTACHVGGGTWTCTAAT-3′) (PCR instrument: GeneAmp 9700, ABI, USA). The PCR products were recovered by 2% agarose gel and purified by AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA). QuantiFluor™-ST (Promega, USA) was used for quantitative analysis. The fecal microbial DNA fragments were sequenced by the Illumina Miseq platform (Illumina, San Diego, CA, USA). The quality control and splicing of the original data were carried out by using Trimmomatic and Flash software. After quality control, the sequences and fuzzy bases less than 50 bp were removed. UPARSE software (version 7.1) was used to cluster the optimized sequences according to 97% similarity; UCHIME software was used to remove chimeras. The taxonomy of each 16S rRNA gene sequence was analyzed by the RDP Classifier algorithm against the Silva (SSU123) 16S rRNA database using a confidence threshold of 70%. Chao1, ACE, Shannon, and Simpson indices were used to reflect α diversity. The core fecal microbiota of each group was shown by the Venn diagram. In β diversity analysis, principal coordinate analysis (PCoA) was used to determine the difference in species composition among samples. According to the composition and sequence distribution of samples at each taxonomic level, the differences in species abundance between groups were compared and tested by the Student t-test. The p < 0.05 was considered to be statistically significant. Microbial biomarkers associated with particular interventions were identified through linear discriminant analysis (LDA) effect size (LEfSe), with an effect size threshold of 3. The effects of SN-6 on the fecal metabolism in Simmental were assayed by LC-MS-based untargeted metabolomics. Fecal samples (50 mg) were accurately weighed, and the metabolites were extracted using a 400 μL of methanol: water (4:1, v/v) solution. The mixture was allowed to settle at –20°C and treated with high throughput tissue crusher Wonbio-96c (Shanghai Wanbo Biotechnology Co., Ltd.) at 50 Hz for 6 min, followed by vortexing for 30 s and ultrasound treatment at 40 kHz for 30 min at 5°C. The samples were placed at –20°C for 30 min to precipitate proteins. After centrifugation at 13,000 g at 4°C for 15 min, the supernatants were transferred to sample vials for LC-MS/MS analysis. Ultra high performance liquid chromatography-mass spectrum (UHPLC-MS) analyses were performed using a Vanquish UHPLC system (Thermo Fisher, Germany) coupled with an Orbitrap Q Exactive™HF-X mass spectrometer (Thermo Fisher, Germany). Samples were injected onto a Hypesil Gold C18 column (100 mm × 2.1 mm, 1.9 μm; Thermo Fisher, Germany) using a 17-min linear gradient at a flow rate of 0.2 ml/min, and the column temperature was maintained at 40°C. The eluents for the positive polarity mode were eluent A (0.1% formic acid in water) and eluent B (Methanol). The eluents for the negative polarity mode were eluent A (5 mM ammonium acetate, pH 9.0) and eluent B (Methanol). The solvent gradient was set as follows: 2% B, 1.5 min; 2–100% B, 12.0 min; 100% B, 14.0 min; 100–2% B, 14.1 min; 2% B, 17 min. Q Exactive™HF-X mass spectrometer via electrospray ionization (ESI) interface was operated in positive/negative polarity mode with a spray voltage of 3.2 kV and capillary temperature of 320°C, sheath gas flow rate of 40 arb, and aux gas flow rate of 10 arb. All results were presented as mean ± standard deviation (SD). A multivariate statistical analysis was performed using ropls (Version 1.6.2) R package from Bioconductor on Majorbio Cloud Platform. To obtain an overview of the metabolic data, an unsupervised method of principle component analysis (PCA) was used, and general clustering, trends, or outliers were visualized. Orthogonal partial least squares discriminate analysis (OPLS-DA) was used for statistical analysis to determine global metabolic changes between comparable groups. Variable importance in the projection (VIP) was calculated in the OPLS-DA model. The p-values were estimated with paired Student’s t-test on single-dimensional statistical analysis. The correlations between the key fecal microbiota and fecal metabolites were assessed by the Spearman’s correlation coefficient and were visualized on a heat map generated by the Python software (Version1.0.0). To clarify the changes in metabolic pathways caused by SN-6 interventions, we characterized potential pathway enrichment analysis using the KEGG pathway. Based on the light yellow hydrolysis circle given on the CMC-Na agar plate, three strains named SN-3, SN-6, and SN-20 were isolated (Figure 1A), and all of them were gram-positive bacteria (Figure 1B). According to the enzyme activity (EA) value (Table 1), the degradation capacity of fiber was SN-3 = SN-20 > SN-6. The 16S rDNA sequence analysis indicated that SN-3 had 99.79% homology to Bacillus pumilus, SN-6 had 99.65% homology to Bacillus paralicheniformis, and SN-20 had 99.79% homology to Bacillus altitudinis. These results demonstrated that SN-3, SN-6, and SN-20 were Bacillus pumilus, Bacillus paralicheniformis, and Bacillus altitudinis, respectively. All three strains could produce amylase, among which SN-6 is the best amylase producer. Both SN-3 and SN-20 could produce protease, and SN-3 was comparatively better than SN-20 (Table 2). Only SN-6 exhibited a reddish-brown oxidation circle on the PDA-guaiacol plate (Supplementary Figure 1), which indicated that SN-6 secreted laccase. None of the three strains could discolor PDA-aniline blue, suggesting that none of them produced Mnp and Lip (data not shown). SN-6 robustly inhibited K99 and S. aureus growth. The inhibition zone diameter was 22.0 mm for E. coli K99 and 24.0 mm for S. aureus. Moreover, SN-3 and SN-20 did not obviously inhibit six indicator bacteria (Supplementary Figure 2). A probiotic with a higher tolerance capability and a good survival rate is often preferred. We found that the survival rate of SN-6 was 91.7% at 70°C for 3 min, and 58.3% at 90°C for 10 min. After culturing SN-6 in simulated gastric juice (pH 3.0) or in neutral simulated intestinal fluid for 4 h, its survival rate was 36.36 and 54.29%, respectively. These results indicated the good survivability of SN-6 in harsh environments. SN-6 was sensitive to antibiotics used in this study except for oxacillin and ceftazidime (Table 3). Enterotoxin-related virulence genes nheA, nheB, nheC, and entFM were amplified in the positive strain (Supplementary Figure 3), while no enterotoxin-related virulence gene was detected in SN-6 (Supplementary Figure 4). It was also observed that the mice, both in the control group and the experimental group, were normal. There was no significant difference in body weight (Figure 2A) and organ index between the two groups (Figure 2B). After 33 days of feeding, Simmental cattle’s body weight in the control and experimental groups was 292.53 and 295.86 kg, respectively. After 61 days of feeding, the body weight of Simmental cattle in the control and experimental groups was 329.48 and 335.62 kg, respectively. It was noted that the SN-6 feeding increased body weight by approximately 3.33 kg/individual (33 days), and 6.14 kg/individual (61 days) compared with the control (p < 0.01) (Figure 3). A total of 475,965 optimized sequences of 16S rRNA of bacteria in 12 fecal samples (six in each group) were obtained. According to 97% sequence similarity, the optimized sequences were clustered by operational taxonomic units (OTU) and 1,338 OTU sequences were obtained. There was no significant difference in the α diversity index between the SN-6 group and the control group (p > 0.05), indicating that SN-6 feeding did not change the fecal flora richness and diversity (Supplementary Figure 5). Likewise, the beta diversity was assessed by principal coordinate analysis (PCoA) based on the Bray–Curtis distance, which was used to study the similarity or differences in sample community composition. As shown in Figure 4A, SN-6 significantly changed the overall community composition of fecal flora compared with the control (p < 0.05), which indicated that the microorganisms in the SN-6 group had distinct clustering. At the phylum level, the fecal microbiota composition of each group is shown in Figure 4B. Firmicutes and Bacteroidetes are the core bacteria with high abundance in ruminants both in the SN-6 group and control group. However, SN-6 feeding increases the abundance of Actinobacteria, which might be due to the increase of Bifidobacterium. At the genus level, significantly increased abundances of Clostridium_sensu_stricto_1, Bifidobacterium, Blautia, and Cellulolyticum were observed in the SN-6 group (*p < 0.05, ***p < 0.001, respectively), while significantly decreased abundances of Monoglobus, norank_f_Ruminococcacea (*p < 0.05) (Figure 4C) were observed. Using LDA and LEfSe analyses for microbial biomarker discovery in two groups, Clostridiaceae, Bifidobacteriaceae, and Lachnospiracese were found enriched in the SN-6 group, while Stackebrandtia and Monoglobus were enriched in the control group (Figure 4D). To get a holistic view of the host metabolism after SN-6 intervention, we used non-targeted metabolomics to identify key metabolites and metabolic pathways that might be altered in the Simmental intestine. A total of 185 metabolites were identified in feces. The OPLS-DA score scatter plots revealed a visible separation between the control and SN-6 groups in positive ion mode (R2Y: 0.982, Q2: 0.635) (Figure 5A). The results of 200 permutations exhibited no over-fitting in OPLS-DA models (Figure 5B). Twenty-six metabolites were found in the fecal sample which met the conditions of p < 0.05 and variable importance in project (VIP) > 1 between the control and SN-6 groups (Figure 5C). The effect of SN-6 on the regulation of some differential metabolites (including 3-indoleacrylic acid, 5-hydroxyindole-3-acetic acid, methyl indole-3-acetate, N-acetyl-D-tryptophan, oleic acid, D-mannose, vitamin A, and kynurenic acid) in the feces is shown in Figure 5D. As shown in Figure 6A, differential metabolites related to different metabolic pathways were mapped. Six potential metabolic pathways were screened according to impact value > 0.1 and p < 0.05. Retinol metabolism (0.38), tryptophan metabolism (0.17), steroid hormone biosynthesis (0.16), nicotinate and nicotinamide metabolism (0.13), pyrimidine metabolism (0.11), and steroid degradation (0.11) are listed in descending order of impact value (Figure 6B). Among them, tryptophan metabolism covered the main differential metabolites, indicating that this pathway might play a vital role in the growth promotion of SN-6. To comprehensively analyze the relations between fecal metabolites and gut microbiota, weight gain associated with altered metabolites was explored. Spearman’s correlation analysis was performed to determine the association between key fecal microbiota and differential metabolites. As shown in Figure 7, indole derivatives (including 3-indoleacrylic acid, methyl indole-3-acetate, 5-hydroxyindole-3-acetic acid), lipids (including vitamin A, oleic acid), and amino acids/peptides (including Val–Ser, L-threonine) were positively correlated with f_Clostridiaceae, f_Lachnospirceae (except for Roseburia), g_Bifidobacterium, unclassified_f_Peptostreptococcaceae, g_Barnesiella, and f_Rike nellaceae, and were negatively correlated with norank_ f_norank_o_Clostridia_vadinBB60_group, norank_f_Ruminoco ccaceae, and Monoglobus. These relations suggested that fecal microbiota could affect fecal metabolites in SN-6-fed Simmental. The rumen is the unique digestive organ of ruminants, and there are about 1010 bacteria per gram of rumen contents. Among them, cellulolytic bacteria are considered to be the main cellulose-decomposing agents (Brulc et al., 2009), accounting for about 10% of rumen microorganisms (Russell et al., 2009). The digestion and absorption of crude fiber in ruminants feed are completely dependent on rumen microorganisms, which exhibit a complex interplay with the host’s functions. Rumen bacteria significantly contribute to food digestion, thus considered potent probiotics and transferring these bacteria to other beef cattle could increase their daily weight gain. That is why, this study aimed to isolate and identify such kinds of bacteria and transfer them to the other cattle to investigate the daily weight gain and their interrelationship with host gut microecology and metabolic homeostasis. Three bacterial strains (Bacillus pumilus SN-3, Bacillus paralicheniformis SN-6, and Bacillus altitudinis SN-20, respectively) were isolated from the rumen of buffaloes. By evaluating SN-3, SN-6, and SN-20 for their antimicrobial properties, the ability to produce fiber digesting enzymes and proliferation, we selected the outstanding SN-6 and focused its possible effects on the Simmental cattle. Typically, SN-6 secrets laccase, which could degrade lignin in crude fiber and releases cellulose from the lignin. Most probably, it significantly contributes to the growth performance of the Simmental. Previous studies have confirmed that rumen microorganisms could secrete antibacterial substances, which could effectively inhibit the growth and multiplication of various pathogenic bacteria (Ren et al., 2019). It is also described that SN-6 had strong inhibitory effects on pathogenic S. aureus and E.coli K99. Therefore, it is reasonable to speculate that SN-6 is capable of preventing diseases caused by pathogenic S. aureus and E. coli K99, such as cow mastitis (Azara et al., 2017), calf diarrhea (Yadegari et al., 2019), piglet diarrhea (Xia et al., 2018), and so on. Interestingly, SN-6 strongly inhibited both gram-negative and gram-positive pathogenic bacteria, suggesting that SN-6 was a promising alternative for food antibiotics in livestock. In addition, the number of living bacteria of SN-6 after fermentation was up to 190 billion/g (data not shown), which demonstrated its tremendous advantage of convenience in preparation. It was reported that probiotics contribute to nutrition and metabolic health (Koh et al., 2016). Cox et al. demonstrated that probiotics promote a stable intestinal microbiota, stimulate digestive rates, and improve intestinal nutritional health (Cox and Dalloul, 2015). Possibly, these probiotics produce a large number of active enzymes during nutrient metabolism, which in turn increase intestinal digestive enzyme activity and promote nutrient absorption (Hu et al., 2018; Gong et al., 2018; Cao et al., 2020). In addition, numerous studies have shown that probiotics improve the feeding efficiency of animals by regulating the intestinal flora, and promoting the growth of dairy cows, lambs, rabbits, and sows (Sun et al., 2013; Jia et al., 2018; Liu L. et al., 2019; Zhang et al., 2020). Various nutrients ingested by the organisms are metabolized by a wide range of gut microbes to maintain complex life activities, where metabolites are transported, absorbed, or excreted through highly dynamic metabolic pathways (Maurice et al., 2013; Lamichhane et al., 2018). Importantly, the produced intestinal metabolites (i.e., tryptophan and short-chain fatty acids) nourish intestinal epithelial cells (Flint, 2016), improve the intestinal lining [He et al., 2022 (Microbiome)], and regulate downstream signaling pathways (Gill et al., 2006; Flint, 2016), acting as a link between the gastrointestinal tract and host health. The current study found that SN-6 could significantly regulate the intestinal flora and increase the average daily weight gain of Simmental, which is consistent with the previous studies. However, the mechanism of promoting growth by probiotics is not completely understood. Therefore, the effects of SN-6 feeding on fecal microbiota and metabolism of Simmental were explored in the present study, and the possible association between fecal microbiota and metabolism was evaluated. The intestinal microbiota has irreplaceable importance in the host’s vital activities and is, therefore, also known as “another organ of the body” (de Vos et al., 2022). The dynamics of the microbiota are influenced by diet, the environment, and other conditions. Probiotic intervention could alter the abundance and composition of microbiota in the gut, which in turn could affect host health (Schepper et al., 2019; Wang et al., 2020). The influences of gut microbiota on the host are highly correlated with complex interactions involving the host–microbe axes series (Xie et al., 2013). Studies on the gut microbiota provide a reference to explore the impact of gut microbiota interactions on organismal health. In this study, we also found that SN-6 feeding significantly influenced β diversity of the host gut microbiota, indicating that SN-6 had significant effects on microbial community structure in the host gut. Also, the differences in specific microorganisms further visualized the intrinsic link between SN-6 addition and gut microbiota composition. The results of this study revealed that SN-6 could increase the relative abundance of potentially beneficial bacteria (i.e., Clostridiaceae, Lachnospirales, and Bifidobacteriales) (Figure 4E), which most probably played an important role in promoting nutrient absorption, preventing diseases, and maintaining host health. Clostridium is a beneficial bacterium against intestinal bacterial infection (Behnsen, 2017; Kim et al., 2017). Lachnospira has a considerable ability to utilize dietary polysaccharides (O Sheridan et al., 2016). Similarly, members of Bifidobacterium are considered to play a critical role in maintaining human health (Di Gioia et al., 2014; Kusada et al., 2017). Meanwhile, the regulation of Bifidobacterium by SN-6 constituted the main factor underlying the increase in phylum Actinobacteria. The above results confirm that SN-6 feeding led to the development of a better structure of the host gut microbiota. In contrast, Monoglobus and Stackbrandtia were enriched in the control group. Monoglobus is often seen in an abnormal inflammatory state and tends to be elevated in the disease groups (Zhang et al., 2021; Miao and Davies, 2010). There are few studies and reports on the function of Stackbrandtia, which belongs to the Actinobacteria, Glycomycetaceae, and is mostly of environmental and soil origin (Zhang et al., 2016; Liu et al., 2018). Although all cattle were healthy, health-threatening microorganisms were still shown to be enriched in the gut of the control group, and these bacteria are likely to be transformed into pathogenic bacteria and involved in intestinal bacterial dysbiosis and disease transmission. Therefore, it is reasonable to speculate that SN-6 increased beneficial bacteria that promote nutrient absorption and helped in disease prevention and inhibited the colonization of potentially harmful bacteria, thus playing a significant role in increasing daily weight gain in Simmental by regulating the intestinal flora. Gut microbes perform a diverse range of metabolic functions including the production of numerous metabolites (Valdes et al., 2018). Increasingly recognized metabolites produced by gut microbes are vital mediators of diet-induced host–microbe interactions. We found that SN-6 affected the fecal metabolic pathways and metabolite concentrations of Simmental, amino acid metabolism, lipid metabolism, and vitamin metabolism were more enriched in the SN-6 group. Moreover, the contents of certain indole derivatives, lipids, and amino acids/peptides in the SN-6 group were significantly higher than those in the control group (Figure 6A). Amino acids are essential precursors for the synthesis of proteins and peptides and have been identified as markers of protein metabolism (Liu C. et al., 2019). Indole acrylic acid plays an essential role in maintaining intestinal homeostasis and barrier integrity (Agus et al., 2018). Kynurenic acid, produced by tryptophan metabolism, might have anti-inflammatory properties in the gastrointestinal tract and participate in immune regulation (Kennedy et al., 2017). In the metabolomic data, we also observed higher oleic acid and mannose contents in the SN-6 group. Oleic acid has natural antioxidant and anti-inflammatory properties. Zhang et al. (2017) found that mannose has an immunomodulatory function, which could specifically induce the differentiation of naive T cells into regulatory T cells (Treg). In mouse models, oral mannose could prevent and inhibit certain autoimmune diseases (Zhang et al., 2017). These findings indicate that SN-6 might exert a growth-promoting effect by elevating the relative concentrations of some positive functional metabolites that promote organismal health and homeostasis. The composition and metabolic pattern of the host–gut microbiota gradually change with the intervention of probiotics (Nealon et al., 2017). In this study, there was a significant correlation between fecal microbes and metabolites (Figure 7). Indole derivatives (including 3-indoleacrylic acid, methyl indole-3-acetate, 5-hydroxyindole-3-acetic acid), lipids (including vitamin A, oleic acid), and amino acids/peptides (including Val–Ser, L-threonine) were positively correlated with f_ Clostridiaceae, f_ Lachnospiraceae (except for Roseburia), g_Bifidobacterium, unclassified_f_Peptostreptococcaceae, g_Bar nesiella, f_ Rikenellaceae, and were negatively correlated with norank_f_norank_o_Clostridia_vadinBB60_group, norank_f_ Ruminococcaceae, and Monoglobus. Many studies have reported that Clostridiaceae, Bifidobacterium, and Peptostreptococcaceae could convert tryptophan into indole and indole derivatives (Aragozzini et al., 1979; Wikoff et al., 2009; Russell et al., 2013; Williams et al., 2014; Dodd et al., 2017; Wlodarska et al., 2017). Studies have shown that tryptophan and its downstream metabolites could bind to aryl hydrocarbon receptor (AHR); the resulting complex is transported into the nucleus, where AHR is activated (Lanis et al., 2017) to regulate intestinal homeostasis, improve gut barrier function, and activate the immune system (Roager and Licht, 2018). These findings also confirmed that SN-6 feeding altered the composition and metabolic pattern of the intestinal flora and thus contributed to the daily weight gain of Simmental. It should be mentioned that additional work is needed to address some limitations of the present study, e.g., verification experiments, and some of the work, such as whether SN-6 modulates gut and rumen microflora metabolism and influences the immunity of the body, is under investigation in our research group. The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material. The animal study was reviewed and approved by the ethics committee of animal welfare and research department, Huazhong Agricultural University. DS, YX, SY, and JL: conceptualization. DS, SY, JL, YC, YX, and ZZ: methodology. SY: software, data curation, and writing – original draft preparation. SY and RW: formal analysis. SY, HL, MA and DS: writing – review and editing. DS: supervision. All authors had read and approved the final manuscript.

PMC9649904

36386704

Márcia Melo Medeiros,Anna Cäcilia Ingham,Line Møller Nanque,Claudino Correia,Marc Stegger,Paal Skyt Andersen,Ane Baerent Fisker,Christine Stabell Benn,Miguel Lanaspa,Henrique Silveira,Patrícia Abrantes

Oral polio revaccination is associated with changes in gut and upper respiratory microbiomes of infants

28-10-2022

OPV-revaccination,non-specific effects,16S rRNA deep sequencing,bacterial microbiota,healthier microbiome composition,upper respiratory microbiome,gut microbiome

After the eradication of polio infection, the plan is to phase-out the live-attenuated oral polio vaccine (OPV). Considering the protective non-specific effects (NSE) of OPV on unrelated pathogens, the withdrawal may impact child health negatively. Within a cluster-randomized trial, we carried out 16S rRNA deep sequencing analysis of fecal and nasopharyngeal microbial content of Bissau–Guinean infants aged 4–8 months, before and after 2 months of OPV revaccination (revaccinated infants = 47) vs. no OPV revaccination (control infants = 47). The aim was to address changes in the gut and upper respiratory bacterial microbiotas due to revaccination. Alpha-diversity for both microbiotas increased similarly over time in OPV-revaccinated infants and controls, whereas greater changes over time in the bacterial composition of gut (padjusted < 0.001) and upper respiratory microbiotas (padjusted = 0.018) were observed in the former. Taxonomic analysis of gut bacterial microbiota revealed a decrease over time in the median proportion of Bifidobacterium longum for all infants (25–14.3%, p = 0.0006 in OPV-revaccinated infants and 25.3–11.6%, p = 0.01 in controls), compatible with the reported weaning. Also, it showed a restricted increase in the median proportion of Prevotella_9 genus in controls (1.4–7.1%, p = 0.02), whereas in OPV revaccinated infants an increase over time in Prevotellaceae family (7.2–17.4%, p = 0.005) together with a reduction in median proportion of potentially pathogenic/opportunistic genera such as Escherichia/Shigella (5.8–3.4%, p = 0.01) were observed. Taxonomic analysis of upper respiratory bacterial microbiota revealed an increase over time in median proportions of potentially pathogenic/opportunistic genera in controls, such as Streptococcus (2.9–11.8%, p = 0.001 and Hemophilus (11.3–20.5%, p = 0.03), not observed in OPV revaccinated infants. In conclusion, OPV revaccination was associated with a healthier microbiome composition 2 months after revaccination, based on a more abundant and diversified bacterial community of Prevotellaceae and fewer pathogenic/opportunistic organisms. Further information on species-level differentiation and functional analysis of microbiome content are warranted to elucidate the impact of OPV-associated changes in bacterial microbiota on child health.

Oral polio revaccination is associated with changes in gut and upper respiratory microbiomes of infants After the eradication of polio infection, the plan is to phase-out the live-attenuated oral polio vaccine (OPV). Considering the protective non-specific effects (NSE) of OPV on unrelated pathogens, the withdrawal may impact child health negatively. Within a cluster-randomized trial, we carried out 16S rRNA deep sequencing analysis of fecal and nasopharyngeal microbial content of Bissau–Guinean infants aged 4–8 months, before and after 2 months of OPV revaccination (revaccinated infants = 47) vs. no OPV revaccination (control infants = 47). The aim was to address changes in the gut and upper respiratory bacterial microbiotas due to revaccination. Alpha-diversity for both microbiotas increased similarly over time in OPV-revaccinated infants and controls, whereas greater changes over time in the bacterial composition of gut (padjusted < 0.001) and upper respiratory microbiotas (padjusted = 0.018) were observed in the former. Taxonomic analysis of gut bacterial microbiota revealed a decrease over time in the median proportion of Bifidobacterium longum for all infants (25–14.3%, p = 0.0006 in OPV-revaccinated infants and 25.3–11.6%, p = 0.01 in controls), compatible with the reported weaning. Also, it showed a restricted increase in the median proportion of Prevotella_9 genus in controls (1.4–7.1%, p = 0.02), whereas in OPV revaccinated infants an increase over time in Prevotellaceae family (7.2–17.4%, p = 0.005) together with a reduction in median proportion of potentially pathogenic/opportunistic genera such as Escherichia/Shigella (5.8–3.4%, p = 0.01) were observed. Taxonomic analysis of upper respiratory bacterial microbiota revealed an increase over time in median proportions of potentially pathogenic/opportunistic genera in controls, such as Streptococcus (2.9–11.8%, p = 0.001 and Hemophilus (11.3–20.5%, p = 0.03), not observed in OPV revaccinated infants. In conclusion, OPV revaccination was associated with a healthier microbiome composition 2 months after revaccination, based on a more abundant and diversified bacterial community of Prevotellaceae and fewer pathogenic/opportunistic organisms. Further information on species-level differentiation and functional analysis of microbiome content are warranted to elucidate the impact of OPV-associated changes in bacterial microbiota on child health. The enormous contribution of vaccines against childhood diseases is well-known (Uwizihiwe, 2015), and much of the decline in child mortality since 1990 (United Nations Inter-agency Group for Child Mortality Estimation [UN IGME], 2019) may be ascribed to the increase in vaccination coverage (National Research Council (US) Working Group on the Effects of Child Survival and General Health Programs on Mortality, 1993). Furthermore, analyses of epidemiological data on childhood mortality have pointed to additional beneficial effects provided by live-attenuated vaccines, such as measles (MV) and Bacillus Calmette-Guérin (BCG) vaccines (Higgins et al., 2016), and, oral polio vaccine (OPV) (Andersen et al., 2018, 2021), by decreasing the all-cause child mortality rate in low income settings (Shann, 2010; Aaby and Benn, 2021; Prentice et al., 2021). This has been confirmed in several randomized clinical trials (Aaby et al., 2010; Lund et al., 2015; Biering-Sørensen et al., 2017). These vaccines seem to confer protection against unrelated pathogens, exhibiting non-specific effects (NSE). The development of a heterologous adaptive host immune response and the so-called, trained immunity or innate immune memory, i.e., epigenetic and metabolic reprogramming of innate immune cells (Netea et al., 2016), have been pointed out as possible immune mechanisms behind NSE. BCG’s capacity to induce a heterologous adaptive immune response and trained immunity has been confirmed in clinical trials with healthy adult volunteers (Kleinnijenhuis et al., 2012, 2014), and a recent retrospective analysis done at Mayo Clinic raised a possible role for trained immunity toward the partial protection provided by OPV against SARS-CoV-2 infection (Pawlowski et al., 2021) and some authors even found that OPV immunization vs. placebo reduced the number of laboratory-confirmed COVID-19 infections (Yagovkina et al., 2022) and improved overall health in males during the COVID-19 pandemic (Fisker et al., 2022). Eradication of polio infection preconizes the withdrawal of OPV from routine immunization programs (World Health Organization [WHO], 2021b). However, considering the beneficial NSE, an increased understanding of them is urgently needed to mitigate the potential impact of OPV withdrawal on child survival in these settings. Gut commensal bacteria are essential for the development and function of innate and adaptive immune systems (Belkaid and Hand, 2014), and potentially play a role in the induction of innate immune memory (Danelishvili et al., 2019). Polioviruses (PV), including the live-attenuated ones present in OPV formulation, replicate in the oropharyngeal, but mainly in gut mucosal cells (Mendelsohn et al., 1989). The specific composition of gut bacterial microbiota seems to be important to drive a strong OPV-induced immune response. For example, Actinobacteria, mainly represented by Bifidobacterium longum, seem to be important to drive specific T-cell and antibody responses to live-attenuated PV (Huda et al., 2014), whilst a more diverse environment with abundance in Firmicutes and Proteobacteria seems to have a negative effect (Zhao et al., 2020). Furthermore, some authors have suggested a role for pathogenic bacteria or environmental enteropathy in the low-immunogenicity levels of oral vaccines in low-income countries (Grassly et al., 2016; Parker et al., 2018, 2021). The OPV-induced immune response seems to play a role against pathogenic gut bacteria by reducing etiology-specific bacterial diarrhea in male infants (Upfill-brown et al., 2020), and possibly favoring innate immune mechanisms that may mediate OPV-NSE, such as the presence of anti-microbial peptides in the stool of newborns 6 weeks after the first OPV dose administered at birth, concomitantly with BCG (Alam et al., 2015). Previous evidence furthermore suggests that OPV may protect against upper respiratory infections, such as acute otitis media, caused by other viruses from the family Picornaviridae, such as Rhinoviruses, through immune mechanisms involving heterologous adaptive response and antiviral innate defense based on type I interferon (Seppälä et al., 2011). All studies above highlight the complex interaction among OPV, bacteria, and/or viruses present in the respiratory tract and gut. We hypothesized that OPV may induce changes in the microbiome of the sites of live-attenuated PV replication, which may mediate OPV-NSE. To assess the effect of OPV revaccination on the diversity and composition of the bacterial microbiota of the gut and upper respiratory tract, we carried out the Micro-OPV study. Micro-OPV is a cohort study within a cluster-randomized trial, the RECAMP-OPV (Varma et al., 2019). RECAMP-OPV had a random allocation of infants to OPV revaccination or no revaccination, and it was conducted by the Bandim Health Project (BHP) in Guinea-Bissau. Infants aged 17–34 weeks participating in RECAMP-OPV and living 2 h-drive from the capital Bissau were recruited to the Micro-OPV. At this age range infants have already been eligible for a set of vaccines as part of the routine immunization plan: BCG and first OPV dose at birth (OPV0), followed by three doses of Pentavalent, Pneumococcal-conjugated, and OPV, both administrated at 6, 10, and 14 weeks of age, and Rotavirus vaccine administrated at 6 and 10 weeks of age (World Health Organisation [WHO], 2021a). Due to randomization design, recruited infants shared similar major microbiome determinants which are detailed below (in Materials and methods). Fecal and nasopharyngeal microbial content of infants were collected at enrollment and after 2 months of follow-up and submitted to 16S rRNA deep sequencing analysis. We compared the changes in the gut and upper respiratory bacterial microbiotas from enrollment to follow-up between 47 OPV revaccinated infants and 47 not OPV revaccinated controls, who completed the 2 months of follow-up. All infants aged 17–34 weeks recruited to the RECAMP-OPV trial (Varma et al., 2019) (NCT03460002 on www.clinicaltrials.gov), and living in village clusters distant 2 h-drive from the capital Bissau were eligible to participate in the Micro-OPV cohort study (see the experimental design of Micro-OPV study in Figure 1A and the Guinea–Bissau map with the covering area of the Micro-OPV study in Figure 1B). This age range was planned because infants at those ages have already been eligible for a set of vaccines as part of the routine immunization plan, except MV (World Health Organisation [WHO], 2021a; Figure 1A). Area delimitation was due to logistic reasons concerning the transportation and preservation of biological samples, which are further detailed below. Biosampling took place from August 2018 to December 2018. In RECAMP-OPV clinical trial, clusters were randomly classified 1:1 as Intervention clusters (with infants revaccinated with an extra dose of OPV in addition to those administrated according to routine immunization plan) or Control clusters (whose infants received no additional OPV) based on externally generated random numbers. All clusters received the same number of visits, at baseline and 2 months post-baseline. Due to the randomization design, recruited infants shared similar major microbiome determinants, such as, at term gestational age, vaginal delivery, and feeding progression. Given the exploratory nature of the Micro-OPV study, we conveniently planned to recruit 60 infants from each cluster to participate. At baseline and 2 months after baseline, in addition to the main trial procedures required by the RECAMP-OPV, including the administration of an extra dose of OPV to infants from the Intervention clusters at baseline, fecal and nasopharyngeal biological content was collected, and clinical data registered. A total of 47 participants out of 60 revaccinated with OPV and 47 participants out of 58 not revaccinated completed the 2 months follow-up in the Micro-OPV study. Study villages are agricultural-based, served by well water and pit-latrines, and just 30–50% of them have access to electricity supply. Households live very close to each other, often sharing the same kitchen space and kitchenware, and people of the same household share the same food container. Also, people live very close to farm animals (chicken, pig, goat, and cow) and pets, such as dogs and cats. Procedures in the field under the Micro-OPV study were carried out in the rainy season (August to December) in 2018. After assurance of all ethical aspects described below, parents or legal guardians of selected infants attended a questionnaire-based interview at both time-points. All interviews were conducted in Créole, the lingua-franca, by the same BHP field worker under the researchers’ supervision. Information on major microbiome determinants and possible cofounders of the effect of OPV revaccination, such as housing conditions, mother’s health during pregnancy and delivery, and infant’s health since delivery to present, including delivery conditions, feeding history, previous infections, hospitalizations, and use of medicines were recorded at enrollment. Feeding progression, health problems, hospitalizations, and use of medicines since the enrollment were registered at follow-up. After the interviews, infants were submitted to a clinical evaluation and measurement of axillary temperature. Anthropometric data [weight, height, and mid-upper-arm-circumference (MUAC)] were also noted. No infant met exclusion criteria at enrollment, namely, fever defined as an axillary temperature ≥38°C, MUAC < 110 mm in an infant older than 6 months or previous history of allergic reaction due to vaccination. Nasopharyngeal (NP) samples and fecal (F) samples were collected from all participants at enrollment and 2 months later at follow-up using ultra minitip flexible and regular flocked swabs (FLOQSWABS®, Copan, Brescia, Italy), respectively. Immediately after the collection procedure, swabs were immersed in 1 ml of DNA/RNA Shield™ (Zymo Research, Irvine, CA, USA) in 15 ml Falcon tubes and kept cold until laboratory processing. At enrollment, after sampling procedures, one OPV-dose (two drops) was administered to infants belonging to the Intervention clusters, by the experienced BHP nurse. Infants belonging to the Control clusters received no intervention. No additional OPV dose was administrated between the two time-points to any of the participants. The raw material was transferred from Falcon tubes to 2 ml Cryotubes in the laboratory facility of the Guinea-Bissau National Institute of Health in sterile conditions, kept at −20°C for 1 month in the BHP premises, transported to the Institute of Hygiene and Tropical Medicine, Universidade NOVA de Lisboa, Portugal (IHMT-NOVA) in dry ice and kept at −80°C until DNA extraction. DNA was extracted from 500 μl of each raw sample through ZymoBIOMICS™ DNA Miniprep Kit (cat. no°.: D4300, Zymo Research, Irvine, CA, USA) following manufacturer’s orientations. Mechanical lysis was standardized through a ZymoBIOMICS™ Microbial Community Standard II (Log Distribution) (cat. n°.: D6310, Zymo Research, Irvine, CA, USA) using a Mini-BeadBeater device (model EUR16, BIOSPEC Products, Inc., Bartlesville, OK, USA). All DNA samples presented A260/A280 ratio >1.8 measured by a NanoDrop™ 1000 Spectrophotometer device (Thermo Fischer Scientific, Waltham, MA, USA) and 100 ng of DNA per sample, measured in a Qubit Fluorometer (Thermo Fischer Scientific, Waltham, MA, USA), were used to prepare the library. Paired (F and NP) extracted DNA samples of 94 participants who completed the 2 months of follow-up (47 OPV revaccinated infants and 47 controls) were deeply characterized by 16S rRNA amplicon sequencing to identify different changes in bacterial diversity and composition of gut (G) and upper respiratory (R) bacterial microbiotas over time. For that, paired-ended amplicon deep sequencing of V3 and V4 variable regions of the gene that codifies the 16S rRNA was performed using previously evaluated primers (341F: 5′-CCTACGGGNGGCWGCAG-3′; 805R: 5′-GACTACHVGGGTATCTAATCC-3′), preceded by heterogeneity spacers (Klindworth et al., 2013). Library construction and sequencing were performed on an Illumina MiSeq instrument (Illumina Inc., San Diego, CA, USA), using a 600 cycle V3 kit. Demultiplexing was done with the bcl2fastq Conversion Software (Illumina Inc., San Diego, CA, USA). Cutadapt (version 2.3) was used to trim off heterogeneity spacers and primers at an error rate of 8% (allowing for one mismatch per primer) (Martin, 2011). The resulting sequences were quality-filtered, truncated, denoised, merged, and chimera-filtered through the DADA2 pipeline (version 1.12.1). High-resolution amplicon sequence variants (ASVs) were inferred from trimmed reads per run with default settings, except for truncation lengths (forward reads: 270 bp, reverse reads: 210 bp) (Callahan et al., 2016). Consensus removal of chimeras was performed on data from all runs. Samples with a read count <10,000 after quality filtering were re-sequenced. Taxonomic assignment of the resulting ASVs was done by means of dada2’s assignTaxonomy() and addSpecies() functions with the Silva reference database and species-level training set (version 132), respectively (Callahan, 2018). Subsequently, the ASV count table and the taxonomic table were integrated with the R package phyloseq (McMurdie and Holmes, 2013) to perform ecological and statistical analyses of microbiota data. The data set was split into F and NP samples and the R package decontam was used to identify and remove contaminants (Davis et al., 2018). For this, controls used in DNA extraction, the DNA/RNA Shield™ (Zymo Research, Irvine, CA, USA) solution and the controls from PCR were used. The decontam method “either” was used in F and NP samples data set adjusted for each one (frequency threshold 0.075, prevalence threshold 0.5 and frequency threshold 0.05, prevalence threshold 0.5, respectively). After computational contaminant removal, read counts of re-sequenced samples were merged. From the F sample data set, 67 ASVs were identified as contaminants and removed. In addition, 40 ASVs classified as Archaea, Cyanobacteria, Planctomycetes, Chloroflexi, Rhizobiales, Rhodobacterales, or Rhodospirillales were removed manually, as our major interest was to observe the behavior of the bacterial microbiota of gut and upper respiratory tract. Moreover, 61 ASVs that were not classified to at least order-level were excluded. Twenty-seven ASVs from the NP sample data set that were identified as contaminants were excluded. In addition, 99 ASVs that were not classified to at least order-level were excluded, as well as 224 ASVs classified as Eukaryota, Archaea, Cyanobacteria, Planctomycetes, Chloroflexi, Deinococcus-Thermus, Kiritimatiellaeota, Rhizobiales, Rhodobacterales, Rhodospirillales, or Oceanospirillales. A number of 2,373 ASVs were obtained from the 94 F samples, while 770 ASVs were obtained from the 94 NP samples, both distributed by seven taxonomic ranks. The average read count per sample after all pre-processing steps was 37,897 (F samples) and 31,139 (NP samples). Statistical significance levels were set at <0.05 for all tests used in analyses of epidemiological or sequencing data unless otherwise stated. General characteristics at the enrollment and follow-up were described by group allocation (Intervention group with OPV revaccinated infants or Control group with not revaccinated controls) as proportions, averages, and medians, as appropriate, in IBM Statistics SPSS 25th version program. Differences in quantitative variables between groups were tested using Student’s t-test or Mann–Whitney U test, while within-group differences from enrollment to follow-up were tested using paired Wilcoxon signed rank test. Differences in qualitative variables between groups were tested using the Chi-square test or the Exact Fisher test, while within-group differences over time were tested using the McNemar test. We did not correct for the cluster, i.e., infants were considered independent in the analysis, which would tend to exaggerate potential group differences. Statistical analyses of microbiota characteristics were performed in R (version 4.0.4) (R Core Team, 2021). Graphs were created with ggplot2 (Wickham, 2016). Indexes of Shannon diversity (H) and Inverse of Simpson (InvSimpson) were used to assess alpha-diversity, i.e., differences in within-sample richness and evenness of microbiotas. Shannon diversity is a measure of alpha-diversity combining richness and evenness, where higher values indicate a high number of different taxonomic groups homogeneously distributed in a community. InvSimpson is a measure of dominance or a complementary measure of evenness, where higher values indicate homogeneous proportions of taxonomic groups and low dominance levels. They were compared between groups at enrollment using Mann–Whitney U test and within-groups over time using paired Wilcoxon signed rank test. A core microbiota was defined as the subset of ASVs with at least five reads in at least two samples. The count data was then Hellinger transformed, that is calculation of sample-wise proportions and subsequent square root transformation, to account for differences in samples’ library sizes and zero inflation. Principal co-ordinates analysis (PCoA) based on the Bray–Curtis distance for visualization of differences in the bacterial composition between OPV revaccinated and controls and between time-points was performed. Differences in bacterial composition between groups and time points were tested with permutational multivariate analysis of variance using distance matrices (PERMANOVA) with the adonis() function from the vegan package (version 2.5-5) (Oksanen et al., 2019). As a pre-requisite, group homogeneity of variances was tested by implementing the betadisper() function. Additionally, pairwise comparisons of within- and between-group Bray-Curtis dissimilarities of all group-time point combinations were made using a pairwise Wilcoxon rank sum test with continuity correction and p-values adjusted with Benjamini–Hochberg correction. Taxonomic heat trees based on relative abundance up to genus level were generated with the package metacoder (Foster et al., 2017) and used to visualize differences in relative abundances between enrollment and follow-up within OPV revaccinated and controls. Differentially abundant taxa between enrollment and follow-up within groups were identified by linear discriminant analysis (LDA) effect size (LEfSe) on count data normalized to the sum of 1e + 06 with an LDA cut-off of 4 (package microbiomeMarker) (Cao, 2020). Taxa identified more than three times in at least 5% of the samples were considered in the analysis. A high LDA score (log10) indicates a high effect size of the respective taxon for explaining group difference. LEfSe allows the identification of differentially abundant taxa on several taxonomic levels (here kingdom to species) by accounting for the hierarchical structure of bacterial phylogeny. The following results gather data from 47 participants in each group who completed the follow-up, according to the experimental design of the Micro-OPV study shown in Figure 1A. No major differences in participant profiles were observed between the 47 (60% female) OPV revaccinated infants (participants in the Intervention group) and the 47 (49% female) controls (participants in the Control group) at enrollment (Table 1). Although a significantly higher mean length was observed for OPV-revaccinated infants at enrollment, it may reflect the slightly higher age of the infants in this group and, possibly missed length measures for six participants in the Control group and for two OPV-revaccinated participants. This difference was no longer observed at follow-up when complete data were recorded from all infants. No major differences in clinical conditions anticipated to affect the microbiome at the time of sampling and during the 2 months follow-up were observed between groups (Tables 1, 2). Participant profiles after excluding infants that used antibiotics during follow-up (seven OPV revaccinated infants and three controls) were similar to that shown for the 47 infants in each group (Supplementary Tables 6, 7). Infants in both groups were similar in characteristics of major microbiome determinants at enrollment, such as, environmental, birth, diet, and clinical factors that influence the microbiome development (see Supplementary Tables 1–4). Supplementary Table 3 describes, specifically, the feeding history of participants at enrollment and follow-up. No significant differences were observed between groups. Breastfeeding coverage was very high among all participants at both time-points. The same was observed regarding porridge feeding and water administration, including median number of doses administrated per week. Regarding another sources of nutrients, fruits were introduced at follow-up for both groups of infants, however, to a lower extent. Significant differences were observed regarding a lower number of residents per house younger than 15 years old (5 vs. 6, p = 0.02) in the group with OPV revaccinated infants, together with a relatively higher use of tap water (5 vs. 0, p = 0.03) and charcoal to cook (6 vs. 0, p = 0.03) in this same group, although most infants in both groups lived in households which used well water and wood to cook. All deliveries in both groups came from at term pregnancies of newborns born vaginally and almost 50% at home. Few mothers and infants in both groups had been tested for human immunodeficiency virus (HIV), and no positive cases had been detected. Chemoprophylaxis to malaria in pregnancy was taken by almost 80% of mothers and few of them reported malaria in pregnancy. Most infants were exclusively breastfed up to the 5th month of life, and then started to receive porridge and water. Few infants had malaria before enrollment, while a higher proportion had already experienced diarrhea episodes. More than 50% of infants in both groups had used medicines up to 1 month before enrollment, mainly Paracetamol and equally few infants in both groups had used antibiotics previously, mainly Amoxicillin. No significant differences were observed regarding the proportion of infants in both groups with complete doses of recommended vaccines according to routine immunization schedule until 14 weeks of age (World Health Organisation [WHO], 2021a). The distribution of Shannon (H) and Inverse of Simpson (InvSimpson) indexes in both groups at enrollment and follow-up is shown in Figure 2. First, we showed that infants in both groups presented similar indexes for both microbiotas at enrollment as observed in Figures 2A,B (for G microbiota) and Figures 2C,D (for R microbiota), evidencing similar exposition to major microbiomes determinants between groups, as previously shown. Then, both indexes significantly increased from enrollment to follow-up in G microbiota of OPV revaccinated infants (H p = 0.001, InvSimpson p < 0.001) and controls (H p = 0.03, InvSimpson p = 0.03), whereas in R microbiota, the H index only significantly increased in OPV revaccinated infants (p = 0.011). Changes in InvSimpson index in R microbiota were not significant for both groups in the same period. Next, we compared the magnitude of changes over time for both microbiotas in each group to test whether they might still be greater in OPV revaccinated infants due to revaccination. A similar magnitude of changes in alpha-diversity over time [comparison (3) in Figure 2] was found in G (Figures 2A,B) and R microbiotas (Figures 2C,D) in infants from both groups. Similar results were obtained after excluding infants in both groups who used antibiotics during follow-up period (data not shown). In summary, alpha-diversity for both microbiotas were similar between groups at enrollment, i.e., before infants in the Intervention group have received OPV revaccination. Two months later, we observed that alpha-diversity increased similarly in both groups over time, despite infants in the Intervention group have received the OPV boost. Visualization of differences in bacterial composition between groups and time-points for G and R microbiota samples were performed using PCoA based on Bray–Curtis dissimilarities (Supplementary Figures 1A,B). After ensuring homogeneity in whitin-group variations by implementing the betadisper() function (ANOVA, p > 0.05), PERMANOVA analyses for G and R microbiotas showed that differences between bacterial composition were based on group and time point for G (p < 0.01) and R (p = 0.001) bacterial microbiotas. When analyses were deepened to identify whether differences were attributable to time point, group or an interaction of both, we identified that differences for G bacterial microbiota were more attributable to time point (p < 0.01) than group (p > 0.05) or an interaction of both (p > 0.05), whilst for R bacterial microbiota differences were attributable to time-point (p < 0.05) and group (p = 0.001) but not due to an interaction of both (p > 0.05). To assess differences in bacterial composition for both microbiotas over time and to test if they were larger in OPV revaccinated infants than controls, firstly, data were split according to time-point and G and R samples from OPV revaccinated and controls at enrollment were analyzed. After Hellinger transformation of data and ensuring homogeneity in within-group variations by implementing the betadisper() function (ANOVA, p > 0.05), PERMANOVA analyses showed no differences between groups regarding bacterial communities in G (p > 0.05) and R (p > 0.05) microbiotas at enrollment, i.e., at baseline, before infants in the Intervention group being revaccinated with an extra-dose of OPV. Next, we split up phyloseq object by group and tested if there was a difference in within-group bacterial composition between samples at enrollment and samples collected 2 months later. After Hellinger transformation of data and testing for homogeneity in whitin-group variation by implementing the betadisper() function (ANOVA, p > 0.05), PERMANOVA analyses showed significant differences over time in bacterial composition in R microbiota samples of OPV revaccinated infants (p < 0.05) and controls (p = 0.01), indicating changes over time in R bacterial microbiota samples in both groups. Using the same methodology, PERMANOVA analyses showed that the bacterial composition of G microbiota samples of controls was different between time points (p < 0.05). Heterogeneity in within-group variation over time in G microbiota samples of OPV revaccinated infants avoided PERMANOVA analyses to verify differences in bacterial composition over time. This fact suggests that changes over time had occurred in G microbiota of OPV revaccinated infants, at least in some samples. To assess whether changes over time in bacterial composition were greater in OPV revaccinated than controls, we carried out pairwise comparisons of within- and between-group Bray–Curtis dissimilarities on Hellinger transformed data. Data for G and R microbiotas are shown in Figures 3A,B, respectively. First, we observed that bacterial communities in G microbiota in both groups at enrollment (first black box, Figure 3A) were no more different from each other than within the two groups at enrollment [first (padjsted = 0.524) and third gray boxes (padjusted = 0.540), Figure 3A]. The same was found for bacterial communities in R microbiota [first and third gray boxes (padjusted = 0.199), Figure 3B]. Then, the Control group was used as a reference for changes in OPV revaccinated infants over time, showing that bacterial communities in G microbiota changed significantly more over time in OPV revaccinated infants than controls (blue box vs. second black box, respectively, padjusted < 0.001, Figure 3A). Results for R microbiota samples were similar, also showing a significantly larger change over time in OPV revaccinated infants than controls (blue box vs. second black box, respectively, padjusted = 0.018, Figure 3B). In summary, results suggest that changes over time in the composition of G and R bacterial microbiotas occurred in both groups and were greater in OPV revaccinated infants. The large within-group variation found in G microbiota samples in the group of OPV revaccinated infants at follow-up may suggest favorable growth of specific taxa in certain samples. Similar results were found after excluding antibiotic users in both groups (data not shown). Differences in the relative abundance of the 50 most abundant genera and higher taxonomic levels for both microbiotas and groups over time were visualized in differential taxonomic heat trees shown in Supplementary Figures 2A–D, fully described in the same Supplementary material. LEfSe analysis was used to identify taxa groups that significantly changed in both microbiotas and groups over time (LDA score ≥4). Figure 4A illustrates which taxa groups significantly decreased over time in OPV revaccinated infants (yellow bars), i.e., that were higher at enrollment, and which ones significantly increased over time (green bars). B. longum, Escherichia/Shigella, Campylobacter, and Clostridium butyricum are the taxa that most decreased in OPV revaccinated (range of padjusted values from 0.0006 to 0.016), whereas Prevotellaceae and Streptococcus were the ones that most increased over time (range of padjusted values from 0.001 to 0.03). A small number of taxa (dark red bars in Figure 4B) significantly decreased in G microbiota in the Control group, mainly B. longum (padjusted = 0.01), whereas a greater number of taxa significantly increased over time (blue bars in Figure 4B), mainly Prevotella_9, Bifidobacterium kashiwanohense and Campylobacter (range of padjusted values from 0.0007 to 0.02). Corynebacterium_1 was the taxon that most decreased in R microbiota in OPV revaccinated (yellow bars in Figure 4C, range of padjusted values from 0.0003 to 0.0006), whereas Haemophilus (padjusted = 0.006) and Staphylococcus (padjusted = 0.002) were those that most increased over time (green bars in Figure 4C). Corynebacterium_1, Staphylococcus, and Fusobacterium were those that most decreased in the Control group, mainly the former (padjusted = 0.0002, dark red bars in Figure 4D), whereas Haemophilus (padjusted = 0.03) and Streptococcus padjusted = 0.001) were the ones that most increased over time (blue lines in Figure 4D). When comparing changes over time among taxa with relative abundance greater than 1%, we can observe more taxa groups in OPV revaccinated (Figure 5A) than in controls (Figure 5B) that significantly changed in G microbiota. The opposite occurred in R microbiota, for OPV revaccinated (Figure 5C) and controls (Figure 5D). A significant decrease in the median proportion of B. longum over time was observed both in OPV revaccinated (25–14.3%, padjusted = 0.0006) and controls (25.3–11.6%, padjusted = 0.01) (Figure 5A). A significant increase in the median proportion of Prevotellaceae (7.2–17.4%, padjusted = 0.005) family was observed in OPV revaccinated infants (Figure 5A), whereas a significant increase in the median proportion of Prevotella_9 genus was observed in the Control group (1.4–7.0%, padjusted = 0.02) (Figure 5B). Additionally, a significant decrease in Escherichia/Shigella (5.8–3.4%, padjusted = 0.01) was observed within OPV revaccinated group over time (Figure 5A), with a significant decrease of an undetermined Escherichia/Shigella species (4.9–3%, padjusted = 0.02). On the other hand, the median proportion of Campylobacterales (0.9–2.6%, padjusted = 0.03) and Lachnospiraceae (0.8–2.2%, padjusted = 0.007) significantly increased over time in G microbiota in the Control group (Figure 5B). The main change in R microbiota of the OPV revaccinated infants over time (Figure 5C) was the significant reduction in the median proportion of a Corynebacterium_1 species (3.4% to 0.8%, padjusted = 0.0006). The same was also observed in the Control group (4.8–1%, padjusted = 0.005) (Figure 5D), but simultaneously with an increase in the median proportion of other several taxa groups (Figure 5D), such as Firmicutes (from 19.2–22.8%, padjusted = 0.04), Streptococcus genus (2.9–11.8%, padjusted = 0.001) and an undetermined Streptococcus species (2–5.9%, padjusted = 0.01), followed by the increase in Pasteurellaceae family (11.4–20.5%, padjusted = 0.03), including organisms from Haemophilus genus (11.3–20.5%, padjusted = 0.03). In summary, taxonomic composition of G and R microbiotas changed differently over time in OPV revaccinated infants vs. controls. A more proportional alternance over time between the most abundant commensals (B. longum and Prevotellaceae) was observed in G microbiota of OPV revaccinated infants. Also, it was observed a more abundant and diversified community of Prevotellaceae in G microbiota of OPV revaccinated infants, followed by a reduction in the median proportion of Escherichia/Shigella and limited increase in the median proportion of Campylobacterales. In R microbiota, it was observed a limited increase in the median proportion of Streptococcus and Haemophilus in OPV revaccinated infants, whereas in controls these taxa were present among those with relative abundance greater than 1% at follow-up. Similar results were found after excluding antibiotic users in both groups (data not shown). This is the first study that used 16S rRNA deep sequence analysis of gut and upper respiratory bacterial content to show changes in the composition of gut and upper respiratory microbiomes after OPV revaccination. Strengths of the study includes the cluster-randomized set-up resulting in similar epidemiological profiles and characteristics of major microbiome determinants in both groups at enrollment. This allowed us to provide unbiased effects of OPV revaccination on the selected microbiomes. High-quality sequencing results reflect the good strategies adopted for sequencing, as well as for sample collection, storage, and DNA extraction. A 2-month follow-up period was planned to maximize detection of changes in both microbiomes after OPV revaccination, as in that period, replication of the live-attenuated PV and an intense activity of host immune system due to revaccination are ongoing (Connor et al., 2021). Also, a longer observation period for weaning age infants could dilute the results. Main weaknesses found in the present study were the intrinsic limitation of 16S rRNA gene deep sequencing analysis for species discrimination and the lack of functional analysis of microbiome content. As the clustering was not considered in the statistical analyses of epidemiological data and due to small number of participants in the Micro-OPV study, there was no power to make any conclusions regarding clinical protection to unrelated pathogens conferred by OPV-associated changes in bacterial microbiota. Regarding gut microbiota composition, our results are in line with the successive colonization events of the intestinal tract, that are described (Moore and Townsend, 2019; Dizzell et al., 2021) for at term vaginally born infants, exclusively breastfed until 6 months of life and on a recent weaning process with the introduction of starches through porridges, such as participating infants in the Micro-OPV study. B. longum was the major representative of the Actinobacteria phylum in G microbiotas for both groups at enrollment, corroborating literature data that point to B. longum, Bifidobacterium breve and Bifidobacterium bifidum as the dominant bacteria in G microbiota of breastfed infants, as these commensal bacteria are the main metabolizers of human milk oligosaccharides (O’Callaghan and van Sinderen, 2016; Le Doare et al., 2018). A high breastfeeding coverage, given as the only nutrient source until 6 months of age, was observed among the participating infants. At follow-up, we found a significant decrease in Actinobacteria and increase in Bacteroidetes for both groups. A higher porridge coverage at follow-up in both groups, as shown in Supplementary Table 3, explains this finding and corroborates literature data which point to an increase in Bacteroidetes after introduction of more solid foods, mainly starches (Houghteling and Walker, 2015; Moore and Townsend, 2019). Also, the marked abundance of Bacteroidetes is in accordance with what has been previously described for infants living in rural African villages (De Filippo et al., 2010). Despite no differences between the OPV revaccinated infants and the control infants regarding feeding history and progression, a more diverse and pronounced increase in relative abundance of the Prevotellaceae family, comprising several Prevotella genera, was observed in OPV revaccinated infants, whereas a less expressive increase of this taxon, restricted to Prevotella_9 genus occurred in controls. Even though a similar greater proportion of infants in both groups at enrollment were fed with porridge and this number increased similarly in both groups at follow-up, a higher diversity in Bacteroidetes was observed in OPV revaccinated infants at follow-up, followed by a more proportional alternance over time between the median proportions of Actinobacteria and Bacteroidetes compared to Control group. A decrease in potentially pathogenic/opportunistic organisms of Proteobacteria, such as Escherichia/Shigella, is expected for weaning infants and was observed in OPV revaccinated infants over time, followed by a decrease in the relative abundance of Streptococcus. The same decrease was not observed in the Control group and, indeed, an increase over time in Campylobacterales was observed together with a more pronounced increase in families such as Lachnospiraceae and Ruminococcaceae known to codominate the fecal bacteria of healthy adults (Ishiguro et al., 2018). These results reveal a potential OPV beneficial effect associated with a greater diversity and abundance of important gut commensal bacteria, such as those from the Prevotellaceae family, known producers of short-chain fatty acids (SCFA), and decrease in potentially pathogenic/opportunistic taxa. SCFA, such as acetate, propionate and butyrate, are the final products from the metabolism of non-digestible dietary fibers and play a role in the modulation of gut immune system (Flint et al., 2014; Deleu et al., 2021). They are involved in epigenetic regulation of innate immune cells and possibly, in the developmental process of innate memory (Danelishvili et al., 2019), strengthening the hypothesis of trained immunity mechanisms behind NSE of OPV and due to changes in bacteria microbiota. In contrast, control infants that did not receive OPV revaccination developed a more imbalanced gut bacterial microbiota composition, with relatively more potential opportunistic bacteria and with some of them expected to be present in older stages of life. We cannot unambiguously determine the exact species of Escherichia/Shigella that was reduced in OPV revaccinated infants, since 16S rRNA sequencing analysis has limitations for species level discrimination. Hence, we cannot confirm a direct OPV effect against pathogenic bacteria in this taxon. R microbiota results are also in line with what has been previously described for nasopharyngeal colonization (Lanaspa et al., 2017). We observed in OPV revaccinated that significant changes in most abundant taxa were restricted to a decrease in the median proportion of an opportunistic Corynebacterium_1 species after OPV revaccination. In the opposite, several taxa increased in the Control group at follow-up, including potential causative genera of upper and lower respiratory infections, such as Streptococcus and Haemophilus. As observed above for changes in G microbiota after OPV revaccination, changes in R microbiota also suggest that OPV revaccination was associated with decrease in some potential opportunistic species under the Corynebacterium genus, without changing the dominance of this taxon, which is expected to be the dominant one in breastfeed infants. Also, our results suggest that OPV revaccination may limit the increase in potential pathogenic organisms of Streptococcus and Haemophilus genera, that are expected to colonize the oropharynx in later stages of life. Here, we presented changes in gut and upper respiratory microbiomes associated with OPV revaccination, characterized by the development of a healthier microbiome composition. Our study was too small to study the clinical implications, so, it remains to be studied whether the OPV-associated changes in bacterial microbiota of gut and upper respiratory tract are associated with beneficial NSE conferred by OPV revaccination. A similar experimental set up including more infants and detailed information on etiological diagnosis of specific clinical endpoints, species level differentiation at microbiota analysis and functional analysis of microbiome content using metabolomics would be helpful to elucidate this question. In conclusion, the Micro-OPV study showed that OPV revaccination vs. no revaccination was associated with changes in gut and upper respiratory bacterial microbiotas toward a healthier composition. This comprises a greater diversity and abundance of gut commensals potentially producers of immune modulatory molecules involved in trained immunity, such as, short-chain fatty acids, and a lower proportion of potential pathogenic/opportunistic genera for both microbiotas. The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/, BioProject ID: PRJNA773649. The studies involving human participants were reviewed and approved by Comité Nacional de tica em Saúde (CNES)-Guinea-Bissau. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. MMM, ABF, CSB, ML, and HS: conceptualization. MMM, ACI, MS, PSA, ABF, ML, and PA: methodology. MMM, ACI, MS, and PA: formal analysis. ABF, ML, and HS: funding acquisition. MMM, LMN, CC, ABF, and ML: investigation. MMM, ML, and PA: project administration. MMM, ML, HS, and PA: supervision. MMM and ACI: writing—original draft. LMN, MS, PSA, ABF, CSB, HS, and PA: Writing—review and editing. All authors have read and agreed to the published version of the manuscript.

PMC9649906

Evandro A. De-Souza,Nadia Cummins,Rebecca C. Taylor

IRE-1 endoribonuclease activity declines early in C. elegans adulthood and is not rescued by reduced reproduction 10.3389/fragi.2022.1044556

28-10-2022

UPR,unfolded protein response,proteostasis,cell stress,stress response,C. elegans,aging,IRE1

The proteome of a cell helps to define its functional specialization. Most proteins must be translated and properly folded to ensure their biological function, but with aging, animals lose their ability to maintain a correctly folded proteome. This leads to the accumulation of protein aggregates, decreased stress resistance, and the onset of age-related disorders. The unfolded protein response of the endoplasmic reticulum (UPRER) is a central protein quality control mechanism, the function of which is known to decline with age. Here, we show that age-related UPRER decline in Caenorhabditis elegans occurs at the onset of the reproductive period and is caused by a failure in IRE-1 endoribonuclease activities, affecting both the splicing of xbp-1 mRNA and regulated Ire1 dependent decay (RIDD) activity. Animals with a defect in germline development, previously shown to rescue the transcriptional activity of other stress responses during aging, do not show restored UPRER activation with age. This underlines the mechanistic difference between age-associated loss of UPRER activation and that of other stress responses in this system, and uncouples reproductive status from the activity of somatic maintenance pathways. These observations may aid in the development of strategies that aim to overcome the proteostasis decline observed with aging.

IRE-1 endoribonuclease activity declines early in C. elegans adulthood and is not rescued by reduced reproduction 10.3389/fragi.2022.1044556 The proteome of a cell helps to define its functional specialization. Most proteins must be translated and properly folded to ensure their biological function, but with aging, animals lose their ability to maintain a correctly folded proteome. This leads to the accumulation of protein aggregates, decreased stress resistance, and the onset of age-related disorders. The unfolded protein response of the endoplasmic reticulum (UPRER) is a central protein quality control mechanism, the function of which is known to decline with age. Here, we show that age-related UPRER decline in Caenorhabditis elegans occurs at the onset of the reproductive period and is caused by a failure in IRE-1 endoribonuclease activities, affecting both the splicing of xbp-1 mRNA and regulated Ire1 dependent decay (RIDD) activity. Animals with a defect in germline development, previously shown to rescue the transcriptional activity of other stress responses during aging, do not show restored UPRER activation with age. This underlines the mechanistic difference between age-associated loss of UPRER activation and that of other stress responses in this system, and uncouples reproductive status from the activity of somatic maintenance pathways. These observations may aid in the development of strategies that aim to overcome the proteostasis decline observed with aging. A fundamental biological question is why organisms lose their ability to sense and respond to stress as they age. This decline in the function of stress responses leads to a buildup of damaged macromolecules, in particular proteins. Proteins are molecules involved in virtually all the cellular reactions that make life possible, and a network of protein quality control pathways exists to ensure that they maintain their optimal conformation and function. With aging, these quality control processes become less efficient, leading to the accumulation of misfolded proteins and aggregates that are a signature of the onset of neurodegenerative diseases. A conserved hub in this protein homeostasis (proteostasis) network is the unfolded protein response of the endoplasmic reticulum (UPRER). The UPRER allows cells to sense and respond to the accumulation of misfolded proteins in the lumen of the endoplasmic reticulum (ER) (Walter and Ron, 2011). In metazoans, the UPRER can be activated through three distinct branches, controlled by the upstream regulatory molecules IRE1, PERK, and ATF6. The IRE1 branch is the only one conserved from yeast to humans, and plays a role in the aging process and in the pathophysiology of several diseases (Labbadia and Morimoto, 2015b). IRE1 has both kinase and endoribonuclease activities, and responds to proteostatic imbalance through the activation of several downstream mechanisms. At least two of these mechanisms depend upon the endoribonuclease activity of IRE1. These include the splicing of a specific regulated intron from the mRNA of a transcription factor, XBP1, which allows spliced XBP1s to be translated and to regulate a variety of target genes that promote proteostasis (Calfon et al., 2002). In addition, IRE1 is also able to degrade a variety of ER-localized transcripts through a process called regulated IRE1-dependent decay (RIDD), which reduces the protein folding load in the ER and may also play specific regulatory roles (Hollien and Weissman, 2006; Bae et al., 2019). The activation of the IRE1 branch of the UPR has been shown to decline with age in the model organism C. elegans, as well as in the murine brain and in human cells, but to date, the molecular mechanisms underlying this decline remain elusive (Ben-Zvi et al., 2009; Taylor and Dillin, 2013; Labbadia and Morimoto, 2015a; Cabral-Miranda et al., 2020; Sabath et al., 2020). The identification of the major mechanisms underlying this inhibition might aid in the design of novel interventions to counter age-associated loss of proteostasis and increase human healthspan. It was previously shown that the onset of reproduction affects the ability to transcriptionally activate stress-related genes in C. elegans (Labbadia and Morimoto, 2015a). An intriguing hypothesis to explain these results is that upon the onset of the reproductive stage, organisms deviate resources from the soma to maintain the integrity of embryos for the next generation (Maklakov and Immler, 2016). In agreement, animals lacking a germline are more resistant to stress and have extended lifespans (Hsin and Kenyon, 1999; Flatt et al., 2008; Labbadia and Morimoto, 2015a). Activation of the cytosolic heat shock response (HSR) is lost at the onset of reproduction and this loss occurs at the level of chromatin, with increased levels of H3K27me3 histone methylation rendering stress genes inaccessible to the HSR transcription factor HSF-1 (Labbadia and Morimoto, 2015a). This loss of HSR activation can be rescued by mutations that remove the germline stem cells, through rescued expression of the H3K27 demethylase jmjd-3.1, levels of which otherwise decline upon the onset of reproduction. In the same study, ER stress-induced expression of the UPRER target genes hsp-3 and hsp-4 were also found to decline during the first day of adulthood, suggesting that the ability to activate the UPRER also collapses very early in C. elegans aging. In addition, a recent study in a human fibroblast model of aging demonstrated that, in these cells, the ability of the UPRER transcription factor XBP1 to regulate its target genes declines with senescence while no loss of IRE1-regulated XBP1 splicing occurs (Sabath et al., 2020). This suggests that the decline in activation of this branch of the UPRER in senescent human fibroblasts may, like the loss of HSR activity in C. elegans, lie at the level of chromatin accessibility. However, it is not clear whether this is also true in non-dividing cells, such as the somatic cells of C. elegans, in which age-associated loss of UPRER activation has previously appeared to lie at the level of xbp-1 splicing (Taylor and Dillin, 2013). In addition, it is not known whether loss of reproductive capacity can delay this loss of UPRER activation in the worm, as it does the contemporaneous loss of HSR activation. We therefore set out to ask when and how the loss of IRE-1/XBP-1 pathway activation occurs in C. elegans. We also aimed to discover whether reduced germline function could rescue this decline. Our findings confirm that, as previously suggested, stress-induced UPRER activation declines within the first day of C. elegans adulthood. We also found that loss of IRE-1/XBP-1 pathway function occurs at the level of IRE-1 activation in this organism, upstream of any changes to chromatin accessibility. Finally, to our surprise we found that, unlike the HSR, age-associated loss of UPRER activation cannot be rescued by a mutation that prevents the formation of the germline. Together, this suggests that the aging-related loss of activation of the UPRER, while occurring at the same time as loss of HSR activation, may happen through a fundamentally different mechanism. This may have important implications for the development of methods to restore UPRER activation in older cells, in order to develop therapies that target the onset of diseases of aging. Worms were kept on NGM plates seeded with OP50-1 E. coli using standard procedures (Brenner, 1974; Imanikia et al., 2019a; Imanikia et al., 2019b). For assays with the glp-1(e2141) strain, to induce sterility, worms were synchronized and kept for 48 h at 25°C, and then moved back to 20°C. A list of the strains used in this work can be found in Supplementary Table S1. Worm images were obtained using a Leica M205 FA microscope. Worms were immobilized with 10 mM of sodium azide (Sigma Aldrich) before imaging. Fluorescence of worms was quantified using ImageJ. Unless otherwise stated, all RNAi clones used were from the Ahringer RNAi library (Kamath et al., 2003) and were sequenced before use. Bacteria containing RNAi vector were grown overnight in LB with 100 μg/ml carbenicillin (Formedium) and used for seeding NGM plates containing 1 mM IPTG (Generon) and 100 μg/ml of carbenicillin. Worms were lysed at the beginning of day 1 or day 2 of adulthood and samples normalized by protein content as measured by BCA assay. 30 μg of protein was loaded per lane for SDS-PAGE and gels were transferred using the iBlot system (Invitrogen) before probing with α-FLAG (Sigma Aldrich) or α-tubulin (Sigma Aldrich) followed by α-mouse-HRP (Abcam). Quantification of bands was carried out using ImageJ. Approximately 150 worms were collected per sample, and RNA extracted. Briefly, a 1:1 mixture of worms and glass beads was added together with a 3x volume of Trizol LS (Life Technologies). The sample was then quickly frozen in liquid nitrogen. Each sample was centrifuged (13,000 g for 1 min), supernatant collected, and an equal volume of ethanol added to the sample. Samples were then processed with a Direct-zol RNA Miniprep kit (Zymo Research). 1 μg of RNA was used for cDNA synthesis using the QuantiTect reverse transcription kit (QIAGEN). The qPCR run was performed in a Vii7 Real-Time PCR machine (ThermoFisher Scientific) and the data was analyzed using the comparative 2ΔΔCt method (Livak and Schmittgen, 2001). The sequence of the primers used in this work can be found in Supplementary Table S2. Worms were treated with 50 ng/μl of tunicamycin for 4 h, RNA extracted, and cDNA prepared as described above. 2 μl of cDNA was used in a PCR reaction to amplify xbp-1 spliced and unspliced products. Samples were run on a 2.5% agarose gel stained with SYBR Safe. Quantification of the percentage of spliced xbp-1 bands was done using ImageJ. The primer sequences used are listed in Supplementary Table S2. All experiments were performed at least three times. Bars represent the mean and error bars represent the SEM (standard error of the mean). For statistical analysis between two groups, unpaired Student’s t test or Mann-Whitney U test were used. For comparisons with more than two groups, One-Way ANOVA with Sidak’s or Tukey’s multiple comparisons tests was used. Activation of the IRE-1/XBP-1 branch of the UPRER (Figure 1A) is known to decline with age, but the exact timing of this decline is unclear. As the ability to activate the HSR has been shown to decrease very early in C. elegans adulthood, around the onset of the reproductive period (Labbadia and Morimoto, 2015a), we asked whether the same was true of UPRER activation. Animals expressing an hsp-4::GFP UPRER reporter transgene were treated with the N-linked glycosylation inhibitor tunicamycin, which induces ER stress, at the beginning of the first day of adulthood before egg laying had begun, or at the beginning of day 2 of adulthood. Specifically, animals treated at day 1 of adulthood were within the first 8 h post-L4 larval stage, with oocytes visible within the reproductive system but before the initiation of egg laying, while day 2 animals were 24 h older and actively laying eggs. Following tunicamycin treatment, fluorescence was assessed. In early day 1 of adulthood, UPRER activation was highly significant, whereas by day 2 it was no longer observable (Figures 1B,C). This suggests that, like the HSR, the ability to activate the UPRER is lost with the onset of reproduction. To confirm this finding, we also measured transcript levels of the XBP-1 target gene hsp-4 and levels of the spliced, active form of xbp-1, xbp-1s, at day 1 and day 2 of adulthood (as above) and found that, again, upregulation upon tunicamycin treatment at day 1 was lost by day 2 (Figures 1D,E). To ask whether age-associated loss of UPRER activation was specific to tunicamycin treatment, we also used genetic tools to activate the UPRER. Exposure of hsp-4p::GFP animals to RNAi against pdi-2, dnj-7 or sams-1 robustly activated the UPRER within 72 h when transfer to plates of RNAi-expressing bacteria occurred at the L1 stage of larval development (Supplementary Figure S1). Some activation was also seen upon transfer at the L3-L4 larval stage. However, transfer at adulthood proved too late to induce UPRER activation, suggesting that age-associated loss of UPRER activation is not specific to ER stress-inducing drug treatment. Finally, we asked whether UPRER activation by heat stress also declined with age. To our surprise, we found that the hsp-4p::GFP transgene was still robustly activated by heat stress at day 2 of adulthood (Supplementary Figure S2A). However, UPRER activation by heat stress proved to be independent of IRE-1 and XBP-1, suggesting that heat stress-induced hsp-4p::GFP activation is mediated through an alternative mechanism, and that only hsp-4p::GFP activation dependent upon IRE-1 and XBP-1 undergoes age-dependent decline at the onset of reproduction (Supplementary Figure S2B). Aged worms constitutively expressing a spliced version of xbp-1 are also capable of inducing hsp-4p::gfp past the age at which UPRER activation usually declines (Taylor and Dillin, 2013), suggesting that loss of UPR activation lies upstream of the transcriptional accessibility of target genes. The absence of spliced xbp-1s transcript following tunicamycin-induced UPRER activation at day 2 of adulthood (Figure 1E) suggests that it is the ability of IRE-1 to splice xbp-1 mRNA, a core function of this enzyme (Figure 2A), that is lost early in adulthood. We confirmed that xbp-1 splicing is lost at day 2 of adulthood, while unspliced xbp-1 is still readily detectable (Figures 2B,C). Given that xbp-1 is still transcribed and therefore available for splicing, one explanation for this is that the endoribonuclease function of IRE-1 can no longer be activated at this age. However, it is also possible that xbp-1 mRNA is no longer recruited to the ER once reproduction has begun, and is therefore not accessible to IRE-1. To distinguish between these possibilities, we asked whether the endoribonuclease-dependent ability of IRE-1 to degrade other ER-localized RNAs, a process known as regulated Ire1-dependent decay (RIDD) (Hollien and Weissman, 2006) (Figure 2A), is also lost by day 2 of adulthood. There is only one confirmed RIDD substrate in C. elegans, the neuropeptide flp-6 (Levi-Ferber et al., 2021). We asked whether flp-6 mRNA levels still decline upon ER stress induction at day 2 of adulthood, and found that this was no longer the case (Figure 2D). We also asked whether the C. elegans homologue of the mammalian RIDD substrate Blos1, blos-1, was degraded upon tunicamycin treatment, but found no evidence of reduced blos-1 transcript levels following tunicamycin exposure (Supplementary Figure S3) (Bae et al., 2019). The loss of flp-6 degradation coupled with the loss of xbp-1 splicing therefore suggests that both of the endoribonuclease functions of IRE-1 decline upon the onset of reproduction, lending weight to the possibility that IRE-1 itself fails to become active upon ER stress as cells age. We then asked whether this loss of IRE-1 activity might be due to reduced transcription of IRE-1, but we found that ire-1 transcript levels do not decrease between day 1 and day 2 of adulthood (Figure 3A). We also wondered whether levels of IRE-1 protein might decline with age. To investigate this, we used CRISPR-Cas9 to insert a 3xFLAG tag onto the N-terminus of IRE-1, and used Western blotting with α-FLAG to determine endogenous protein levels. We found no decrease in IRE-1 protein levels between day 1 and day 2 of adulthood that would explain the loss of IRE-1 activity between these ages (Figure 3B). Finally, we wondered whether reduced levels of spliced xbp-1 might result from a failure to re-ligate the xbp-1 mRNA following splicing by IRE-1, leading to degradation. However, we saw no reduction in transcription of the xbp-1 ligase rtcb-1, suggesting that the loss of spliced xbp-1 at day 2 of adulthood cannot be explained by reduced expression of this enzyme (Figure 3C). These data therefore suggest that IRE-1 protein is present in aged animals but can no longer mediate its endoribonuclease functions. We then turned our attention to whether this age-associated loss of IRE-1 activity can be prevented or reversed. It has been previously shown that glp-1 mutant animals, which lack a functioning germline, are still capable of activating the HSR at day 2 of adulthood, and have generally higher levels of stress resistance (Labbadia and Morimoto, 2015a). We observed wild type levels of xbp-1 splicing upon tunicamycin treatment in glp-1(e2141) animals at the beginning of day 1 of adulthood; however, by day 2, xbp-1 splicing had been completely lost (Figure 4A). In addition, no activation of hsp-4p::GFP was seen in glp-1(e2141) animals at day 2 (Figure 4B). Furthermore, reducing levels of the germline-regulated H3K27 demethylase jmjd-3.1 to mimic its decreased expression upon the onset of reproduction (Labbadia and Morimoto, 2015a), did not affect the ability of neuronally-overexpressed xbp-1s to regulate hsp-4p::GFP (Supplementary Figure S4). This suggests, surprisingly, that unlike other stress response pathways which also decline at the onset of the reproductive period, age-associated loss of UPRER activation cannot be rescued by eliminating the formation of the germline, and does not seem to depend upon changes in chromatin accessibility at target gene promoters. Finally, we asked whether IRE-1 activation after day 1 of adulthood could be rescued by treatment with a drug, IXA4, shown to enhance activation of mammalian Ire1 (Grandjean et al., 2020). Treatment with 200 μM IXA4 at the L3/L4 larval stage induced a modest degree of UPRER activation (Supplementary Figure S5A). However, treatment at day 2 of adulthood did not activate the UPRER (Supplementary Figure S5B). We also treated animals with tunicamycin in addition to IXA4, to ask whether the drug could potentiate UPRER activation in response to ER stress. While there was a trend towards higher hsp-4::GFP activation in this case, the difference was not statistically significant (Supplementary Figure S5B). This suggests that IXA4 treatment cannot itself activate IRE-1 in aged animals, nor can it unambiguously reverse the age-associated loss of UPRER activation upon ER stress in this system. The loss of IRE-1 activation with age in C. elegans therefore remains refractory to our efforts to restore it. These findings reveal that the ability to activate the UPRER is lost at the beginning of the reproductive period in C. elegans; that this arises from a failure to activate the regulatory enzyme IRE-1, leading to a loss of both xbp-1 splicing and RIDD activity; and that this loss cannot be rescued by a failure to develop a germline. These results have several implications. This work as well as previous studies shows that the onset of reproduction is a time of general stress response decline in C. elegans, suggesting that this period of early adulthood plays a key role in aging (Labbadia and Morimoto, 2015a). This is consistent with germline/soma tradeoff models that explain the evolution of aging by the diversion of resources from somatic maintenance in order to focus an organism’s efforts on the production of progeny (Maklakov and Immler, 2016). A corollary of this is that preventing reproduction can delay aging by allowing energy to be invested instead into the maintenance of the soma. However, we show here that loss of UPRER activation cannot be delayed by inhibiting the development of the germline, uncoupling the loss of somatic maintenance from investment in progeny production. It also demonstrates that the mechanistic change underlying UPRER decline cannot be rescued by the endocrine signals that arise from the proliferating germline stem cells in glp-1 mutants, differentiating it from the HSR, activation of which is restored in glp-1 animals (Arantes-Oliveira et al., 2003; Berman and Kenyon, 2006; Labbadia and Morimoto, 2015a). The mechanistic basis for loss of UPRER activation in the somatic cells of C. elegans seems to lie upstream of the transcriptional activation of UPRER targets. This again differentiates it from the HSR, in which loss of chromatin accessibility at promoters of target genes explains the loss of HSR activation at the beginning of the reproductive period (Labbadia and Morimoto, 2015a). Changes in chromatin structure have also been proposed to explain the loss of UPRER activation in a human fibroblast model of aging (Sabath et al., 2020). In this human cell model, Xbp1 is still spliced following ER stress in aged cells, but downstream target genes are no longer upregulated. In C. elegans, however, XBP-1 target genes can still be regulated in aged animals when xbp-1s is overexpressed, suggesting that the promoters of these genes remain accessible, and therefore that age-associated failure to activate the UPRER in these animals involves an upstream event (Taylor and Dillin, 2013). Indeed, we show here that xbp-1 splicing itself is lost in early adulthood, as is RIDD activity against flp-6 (Figures 2B,C). Interestingly, this is the first description of a decline in RIDD activity during early adulthood; as RIDD is relevant to several aspects of mammalian physiology, including adaptive immunity, lipid metabolism, cellular differentiation, and insulin regulation (Han et al., 2009; Lee et al., 2011; So et al., 2012; Osorio et al., 2014; Wang et al., 2018), this loss of RIDD activity with age has significant implications for mammalian healthspan. Together, these findings strongly suggest a failure in IRE-1 activation with age, which cannot be explained by loss of IRE-1 expression (Figure 3). This raises the possibility that the mechanisms underlying age-associated loss of UPRER activation might vary, either between species, or in dividing (e.g., fibroblasts) vs. non-dividing cells (e.g., somatic cells of C. elegans). This latter idea is supported by recent work showing that the induction of Xbp1 splicing by tunicamycin exposure is partially compromised in the hippocampus of aged mice (Cabral-Miranda et al., 2020). However, the reasons for this variation are unclear, as is the mechanistic basis for loss of IRE-1 activation in C. elegans cells. In addition, how these different mechanisms that lead to the collapse of proteostasis pathways are orchestrated to occur during the same narrow window in early adulthood remains an open question. One possibility is that these events may share the same underlying molecular cause; for example, alterations in the redox environment in C. elegans can perturb both IRE-1 RNAse activity and histone methylation events (Hourihan et al., 2016; Bazopoulou et al., 2019). Understanding this underlying mechanism would facilitate the design of therapeutics to reactivate Ire1, which may have applications in treating age-associated disease, including neurodegenerative disorders that are associated with dysregulated UPRER activation (Hetz and Mollereau, 2014). Neither strategy that we deployed to delay or restore UPRER activation in aged animals–germline inhibition or treatment with the Ire1-activating drug IXA4 (Grandjean et al., 2020)—significantly rescued UPRER activation in animals at day 2 of adulthood. As IXA4 was identified through screening based on activation of mammalian Ire1, which has notable differences from C. elegans IRE-1, it is possible that the drug is either not delivered to the relevant cells in our assays, or that it is not activating C. elegans IRE-1 as effectively as mammalian Ire1. Understanding the molecular basis for the inactivation of this enzyme would allow the design of more targeted approaches—for example, a failure in kinase domain activation would suggest the trial of specific kinase activating molecules. Finally, a previous study showed that a mutation in eat-2, proposed to act as a mimetic of dietary restriction, is capable of partially rescuing the age-related decline in UPRER activation (Matai et al., 2019). This suggests that studying downstream effectors of the eat-2 pathway, such as the transcription factors PHA-4 and SKN-1, might help us to understand the basis of age-related decline in IRE-1 activity and how it might be mitigated (Park et al., 2010; Glover-Cutter et al., 2013; Hourihan et al., 2016; Matai et al., 2019). Future work to identify the mechanism underlying molecular failure of IRE-1 in older cells, the mechanistic basis of UPRER rescue, and the relevance of these mechanisms in non-dividing human cells such as neurons, is likely to be of significant interest in tackling age-related disease.

PMC9649909

36386148

Cédric Boudou,Luce Mattio,Alexey Koval,Valentin Soulard,Vladimir L. Katanaev

Wnt-pathway inhibitors with selective activity against triple-negative breast cancer: From thienopyrimidine to quinazoline inhibitors 10.3389/fphar.2022.1045102

28-10-2022

Wnt signaling,triple-negative breast cancer,β-catenin,cancer survival,medicinal chemistry,structure activity relationship,thienopyrimidine,quinazoline

The Wnt-pathway has a critical role in development and tissue homeostasis and has attracted increased attention to develop anticancer drugs due to its aberrant activation in many cancers. In this study, we identified a novel small molecule series with a thienopyrimidine scaffold acting as a downstream inhibitor of the β-catenin-dependent Wnt-pathway. This novel chemotype was investigated using Wnt-dependent triple-negative breast cancer (TNBC) cell lines. Structure activity relationship (SAR) exploration led to identification of low micromolar compounds such as 5a, 5d, 5e and a novel series with quinazoline scaffold such as 9d. Further investigation showed translation of activity to inhibit cancer survival of HCC1395 and MDA-MB-468 TNBC cell lines without affecting a non-cancerous breast epithelial cell line MCF10a. This anti-proliferative effect was synergistic to docetaxel treatment. Collectively, we identified novel chemotypes acting as a downstream inhibitor of β-catenin-dependent Wnt-pathway that could expand therapeutic options to manage TNBC.

Wnt-pathway inhibitors with selective activity against triple-negative breast cancer: From thienopyrimidine to quinazoline inhibitors 10.3389/fphar.2022.1045102 The Wnt-pathway has a critical role in development and tissue homeostasis and has attracted increased attention to develop anticancer drugs due to its aberrant activation in many cancers. In this study, we identified a novel small molecule series with a thienopyrimidine scaffold acting as a downstream inhibitor of the β-catenin-dependent Wnt-pathway. This novel chemotype was investigated using Wnt-dependent triple-negative breast cancer (TNBC) cell lines. Structure activity relationship (SAR) exploration led to identification of low micromolar compounds such as 5a, 5d, 5e and a novel series with quinazoline scaffold such as 9d. Further investigation showed translation of activity to inhibit cancer survival of HCC1395 and MDA-MB-468 TNBC cell lines without affecting a non-cancerous breast epithelial cell line MCF10a. This anti-proliferative effect was synergistic to docetaxel treatment. Collectively, we identified novel chemotypes acting as a downstream inhibitor of β-catenin-dependent Wnt-pathway that could expand therapeutic options to manage TNBC. The β-catenin-dependent Wnt-pathway regulates various physiological processes, from embryonic development to tissue homeostasis, and regeneration in adults (Parsons et al., 2021). However, aberrant activation of the Wnt-pathway plays a critical role in cancer cell proliferation, survival, and metastasis as well as in maintenance of cancer stem cells. This overactivation is predominantly driven by mutations, such as loss-of-function mutations for negative regulators like APC, AXIN1 or AXIN2, gain-of-function mutations for β-catenin (CTNNB1) or TCF, and eventually dysregulation of Wnt receptor abundance through mutations of RNF43, ZNRF3 or RSPO (Bugter et al., 2021; Parsons et al., 2021). In addition, several cancers have been linked to overexpression of the Wnt-pathway components, such as Wnts, their receptors (FZDs and LRP5/6), or Porcupine—the enzyme regulating Wnt secretion and activity (Stewart, 2014; Zhan et al., 2017; Shaw et al., 2019a; Wang et al., 2021). As a result of such aberrations, β-catenin accumulates in cell nucleus and promotes transcription of oncogenic target genes. Since the discovery of the first WNT family member in 1982, Wnt signaling has attracted increased attention in the drug discovery field. Targeting the β-catenin-dependent pathway has been the focus of multiple drug discovery programs both in academia and industry, but only a few have reached clinical trials (Blagodatski et al., 2014; Liu et al., 2021). Among them, there are five small molecules acting as Porcupine inhibitors (e.g., 1 and 2, in Figure 1), which are profiled in phase 1/2 clinical trials for different cancer indications (Flanagan et al., 2022), while CWP232291, PRI-724, and SM08502 target downstream components of the Wnt-pathway (Shaw et al., 2019a). CWP232291 is a peptidomimetic targeting Sam68, an RNA-binding protein, and is currently profiled on acute myeloid leukaemia in phase 1/2. PRI-724 (3, in Figure 1) affecting interaction between β-catenin and the transcription co-activator CBP, has been evaluated in clinical trials against multiple cancers and cirrhosis. Finally, SM08502, acting through CDC-like kinase, reduces the Wnt-pathway-related gene expression and is profiled in phase 1 (Tam et al., 2020). Our research has been aimed at discovering novel Wnt-pathway inhibitors (Koval and Katanaev, 2012; Katanaev et al., 2021; Larasati et al., 2022). We have previously developed a High-Throughput Screening (HTS) assay based on a triple-negative breast cancer (TNBC) cell line BT-20 that could be adapted to any β-catenin-dependent cancer cell line. In this assay, BT-20 cells are stably transfected with the Wnt-responsive TopFlash reporter, and activated by the in-house purified Wnt3a (Shaw et al., 2019b). A pilot screening of a commercial library of 1,000 compounds led to the identification of FSA as selective inhibitor of the Wnt signaling and cancer cell proliferation, in vitro and in vivo (Katanaev and Koval, 2021). This assay has also been used to drive the drug design of clofazimine derivatives acting as downstream inhibitors of the Wnt-pathway (Koval et al., 2021). In continuation of our efforts to identify novel chemotypes acting as β-catenin-dependent Wnt inhibitors and potential anticancer drugs, we here report a novel chemical series leading to compounds endowed with submicromolar activity in cancer cell lines, inhibiting the β-catenin-dependent Wnt signaling at downstream levels of the pathway. The advantage of our platform lies in its ability to discriminate between inhibitors of the β-catenin-dependent Wnt signaling acting at upstream vs. downstream levels of the pathway, by stimulating the pathway at the different levels either with Wnt3a or with a GSK3β inhibitor, like CHIR99021 (Shaw et al., 2019b). Both stimulations abolish formation of the destruction complex, a multiprotein complex consisting of Axin, adenomatous polyposis coli (APC), glycogen synthase kinase three beta (GSK3β), and casein kinase one alpha (CK1α). Upon inactivation of the complex, β-catenin accumulates in the cytoplasm and translocates into the nucleus where it interacts with transcription factors. These downstream events are quantified by a Luciferase transcription-based readout assay (TopFlash assay). During our investigation toward novel Wnt-inhibitors, we identified a thienopyrimidine 4a as a β-catenin-dependent Wnt-pathway inhibitor with micromolar potency in the TNBC cell line HCC1395. Both means of the pathway activation—at the upstream levels with Wnt3a and at the downstream levels with a GSK3β inhibitor CHIR99021—revealed a similar potency of 4a to inhibit the signaling, respectively 8.31 µM and 8.47 µM (Figure 2). This result led to the conclusion that 4a acts downstream of GSK3β in the Wnt-pathway. We considered 4a as a good starting point for a drug discovery exploration. The compound has no cytotoxic activity (measured by Renilla luciferase expressed under the Wnt-independent CMV promoter, Figure 2) and has a good potency with respect to its low molecular weight (241 g/mol). The simplicity of the scaffold combined with synthetic accessibility allows building a library of compounds and performing structure activity relationship (SAR) studies of the three main parts of the molecule: the top part, the linker and the core (Figure 2). The synthesis of thieno [2,3-d]pyrimidine derivatives is illustrated in Scheme 1. Commercially available 4-chlorothieno [2,3-d]pyrimidine was subjected to aromatic substitution with different primary and secondary amines to afford the benzyl derivatives (4a-e, 4g, 4i-l and 4o-q) and some of the linker replacement 5a-n in the 29%–81% yield. In some cases, formation of side products did not permit purification of the final compounds, or the corresponding amines were not available. Therefore, as an alternative approach, 4-chlorothieno [2,3-d]pyrimidine was converted into 4-aminothieno [2,3-d]pyrimidine (compound 6) with ammonium hydroxide in the 75% yield without purification. This second building block was reacted with differently substituted benzyl bromides or benzoyl chloride to afford compounds 4f, 4k, 4m, and 7 in moderate yields. With these synthetic approaches, it was possible to obtain a series of analogues of 4a modified in the top part and in the linker portions (Figure 2). To perform core modifications, the thienopyrimidine nucleus was replaced with a series of heterocycles. For the quinazoline core, the synthetic approach used was similar to the one used for the thienopyrimidine derivatives. Commercially available 2,4-dichloroquinazoline was subjected to the nucleophilic aromatic substitution with different primary amines to afford the intermediates 8a-i in moderate to good yields (Scheme 2). Interestingly, although both C2 and C4 positions are electron-deficient, the reaction is regioselective toward the C4 position (Mohamed and Rao, 2015). In the second step, chlorine in position C2 was removed using reductive conditions. For intermediates 8a-f, 8f and 8h-i, ammonium formate in presence of palladium on charcoal at reflux was used to afford the expected compounds in the 11%–74% yield. For compounds 8e and 8g, bearing chlorine on the aromatic ring in the top part, a regioselective reduction of the chlorine in C2 of the quinazoline derivatives had to be performed. Therefore, compounds 9e and 9g were obtained in milder conditions with zinc, N,N,N′,N′-tetramethyl ethylenediamine, acetic acid in methanol at 45°C overnight. Preparation of pyrimidines (11 and 12), isoquinoline, 13 and quinoline 16 derivatives followed similar synthetic strategies and are described in the Supplementary Material. The synthesized compounds were used to explore SAR of three defined areas of investigation (Figure 2). All compounds were tested in the HCC1395 TNBC cell line using the GSK3β inhibitor to induce β-catenin signaling since 4a acted downstream of GSK3β (Figure 2). Both inhibition of β-catenin signaling measured by the TopFlash reporter and cytotoxicity measured by the Renilla reporter were assessed. SAR of benzyl substitution is depicted in Table 1. Overall, ortho substituents were not tolerated or resulted in toxicity, like the 2-fluoro derivative 4e. The only exception is the 2-methoxy analogue 4b, which showed a 3-fold improvement in the activity (IC50 = 2.79 µM) compared to the unsubstituted benzyl derivative 4a (IC50 = 8.31 µM), although with a partial inhibition of 40.7% compared to 72% of the parental compound. In the meta position, electron withdrawing groups were not tolerated (compounds 4f, 4i), except for compound 4i bearing chlorine, which resulted in the activity similar to that of 4a. However, potency was modestly improved by introducing electron donating groups, such as methoxy (4c) or methyl (4n) groups. In the para position, electron donating (methoxy, 4d) or withdrawing (cyano, 4m) groups abrogated the activity, whereas fluorine 4g, chlorine 4j and methyl 4o groups showed a 2-fold improvement in the activity compared to 4a. Therefore, derivatives combining the substituents which gave the best results in the inhibitory activity were designed and synthesized. Pleasingly, compounds 4p and 4q, bearing a methoxy group in the ortho position and fluoro or chloro in the para position, respectively, resulted in the IC50 of 2.59 µM and 1.33 µM. To investigate contribution of the linker between the thienopyrimidine core and the aromatic rings, we initially probed the influence of an additional methyl group, preparing a series of 1-phenylethylamine analogues (Table 2). Except for the 3-methoxy analogue 5b, which was inactive, the other derivatives resulted in 2 to 7-fold more potent compounds. In particular, 5d (4-chloro), 5e (4-methyl) and 5a (2-methoxy) showed submicromolar activities of 0.6, 0.63, and 0.77 µM, respectively. Unfortunately, this modification of the linker resulted in partial inhibition (<50% efficacy). Polar groups such as an amide linker (Liu et al., 2021) were not tolerated (Table 3). Alkylation of the amino group provided different results with benzyl or 2-phenylethan-1-amine top parts. Addition of ethyl group on compound 4c resulted in toxicity at 8.30 µM with compound 5i, whereas addition of a methyl group on compound 5g resulted in the similar potency with compound 5j (IC50 = 2.04 µM, 100% efficacy) (Table 3). This loss in potency could derive from different SAR between benzyl and 2-phenylethan-1-amine. Then, we investigated the impact of the linker length (Table 3) and found that compounds without a linker between the core and the top part, like 5f, or bearing a three-carbon chain (5h) were inactive. On the other side, a two-carbons linker (5g) was three times more potent than 4a with 89% of inhibition (IC50 = 2.64 and 8.31 µM respectively). Attracted by the potency of 5g, we decided to introduce several in-house available substituted 2-phenylethan-1-amines. As for the benzyl derivatives, chlorine substitution was not tolerated in the ortho position (5k, IC50 = 42.87 µM), but resulted in good potency in the meta (5l, IC50 = 3.07 µM) and para (5l, IC50 = 1.79 µM) positions. Unlike the benzyl analogue 4c, the methoxy group in the meta position was not tolerated (5n). We finally evaluated replacement of the core and, without surprise, the bioisosteric replacement of thieno [2,3-d]pyrimidine by quinazoline was well tolerated (Table 4). Other modifications of the bottom part led to drastic loss of potency (Table S1). Substituted quinazolines with 2-chloroquinazoline (8a), 7-dimethylthieno [3,2-d]pyrimidine (10) and single heteroaromatics 2-chloropyrimidine (11) or pyrimidine (12) were no longer active. Moreover, we observed that both nitrogens of the pyrimidine ring from quinazoline were necessary for the activity, since the isoquinoline 13 and quinoline 16 derivatives were inactive or toxic at 18.67 µM. Attracted by the potency of quinazoline 9b with IC50 of 3.58 µM compared to 4.96 µM for thieno [2,3-d]pyrimidine 4c, we evaluated SAR transfer synthesizing quinazoline derivatives bearing the best modifications obtained from the thieno [2,3-d]pyrimidine analogues previously described. With few exceptions, the resulting compounds had similar or better potencies and better efficacies (Table 4). For substituents of the phenyl ring, compounds bearing a substituent in the para position showed a good activity: 4-methoxy was active for the quinazoline version (9c, with IC50 = 4.76 µM, 81.5%) and 4-fluoro (9b), 4-chloro (9e) and 4-methyl (9f) were more potent with IC50 of 0.67, 2.18, and 1.22 µM, respectively. Also, with the quinazoline core, a longer linker did not lead to activity improvement, as we observed with the 4-chlorophenylethyl derivative 9g, found to be inactive. On the other hand, addition of methyl to provide 1-phenylethylamine derivatives resulted either in a drastic reduction of activity (9h), or in compounds with similar activities, if we compare 9f with 9i with 1.22 µM and 1.88 µM, respectively. In general, this investigation of SAR of the three areas of the hit compound led to up to 10-fold improvements from the initial compound 4a (IC50 = 8.31 µM) to compound 9d with IC50 of 0.67 µM. To assess the translation of the compounds’ anti-Wnt potency—ranging from two-digits micromolar to submicromolar values—into inhibition of cell proliferation, we applied the classical MTT assay. As illustrated in Figure 3A, there was a good correlation for HCC1395 cells between the TopFlash and MTT assays with r = 0.648. Interestingly, the partial efficacy observed for some compounds in the TopFlash assay appeared not to influence the potency in the MTT survival assay. We further evaluated whether our compounds could inhibit proliferation of a large panel of TNBC cell lines and of the non-cancerous breast fibrocystic MCF10a cells (Figure 3B and Supplementary Table S2). We analyzed the obtained set of IC50’s by independent clustering that resulted in a clear separation of the cell lines tested into 2 groups: highly sensitive ones (HCC1395 and MDA-MB-468 lines) and poorly sensitive ones (BT20, HCC1806 and MDA-MB-231). Cell lines with high susceptibility to the compounds’ treatment systematically demonstrated lower IC50s as compared to the others. Those with the low antiproliferative sensitivity also clustered with the control non-cancerous MCF10a cells, which indicates strong selectivity of the compounds towards a subset cancer cells. In this regard, only few compounds such as 5j, 5h and 8a (Figure 3A in red, see Supplementary Table S2) were considered to have promiscuous effects with no selectivity over the control cells (less than 3-fold difference in IC50), although 5h should rather be considered as inactive (IC50 > 50 µM). We continued by analysis of the compounds’ capacity to suppress Wnt signaling in TNBC cells independently of an external Wnt ligand and beyond the artificial TopFlash reporter system. To this end, we assessed the expression levels of the transcription factor c-Myc, a well-proven key target gene of the pathway (Rennoll and Yochum, 2015; Koval et al., 2018; Ahmed et al., 2019). Effects of the compounds on c-Myc levels in HCC1395 cells upon treatment by the indicated compounds, chosen based on their higher potency (IC50 between 0.6 µM and 2.59 µM in the TopFlash assay), are summarized in Figure 4. As expected, Wnt inhibition by the compounds resulted in profoundly decreased endogenous c-Myc levels. Compound 5d (IC50 = 0.6 µM, Table 2) was used as a representative to analyze the selectivity of the chemotype towards the Wnt-pathway, by profiling it against a panel of different transcriptional reporters with significant basal activities in HCC1395 cells (Figure 5). These results demonstrate that none other signaling pathways are affected by the compound. We thus conclude that the anti-proliferative effect of the compound is related exclusively to the suppressed Wnt signaling. The fact that some TNBC cell lines were highly sensitive to the compounds, while others were much less so (see Figure 3), brought us to hypothesize that different gene expression determinants existed between these two sets of cancer lines. We took advantage of the RNAseq-derived gene expression datasets in CCLE (Cancer Cell Line Encyclopedia (Barretina et al., 2012)) available for the five TNBC lines used in our studies. In order to pinpoint the genes most likely responsible for the differential response among the two groups of cell lines, we extracted 419 genes that had at least 4-fold higher or lower average expression levels in the HCC1395 and MDA-MB-468 lines as compared to the other cell lines. Pathway enrichment analysis of this list performed with the DAVID online tool (Huang et al., 2009; Sherman et al., 2022) identified the Wnt-pathway among the top over-represented pathways in this list, with 10 genes (Figure 6A and Supplementary Table S3). Indeed, expression levels of these 10 Wnt component and target genes cluster the 5 TNBC cell lines into the same two groups as by the potency to the compound treatment (Figure 6B). Plotting StringDB (Szklarczyk et al., 2021) interactions (physical and genetic) among these genes place them in a tightly interconnected network (Figure 6C). Taken together, these data indicate that the differential capacity of the compounds to inhibit proliferation of these lines is driven by peculiarities of the Wnt-pathway organization in them—the phenomenon previously predicted by us (Koval and Katanaev, 2018). The Wnt-pathway is implicated in chemoresistance to conventional chemotherapies and drives a poor chemotherapy response in a broad range of tumors (O’Reilly et al., 2015; Martin-Orozco et al., 2019; Zhong and Virshup, 2020). To assess how combination with our compounds affects the anti-proliferative capacity of docetaxel—one of the first-line treatments for TNBC (Mandapati and Lukong, 2022), we performed the Loewe synergism assessment over a complete dose-response range of both compounds followed by analysis in the COMBENEFIT software in Matlab (Figure 7 and Supplementary Figure S1) (Di Veroli et al., 2016). The weighted summary synergism and antagonism scores indicate that compounds 4p, 5a, 5m, 9d, 9f, and 9i exert exclusively synergistic effects when used as a co-treatment with docetaxel. This effect is overall similar among the compounds (best for 9i) and can be characterized as mild since it scores 15–30 on a scale of 0–100. The dose-by-dose matrices calculated for each compound presented at Supplementary Figure S1 indicate that for most of the compounds, major zones of synergism can be seen at the highest concentrations of our compounds and intermediate concentrations of docetaxel. This serves as a strong indication that application of these compounds in vivo might be useful in combination with a conventional chemotherapy through reduction of the chemotherapy dose necessary to obtain a complete response. From the initial thienopyrimidine hit with an IC50 of 8.31 µM, SAR exploration of three portions of the scaffold led to a 10-fold improvement with four compounds active below the µM range in HCC1395 cells. Although further SAR exploration would be necessary to fully understand the properties driving potency in the HCC1395 cell-based assay, elements of SAR were identified. From exploration of the top part (Table 1, Table 2 and Table 4), electron density of the phenyl does not seem to be critical as electron withdrawing or donating groups have little influence on the potency. Lipophilic or polar interactions are more important in this region. Also, for the thieno [2,3-d]pyrimidine derivatives, the most potent compounds are 1-phenylethylamine analogues (compounds 5a, 5d and 5e; Table 2), of which only racemic mixture were tested. It would be interesting to prepare and test each enantiomer. For the core modifications, a better activity was achieved with the quinazoline nucleus replacing the thienopyrimidine core. We also observed some differences between thieno [2,3-d]pyrimidine and quinazoline SAR (Table 1 and Table 4). Although we expected a good shape similarity for both cores, we could speculate that the subtle electron density difference resulting from the bioisosteric replacement of the thiophene part by phenyl affects the interaction with the biological target. Based on the assay paradigm, we could conclude that these compounds act below the destruction complex within the Wnt-pathway, as they inhibited the β-catenin-dependent signaling induced by an inhibitor of GSK3β, CHIR99021. Moreover, our exploration of the genetic differences between the cancer lines with strong and weak proliferative response to the derivatives highlighted a set of the Wnt signaling component genes likely responsible for this effect, where TCF7 (a.k.a TCF-1) and TCF7L1 (a.k.a TCF-3) and their immediate regulators and co-factors TLE4 and RUNX3 occupy the central place. Taken together with a significant evidence of non-redundant relationships among TCF/LEF transcriptional factors (Arce et al., 2006; Mao and Byers, 2011), it beckons us that the molecular target of our compounds might be hidden among unique partners of these transcriptional factors. This proposition is also reinforced by the clear specificity of action for our compounds to Wnt signaling with no significant inhibition of other signal transduction pathway observed in the HCC1395 cell line. The search for such factors—molecular targets of our compounds—is essential for further development of this promising series of compounds. The selective mechanism of action of our compounds might be a double-edged sword: while it is likely behind their capacity to avoid strong effects on healthy cells represented by MCF-10a in our study, on the other hand it might also limit the scope of therapeutic applications of the compounds. With proliferation of 2 out of 5 TNBC lines strongly affected by the compounds, deciphering of the molecular target and of the full mechanism of action will be a key for evaluation of their scope within TNBC and other cancers, as well as for eventual patient stratification for the application of this novel therapy. Immortalized cancer cell lines as well as 2D cultures have certain limitations in translation of the results obtained to a more complex 3D milieu of the tumor in the patients. However, our findings represent a promising step forward in the development of the long-desired Wnt-targeting agents. Future investigations will follow the logical steps of the preclinical development using animal models, and should further provide interesting insights into the roles of Wnt signaling in the tumor resistance, given the synergistic relationships between our compounds and the conventional chemotherapeutic docetaxel. From the initial potency of the thienopyrimidine 4a with IC50 of 8.31 µM on the HCC1395 cell line, investigation of three portions of the molecule to generate a library of compounds led to identification of submicromolar inhibitors. Through orthogonal readouts, we confirmed that these thienopyrimidine derivative compounds act as downstream inhibitors of the β-catenin-dependent Wnt-pathway. Targeting downstream components would be more efficacious for the tumors where the pathway is induced by loss-of-function mutations in APC or Axin or by gain-of-function mutations in β-catenin or TCF, as compared to the upstream pathway inhibitors such as Porcupine inhibitors. Inhibition of the Wnt-pathway is translated into differential efficacy on inhibition of TNBC cell proliferation. Although we have not yet identified the exact molecular target, our analyses indicate the target to be among the partners of β-catenin-dependent transcriptional factors. Identification of the molecular target would unlock further development through the target-based assay to support further SAR exploration to identify compounds suitable for in vivo evaluation. Such findings will also be key to define the scope of application, determine potential biomarkers, and permit patient stratification. The Wnt3a or CHIR99021-stimulated luciferase activity in HCC1395 TNBC cell line stably transfected with the TopFlash reporter was analyzed essentially as described for the BT-20 TNBC cell line (Shaw et al., 2019b; Koval et al., 2021). Reporter cells were seeded at 10,000 cells/well in a white opaque 384-well plate in the final volume of 20 μl of DMEM medium supplemented with 10% FCS. The cells were maintained incubated at 37°C, 5% CO2 overnight for attachment. Subsequently, they were transfected by a plasmid encoding constitutively expressed Renilla luciferase under the CMV promoter overall as described in the manufacturer’s protocol using 12 μg/ml of DNA and 40 μl/ml XtremeGENE nine reagent. The next day, the medium in each well was replaced with a 10 μl of fresh medium containing compound of interest, and, after 1 h of preincubation with the compounds, additional 10 μl of medium supplemented with Wnt3a [purified as described (Willert et al., 2003; Koval and Katanaev, 2011)] or CHIR99021 were added creating respective final concentrations of 2.5 μg/ml or 5 μM of each. In case of other signal transduction reporters, no additional stimulation was performed. Compound dilutions were prepared by serial dilution in DMSO and diluted with the amount of medium necessary to obtain their final concentrations indicated on the figures and tables and maintain concentration of DMSO of 0.5% in all assay points. After overnight incubation, the supernatant in each well was removed, and the luciferase activity was measured as described (Dyer et al., 2000; Huber et al., 2021; Pellissier et al., 2021). The culture medium was completely removed from all wells of the plate. Finally, the luciferase activity of the firefly and Renilla luciferases was detected sequentially in individual wells of a 384-well plate through injection of corresponding measurement solutions and immediate reading (400 ms integration time) in Infinite M Plex multifunctional plate reader with injection module. Indicated TNBC cell lines were detached and resuspended at 120,000 cells/ml and added into each well of a transparent 384-well plate in the final volume of 20 µl/well. The cells were maintained in DMEM containing 10% FBS at 37°C, 5% CO2 overnight. Next day, the medium in each well was replaced by 40 μl of the fresh one containing the indicated concentrations of compounds. In case of drug combination experiments, each compound dilution was delivered in corresponding well in 20 μl of medium. Compound dilutions were prepared by serial dilution in DMSO and diluted with the amount of medium necessary to obtain their final concentrations indicated on the figures and tables and maintain concentration of DMSO of 0.5% in all assay points. After incubation for 3–4 days, depending on the cell line, the medium in each well was replaced by 25 μl of 0.5 mg/ml Thiazolyl blue solution in 1xPBS following by incubation for 3 h at 37°C. Then the solution was removed and 25 μl DMSO was added into each well. Absorbance at 510 nm was measured in a Tecan Infinite M200 PRO plate reader. HCC1395 TNBC cell line was seeded at 100,000 cells/well in 24 well plates. On the next day, the medium was replaced by the fresh one pre-warmed at 37°C containing the indicated compounds. After 24 h incubation, the medium was removed, followed by washing twice with 500 μl of 1x PBS. The cells were lysed in the well by addition of 30 μl of ice-cold RIPA buffer (1x TBS, 4 mM EDTA, 1% Triton, 0.1% SDS, 1x Protease inhibitor cocktail (Roche)) and incubated on ice for 10 min. The samples were resuspended and then centrifuged at 18,000 g at 4°C to remove debris. 15 μl of the supernatants each were further analyzed by Western blot with antibodies against c-Myc (Abcam) and α-Tubulin (Sigma) at 1:1,000 dilutions. The data was visualized analyzed in either GraphPad Prism nine using indicated statistical analysis or in R using gplots package (Warnes et al., 2009). Dataset from CCLE (Barretina et al., 2012) were used in the study and pathway enrichment analysis was performed by DAVID online tool (Sherman et al., 2022). Drug combination data was treated in COMBENEFIT software (Di Veroli et al., 2016). Network was built using Cytoscape with StringDB plugin (Shannon et al., 2003; Szklarczyk et al., 2021).

PMC9649917

Rui-Jun Bai,Yu-Sheng Li,Fang-Jie Zhang

Osteopontin, a bridge links osteoarthritis and osteoporosis

28-10-2022

bone metabolism,inflammation,osteoarthritis,osteoporosis,osteopontin

Osteoarthritis (OA) is the most prevalent joint disease characterized by degradation of articular cartilage, inflammation, and changes in periarticular and subchondral bone of joints. Osteoporosis (OP) is another systemic skeletal disease characterized by low bone mass and bone mineral density (BMD) accompanied by microarchitectural deterioration in bone tissue and increased bone fragility and fracture risk. Both OA and OP are mainly affected on the elderly people. Recent studies have shown that osteopontin (OPN) plays a vital role in bone metabolism and homeostasis. OPN involves these biological activities through participating in the proliferation, migration, differentiation, and adhesion of several bone-related cells, including chondrocytes, synoviocytes, osteoclasts, osteoblasts, and marrow mesenchymal stem cells (MSCs). OPN has been demonstrated to be closely related to the occurrence and development of many bone-related diseases, such as OA and OP. This review summarizes the role of OPN in regulating inflammation activity and bone metabolism in OA and OP. Furthermore, some drugs that targeted OPN to treat OA and OP are also summarized in the review. However, the complex mechanism of OPN in regulating OA and OP is not fully elucidated, which drives us to explore the depth effect of OPN on these two bone diseases.

Osteopontin, a bridge links osteoarthritis and osteoporosis Osteoarthritis (OA) is the most prevalent joint disease characterized by degradation of articular cartilage, inflammation, and changes in periarticular and subchondral bone of joints. Osteoporosis (OP) is another systemic skeletal disease characterized by low bone mass and bone mineral density (BMD) accompanied by microarchitectural deterioration in bone tissue and increased bone fragility and fracture risk. Both OA and OP are mainly affected on the elderly people. Recent studies have shown that osteopontin (OPN) plays a vital role in bone metabolism and homeostasis. OPN involves these biological activities through participating in the proliferation, migration, differentiation, and adhesion of several bone-related cells, including chondrocytes, synoviocytes, osteoclasts, osteoblasts, and marrow mesenchymal stem cells (MSCs). OPN has been demonstrated to be closely related to the occurrence and development of many bone-related diseases, such as OA and OP. This review summarizes the role of OPN in regulating inflammation activity and bone metabolism in OA and OP. Furthermore, some drugs that targeted OPN to treat OA and OP are also summarized in the review. However, the complex mechanism of OPN in regulating OA and OP is not fully elucidated, which drives us to explore the depth effect of OPN on these two bone diseases. Osteoarthritis (OA), the most common aging-related joint pathology is a degenerative disease affecting all the structures of the joints. OA is mainly characterized by articular cartilage destruction along with changes occurring in other joint components including bone, menisci, synovium, ligaments, capsule, and muscles (1). Worldwide estimates that 9.6% of men and 18.0% of women aged over 60 years have symptomatic OA (2). Radiographic evidence of OA occurs in the majority of people by 65 years of age and in about 80% of those aged over 75 years (3). Osteoporosis (OP) is defined as a systemic skeletal disease characterized by low bone mass and micro-architectural deterioration of bone tissue with a consequent increase in bone fragility and susceptibility to fracture (4). OP is a major risk factor for fractures of the hip, vertebrae, and distal forearm. It is thought of as a disease of old age which is present in 15% of those 50–59 years of age, but these figures increase quickly to 70% of those over 80 years of age (2). The relationship between OA and OP is complex and controversial. Kim et al. (5) conducted a meta-analysis and finally found the frequency of OP overall in men and women with OA was no different. However, according to the site of bone mineral density measurement, there was a higher prevalence of OP in the lumbar spine in both men and women compared to the matched controls. A cross-sectional study which was aimed to reveal the relationship between radiographic features of OA and bone mineral density (BMD) found that hand osteophytes and sclerosis exhibited a positive relationship with the BMDs of the lumbar spine and femoral neck while the knee and hand joint space narrowing presented a negative tendency to the BMD of the lumbar spine and femoral neck (6). Kasher et al. (7) confirmed that a higher hand OA score was significantly negatively correlated with arm and hand BMD measurements in males and females accompanied by a higher prevalence of wrist fracture, but the knee OA affection was positively associated with the elevated hip, spine, and total body BMD levels. A previous study also found the relationship between OA and OP was sexually different for the reason that femur neck and lumbar BMD and OA showed a positive tendency in women while BMD of the lumbar and pelvis in men was negatively correlated with OA (8). Hence, illuminating the relationship between OA and OP may uncover the relationship between OA and OP as well as the molecular or pathological factors that influence them that could help treat these diseases. Some factors have been found to affect the progress of both OA and OP, including sex hormones, ethnicity, age, nutritional factors, genetic factors and physical activity. Aging is a predictor of radiographic OA, bone loss, development of OP, and fracture (2). Recently, a few studies found that the expression of OPN mRNA isolated from human OA cartilage was enhanced as compared with normal cartilage. OPN was shown to be upregulated in human OA chondrocytes (9). OPN in plasma, synovial fluid and articular cartilage is associated with progressive joint damage and is likely to be a useful biomarker for determining disease severity and progression in knee OA (10, 11). Additionally, high serum OPN level was also a significant risk factor causing menopausal OP and serum OPN levels could be used as a biomarker for the early diagnosis of OP in postmenopausal women (12). Thus, OPN may be involved in the molecular pathogenesis of OA and OP, it may play a role as a bridge between OA and OP. Herein, we summarize current understandings of the molecular mechanism of OPN in OA and OP, focusing on recent results that have examined the role of OPN between OA and OP. OPN is a 44-75 KD multifunctional phosphoprotein secreted by many cell types such as osteoclasts, chondrocytes, synoviocytes, macrophages, lymphocytes, epithelial cells and vascular smooth muscle cells (SMC) and is present in the extracellular matrix of mineralized tissues and extra-cellular fluids, at sites of inflammation (13–15). It is encoded by the SPP1 gene and maps as a tandem array to the long arm of chromosome-4 (16), producing splicing variants of mRNA, full-length OPN and spliceosomal OPN (missing exons 4 or 5) (17, 18). The full-length OPN is cleaved by thrombin to form the thrombin-cleaved OPN (19). Both the native OPN and cleaved OPN could interact with integrin and play an important role in regulating inflammation, biomineralization, bone remodeling, immune functions, chemotaxis, and cell apoptosis (19, 20). Gene structure and chromosomal location identify OPN as a member of the small integrin-binding ligand N-linked glycoprotein (SIBLING) family. This protein also known as early T cell activation gene-1 (Eta-1) is abundant in bone, where it mediates important cell-matrix and cell-cell interactions (21). OPN expression is one of the important events involved in cartilage-to-bone transitions in fracture repair during the period of chondrocyte maturation (22, 23). OPN facilitates the attachment of osteoclasts to the bone matrix via interaction with cell surface αvβ3 integrin and CD44 (24, 25). OPN interacts with receptors such as integrin and CD44 to regulate physiological and pathological processes of proliferation, differentiation, inflammation, metabolism and tumor metastasis in chondrocytes, osteoblasts and osteoclasts (26–28). Previous studies have shown that OPN is involved in the occurrence, repair and maintenance of cartilage and subchondral bone metabolic homeostasis, suggesting that OPN is an important regulator of OA and OP and plays an important role in chondrocyte and osteocyte metabolism by regulating the extracellular matrix components of articular cartilage subchondral bone components (10, 29, 30). OPN is widely distributed in many cells and tissues such as chondrocytes, plasma, synovial, osteoblasts and osteoclasts and plays regulatory roles in many diseases. Clinical studies have shown that OPN is involved in bone strength and remodeling, suggesting that serum OPN is positively correlated with the severity of OP and could be targeted as a biomarker for early diagnosis of postmenopausal osteoporosis (31, 32). Further, OPN is related to bone turnover and bone mineral density (BMD) and influences morphological formation and reconstruction (33–35). The high serum OPN level could result in a low BMD and OP in postmenopausal women (34, 35), moreover, the high level of OPN is associated with osteoporotic fractures in postmenopausal women, particularly at the lumbar spine (32). In addition, researchers have examined the relation between obesity and OP, and suggested obesity could lead to a large number of adipocytes and adipose tissue which may be greatly related to OP (36–38). Dai et al. (39) found that OPN secreted by the macrophages in epididymal white adipose tissue regulated the bone metabolism in high-fat diet (HFD)-induced obesity. The OPN selectively circulated to the bone marrow and promoted the degradation of the bone matrix by activating osteoclasts, both surgical removal of epididymal white adipose tissue and local injection of OPN-neutralizing antibodies or drugs aimed to deplete macrophages could ameliorate HFD-induced bone loss in mice. There is also evidence indicating that OPN is one of the most overexpression genes in the adipose tissue-derived from obese patients (40). OPN is a critical intrinsic regulator that plays an important role in the pathogenesis and progress of OA. The plasma and synovial fluid level of OPN in OA patients are higher than in healthy adults, suggesting that OPN may be correlated with the severity of joint lesions in OA (11). Min et al. (41) investigated the serum of 249 people and found serum OPN was significantly increased in OA patients compared with control group, furthermore, they found there existed a gender difference in the concentration of OPN between the OA group and the control group. The gender difference in OPN expression was mainly presented that serum OPN level in the control group was lower in OA patients with the exact level of 2539.9 (pg/ml) and 5538.1 (pg/ml) in males, while the results of serum OPN level in the control group compared to OA patients were 1632.0 (pg/ml) and 4545.8 (pg/ml) in female. In addition, other researchers also found higher OPN mRNA and protein expression in synovial fluid of OA patients is closely related to the occurrence and development of OA (42), while OPN gene polymorphism could decrease the risk of OA (43). Recently, studies also show the expression of OPN is greatly increased in both the superficial zone and deep zone of articular cartilage of OA patients (9, 44). Animal experiments show that increased expression of OPN accelerates the turnover and remodeling of OA subchondral bone, promotes vessels formation in subchondral bone, mediates articular cartilage degeneration induced by subchondral bone metabolism, and accelerates the progression of OA (45). Some contradictory findings show OPN has a therapeutic value and mechanism in OA treatment. The mRNA and protein expression of OPN, CD44 is upregulated in OA chondrocytes, moreover, the upregulated OPN could delay chondrocyte degeneration and reduce cartilage matrix component loss by binding to CD44 and integrin via OPN/CD44/PI3K signaling pathway (46). CD44 is a cell surface protein that interacts with a variety of extracellular matrix (ECM) components whose principal ligands mainly include hyaluronan (HA) and OPN (47). The CD44 variants containing the exons v6 and v7 bind to the N- and C- terminal portions of OPN in an arginine-glycine-aspartic independent manner and further regulate the effect of OPN on bone metabolism (48). Zhang et al. (15) indicated that HA could upregulate OPN mRNA expression in OA fibroblast-like synoviocytes, and the high expression of OPN mRNA in OA may be a result of increased HA level of OA synovitis, finally, alleviating the severity and improving the symptoms of OA. OPN deficiency may enhance the senescence and apoptosis of OA chondrocytes, up-regulated the expression of pro-inflammation, and decrease the expression of COL2A1, finally accelerating the progress of chondrocytes and OA severity (27). OPN is highly expressed in the process of inflammation. Many studies have characterized the function of OPN in inflammation activity, and it plays an important role in the pathogenesis of various inflammatory diseases such as OA. The previous study has shown that OPN is a proinflammatory cytokine that plays an important role in the pathogenesis of arthritis as summarized in Table 1 . There presents an elevated level of OPN in synovial fluid samples from RA patients, and the increased expression of OPN is correlated with the high level of multiple inflammatory cytokines including tumor necrosis factor-alpha (TNF-α) and interleukin-6(IL-6) (62). Another study also confirms the high protein expression of OPN in synovial fluid derived from advanced OA patients, significant differences and correlations are found among the thrombin-cleaved OPN, synovitis, and cartilage damage indicated by higher Kellgren-Lawrence scores (63). A long-term cohort study has shown the inflammatory mediators mainly including CRP, IL-6 accompanied by OPN significantly increased in arthroplasty patients the day after surgery and returned to and baseline six weeks later. It also found other inflammatory factors IL-1β and IL-8 showed the same up-regulated trend and reached the peak at 5 years post-surgery and returned to normal level at 10 years while the TNF-α level did not show any change preoperative or postoperative (53). The serum sample collected from total joint arthroplasty patients showed OPN and matrix metalloproteinase-9 (MMP-9) was greatly up-regulated, while MMP-9 was known to cleave OPN aimed to degrade Cola2a1 and upregulated by inflammatory markers such as IL-1, IL-6 and CRP (49). These biomarkers of inflammation were correlated with the progress of OA and total joint arthroplasty. The OPN level also significantly increased in synovial fluid samples from symptomatic primary knee osteoarthritis with ultrasound-confirmed joint effusion, moreover, the up-regulated OPN level presented associations between IL-8 and TNF which is responsible for pain, cartilage damage, clinical severity, and progression of OA (54). Wang, et al. (50, 51) also revealed that OPN could regulate the expression of various inflammatory factors including MMP-13, IL-6 and IL-8 which were significantly upregulated in OA tissues and associated with the pathogenesis of OA. Furthermore, many drugs have been devoted to investigating the relationship between OPN and inflammation factors in OA. Isorhamnetin could reduce knee swelling and alleviate cartilage damage in MIA-induced rats. The therapeutic effect was achieved mainly through inhibiting the expression of OPN and C-terminal telopeptide of type II collagen (CTX-II) accompanied by decreased levels of NO, PGE2, iNOS and COX- 2 (52). However, some controversial results found that OPN plays a protective role in OA. Tian et al. (27) compared the mRNA expression level of OPN from human OA chondrocytes and normal chondrocytes, and found the mRNA level of OPN was greatly suppressed in normal chondrocytes. OPN deficiency increased the expressions of Col10a1, IL-1β, TNF-ɑ, MMP-13, and ADAMTS-5 but decreased the expression of Col2a1, finally leading to higher rates of senescence and apoptosis of chondrocytes. Other researchers also found that OPN deficiency led to the induction of MMP-13 in instability-induced and aging-associated OA, which degrades a major component of the cartilage matrix protein type II collagen, indicating OPN plays a pivotal role in the progression of OA (55). Therefore, OPN might be a critical biomarker in the inflammatory activity of OA for either up-regulated OPN level or OPN deficiency could stimulate the secretion of inflammatory factors, damage articular cartilage, and aggravate the severity of OA. Osteoporosis is a common disease in the aging population and multiple studies have shown that inflammation cytokines greatly influence the pathogenesis of OP (64, 65). Researchers have shown that many pro-inflammatory mediators including cytokines such as TNF-α, IL-1, IL-6, and IL-10 up-regulated in OP and participate in the process of bone resorption or bone mineral density (65–67). As a secreted protein, OPN is greatly related to the inflammatory response in OP, which further influences the process of bone remodeling as the Table 1 shown. The Animal experiment has shown that OPN and osteocalcin (OCN) level declined in inflammation-mediated osteoporosis (IMO) of ovariectomized rats. The IMO rats represented higher serum tartrate-resistant acid phosphatase (TRAP), CTX-I, and pro-inflammatory factors TNF-α, IL-1β, and IL-6 levels, accompanied by decreased femur BMD, bone mineral content (BMC) and distal femur cancellous bone in IMO rats (56). Gao, et al. (58) demonstrated rheumatoid arthritis induced bone loss and bone quality deterioration, with high bone turnover in collagen-induced arthritis (CIA) rats. The CIA rats showed higher levels of IL-6, and TNF-α in serum accompanied by decreased mRNA and protein levels of Osx (Osterix), OPN which is induced by a TNF-α-induced inflammatory medium in vitro. Osteoclasts are multinucleated cells essential for bone resorption and play a central role in the development of OP. The osteoclasts intervened by knock-down angiopoietin-like protein2 could promote expressions of pro-inflammatory cytokines including TNF-α, IL-1β, IL-6, and CCL-2, reduce Runx2, OPN, and Colla1 levels, finally improve bone loss and BMD in osteoporosis mice induced by ovariectomy (59). Some studies found many drugs could regulate the OPN level and inflammation response to alleviate the severity of OP. Monotropein significantly inhibited bone mass reduction and improved bone micro-architectures by enhancing bone formation in osteoporotic mice by suppressing the secretion of IL-6 and IL-1β induced by LPS. The experiment in vitro also increased the expression and activity of alkaline phosphatase (ALP) and OPN in osteoblasts (61). Other researchers also confirmed that increasing the expression level of OPN and Runx2 could promote bone remodeling and reduce bone loss in OP accompanied by inhibiting the expression of inflammatory markers TNF-α, iNOS, or IL-1β in LPS-stimulated osteoblastic cells (57, 60). Taking together, the above results and studies mainly investigated the role of OPN and its regulating effect on inflammation activity in OA and OP, and further indicated overexpression of OPN or declined OPN levels in chondrocytes, synoviocytes, and osteoblasts could lead to metabolic disturbance in bone diseases as shown in Figure 1 . OPN is secreted by many types of cells, such as macrophages, lymphocytes, epithelial cells, vascular smooth muscle cells, chondrocytes, and synovial cells, which exist in a large number of cells and tissues (68–70). It is well known that abnormal expression of OPN in mRNA or protein level is correlated to the onset of OA. Up-regulated OPN and the phosphorylation of OPN in chondrocytes and synovial fluid greatly contribute to and aggravate the severity of OA (10, 19, 68). A previous study shows overexpression of OPN could promote the proliferation of chondrocytes, but suppressed their apoptosis through the NF-κB signaling pathway to regulate the pathological process of OA (70). The elevated OPN serum level is coincident with the tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts and the extent of bone erosion in CIA mice, silencing of OPN using lentiviral-OPN short hairpin RNA presents an opposite result (71). Elmazoglu, et al. (72) found OPN and bone morphogenetic protein-7(BMP-7) were inhibited by S-allylcysteine accompanied by the suppressed inflammation response and decreased IL-1β, IL-6 in chondrocytes which finally alleviate the severity of OA. OPN also mediates subchondral bone remodeling and cartilage degeneration in OA. Animal study shows that higher level of OPN secreted by pre-osteoblasts and osteoblasts in subchondral bone could promote osteoclastogenesis and blood vessels formation in subchondral bone, accelerate the turnover and remodeling of subchondral bone and mediate articular cartilage degeneration induced by subchondral bone metabolism in anterior cruciate ligament transection and destabilization of the medial meniscus (ACLT + DMM) OA model (45). Furthermore, OPN is also on behalf of the osteogenic and chondrogenic differentiation progress, abnormal expression of OPN along with Runx2, Sox9, and OCN may induce metabolic disturbance and cartilage damage in OA (51, 73). Therefore, OPN plays a critical role in regulating the metabolism in chondrocytes, synovial cells, and subchondral bone which greatly influence the physiological process of joint, it should be precise modulated at a proper level for the prevention of onset and aggravation of OA as summarized in Table 2 and shown Figure 2 . Osteoporosis is a systemic skeletal disorder characterized by systemic damage to bone mass and microstructure, which is caused by bone metabolism disorders and is the main reason for fragility fractures in aged people (30, 86, 87). Notably, OPN plays an important role in bone strength and bone remodeling. The role of OPN involved in cartilage and bone metabolism of OP has been greatly investigated in many previous studies and the effect of OPN levels on osteoporosis is gaining more and more attention, especially those OPN exits in serum, chondrocytes, osteoclasts, and osteoblasts. A large number of previous studies through animal experiments have shown that OPN has a protective effect on osteoporosis and OPN-deficient mice by oophorectomy are resistant to osteoporosis (88–90). Chondrocyte-specific genes knockout mice also showed impaired cartilage formation, decreased bone density and an osteoporotic phenotype with the decreased osteoblast marker genes including OPN and OCN in osteoblasts, as well as the expression of osteoblast differentiation regulation genes such as Osx, Runx2 and ATF4 (74). Ovariectomy mice presented an osteoporosis phenotype mainly on osteochondral remodeling accompanied by increasing the endplate porosity and decreasing the bone volume, the changes of OPN, OCN, Osx in osteoblast and serum which influence the osteoblast differentiation is responsible for this pathological process (75, 76). Clinical studies have found that serum level of OPN, as biomarkers for early diagnosis of osteoporosis in postmenopausal women, is positively related to the severity of osteoporosis. The postmenopausal women have a higher OPN serum level compared to those premenopausal women, meanwhile the higher OPN level shows a negatively correlated relationship with the weight, height, and BMD (31, 32, 35). Previous scholars found the serum OPN level in the menopausal females was 15.4 (ng/ml) compared to non-menopausal females with a lower OPN level of 7.8 (ng/ml), moreover, the higher OPN level also indicated an approximately 2.97-fold risk of osteoporosis compared with the persons with low serum OPN levels (12). Vancea et al. (33–35) also confirmed the negatively relationships between OPN and BMD in the postmenopausal women, they considered postmenopausal women exhibited higher serum OPN levels may have a lower BMD and higher risk of OP than the premenopausal women. Wu et al. (82) adopted an extreme sampling design to systemically screen the OP-related cytokines in serum of postmenopausal women, further verified the OPN and BLC modulate the bone metabolism by inhibiting bone formation and promoting bone resorption. Since OPN has a major effect on osteoclasts through regulating the bone metabolism and bone remodeling, plenty of scholars have devoted great energy to investigating the role of OPN in OP and found new drugs targeting OPN to treat osteoporosis. Dong et al. (79) investigated the effects of anagliptin on the differentiation and mineralization of osteoblasts and discovered that anagliptin significantly increased matrix deposition and mineralization by increasing the activity of osteoblasts as evidenced by elevated levels of ALP, OCN, OPN, and BMP-2. Other studies also confirmed increased expression of OPN in mRNA or protein level of osteoclasts protects the osteoporosis to promote osteogenic differentiation and inhibits osteoclastogenesis. The potential therapeutic effects of elevated OPN activity on osteoporosis mainly reflected on increasing body weight gain and BMD, with the assistance effect of up-regulated level of ALP, OCN, BMP-2 and reduced tartrate-resistant acid phosphatase (TRAP) activity as well as C-terminal telopeptide of type I collagen (CTX-I) level (81, 85, 91). In addition, some scholars also found traditional Chinese medicine tonifying prescriptions have potential therapeutic effects on osteoporosis in ovariectomy-induced rat model. The traditional Chinese medicine mainly targeted to increase the expression of OPN, BMP-2 and Runx2 which finally resulted in reducing the broken trabecular bones in femur bones and increasing the activity of ALP combined with enhanced the content of total bone mineral density in osteoporosis rats (77). Due to the important biological effect of OPN on osteoporosis, its expression and physiological effect in mesenchymal stem cells (MSC) has been greatly studied. Recently, researchers confirmed that the therapeutical effect of MSCs on osteoporosis relied on the elevated extracellular Ca2+ promoted cell proliferation and matrix mineralization of MSCs. Moreover, the enhancement of MSCs under up-regulated extracellular Ca2+ condition is induced by the elevated OPN level (92). It is further found OPN not only significantly promoted the proliferation of MSCs, but also activated the migration and regulate the cell stiffness through integrin and the Wnt signaling pathway which accelerates the osteoblasts differentiation of MSCs indicated by overexpression of osteoblasts differentiation related proteins (78, 93, 94). The complex effect of OPN means that regulating its expression is a complex and challenging process. Recently, some new and classic drugs have investigated the effect of OPN on OA and OP. Pharmaceutical interventions of OPN in OA have hitherto mainly focused on the OPN level in chondrocytes, synovial tissue and synoviocytes. Elmazoglu et al. (72) found that S-allylcysteine could inhibit the IL-1β, IL-6 and OPN in chondrocytes by suppressing the activity of the p-JNK/pan-JNK signaling pathway. The inhibition of OPN and these cytokines finally lead to the up-regulated level of peroxidase and type-II-collagen to delay the progress of OA. Li et al. (51) investigated the effect of chitosan oligosaccharides which are packed into the extracellular vesicles in OA. They verified that chitosan oligosaccharides could reverse IL-1β induced chondrocytes apoptosis and the inhibition of viability and migration of chondrocytes, furthermore, they also discovered the therapeutical effect of chitosan oligosaccharides on cartilage damage was by down-regulating the level IL-1β, OPN, and p53 accompanied by the increased level of Col2a1, OCN, and Runx2 via PI3K-Akt pathway. Isorhamnetin also reduced MIA-induced knee swelling by significant reduction of articular cartilage damage through regulating the OPN level in rats. The protective effect of isorhamnetin on OA was through the inhibition of OPN, NO, PGE2, iNOS and COX-2 (52). Slovacek et al. (49) showed that OPN and MMP-9 levels were significantly elevated in OA, the increased level of OPN and MMP-9 is stimulated by inflammatory markers, such as IL-1, IL-6, and CRP. Previous researchers found adiponectin aggravates bone erosion by promoting OPN production in synovial tissue while silencing of OPN with lentiviral-OPN short hairpin RNA could reduce the number of TRAP-positive osteoclasts and the extent of bone erosion in CIA mice (71). In addition, the microRNA(miR) could also influence the expression of OPN in OA. MiR-181c could greatly inhibit the effect of OPN on stimulating the secretion of MMP-13, IL-6, and IL-8 along with repressing synoviocyte proliferation which finally alleviated local inflammation activity and cartilage damage in joints, severed as a therapeutic strategy for OA (50). The multiple effects of OPN on different types of cells indicates that modulation of OPN in OP represents a definite challenge. Our current understanding of pharmaceutical interventions of OPN in OP is discussed in the following sections and summarized in Table 3 . Gao et al. (58) found salvianolic treatment ameliorated osteopenia and improved bone quality in rats, the potential mechanism may rely on increasing Osx, OPN, and Col1a1 levels in TNF-α-induced MC3T3-E1 osteoblasts through regulating the RANKL/RANK/OPG signaling pathway. Monotropein was also found to play an important role in regulating the OPN metabolism in MC3T3-E1 cells. Research indicated that monocropping was able to increase the proliferation and activity of ALP, bone matrix mineralization and the expression of bone matrix protein OPN accompanied by decreasing the production of IL-6 and IL-1β via suppressing the activity of the NF-κB signaling pathway (61). Choi et al. (96) found that Palmul-tang could also increase bone mineral content and improve bone mineral density to treat OP. Furthermore, they indicated that the therapeutic effect of Palmul-tang mainly depended on up-regulating the expressions of BMP-2, Runx2, and Osx with its downstream factors ALP and OPN through the BMP-2 signaling pathway. Increased level of OPN also influenced the osteoblasts differentiation, matrix deposition and mineralization along with the high level of ALP, OCN, OPN, and BMP-2 which was related to the Wnt/β-catenin signaling pathway with the interventions of anagliptin and artesunate (78, 79). Apart from the above biomarkers such as BMP-2, Runx2, Osx and ALP, the serum OPN level was also negatively correlated with bone turnover markers mainly including parathyroid hormone, lumbar spine BMD, femoral neck BMD and positively associated with type I procollagen amino-terminal propeptide (PINP), carboxy-terminal cross-linking telopeptide of type I collagen (CTX) in the OP patients (35). Furthermore, they indicated CTX was independent predictor of serum OPN while vitamin D was not correlated to OPN in adults (32, 97). The animal model showed that up-regulated OPN could enhance the activity of ALP, OCN, and BMP-2 in dexamethasone-induced OP, the anti-osteoporosis function of myricetin in vivo may be due to the promoting osteogenic differentiation and matrix mineralization effect caused by OPN (83). Taking together, the pharmaceutical interventions of OPN in bone metabolism present different biological effects due to the multiple types of cells. Generally speaking, the current drug treatment mainly depended on reducing the OPN level in chondrocytes and synovial cells to suppress inflammatory activity alleviate cartilage erosion, and delay the course of OA. While it referred to OP, the pharmaceutical effect of drug intervention on OPN is to increase the expression level of OPN in osteoclasts and stem cells. It is contrary to the study of drug intervention in OA, the increased OPN level, as well as its related cytokines BMP-2, OCN, and Runx2, could promote bone formation and enhance bone mineral density to achieve the purpose of treating OP. In recent years, there has been growing interest in trying to identify the physiological effect of OPN on skeleton diseases. OPN acts as a secreted protein that participates in the progress of bone remodeling and bone metabolism which are relevant to many bone metabolism disturbance diseases. OPN regulates the inflammation activity both in OA and OP, and plays controversial roles in the inflammation response for either overexpression of OPN level in chondrocytes and synoviocytes or OPN-deficiency in cells residents in joints may stimulate and enhance the inflammation activity in OA as the Figure 1 shown. While the higher level of OPN could suppress the secretion of proinflammation cytokines and inflammation activity, alleviate bone mass reduction, and promote bone formation as well as matrix calcification in OP. As referred to as the role of cartilage and bone metabolism, OPN is a biological marker of OA severity. Overexpression of OPN enhances the proliferation of chondrocytes and TRAP-positive osteoclasts along with the osteogenic and chondrogenic differentiation progress, which finally leads to bone erosion and cartilage degeneration in OA. High levels of OPN could increase matrix deposition and enhance bone remodeling by promoting osteoblasts differentiation along with the elevated level of ALP, OCN, and BMP-2 which finally resulted in reducing the pain and pathological changes of osteoporosis. Low activity of OPN could increase fracture sensitivity in patients with osteoporosis, especially for those postmenopausal women as Figure 2 shown. Due to the above physiological role of OPN in OA and OP, we do deem that OA and OP influenced each other for the reason that OA is associated with bone formation as seen in subchondral sclerosis and osteophyte formation, and the tendency to accumulate bone in the subchondral area could increase the onset of OA. In contrast, once OA is established, the pain and reduced mobility reduce bone mass, particularly in the affected limb. The pain and loss of joint function in OA patients could result in muscle loss and postural instability, which subsequently increased fracture risk. OPN may serve as a bridge role in regulating bone metabolism and signal transduction between the disease of osteoarthritis and osteoporosis, and it should be precisely modulated at a proper level for either higher OPN level or lower OPN activity may induce to disrupt the balance of bone tissue. However, there are still many details and effects of OPN on bone metabolism and bone-related disease not been elucidated. Further in-depth studies on the physiological function of OPN provide new ideas and directions for exploring and clarifying the pathogenesis of bone metabolic diseases, targeting OPN may provide new clinical therapeutic orientations and values for OA and OP. All the authors researched the data for the article, provided substantial contributions to discussions of its content, wrote the article and reviewed and/or edited the manuscript before submission. All authors contributed to the article and approved the submitted version. This work was supported by the National Key R&D Program of China (No.2019YFA0111900), the National Natural Science Foundation of China (No.81501923 and No. 82072506), Provincial Natural Science Foundation of Hunan (No.2020JJ3060), Administration of Traditional Chinese Medicine of Hunan Province (No.2021075), InnovationDriven Project of Central South University (No.2020CX045), Wu Jieping Medical Foundation (No.320.6750.2020-03-14), Rui E (Ruiyi) Emergency Medical Research Special Funding Project (No.R2019007), and Independent Exploration and Innovation Project for Postgraduate Students of Central South University (No.2021zzts1037). The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649921

Jie Yang,Yue Sun,Jinling Chen,Yu Cheng,Haoran Zhang,Tengqi Gao,Feng Xu,Saikun Pan,Yang Tao,Jing Lu

Fermentation of ginkgo biloba kernel juice using Lactobacillus plantarum Y2 from the ginkgo peel: Fermentation characteristics and evolution of phenolic profiles, antioxidant activities in vitro, and volatile flavor compounds

28-10-2022

ginkgo biloba kernel juice,lactic acid bacteria fermentation,phenolic substances,free amino acids (FAAs),volatile flavor substances

In this study, a strain of Lactobacillus plantarum Y2 was isolated from the ginkgo peel, and showed adequate adaptation to the ginkgo biloba kernel juice. After 48 h of fermentation, the number of viable cells in the stable growth phase was remained at 10.0 Log CFU/mL, while the content of total organic acid increased by 5.86%. Phenolic substances were significantly enriched, and the content of total phenolic substances increased by 9.72%, and the content of total flavonoids after fermentation exceeded 55.33 mg/L, which was 3.6 times that of the unfermented ginkgo juice. The total relative content of volatile flavor compounds increased by 125.48%, and 24 new volatile flavor substances were produced. The content of total sugar, total protein, and total free amino acid decreased to 44.85, 67.51, and 6.88%, respectively. Meanwhile, more than 82.25% of 4′-O-methylpyridoxine was degraded by lactic acid fermentation, and the final concentration in ginkgo biloba kernel juice was lower than 41.53 mg/L. In addition, the antioxidant and antibacterial activities of fermented ginkgo biloba kernel juice were significantly enhanced. These results showed that LAB fermentation could effectively improve the nutritional value and safety of ginkgo biloba kernel juice.

Fermentation of ginkgo biloba kernel juice using Lactobacillus plantarum Y2 from the ginkgo peel: Fermentation characteristics and evolution of phenolic profiles, antioxidant activities in vitro, and volatile flavor compounds In this study, a strain of Lactobacillus plantarum Y2 was isolated from the ginkgo peel, and showed adequate adaptation to the ginkgo biloba kernel juice. After 48 h of fermentation, the number of viable cells in the stable growth phase was remained at 10.0 Log CFU/mL, while the content of total organic acid increased by 5.86%. Phenolic substances were significantly enriched, and the content of total phenolic substances increased by 9.72%, and the content of total flavonoids after fermentation exceeded 55.33 mg/L, which was 3.6 times that of the unfermented ginkgo juice. The total relative content of volatile flavor compounds increased by 125.48%, and 24 new volatile flavor substances were produced. The content of total sugar, total protein, and total free amino acid decreased to 44.85, 67.51, and 6.88%, respectively. Meanwhile, more than 82.25% of 4′-O-methylpyridoxine was degraded by lactic acid fermentation, and the final concentration in ginkgo biloba kernel juice was lower than 41.53 mg/L. In addition, the antioxidant and antibacterial activities of fermented ginkgo biloba kernel juice were significantly enhanced. These results showed that LAB fermentation could effectively improve the nutritional value and safety of ginkgo biloba kernel juice. Ginkgo biloba L., one of the oldest living tree species on earth (1), has been reported to have great development potential in the field of functional food (2). Meanwhile, ginkgo biloba kernel juice contains various active ingredients such as ginkgo phenolic acids, flavonoids, and polysaccharides, etc., making it widely used in food, health products, cosmetics, medicine and other fields (3). In particular, the special physiologically active components such as flavonoids and ginkgoid acid contained in it have the functions of anti-oxidation, anti-aging, anti-inflammatory, anti-allergic, and inhibiting nerve cell apoptosis (4). Recent studies on ginkgo have mainly focused on ginkgo leaves and relatively few studies on ginkgo kernels (5). Therefore, it is necessary and meaningful to further develop and utilize ginkgo kernels. 4′-O-methylpyridoxine (MPN) is a vitamin 6 derivative which can cause in pregnancy, radiation sickness, seborrheic dermatitis, and other pathologies. Interestingly, only a threshold concentration of MPN is beneficial to the human body; a higher concentration can lead to toxic reactions in the human body. The MPN content in medicine ranged from 0 to 9.77 μg/mL (6), and in ginkgo kernels ranged from 172.8 to 404.2 μg/g, indicating that excessive consumption of ginkgo kernels was harmful to human health (7). As MPN concentrations severely limit the development and utilization of ginkgo kernels, MPN degradation during the production process of ginkgo is a vital concern that needs to be addressed. Lactic acid bacteria (LAB) have been widely used in the fermentation of pickles, yogurt, soy sauce and tempeh, owing to their ability to regulate the human intestinal flora, promote the absorption of nutrition substances, kill harmful flora and the toxins, and ameliorate food flavor (8). Moreover, the secondary metabolites in the fermentation process with LAB have several health benefits, such as promoting the activity of antioxidant enzymes in cells, and increasing the content of beneficial substances. For example, Verón found that LAB fermentation could increase the activities of ferulic acid, caffeic acid derivatives and intracellular antioxidant enzymes, and enhance the overall antioxidant capacity of the fermentation broth (9). Currently, studies on ginkgo mainly focuses on ginkgo leaves, while there are relatively few studies on ginkgo kernels. Ginkgo products mainly include canned ginkgo kernel, ginkgo wine, and beverages (10, 11). However, only a few studies had been conducted on fermented silver almonds. In this study, a LAB strain was selected from ginkgo peel to ferment ginkgo biloba kernel juice. The physiological and biochemical indices of fermented ginkgo biloba kernel juice were analyzed, and the antioxidation activity and bacteriostatic ability were evaluated. The results of this study will enrich the physiological and biochemical studies of LAB-fermented ginkgo biloba kernel, and improve the reference for the development of ginkgo products. Ginkgo was sourced from the campus of Jiangsu Ocean University (Lianyun Gang, Jiangsu province), and the LAB strains were screened from the surface of the ginkgo fruit. The ginkgo fruit was inoculated in MRS culture medium (g: mL, 1:1.5) and cultured at 37°C for 24 h. The fermentation broth was spreaded on MRS plates to obtain single colonies which were underlined on MRS plates and placed in a 37°C incubator for 24 h to obtain single colonies. The obtained single colonies were then grown in MRS medium, with 40% glycerol stocks stored and frozen in a −40°C refrigerator. These LAB strains were identified using 16S-rRNA, physiological and biochemical tests in accordance with conventional methods (12). The ginkgo fruit was boiled for 5–10 min, and the shell and seed coat were removed. After washing with water, the shelled ginkgo kernels were crushed and broken up in a certain proportion (ginkgo biloba kernels: distilled water = 1:2.2, g:mL), and the starch dissolution rate was required to reach more than 80%. The α-amylase (20 U/g, ginkgo mass) and saccharification enzyme (30 U/g, ginkgo mass) were added, and then ginkgo biloba kernel juice was saccharified at 50°C for 2 h. After the enzymatic hydrolysis the ginkgo biloba kernel juice was completed. It was filtered through a 100-mesh filter, and pasteurized in a water bath at 90°C for 20 min (13). In a 500 mL sealed Erlenmeyer flask, 1% (v/v) inoculum of ginkgo biloba kernel juice was inoculated to ensure an initial viable count of approximately 5.0 Log CFU/mL. The fermentation process was performed in an incubator at 37°C for 48 h. At the end of fermentation, the bacterial cells were removed by centrifugation (10,000 g, 10 min, 4°C), and the supernatant was collected for further chemical analysis. Viable cell counts were determined using the standard plate method (14). The fermentation broth was first diluted to an appropriate concentration, then the diluted liquid was spread on a plate and the colonies that grew were counted. A precision pH meter (PHS-3C, Shanghai INESA Scientific Instrument Co., Ltd., Shanghai, China) was used to measure the pH value of each sample of ginkgo biloba kernel juice fermentation broth and the bacteriostatic activity was assessed by measuring the diameter of the outward inhibition zone of the Oxford cup using a Vernier caliper (15). The centrifuged supernatant (10 mL) was concentrated in vacuo to 2 ml, and then screening through a 0.22 μm membrane screening. The bacterial pathogens were Escherichia coli CICC 10003, Staphylococcus aureus CICC 23656, and Bacillus cereus s CICC 23828. The total sugar content was determined using the sulfuric acid phenol method (16). The principle of this method was that polysaccharides were hydrolyzed into monosaccharides under the action of sulfuric acid, and then rapidly dehydrated to form uronic derivatives, which were then combined with phenol to form orange-yellow compounds. The absorbance values at 470 nm of the orange-yellow compounds were linearly related to the monosaccharide concentration. A standard curve was established using anhydrous glucose and the results were expressed in μg/mL of glucose equivalents. Protein was detected using the Coomassie brilliant blue method G250 (17). A standard curve was established using a Bovine Serum Albumin (BSA) standard solution, and the total protein equivalent was expressed in mg/mL. The organic acid spectrum was analyzed using a the Shimadzu LC-2010A system (Shimadzu, Tokyo, Japan), following the method of Luo Ke with some modifications (18). The chromatographic column was an Agilent TC-C18 column (4.6 × 25 mm, 5 μm), the detection wavelength was 210 nm, 0.08 M KH2PO4 solution (adjusted to pH 2.5 with phosphoric acid) was used for isocratic elution. The column temperature was 30°C, the flow rate was 0.7 mL/min, and the injection volume was 20 μL. The free amino acids were quantified using Adeyeye’s method with slight modifications. The analytical column was 2622PH 4.6 mm I.D. × 60 mm, the flow rate was 0.40 mL/min, the column temperature was 57°C, the reaction temperature was 135°C, and the detection wavelength was 570 nm, and the injection volume was 20 μL. The total flavonoids was determined according to the method of Kwaw et al. (19). A calibration curve was constructed with rutin and the results were expressed in μg/mL of rutin equivalents. The total phenolic content of crude polyphenols in fermented or unfermented ginkgo biloba kernel was determined using the Folin-Ciocalteu method. A calibration curve constructed using gallic acid and the results were expressed in mg/L of gallic acid equivalents. Phenolic acid and flavonoids content were determined according to the method described by Chiang et al. (20). The column was Inertsil ODS-3 5 μm (4.6 × 250 mm, 5 μm), and the column temperature was 25°C. The mobile phase was composed of solution A (1% acetic acid in water) and the solution B (1% acetic acid in methanol), and the flow rate was 0.6 mL/min. The gradient set was as follows: 0–10 min, 10–26% B; 10–25 min, 26–40% B; 25–45 min, 40–65% B; 45–55 min, 65–95% B; 55–58 min, 95–10% B; 58–65 min, 10% B. The wavelength for determination of phenolic acids and flavonoids was 280 and 350 nm, respectively. The concentration of each phenolic compound was calculated according to the standard calibration curve, and the results were expressed in mg/mL. Total antioxidant capacity was measured in the same way as previously described (21). The total antioxidant capacity was calculated as follows: total antioxidant capacity = (Measured OD − Control OD)/0.01/30 × total reaction volume/sampling volume × sample dilution ratio before testing. The ABTS+-SA assay was performed using the method described by Tao et al. (21). Trolox was used as the standard, and the results were expressed in mmol/LTrolox equivalents. The Iron Reducing Antioxidant Ability (FRAP) of the fermented samples was evaluated by the method of Yan et al. (22). The volatile components were separated, collected, and analyzed using a triple quadrupole GC-MS (Trace 1310/TSQ 9000, Thermo Scientific) (23). The headspace vial was filled with nitrogen to expel the air in the vial, and 5 mL tested sample was taken into the Thermo RSH autosampler for extraction. The extraction head was placed in the gas chromatography injection port, the injection port temperature was 250°C, the aging was 30 min, the oven temperature was set to 80°C, the water bath was set for 20 min, the adsorption extraction was 30 min, and the extraction head was inserted into the injection port for analysis for 5 min. The MPN content was determined reference to the method described by Yoshimura using the Agilent 1260 fluorescence detector (detection wavelength was 291 nm) (24). All experiments were performed three times in the same way, and all samples were analyzed in triplicate. The significance analysis was performed using SPSS Statistics 20 (IBM Corp., NY, USA), and the significance was assessed at p < 0.05 value. The principal component analysis (PCA) was performed using Origin 2018 (Origin Lab Corp., UK), and all data results were expressed as mean ± standard deviation. In this study, nine LAB strains were isolated from the surface of ginkgo fruit. The developmental tree established by the BLAST program and the results of the physiology and biochemistry of the strains confirmed that all the selected strains were all Lactobacillus plantarum (Table 1 and Figure 1). Next, the acid-producing and antibacterial activities of the nine strains were evaluated. The pH of ginkgo biloba kernel juice, fermented by different strains, decreased, indicating that these strains had good acid production capacity. Among them, the pH of ginkgo biloba kernel juice fermented by the strain Y2 had a lower pH than the other strains, indicating that Y2 has a stronger growth and metabolism in ginkgo biloba kernel juice (25). In addition, the pH of ginkgo biloba kernel juice fermented by the strain Y2 had the best antibacterial effect on pathogenic bacteria (E. coli, S. aureus, and B. cereus), with inhibition zone diameters of 13.77 ± 0.32, 21.54 ± 0.58, and 15.57 ± 0.53 mm, respectively. When the acidity of fermented ginkgo biloba kernel juice is high, the pH value of the medium decreases, reducing conductivity for pathogenic bacterial growth (26). Lactic acid and compounds such as bacteriocins and synthetic hydrogen peroxide produced by the fermentation of lactic acid bacteria inhibit microbial growth (27). In conclusion, the strain Y2 had a good growth and metabolic status in ginkgo biloba kernel juice, and was selected as an effective fermentation strain for further experiments. Cell viability is a functional feature used to evaluate bacterial growth (28). The strain Y2 grew well in ginkgo biloba kernel juice without any nutritional supplements according to the number of viable cells of strain Y2 in the fermenting process, showing that ginkgo biloba kernel juice could be used as a fermentation substrate (Figure 2A). The strain Y2 grew rapidly in the logarithmic phase from 4 to 8 h, and the number of viable bacteria increased significantly, indicating that Y2 was suitable for growth in ginkgo biloba kernel juice. The number of viable cells started at 5.0 ± 0.02 log CFU/mL and grew rapidly to 10.0 ± 0.03 log CFU/mL after 8 h of fermentation. After 24 h of fermentation, the number of viable cells stabilized at 12.1 ± 0.08 log CFU/mL. After 48 h of fermentation, the number of viable cells of LAB Y2 remained at 8.4 ± 0.03 log CFU/mL. The initial pH of ginkgo biloba kernel juice was 6.2 ± 0.04 (Figure 2B). In the early stage of fermentation, a large amount of lactic acid was produced, and the pH value of the fermentation broth decreased rapidly. After 12 h of fermentation, the rate of pH continued to gradually slow down reaching a final pH of 3.27 ± 0.07 at 48 h. According to the characteristics of lactic acid bacteria fermentation and acid production, Y2 was a LAB with strong acid production ability and good fermentation effect. The changes in antioxidant activity during the fermentation of ginkgo biloba kernel juice were shown in Figures 2C,D. The ABTS+ radical scavenging capacity and FRAP reducing capacity increased with time. After 48 h of fermentation, the free radical scavenging rate of ABTS+ increased from 4.52 ± 0.72 to 13.69 ± 0.32 mM, and the reducing capacity of FRAP increased from 0.35 ± 0.01 to 0.62 ± 0.02 mM. After 0–4 h of fermentation, the ABTS+ free radical scavenging rate and FRAP reducing ability increased significantly, and the antioxidant activity of the fermentation broth improved significantly. After 24 h of fermentation, the lifting speed approached slowly, and the antioxidant activity was basically stable at the end of fermentation. Carbon source catabolism could provide energy for the growth of Lactobacillus and indirectly promote an increase in organic acid content (29). The total sugar content decreased significantly from 12 to 36 h of fermentation, indicating that the lactic acid bacteria in the stable growth period consumed large amounts of carbon sources (Table 2). As fermentation entered the decay period (36–48 h of fermentation), lower substrate pH and reduced nutrients limited the growth of Y2 and indirectly affected its sugar metabolism, resulting in a stabilization of the total sugar contents (30). The consumption rate exceeded 44.85% after the fermentation ended. During the fermentation, the organic acids were transformed into each other over time, and the total organic acid content increased by about approximately 5.86% after 48 h of fermentation. The components of all fermentation time periods were well separated from the control samples according to the component map (Figure 3A), showing that the fermentation of Y2 significantly changed the organic acid composition in ginkgo biloba kernel juice. According to the loading diagram (Figure 3B), malic acid, tartaric acid, citric acid, pyruvic acid and quinic acid all had high positive values for PCI, indicating that the content of these substances decreased during the fermentation process. While the decrease in malic acid was particularly obvious, levels of oxalic acid did not observe any changes during fermentation. Lactic acid, shikimic acid, fumaric acid, and succinic acid had high negative values for PCI, reflecting a general increase in content after fermentation. Ginkgo kernel was rich in a variety of natural organic acids, of which malic acid had the highest content (926.91 ± 5.35 μg/mL), followed by pyruvic acid (80.27 ± 0.27 μg/mL), citric acid (43.77 ± 0.55 μg/mL) and tartaric acid (632.34 ± 19.98 μg/mL) (Table 3). Lactic acid was the main organic acid produced by the consumption of sugars by lactic acid bacteria in ginkgo biloba kernel juice, and the lactic acid content continued to increase throughout the fermentation process. The initial lactic acid content was 403.33 ± 17.77 μg/mL and increased to 722.58 ± 29.76 μg/mL after 48 h of fermentation. The production of abundant lactic acid reduced the pH value of the fermentation substrate. Pyruvate and citric acid could be decompose into various products such as lactic acid and acetic acid during fermentation (31). Malic acid, which was a good carbon source, observed the highest proportion of total organic acids. The content of malic acid decreased by about 25.25% during the fermentation process, while the content of lactic acid and succinic acid increased. Tartaric acid could be decomposed into various products, such as gluconic acid, which was then oxidized to 2-keto-D-gluconic acid (2-KGA) and 5-keto-D-gluconic acid (5-KGA) (32). During fermentation, the tartaric acid content gradually decreased, and the reduction rate was about 37.42%. In conclusion, lactic acid bacteria fermented ginkgo biloba kernel juice could transform and generate a variety of organic acids. Organic acid was the base of taste substance of ginkgo biloba kernel juice, and was the precursor of many flavor substances. Volatile flavor substances such as acids, alcohols, and aldehydes will be produced in the process of microbial metabolism. Therefore, we evaluated the changes of flavor substances before and after fermentation (Table 5). Lactic acid bacteria could degrade macromolecular proteins into small peptide chains, amino acid proteases and peptidases through their own reproductive metabolism and hydrolysis during fermentation (33). Under the catalytic reaction, the proteases could degrade the protein into smaller peptides that are degraded into the small-molecule free amino acids by the action of the peptidase (34). The total protein content of ginkgo biloba kernel juice decreased continuously during the fermentation process, and approximately 67.51% of the metabolism was consumed at the end of fermentation (Table 2). Among all free amino acids, glutamic acid had the highest content (74.22 ± 0.01 μg/mL), followed by arginine (55.68 ± 0.01 μg/mL) and alanine (20.83 ± 0.05 μg/mL) (Table 4). Glutamate, free histidine and alanine levels decreased significantly after fermentation. During the fermentation process, under the action of lactic acid bacteria decarboxylase, Glu is converted into γ-aminobutyric acid, and Asp is converted into Ala (35). The combination of aspartic acid, glutamic acid and Na+ makes the sample umami, which would give ginkgo biloba juice rich flavor substances. In addition, the content of aromatic amino acids (Phe) decreased significantly after fermentation, while the content of another aromatic tyrosine (Tyr) increased, owing to the effect of amino acid invertase on the fermentation process (36). The content of branched-chain amino acids (Leu, Ile) also decreased significantly after fermentation, then were metabolized into keto acids, alcohols and fatty acids through amino acid converting enzymes. Phe and Thr could be converted into benzyl alcohol, phenylethanol and acetaldehyde with certain flavor under the action of an amino acid converting enzyme. Arginine is converted to citrulline by the enzyme arginine deiminase during lactic acid fermentation, which is further broken down into ornithine and free ammonia (37). The total free amino acid content increased by approximately 6.88% after 48 h of fermentation. The effects of lactic acid fermentation on free amino acid profiles were studied using principal component analysis (PCA). In principal component analysis, PC1 and PC2 accounted for 98.2 and 1.8% of the data variance, respectively. As shown in the composition diagram (Figure 3C), the samples of unfermented ginkgo biloba kernel juice were located on the positive side of PC1, and the samples fermented for 48 h were distributed on the negative side. This indicated that lactic acid bacteria had a significant effect on the change in free amino acid content during the fermentation of ginkgo biloba kernel juice. Specific changes in the free amino acids were observed through in the loading diagram (Figure 3D). Phe, Thr, Leu, Ile, Asp, Ser, Ala, and free histidines were all positively correlated with PC1, indicating that these amino acids would be affected by the fermentation process. Whereas, Lys, Val, Met, Gly, Cys, Tyr, and Arg were all negatively correlated with PC2, indicating that their content increased with the fermentation time. Ginkgo kernels were also rich in a variety of phenolic substances, which led to the production of many secondary metabolites during fermentation, including phenolic acids, stilbenes, lignins, and aromatic compounds (38). Therefore, it is necessary to study the effect of LAB fermentation on the phenolic components in ginkgo biloba kernel juice. As shown in Table 1, the total phenolic content of ginkgo biloba kernel juice at the beginning of fermentation was 116.51 ± 0.51 mg/L. It decreased significantly within 4 h of fermentation, and then increased slowly. After 48 h of fermentation, the total phenolic content of the samples was approximately 9.72% higher than that of the unfermented samples. Procyanidin B2 is an oligomer, dimer and polymer of catechin (39). The highest content of flavonoids in ginkgo kernel was about 7.32 ± 0.06 μg/mL of catechin and about 125.92 ± 0.56 μg/mL of procyanidin B2 (Table 5). Procyanidin B2 increased by approximately 20.93%, and catechin increased by approximately 30.46% after 4 h of fermentation. The increase in the total flavonoid content could result in an increase in the antioxidant activity of the fermented samples. In the process of fermenting ginkgo biloba kernel juice, the improvement in antioxidant activity was highly correlated with the transformation of phenolic substances (40). Previous studies have reported that phenolic compounds could be used as reducing agents as well as free radical scavengers and singlet oxygen quenchers (41). Phenols could be converted into each other, and the antioxidant activity of the fermentation broth can be improved (42). It was also shown that the antioxidant capacity of phenolic substances in ginkgo biloba kernel juice was significantly improved after Y2 fermentation (Figures 2C,D). There is also a certain relationship between FRAP and phenolic content, and the presence of phenolic substances can effectively reduce Fe3+ to Fe2+ form (43). Therefore, the increase in phenolic substances was an important factor affecting the antioxidant activity of fermentation. As shown in Figures 3E,F, principal component analysis (PCA) was used to analyze the effect of Y2 fermentation on phenolic components in ginkgo biloba kernel juice. After 48 h of fermentation, phloridzin, epicatechin and gallate in the samples were metabolized to form gallic acid and phloretin with a strong antioxidant quality (44). L. plantarum fermentation could produce some phenolic acid decarboxylases, such as ferulic acid. These extracellular enzymes act on phenolic components, and from a chemical point of view, glycosylation, methylation and other types of substitutions are carried out, resulting in structural changes. This lead to the mutual conversion of phenolic substances in the sample, thereby changing their oxidative activity (45). The p-hydroxybenzoic acid content decreased by approximately 17.95% after fermentation. Further, the content of chlorogenic acid was the highest at 4 h of fermentation (3.64 ± 0.08 μg/mL), and it increased by about 20.17% at the end of fermentation. The increase in chlorogenic acid content leads to an increase in pharmacological functions, such as antioxidant, antiviral, anti-cancer and antibacterial capacities (46). During the fermentation period of 4–8 h, the content of chlorogenic acid gradually decreased, potentially under the action of extracellular enzyme hydrolysis, which indirectly led to an increase in caffeic acid, p-coumaric acid and ferulic acid. Phenolic acids not only have various pharmacological effects such as anti-inflammatory, antioxidant, anti-mutation, and anti-cardiovascular disease, but also have co-color effects (such as ferulic acid), which can enhance the stability of their own colors after fermentation (47). There are few studies on the volatile flavor compounds after fermentation, and further research is needed. A total of 64 volatile flavor compounds were identified in ginkgo biloba kernel juice before and after fermentation, including three kinds of aldehydes, 18 kinds of alcohols, 9 kinds of ketones, 8 kinds of esters, 29 kinds of hydrocarbons, 6 kinds of phenols, 3 kinds of heterocycles, 7 kinds of acids, and two other types. From the analysis results, the fermented samples and the unfermented samples were quite different, and the similarities of volatile flavor substances were low. The relative content and species of ginkgo biloba kernel juice in alcohols, acids, and heterocyclic compounds were significantly increased after 48 h of fermentation (Figure 4). The main aroma substances in unfermented ginkgo biloba kernel juice were 2-ethyl-1-hexanol, 1-octanol, and n-octanol (Table 6). The total relative content of volatile flavor compounds increased by 125.48% after fermention. Among them, the total relative content of alcohol increased by about 104.69%, potentially due to the emergence of new generation of various volatile flavor substances. After 48 h of fermentation, the contents of 1-octanol and n-octanol increased significantly, the contents of 2-ethyl-1-hexanol and 1-octanol decreased, and 1-heptanol, 1-octen-3-ol and 2-octen-1-ol newly formed. The relatively high contents of newly generated alcohols were geraniol (1.81%), 1-octen-3-ol (2.65%) and 1-heptanol (3.24%). These alcohols were mainly produced by metabolism during microbial processes, degradation of unsaturated fatty acids, and reduction of carbon-based compounds during fermentation. Aldehydes are unstable compounds, and the relative content of aldehydes increased by about 126.5%, among which the highest content was 16.15% 2,5-dimethylbenzaldehyde. The relative content of hydrocarbons increased by about 32.92%, while the types of hydrocarbons decreased after fermentation. Meanwhile, the relative content of newly generated aromatic compounds was higher, of which 1,6-dimethyl-4-(1-methylethyl)-naphthalene accounted for about 2.18% of the total relative content after fermentation. The increase of total content of alcohols and esters in ginkgo biloba kernel juice could mpart stronger fruity and floral aromas (48). Under the metabolic activity of lactic acid bacteria, more aldehydes are reduced to alcohols and acids, and the increase of alcohols will increase the corresponding esters, which is also the reduction of aldehydes in fermented ginkgo biloba kernel juice, while alcohol, the reasons for the increase in esters (49). The content of acids increased by about 786.01%, and the relative content of acids increased significantly. The strain Y2 produced a large amount of acids during the fermentation process, among which the new acids were 2.16% methoxy-acetic acid, 1.02% dodecanoic acid, 0.29% 2-methylbutyric acid, 0.24% 2-ethylhexanoic acid. Among them, the relative contents of L-lactic acid and propionic anhydride decreased after the fermentation, while the relative content of acetic anhydride was high (4.18%). Meanwhile, the relative content of phenols changed scarcely, and eugenol (7.48%) and 3-ethylphenol (0.29%) were newly formed at the end of fermentation. The former was produced by the reaction of guaiacol and allyl alcohol, and had strong antibacterial activity. The latter had an aromatic odor and was widely used in organic synthesis and as a solvent (50). Future research will focus on the transformation mechanism between these molecules and their effect on the flavor of ginkgo biloba kernel juice. 4′-O-methylpyridoxine (MPN), also known as ginkgo toxin, is an anti-vitamin B6 compound, and one of the toxic components in ginkgo kernels. If ingested in large amounts, adverse symptoms, such as vomiting, clonic seizures, and loss of consciousness could occur (51). Relevant studies had shown that MPN could be effectively converted into 4′-O-methylpyridoxine-5-glycoside (MPNG) after heating, microwave irradiation and boiling (52). In this study, the MPN content in ginkgo biloba kernel juice was identified, by comparison with standard chemicals determined by high performance liquid phase (HPLC) analysis (Figure 5). The content of MPN was reduced by roughly 82.25% after fermentation for 48 h, indicating that strain Y2 could effectively gradate MPN. Relevant studies have shown that lactic acid bacteria have the function of adsorbing toxins during the fermentation process. During the growth of lactic acid bacteria, a large number of polysaccharides and peptidoglycan were produced on the cell wall, which played a key role in adsorbing the toxic substance 4′-O-methylpyridoxine in ginkgo biloba kernel juice (53). It also had been reported that MPN could be phosphorylated to form 4′-O-methylpyridoxine-5′-phosphate (MPNP) under the action of phosphoric acid produced during LAB fermentation (54). However, there are few studies on the degradation mechanism of 4′-O-methylpyridoxine content by LAB fermentation, and we will further investigate how strain Y2 bio-transformed MPN. This work showed that LAB fermentation was a potential way to improve the beneficial value and safety of ginkgo biloba kernel juice. The total relative content of volatile flavor substances increased by about 125.48%, and 24 new substances were added. The biotransformation and biosynthesis content of the phenols was increased, and the total phenolic concentration increased by about 9.72%, which indirectly led to the enhance of antioxidant capacity of the fermentation. The increase of the aromatic amino acids and volatile flavor substances reflected the metabolism of lactic acid bacteria in fruit juice. In addition, 4′-O-methylpyridoxine, as one of the toxic substances in ginkgo biloba kernel, showed a degradation rate above 82.25%, and the total ginkgo acid content in the final product was less than 41.53 mg/L. As few reports about the degradation of 4′-O-methylpyridoxine by fermentation, further studied will focus on the degradation mechanism. Our results showed that the LAB fermented ginkgo biloba kernel juice had broad application prospects. The original contributions presented in this study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s. JY and YS performed the experiments and analyzed the data. JC, YC, HZ, and TG drafted the manuscript. FX and SP analyzed and discussed the data. YT and JL contributed to the writing—review, editing, and funding acquisition. All authors have read and agreed to the published version of the manuscript.

PMC9649922

Zhuoran Tang,Qi Wang,Peixin Chen,Haoyue Guo,Jinpeng Shi,Yingying Pan,Chunyu Li,Caicun Zhou

Computational recognition of LncRNA signatures in tumor-associated neutrophils could have implications for immunotherapy and prognostic outcome of non-small cell lung cancer 10.3389/fgene.2022.1002699

28-10-2022

non-small cell lung cancer,tumor-associated neutrophils,long noncoding RNA,immunotherapy,computational recognition

Cancer immune function and tumor microenvironment are governed by long noncoding RNAs (lncRNAs). Nevertheless, it has yet to be established whether lncRNAs play a role in tumor-associated neutrophils (TANs). Here, a computing framework based on machine learning was used to identify neutrophil-specific lncRNA with prognostic significance in squamous cell carcinoma and lung adenocarcinoma using univariate Cox regression to comprehensively analyze immune, lncRNA, and clinical characteristics. The risk score was determined using LASSO Cox regression analysis. Meanwhile, we named this risk score as “TANlncSig.” TANlncSig was able to distinguish between better and worse survival outcomes in various patient datasets independently of other clinical variables. Functional assessment of TANlncSig showed it is a marker of myeloid cell infiltration into tumor infiltration and myeloid cells directly or indirectly inhibit the anti-tumor immune response by secreting cytokines, expressing immunosuppressive receptors, and altering metabolic processes. Our findings highlighted the value of TANlncSig in TME as a marker of immune cell infiltration and showed the values of lncRNAs as indicators of immunotherapy.

Computational recognition of LncRNA signatures in tumor-associated neutrophils could have implications for immunotherapy and prognostic outcome of non-small cell lung cancer 10.3389/fgene.2022.1002699 Cancer immune function and tumor microenvironment are governed by long noncoding RNAs (lncRNAs). Nevertheless, it has yet to be established whether lncRNAs play a role in tumor-associated neutrophils (TANs). Here, a computing framework based on machine learning was used to identify neutrophil-specific lncRNA with prognostic significance in squamous cell carcinoma and lung adenocarcinoma using univariate Cox regression to comprehensively analyze immune, lncRNA, and clinical characteristics. The risk score was determined using LASSO Cox regression analysis. Meanwhile, we named this risk score as “TANlncSig.” TANlncSig was able to distinguish between better and worse survival outcomes in various patient datasets independently of other clinical variables. Functional assessment of TANlncSig showed it is a marker of myeloid cell infiltration into tumor infiltration and myeloid cells directly or indirectly inhibit the anti-tumor immune response by secreting cytokines, expressing immunosuppressive receptors, and altering metabolic processes. Our findings highlighted the value of TANlncSig in TME as a marker of immune cell infiltration and showed the values of lncRNAs as indicators of immunotherapy. Lung cancer is related with high mortality rates in China with non-small cell lung cancer (NSCLC) accounting for >80% of lung cancers (Zhu et al., 2017). The administration of immune checkpoint inhibitors (ICIs) in cancer therapy has had remarkable results (Yue et al., 2018; Dolladille et al., 2020; Galluzzi et al., 2020). For advanced non-small cell lung cancer (NSCLC), several clinical trials have confirmed that as first- or second-line treatment, ICIs are superior to platinum-based chemotherapy (Ko et al., 2018; Vansteenkiste et al., 2019; Chen et al., 2020). However, only 20%–40% of advanced NSCLC patients achieve sustained clinical benefits from PD-(L)1 inhibitor therapy, with most patients having primary or acquired resistance to immunotherapy (Socinski, 2014). Moreover, those who do not respond to immunotherapy may suffer immune-related adverse events (IRAE) and the high costs of anti-PD-(L)1 monoclonal antibody therapy (Khoja et al., 2017; Das and Johnson, 2019; Schoenfeld et al., 2019). Thus, effective biomarkers that distinguish potential responders from non-responders, and indicate patient clinical response in real-time are urgently needed to improve treatment outcomes. The TME is comprised of a complex cell population that includes tissue-resident lymphocytes, fibroblasts, endothelial cells, and neurons that are present before tumorigenesis, as well as blood-derived cells recruited to tumor sites (Butturini et al., 2019). Immune cells are the main cellular components in tumors. Tumor-infiltrating myeloid cells, including tumor-associated macrophages (TAM), regulatory dendritic cells, tumor-associated neutrophils (TAN), myeloid-derived suppressor cells (MDSC), as well as tolerogenic dendritic cells (TOL-DC), facilitate the formation of immunosuppressive microenvironments (Schupp et al., 2019). These cells directly or indirectly inhibit the antitumor immune response by secreting cytokines, expressing immunosuppressive receptors, and altering metabolic processes, leading to tumor immune escape. Tumor-associated neutrophils (TANs) are a key part of tumor-infiltrating myeloid cells and are regularly detected in the TME. Clinically, TANs can be used to predict treatment outcomes and immunotherapy response (Nielsen et al., 2021). Transcriptomic studies have identified gene expression biomarkers as well as signatures for quantitative assessment of TANs, as well as for stratification based on prognoses and immunotherapeutic response (Lecot et al., 2019; Wu and Zhang, 2020). Long non-coding RNA (lncRNAs) influence almost all biological processes and pathways, and their dysregulation is associated with various diseases. Additionally, lncRNAs have wide functional diversity due to their influence on gene expression levels at transcriptional, post-transcriptional and epigenetic levels (Rinn and Chang, 2012; Fatica and Bozzoni, 2014; Marchese et al., 2017; Bao et al., 2020). The correlation between lncRNAs and immune function has been reported. Recent studies have shown that lncRNAs are abundant with cell type specificity in various immune cell subsets (Rinn and Chang, 2012; Atianand et al., 2017; Chen et al., 2017; Zhou et al., 2017; Zhou et al., 2018). LncRNAs expression pattern has been correlated with infiltrations of immune cells into the TME (Hu et al., 2013; Ranzani et al., 2015; Sage et al., 2018; Wang et al., 2018; Zhao et al., 2021). Nevertheless, neutrophil-specific lncRNAs as well as their significance in assessing TANs and prediction of clinical outcomes and immunotherapeutic responses require further study. Here, a computational framework is proposed for determining neutrophil-specific lncRNA expression levels and lncRNA signatures for TANs (TANLncSig) via integrative immune, lncRNA, and clinical profiling analyses. The TANLncSig’s ability to predict clinical outcome and response to immunotherapy by NSCLC patients was also investigated. The data set can be obtained from the GEO database with series accession number GSE28490 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE28490), These included chip data on the expression of nine human immune cells (neutrophils, monocytes, B cells, eosinophils, CD4 T cells, NK cells, mDCs, CD8 T cells, and pDCs). The GEO2R tool from GEO was used for differential expression analysis. Using adjusted p = <0.05 and logFC >1 as cutoff thresholds identified 17 lncRNAs with high neutrophil-specific expression. Clinical data and TCGA RNA-seq datasets for LUSC and LUAD were downloaded by the UCSC Xena browser (https://xenabrowser.net/). Lusc-LINC01272-neutrophils malignant/Luad-LINC01272-neutrophils malignant results from single cell sequencing datasets. First, a monovariate Cox regression analysis was used to find neutrophil-specific lncRNAs with prognostic value in LUSC and LUAD, and LASSO Cox regression was used to determine their risk scores. The multivariate Cox regression analysis (age, risk score, tumor stage, gender), Kaplan-Meier (KM) survival analysis and 3, 5, and 10 years survival AUCs were used to evaluate risk score. Multivariate ANOVA was used to analyze differences between neutrophil-specific, highly expressed lncRNA and risk score in LUSC and LUAD samples at various TNM stages. In LUSC and LUAD samples, genes with mean expression levels >1 were identified and their correlation with risk score analyzed. 1,000 genes with the highest absolute correlation coefficient value were selected from those with positive correlation coefficients (>0, p = <0.05) and those with negative correlation coefficients (<0, p = <0.05). ClusterProfiler for R was used to analyze GO terms of biological process (BP), Molecular function (MF), cellular component (CC), and KEGG pathway enrichment analyses. After gene enrichment, the adjusted p-value < 0.05 and the smallest TOP10 was selected for mapping. Pearson correlation analysis was used to determine correlations between risk score, neutropen-specific lncRNAs, and the expression of common immune checkpoint inhibitors and correlation heat maps drawn, with * denoting p ≤ 0.01 while + denotes p ≤ 0.05. To recognize neutrophil-specific lncRNAs, dataset GSE28490 was downloaded from GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE28490). This dataset includes chip data on expressions of nine human immune cells (CD4+ T cells, neutrophils, monocytes, B cells, eosinophils, CD8+ T cells, NK cells, mDCs, and pDCs). Using GEO2R, 17 lncRNA specifically highly-expressed in neutrophils (p =<0.05, log2>1) were identified. These neutrophil-specific lncRNAs are referred to as TAN-associated lncRNAs (TANlncRNA) (Figure 1). To develop a neutrophil-specific lncRNA risk score for predicting prognosis, the TCGA NA-SEQ dataset, TCGA lung squamous cell carcinoma (LUSC) as well as adenocarcinoma (LUAD) gene expression data, clinical features, and prognosis data were downloaded from UCSC Xena. First, univariate Cox regression analyses were used to establish neutrophil-specific lncRNAs with prognostic value in LUSC and LUAD. The final signature named TANlncSig (Table 1). This analysis identified three lncRNAs with prognostic value in LUSC (LINC01272, LINC00261, LINC00668, p = <0.05). Using these three lncRNAs, the expression value of lncRNA was weighted using multivariate Cox regression coefficient to obtain risk scores via the formula: risk score = 0.09 * LNC00668 + 0.17 * LNC00261. Then, TANlncSig scores for every patient in the discovery dataset were determined, after which the 542 patients were grouped into the high (n = 271) or low (n = 271) risk groups. Low risk group patients were found to have longer overall survival (OS) relative to the high-risk group patients (p = 0.039, ≤0.05, Figure 2A). Multivariate Cox regression analyses revealed that risk score (p < 0.001), stage (p < 0.001), age (p = 0.037, ≤0.05), and gender (p = 0.007, ≤0.01) significantly affected the prognostic outcomes of LUSC patients. The p-value and hazard ratio of TANlncSig were better than those of stage and age (Figure 2B). That said, TANlncSig has the potential to be a good predictor of efficacy. The predictive capacity of TANlncSig was authenticated using the TCGA internal testing dataset and revealed the 3-, 5-, and 10-year OS rates for low-risk group patients to be 60.42, 54.47, and 54.23%, respectively (Figure 2C). Indicating that risk score significantly correlates with OS in LUSC. Similar analysis was done for LUAD. First, three lncRNAs with prognostic values (LINC00528, LINC00967, and LINC00261) were identified using univariate Cox analysis. Using the above three lncRNAs, lncRNAs expression value was weighted by multivariate Cox regression coefficient to determine risk score using the formula: risk score = −5.32 * LINC00967-0.16 * LINC00261-0.74 * LINC00528. Patients with LUAD in the low-risk group had longer OS relative to high-risk group LUAD patients (p = 0.0029, ≤0.01, Figure 3A). Cox multivariate regression analyses revealed that risk score (p < 0.001) and stage (p < 0.001) significantly correlated with LUAD prognosis. In lung adenocarcinoma, the p-value and hazard ratio of TANlncSig were equally better than those of stage and age (Figure 3B). The 3-, 5-, and 10-year OS rates in low-risk group patients were 61.01, 61.20, and 65.30%, respectively (Figure 3C). These results indicate that risk scores in the LUAD dataset significantly correlate with patients’ OS. Clinical phenotypic correlation analysis of single prognostic lncRNA and risk score (tumor stage, T, N, and M staging) was performed in lung adenocarcinoma as well as squamous cell carcinoma. According to statistical analysis, the risk score in different tumor stages of lung squamous cell carcinoma showed significant statistical differences, and the statistical results showed that p = 0.0013, <0.01 (Figure 4A). The risk score in different tumor stages of lung adenocarcinoma also showed significant statistical differences (p = 0.0081, <0.01) (Figure 4B). The TNM staging system is the most widely used tumor staging system, worldwide. T denotes tumor sizes and local invasion range, N denotes lymph node involvement, and M denotes distant metastasis. TNM staging has great clinical value in prognosis prediction (Ficarra et al., 2007; Moch et al., 2009). The risk score lack of significance in different T stages and N stages of lung squamous cell carcinoma (Figures 5A,B). The risk score has significant statistical difference in different M stages of lung squamous cell carcinoma (p = 0.011, <0.05) (Figure 5C). The risk score has significant statistical difference in different T stages (T1, T2, T3, and T4 stages) of lung adenocarcinoma (p = 0.0023, <0.01) (Figure 5D). Similarly, the risk score has significant statistical difference in different N stages (N0, N1, N2, and N3 stages) of lung adenocarcinoma (p = 0.013, <0.05) (Figure 5E). The risk score lack of significance in different M stages of lung adenocarcinoma (Figure 5F). In LUSC and LUAD samples, genes with average expression levels >1 were identified and their risk scores analyzed. 1,000 genes with the largest absolute correlation coefficient values were selected from positive (correlation coefficient >0, p ≤ 0.05) and negative (correlation coefficient <0, p ≤ 0.05) and correlation genes and pathway enrichment analysis done using cluster profiler on R. In LUSC, positive correlation genes are mainly associated with biological processes (BP) associated with T-cell activation, leukocyte proliferation, and leukocyte cell-cell adhesion. For cellular component (CC) they were enriched in endocytic vesicle, tertiary granule, and secretory granule membrane. For molecular function (MF), they were enriched in immune receptor activity and cytokine binding. KEGG pathway analysis revealed enrichment mainly for cell adhesion molecules cams (Figure 6A). Negative correlation genes in lung squamous cell carcinoma are mainly enriched for biological processes (BP) associated with skin development, epidermis development, and cornification. For cellular component (CC), they were enriched for cornified envelope, desmosome, and cell-cell junction. For molecular function (MF), they were enriched for microtubule binding and tubulin binding. For KEGG pathways, they were enriched for basal cell carcinoma (Figure 6B). Positive correlation genes in lung adenocarcinoma were mainly enriched in biological processes (BP) associated with translational termination and adenocarcinoma. For cellular component (CC), they were enriched for ribosomal subunits, ribosome and large ribosomal subunit. For molecular function (MF) they were enriched for structural constituent of ribosome and cadherin binding. For KEGG pathways, they were enriched for ribosome and cell cycle (Figure 6C). Genes associated with negative correlations in LUAD are involved in biological processes (BPs) associated with lymphocyte differentiation, leukocyte proliferation, and antigen receptor-mediated signaling. For cellular component (CC), they were enriched for external side of plasma membrane and immunological synapse. For molecular functions (MFs), they were enriched for guanyl-nucleotide exchange factor activity. For KEGG pathways, they were enriched for primary immunodeficiency and B-cell receptor signaling pathway (Figure 6D). In accordance with previously reported expression levels of the immune cell specific marker genes, cibersort (https://cibersort.stanford.edu/) was further used to evaluate the levels of immune infiltration of 22 immune subpopulations in high-risk and low-risk patient groups. t-test was performed to determine the difference in lymphocyte infiltration levels between the two groups. As shown in Figures 7A,B, in both LUSC and LUAD, high-risk patients were significantly enriched in 12 immune subpopulations, while low-risk patients were enriched in 10 immune subpopulations. Additionally, mononuclear immune cells, including neutrophils, were found to infiltrate significantly more in the high-risk patient group than in several other groups. Single-cell sequencing data of LUSC and LUAD downloaded from GSE127465, cell type notes downloaded from TISCH (http://tisch.comp-genomics.org/). The homologous expression levels of LINC01272 of the TANlncSig in neutrophil cell lines differed significantly from those of malignant cell lines according to a subsequent analysis of neutrophil cell lines (Figures 7C,D). This indicates that these lncRNAs are expressed differently in neutrophils compared with malignant cells. In the above study, we found that the TANlncSig was not only associated with patient prognosis but also as a TAN indicator. TANlncSig was further validated in independent datasets by the microarray platform in order to verify versatility and robustness of TANlncSig. The Affymetrix HG-U133 Plus 2.0 platform was used to analyze 83 LUAD patients from the GSE30219 dataset. As demonstrated again, TANlncSig can distinguish between patients who have high and low survival risk. A total of 83 patients were stratified into 41 high-risk patients and 42 low-risk patients in the GSE30219 dataset. Furthermore, patients in the high-risk group had a marginally poorer outcome than those in the low-risk group (p = 0.0024, ≤0.01; log-rank test) (Figure 8A). The AUC of ROC curve at 3, 5, and 10 years were 64.13, 66.87, and 60.58% respectively (Figure 8B). The results show that TANlncsig can accurately predict the 5-year overall survival of patients, indicating that TANlncsig has good efficacy and certain stability. In order to investigate whether TANlncSig is an independent prognostic factor, a multivariate Cox regression analysis was conducted in patient cohorts. In the independent GSE30219 dataset, the TANlncSig still maintained a significant correlation with OS in the multivariate analysis (HR = 6.74, 95% CI 1.283-35.5, p = 0.024, ≤0.01). Thus, these results demonstrate that the TANlncSig helps predict OS independently of other conventional clinical factors (Figure 8C). Next, prognostic lncRNAs and risk score were correlated with immune checkpoint molecules expression in LUSC and LUAD patients. In LUSC, risk score, LINC01272, and LINC00261 positively correlated with the expression of most ICBs, while LINC00668 had negative correlations with the expression of most ICBs (Figure 9A). In LUAD, risk score had negative correlations with expression levels of most ICBs, while LINC00528 positively correlated with expression levels of most ICBs (Figure 9B). The expressions of risk score were divided into high and low groups and combined according to the median. The combination was used to analyze the prognosis of immunotherapy for non-small cell lung cancer. In lung squamous cell carcinoma, the combination of CEACAM1, TNFSF4, gem, CD47, vtcn1 and risk score can well stratify the prognosis of patients. In lung adenocarcinoma, all ICB molecules combined with risk score can well predict the prognosis of patients. These results suggest that risk score can be used as an index to predict the response of patients to immunotherapy. In the peripheral blood, neutrophils are the most abundant white blood cells (Dinh et al., 2020). They have a central role in human non-specific immunity. Previous studies suggest that neutrophils inhibit tumors by secreting cytokines and producing reactive oxygen species (Vaughan and Walsh, 2005; Mishalian et al., 2013; Coffelt et al., 2015; Ponzetta et al., 2019). However, other studies indicate that neutrophils in the tumor microenvironment (TME) promote tumorigenesis. Cytokines and chemokines production by invasive neutrophils might affect the recruitment and activation of inflammatory cells in the TME, create an immunosuppressive microenvironment that is conducive for tumorigenesis, regulate tumor growth, metastasis and angiogenesis, and influence patient prognosis. Traditional methods for quantifying tumor immune cells infiltration based on histology or immunohistochemistry may have bias and variabilities (Yoshihara et al., 2013; Gibney et al., 2016; Spranger and Gajewski, 2018; Zhang et al., 2020; Sanchez-Pino et al., 2021). More recently, RNA-seq analyses have shown that lncRNAs exhibit a better degree of cell type specificity, relative to protein-coding genes in immune cells, highlighting their potential as subpopulation-specific immune cells molecular markers (Huang et al., 2018; Chen et al., 2019; Zhou et al., 2021). Here, we used a machine learning-based computational framework to identify lncRNA features for evaluating TANs and explored their clinical significance using a combination of lncRNA, immune, and clinical spectrum analyses. The computational framework was used on the TCGA discovery dataset of NSCLC to identify a lncRNA signature (TANlncSig) comprised of 17 lncRNAs obtained from a list of neutrophil-specific lncRNAs using machine learning. Functional enrichment analysis of TANlncSig-related mRNAs showed that TANlncSig is highly correlated with cancer markers of immune response and sustained proliferative signals. Recent experimental evidence on some TANlncSig components is consistent with functional annotations using bioinformatics. It appears that Mir-1303, which is upregulated in tumor tissues, acts as a sponge for LINC01272 and negatively correlates with its expression. A reduction in LINC01272 expression in tissues and cells of NSCLC patients may serve as an independent prognostic marker. LINC01272 overexpression may inhibit NSCLC cells proliferation, migration, and invasion by inhibiting MI-1303 (Zhang and Zhou, 2021). LINC00261 downregulation in gastric cancer is associated with poor prognosis. Ectopic LINC00261 expression disrupts cell migration and invasion, inhibiting metastasis in vitro as well as in vivo. LINC00261 downregulation promotes cell migration and invasion in vitro. LINC00261 overexpression influences epithelial-mesenchymal transition (EMT) through the regulation of E-cadherin, Vimentin and N-cadherin (Liu et al., 2020; Zhai et al., 2021). LINC00668 expression is significantly upregulated via STAT3 signaling in NSCLC tissues as well as cell lines. Clinical studies show that upregulated LINC00668 correlates with histological grade, advanced TNM stage, and lymph node metastasis. Additionally, multivariate analyses established that LINC00668 as an independent marker of overall survival (OS) in patients with NSCLC. LINC00668 downregulation inhibits proliferation, migration, and invasion of NSCLC cells and promotes apoptosis. Mechanistically, LINC00668 is a direct target of miR-193a, leading to down-regulation in the expression of its target gene KLF7. STAT3-initiated LINC00668 promotes NSCLC progression by upregulating KLF7 via sponging Mir-193a. Therefore, it may serve as a prognostic marker and therapeutic target for NSCLC (An et al., 2019). From the perspective of lncRNA, TANlncSig seems to be a transcriptional marker as a potentially measurable indicator of neutrophil activity and prognosis. To further assess TANlncSig’s role in clinical risk stratification, we evaluated its relationship with survival in patients with NSCLC. When applied to the TCGA RNAseq patient dataset, TANlncSig significantly correlated with patient survival. In TANlncSig, three lung squamous cell carcinoma, neutrophil-specific lncRNAs (LINC01272, LINC00261, and LINC00668) were markedly associated with prognostic outcomes. In lung adenocarcinoma, three neutrophil-specific lncRNAs (LINC00528, LINC00967, and LINC00261) significantly correlated with prognosis. In squamous cell carcinoma and lung adenocarcinoma, correlation analysis of individual lncRNAs and risk score with clinical features (TNM staging) revealed that risk score varied significantly with tumor stage. After adjusting for traditional clinical factors, TANlncSig was verified to be an independent prognostic marker for differentiating between poor and good survival outcomes across patient datasets. Immune checkpoint inhibitors (ICIs) have emerged as effective lung cancer immunotherapies (Suresh et al., 2018; Iams et al., 2020). Some of the drugs acting on the immune checkpoints, CTLA4 and PD-1/PD-L1, have excellent performance against various tumors. Although significant breakthroughs have been made on CTLA4 and PD-1/PD-L1 antibodies, single-drug effective rates are only about 20%, and they benefit a limited proportion of patients (Magiera-Mularz et al., 2017; Lingel and Brunner-Weinzierl, 2019; Rotte, 2019; Yang and Hu, 2019; Liu and Zheng, 2020). The limited efficacy is attributable to the immune system’s complexity. Indeed, immune cells, cytokines, and immune adjuvants in the TME interact with each other, limiting the effects of drugs on individual targets. Thus, drugs that target different links and aspects of tumor immunity are needed to enhance immunotherapy outcomes. Up to 29 immunoglobulin superfamily members and 26 members of the tumor necrosis factor receptor superfamily are expressed on T-cell surfaces alone, and there have been preclinical or clinical studies on related immune targets and drugs. Specific immune checkpoints include lymphocyte activating gene 3 (LAG-3), T-cell immunoglobulin mucin 3 (TIM-3), and V region Ig inhibitor (VISTA). Non-specific immune checkpoints include human killer cell immunoglobulin like receptor (KIR), indoleamine 2, 3-dioxidase (IDO), and CD47, these novel immune checkpoint molecules are expected to provide hints for clinical and basic research (Manser et al., 2015; Munn and Mellor, 2016; Burugu et al., 2018; Huang et al., 2020; Logtenberg et al., 2020). VISTA, (B7-H5, PD-1H) is an immunomodulatory receptor that inhibits T-cell response. VISTA is overexpressed on CD11b myeloid cells (e.g., macrophages, monocytes, neutrophils, and dendritic cells) and it is found that in humans and mice at a lower level in primitive CD4+ and CD8+ T-cells as well as Tregs. With two potential protein kinase C binding sites and proline residues acting as docking sites in its cytoplasmic tail domain, VISTA can serve as both a receptor and a ligand (Huang et al., 2020; Mutsaers et al., 2021). OX40 (TNFRSF4) has been found to be expressed in activated NK cells, T-cells, NKT cells, as well as neutrophils, and acts as an auxiliary costimulatory immune checkpoint (Curti et al., 2013; Aspeslagh et al., 2016; Buchan et al., 2018). Combining immune checkpoint genes and TANlncSig showed combined prognostic effects on patient survival, in line with previous findings that immunomotor interactions between neutrophilic infiltration and expression levels of checkpoint genes affect the outcome of cancer patients and immunotherapy may also be associated with this condition. In combination with earlier findings, it appears that TANlncSig is correlated with immunosuppressive phenotypes and could predict ICI response. Together, these results indicate that TANlncSig can complement and/or add information to existing immune checkpoint genetic markers. Due to few gene mutations, lung squamous cell carcinoma is less selective than adenocarcinoma with regards to treatment options, and its survival time (about 1 year) is shorter than that of adenocarcinoma (Travis et al., 2021). Thus, novel, effective advanced lung squamous cell carcinoma treatments are needed to improve patient outcomes. The emergence of immune checkpoint inhibitors in recent years has markedly improved treatment options for advanced lung squamous cell carcinoma patients. Immune checkpoint inhibitors have substantially changed advanced lung squamous cell carcinoma treatment, leading to a shift from retro line immunotherapy to front-line treatment options. Originally approved as second-line treatment after platinum-based dual therapy, palivizumab is now recommended as a single-agent first-line treatment or in combination with chemotherapy. Although treatments targeting the immune checkpoints PD-1 and CTLA4 are successful in many cancers, not all patients benefit from them. Our findings indicate that the combination of CEACAM1, TNFSF4, GEM, CD47, VTCN1, and TANlncSig in squamous cell carcinoma can effectively stratify patients by prognosis, highlighting these immune checkpoint receptors as potential therapeutic targets against advanced lung cancer. In conclusion, we used a machine learning-based computational framework to identify lncRNA features of TANs (TANlncSig) via comprehensive analyses of lncRNA, immune, as well as clinical features. TANlncSig revealed a substantial and repeatable correlation with outcomes, even after adjustments of clinical covariates. Analysis of correlation between prognostic lncRNAs and risk score with the expression of immune checkpoint molecules demonstrated that TANlncSig can predict immunotherapy. The study is the first to define lncRNA characteristics of tumor-associated neutrophils, highlighting the importance of lncRNAs in immune responses and the potential for more precise and personalized treatment cancer immunotherapy.

PMC9649924

Ziling Wang,Xiaoying Hou,Min Li,Rongsheng Ji,Zhouyuan Li,Yuqiao Wang,Yujie Guo,Dahui Liu,Bisheng Huang,Hongzhi Du

Active fractions of golden-flowered tea (Camellia nitidissima Chi) inhibit epidermal growth factor receptor mutated non-small cell lung cancer via multiple pathways and targets in vitro and in vivo

28-10-2022

golden-flowered tea,Camellia nitidissima Chi (CNC),non-small cell lung cancer (NSCLC),natural product,epidermal growth factor receptor (EGFR)

As a medicine-food homology (MFH) plant, golden-flowered tea (Camellia nitidissima Chi, CNC) has many different pharmacologic activities and is known as “the queen of the tea family” and “the Panda of the Plant world”. Several studies have revealed the pharmacologic effects of CNC crude extract, including anti-tumor, anti-oxidative and hepatoprotective activity. However, there are few studies on the anti-tumor active fractions and components of CNC, yet the underlying mechanism has not been investigated. Thus, we sought to verify the anti-non-small cell lung cancer (NSCLC) effects of four active fractions of CNC. Firstly, we determined the pharmacodynamic material basis of the four active fractions of CNC (Camellia. leave. saponins, Camellia. leave. polyphenols, Camellia. flower. saponins, Camellia. flower. polyphenols) by UPLC-Q-TOF-MS/MS and confirmed the differences in their specific compound contents. Then, MTT, colony formation assay and EdU incorporation assay confirmed that all fractions of CNC exhibit significant inhibitory on NSCLC, especially the Camellia. leave. saponins (CLS) fraction on EGFR mutated NSCLC cell lines. Moreover, transcriptome analysis revealed that the inhibition of NSCLC cell growth by CLS may be via three pathways, including “Cytokine-cytokine receptor interaction,” “PI3K-Akt signaling pathway” and “MAPK signaling pathway.” Subsequently, quantitative real-time PCR (RT-qPCR) and Western blot (WB) revealed TGFB2, INHBB, PIK3R3, ITGB8, TrkB and CACNA1D as the critical targets for the anti-tumor effects of CLS in vitro. Finally, the xenograft models confirmed that CLS treatment effectively suppressed tumor growth, and the key targets were also verified in vivo. These observations suggest that golden-flowered tea could be developed as a functional tea drink with anti-cancer ability, providing an essential molecular mechanism foundation for MFH medicine treating NSCLC.

Active fractions of golden-flowered tea (Camellia nitidissima Chi) inhibit epidermal growth factor receptor mutated non-small cell lung cancer via multiple pathways and targets in vitro and in vivo As a medicine-food homology (MFH) plant, golden-flowered tea (Camellia nitidissima Chi, CNC) has many different pharmacologic activities and is known as “the queen of the tea family” and “the Panda of the Plant world”. Several studies have revealed the pharmacologic effects of CNC crude extract, including anti-tumor, anti-oxidative and hepatoprotective activity. However, there are few studies on the anti-tumor active fractions and components of CNC, yet the underlying mechanism has not been investigated. Thus, we sought to verify the anti-non-small cell lung cancer (NSCLC) effects of four active fractions of CNC. Firstly, we determined the pharmacodynamic material basis of the four active fractions of CNC (Camellia. leave. saponins, Camellia. leave. polyphenols, Camellia. flower. saponins, Camellia. flower. polyphenols) by UPLC-Q-TOF-MS/MS and confirmed the differences in their specific compound contents. Then, MTT, colony formation assay and EdU incorporation assay confirmed that all fractions of CNC exhibit significant inhibitory on NSCLC, especially the Camellia. leave. saponins (CLS) fraction on EGFR mutated NSCLC cell lines. Moreover, transcriptome analysis revealed that the inhibition of NSCLC cell growth by CLS may be via three pathways, including “Cytokine-cytokine receptor interaction,” “PI3K-Akt signaling pathway” and “MAPK signaling pathway.” Subsequently, quantitative real-time PCR (RT-qPCR) and Western blot (WB) revealed TGFB2, INHBB, PIK3R3, ITGB8, TrkB and CACNA1D as the critical targets for the anti-tumor effects of CLS in vitro. Finally, the xenograft models confirmed that CLS treatment effectively suppressed tumor growth, and the key targets were also verified in vivo. These observations suggest that golden-flowered tea could be developed as a functional tea drink with anti-cancer ability, providing an essential molecular mechanism foundation for MFH medicine treating NSCLC. Golden-flowered tea (Camellia nitidissima Chi—CNC) as an edible and medicinal plant (EMP) is an evergreen shrub belonging to the family Camellia (1). Golden-flowered tea has been known as “the panda of the plant world” and “the queen of the tea family” for its great ornamental and medicinal value. According to the “Guangxi Zhuang Autonomous Region Zhuang Medicine Quality Standard” (2), CNC has been used to treat various diseases such as pharyngitis, dysentery, liver cirrhosis and cancer for a long time. Most recently, CNC has been introduced and cultivated in Australia, Japan, the United States, and other countries (3). Moreover, a plethora of researchers are interested in the anti-cancer effects of CNC as a functional food. Previous studies of CNC pharmacological effects had emphasized anti-tumor, anti-obesity and hypolipidemic effects (4, 5). In the last two decades, much of the research about CNC has explored the pharmacological effects of flower fractions, while the studies of leaf fractions are extremely rare. In fact, the leaves of CNC have been used as tea for a long time (6). Although, several studies have revealed the anti-cancer effects of CNC leaves crude extract, there are few reports on the anti-tumor active fractions and components of CNC (7). Thus, it is necessary to explore the differences in the pharmacological effects of different active fractions of CNC. Previously, our research for the first time confirmed that the four active fractions of CNC (Camellia. leave. saponins, Camellia. leave. polyphenols, Camellia. flower. saponins, Camellia. flower. polyphenols) effectively inhibited the proliferation, metastasis and invasion of anti-NSCLC in vitro (8), while the anti-cancer mechanism remains to be revealed. Lung cancer is the leading type of cancer death worldwide, with NSCLC being the most common sub-type (9–12), accounting for approximately 85% (13). Among the emerging oncology therapies, molecular targeted drugs have become the first choice for treating NSCLC (14). Approximately 10–40% of NSCLC patients worldwide have tumor cells carrying epidermal growth factor receptor (EGFR) activating mutations (15). The epidermal growth factor receptor-tyrosine kinase inhibits (EGFR-TKI) targeted therapy is a milestone in tumor treatment with remarkable effects (16). However, NSCLC frequently develops acquired resistance when treated with NSCLC owing to factors such as tumor mutational burden (17), immune evasion and tumor microenvironment (TME) (18, 19). Therefore, the search for new therapeutic agents for drug-resistant NSCLC and the analysis of medicinal treatment mechanisms are frontier issues in oncology science, which have scientific value and clinical guidance significance for the treatment of NSCLC. Thus, we attempt to explore the molecular mechanism to provide more scientific evidence for the application of golden-flowered tea in the treatment of NSCLC. In brief, the component difference between the four fractions of CNC was first reported in this study. Then, we evaluated the anti-tumor activity of four fractions of CNC on three different NSCLC cell lines. To determine the programmed cell death effect on non-small cell lung cancer cells, we investigated whether CLS treatment induces the apoptosis of NCI-H1975 cells by TdT-mediated dUTP Nick-End Labeling (TUNEL) assay, Annexin V and propidine iodide (PI) staining, reactive oxygen species (ROS) measurement, superoxide dismutase (SOD) measurement, SEM examination and lactate dehydrogenase (LDH) release. Transcriptomics analysis was employed to probe the genetic changes after treatment of NSCLC cells with CLS. Subsequently, RT-qPCR and WB confirmation were performed for the candidate pathways. Finally, Xenograft models assay also proved the inhibitory effect of CLS in vivo. Taken together, our study investigated the inhibitory effect of different fractions of CNC on NSCLC (Figure 1). Importantly, our work will facilitate the study of the anti-tumor effect and mechanism of CNC as a functional tea. The leaves and flowers of CNC were collected from Fangchenggang, Guangxi Province, China. The extraction method was based on the previous research of our team (7, 20). The leaves and flowers were air-dried and grinded into powder. The powder of leaves (6.3 kg) and flowers (6.0 kg) were separately refluxed with 95% ethanol for 3 times (3, 2 and 1 h). The extracts were combined and evaporated in a rotary evaporator to obtain ethanol extracts. Finally, four different active fractions of CNC (Camellia. leave. polyphenols, Camellia. flower. polyphenols, Camellia. leave. saponins, Camellia. flower. saponins) were obtained by macroporous resin purification process (21, 22). Total polyphenols were determined by Follin-Ciocalteu (FC) assay. The FC reagent (diluted 1:10 in water) and aqueous Na2CO3 (10%) were added to the two fractions of CNC (Camellia. leave. polyphenols, Camellia. flower. polyphenols) in sequence. Gallic acid control solution was prepared to draw the calibration curve. The absorbance value was measured at 765 nm after constant shaking at 37°C for 30 min. Total saponins were determined by Vanillin-acetate method. Firstly, 5.0 g of vanillin was weighed to configure a 5% solution of vanillin acetate. Ginsenoside Re control solution was prepared to draw the calibration curve. Prepared 1 mL of 1 mg/mL of the solution (Camellia. leave. saponins, Camellia. flower. saponins) to be measured in the test tube in a water bath to evaporate. The 5% vanillin-acetate solution and Perchloric acid were added to the two fractions of CNC in sequence. After heating the test tubes at 60 degrees for 10 min, the test tubes were cooled with ice water and 5 mL of glacial acetic acid was added. The absorbance value was measured at 560 nm. The four active fractions of CNC were identified by the UPLC-Q-TOF-MS/MS. After being dissolved in methanol, the sample was filtered through a 0.22 μm microfiltration membrane for analysis. The UPLC-Q-TOF-MS/MS has equipped with an Agilent SB-C18 (1.8 μm, 2.1 mm × 100 mm) column. The mobile phase is composed of solvent A, 0.1% formic acid in water and solvent B, 0.1% formic acid of acetonitrile. The elation gradient procedure was performed: 0–9 min, 5–95% B; 9–10 min 95% B; 10–11.10 min 95–5% B; 11.10–14 min 5% B. The flow rate was 0.35 mL/min and the sample injection volume was 4 μL. The effluent was alternatively connected to an ESI-triple quadrupole-linear ion trap (QTRAP)-MS. The ESI source operation parameters were as follows: an ion source, turbo spray; source temperature 550°C; ion spray voltage (IS) 5,500 V (positive ion mode)/-4,500 V (negative ion mode); ion source gas I (GSI), gas II (GSII), curtain gas (CUR) was set at 50, 60, and 25.0 psi, respectively; the collision-activated dissociation (CAD) was high (23). NCI-H1975 cells, A549 cells and HCC827 cells were grown in RPMI Medium 1,640 basic (1×) supplemented with 10% fetal bovine serum(GEMINI BIO-Products)in a humidified chamber with 5% CO2 and 37°C. The cell culture method is the same as the previous culture method of our team (24). NCI-H1975 cells, A549 cells and HCC827 cells were seeded at 3,000, 5,000, and 3,000 cells per well of 96-well plates in triplicate. Cell viability was measured at 72 h by using a 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyl tetrazolium bromide (MTT) assay. NCI-H1975 and HCC827 cells were seeded at 500 cells per well of 6-well plates in a medium containing Penicillin-Streptomycin-Gentamicin Solution (Solarbio, P1410). After 24 h of incubation, the cells were treated with different concentrations of CNC every 5 days until colonies formed in 10 days. The remaining colonies were stained with crystal violet. After NCI-H1975 and HCC827 cells were treated by different fractions of CNC in 12-well plates for 24 h, the cells were cultured with 10 μM EdU (KevGEN BioTECH, KGA331-500) for 2 h, followed by incubation with 4% polychloraldehyde for 15 min. Washed by 3% BSA in PBS twice, the cells were incubated with 0.5% Triton X-100 (Solarbio, 9002-39-1) in PBS for 20 min. The cell plates were washed twice with 3% BSA in PBS and incubated with a 1 × Click-iT reactant mixture for 30 min. The cells treated with 1 × Click-iT reactant mixture were incubated with 1 × Hochest 33342 for 30 min under dark conditions. The proliferating cells (green) and the nuclei of all cells were observed under a laser confocal microscope under dark conditions. Different visual fields were randomly taken for image collection and synthesis analysis. Finally, the proliferation rate was calculated. After NCI-H1975 were cultured in 12-well plates for 24 h, the cells were treated with different concentration of CLS for 24 h. Cells were subsequently incubation with 4% polychloraldehyde for 30 min. Washed by PBS twice, the cells were incubated with 0.3% Triton X-100 (Solarbio, 9002-39-1) in PBS for 10 min. The cell plates were washed twice with PBS and incubated with a TUNEL reactant mixture for 60 min at 37°C (Beyotime, C1086). Different visual fields were randomly taken for image collection and synthesis analysis. And the TUNEL positive rate was calculated and normalized to that of the control group. The apoptosis rate of NCI-H1975 cells using Annexinv-fluorescein isothiocyanate (FITC) and PI double staining technique (KevGEN BioTECH, KGA107). NCI-H1975 cells were processed at different concentrations of CLS for 48 h. The cells were collected by digestion with EDTA-free trypsin and washed twice with PBS. After processing according to the steps in the instructions, all groups were measured by flow cytometer. The effect of different concentrations on CLS-mediated ROS production in NCI-H1975 cells was determined using the cell-permeable fluorescent probe 2’,7’-dihydrofluorescein-diacetate (DCFH2-DA). NCI-H1975 cells were incubated in different concentration of CLS for 24 h, the cells were cultured with 1 μM DCFH2-DA (Solarbio, D6470) for 30 min at 37°C. And the ROS positive rate was calculated and normalized to that of the control group. The effect of different concentrations on CLS-mediated SOD production in NCI-H1975 cells was determined using the SOD reagent kit (Njjcbio, A001-3-2). After being processed at different concentrations of CLS for 48 h, the proteins of NCI-H1975 cells were extracted. The relative content of SOD was determined by the reagent kit. The SOD positive rate was calculated and normalized to the control group. After NCI-H1975 cells were treated in accordance with the above-described experimental design, SEM was used to observe the difference between the treated and control groups. After cell crawling was washed with PBS, electron microscope fixative (Servicebio, G1102) was added and placed in a four-degree refrigerator for 1 h. Ethanol gradients were used to remove water from the samples, with dehydrating agent concentrations of 30, 50, 70, 80, 90, and 100% (twice) in order, with each dehydration time of 5 min. Finally, the samples were dried in the desiccator for 1.5 h and then sprayed with gold and photographed. The release of IL-1β and LDH can be detected during the onset of pyroptosis (25). After NCI-H1975 cells were treated in accordance with the above-described experimental design, LDH release was measured by LDH assay kit (Njjcbio, A020-2) to observe the difference between the treated and control groups. The absorbance was measured at a wavelength 450 nm using microplate reader. RNA degradation and contamination were monitored on 1% agarose gels. The preparation of each RNA sample requires 3 μg of RNA as input material (26). Sequencing libraries were generated using the NEBNext® UltraTM ® RNA Library Preparation Kit (NEB, USA), and index codes were added to the attribute sequences of each sample. After the library inspection is qualified, the different libraries are pooled according to the requirements of effective concentration and target data volume. And illumine sequencing is performed, and the generated 150 bp paired-end reads. Differential expression analysis was performed for two conditions/groups (two biological replicates per condition) using the DESeq2 R package (1.16.1). Genes identified by DESeq2 with adjusted p-values < 0.05 were designated as differentially expressed genes. Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of differentially expressed genes was implemented by the cluster profile R package (27, 28). Total RNA was extracted from cell culture samples using the TRNzol Universal Reagent (Tiangen, W9712) according to the manufacturer’s instructions. The cDNA was synthesized from total RNA (1 μg) using reverse transcription (Vazyme, R323-01). Primer sequences were as follows in Table 1. PCR amplification was executed by the SYBR Green PCR master mix (LightCycler 480, 30408), and the PCR-amplified gene products were analyzed. After the cells were treated with CLS for the indicated time, cell lysates were lysed by RIPA buffer supplemented with a complete protease and phosphatase inhibitor mixture (Beyotime, 45482). Samples of mouse tumor tissues were stored in a −80°C refrigerator and homogenized with RIPA (Solarbio, 676). Proteins were separated on a 6–10% SDS-PAGE system and transferred to a polyvinylidene fluoride (PVDF) membrane. WB was performed according standard protocol with following primary antibodies: Anti-ACTB (Abclonal, AC004; 1:10,000), Anti-INHBB (Abclonal, A8553;1:1,000), Anti-TGFB2 (Abclonal, A3640; 1:1,000), Anti-ITGB8 (Abclonal, A8433; 1:1,000), Anti-PIK3R3 (Abclonal, A17112; 1:1,000), Anti-TrkB (Abclonal, A2099; 1:1,000), Anti-CACNA1D (Abclonal, A16785; 1:1,000), HPR Goat Anti-Mouse (Abclonal, AS003; 1:2,000) or HPR Goat Anti-Rabbit (Abclonal, AS014; 1:2,000) secondary antibodies were used. Protein bands were visualized by chemiluminescence reagents (Meilunbio, MA0186-1) and were performed using a luminescent image analyzer (Proteinsimple, 601577). Raw data were analyzed by using Fuji film v3.0. Thirty 5-week-old BALB/c-nude mice were obtained from the Biont (Jiangsu, No.320727210100432325). All mice were housed in a temperature-controlled environment (24 ± 2°C) with a 12/12 h dark/light cycle at the Animal Center of Hubei university of Chinese medicine. The standard rat chow and water used for animal feeding and all animal experiments were conducted by the animal ethics-related regulations of Hubei University of Traditional Chinese Medicine, permission number: SYXK2017-0067-ZYZYZX2022-2. BALB/c-nude mice were injected subcutaneously in the armpit with NCI-H1975 (5 × 106 cells/mice) in 150 μL PBS. After the mean tumor volume reached 50 mm3, BALB/c-nude mice were randomly divided into model control group (n = 6), tax group (n = 6), low-dose group (n = 6), medium-dose group (n = 6), high-dose group (n = 6). Low-dose orally took 100 mg/kg CLS every day, medium-dose orally took 200 mg/kg CLS every day, and high-dose orally took 400 mg/kg CLS every day. And taxol (anhydrous ethanol: castor oil = 1:1) was injected at 20 mg/kg every 2 days in the tail vein. Tumor volume was monitored by vernier calipers throughout the experiment. All mice were executed and tumors were removed on day 13. , where V0 represents the tumor volume of day 1 (the day of CLS first administration), Vt represents the tumor volume of day 13 (29). Samples from the tumor xenografts and liver were dissected, formalin-fixed and paraffin-embedded. Paraffin blocks were placed on the pre-cooling table and adjusted the knees to 4-M m thickness. Sections were incubated with citric acid (pH 6.0) antigen retrieval buffer (Beyotime, P0085) for antigen retrieval in a microwave oven. After blocking endogenous peroxidase with 3% hydrogen peroxide, and serum sealing by 3% BSA (Beyotime, P0007), sections were then incubated by Ki67 antibody (Abcam, Ab16667) and further processed with secondary antibody (Abcam, Ab6721). The chromogenic reaction was performed by DAB (Solarbio, DA1010). Sections were counterstained with hematoxylin (Solarbio, H8070) and observed under a microscope. The nucleus of hematoxylin stained is blue, and the positive expression of DAB is brownish yellow. All graphs were generated using GraphPad Prism 8.0 (SanDiego, CA, USA). One-way ANOVA with Bonferroni correction was used for statistical analyses. Statistical significance was set at *p < 0.05, and **p < 0.01 compared to control unless stated differently. As is known, total polyphenols and total saponins are important active ingredients in tea beverages. Therefore, we first determined the contents of total polyphenols and total saponins in each of the four active fractions of CNC. The polyphenols contents of Camellia. leave. polyphenols (CLP) and Camellia. flower. Polyphenols (CFP) in terms of gallic acid equivalent (standard curve equation: y = 4.2285x+0.0597, r2 = 0.999) were from 20 to 100 μg/mL and listed in Table 2. The polyphenols contents in CLP were 136.89 ± 3.18 mg/g and the phenolic contents in CFP were 327.03 ± 4.03 mg/g. Table 2 also showed the content of total saponins reported as Ginsenoside Re equivalent (standard curve equation: y = 1.4807x+0.0223, r2 = 0.99), which were from 0.02 to 0.14 mg/mL. Saponin contents were 38.83 ± 0.57 mg/g in CLS and 56.53 ± 0.83 mg/g in CFS as shown in Table 2. The results indicated that total saponins and total polyphenols might be important active components in goldenrod tea. To confirm the material basis of CNC, the active chemical components of the four active fractions of CNC were detected separately by UPLC-QTOF-MS/MS. As is shown in Figure 2A and Table 3, the three components with the highest content in CLS are isoschaftoside, hyperin and vicenin-2. In CLP, the three main effective compounds are 6-O-Feruloyl-β-D-glucose, epicatechin glucoside and isoschaftoside (Figure 2B and Table 3). However, astragalin, isoschaftoside and brevifolin carboxylic acid are the most abundant substances in CFS (Figure 2C and Table 4). And in CFP, 6-O-Galloyl-β-D-glucose, 3-O-Galloyl-D-glucose and isoschaftoside demonstrated extremely high content (Figure 2D and Table 4). In conclusion, we initially revealed the specific chemical composition of different fractions of CNC, which laid the foundation for the subsequent activity study. To determine the anti-cancer effect on different non-small cell lung cancer cells, we firstly investigated whether CNC treatment inhibits the proliferation of NSCLC by MTT assay. Treating with CNC significantly suppressed the proliferation of NCI-H1975, A549 and HCC827 cells (Figures 3A,B). After 72 h of treatment, the results confirmed that CLS, CLP, CFS and CFP could significantly inhibit the proliferation of NCI-H1975, A549, and HCC827 cells. It was worth noting that the active fractions of CNC exhibited high inhibitory effect on 3 NSCLC cell lines, especially on EGFR mutant cells NCI-H1975. Combining the information from the previous MTT assay, we selected NCI-H1975 and HCC827 cells as the main research object. Thus, we performed a colony formation assay by giving CNC every 5 days for 10 days into NCI-H1975 and HCC827 cells. The results demonstrated that CNC treatment significantly restrained anchorage-dependent colony formation of NCI-H1975 and HCC827 cells (Figure 3C and Supplementary Figure 1). At low doses, the NCI-H1975 cells eventually all died as well, demonstrating the remarkable anti-tumor activity of CNC. Furthermore, the EdU assay is one of the most accurate and direct methods for detecting cell proliferation. Observed by laser confocal microscope, the proportion of proliferating NCI-H1975 and HCC827 cells (green) was significantly lower than the control group after 24 h of different concentrations of CNC treatment (Figure 4 and Supplementary Figure 1). Expectedly, CNC treatment led to the significantly decreasing proliferation of NCI-H1975 cells. Simultaneously, we found that CLS had a higher proliferating inhibitory effect on NCI-H1975 cells. These collective data indicated that CNC inhibited the proliferation of NSCLC, supporting that CNC is a new anti-cancer EMP with promising research prospects. To determine the programmed cell death effect on non-small cell lung cancer cells, we investigated whether CLS treatment induces the apoptosis of NCI-H1975 cells by TUNEL assay. Treating with CLS significantly induced the apoptosis of NCI-H1975 cells (Figure 5A). Subsequently, we found that CLS inhibited ROS production, suggesting that NCI-H1975 may not induce programmed cell death through ferroptosis (Figure 5B). These results might originate from the antioxidant effect of CNC related (6). Annexin V-FITC/PI assay results showed that CLS treatment significantly unregulated the appearance of labeled cells in Q3 (from 1.98 ± 0.51 to 5.27 ± 2.87) suggesting that there was an increased early apoptosis in NCI-H1975 cells (Figure 5C). Moreover, scanning electron microscope (SEM) showed that the cell in the control group were normal and cell membranes were intact. By contrast, the cell in CLS group showed the damaged cell membranes and evidently increased number of scorched corpuscle (Figure 5D). To further confirm whether the cells underwent pyroptosis, we examined the levels of LDH in the cell supernatant and the relative mRNA expression of IL-1β. These results showed LDH content and IL-1β expression increased with increasing concentrations of CLS administration, which gave the best agreement with CLS induced programmed cell death through pyroptosis (Figure 5E). The above studies confirmed the anti-tumor activity of CLS, yet the mechanism of CLS treatment is unknown. Therefore, we applied transcriptome analysis to initially study the mechanism of CLS treatment. The global gene expression changes induced by CLS treatment were determined by comparing the gene profiled NCI-H1975 cells based on microarray data. We found that 1,008 were significantly down-regulated after CLS treatment, while 1,077 genes were significantly up-regulated (Figure 6A), suggesting the global transcriptome changes after CLS treatment. The pathway-enrichment analysis was annotated based on different databases (GO and KEGG) for homologous alignment to classify the function of differentially expressed genes (DEGs) between control and CLS-treated groups. String-based GO pathway analysis revealed several enriched pathways, including leukocyte migration, chemotaxis, taxis, proteinaceous extracellular matrix, and cytokine activity (Figure 6B). Moreover, “cytokine activity” was the most abundant term for DEGs in the metabolic process which indicated cytokine-mediated signaling pathway could be in-depth investigate (30). Based on the KEGG, we attempted to perform a standard pathway enrichment analysis to identify the major active pathways for the inhibitory effect of CLS on NCI-H1975 cells. According to the pathway-enrichment analyses of these DEGs (Q-value<0.05), the most significantly enriched pathways are “Cytokine-cytokine receptor interaction” and “PI3K-Akt signaling pathway” (Figure 6C). Specifically, 22 DEGs were relevant to “Cytokine-cytokine receptor interaction,” and 23 DEGs of the PI3K-Akt signaling pathway were involved in the anti-tumor process (Figures 6D,E). Transform growth factors (TGF), as an important class of cytokines, have been identified as mediators of a large number of diseases and can regulate the TME (31). TGF can also activate TGFB receptors on the MAPK signaling pathway, thereby affecting the MAPK signaling pathway. Besides, a large number of papers confirmed the existence of multi-level crosstalk between Ras/MAPK and PI3K/Akt signaling pathways (Figure 6H) (32, 33). Moreover, CLS treatment regulated 16 DEGs in the “MAPK signaling pathway” while suppressing NCI-H1975 cells growth (Figure 6F). These results indicated that CLS inhibited NSCLC cells growth via multiple targets and pathways, especially by inhibiting signal transduction of “Cytokine-cytokine receptor interaction,” “PI3K-Akt signaling pathway” and “MAPK signaling pathway,” while the involvement of the above pathways needs further experimental verifications (Figures 6H,I). To verify the results of transcriptome analysis, we used RT-qPCR to validate the key genes of these three pathways separately. As shown in Figure 7A, the results showed that CLS could concentration-dependently inhibit the mRNA levels of TGFB2, INHBB (Cytokine-cytokine receptor interaction), PIK3R3, ITGB8 (PI3K-Akt signaling pathway), NTRK and CACNA1D (MAPK signaling pathway) were relatively significant and concentration-dependent (p < 0.05) compared to the control group. The transcriptomic data and RT-qPCR validation unveiled that TGFB2, INHBB, PIK3R3, ITGB8, NTRK (TrkB) and CACNA1D might be a critical targeted gene for NSCLC inhibition by CLS. To further verify the changes in NSCLC protein expression after CLS treatment, we measured the expression of proteins encoded by crucial genes in NSCLC using the Western Blot assay (Figure 7B). Compared to the control group, the expression of these six key genes was observably decreased. These results were consistent with the result of RT-qPCR, indicating that CLS inhibited Cytokine-cytokine receptor interaction, PIK-Akt and MAPK signaling pathways. To further explore the effect of CLS treatment on tumor development in vivo, tumor-bearing mice were treated with different CLS concentrations. NCI-H1975 cells were inoculated subcutaneously into an underarm flank of athymic mice. Intriguingly, the body weight of low-dose, medium-dose and high-dose treated mice were slightly lower than the untreated animals (Figure 8B), but all vital signs such as activity status were normal. In addition, H&E results and physiological and biochemical results of the liver and kidney showed no significant hepatic or renal toxicity with CLS treatment (Supplementary Figure 2). This result might be based on the ability of CNC to inhibit lipase activity (34, 35). Compared with untreated mice in the control group, the CLS 100 mg/kg showed a tumor inhibition effect since day 9 (Figure 8C) while CLS 200 mg/kg and 400 mg/kg treated group significantly suppressed tumor growth starting at day 5 (Figures 8C,D). In addition, the tumor weights of the mice were significantly different in all the administered groups after execution compared to the control group (Figure 8E). Suppression of Ki67 by CLS was also confirmed in vivo by IHC analysis in CLS-treated BALB/c-nude mice tumors (Figures 8F,G). Based on the in vitro transcriptome analysis and validation of “cytokine-cytokine receptor interaction,” “PI3K-Akt signaling pathway” and “MAPK signaling pathway,” we measured the relative protein expression of TGFB2, INHBB, PIK3R3, ITGB8, NTRK, and CACNA1D in tumor tissue. Interestingly, we found that the paramount targets previously validated by cell samples in Western blot experiments were also corroborated in tumor samples (Figures 8H,I). Taken together, these results proved that CLS treatment could effectively inhibit the growth of NCI-H1975 tumor xenografts in a dose-dependent manner through cytokine-cytokine receptor interaction, PIK-Akt, MAPK signaling pathways. Although EGFR-TKI have become a first-line inhibitor of EGFR mutation-positive NSCLC, nearly half of NSCLC patients are resistant to EGFR-TKI-based chemotherapies. Thus, it is an urgent need for development of drugs that could inhibit NSCLC with EGFR mutations. In this study, we identified inhibitory effects of different active fractions of CNC on NSCLC cell lines. Four fractions of CNC demonstrated remarkable anti-NSCLC effect. Intriguingly, upon treatment with CLS on NCI-H1975 cells, CLS suppressed the cytokine-cytokine receptor interaction, PIK-Akt and MAPK signaling pathways, leading to growth inhibition of the tumor in vitro and in vivo. Briefly, this study suggested that CNC, as a functional food, could provide a more efficient treatment in EGFR mutated NSCLC. As a plant with both medicinal and ornamental value, the pharmacological effects of its flower fractions have been thoroughly studied (34, 36, 37), while the research on leave fractions was rare. Furthermore, few studies have systematically investigated the differences in the chemical composition of the four active fractions of CNC. In this study, we firstly identified the three main components of CLS are isoschaftoside, vicenin-2 and hyperin (Table 3). And recently reported, Vicenin-2 and Hyperin were identified as two novel nature medicine against NSCLC, indicating that CLS possess anti-NSCLC properties and play a crucial role in patients’ defense against tumor (38, 39). However, the pharmacology effects of isoschaftoside have been rarely reported. Hence, further investigation is needed to evaluate the effects of isoschaftoside on NSCLC cells. Besides, the chemical composition of the four active fractions of CNC was found to be various by UPLC-QTOF-MS/MS (Table 4). In the two extracted fractions of the leaves of CNC, the content of geniposide (CLS 1.89 %, CLP 1.68 %) was high, while the content of the two extracted fractions of the flowers was almost absent. Extensive experiments and analysis demonstrate that geniposide possesses relatively strong anti-tumor activity (40, 41) and pulmonary protective effect (42). This suggests that the variation in geniposide content could contribute to the difference in anti-tumor activity between CNC leaf fractions and flower fractions. It is reasonable to make assumptions that the anti-tumor activity of different fractions of CNC also differed based on the composition differences. Nowadays, few studies have investigated the effect of different active fractions of CNC on NSCLC. In this study, we examined the inhibitory of four active CNC fractions on three cell lines of NSCLC (NCI-H1975, HCC827, A549). These results showed that the four active fractions of CNC possessed remarkable inhibitory on NSCLC cell lines, especially on EGFR-T790-mutated NCI-H1975 cells (Figure 3A). To further confirm the inhibitory effect on EGFR-T790-mutated NCI-H1975 cells, colony formation assay and EdU incorporation assay were performed (Figures 3B, 4). Combining the above results, we concluded that the active fractions of CNC could effectively inhibit the proliferation of NSCLC, among which CLS had the better inhibitory effect on NCI-H1975 (Supplementary Figure 1). In addition, we found that CLS could induced programmed NSCLC death through pyroptosis (Figure 5). Altered levels of NSCLC-related genes have been inspected by transcriptome analysis after CLS treatment. Interestingly, CLS appears to suppress tumor cell growth via “Cytokine-cytokine receptor interaction,” “PI3K-Akt signaling pathway” and “MAPK signaling pathway”(Figure 6). According to transcriptomic results, the expression of TNFRSF10C, INHBB, TGFB1, TGFB2, and TGFB3 were down-regulated after CLS stimulation (Figure 6D). Studies of cytokines suggested that chemokines as a cytokine can promote anti-tumor immunity to NSCLC (43). At present, anti-cytokine antibodies and cytokine blockers have been extensively studied in tumor therapy (44). Transform growth factors (TGF), as an important class of cytokines, have been identified as mediators of a large number of diseases and can regulate the TME (31). Consequently, we choose TGF-β, Activin and TRAIL as the key cytokines and cytokine receptors. In this research, CLS treatment might promote T cell differentiation and tumor immune response by inhibiting the expression of three TGF phenotypes (TGF-β1, TGF-β2, and TGF-β3), thereby inhibiting tumor angiogenesis and invasion. At the same time, by inhibiting TNFRSF10C competitively binding with tumor necrosis factor-associated apoptosis-inducing ligand (TRAIL), TRAIL-induced tumor cell apoptosis can be promoted, thus possibly inhibiting the proliferation and metastasis of tumor cells (Figures 6D,H). Alternatively, tumor cells promote the expression of cytokines to escape the immune response. Overall, the inhibitory of CLS on tumor growth might be exerted by affecting the interaction of cytokines-cytokines receptors. Among them, TGF-β and TNFRSF10C might be the two most critical targets for suppressing NSCLC growth. Therefore, remodeling the immune microenvironment of NSCLC through inhibiting Cytokine-cytokine receptor interaction provides new perspectives for the treatment of NSCLC. Besides, our RT-qPCR and WB results also suggested that the down-regulation of the levels of TGFB2 and INHBB genes might be important targets that suppress the proliferation of NCI-H1975 cells (Figures 7A,B). TGF-β induces EMT in tumor cells through Smad and non-Smad signaling pathways, whereas non-Smad includes signaling pathways such as PI3K, MAPK (45). Based on transcriptomic data, we found that ITGB8 down-regulation cascades its downstream signal PIK3R3 expression to be suppressed (Figures 6E,I). We hypothesized that the up-regulation of PPP2R5B combined with the down-regulation of PIK3R3 resulted in the inhibition of the Akt-mTOR signaling pathway (Figures 6E,I). As a result, eukaryotic initiation factor 4E (elF4E) was down-regulated. Furthermore, numerous studies have shown that ITGB8 and EIL4E proteins are associated with cancer migration, invasion and metastasis and autophagy (46, 47). We speculate that CLS may induce autophagy-mediated cell death by inhibiting the PIK3R3-Akt-mTOR axis through ITGB8 down-regulation. Mitogen-activated protein kinase (MAPK) signaling pathway, as one of the key pathways to induce tumor production, is involved in a series of cell physiological activities such as cell growth, differentiation and apoptosis (48). According to the transcriptomic analysis, CLS inhibited both CACN receptors (CACNA1D, CACNG4, CACNA1I) and RTK receptors (NTRK2, PDGFRB) and cascade RAS was inhibited (Figures 6F,I). The loss of Ca2+ caused by CACN receptors repression also resulted in the down-regulation of RAS and MEF2C expression. It is worth mentioning that intracellular Ca2+ can be considered a major regulator of autophagy. Therefore, we selected six genes related to PIK-Akt and MAPK signaling pathways for RT-qPCR validation (PIK3R3, ITGB8, EIF4E1B, TrkB, CACNA1D, and MEF2C), and the results indicated that PIK3R3, ITGB8, NTRK, and CACNA1D could be used as new targets for NCI-H1975 (Figures 7A,B). This observation supports the hypothesis that CLS could induce cell autophagy and inhibit tumor growth via PI3K and MAPK signaling pathways. However, the mechanism by which PIK3R3, ITGB8, TrkB, and CACNA1D induce cell autophagy and inhibit tumor growth remains to be elucidated. We speculate that suppression of TGFB2, CACNA1D, TrkB, and ITGB8 could result in reduced PI3K expression, which ultimately would inhibit the metastatic and invasive ability of NCI-H1975 (Figure 6I). In this regard, it has been reported that Vicenin-2 (the content in CLS is 2.36%) inhibited the expression of key proteins of PI3K/Akt and TGF-β/Smad signaling pathway in A549 and NCI-H1299 cells, resulting in reduced EMT (39). These observations suggesting TGFB2, INHBB, PIK3R3, ITGB8, NTRK, and CACNA1D as major mediators in CLS-induced NCI-H1975 cell death. Importantly, these results indicated that CNC as a functional food has the advantage of being multi-channel and multi-target against NSCLC. Based on the observed effects of CLS on NSCLC, we also constructed the xenograft models assay in nude mice to verify whether CLS is effective in vivo NSCLC models. In this work, we found that CLS significantly inhibited the growth of transplanted tumors in nude mice in a concentration-dependent manner (Figure 8A). Interestingly, the slight change in body weight of nude mice with increasing drug doses. However, physiological and biochemical results and H&E sections of the liver and kidney confirmed the absence of significant drug toxicity, suggesting CLS may play a role in lipid-lowering (Figure 8B and Supplementary Figure 2) (35). Also, we performed WB validation in tissue samples of the proteins screened in the previous experiments, with results generally consistent with cell samples (Figures 8H,I). Also, we verified the pathological patterns of tumors and the expression level of Ki67 in tumors by H&E and IHC, which indicating tumor proliferation rate was significantly suppressed (Figures 8F,G). Collectively, our findings identified CLS as a new EMP for NSCLC, providing an essential molecular foundation for enhanced understanding of CNC treatment for NSCLC. In this work, we preliminarily elucidated the anti-tumor effect by which the four active fractions of CNC against NSCLC and the anti-tumor mechanism of CLS. However, this investigation has several limitations. As an essential method to evaluate pharmacological effects of Chinese medicine, Serum Pharmacology is an important auxiliary analysis method (49, 50). Despite the compositions of different fractions of CNC having been identified, the compounds in serum after oral administration of CLS in mice still require in-depth research. In addition, the specific compounds that affect these signaling pathways and targets still require further corroboration. Although the details of effective compounds and their mechanism in CNC remain unknown, our findings revealed a basic mechanism for the anti-NSCLC effect of CLS, providing scientific support for the application of CNC as a functional food with anti-cancer activity. The original contributions presented in the study are publicly available. This data can be found here: https://www.ncbi.nlm.nih.gov/bioproject/PRJNA891478/. The animal study was reviewed and approved by Hubei University of Chinese Medicine. ZW: methodology, data curation, formal analysis, study design, investigation, writing, and validation. XH: study design and writing – original draft. ML: writing – original draft. RJ: software, methodology, and supervision. ZL: software, data curation, and supervision. YW: validation, investigation, and methodology. YG: data curation. DL: formal analysis, software, and supervision. BH: resources, supervision, and writing – original draft. HD: resources, supervision, investigation, writing – original draft and editing, and project administration. All authors contributed to the article and approved the submitted version.

PMC9649934

Wajiha Kanwal,Abdul Rehman

High prevalence of vitamin D deficiency in HIV-infected individuals in comparison with the general population across Punjab province, Pakistan

31-10-2022

Hypovitaminosis D,HIV/AIDS,HBV/HCV,ELISA,CD4 count

The status of vitamin D in individuals infected with human immunodeficiency virus (HIV), particularly in naïve as well as treated patients, has never been reported in the Pakistani population. A cross-sectional study was performed to measure vitamin D in individuals infected with HIV living in various districts of the Punjab, Pakistan. 1000 persons attending various treatment centers of the Punjab were screened for HIV, hepatitis B virus (HBV), hepatitis C virus (HCV), and Syphilis. Total 398 patients met inclusion criteria and vitamin D level was measured in respective cases by using enzyme-linked immunosorbent assay (ELISA) technique. 232 samples from the healthy population were also included in present research. Demographic history and clinical parameters regarding HIV disease were evaluated. Comparison of variables was done to find out the link between vitamin D levels and characteristics of HIV infected persons and comparison to that of healthy individuals was performed. Among 398 HIV patients vitamin D deficiency and insufficiency was found among 15 % and 39 % while majority of the control participants had sufficient levels of vitamin D (78 %). Most of the HIV infected individuals were males (68.6 %) and had age between 24 and 47 years (67.8 %). A significant relationship was found for vitamin D level, lifestyle and CD4 count among HIV + ve non acquired immunodeficiency syndrome (AIDS) subjects (95 % CI; p < 0.001, p = 0.09). For HIV + ve AIDS patients vitamin D had a significant relationship with lifestyle along with HIV viral load and CD4 count. Hypovitaminosis D prevails among the HIV infected population of Punjab, Pakistan.

High prevalence of vitamin D deficiency in HIV-infected individuals in comparison with the general population across Punjab province, Pakistan The status of vitamin D in individuals infected with human immunodeficiency virus (HIV), particularly in naïve as well as treated patients, has never been reported in the Pakistani population. A cross-sectional study was performed to measure vitamin D in individuals infected with HIV living in various districts of the Punjab, Pakistan. 1000 persons attending various treatment centers of the Punjab were screened for HIV, hepatitis B virus (HBV), hepatitis C virus (HCV), and Syphilis. Total 398 patients met inclusion criteria and vitamin D level was measured in respective cases by using enzyme-linked immunosorbent assay (ELISA) technique. 232 samples from the healthy population were also included in present research. Demographic history and clinical parameters regarding HIV disease were evaluated. Comparison of variables was done to find out the link between vitamin D levels and characteristics of HIV infected persons and comparison to that of healthy individuals was performed. Among 398 HIV patients vitamin D deficiency and insufficiency was found among 15 % and 39 % while majority of the control participants had sufficient levels of vitamin D (78 %). Most of the HIV infected individuals were males (68.6 %) and had age between 24 and 47 years (67.8 %). A significant relationship was found for vitamin D level, lifestyle and CD4 count among HIV + ve non acquired immunodeficiency syndrome (AIDS) subjects (95 % CI; p < 0.001, p = 0.09). For HIV + ve AIDS patients vitamin D had a significant relationship with lifestyle along with HIV viral load and CD4 count. Hypovitaminosis D prevails among the HIV infected population of Punjab, Pakistan. Vitamin D, a steroid hormone, can be synthesized endogenously or exogenously via food or dietary supplements (Ali, 2020) and is a standard hormone characterized for its pivotal role in skeletal and mineral homeostasis. After exposure to UV rays from the sun, 7-dihydrocholesterol in the skin is converted to preVitD3. PreVitD3 is then transformed into cholecalciferol via spontaneous isomerization (VitD3) and is carried to the kidney and gets hydroxylated to form 1, 25 dihydroxycholecalciferol (1,25(OH)2D) or calcitriol and locally activated by CYP27B1 (Kim et al., 2020) in numerous tissues including immune cells, brain, breast, prostate and smooth muscles (Dusso et al., 2005, Jiménez-Sousa et al., 2018). The major hormonal activity of vitamin D is linked with intestinal calcium transport, renal calcium absorption, insulin secretion, osteogenesis, blood pressure regulation, and apoptosis (Skrobo et al., 2018). In order to initiate all biological processes and to be transported to all tissues 1,25(OH)2D form a complex with vitamin d-binding protein (DBP), an essential substrate to maintain the level of vitamin D in the body (Callejo et al., 2020). Vitamin D action process is interceded by interaction with high-affinity transcription factor VDR. The d-VDR complex modulates expression of gene at the transcriptional stage (Prietl et al., 2013). In addition to classical function, vitamin D also involves the following non-classical functions; immune responses, hormonal secretion, cell proliferation, and maturation (Xu et al., 2020, Dimitrov et al., 2021). Vitamin D acts as a potent immunomodulator expressing intracellular VDR on macrophages, monocytes, and lymphocytes (T and B cells) (Holick, 2007, Bishop et al., 2020). The first evidence of vitamin D's effect on the immune cells in both innate and adaptive immune responses emerged approximately-three years ago (Penna et al., 2005, Bikle, 2009). It regulates the innate immune system by increasing the synthesis of β2 defensins and cationic antimicrobial peptide (CAMP), cathelicidin by macrophages and monocytes, hence enhancinptig their antibacterial activity (Dai et al., 2010). The immunomodulatory effect of vitamin D in adave immune response suggests that it inhibits T-helper1 cell (Th1) activation and synchronizes T-helper2 cell (Th2), T regulatory cell (Treg) and T-helper17 cell activity (Th17) (Boonstra et al., 2001). Vitamin D effect on cell differentiation is affirmed through antigen-presenting (APC) cells and dendritic cells, involved in T cells differentiation (Bishop et al., 2020). Dendritic cell differentiation within a vitamin-D microenvironment regulates a “tolerogenic state” characterized by decreased inflammatory cytokine levels (i.e., IL-12 AND TNF-) and enhanced anti-inflammatory cytokine levels (IL-10), which predominantly upregulate the development of regulatory T cells and promote cell death of autoreactive T-cells (Unger et al., 2009, Bishop et al., 2020). Vitamin D importance in regulating immune responses revealed that its insufficiency is linked with an increased risk of a large number of comorbidities, including metabolic disorders, cardiovascular disorders, osteoporosis, diabetes, autoimmune diseases, cognitive disorders, and certain malignancies (Colotta et al., 2017, Charoenngam and Holick, 2020). Vitamin D deficiency is a worldwide condition with a high occurrence in the general population of both developing (Middle East and subcontinent) and developed nations (Raza et al., 2019). Clinically, hypovitaminosis is described as a serum D (25[OH]D) concentration of 20 ng/ml (Holick, 2007) in both the general population and HIV-infected individuals (Lappe et al., 2007). Sunlight deprivation, black race, malnutrition, low nutritional intake, obesity, high altitude living, and usage of medicines that stimulate catabolism of vitamin D, like glucocorticoids and anticonvulsants, are all well known vitamin D deficiency risk factors (Bischoff-Ferrari et al., 2009). The deficiency of vitamin D is also linked with the prevalence of opportunistic infections and HIV exacerbation, leading to death in untreated patients (Bischoff-Ferrari et al., 2009). Vitamin D helps to increase the lifespan of individuals infected with HIV through a therapy i.e., highly active antiretroviral therapy (HAART) but the risk of comorbidities remains elevated in comparison to the general population, most likely due to immune-suppression (Collins et al., 2020). Another study described that vitamin D and CD4 + T cells made an undeviating association and decreased absolute CD4 + T cells were found among HIV-positive people who were vitamin D deficient after starting HAART (Rosenvinge et al., 2010). A metabolite of vitamin D, (25[OH]D), is served to determine the serum vitamin D concentration in humans (DeLuca, 2004). Although the 1,25-dihydroxy vitamin D (1,25[OH]2D) is the active vitamin D metabolite, routinely it can not be estimated due to its stringent regulation and short half life (Wasserman and Rubin, 2010). Quantifying 1,25[OH]2D alongside 25[OH]D may help to identify other diseases of 1- hydroxylation, 25[OH]D renal conversion to its active form, and irregular extra-renal synthesis of 1,25[OH]2D (systemic infections, sarcoidosis, and hematological disorders) (Wasserman and Rubin, 2010). There is a pressing need to investigate the vitamin D levels of the HIV-infected population. Immunocompromised population has an increased risk of suffering hypovitaminosis-related consequences. Consequently, the present study intends to investigate vitamin D levels in immunocompromised populations, i.e. HIV-infected individuals from different areas of Punjab, Pakistan. In order to conduct this study, n = 1000 subjects were recruited between 2018 and 2019 from Voluntary Confidential Counseling and Testing (VCCT) Centers in Punjab, including Allied hospital Faisalabad, DHQ hospital Chiniot, Aziz Bhatti Shaheed Hospital Gujrat, Benazir Bhutto Shaheed hospital Rawalpindi, DHQ hospital Bahawalnagar, and DHQ hospital DG Khan (Fig. S1). Control blood samples (n = 232) were drawn from healthy populations that have not any known infectious or metabolic disorder. An informed consent about the study was obtained from patients infected with HIV and controls prior to sample collection. A questionnaire was created per the sample collecting inclusion and exclusion criteria. HIV positive status was a prerequisite for inclusion criteria. The co-infected patients with HBV, HCV and syphilis were excluded. The research questionnaire assessed demographic history (age, gender, district, and history of HIV treatment) and lifestyle (urban and rural). The 5CC Ethylenediaminetetraacetic acid (EDTA) whole blood was taken using the World Health Organization (WHO)-recommended venipuncture method. The whole blood was used to obtain a CD4 count, after which plasma was isolated for HIV-PCR by centrifugation. These specimens were then processed in the Advanced diagnostic laboratory. Following the rapid screening test, samples infected with hepatitis B, C, and/or syphilis were excluded from this investigation. Also, subjects who were drug addicts or had infections following surgery were excluded. The present research work was recommended by the research ethics and biosafety committee of the Institute. Initially, samples were tested for HIV Ag/Ab, HBV, and HCV using “AlereTM HIV Ag/Ab,” “AlereTM HBs Ag,” and “SD Bioline anti-HCV Ab” quick diagnostic machines, respectively. The CD4 count was performed on “PIMATM analyzer” using 25 µl whole blood in PIMATM CD4 cartridge. It is an image-based, automated immunological assay intended for rapid in vitro measurement of CD4 cells (T helper cells). Selected samples were treated for RNA extraction using a kit (QIAamp Viral RNA Mini Kit), and a buffer containing guanidine-thiocyanate was used to lyse and homogenize the samples in order to purify and collect the intact RNA. To ensure the appropriate binding conditions, ethanol was added and then any contaminants present were washed away using the RNeasy Mini spin column. Then, high-quality RNA was eluted in 70 ul of AVE buffer. HIV RNA was amplified using the Artus HI Virus-1 RG RT-PCR Kit (the limit of detection for HIV viral load was 71 IU/ml). For PCR amplification, a 50 µl reaction mix was prepared in the “Rotor Gene” cup by adding 30 µl MM (master mix) and 20 µl of extracted RNA. “Rotorgene” was used to amplify the reaction mixture. After performing RT-PCR for HIV viral load, 398 HIV-positive samples were chosen for additional vitamin D analysis. The vitamin D assay was performed by EUROIMMUN 25-OH vitamin D test kit (Noori et al., 2018). The vitamin D status was categorized as sufficient (≥30 ng/ml), insufficient (11–29 ng/ml), or deficient (≤10 ng/ml) (Ginde et al., 2009). Using the laboratory management information system (LMIS) software of PACP-ADL, the patient's age, gender, and area of residence were recorded. SPSS 28.0.0.0 was used to tabulate and analyze the data for all conducted tests and demographic variables. Chi square test was done to see the association of serum 25(OH) D and p-value<0.05 was considered as significant. Blood samples received from various HIV treatment centers of the Punjab i.e., Aziz Bhatti Shaheed hospital (ABS) Gujrat, Allied hospital Faisalabad, Benazir Bhutto Shaheed hospital (BBSH) Rawalpindi, DHQ hospital Bahawalnagar, DHQ hospital Chiniot, DHQ hospital DG Khan and various places from Lahore (Fig. S1). Most of the HIV samples in Lahore (n = 196) region were collected at Jinnah Hospital Lahore 56 (28.6 %) followed by Services Hospital Lahore 51 (26 %) (Fig. 1). A total number of 789 HIV positive samples were obtained after HIV ICT screening out of total 1000 samples and 181 samples were rejected for HCV, HBV and Syphilis co-infection as an exclusion criteria. A subject was considered hepatitis C or B positive if one had anti-HCV antibody or HBsAg positive on ICT device. The remaining 607 samples were further processed for PCR amplification to rule out for the detection of viral load. After following all the screening steps, 398 samples with HIV viral load were preferred for CD4+ count assessment and Vitamin D analysis (Fig. 2). Demographic characteristics of the HIV positive patients such as age, gender, lifestyle and use of cART were investigated. Most of the studied patients were males 275 (69.09 %). TGs were also reported in present study but in low numbers i.e., 6 (1.5 %). Most of the studied population 270 (68 %) was aged between “24–47 years” followed by “1–23 years” 92 (23 %) and was mostly from urban areas i.e., 256 (64.30 %). HIV risk groups were also identified in the studied population. Among them 138 (34.6 %) were intravenous drug users (IDUs) and the general population also had high frequency 90 (22.6 %) after IDUs. In the present study, 279 (70.1 %) samples were of newly diagnosed cases. Controls were also categorized for age and gender (Table 1). Most of the controls were males 155 (67 %) and 139 (60 %) aged between “24–47 years”. In the present study, most of HIV cases have viral load ≤ 10,000 i.e., 233 (56.03 %). CD4 count was also categorized and most of the patients had “200–499 count/µl” i.e., 154 (38.7 %). In the current investigation, deficiency and insufficiency of vitamin D was found in 62 (15.6 %) and 156 (39.2 %) patients, respectively. Most of the control participants had sufficient levels of vitamin D 182 (78 %). Results of statistical analysis indicate a noteworthy link of vitamin D deficiency with HIV cases (p < 0.001). The concentration of 25(OH) D in the population infected with HIV and non infected population (controls) has been shown in Fig. 3. Among HIV positive, non AIDS cases, all vitamin D categories including vitamin D sufficiency, insufficiency, and deficiency were found in 130, 104, and 33 subjects, respectively. The male population had insufficient 25(OH)D levels (65.6 %). A comparable vitamin D association was found in relation to lifestyle and CD4 count of the participants (p < 0.001 and p = 0.009) as shown in Table 2. Various parameters of AIDS patients (CD4 count 〈2 0 0) in different categories with respect to vitamin D were analyzed by using SPSS version 28.0. The categorization was done on the basis of amount of vitamin D such as ≤ 10 ng/ml was considered as deficient patient (n = 27), 11 to 29 ng/ml was insufficient patients (n = 51) and ≥ 30 ng/ml was considered as sufficient patients (n = 53). In the category of gender, male and female were included. Deficiency and insufficiency was prevalent among males (70.37 % & 76.4 %) as compared to female patients (29.6 % & 23.5 %). Lifestyles of AIDS patients were classified in terms of rural and urban. Vitamin D deficiency was more prevalent among rural area participants (70.4 %) while insufficiency was seen among urban cases in high frequency (70.5 %). Lifestyle of AIDS patients was significantly linked with concentration of vitamin D (p < 0.001). Among HIV viral load and CD4 count categories, a significant relationship was determined in relation to 25(OH)D with lower CD4 count and viral load ≥ 10,0000 copies/ml (p = 0.01) (Table 3). Pearson correlation depicted the weak negative correlation of vitamin D with viral load and CD4 count (-0.02, −0.025). However, the correlation was non-significant at CI = 95, p > 0.05 (Table S1). A condition in which serum vitamin D levels are low known as hypovitaminosis D which correlates a number of medical conditions significantly transitioning bone health. In clinical practice, hypovitaminosis D is generally assessed by insufficient 25-hydroxy vitamin D status in serum. The worldwide cumulative agreement suggests that below the optimal range (<30 ng/ml) is considered as vitamin D deficiency (Holick, 2007, Vieth et al., 2007). However, some studies propose the 25–80 ng/ml as optimal range and consider < 20 ng/ml as more apt to explicate deficiency (Ross et al., 2011). The current work reports that deficiency (15 %) and insufficiency (39 %) of vitamin D was higher among HIV infected individuals than local Pakistani population. Oyedele and Adeyemi reported the variance in prevalence rate of deficiency of vitamin D from 10 to 73 % in the population infected with HIV (Oyedele and Adeyemi, 2012). Above and beyond the certain discrepancies in heterogeneous study population, the researchers also ruled out the geographic and demographic differences distressing the concentrations of vitamin D in the individuals under trials. They also appraise the schematic approaches and the controversial cut-off points for low vitamin D levels as significant limitations in cross-sectional research involving HIV patients. Other trials stated that HIV-positive carriers frequently encounter vitamin D deficiency (Rodríguez et al., 2009, Mueller et al., 2010) even on successful combined antiretroviral therapy (cART). Well-known risk factors impairing bone maintenance and attributing to low levels of 25(OH)D in HIV infection include minimized exposure to sunlight and nutritive intake, insufficient absorption, fatty liver, and renal damage encouraged anomalous vitamin D activation, altered bioavailability of non-hydroxylated vitamin D in adipose tissue, and the antiretroviral treatment intervening vitamin D metabolism (Cozzolino et al., 2003). Though progressive age is a typical indicator for hypovitaminosis D (Vescini et al., 2011), this relationship was not significant in the present work (p = 0.29; p = 0.09), and is consistent with prior findings in HIV carriers (Wasserman and Rubin, 2010, Dao et al., 2011, Allavena et al., 2012). Allavena et al. did not observe substantial correlation between age and VDD, one of the justifications may comprise the samples from younger HIV cohorts with few patients over 60 years of age, identical to the current research, which consisted majority of patients ranging within 23–47 years of age (Allavena et al., 2012). The current work revealed that insignificant results of gender and hypovitaminosis D (p = 0.471) consistent with other studies conducted on HIV infected individuals (Wasserman and Rubin, 2010, Allavena et al., 2012, Kwan et al., 2012). The majority of the population in study belonged to urban areas 257 (64.6 %). A notable link was observed between lifestyle and vitamin D status of HIV/AIDS & HIV/non AIDS patients (p < 0.001). Vitamin DD was mostly observed among rural populations. Although there are certain confounding factors for VDD in urban populations such as the sedentary lifestyle, growing urbanization and industrial development may also reduce contact with sunlight but the people usually take supplements in order to meet vitamin deficiencies. The association among 25(OH)D concentrations along with viral load and CD4+ T-cell count is contradictory. Few clinical trials approved a positive correlation (Aziz et al., 2013) while some failed to reveal a significant association (Gedela et al., 2014)). Another study described that lower CD4+ T cell count is linked with deficiency of vitamin D in individuals infected with HIV (Zhang et al., 2017). Present study findings are in good agreement with the results of Zhang et al. (2017) i.e., a significant relationship is established in CD4 count and concentration of vitamin D among HIV/AIDS (p = 0.01) in the current study. Several other investigations have also associated lower 25(OH)D concentration to lower CD4+ T cell counts (Welz et al., 2010, Sudfeld et al., 2012). Because of complexity in their nature, several mechanisms involved in elucidating the link between chronic HIV disease and resultant 25(OH)D decrease are unclear. The individuals with immunosuppression are usually susceptible to a number of clinical complications; one may hypothesize a cause of HIV infection related to VDD. Secondly, as the viral infection progresses to chronic inflammation, elevated certain pro-inflammatory cytokines like TNF-α may interfere with the parathyroid hormone secretion (PTH). The PTH is the stimulatory hormone for the active form of vitamin D i.e. 1,25-dihydroxyvitamin D and its reduced release may result in renal 1α-hydroxylase impairment. Third, patients with low immunity and decreased CD4+ immune cells are more prone to infectious complications and they may have a reduced sun exposure as a contributing factor for hypovitaminosis D. Hospitalization may also be considered an option for food intake limitations, insufficient absorption and eventually malnutrition (Mansueto et al., 2015). In the current work, viral load showed no significant relationship with vitamin D status in HIV/Non-AIDS (p = 0.53) however, HIV/AIDS subjects had a link among these parameters (p = 0.01). Bearden et al. (2013) reported a delicate association, statistically insignificant (p = 0.36), between HIV viral load and vitamin D levels. Among HIV/Non-AIDS subjects, most of the patients had viral load ≤ 10000 while the majority in HIV/AIDS group had > 10000. So increased prevalence VDD in the HIV/AIDS subjects could be explained by the likelihood of interactions between lipopolysaccharide (LPS), toll-like receptor (TLR) signaling pathways, HIV viremia, and proinflammatory cytokines with activation of 25-hydroxyvitamin D-1α-hydroxylase (CYP27B1) in macrophages. Uncontrolled HIV viral loads elvate LPS and proinflammatory cytokines through impaired gut-linked tissues (Anselmi et al., 2007, Brenchley and Douek, 2008). The up-regulation of 1,25(OH)2D receptor-specific Cyp24 mRNA and cathelicidin mRNA was not determined in the absence of vitamin D [25(OH)D] indicating the crucial role of vitamin D in the regulation of such molecules (Liu et al., 2006). In the current investigation, a remarkable borderline relationship was determined in 25(OH)D levels between naïve and patients under antiretroviral therapy among HIV/AIDS subjects (p = 0.05). Aziz et al. (2013) findings showed recovered 25(OH)D3 levels in HAART-treated HIV carriers as compared to the newly infected patients. This recovery is ascribed to the sufficient vitamin D supplementation in patients infected with HIV. There are certain confounders such as the kind of ART given to the patient. Moreover, the present study didn’t include certain individual antiretroviral drugs. Efavirenz interacts with the enzymes responsible for vitamin D metabolism (cytochrome P450 monooxygenases) and potentially induces the enzymes involved in hydroxylation of vitamin D3 to 25(OH)D3 (CYP3A4) and catabolism of 1,25(OH)2D to inactive forms (CYP24) along with reduced CYP2R1 transcription in HIV populations (Kim et al., 2012). Hypovitaminosis D is a risk factor for developing comorbidities including infectious diseases as well as immune disorders in patients infected with HIV. The status of vitamin D of a large cohort of HIV infected patients in comparison to that of the general population particularly in naïve as well as treated patients, is under researched in Pakistani population. In conclusion, the present work documents the pervasiveness of vitamin D insufficiency or deficiency among a large (n = 398) group of HIV-infected subjects in comparison with that in the general population. HIV subjects have a greater pervasiveness of VDD deficiency in comparison to the local population. Vitamin D deficiency occurrence rate was greater in male population infected with HIV and those living in urban areas of Punjab, Pakistan. A strong relationship was determined between the lifestyle and vitamin D level. The significant associations of low concentrations of 25(OH)D, HIV viral load, and CD4 count among AIDS individuals culminate the requirement to assess insufficiency or deficiency of vitamin D as routine care of HIV infected population. No funding was obtained. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

PMC9649960

Xing Guo,Kai Zhong,LongFei Zou,Hao Xue,ShuLing Zheng,Jiang Guo,Hui Lv,Ke Duan,DengHua Huang,MeiYun Tan

Effect of Lactobacillus casei fermented milk on fracture healing in osteoporotic mice

28-10-2022

antibiotics,gut microbiota,probiotics,osteoporosis,fracture healing,renin-angiotensin system,RANKL pathway

The interaction between the gut microbiota and the host has been described experimentally by germ-free animals or by antibiotic-disturbed gut microbiota. Studies on germ-free mice have shown that gut microbiota is critical for bone growth and development in mice, emphasizing that microbiota dysbiosis may interfere with normal bone development processes. This study aimed to clarify the effect of antibiotic treatment on disturbed gut microbiota on bone development in mice and to investigate the effect of probiotic treatment on fracture healing in mice with dysbiosis. Our results showed that 4 weeks old female Kunming mice showed significantly lower abundance and diversity of the gut microbiota and significantly lower bone mineral density after 12 weeks of antibiotic treatment and significantly increased levels of RANKL and Ang II in serum (p<0.05). Mice with dysbiosis received 5 mL of Lactobacillus casei fermented milk by daily gavage after internal fixation of femoral fractures, and postoperative fracture healing was evaluated by X-ray, micro-CT scan, and HE staining, which showed faster growth of the broken ends of the femur and the presence of more callus. Serological tests showed decreased levels of RANKL and Ang II (p<0.05). Similarly, immunohistochemical results also showed increased expression of α smooth muscle actin in callus tissue. These results suggest that oral antibiotics can lead to dysbiosis of the gut microbiota in mice, which in turn leads to the development of osteoporosis. In contrast, probiotic treatment promoted fracture healing in osteoporotic mice after dysbiosis, and the probiotic effect on fracture healing may be produced by inhibiting the RAS/RANKL/RANK pathway.

Effect of Lactobacillus casei fermented milk on fracture healing in osteoporotic mice The interaction between the gut microbiota and the host has been described experimentally by germ-free animals or by antibiotic-disturbed gut microbiota. Studies on germ-free mice have shown that gut microbiota is critical for bone growth and development in mice, emphasizing that microbiota dysbiosis may interfere with normal bone development processes. This study aimed to clarify the effect of antibiotic treatment on disturbed gut microbiota on bone development in mice and to investigate the effect of probiotic treatment on fracture healing in mice with dysbiosis. Our results showed that 4 weeks old female Kunming mice showed significantly lower abundance and diversity of the gut microbiota and significantly lower bone mineral density after 12 weeks of antibiotic treatment and significantly increased levels of RANKL and Ang II in serum (p<0.05). Mice with dysbiosis received 5 mL of Lactobacillus casei fermented milk by daily gavage after internal fixation of femoral fractures, and postoperative fracture healing was evaluated by X-ray, micro-CT scan, and HE staining, which showed faster growth of the broken ends of the femur and the presence of more callus. Serological tests showed decreased levels of RANKL and Ang II (p<0.05). Similarly, immunohistochemical results also showed increased expression of α smooth muscle actin in callus tissue. These results suggest that oral antibiotics can lead to dysbiosis of the gut microbiota in mice, which in turn leads to the development of osteoporosis. In contrast, probiotic treatment promoted fracture healing in osteoporotic mice after dysbiosis, and the probiotic effect on fracture healing may be produced by inhibiting the RAS/RANKL/RANK pathway. Bacteria, viruses, fungi, and protozoa colonized in the gut together form the gut microbiota. In recent years, these microorganisms have been shown to have a strong association with the host, including intestinal physiology, metabolic function, immune system function, and inflammatory processes (1–4). And surprisingly, the gut microbiota has also been found to influence bone growth and development, for example, germ-free mice have lower levels of osteoclastogenic and higher bone mass (5), while male mice with intestinal infections caused by pathogenic bacteria were shown to have increased bone loss (6), and probiotic treatment reduced bone loss in ovariectomized female mice (7, 8) and type 1 diabetic male (9). These studies suggest that dysbiosis of the gut microbiota can lead to bone loss and that probiotic treatment can prevent bone loss due to some factors. However, the effect of antibiotic treatment on bone growth and development in mice remains unclear, and this effect seems to be related to the strain, sex, and age of the mice. Therefore, in this study, female Kunming mice at 4 weeks were selected for study, and antibiotic treatment was applied for 12 weeks to clarify the effect of antibiotic treatment on bone metabolism. Oral probiotics are currently the most commonly used treatment for gut microbiota dysbiosis, and available studies have shown that probiotic therapy can alleviate multiple causes of osteoporosis, such as glucocorticoid-induced osteoporosis (10), postmenopausal osteoporosis (7, 11) and alveolar bone loss due to periodontitis (12). It is well known that osteoporosis can lead to the slow growth of fracture ends and is an important cause of delayed fracture healing. The effect of probiotic therapy on fracture healing in osteoporotic states is still unclear and deserves further study. Lactobacillus casei, one of the most widely researched and used probiotics, the Lactobacillus casei treatment has been shown to alleviate bone loss due to a variety of factors such as type 1 diabetes (13), rheumatoid arthritis (14), ovariectomized (15), etc. Experiments have demonstrated that Lactobacillus fermented milk can affect bone metabolisms, such as promoting osteoblast bone formation in vitro and alleviating osteoporosis in spontaneously hypertensive rats and de-ovulated rats (16–18). Studies have shown that Lactobacillus fermented milk produces valinyl-prolinyl-proline (VPP) and the bioactive peptide isoleucine-prolinyl-proline (IPP), and in addition, these small peptides have angiotensin-converting enzyme (ACE) inhibitory activity that blocks the conversion of angiotensin I (AngI) to angiotensin II (AngII) and inhibits the breakdown of bradykinin by inhibiting ACE (16–18). Recent studies have shown that the expression of components of the renin-angiotensin system (RAS), such as renin, ACE, and Ang II receptors, are present in the local bone microenvironment and callus and are important for bone growth and development (19–21). Activation of RAS in the bone microenvironment has been shown to contribute to osteoporosis by stimulating the release of receptor activator for nuclear factor-κ B Ligand (RANKL) from osteoblasts (22–24), while inhibition of local RAS activation can alleviate bone loss and accelerate bone healing and remodeling (21, 25). Therefore, we hypothesized that dysbiosis of the gut microbiota due to antibiotic treatment could lead to osteoporosis in mice and that oral treatment with Lactobacillus casei fermented milk could alleviate bone loss and accelerate fracture healing in mice, and this effect might be produced by modulating the local RAS. Herein, our results show that 12 weeks of antibiotic treatment leads to osteoporosis in mice. Oral treatment with Lactobacillus fermented milk significantly reduced Ang II and RANKL levels in serum, alleviated osteoporosis, and promoted fracture healing in mice. Eighty-4-week-old female Kunming mice from Chengdu Dasuo Experimental Animal Co., Ltd. were randomly divided into control (n=40) and experimental groups (n=40). Mice were kept in a special pathogen-free environment with free access to sterilized food and autoclaved water. In the experimental group, broad-spectrum antibiotics (0.5 g/l neomycin, 1.0 g/l ampicillin) were added to the drinking water starting at 4 weeks of age, and the water with or without antibiotics was renewed every two days. After 12 weeks of treatment (age: 16 weeks), the bone mineral density of the lumbar vertebrae (L3-4) of mice was measured by dual-energy X-ray scanning to clarify the effect of antibiotic treatment on bone metabolism in mice and to establish a model of osteoporosis after gut microbiota dysbiosis ( Figure 1A ). After confirming the model establishment, the mice were anesthetized by intraperitoneal injection of 1% sodium pentobarbital solution at a dose of 50 mg/Kg, and the right femur of the mice was surgically exposed to the middle segment, and the femoral stem was cut off transversely and flatly with a scalpel to cause transverse fracture of the femur. Then the femoral stem was repositioned so that both ends were fixed and the knee joint was in a flexion-neutral position, and a 1 mL sterile syringe needle was inserted into the medullary cavity parallel to the femoral stem using the center of the intercondylar fossa as the entry point to confirm that the fracture end was well aligned and the needle was firmly fixed, and the remaining needle was cut to make it completely fixed in the femur. After surgery, the experimental mice were divided into the Water group (n=11), Milk group (n=11), and Lactobacillus casei fermented milk group (n=11), and were gavaged with sterile water, milk and Lactobacillus casei fermented milk (5mL/d) for four weeks ( Figure 1B ). All animal experiments were approved by the Animal Research and Care Committee of Southwestern Medical University. Briefly, Lactobacillus casei ATCC393 was added to sterile MRS broth medium for expanded culture and added to skim milk powder solution (11%) at 4% concentration (1.5×108 CFU/mL) and cultured at 37°C for 24 hours. After the skim milk powder solution changed from liquid to viscous, the solution was centrifuged at 4°C for 10 min at 2500 rpm and the supernatant was taken. The pH of the supernatant was then adjusted to 7.5 with a 10% sodium hydroxide solution. After 12 weeks of feeding, feces were collected from all mice using sterile tubes with at least two feces per mouse (>0.1g/serving) and stored immediately at -80°C. Fecal samples from 8 mice in the experimental and control groups were randomly selected for 16S rRNA sequencing using MiSeq technology (26). Briefly, DNA was extracted and the 16S’V4-V5’ region was amplified using specific primers (515F 5’-GTG CCA GCM GCCGCG GTAA-3’; 926R 5’-CCG TCA ATT CMT TTG AGT TT-3’). The amplified products were purified and quantified to create a 16S rRNA library and sequenced (Illumina). Sequences were then quality trimmed to identify and remove chimeric sequences (Trimmomatic and mothur), and sequences were classified using USEARCH software to remove those classified as eukaryotic, archaeal, chloroplast, mitochondrial, or unknown. Finally, the sequence data were filtered to remove any sequences that appeared only once in the dataset, and the clean tags processed above were clustered OTU, and the sequences were clustered to operational taxonomic units (OTUs) with a 97% similarity using USEARCH. The community richness and diversity were analyzed by Mothur to explore the alpha diversity of mouse gut microbiota and the community richness was assessed by ACE estimator and CHO estimator, and the community richness was assessed by Shannon estimator and Simpson estimator. To visually investigate the similarity or difference of the data, the eigenvalues and eigenvectors between samples were determined by calculating ecological distances to perform a principal coordinate analysis based on evolutionary relationships and to create a matrix heat map. Mice were anesthetized and fixed on a test bench, and bone mineral density of the lumbar spine was measured (scan speed 1 mm/s, resolution 0.5x0.5 mm) using dual-energy X-rays (GE Healthcare, Piscataway, NJ, USA). The instrument was calibrated with a companion model before each measurement. the mean BMD of L3 and L4 was considered to be the BMD of the lumbar spine. After 8 weeks of antibiotic feeding and 4 weeks after femur fracture surgery, 1 mL of blood was taken from the heart under anesthesia, placed at 4°C overnight, and then centrifuged at 4°C for 10 min at 8000 rpm. The supernatant was extracted and tested for mouse Angiotensin II (Ang II), and Receptor Activator for Nuclear Factor-κ B Ligand (RANKL) ELISA kits (Elabscience Biotechnology Co., Wuhan, Hubei, China) after returning to room temperature according to the manufacturer’s instructions (27). Tibial and femoral samples were collected 4 weeks after femoral fracture surgery, after execution of the mice, the intramedullary pins were removed and immediately placed in 10 formalin for 24 hours for fixation, and the bone was then transferred to a 70% ethanol solution for scanning with a Siemens Inveon Micro-PET/CT system. Voxel resolution was 20 μm. each run included the bones of mice under experimental conditions and calibrated models to standardize grayscale values and maintain consistency between analyses. The growth of the callus at the fracture was assessed by imaging and 3D reconstruction. The bone was also separated from the bone marrow using a fixation threshold (980), and bone mineral density (BMD), trabecular thickness (Tb. Th), trabecular number (Tb. N), and spacing (Tb. Sp) were assessed in the area of secondary trabeculae 1-1.5 mm from the proximal tibial growth plate, using hand-drawn contour lines to identify the area of interest. The data were then analyzed and recorded using Inveon Research Workplace software (Siemens, Germany). Surface images such as femoral trabeculae were taken from an area of the femur with a length of 1.0 mm and a diameter of 1.0 mm for analysis. All bone analyses were performed without regard to experimental conditions. At weeks 1, 2, and 4 after femoral fracture surgery, mice were anesthetized with 1% sodium pentobarbital solution and fixed on a test bench, and the growth of bone callus was evaluated using X-ray scans. The scoring criteria used were the Radiological United Score of the Tibial(RUST) (28) fracture healing score as follows: visible fracture line without callus Score = 1; visible fracture line with healing tissue formation Score = 2; no visible fracture line with bridging callus Score = 3. The above criteria were evaluated at the anterior, posterior, medial, and lateral aspects of the fracture site, with a total score of 4-12. Four weeks after femoral fracture surgery, femoral specimens were collected after execution of the mice, the intramedullary pins were removed, repeatedly rinsed with saline, and immersed in ethylenediaminetetraacetic acid solution (10%) for one month (sample volume: solution volume >20:1), and the solution was changed every 3 days. The decalcified samples were embedded in paraffin, cut into slices (thickness: 10 μm; 1 slice/sample), and stained with hematoxylin-eosin (H&E). Histological evaluation of stained sections of fracture ends using light microscopy. After preparing the femoral sections using the same procedure as above, the sections were immunohistochemically processed using an anti-alpha smooth muscle actin antibody according to the manufacturer’s instructions (Abcam, United Kingdom). Briefly, sections were washed with TBS solution containing 0.025% Triton X-100 and then closed for 2 h at room temperature using TBS solution containing 10% normal serum and 1% BSA. rabbit antisera against mouse alpha-smooth muscle (Abcam, United Kingdom) applied to sections overnight at 4°C. To visualize the antigen-antibody reaction, the slides were incubated in a TBS solution containing 0.3% H2O2 for 15 min. Finally, sections were counterstained with hematoxylin. Statistical analysis was performed by using Graph Pad Prism. Measurements that conformed to a normal distribution were compared between two groups using the t-test and between three groups using one-way ANOVA. For measurements that did not conform to a normal distribution, the Wilcoxon test was used to make comparisons between the two groups. Differences of P<0.05 were considered statistically significant. To investigate the effect of antibiotic treatment on the gut microbiota of mice, we performed 16S RNA sequencing of fecal samples to analyze the composition of the mouse gut microbiota at three levels: phylum, class, and order ( Figure 2 ). The results showed that after 12 weeks of antibiotic treatment, profound changes in the composition of the gut microbiota occurred at all three levels. Compared with the control group, the antibiotic group showed a significant increase in the relative abundance of Bacteroidetes phylum and a significant decrease in the relative abundance of Firmicutes phylum at the phylum level, and an increase in the relative abundance of Bacteroidales and a decrease in the relative abundance of Clostridiales at the class level, as well as an increase in the relative abundance of Bacteroidales and a decrease in the relative abundance of Clostridiales at the order level. ( Figure 2A ). The non-parametric test results also showed that the above populations were the most diverse at their corresponding levels ( Figure 2B ).To further investigate the effect of antibiotic treatment on the species richness of the gut microbiota, we used the Shannon, Simpson, Ace, and Chao estimators to explore the alpha diversity of the mouse GM, and the results showed that antibiotic treatment significantly reduced the Shannon, Ace and Chao indices, while the Simpson index was significantly increased, suggesting that antibiotics treatment significantly decreased the species richness of mouse GM ( Figure 2C ). In addition, we also multi-beta diversity was analyzed, as shown in Figure 2D The thermal matrix plot and PCoA analysis both showed dispersed samples from different groups and concentrated samples from the same group. In conclusion, the above results suggest that antibiotic treatment significantly altered the composition and reduced the complexity of the gut microbiota in mice. To investigate the effect of antibiotic treatment on bone quality in mice, we assessed bone mineral density (BMD) of the spine using dual-energy X-rays. Dual-energy radiographs showed that the grayness of the spine was significantly lower in antibiotic-treated mice than in controls ( Figure 3A ). Correspondingly, the quantitative analysis of L3-L4 showed a significant decrease in BMD in the antibiotic group compared to the control group ( Figure 3B ). To investigate whether antibiotic-induced osteoporosis is related to RAS, we measured the serum RAS fractions and RANKL levels in mice after 12 weeks of antibiotic treatment. The results showed that Ang II and RANKL levels were significantly increased in the serum of mice in the antibiotic group, suggesting that osteoporosis caused by antibiotic treatment may be produced by activating the RAS ( Figure 4 ). To investigate the effect of Lactobacillus casei fermented milk on fracture healing in osteoporotic mice, we performed X-ray scans on mice at 1, 7, 14, and 28 days after surgery to assess postoperative fracture healing. As shown in Figure 5A , the fracture alignment was satisfactory and there was no obvious displacement of the intramedullary pin on postoperative day 1; on postoperative day 7, the fracture line was visible and healing tissue formation was visible in some mice, but no obvious callus growth was observed; on a postoperative day 14, the fracture line was still clearly visible and callus formation was observed in the stress measurement, and callus formation was significantly more in the Lactobacillus casei fermented milk group; on postoperative day 28, the fracture line was slightly blurred and the callus formation was significantly increased. Significant callus formation was observed in the milk and Lactobacillus casei fermented milk groups on both the stress side and the tension side, but more pronounced callus formation was observed in the water group only on the stress side. We quantitatively assessed the fracture healing using the RUST scale, and the results showed that the fracture healing was significantly better in the Lactobacillus casei fermented milk group than in the water group at all time points ( Figure 5B ). In addition, we performed the histological assessment of fracture healing in mice by HE staining, as shown in Figure 6 . The bone trabeculae in the callus at the fracture in the water group were sparsely and irregularly arranged, those in the callus in the milk group were slightly dense and regular, and the fibrous callus in the Lactobacillus casei fermented milk group disappeared and the bone marrow cavity was partially recanalized, which was significantly better than the water and milk groups. To further investigate the effect of Lactobacillus casei on fracture healing, we performed Micro-CT scans of the tibia and femur of mice at 28 days postoperatively and performed 3D reconstruction to assess callus growth and bone trabecular microparameters. The results showed that only a few calluses formed at the fracture in the water group and partial separation of the fracture were observed, whereas the calluses at the fracture in the mice in the Lactobacillus casei fermented milk group were significantly more numerous and connected into pieces, which was significantly better than the milk and water groups ( Figure 7A ). In addition, the evaluation of tibial trabecular parameters in mice showed that the BV/TV, Tb. Th, and Tb. N of mice in the Lactobacillus casei fermented milk group was significantly higher than those in the water and milk groups, and the corresponding Tb. Sp was significantly lower than that in the water and milk groups( Figure 7B ). α smooth muscle actin (αSMA), a specific marker of osteochondral progenitor cells during fracture healing, is one of the important indicators for assessing fracture healing (29). We explored its expression in callus by immunohistochemistry, and the results showed that αSMA expression was significantly higher in bone scabs of mice in the Lactobacillus casei fermented milk group compared to the water and milk groups ( Figure 8 ). To investigate whether the osteoprotective effect of Lactobacillus casei fermented milk was related to RAS, we measured the serum levels of Ang II and RANKL in mice 28 days after surgery. The results showed that the serum levels of Ang II and RANKL were significantly lower in the Lactobacillus casei fermented milk group compared with the water and milk groups. It is suggested that the osteoprotective effect of Lactobacillus casei fermented milk may be related to the inhibition of RAS activity ( Figure 9 ). Our study showed that antibiotic treatment for 12 weeks led to significant changes in the gut microbiota and osteoporosis in mice, and the levels of Ang II and RANKL in serum were significantly increased, suggesting that RAS may play an important role in bone loss due to gut microbiota dysbiosis. And that Lactobacillus casei fermented milk treatment alleviated osteoporosis and promoted fracture healing in mice, and that this effect may have been produced by inhibiting Ang II and RANKL production. The gut microbiota has recently been shown to affect multiple systems in the host, including the skeleton (1–5). Studies on germ-free mice have shown the importance of gut microbiota for bone metabolism, but the problem of defective immune system development in germ-free mice cannot be avoided (5). Therefore, perturbing the gut microbiota through antibiotic therapy may be a better option (30–33) It is well known that antibiotics are widely used to treat and prevent bacterial infections. However, it has been shown that antibiotic treatment can have lasting negative effects on the host by depleting commensal bacteria leading to dysbiosis and long-term changes in the gut microbiota (34–36). This provides an effective means for exploring the link between gut microbiota and bone metabolism. However, the gut microbiota is usually in a dynamic homeostatic transient state, with strong resistance to external influences. Studies have shown that while short-term antibiotic treatment (1-2 weeks) can perturb the gut microbiota, the gut microbiota can quickly return to its pre-treatment state after antibiotic treatment is discontinued (37). For this reason, our study used neomycin and ampicillin for a longer-term (12 weeks) antibiotic treatment to cause long-term and sustained disruption of the gut microbiota (38). To effectively limit the parenteral effects of antibiotic treatment, two antibiotics, neomycin, and ampicillin, with low intestinal absorption in rodents, were used in this study to minimize parenteral effects (39, 40). Although it is rare for a human to be treated with antibiotics throughout growth and development, there are cases of long-term and significant changes in the gut microbiota due to diet or metabolism (37). Previous studies have demonstrated that antibiotic treatment can have a significant effect on the gut microbiota (30–33), and our findings similarly demonstrate this. Comparing the differences in gut microbiota structure between the antibiotic-treated and control groups by 16s RNA sequencing, our results showed significant alterations in the gut microbiota of mice after antibiotic treatment at multiple levels from phylum to order. Previous studies have shown that antibiotic treatment can cause a decrease in the abundance of the phylum Bacteroidetes, a major phylum in healthy mice and humans and that a decrease in the abundance of Bacteroidetes is associated with many diseases, including IBD and type 2 diabetes (41, 42). And interestingly, in our study, alterations in the gut microbiota of mice after antibiotic treatment were mainly manifested at the phylum level by an increase in the relative abundance of the Bacteroidetes phylum and a decrease in the relative abundance of the Firmicutes phylum. We note the presence of Lactobacillus, a widely studied and used bacterium in the Firmicutes phylum, and studies have shown that Lactobacillus treatment is beneficial in increasing bone mineral density (7, 10, 43). And the relative abundance of Lactobacillus also showed a significant decrease in our study. Therefore, we speculate that the decrease in the relative abundance of Lactobacillus may be one of the reasons for bone loss due to antibiotic treatment. However, it is noteworthy that the altered abundance of the Bacteroidetes phylum in our study appeared significantly different from previous studies, which we believe may be related to the type of antibiotic, dose, duration of treatment, strain, sex, and survival environment of the mice. Many studies have shown that the gut microbiota is critical to bone metabolism and that dysbiosis of the gut microbiota can lead to bone loss (6). This was also demonstrated in our study, where antibiotic treatment resulted in a significant reduction in spinal bone mineral density in mice. Although antibiotic therapy has been widely used to explore the link between gut microbiota and bone metabolism, we note that many studies have shown mixed results. For example, female mice (C57BL/6J) showed a significant increase in BMD after 3 weeks of antibiotic treatment, but no effect after 7 weeks (30); Low doses of penicillin increased BMD in female mice, while males receiving the same treatment showed decreased BMD (44). In conclusion, antibiotic treatment can affect bone metabolism, but the effect depends on the duration of antibiotic treatment as well as the age, sex, and strain of the mice. Oral probiotics are the most common treatment for gut microbiota dysbiosis, and many studies have demonstrated that probiotic therapy can alleviate multiple factors contributing to osteoporosis (7, 8, 45). Lactobacillus is the most widely studied and used, and is available through diet (e.g., yogurt). Therefore, for our study, we chose to use Lactobacillus casei ATCC393 for the treatment of osteoporotic mice after dysbiosis, and in addition, we added Lactobacillus casei to skim milk to make Lactobacillus casei fermented milk to imitate the most common way of obtaining Lactobacillus in humans. Our study showed that Lactobacillus casei fermented milk treatment significantly improved the microstructure of bone trabeculae in osteoporotic mice caused by dysbiosis of the gut microbiota, furthermore, it had a significant effect on the healing of fractures in osteoporotic mice. Our data showed that BV/TV, Tb. Th and Tb. N was significantly higher and Tb. Sp was significantly lower in mice treated with Lactobacillus casei fermented milk, and histomorphology and imaging showed significantly faster and more callus formation and significantly faster healing at the femur fracture. And the expression of αSMA in callus was significantly increased in mouse callus after Lactobacillus casei fermented milk treatment. To investigate the possible mechanism of gut microbiota affecting bone metabolism in this study, we measured the levels of Ang II and RANKL in mice serum, and the results showed that the levels of Ang II and RANKL in mice serum were significantly increased after antibiotic treatment, while Lactobacillus casei fermented milk treatment significantly decreased the levels of Ang II and RANKL in serum. Ang II acts as a major effector protein of RAS, and our results suggest that RAS may have an important role in osteoporosis caused by dysbiosis of the gut microbiota. Tissue localized RAS is called tissue RAS (46), studies have shown that the local RAS is thought to be involved in a variety of physiological and pathological processes, such as insulin secretion, glomerulosclerosis, nephritis, and atherosclerosis (47–50). Recent studies have found that RAS components such as ACE and Ang II receptors are expressed in the local environment of bone tissue and callus (19–21). Many studies have shown that local RAS activation was found to be involved in age-related osteoporosis in mice (22), onset and development of postmenopausal osteoporosis in an ovariectomy model (23, 24), and osteoporosis due to overuse of glucocorticoids (20), suggesting an important role for RAS in the development of osteoporosis. In addition, Shimizu (23) and Zhou (24) found that Ang II promotes osteoclast proliferation and differentiation by inducing RANKL release from osteoblasts, which in turn leads to osteoporosis. Our results also show that alterations in the gut microbiota result in significant changes in serum Ang II and RANKL with the same trend and, more importantly, that these changes can affect bone mineral density and fracture healing rate. This suggests that RAS may be one of the important pathways through which the gut microbiota affects bone metabolism. We note that studies have shown that Lactobacillus fermented milk produces valinyl-prolinyl-proline (VPP) and the bioactive peptide isoleucine-prolinyl-proline (IPP), and in addition, these small peptides have angiotensin-converting enzyme (ACE) inhibitory activity that blocks the conversion of Ang I to Ang II and inhibits the breakdown of bradykinin by inhibiting ACE (16–18). We believe that this may be one of the reasons why gut microbiota affects RAS. And in our study, we found a significant reduction of Lactobacillus with osteoporosis after antibiotic treatment, while Lactobacillus casei fermented milk treatment resulted in significant improvement of bone trabecular microparameters, alleviation of osteoporosis, and a significant acceleration of fracture healing, which seems to corroborate this view, but further studies are needed. And in our study, we found a significant reduction of Lactobacillus with osteoporosis after antibiotic treatment, while Lactobacillus casei fermented milk treatment resulted in significant improvement of bone trabecular microparameters, alleviation of osteoporotic symptoms, and a significant acceleration of fracture healing, which seems to corroborate this view, but further studies are needed. In conclusion, our study shows that antibiotic treatment leads to significant alterations in the gut microbiota and causes the development of osteoporosis in mice, that Lactobacillus casei fermented milk treatment alleviates this osteoporosis and accelerates fracture healing, and that this effect is produced by modulating RAS activity and thus affecting RANKL release. However, this study has several limitations. First, only one concentration of antibiotic treatment was studied, which does not adequately determine the relationship between antibiotic treatment and bone loss and RAS. Second, our results contradict previous studies. The effect of antibiotic treatment on the gut microbiota still needs further study. Third, although 16s RNA sequencing of fecal specimens is currently the predominant way to assess the composition of the gut microbiota, there is a possibility that the fecal microbiota does not reflect the entire gut microbial environment. In addition, we only examined the circulating RAS fractions and RANKL in mice and did not perform the local examination of bone tissue. Therefore, our findings still require further studies to better elucidate the link between gut microbiota and bone metabolism. The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author. The animal study was reviewed and approved by Animal Research and Care Committee of Southwestern Medical University (NO. SWMU20220091). XG, KZ, LZ, contributed equally to this study. All authors contributed to the article and approved the submitted version. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

PMC9649966

Karine Sarkisova,Gilles van Luijtelaar

The impact of early-life environment on absence epilepsy and neuropsychiatric comorbidities

09-11-2022

Early environment,Absence epilepsy,Neuropsychiatric comorbidity,Epigenetic modification,Animal epilepsy model,WAG/Rij

This review discusses the long-term effects of early-life environment on epileptogenesis, epilepsy, and neuropsychiatric comorbidities with an emphasis on the absence epilepsy. The WAG/Rij rat strain is a well-validated genetic model of absence epilepsy with mild depression-like (dysthymia) comorbidity. Although pathologic phenotype in WAG/Rij rats is genetically determined, convincing evidence presented in this review suggests that the absence epilepsy and depression-like comorbidity in WAG/Rij rats may be governed by early-life events, such as prenatal drug exposure, early-life stress, neonatal maternal separation, neonatal handling, maternal care, environmental enrichment, neonatal sensory impairments, neonatal tactile stimulation, and maternal diet. The data, as presented here, indicate that some early environmental events can promote and accelerate the development of absence seizures and their neuropsychiatric comorbidities, while others may exert anti-epileptogenic and disease-modifying effects. The early environment can lead to phenotypic alterations in offspring due to epigenetic modifications of gene expression, which may have maladaptive consequences or represent a therapeutic value. Targeting DNA methylation with a maternal methyl-enriched diet during the perinatal period appears to be a new preventive epigenetic anti-absence therapy. A number of caveats related to the maternal methyl-enriched diet and prospects for future research are discussed.

The impact of early-life environment on absence epilepsy and neuropsychiatric comorbidities This review discusses the long-term effects of early-life environment on epileptogenesis, epilepsy, and neuropsychiatric comorbidities with an emphasis on the absence epilepsy. The WAG/Rij rat strain is a well-validated genetic model of absence epilepsy with mild depression-like (dysthymia) comorbidity. Although pathologic phenotype in WAG/Rij rats is genetically determined, convincing evidence presented in this review suggests that the absence epilepsy and depression-like comorbidity in WAG/Rij rats may be governed by early-life events, such as prenatal drug exposure, early-life stress, neonatal maternal separation, neonatal handling, maternal care, environmental enrichment, neonatal sensory impairments, neonatal tactile stimulation, and maternal diet. The data, as presented here, indicate that some early environmental events can promote and accelerate the development of absence seizures and their neuropsychiatric comorbidities, while others may exert anti-epileptogenic and disease-modifying effects. The early environment can lead to phenotypic alterations in offspring due to epigenetic modifications of gene expression, which may have maladaptive consequences or represent a therapeutic value. Targeting DNA methylation with a maternal methyl-enriched diet during the perinatal period appears to be a new preventive epigenetic anti-absence therapy. A number of caveats related to the maternal methyl-enriched diet and prospects for future research are discussed. Generalized epilepsies have a strong genetic component. The variety of mutations and involvement of different combinations of genes, the genetic heterogeneity, and the heterogeneousness of the various epileptic phenotypes hamper quick progress and understanding of the different pathologies underlying generalized genetic epilepsies (GGE). Absence epilepsy is classified as an epilepsy syndrome with a genetic cause according to the current International League Against Epilepsy (ILAE) classification system (Scheffer et al., 2017). In previous versions of the classification system, absence epilepsies were referred to as idiopathic syndrome. The GGE represents 15–35 % of the population of people with epilepsy (Jallon and Latour, 2005). Absence epilepsy is the most common form of generalized epilepsies during childhood. In large cohorts, the frequency of the most prevalent type of absence epilepsy, childhood absence epilepsy (CAE), varies from 1.5 % to 12.1 % among the genetic generalized epilepsies, and this large difference depends largely on the mode and source of the case definition. The incidence of CAE has been estimated to range from 0.7 to 8 per 100,000 persons, and the prevalence rates of CAE in children’s cohorts ranged from 0.1 to 0.7 per 1000 persons (Jallon and Latour, 2005, Matricardi et al., 2014). In adults (> 17 years), the average GGE incidence is much lower: only 2.9/100 000 inhabitants per year and the prevalence of absence epileptic patients is 0.3 among 1000 persons (Gesche et al., 2020). Therefore, the absence epilepsy is most commonly, although not exclusively, considered a pediatric disease. For a long time, CAE was considered to be a benign form of epilepsy given the high (60–70%) complete remission rate (Berg et al., 2014, Bouma et al., 1996) and the absence of comorbid pathologies. However, subsequent studies have shown that CAE is associated with cognitive impairments, among other things, attention disturbances, and psychiatric comorbidities, such as attention-deficit/hyperactivity disorder (ADHD), anxiety, and depression (Barone et al., 2020, Cheng et al., 2017, Gruenbaum et al., 2021, Masur et al., 2013, Tenney et al., 2013, Tenney and Glauser, 2013, Vega et al., 2011). Cognitive impairments, including recognition memory and attention disturbances, were also found in the WAG/Rij model of CAE (Fedosova et al., 2022, Fedosova et al., 2021, Leo et al., 2019). Two animal models are at the forefront of neurobiological research towards mechanisms and processes that govern generalized absence epilepsy. The GAERS and WAG/Rij models, both models first discovered in the eighties of the last century (van Luijtelaar and Coenen, 1986, Vergnes et al., 1982), are well described and considered as models with a high predictive, construct and face validity (Coenen and van Luijtelaar, 2003, Depaulis and Charpier, 2018, Marescaux and Vergnes, 1995). The WAG/Rij strain was already fully inbred before the discovery that they showed spontaneous SWDs concomitant with clinical signs (van Luijtelaar and Coenen, 1986), while the GAERS, originally a selection line, were later fully inbred. Current theories on the origin of absence epilepsy in humans have been molded by discoveries in these models (Meeren et al., 2002, Polack et al., 2007), have led to the cortical focus theory on absence epilepsy, and inspired research in other species (Ding et al., 2019, Zobeiri et al., 2019) and absence epileptic patients (Crunelli et al., 2020, Gupta et al., 2011, Miao et al., 2014, Moeller et al., 2008, Ossenblok et al., 2019, Tenney et al., 2013, Westmijse et al., 2009, Youssofzadeh et al., 2018). When the animals of these two inbred strains are born, they do not have seizures. The spike-wave discharges (SWDs) are expressed age-dependently in both WAG/Rij rats (Coenen and van Luijtelaar, 1987, Gabova et al., 2020) and GAERS (Jarre et al., 2017). In WAG/Rij rats, the number of animals expressing SWDs, the number of SWDs per hour, as well as the mean duration of individual SWDs increases with age from about 2 months of age (Coenen and van Luijtelaar, 1987, Gabova et al., 2020, Schridde and van Luijtelaar, 2005). SWDs in the rodent rat models have a frequency of about 7–11 Hz, mean duration of about 5 s, and the number of daily SWDs in 6-month-old WAG/Rij rats is about 16–18 per hour (van Luijtelaar and Coenen, 1986, van Luijtelaar and van Oijen, 2020). GAERS have more and longer SWDs, as well as an earlier onset of the fully developed SWDs (Akman et al., 2010). However, the SWDs incidence, number and mean duration may vary in different colonies of WAG/Rij rats, as has been found in GAERS (Powell et al., 2014). In WAG/Rij rats, age-dependent increases in the number and mean duration of SWDs occurred unevenly: from 2 to 6 months of age, these SWDs characteristics increase by 20 and 6 times, respectively, and from 6 to 12 months of age only by 1.5 times. In general, the epileptic phenotype in WAG/Rij rats is fully expressed by 6–7 months of age. However, the SWDs amplitude and asymmetry index continues to increase up to 8–9 months of age, indicating that the evolution of SWDs in WAG/Rij rats is completed by 8–9 months of age (Gabova et al., 2020). It is not excluded that age-related dynamics of SWDs development in other colonies of WAG/Rij rats may have some differences as well. Detailed recent studies in GAERS (Jarre et al., 2017) and WAG/Rij rats (Gabova et al., 2020) have described the precursors of SWDs, and not mature SWDs were retrospectively present as early as 2 months (WAG/Rij) and 30 days (GAERS) of age. Then, with age, immature discharges successfully undergo three stages of “maturation” (Gabova et al., 2020, Jarre et al., 2017), which reflects progressive electrophysiological changes in the somatosensory cortex (Jarre et al., 2017), a brain region that is related with the initiation, spreading and generalization of SWDs in the WAG/Rij and GAERS models (Meeren et al., 2002, Polack et al., 2007). The vulnerability of the immature developing rat brain towards external influences has been known for a long time, including the age-dependent decrease during ontogeny (Wasterlain and Plum, 1973), and these external environmental factors might have long-term consequences, including the phenotypic expression of genetically determined epilepsies. Early quantitative genetic studies, a complete Mendelian cross-breeding study between WAG/Rij rats and another fully inbred strain, Agouti Copenhagen Irish (ACI), not only confirmed a genetic transmission of genes responsible for the occurrence of SWDs, but equally demonstrated epigenetic and early environmental influences, including maternal effects, for SWD characteristics such as incidence, mean duration, and number of SWDs. Maternal effects were evident since rats of the F1 generation raised by ACI versus WAG/Rij mothers showed different characteristics of SWDs (Peeters et al., 1992, Peeters et al., 1990). The contribution of maternal effects on SWDs was confirmed in a second Mendelian cross-breeding study between BN and F344 rats (Vadasz et al., 1995). Later, the role of maternal behaviour as an early environmental factor contributing to the development of absence epileptic phenotype in WAG/Rij rats was established (Sarkisova and Gabova, 2018) and will be discussed further in Section 3.5 “Maternal care”. Previous data have shown that WAG/Rij rats develop also a mild depression-like (dysthymia) phenotype in parallel with spontaneous absences (Sarkisova et al., 2003), and, importantly, that prevention of epileptogenesis by early and chronic drug administration also prevented the development of a mild depression-like phenotype, suggesting a close and causal relationship between them (Russo et al., 2016, Russo et al., 2011, Sarkisova and van Luijtelaar, 2011, Sarkisova et al., 2010, van Luijtelaar et al., 2013). Interestingly, no phenotypic expression of behavioural depressive-like symptoms was found in pre-symptomatic (36-day-old) WAG/Rij rats. Behavioural depression-like symptoms appear at the age of 3 months when SWDs start to be clearly expressed. Then, with age, depressive symptoms increase, as far as absence epileptic seizures aggravate (Sarkisova et al., 2014). In all, both aspects of the phenotype, the spontaneous absence seizures combined with the depressive-like behaviour, are co-occurring in this model and are under the influence of external factors, including chronic drug administration. Since the early nineties, quite a number of studies have been done regarding the effects of environmental factors on the phenotypic expression of SWDs and psychiatric comorbidities in WAG/Rij rats, here we will review these studies, and we will emphasize studies in WAG/Rij rats, but not exclusively. Moreover, early intervention studies in other epilepsy models might be discussed as well, since they may point towards putative environmental factors that might play a role in the epigenetic effects on SWDs and comorbid depression-like behaviour in WAG/Rij rats. This review will discuss prenatal, perinatal, and postnatal factors, with an emphasis on early postnatal influences, considering that most studies were done in the most vulnerable period of the immature brain. Environmental exposures must occur during the window of susceptibility (at specific times for different cell types’ maturation) to produce alterations in offspring, which can persist throughout life. This differential sensitivity to environmental factors during development is a well-known fact, which is based, among other things, on an early observation (Wasterlain and Plum, 1973). These authors compared the effects of a supramaximal electroconvulsive seizure per day for 10 days between days 2 and 11, days 9 and 18, and days 19–28 of their postnatal life (Post Natal Day, PND). The seizures as presented in the youngest group had a larger effect on brain weight, and brain cell number, the slightly older treated group had a smaller reduction in brain weight, brain protein, and brain RNA without a fall in brain DNA, suggesting a reduction of cell size without a change in cell number. The brains of the oldest shocked group showed no change in brain weight, cell number, or cell size. All this suggested that the immature rat brain is more vulnerable to seizures than the older brain. Now it is widely accepted that the brain growth spurt in rats and mice occurs postnatally, with peak growth velocity on PND7-PND10, and ends in the third week, suggesting that these early weeks of life are the most sensitive to disturbances due to adverse environmental factors. The fact that the brain in the offspring is vulnerable during the perinatal period is important clinically since it is now established that the majority of anti-seizure medications (ASMs) cause apoptotic neurodegeneration in the developing rat brain at doses and plasma concentrations relevant for the anticonvulsant treatment of the pregnant mothers with epilepsy (Ikonomidou, 2010). Prenatal and early postnatal exposure to environmental factors primarily associated with the mother and including early-life stress, maternal care, and nutrition, taking pharmaceutical drugs or medication can exert adverse effects on offspring and promote the development of different diseases later in life (Tchernitchin and Gaete, 2015). These early environmental factors that affect an organism's phenotype result from gene-environment interactions, which are mediated by epigenetic modifications of the genome. Epigenetic modifications regulate gene expression without altering the DNA sequence. Epigenetic modifications can include DNA methylation, histone modifications, chromatin remodeling, and non-coding RNA, as well as more recently identified mechanisms such as hydroxymethylcytosine residues (Gräff et al., 2011, Kriaucionis and Heintz, 2009, Skinner et al., 2010). DNA methylation refers to the process by which a methyl group attaches to DNA via cytosine at specific locations in the genome called CpG sites (Razin, 1998). DNA methylation causes a stable but potentially reversible change in gene expression. DNA methylation usually results in the silencing of a gene, but recent evidence indicates that methylation may also be associated with gene activation (Métivier et al., 2008). Epigenetic modifications of the genome provide a mechanism that allows the stable propagation of gene activity from one generation of cells to the next (Jaenisch and Bird, 2003). Unlike genetic changes, epigenetic modifications are more dynamic and are often reversible, and allow the phenotype to adapt to alterations in the environment throughout life; it is the ultimate solution of nature for proper adaptation to different circumstances in normal neurodevelopmental processes in all living species. However, variations in epigenetic modifications can contribute not only to phenotypic diversity (variability of normal phenotypes adapted to a specific environment) but also to the development of pathologic phenotypes such as psychiatric, neurodegenerative, and neurological diseases, including epilepsy and associated behavioural comorbidities in predisposed individuals (Fig. 1). Early environmental factors may impact brain cell programming and through epigenetic mechanisms can produce susceptibility or, on the contrary, resistance to the development of various neurological and neuropsychiatric diseases, including epilepsy and its comorbidities. Therefore, knowledge of epigenetic mechanisms of the different disorders could lead to the development of new preventive or protective therapeutic strategies. Epigenetics of epileptic and psychiatric diseases is a new promising area of research. Genomic analysis is very important to understand the risk for the development of any pathology. However, it should be emphasized that the genome may be altered by epigenetic mechanisms, which can modify the susceptibility to develop diseases. Prenatal effects of substances, behavioural procedures, and the occurrence of seizures may lead to modifications in various neurotransmitter systems, neural excitability, and seizures. Well-known are the prenatal cocaine administration studies of Baraban et al. (1997) and Keller and Snyder-Keller (2000). They found that prenatally administered cocaine reduced the cocaine-induced seizure threshold later in life in a sex-dependent manner since only females showed an increased sensitivity to cocaine-induced seizures. This sex difference was not present in the same pretreated dams when the seizures in adulthood were induced by acute injection of pentylenetetrazole (PTZ), indicating that their increased seizure susceptibility was specific to cocaine. It was thought that limbic structures such as the piriform cortex, amygdala, and hippocampus are heavily involved since the PTZ seizure was associated with intense c-fos induction in these limbic structures. Ethanol is a psychoactive substance that exerts a detrimental effect on the brain, especially during neurodevelopment. Prenatal exposure to alcohol can cause a number of physiological, cognitive, and behavioural abnormalities in offspring, called Fetal Alcohol Syndrome (FAS), which is also associated with increased susceptibility to convulsive seizures (Cho et al., 2017) and predisposition to neuropsychiatric disorders, in particular, to anxiety and depression (Easey et al., 2019). The degree to which alcohol affects offspring’s development depends of a variety of factors, such as timing, level of alcohol exposure, and genetic background. Recent studies have shown that epigenetic modifications, such as DNA methylation and post-translational histone modifications, can mediate the effects of prenatal alcohol exposure on gene expression in the brain structures and, as a consequence, on the phenotype of offspring. There is evidence that persistent changes of DNA methylation profiles and histone modifications may be associated with the long-term physiological and neurobehavioural alterations observed in FAS (Lussier et al., 2018, Mandal et al., 2017, Zhang et al., 2015). Human and animal studies indicate that prenatal alcohol exposure results in devastating neuropathological consequences in the offspring, including suppression of the migration of cortical GABAergic interneurons, a reduction in the number of cortical neurons, as well as alterations in radial glial cells (Luhmann, 2016). In postnatal studies in humans, aplasia or hypoplasia of the corpus callosum, reduced cortical thickness, profound changes in the frontal cortex, cerebellum, and a reduced fronto-thalamic distance were detected (Kfir et al., 2009, Riley et al., 2004, Sowell et al., 1996). Interestingly, all these brain structures are involved in genetic generalized epilepsy of the absence type. Few studies have described the effects of prenatal exposure to alcohol on epileptic seizures in genetic models. We found the only study of the effect of perinatal alcohol exposure on genetic models of generalized epilepsy (Russo et al., 2008). In this research, the effects of ethanol given during the last 2 weeks of gestation and for 1 week after delivery were compared between Genetically Epilepsy Prone Rats (GEPR), a genetic epilepsy model with convulsive seizures elicited by audiogenic stimulation (Ribak, 2017), and WAG/Rij rats with genetic absence epilepsy. Two non-epileptic strains were used as controls. In addition, susceptibility to PTZ-induced seizures was assessed in ethanol-treated and control animals. It was found that perinatal exposure to ethanol increases the number and intensity of audiogenic seizures in GEPR, but does not affect the susceptibility to PTZ-induced seizures in a non-epileptic control group. In WAG/Rij rats, the effects were opposite: perinatal exposure to ethanol decreased the number of SWDs, their mean and total duration. Control Wistar rats were virtually without SWDs. In Wistar rats exposed to ethanol, SWDs were completely absent. Perinatal ethanol exposure had a protective effect on PTZ-induced seizures in WAG/Rij rats. This effect was the most prominent following administration of PTZ at a dose of 50 mg/kg. The authors suggest that the disruption in the cortico-thalamo-cortical circuit and specifically alterations in the somatosensory cortex may underlie a reduction in the number and duration of SWDs in WAG/Rij rats perinatally exposed to ethanol. The increased sensitivity to audiogenic seizures in GEPR prenatally exposed to ethanol can be attributed to GABAergic abnormalities in the inferior colliculus contributing to initiation and spread seizure activity during audiogenic stimulation (Ribak, 2017). The data of Russo et al. (2008) demonstrate a clear gene-environment interaction, namely the opposite effect of the same treatment on 2 genetic epilepsy models, one convulsive and another non-convulsive. The second interesting fact is that the long-term effects of perinatal alcohol exposure on PTZ-induced seizures were the same irrespective of the genotype: in both cases, antiepileptic, similar to the effect of perinatal alcohol exposure on SWDs. This suggests that non-convulsive absence seizures and PTZ-induced convulsive seizures may use some of the same cortico-thalamo-cortical circuits. ASMs are often used to treat epilepsy and neuropsychiatric disorders in women during prenatal, perinatal, and postnatal periods because these illnesses almost always require continued pharmacological treatments. The evidence that some of the ASMs, and certainly valproic acid (VPA), are causing adverse neurodevelopmental effects is overwhelming (Daugaard et al., 2020, Wlodarczyk et al., 2012). ASMs cross the placenta, producing substantial fetal medication exposure. Preclinical studies showed that some ASMs, such as phenobarbital, phenytoin, vigabatrin, and VPA can induce neuronal apoptosis and impaired neurogenesis in the immature brain (Bittigau et al., 2002). Lamotrigine, VPA, and vigabatrin may impair neuronal migration in the developing brain (Manent et al., 2008). Prenatal VPA produces alterations of multiple gene expressions in offspring by affecting epigenetic mechanisms such as histone acetylation and DNA methylation, participating in fundamental developmental and regulatory processes. VPA acts through several mechanisms including an increase of gamma-aminobutyric acid (GABA) in the brain due to the inhibition of its catabolism, the suppression of voltage-sensitive channels, and the inhibition of histone deacetylase. VPA is a powerful epigenetic modifier that mainly affects brain development in offspring (Ornoy et al., 2020). VPA interferes with one-carbon metabolism (Fig. 2), including the transport of methylfolate into the brain and placenta, and the effect of VPA on folate metabolism is thought to be associated with increased risk for VPA-induced fetal developmental abnormalities. However, genetic factors, notably polymorphisms related to one-carbon metabolism, contribute to the vulnerability to VPA-induced developmental risks (Bottiglieri et al., 2000, Reynolds and Green, 2020). Decreased DNA methylation was found in neonates of mothers who took ASMs during pregnancy (Smith et al., 2012). Epigenetic modifications of gene expression induced by prenatal ASMs exposure might underlie increased susceptibility to seizures and epilepsy in offspring later in life. Prenatal exposure to VPA was shown to increase susceptibility to kainic acid-induced seizures in adult mice offspring due to aberrant and/or ectopic hippocampal neurogenesis associated with significant alterations in the expression of neuron differentiation and nervous system development-related genes. VPA disrupted the expression of multiple genes, including neural stem/progenitor cells genes associated with cell migration, consequently producing ectopic localization of newborn neurons in the hilus (Sakai et al., 2018). Ectopically located granular cells in the hilus of the hippocampal dentate gyrus are more excitable than those located normally in the granular cell layer (Zhan et al., 2010). VPA also influenced another cell type in the immature brain: it selectively suppressed inhibitory synaptic formation by repressing the expression of a vesicular GABA transporter and glutamate decarboxylases in neurons, leading to a shift in balance towards more excitation in rat cortical neurons (Fukuchi et al., 2009, Kumamaru et al., 2014). VPA exposure resulted in distinct defects in synaptic differentiation of neocortical neurons and retardation of axonal growth specific to GABAergic neurons (Iijima et al., 2016). Disturbance of the excitatory and inhibitory balance resulting from neurodevelopmental dysfunction in GABAergic circuitry might be a common cause not only of increased seizure susceptibility and epilepsy but also of multiple comorbid behavioural, emotional, and cognitive impairments, autistic spectrum disorder (ASD) and ADHD (Christensen et al., 2013, Genovesi et al., 2011, Gesche et al., 2020, Iijima et al., 2016, Williams et al., 2016). Prenatal VPA administration in Wistar rats modified the susceptibility to generalized tonic-clonic seizures induced by PTZ, but not to status epilepticus (SE) induced by lithium-pilocarpine in the 2-week-old offspring. Two subgroups, with different PTZ-induced seizure susceptibility, were found after prenatal exposure to VPA. The highly susceptible subgroup exhibited an increased duration of generalized tonic-clonic seizures and developed SE, while the low susceptible subgroup exhibited only minimal seizures. The authors speculate that the seizure protective effect of prenatal VPA may be due to an increase in fetal levels of GABA and changes in drug metabolism promoting an increased degradation of PTZ. The reason for being highly susceptible to PTZ was thought to be an under-development of the GABAergic system (Puig-Lagunes et al., 2016). Very few studies were done on the genetic epilepsy models. A recent study showed that exposure to VPA for 2 weeks before conception that was continued during pregnancy in GAERS, Non-Epileptic Controls (NEC), and random bred non-epileptic Wistars caused a significant reduction in birth weight and length via C-birth born pups in which the mothers were VPA-treated. Moreover, there were altered intravertebral distances and a delay in the development of the vertebral arches in the VPA groups. These neurodevelopmental and vertebrate impairments were not affected by strain (Jazayeri et al., 2020). The outcomes showed, besides the well-known teratogenicity of VPA, that the epileptic genotype was not a contributing factor in this adverse effect. Klioueva et al. (2001) investigated the effects of daily subconvulsive (40 mg/kg) injections of PTZ (7–10 injections) in pregnant WAG/Rij dams on the susceptibility to spontaneous occurring SWDs in offspring when they were 4–5 months old. EEG recordings in the adult offspring showed that prenatally treated rats have three times fewer SWDs during a two-hour recording period at PND130. The authors proposed that the decrease in SWDs in prenatally treated subjects is due to reduced functional GABAergic activity. Evidence for this came a few years later from a study from Naseer et al. (2009): they showed that maternal PTZ-induced seizures indeed decreased GABAB1 receptors, next to neuronal death via caspases-3 in hippocampal neurons. The hippocampus is not a primary brain region in the typical absence epilepsy (Inoue et al., 1993, Onat et al., 2013), although it may modulate the occurrence of SWDs (Tolmacheva and van Luijtelaar, 2007), and during SWDs there is an increased network coupling between the somatosensory cortex, containing the SWD initiation site, and the hippocampal dentate gyrus (Papp et al., 2018). However, a decrease in SWDs was obtained when the GABA reuptake blocker tiagabine was locally administered in the hippocampus, suggesting that an increased GABAergic hippocampal functioning, not a diminished GABAergic functioning, could be the cause of a reduced number of SWDs in the offspring of prenatally PTZ-treated dams. Therefore, it is more likely that functional disturbances are the key players in the absence epilepsy essential circuits, the networks between cortex and thalamus affected by prenatal administration of PTZ-induced decreases in SWDs. It is not known, however, whether there is also a decreased functional activity of GABAB receptors, induced by prenatal PTZ exposure, in the cortex and thalamus. GABAB receptors are not only highly expressed in the hippocampus, but also cortex and thalamus (Bischoff et al., 1999). A role of GABAB receptors in the increased cortical excitability has been demonstrated in WAG/Rij rats: levels of mRNA for most GABAB receptors are diminished in the neocortex of WAG/Rij rats, and higher doses of the GABAB agonist baclofen are required to depress pharmacologically isolated, stimulus-induced IPSPs generated by cortical neurons. It is therefore thought that a decreased function of presynaptic GABAB receptors in the neocortex may contribute to neocortical hyperexcitability and the occurrence of SWDs (Inaba et al., 2009). Given the functional deficiency of GABAB receptors in the somatosensory cortex in WAG/Rij rats (Merlo et al., 2007), it is not likely that the diminishment in SWDs in WAG/Rij rats found by Klioueva et al. (2001) after prenatal PTZ exposure is due to a further cortical deficiency of GABAB receptors. The last major candidate to explain the decrease in SWDs by prenatal PTZ administration is the thalamus. GABAB receptors in the thalamus may underlie the generation of SWDs since they play a role in burst firing by activating low-threshold calcium currents (LTCC) (Crunelli and Leresche, 1991). Moreover, Liu et al. (1992) showed that injections of the selective GABAB agonist R-baclofen into the ventrolateral thalamus increased the number of SWDs in GAERS in a dose-dependent way, whereas the same but also systemic injections with the GABAB antagonist CGP 35–348 decreased SWDs in old Wistar rats with lots of SWDs (Puigcerver et al., 1996). This might imply that a diminished GABAB receptor function in the thalamus may indeed be responsible for the reduction in SWDs after prenatal administration of PTZ in the pregnant dams. Nowadays it is widely assumed that an increased tonic inhibition in the thalamus is a major subcortical determinant of SWDs in various genetic absence epilepsy and seizure models, including the GAERS, stargazer, and lethargic mice models (Cope et al., 2009, Crunelli et al., 2020). In case prenatal PTZ is indeed causing a functional deficiency of GABAB receptors in the thalamus in the offspring, this may lead to a diminishment of tonic inhibition and fewer SWDs. Although GABAA receptors are the constituent receptor types for tonic inhibition, studies in rat brain slices have shown that activation of postsynaptic GABAB receptors enhances the magnitude of the tonic GABAA current recorded in thalamocortical cells. This demonstrates postsynaptic crosstalk between GABAB and GABAA receptors. A consequence of this cross-talk could be that also GABAB receptors contribute to tonic GABAergic inhibition, allowing an explanation for the modulatory effects of GABAB agonists and antagonists in increasing and decreasing SWDs respectively, and for the prenatal effects of PTZ (Connelly et al., 2013). The direction of the prenatally-induced effects of PTZ on SWDs, specifically a reduction in SWDs, is opposite to the commonly reported pro-epileptic effects of prenatal seizures in the offspring later in life. Once more, a reduction of thalamic and/or hippocampal GABAergic functioning might explain both the effects of an increase in limbic convulsive seizures and the decrease in SWDs in adult rats. Early-life stress is one of the strongest environmental factors that can produce epigenetic modifications that persisted into adulthood and could be transmitted across generations. Abnormal epigenetic modifications, in particular DNA methylation, are thought to be a cause of many diseases in humans (Gräff et al., 2011, Kiss et al., 2016, Liu et al., 2008, Nilsson et al., 2018, Robertson, 2005, Skinner et al., 2010). It has been shown that early-life stress-induced epigenetic modifications, leading to behavioural disorders in adult animals, can be transmitted across generations (Franklin et al., 2011, Franklin et al., 2010, Razoux et al., 2017, Stenz et al., 2018). Neonatal maternal separation, neonatal handling, neonatal isolation, poor maternal care, or exposure neonates to aggressive foster mothers and/or fathers (early-life maltreatment models) can be regarded as powerful early-life stressors, and therefore all of them could produce epigenetic modification of DNA and, as a consequence, trigger and/or provoke (in genetically susceptible individuals) seizures, epilepsy and comorbid behavioural abnormalities later in life. There is no conclusive evidence on whether early-life stress can provoke epileptic seizures and comorbidities in humans possibly because most human studies were based on retrospective self-repots. Only a single study was found in which epileptic children raised in warzones and non-warzones in Croatia were compared and showed that in the warzone the assumed higher stress levels in children were accompanied by a greater number of generalized seizures (both absence and tonic-clonic seizures), while in non-warzone a much smaller increase in the number of partial seizures was seen. Moreover, a first epileptic seizure during the war (without a subsequent diagnosis of epilepsy) was directly linked in time to a stressful event, suggesting that severe stress can potentially provoke epileptic insults without a primary process of epileptogenesis or predisposition (Bosnjak et al., 2002). Early in development, the brain is thought to be more prone to seizures, probably caused by an age-related imbalance between excitation and inhibition, which may be associated, among other things, with the initially excitatory effects of the neurotransmitter GABA (Holmes, 1997). Because of this, the effect of stress on epilepsy and epileptogenesis is expected to be more apparent during brain development. The brain structures, which regulate the stress response, among others the hippocampus, are also often involved in epilepsy, and therefore stress, especially strong, long-lasting, and acting very early in life (in a critical window for epigenetic modifications) could play a role in triggering seizures (Huang, 2014). Early-life stress can affect brain excitability (Dubé et al., 2015) and/or connectivity (Nephew et al., 2017, Razoux et al., 2017) and can provoke seizure generation and epilepsy (Dubé et al., 2015, Gunn and Baram, 2017, Salzberg et al., 2007). In animal studies, using the amygdala kindling model of limbic epileptogenesis, it has been shown that burst firing in the thalamic reticular nucleus was significantly increased in kindled rats previously subjected to maternal separation-induced early-life stress compared to kindled rats previously subjected to neonatal handling. Maternal separation also enhanced the burst firing of hippocampal pyramidal neurons. Following kindling-induced seizures, somatosensory cortical neurons exhibited a more pronounced increase in burst firing in rats subjected to early maternal separation than in rats subjected to neonatal handling. Results suggest that early-life stress enhances vulnerability to limbic epilepsy in adulthood, as evidenced by changes in firing patterns in thalamocortical and hippocampal regions (Ali et al., 2013) and by reduced electrical seizure thresholds and prolonged seizure duration during kindling epileptogenesis (Koe et al., 2014). Thus, an early stressful environment appears to promote a vulnerability to epilepsy development due to alterations in brain excitability (Jones and O’Brien, 2013). It may be assumed that early-life stress-induced changes in brain connectivity (Liu et al., 2016, Nephew et al., 2017) and region-specific structural abnormalities and dysfunction (Yang et al., 2015) could underlie behavioural impairments, including anxiety and depression, associated with epilepsy. Early-life social isolation stress (single housing) was shown to produce increased seizure susceptibility, epileptogenesis, and neurochemical alterations leading to anxiety and depression (Mumtaz et al., 2018). Socially isolated mice had a higher network degree, suggesting higher overall connectivity, as well as abnormal network structure and white matter microstructure (Liu et al., 2016). Changes in the resting functional connectivity were found in the genetic absence epilepsy both in humans and animal models. However, it is unknown whether early-life stress induced by insufficient maternal care provided by depressive WAG/Rij mothers (see Section 3.5) could contribute to the alterations of network structure in adult WAG/Rij rats genetically predisposed to absence epilepsy. Absence seizures in the WAG/Rij rat model were associated with changes in network resting functional connectivity. A high degree of cortical-cortical correlations (when SWDs were present), but not in non-epileptic controls was observed. The strongest connectivity was seen between regions involved in seizures, mainly in the somatosensory and adjacent cortices. Resting inter-hemispheric cortical-cortical correlations were significantly higher in WAG/Rij rats compared to non-epileptic Wistar rats (Mishra et al., 2013). Microstructural changes in white matter pathways interconnecting the regions of seizure discharges were reported in two genetic models of absence epilepsy (WAG/Rij and GAERS) (Chahboune et al., 2009). In human patients with typical childhood absence epilepsy, abnormally increased resting connectivity between the two hemispheres, most evident in cortical areas, was also found (Bai et al., 2011). Stress and elevated level of corticosteroid hormones affect neuronal excitability in various brain regions (Joëls et al., 1995) and can increase susceptibility not only for convulsive seizures but also for non-convulsive absence seizures (Tolmacheva et al., 2012, Schridde and van Luijtelaar, 2004b). WAG/Rij rats showed elevated resting corticosterone concentration compared to age-matched Wistar rats, a larger and quicker rise after foot-shock stress than Wistars, and a decline to the lower than baseline corticosterone level 60 min after exposure to a stressor. The results suggest that the hypothalamic-pituitary-adrenal (HPA) axis in epileptic WAG/Rij rats is rather different from non-epileptic Wistar rats (Tolmacheva et al., 2012). These data indicate that stress reactivity of the HPA axis might be involved in the regulation of the genetic absence epilepsy. Because maternal care affects the development of HPA stress reactivity in the offspring (Liu et al., 1997), there is a reason to assume that larger HPA stress responses in adult WAG/Rij rats were programmed early in life by poor maternal care (see Section 3.5). More studies are needed to elucidate this issue further. Although various preclinical epilepsy models have shown increased seizure susceptibility in naive rodents after prenatal and early postnatal stress exposure, a causal relationship between stress and epileptogenesis in epileptic patients has not been fully resolved for a long time. However, subsequent studies have shown that early-life stress can be a seizure precipitant and a risk factor for epileptogenesis in humans as well (Novakova et al., 2013, van Campen et al., 2014), but only in people with a stress-sensitive type of epilepsy. Stress sensitivity is more common in children who experienced early-life stress. Children with stress-sensitive seizures also show an altered release of cortisol in response to stress (van Campen et al., 2015b). A negative correlation of cortisol level with global functional connectivity was found only in people with stress-sensitive seizures, not in those without stress-sensitivity of seizures (den Heijer et al., 2018). Based on the data that epilepsy is associated with enhanced functional connectivity, authors speculate that increased functional connectivity in epilepsy is somehow a protective mechanism against seizure generation, which fails in a stress-sensitive subgroup of subjects as cortisol level increases. Put in other words, a change in the brain’s functional connectivity induced by stress hormones can facilitate the generation of seizures. In stress-sensitive individuals, early-life stress can be a risk factor for childhood epilepsy as well (van Campen et al., 2012). HPA-related stress hormones can affect excitatory and inhibitory processes in the brain structures that are critically involved in seizure generation. Early-life stress might provoke vulnerability to seizure generation and epileptogenesis via altering glucocorticoids level (Kumar et al., 2007), HPA-axis (Joëls, 2009), membrane receptors, for instance, GABA (Reddy, 2013), NMDA and AMPA (Olney et al., 1991, Rogawski, 2013), neurogenesis (McCabe et al., 2001), brain structures connectivity (Nephew et al., 2017, Wang and Meng, 2016), monoaminergic brain systems (Matthews et al., 2001), dendritic spine morphology (Wang et al., 2013, Wong and Guo, 2013), and ion channels (Jones et al., 2011, Russo et al., 2016). Ion channels are a class of gene products that influence neuronal and network excitability. Channelopathies occur in epilepsy, leading to the disruption of neuronal and network function. The hyperpolarization-activated cyclic nucleotide-gated (HCN) ion channels, particularly the major isoform, HCN1, have been reported to accompany, and perhaps contribute to, the epileptogenic processes in many temporal lobe epilepsy (TLE) models (McClelland et al., 2011). HCN1 ion channels in the somatosensory cortex play a special role in the pathogenesis of absence epilepsy in the WAG/Rij and GAERS genetic models, and they are very sensitive to environmental manipulations (see 3.3, 3.4, 3.6, 4.2). Early-life stress-induced region-specific abnormalities in dendritic spine formation (Wong and Guo, 2013) may alter synaptic differentiation and lead to impaired synaptic function/plasticity and abnormal network connectivity, which could underlie the development of epilepsy and related comorbid behavioural abnormalities. In rodents, HPA axis formation, differentiation, and maturation of many brain structures, including cortex and hippocampus, synaptogenesis, neurotransmitter systems development as well as the establishment of connectivity between brain structures, take place during early postnatal life, and that is why this period of ontogeny is so susceptible to environmental influences. The favorable early environment promotes good self-regulation of the HPA axis later in life, but early-life adversity disrupts normal HPA axis development and “programs” poor HPA self-regulation, which plays an important role in the pathogenesis of epilepsy and its psychiatric comorbidities, depression, and anxiety in particular. Interestingly, corticotropin-releasing factor (CRF) and its type 1 receptor were shown to play a critical role in modulating adverse effects of early-life stress such as dendritic remodeling in the cortex and hippocampus and memory impairment (Martin and Wellman, 2011, Wang et al., 2013, Yang et al., 2015). The role of particular genes, including the CRF gene, has been demonstrated in triggering early adversity-associated pathological conditions such as anxiety and depression (Vaiserman and Koliada, 2017). Early-life stress may be a shared causal environmental factor for both epilepsy and psychiatric comorbidities often accompanying it. Dysregulation of HPA activity is one of the most commonly observed neuroendocrine symptoms of depression in humans (Holsboer, 2000). HPA dysregulation was found in different animal epilepsy models (Daniels et al., 1990, Mazarati et al., 2009, Szafarczyk et al., 1986, Tolmacheva et al., 2012) and human studies (Zobel et al., 2004), as well as in depressive disorders (Keller et al., 2017). Therefore, hyper(re)activity of the HPA axis, which develops in epilepsy, could be regarded as one of the possible mechanisms underlying the co-morbidity between epilepsy and depression. Put into other words, the dysregulation of the HPA system might be proposed as a common pathophysiological mechanism of epilepsy and depression co-morbidity (Mazarati et al., 2009). Recently it was found that the neuropeptide hormone ghrelin is altered in children with generalized epilepsies (Costa et al., 2022) and is associated with stress response (Meyer et al., 2014, Stark et al., 2016). Ghrelin regulates the HPA axis and associated stress-related mood disorders, such as fear, anxiety, and depression (Spencer et al., 2015). This suggests that ghrelin could serve as a common pathophysiological mechanism of epilepsy and its neuropsychiatric comorbidities. Epigenetic modifications, particularly DNA methylation, could be implicated as a molecular mechanism underlying the impact of early-life stress on epilepsy and its behavioural comorbidities. Accumulating evidence suggests that early-life stress can induce region-specific dysregulation of the DNA methylation pattern of multiple genes in the brain, which can lead to anxiety (Elliott et al., 2016) and depression-like behaviour (Zhang et al., 2019), and these changes in DNA methylation can be transmitted across generations (Murgatroyd et al., 2009, Roth et al., 2009). Interestingly, a substantial body of evidence also indicates that dysregulation of epigenetic mechanisms occurs in several human epilepsy syndromes (Kobow and Blümcke, 2018). DNA methylation has been highlighted as the methylation hypothesis of epileptogenesis (Kobow and Blümcke, 2012). Epigenetic mechanisms can influence the expression profile of candidate genes that persist in epilepsy. Epigenetic modifications can impact seizures and epilepsy in several ways (Lubin, 2012, Roopra et al., 2012). Firstly, seizure activity can be a result of histone acetylation-induced gene expression changes, including alterations in mRNA levels, for glutamate receptors 2 (GluR2 and GRIA2) and BDNF, the well-characterized epileptogenesis-related genes. For instance, histone acetyltransferase-mediated increases in histone acetylation levels at the promoter regions of the glutamate receptor 2 and BDNF genes have been found to correlate with these gene expression changes associated with seizures in an experimental animal TLE model (Huang et al., 2002). Histone acetylation is also involved in epileptogenesis in human epileptic disorders (Qureshi and Mehler, 2010), and may have a crucial role in the development of absence epilepsy and depression-like comorbidity in the WAG/Rij rat model (Citraro et al., 2020). Secondly, seizures can be a result of gene expression changes induced by alterations in the DNA methylation level. DNA methyltransferase enzymes 1 and/or 3a were shown to increase in the brain of patients with TLE (Zhu et al., 2012) and a rat model of TLE (Williams-Karnesky et al., 2013). The DNA methyltransferase (DNMT) enzyme DNMT3a in the prefrontal cortex and nucleus accumbens has also been shown to affect anxiety (Elliott et al., 2016) and depression-like behaviour (LaPlant et al., 2010). Ethosuximide treatment results in alterations in the expression of DNMT enzymes that catalyze DNA methylation, leading to reduction of epileptogenesis and behavioural comorbidity (elevated anxiety) in the GAERS model (Dezsi et al., 2013), indicating that the absence epilepsy and its behavioural comorbidity may share common epigenetic mechanisms. Thirdly, seizures and epileptogenesis might be mediated by transcription factors both in epileptic patients and epilepsy models. Repressor element-1 silencing transcription factor (REST) and neuronal restrictive silencer factor (NRSF) repress gene expression through dynamic recruitment of epigenetic complexes (Qureshi and Mehler, 2009). Of note, REST is involved in the regulation of multiple epileptogenesis-associated factors, including growth factors, neurotransmitter receptors, ion channels, circuit excitability, and neurogenesis (McClelland et al., 2014, McClelland et al., 2011, Roopra et al., 2012). Fourthly, methyl-CpG-binding protein 2 can regulate neuronal activity (Roopra et al., 2012). The expression of several ion channels has been identified to contribute to the aetiology of epilepsy. These channels are critical for electrical signaling between neurons and are responsible for the regulation of neuronal excitability. Dysregulation of ion channel expression is highly associated with epilepsy. Although the molecular mechanisms that underlie these changes in gene expression are not understood yet, it is known that many of these genes can be regulated by NRSF (McClelland et al., 2014, McClelland et al., 2011). Epileptogenesis causes the downregulation of genes that are involved in epilepsy, for instance, the HCN1 ion channel gene in both TLE (McClelland et al., 2011) and genetic absence epilepsy (Nishitani et al., 2019, Strauss et al., 2004) models. Dysregulation of NRSF seems to be associated with epilepsy. However, specific mechanisms are still unknown. There is evidence that HCN1 channelopathy derives from NRSF-mediated transcriptional repression that may contribute to epileptogenesis Thus, therapeutic interventions targeting NRSF to restore HCN1 gene expression can slow down the progression of epilepsy, as has been shown in a mouse model of TLE (McClelland et al., 2011). Whether targeting NRSF to restore HCN1 gene expression could slow down the development of genetic absence epilepsy in the WAG/Rij rat model remains to be investigated. Taken together, early-life stress can prime seizure occurrence and epileptogenesis. In addition, epigenetic modifications can be regarded as a shared pathogenic mechanism underlying the impact of early-life stress on epilepsy and its psychiatric comorbidities. The separation between offspring or children from their mothers has been seminal for understanding the development of psychopathology (Bowlby, 1951), and Bowlby’s ideas have brought about significant changes in perceptions of separations between children and their mothers. Later, maternal separation was introduced on a large scale in the biological and biopsychological literature. It is one of the most commonly used laboratory methods to manipulate and study early-life stress effects on the development of neurological and psychiatric disorders in adulthood. Maternal separation has been induced in several ways, ranging from a single 24 h separation to repeated episodes of separation lasting 3, 6, or 12 h. There was a considerable amount of variability between studies likely related to the differences in the maternal separation protocol due to the lack of standardization (Wang et al., 2020). For instance, the number of days for which the mother was separated from their pups varied in different experiments. A most common procedure is to keep the newborns together as a litter in their familiar environment and remove the mother. This procedure leaves the pups without care but in a familiar environment and the presence of their siblings. According to another procedure, the litter is transferred into a new environment in which pups are deprived not only of the mother but also of their familiar environment, the mother remains in the home cage. The former procedure is usually called “maternal separation” in contrast with the more stressful for pups “maternal deprivation”. Of interest, the maternal separation protocol, which is more stressful for mothers, leads to increased maternal motivation and maternal care after reunion (Bailoo et al., 2014), while maternal deprivation is more stressful for newborn pups, leading to more adverse and long-lasting consequences in later life (Zimmerberg and Sageser, 2011). Studies have shown that maternal separation of sufficient duration (typically 3 h/day during the first 2 postnatal weeks of life) increases anxiety- and depressive-like behaviour in rats during adulthood (Kambali et al., 2019, Lee et al., 2007, Matthews and Robbins, 2003). Neonatal maternal separation is considered as a model of human depression as evidenced by rodent (Vetulani, 2013) and primate (Sánchez et al., 2001) studies. However, other studies in both rats and mice have shown considerable variability of behavioural outcomes induced by neonatal maternal separation. Although most of the reports highlight the harmful effects, such as the increased risk for psychopathology, several studies showed no effects of maternal separation on anxiety- and depression-related phenotypes in different mouse strains (Millstein and Holmes, 2007) or even positive effects: reduced anxiety and improved cognitive ability (Savignac et al., 2011) or increased resilience to later-life stressful events (Santarelli et al., 2017). Inconsistency of the data could be related to differences in genetic background and protocol parameters used. Interestingly, in Wistar-Kyoto rats, known for their stress reactivity, anxiety, and depression-like phenotype, neonatal maternal separation (3 h/day from PND1 to PND14) induced anxiolytic and antidepressant-like behavioural effects in adult offspring. Positive behavioural effects of maternal separation were associated with DNA hypermethylation specifically in the hippocampus. Of note, enhancing DNA methylation in Wistar-Kyoto rats by using dietary methyl-donor supplementation improved anxiety/depression-like phenotype similar to neonatal maternal separation (McCoy et al., 2016). Negative behavioural effects of neonatal maternal separation are thought to be associated with abnormal HPA axis development (Sheng et al., 2021), persistent alterations in the monoaminergic and GABAergic brain systems (Arborelius and Eklund, 2007, Caldji et al., 2000), and epigenetic modifications of genes affected by early-life stress - corticotrophin-releasing hormone (Chen et al., 2012), glucocorticoid receptor (Park et al., 2017), BDNF (Park et al., 2018) and GABAR (Hsu et al., 2003), specifically in the hippocampus. In addition, another variable that should also be taken into account is the so-called “stress hyporesponsive period”, when the adrenocortical response to stress-inducing stimuli is still hypofunctional (the first week of life). This means that separation from the mother and/or siblings during the first postnatal week could have consequences different from those occurring when protocols with the separation in the second week of life are used. For instance, neonatal maternal separation during the first postnatal week was found to modify the expression of glucocorticoid receptors in the CA1 hippocampal subfield to promote increased secretion of corticosterone later in life, indicating an impairment of adrenocortical control and persistence of HPA-axis stimulation (Biagini et al., 1998). Neonatal maternal separation strongly modifies the stress response, a critical factor involved not only in the development of behavioural impairments such as anxiety and depression, but also in seizure induction. However, it should be noted that most attention in this area of investigation was paid to behavioural impairments induced by neonatal maternal separation, and very little research has been done on the effects of maternal separation on epileptogenesis and epilepsy. The effect of neonatal maternal separation on genetic absence epilepsy was studied in the WAG/Rij rat model. WAG/Rij dams were separated from their pups daily for 3 h from PND1 to PND21 to establish the effects of maternal separation on SWDs later in life. At the age of 4–5 months, rats separated from the mother showed 35 % fewer SWDs compared to untreated controls, with no effect on the mean duration of SWDs. The morphology of the SWDs was different in the separated from the mother group as well: reduced amplitude of the peaks of the SWDs, less energy in the 7–12 Hz band, next to a lower peak frequency (Schridde et al., 2006). In sharp contrast to this anti-absence action in adult WAG/Rij rats are the effects of maternal separation (3 h/day from PND2 to PND14) in Wistar rats: separation from the mother increased the vulnerability to limbic epilepsy (Kumar et al., 2011, Salzberg et al., 2007). Neonatal maternal separation in Wistar rats led to accelerated kindling rates in young adulthood, heightened corticosterone responses during and after kindling in females and a similar trend in males, and a reduced number of pyramidal cells in CA3 post-kindling, as well as significantly increased dentate granule cell neurogenesis in female rats subjected to maternal separation compared to their control group (early handling). The enhanced amygdala kindling rate after 3 h maternal separation at PND2-PND17 was confirmed in prepubertal rats at PND18 (Zhou et al., 2010). Kumar and coworkers concluded that these data showed that early life stress results in enduring enhancement of HPA axis responses to limbic seizures, with increased hippocampal CA3 cell loss and augmented neurogenesis, in a sex-dependent manner. This implicates important hippocampal candidate mechanisms through which early-life stress may promote vulnerability to limbic epileptogenesis in rats, as well as to human MTLE and its associated psychiatric disorders. Next, the Kumar et al. (2011) and Schridde et al. (2006) data once more illustrate the opposite effects of maternal separation in limbic versus absence epilepsy. How neonatal maternal separation may have pro-epileptic effects regarding TLE and at the same time an anti-absence epilepsy action is not quite clear. The hippocampus is not a part of the cortico-thalamo-cortical network in which SWDs are initiated and maintained (Tancredi et al., 2000). However, the hippocampus plays a modulatory role in the occurrence of SWDs, as intrahippocampal administration of GABA-mimetic drugs reduces the occurrence of SWDs (Tolmacheva and van Luijtelaar, 2007). A more excitable hippocampus, as obtained through maternal separation, might facilitate the occurrence of SWDs, similar to facilitating the hippocampal kindling rate. However, this mechanism is less likely, considering that maternal separation reduces SWDs instead of enhances SWDs. Schridde et al. (2006) proposed that changes within the cortico-thalamo-cortical system may be responsible for the maternal separation-induced reduction of SWDs. Considering that WAG/Rij rats have marked impairment in Ih in the somatosensory cortex, as has been established in pyramidal neurons, using RT-PCR, in situ hybridization, immunohistochemistry and Western blot (Strauss et al., 2004), and the crucial role of Ih in cortical and thalamic bursting (Gloor and Fariello, 1988, Lüttjohann and Pape, 2019, Pape, 1996, Pinault and O’Brien, 2005) and neural excitability (Poolos et al., 2002), the authors focused on Ih. Maternal separation caused an upregulation of the fast component of Ih and HCN1in the somatosensory cortex in the experimental group of WAG/Rij rats compared to the control group. No group differences in HCN2 and HCN4 proteins or changes in mRNA of any of the channel subunits were found. These data are in agreement with the results of a comparative study by Strauss et al. (2004), indicating that only cortical HCN1 channels seem to be critical for SWDs (Strauss et al., 2004). Therefore, it can be proposed that maternal separation rescues the genetic deficit in Ih channel functions in the somatosensory cortex, causing less cortical excitability and, as a result, a reduction in the number of SWDs. Kole et al. (2007) confirmed the important role of cortical HCN1 in WAG/Rij rats in a developmental study and showed that the changes in cortical Ih precede the onset of SWDs. Blumenfeld et al. (2008) not only confirmed the contribution of HCN1 in cortical excitability in WAG/Rij rats but also demonstrated that antiepileptogenesis by chronic and early treatment prevents the impairment of cortical Ih and age-dependent increase in SWDs. In all, the results of Schridde et al. (2006) provided the first evidence that relatively mild changes in the neonatal environment have a long-term impact on the absence seizures, Ih, and HCN1 and suggested that an increase in Ih and HCN1 is associated with absence seizure reduction. These findings demonstrate that genetically determined SWDs are quite sensitive to early interventions and that the evidence for cortical Ih as a controlling mechanism for SWDs seems to be strong. The direction of the maternal separation effects in WAG/Rij rats is a reduction in SWDs, while in TLE models the effects are opposite. This might point towards the role of GABA, since GABA also acts both as a pro-absence and anticonvulsant agent. Considering that dams display increased maternal care and attentiveness to pups upon reunion after separation (Bailoo et al., 2014) and that a heightened level of maternal care reduces SWDs in WAG/Rij offspring (Sarkisova and Gabova, 2018; see also Section 3.5), we cannot exclude the role of improved maternal care in anti-absence effects of neonatal maternal separation. The genetic background of the WAG/Rij rats may also contribute to this effect. Maternal care can program HPA axis development (Sheng et al., 2021) and thereby can dampen a hyper-activation of the HPA axis leading to the prevention of the development of adverse effects, which could be induced by neonatal maternal separation. Although the effect of maternal separation on comorbid depression has not been studied, it can be assumed that maternal separation may not only reduce SWDs but also improve depression-like behaviour in WAG/Rij rats. The fact that neonatal maternal separation reduced depression-like behaviour in innately depressive Wistar-Kyoto rats (McCoy et al., 2016) supports this assumption. However, further studies are needed to find out if this is the case in WAG/Rij rats. The developing brain is extremely sensitive to even minor environmental perturbations. Neonatal handling is a form of early-life impact which can result in long-term consequences. The procedure commonly involves removing the mother and then the rat pups from their cage, placing the pups in a small container with sawdust, and returning all of them after 15 min back to their cage and this is followed by the return of their mother (Caldji et al., 2000). However, in different studies, the duration of the handling procedure may be different (1, 3, or 15 min), as well as can be repeated a different number of times (for 10, 15, or 21 days). Despite this variability in the parameters of the procedure, the main effects of neonatal handling, such as long-term alterations in brain functions, reduced anxiety/emotionality, and reduced stress responses later in life are robust and well-reproducible (Raineki et al., 2014). The most common duration of the manipulation, 15 min, is much shorter than for maternal separation, typically 3 h. Although this manipulation is short, it may induce changes in the brain and behaviour that persist long into adulthood (Pryce et al., 2001). Repetitive brief handling in neonatal rats has been shown to cause decreased glucocorticoid responses to stress due to a permanent increase in glucocorticoid receptors density and binding in the hippocampus, a critical region for HPA regulation, especially for the negative-feedback inhibition of adrenocortical activity. Moreover, hippocampal cell loss and spatial memory impairments, which emerged with age in the non-handled rats, were almost absent in the handled rats, indicating that neonatal handling can retard the development of age-related pathological processes (Meaney et al., 1988). The molecular mechanisms underlying neonatal handling-induced behavioural and cognitive changes are not fully understood. However, it is assumed that alterations in GABA, the major inhibitory neurotransmitter, which regulates both behavioural and neuroendocrine responses to stress, may contribute to the effects of neonatal handling (Hsu et al., 2003). Abnormalities in GABAergic function have been observed in acquired and genetic models of epilepsy (Treiman, 2001). Acquired alterations or genetic defects in GABA receptor channels cause epilepsy (Chuang and Reddy, 2018). This means that neonatal handling which can modulate the GABAergic brain system may also modify acquired or genetic epilepsy. Neonatal handling-induced changes are reported not only in the GABAergic (Caldji et al., 2000) brain system, but also in monoaminergic (Durand et al., 1998, Papaioannou et al., 2002a, Papaioannou et al., 2002b), cholinergic (Pondiki et al., 2006), and opioid neurotransmitter systems (Ploj et al., 1999), all known to modulate epilepsies, including the frequency of occurrence of SWDs in WAG/Rij rats (van Luijtelaar and Zobeiri, 2014, Russo et al., 2016). Neonatal handling could modify the development of the HPA axis and, as a consequence, its response to different stressors. It has been shown that neonatal handling induces better regulation of HPA axis activity due to the increased negative feedback efficacy (Myers et al., 2012). When performed daily during the first 14 or 21 days of postnatal life, it causes in adult animals a reduced release and production of ACTH and corticosterone due to an enhanced glucocorticoid negative feedback sensitivity, meditated by an increased hippocampal glucocorticoid receptor mRNA expression. This leads to a reduced ACTH and corticosterone response to stressors later in life (Meaney et al., 1993, Raineki et al., 2014). Of interest, not only neonatal handling but also long-lasting (6 weeks) handling in adolescent rats may also produce beneficial effects. Improvements in learning and memory, including spatial memory impairments caused by neurodegeneration, and decreases in anxiety levels have been reported (Costa et al., 2012, Stara et al., 2018). Of note, handling for 7 days, 5 min per day in 60-days-old Wistar rats reduced time spent in the dark box as a measure of anxiety in the light-dark choice test (Aulich et al., 1974). One of the most frequently described effects of neonatal handling is a reduction in anxiety-like behaviour, which is most often established in the open field (more time spent in the center area), in the elevated plus, or zero mazes (more time spent in the open sections (Caldji et al., 2000, Durand et al., 1998, Meerlo et al., 1999, Río-Alamos et al., 2015, Río-Álamos et al., 2017). It is generally accepted that early postnatal handling has beneficial consequences, such as the ability to cope with stressors and improved adaptation to the environment. However, other behaviours may show negative effects of early handling. For instance, neonatal handling negatively affected neurocircuitry that supports social behaviour leading to deficits in social and play behaviours (Raineki et al., 2014). The effects of early neonatal handling on epileptogenesis and seizure susceptibility were most often studied regarding the assumption that early handling is stressful, for an overview see Jones et al. (2014). The effects of neonatal handling in comparison with standard husbandry or completely undisturbed group on epileptiform activity were rarely studied, and if studied, the focus was on hippocampal-related epilepsies and the role of the HPA-axis. And in that case, early handling was proconvulsive since it decreased seizure onset time in the lithium-pilocarpine model (Persinger et al., 2002). In a genetic absence epilepsy model, it has been found that neonatal handling for 15 min per day during PND1-PND22 reduced (35 %) the number of SWDs in adult WAG/Rij rats, while the mean duration of SWDs was not affected. The morphology of the SWDs was also changed as a consequence of neonatal handling: the power of the 7–12 Hz and beta (12.5–25 Hz), as well of the peak frequency, were reduced compared to the untreated control group, while the background EEG showed no group differences (Schridde et al., 2006). It is not easy to attribute the reduction in number and changes in the morphology of SWDs to a certain mechanism; several non-exclusive mechanisms might be proposed, such as the role of the monoamine system, GABA, for the HPA-axis via a reduction in GABAA receptor function and expression in the hippocampal dentate gyrus (Hsu et al., 2003). However, another possibility is considering that the morphological changes were reminiscent of SWDs seen in young WAG/Rij rats, indicating that the process of epileptogenesis is delayed and that changes in the cortico-thalamo-cortical pathways are involved or caused these changes. Schridde et al. (2006) also analyzed HCN channels, previously demonstrated to play a role in the increased focal excitability at the initiation site, the somatosensory cortex, in the WAG/Rij model (Strauss et al., 2004), and showed that in brain slices of neonatally handled rats the typical HCN response was more pronounced and faster in onset compared to a control group that was exposed to a single disturbance in the weaning period for cage maintenance. Importantly, the decreased HCN function is a marker for cortical epileptogenesis in this model, and this decrease was rescued by an antiepileptogenic treatment (Blumenfeld et al., 2008). The combined results allow us to suggest that neonatal handling delayed epileptogenesis, considering that less number of SWDs was initiated and that also the less mature morphology of the SWDs pointed in that direction, and that cortical focal Ih channels are a mediating factor in this process. In line with this role of HCN channels is that HCN channel blocker ZD7288 increased the input resistance, diminished the HCN typical voltage sag, and prevented the rebound depolarization, and this effect was again more pronounced in neonatally handled animals (Schridde et al., 2006). Current-density analyses showed h-channel availability in neonatally handled rats. In situ hybridization and Western blot analyses showed that HCN1 protein expression was increased in the somatosensory cortex, and not the mRNA expression. This implies that post-transcriptional or post-translational factors are playing a role, affecting the location or amount of HCN1 proteins. Other HCN subunits were not affected by neonatal handling. The electrophysiological and molecular evidence strongly suggests a selective increase of HCN1 subunits in layer V of the somatosensory cortex in neonatally handled WAG/Rij rats. Subsequent studies showed that the age-dependent increase in SWDs was accompanied or even caused by the age-dependent decrease in HCN1 channels, which preceded the increase in SWDs (Blumenfeld et al., 2008, Kole et al., 2007). It is therefore suggested that the reduction in the number of SWDs, as reported here, as well as the morphological changes (less mature SWDs were found) induced by neonatal handling, are caused by a smaller reduction in HCN channel functioning and by a diminished increase in cortical excitability. Epigenetic modifications of HCN1 channel gene expression might contribute to these effects (see Section 4.2). The seminal work of Meaney and coworkers showed that differences in maternal care after neonatal handling rather than the handling procedure itself may be the critical factor in inducing neuroplastic changes, among those that are involved in the reduced endocrinological and behavioural response to stressors later in life (Caldji et al., 2000, Hsu et al., 2003). Handling of the pups altered the behaviour of the mother: it increased the level of licking and grooming of pups (Garoflos et al., 2008, Pryce et al., 2001, Reis et al., 2014, Villescas et al., 1977) and did not change substantively the arched-back nursing (de Azevedo et al., 2010). Mothers of non-handled pups demonstrated stable maternal care, including licking/grooming behaviour. Of interest, in the tactile stimulation group, in which the mother was removed from the nest and the pups remained in their home cages and were stimulated with a brush for 10 min/day within the nest, no changes in maternal care were observed after her return to the nest (de Azevedo et al., 2010). This indicates that tactile stimulation, which mimics normal maternal care, prevents neonatal handling-induced increases in pup licking. Taken together, the findings allow us to assume that increased maternal care and the immediate effects of infantile tactile stimulation (see Section 4.1) following the return of pups to the nest could be considered as a mediating mechanism for the beneficial effects of neonatal handling on behaviour, as well as on epileptogenesis in a genetic absence epilepsy model. Therefore, maternal behaviour will be discussed in the next paragraph. Maternal care is the most relevant environmental factor influencing the later-life phenotype in offspring. Deficits in maternal care can induce epigenetic modifications in the offspring leading to neurologic and psychiatric diseases, including epilepsy and depression. Increased risk for epilepsy and psychiatric disorders in offspring may be due to genetic predisposition. However, maternal care has also been shown to be very important. Likely, a combination of genetic, epigenetic, and environmental factors could more accurately explain the link between maternal care and the development of epilepsy and psychiatric disorders in offspring. Maternal anxiety and depression may lead to unresponsive or inconsistent care by the mother toward the child leading to insecure attachment (Campbell et al., 2004) which has been linked to increased risk for anxiety and depression in the offspring (Brumariu and Kerns, 2010, Wan and Green, 2009). Clinical data suggest that reduced maternal care associated with maternal depression increases several times the risk of the development of depression and epilepsy in offspring (Asselmann et al., 2015, Ekinci et al., 2009, Sellers et al., 2013). Animal studies support these findings indicating that the early maternal environment is critical for the later-life phenotype of offspring (Champagne et al., 2003, Francis et al., 1999a, Weaver et al., 2004). So, in the rat, it has been shown that variations in maternal care, particularly in licking/grooming, influence the growth and survival of the offspring, as well as the development of an endocrine, emotional and cognitive response to stress (Champagne et al., 2003). Interestingly, individual differences in maternal behaviour can be transmitted from the mother to her female offspring (Champagne et al., 2003, Francis et al., 1999b). Depressive rat mothers, similarly to humans, exhibit reduced maternal care compared with non-depressive dams, as was found in Flinders Sensitive (Lavi-Avnon et al., 2008) and WAG/Rij (Sarkisova et al., 2017a) rat strains. Poor mothering behaviour can be also observed in the anxious BALB strain of mice (Prakash et al., 2006). Females of BALB mice performed less arched-back nursing and licking/grooming pups and have longer latencies to retrieve pups to the nest (Tarantino et al., 2011), similar to depressive WAG/Rij females genetically predisposed to absence epilepsy (Dobriakova et al., 2014, Sarkisova and Gabova, 2018). In mammals, the quality of early life is primarily dependent on maternal care. In rats and mice, maternal care is manifested in the form of arched-back nursing, non-arched-back nursing (non-nutritive contacts with the pups) in ‘blanket’ or passive posture, and licking/grooming. Arched-back nursing and licking-grooming pups are thought to be the most important types of maternal behaviour exhibited by most rodent species with a great variation between different strains (Champagne et al., 2003). It has been found that these two types of maternal care expressions critically influence the offspring’s later-life phenotype and shape their responsiveness to stress and their level of anxiety (Sakhai et al., 2013). However, recent studies have shown that non-arched-back nursing, which represents tactile (skin-to-skin) contact with pups, in addition to licking/grooming, is also an important or even more important component of maternal care compared with arched-back nursing (see also Section 4.1). This view is supported by the fact, that rearing by foster Wistar mothers with a high level of non-arched-back nursing normalized the absence epileptic phenotype in WAG/Rij rat offspring (Sarkisova et al., 2017a, Sarkisova and Gabova, 2018) and the schizophrenia-like phenotype in apomorphine-susceptible (APO-SUS) rat offspring (van Vugt et al., 2014), as was established in cross-fostering studies. The WAG/Rij mothers exhibited depression-like behavioural symptoms and a low level of the active type of maternal care compared with control Wistar dams irrespective of the specificity of their pups (own or foster): a longer latency to retrieve pups to the nest, a shorter duration of the tactile non-nutritive contacts with pups (non-arched-back nursing), a lesser number of licking/grooming episodes, and a longer duration of arched-back nursing in immobile/passive posture (Sarkisova and Gabova, 2018). Moreover, WAG/Rij dams did not show a preference for a pup-associated compartment in the place preference test (Sarkisova et al., 2017b), indicating that reduced maternal care in WAG/Rij mothers might be due to depression-associated deficits in pup-induced maternal reward. Similarly, the APO-SUS mothers provided less maternal care than their control, apomorphine-unsusceptible (APO-UNSUS) dams. Pups of APO-SUS mothers had a reduced body weight. The APO-SUS mothers were more involved in self-grooming, spent less time in contact with pups and, like the WAG/Rij mothers, exhibited less non-arched-back nursing. The APO-SUS rats had, besides the schizophrenia-like phenotype, also many SWDs (Cools and Peeters, 1992, De Bruin et al., 2000). Therefore, it can be assumed that, although the effects of cross-fostering on EEG and SWDs were not investigated, improved maternal care given by APO-UNSUS mothers might reduce the epileptic phenotype in APO-SUS rats similar to WAG/Rij rats. Further studies are needed to verify this assumption. The responsiveness to stress is regulated by glucocorticoids and glucocorticoid receptors. High levels of circulating glucocorticoids increase the stress response, while lower levels attenuate it. Conversely, high levels of glucocorticoid receptors in the forebrain, in particular in the hippocampus, provide negative feedback that reduces the production of glucocorticoids and thereby dampens the stress response (Seckl, 2007). Interestingly, offspring of mothers with a high level of arched-back nursing and licking-grooming show increased glucocorticoid receptors expression and reduced reactivity to stress, whereas offspring of mothers with a low level of arched-back nursing and licking-grooming demonstrate decreased glucocorticoid receptors expression and increased stress reactivity (Liu et al., 1997). Glucocorticoids act via mineralocorticoid and glucocorticoid receptors, the first of them is involved in stress response onset but the second one participates in response termination and is essential for stress-coping strategy (active or passive). The imbalance between mineralocorticoid and glucocorticoid receptors is thought to be associated with stress vulnerability leading to the development of pathologic phenotypes (Franklin et al., 2012). Put in other words, a high level of active maternal behaviours, such as licking/grooming and arched-back nursing, has beneficial effects on the later-life phenotypes of offspring, but a low level can lead to stress vulnerability, depressive-like behaviour, anxiety, and altered cognitive and social behaviours (Myers-Schulz and Koenigs, 2012). Similarly in humans, maternal attachment to, the favorable environment in childhood predisposes individuals to stress resilience (Jaffee, 2007), while maternal neglect, physical maltreatment, and abuse or traumatic early-life events increase the risk for affective disorders later in life (Dietz et al., 2011, Hulme, 2011). Changes in maternal care can occur naturally due to individual variability in maternal care but can also be produced experimentally using specific manipulations in rodents. One of the ways to change maternal care to investigate the contribution of maternal care to seizure susceptibility and epileptogenesis is a cross-fostering procedure. The procedure of cross-fostering includes the removal of newborn pups from their biological mother and placing them with a foster mother with a different from the biological mother’s manifestation of maternal behaviour, commonly with a higher level of expression. To control whether a cross-fostering procedure per se can cause its own effect, an in-fostering procedure is usually used (pups are cross-fostered to foster mothers of the same strain with approximately the same expression of maternal care). Of interest, the cross-fostering approach was effective in improving depressive phenotype in Flinders Sensitive line rats fostered by Flinders Resistant dams with a higher level of arched-back nursing and licking/grooming (Malkesman et al., 2008). Anxious BALB mice fostered by non-anxious B6 mothers have been reported to express decreased anxiety levels in the elevated plus-maze test (Priebe et al., 2005). Cross-fostering was also found can ameliorate the expression of genetically determined pathologies, such as arterial hypertension in NISAG rats (Yakobson et al., 2001), catalepsy in genetic cataleptic GK rats (Kolpakov et al., 2002), and high anxiety in high responders to novelty (bHR) selection line (Cohen et al., 2015). However, cross-fostering did not exert a substantial effect on the expression of pathologic phenotypes, such as hyperactivity in spontaneously hypertensive (SHR) rats (Howells et al., 2009) and in hyperactive (High-Active) rats (Majdak et al., 2016), in two animal models of ADHD. Females of seizure-prone EL mice exhibit a reduced level of maternal care. They are slower than the control to initiate pup retrieval and spend less time in arched-back nursing but longer time in non-maternal behaviours, such as exploration and self-grooming (Bond et al., 2003). Cross-fostering of genetically susceptible to seizures El pups to seizure-resistant CD-1 mothers with a higher level of maternal care delayed seizure onset and seizure frequency, indicating that the maternal environment plays an important role in shaping the adult seizure phenotype (Leussis and Heinrichs, 2009). Paradoxically, El pups reared in the biparental environment, in which they received more parental care from both biological parents compared with El pups raised by only the El mothers, showed earlier development of seizures and increased seizure susceptibility later in life (Orefice and Heinrichs, 2008). Results indicate that the presence of a second care provider in addition to the dam constitutes a form of stressor exposure in El pups and, as a consequence, accelerates the development of seizures in genetically susceptible offspring. Cross-fostering of WAG/Rij pups genetically predisposed to absence epilepsy with comorbid depression to control Wistar mothers with a high level of maternal care showed that improvement of early caregiving environment can exert disease-modifying effects on epileptogenesis and behavioural comorbidities in genetic absence epilepsy. WAG/Rij offspring reared by Wistar dams with a high level of non-arched-back nursing and licking/grooming exhibited less and shorter SWDs and reduced depression-like behaviour (immobility in the forced swimming test) in adulthood compared with age-matched WAG/Rij offspring reared by their own or foster WAG/Rij mothers with a low level of maternal care. Moreover, a high level of maternal care of foster Wistar mothers decelerated the absence epilepsy appearance and its progression in WAG/Rij offspring, indicating an anti-epileptogenic effect. In 30% of adult (7–8 month old) WAG/Rij offspring reared by foster Wistar mothers no typical well-developed (mature) SWDs were found. Only immature SWDs, which were similar to immature SWDs commonly recorded in young (2–3 month old) WAG/Rij rats, were observed. Of note, adult WAG/Rij rats, adopted by Wistar dams, were behaviourally undistinguishable from aged-matched normal Wistar rats reared by their own or foster Wistar mothers, indicating the absence of behavioural symptoms of comorbid depression. Interestingly, the adoption by WAG/Rij dams with a low level of maternal care did not change EEG and behaviour in Wistar offspring (Sarkisova et al., 2017a, Sarkisova and Gabova, 2018). Results of these studies suggest that early-life environment can interact with a genetic predisposition to shape later-life seizure phenotype and associated behavioural comorbidities in the offspring. Disease-modifying effects of improved maternal care on genetic absence epilepsy and comorbid depression persisted into adulthood suggesting a possible role of epigenetic mechanisms, such as DNA methylation, which affects gene expression. Mounting evidence suggests that maternal care leads to epigenetic modifications of gene expression in the offspring. So, adult rat offspring exposed to poor maternal care early in development exhibited increased methylation in the promoter region of the hippocampal glucocorticoid receptor gene (GR 17) that resulted in decreased gene expression and enhanced response to stress. Conversely, a high level of maternal care was associated with decreased promoter methylation of the hippocampal GR 17 gene and increased gene expression (Weaver et al., 2004). Significantly elevated levels of methylation were detected in the estrogen receptor alpha gene promoter region in adult offspring of low licking/grooming dams compared with offspring of high licking/grooming dams, and cross-fostering reversed this effect (Champagne et al., 2006). Epigenetic modulation of gene expression has been implicated in the long-lasting impact of positive caregiver experiences on the offspring’s adult phenotype, namely, on stress reactivity and maternal behaviour (Weaver et al., 2004). It has been shown that gene expression differences induced by distinct epigenetic (specifically DNA hypomethylation) patterning in the amygdala might underlie downstream behavioural differences such as anxiety and depression-like phenotype (McCoy et al., 2017). Cross-fostering to foster mothers with a different style of maternal care shifted developmental gene expression in the amygdala (not hippocampus) and reduced adult anxiety and depression-like behaviours (Cohen et al., 2015). It might be hypothesized that a high level of maternal care of foster Wistar mothers counteracts the manifestation of pathologic phenotype in adult WAG/Rij offspring by changes in the activity of DNA methyltransferases (DNMTs), enzymes that catalyze DNA methylation, leading to epigenetic modifications in the expression of genes pathogenetically relevant for absence epilepsy and comorbid depression in certain brain structures. In favor of this assumption, there is evidence that a maternal methyl-enriched diet, which impacts DNA methylation, increased DNMT1 and HCN1 ion channel gene expression in the somatosensory cortex responsible for the generation of SWDs, and reduced absence seizures and depression-like comorbidity in adult offspring of WAG/Rij rats (see Section 4.2). We can also assume that changes in DNA methylation induced by a high level of maternal care could prevent the age-related development of the mesolimbic DA deficiency responsible for the manifestation of behavioural symptoms of comorbid depression in adult offspring of WAG/Rij rats. A reduced DAergic tone has been suggested to be a neurochemical mechanism of comorbid depression-like behaviour in WAG/Rij rats (Sarkisova et al., 2013, Sarkisova et al., 2014). Moreover, the absence epileptogenesis could also be associated with a reduced DAergic tone. Decreased tonic release of DA has been suggested to act as a facilitating and self-sustaining factor, which may lead to increased excitability of the somatosensory cortex (Cavarec et al., 2019) and thereby contribute to the generation of SWDs. Interestingly, neurochemical alterations of the DAergic brain system begin to appear before the occurrence of SWDs in WAG/Rij rats (Sarkisova et al., 2014). Maternal care was found to affect the development of the midbrain DA pathway. Tyrosine hydroxylase (TH) immunoreactivity in the ventral tegmental area was elevated by PND6 in response to maternal licking/grooming, and this effect was sustained into adulthood. The differences in the TH cell counts in the ventral tegmental area associated with different levels of postnatal maternal care suggest that the developing DAergic brain system is shaped by maternal care. Maternal licking/grooming altered the expression of genes critical for midbrain DA neuron differentiation and maintenance. Moreover, the offspring of mothers with a high level of licking/grooming had elevated DA receptor mRNA in the nucleus accumbens. TH gene DNA methylation increased with age but did not vary as a function of maternal licking/grooming (Peña et al., 2014). Whether the development of the mesolimbic DAergic brain system, in particular TH gene expression, is shaped by maternal care in the WAG/Rij rat model of absence epilepsy with comorbid depression-like behaviour remains to be established. Due to epigenetic mechanisms, variations in postnatal maternal care can predispose to or, on the contrary, prevent offspring from developing later-life pathologies, including the absence epilepsy and psychiatric comorbidities. Taken together, results suggest that maternal care can make a significant contribution to the trajectory of brain development and thereby substantively impact the adult phenotype of the offspring, even in the case of genetically determined pathology, such as the absence epilepsy and comorbid depression. Donald O. Hebb, dated back to 1947, observed that rats reared in an enriched environment performed generally better on behavioural tasks than those reared under standard laboratory conditions. This observation allowed him to formulate the scientific concept of an ‘enriched environment’. In animal studies, environmental enrichment refers to housing conditions designed to enhance sensory, motor, social, and cognitive stimulation, compared to control conditions. ‘Environmental enrichment’ implies that the interaction of various factors, not a single contributing factor, is an essential element of housing conditions (van Praag et al., 2000). Cage enrichment has a long tradition in inducing neuroplastic changes and is commonly applied in rats post-weaning for several weeks, 24 h/day. It is very well established that housing condition has long-lasting and profound effects across the lifespan on a variety of brain-behavior-related variables. Living in an enriched environment stimulates neurogenesis, increases the expression of neurotrophic factors, modifies brain structure and circuitry, improves cognitive functions, favors alterations in brain chemistry, and may protect against seizures or against the effects of seizures (Kempermann et al., 2002, Rosenzweig and Bennett, 1972, van Praag et al., 2000, Young et al., 1999). The seminal review (Dhanushkodi and Shetty, 2008) contained a small section regarding the effects of enrichment strategies for functional recoveries in TLE. The section was small due to the limited literature on this topic. They concluded based on two studies that animals reared in environmentally enriched conditions show a decreased susceptibility to seizures and hippocampal degeneration when acutely challenged with kainic acid (Young et al., 1999) or showed an increased threshold for amygdala kindling, also after prior housing in an enriched environment (Auvergne et al., 2002). A third study examined whether exposure to an enriched environment after the induction of acute seizures or SE is efficacious for preventing chronic epilepsy. Rutten et al. (2002) used the lithium-pilocarpine SE model in 20-day-old rats, but the effect of post-SE 30-day enriched housing on seizure prevention was not conclusive, perhaps due to the young age of the animals (Rutten et al., 2002). We identified a few more studies, all in TLE models. In the kainic acid seizure model, 20–25 days old animals were used, and the effects of early life environment on seizure-induced behavioral deficits, neuronal injury, and the inflammatory reaction in young rats were investigated. Two rearing conditions, maternal separation for 3 h daily during the first 2 weeks of their lives followed by single housing versus maternal care in an enriched environment for 7–10 days, followed by group housing in an enriched environment, were compared. A significant reduction in DNA fragmentation and microglial activation in the enriched compared to maternally separated animals. These results suggested that a nurturing early enriched environment can enhance the ability of the developing brain to recover from seizures and provide a buffer against their damaging effects (Kazl et al., 2009). The same group of authors showed earlier that environmental enrichment may reverse the adverse effects of early-life seizures on exploratory behaviour and the expression of genes involved in synaptic plasticity (Koh et al., 2005). Of note, not only kainic acid-induced seizures but also depression-like behaviour in the forced swimming test was reduced in juvenile rats reared in an enriched environment. Changes in 5-HT receptor gene expression were paralleled by decreased mobility in the forced swimming test, but environmental enrichment reversed both depression-like behaviour and gene expression. This means that seizures lead to increased susceptibility to depression through transcriptional regulation, while environmental enrichment can interact with gene expression to influence the behavioural comorbidity in epilepsy (Koh et al., 2007). Vrinda et al. (2017) used the lithium pilocarpine SE model in young adult Wistar rats. 6 weeks after SE, rats have been housed 6 h/day in an enriched cage or their regular home cages. Depression-like behaviour, anxiety, spatial learning, and memory were assessed using the sucrose preference test, elevated plus maze, and Morris water maze, respectively. Enriched housing significantly reduced seizure episodes and seizure duration in epileptic rats, next to normalization of hippocampal delta and theta power. In addition, environmental enrichment alleviated depression-like behaviour and hyperactivity. However, environmental enrichment neither ameliorated epilepsy-induced spatial learning and memory deficits nor restored cell density in the hippocampal CA1 region. In another study, it has been shown that environmental enrichment significantly improves hippocampal neurogenesis (increases cell proliferation and survival, extends the apical dendrites), decreases long-term seizure activity, and improves cognitive impairments in adult rats after SE (Zhang et al. (2015)). Positive effects of environmental enrichment were found in other epilepsy models, such as the AY-9944 mouse model of childhood atypical absence epilepsy (Stewart et al., 2012) and the Q54 transgenic mouse model of TLE (Manno et al., 2011). Enriched Q54 mice displayed a reduced frequency of epileptic discharges and reduced hippocampal damage (Manno et al., 2011). Another study demonstrated that AY-9944 mice from enriched housing conditions exhibited less behavioural hyperactivity and anxiety, improved olfactory recognition and spatial learning, but no significant reduction in the number of epileptic discharges in comparison with their non-enriched cohorts (Stewart et al., 2012). Environmental enrichment may facilitate amygdala kindling, but reduce kindling-induced anxiety (Young et al., 2004), indicating the opposite effects on epileptic seizures and behavioural comorbidity. On the other hand, in a genetic model, El mice, enriched housing from PND21 to PND 49, produced a 100 % decrease in seizure susceptibility relative to El controls (Korbey et al., 2008). Thus, the vast majority of the evidence shows positive effects of a post-weaning enriched environment on seizure characteristics in TLE models, both regarding the reduction of seizure susceptibility or inhibition of epileptogenesis. Moreover, there is evidence that the positive effects of environmental enrichment are not restricted to only the post-weaning SE models, or to the manipulations in the silent period, but that also later in life environmental effects can be found. In sharp contrast are the outcomes of enriched housing in the WAG/Rij rat model of absence epilepsy (Schridde and van Luijtelaar, 2005, Schridde and van Luijtelaar, 2004a). The enriched/impoverished housing conditions resembled those used by Rosenzweig and Bennett (Rosenzweig and Bennett, 1972), comprising a pool of weekly changed objects and providing social and inanimate enrichment in a large cage with 8–10 animals per cage, whereas impoverished housing implied a singly housed rat in a standard colony cage. ACI rats were used as a control: they have very few SWDs and are mainly of type II. For the differences between type I and type II SWDs see Midzianovskaia et al. (2001) and van Luijtelaar and Coenen (1986). Half of the rats of each strain (group size n = 20) were housed from post-weaning day 30 to PND90 in the enriched environment, and the other half of the animals were housed in impoverished conditions. EEG recording showed, at PND90, that enriched housing increased the mean duration of the SWDs type I (the classical SWDs), and the number of type II SWDs. Then, half of each group was transferred to the other environment and stayed there for another 60 days, and half of the rats remained in their original housing condition, also for 60 more days. EEG recordings after the second housing period revealed that again the mean duration of type I was increased in the groups of WAG/Rij rats that have spent two or four months in the enriched conditions, and again the number of type II SWDs was enhanced in the WAG/Rij’s that were housed during the last 60 days in the enriched environment (Schridde and van Luijtelaar, 2005). The results are striking because of three reasons: first, the effects are opposite to the literature on the TLE models that found a positive effect of an enriched environment, here a negative, implying an increase in the mean duration (type I) and number (type II) of SWDs. Second, the effects on the number and mean duration of either type I or type II was different. The increase in the mean duration of SWDs might point towards a role of the thalamic reticular nucleus (RTN), a structure involved in determining the duration of SWDs (Lüttjohann and Van Luijtelaar, 2015, Sohal et al., 2000). Third, the lack of an increase in the number of types I SWDs excludes the role of the somatosensory cortex, the trigger zone of SWDs (Meeren et al., 2002). The number of type II SWDs was increased, type II has a more occipital and localized spatial distribution and is, therefore, less generalized compared to type I SWDs, however, the source of this type of SWDs is not known. Therefore, but not based on strong evidence, it is thought that the RTN is a primary candidate for the effects of enriched housing post-weaning on SWDs. The effects of enriched housing were also studied in the GAERS model of absence epilepsy (Dezsi et al., 2016): cage enrichment started immediately the following weaning at PND21, the end of week 3. At weeks 9, 10, 11, 14, and 20, the EEG was recorded and the rats were tested in the open field. Enrichment delayed the onset of epilepsy, as evidenced by a reduced proportion of GAERS who had developed epilepsy by 9 weeks of age. Moreover, recording sessions over weeks showed that enriched GAERS had a lower SWD incidence, and a shorter mean duration compared with standard housed controls. Enrichment also reduced the indices of anxiety as obtained from a brightly lit open field (increased the number of entries into the inner area, as well as the total time spent in the inner area). In addition, it was investigated whether environmental enrichment could improve disease outcomes in adult symptomatic GAERS when absence epilepsy had already developed. Six weeks of housing in an enriched environment reduced the incidence of SWDs, not the mean duration. The authors established as well that the effects of environmental enrichment on SWDs and anxiety were genetically transmitted through the male germ line into the next generation, which also benefitted from the enriched experience of the father. Reduced CRH mRNA expression was associated with these phenotypic improvements, but this was not due to changes in DNA methylation (Dezsi et al., 2016). The reduction in SWDs and CRH, as found in this study, agrees with that cortisol enhances SWDs in the WAG/Rij model (Schridde and van Luijtelaar, 2004b). The reasons for the different effects of enrichment in the two genetic absence models are not immediately clear, but there were some differences between the experiment in GAERS and the WAG/Rij model. The enriched cages were much smaller in the GAERS study and the number of rats in the enriched cage was not mentioned, next the GAERS’s impoverished condition contained two rats, in the WAG/Rij study impoverished housing implied single housing, while the WAG/Rij enrichment cages, measuring 75 × 150 × 80 cm contained 8–10 rats. Therefore, the reasons for the results at variance are difficult to pinpoint considering a large number of differences. It clear is as well, that there are some differences between GAERS and WAG/Rij rats regarding the epileptic phenotype, for a direct comparison see (Akman et al., 2010) regarding the number of SWDs per hour, the frequency of the SWDs, the age of onset, and the genes involved. Epilepsy in both models is genetically determined but driven by different chromosomal locations in the WAG/Rij strain compared to GAERS (Gauguier et al., 2004, Rudolf et al., 2004), and single nucleotide (G to C) mutation in the Cav 3.2 T-type calcium channel gene (Cacna1h) was found in GAERS and not in rats of the WAG/Rij strain (Powell et al., 2009). It is therefore rather likely that not only differences between the type of enrichment and impoverishment, but also different genetic causes are interacting with the environmental factors in contrasting manners to differentially alter SWD characteristics. In general, inconsistency of data concerning the effects of environmental enrichment on epileptogenesis and psychiatric comorbidities may be explained not only by differences in epilepsy models but also by differences in experimental protocols used (Harland and Dalrymple-Alford, 2020). Possibly due to differences in experimental procedures environmental enrichment can not only reduce seizures, but also exaggerate them, and not only absence seizures, as in the WAG/Rij model, but also other types of seizures, for instance, seizures induced by PTZ kindling in Wistar rats. Environmental enrichment improved learning and memory in the Morris water maze test, but lead to the exaggeration of PTZ-induced seizures (Keloglan, 2016). One of the reasons for the seizure aggravation effects of enriched housing could be the stress caused by the change from the familiar and safe environment to a new unfamiliar one enriched with a variety of multimodal stimuli. Multi-sensory stimulation or intense and repeated stimulation of a single modality may be stressful for animals and can lead to seizure aggravation. Stress-induced seizure aggravation effects in epilepsy models are well known (Joëls, 2009), including WAG/Rij rats (Tolmacheva et al., 2012). There is a single paper in which it was demonstrated that exposure to intense 20–40 strobe trains per day for 3 days in adult Sprague-Dawley rats changes the common response to strobe trains in 34/36 rats (Uhlrich et al., 2005). Over the stimulation sessions, a high-amplitude spike-wave response developed not seen before the onset of stimulation, and this spike-wave response emerged fully by the third day of photic exposure, first at the occipital cortex, later it spread to the frontal cortex, and by the end of treatment, the oscillations were mimicking SWDs. The results indicate that visual stimulation, by itself, can induce in adult rats an enduring sensitization of visual response with epileptiform characteristics. The site or origin of these SWD-like oscillations, also sensitive to ETX, is topographically different from the classical SWDs in WAG/Rij and GAERS, which have an initiation site at the somatosensory cortex (Meeren et al., 2002, Polack et al., 2007). Here the visual cortex, interconnected with the visual thalamus seems to be the excitable area and it might be the initiation site of SWDs. Next, the visual pathways between the visual cortex, and the visual thalamus, including the higher-order nuclei, might be the primary circuit for the visual stimulation-induced SWDs. The data, as obtained by Uhlrich, raise the question of whether intense stimulation of other sensory modalities might have the same proepileptic effect, and or whether this is only in epilepsy-prone subjects. Although Sprague-Dawley (SD) rats are not considered as a genetic absence model, there are indications that SD rats are not devoid of SWDs (Komoltsev et al., 2021, Pearce et al., 2014, Willoughby and Mackenzie, 1992). Therefore, it might be possible that the vulnerability to cortico-thalamo-cortical oscillations primes SD rats for SWDs in case they are challenged and exposed to repetitive and intense visual stimulation. Whether similar intense stimulation in young GAERS or WAG/Rij rats may have similar neuroplastic effects remains to be investigated. Sensory experience during early postnatal life modulates cortical development, including morphological and functional characteristics of neurons. Brain circuits are particularly sensitive to alterations produced by sensory stimuli during a certain time window called a critical period (Reha et al., 2020). Deprivation of sensory inputs during the critical period, when functional and structural characteristics of cortical neurons are most susceptible to alterations, can induce substantial impairments of axonal and dendritic morphology, and synaptic connectivity of neural circuits (Briner et al., 2010, Lee et al., 2009). The main regulator of the experience-dependent activity of sensory systems is the balance of excitation and inhibition. In some cases, sensory deprivation may potentiate inhibition, which can suppress responses to deprived sensory inputs. In other cases, sensory deprivation may weaken the inhibition, leading to the restoration of sensory responsiveness (House et al., 2011). The time course of experience-dependent sensory development is specific for each sensory system. The critical period of the somatosensory system precedes the critical periods of the visual and auditory systems (Li et al., 2009). However, postnatal deprivation of sensory input in one modality can result in compensatory cross-modal plasticity that increases activity in the remaining intact senses (Dooley and Krubitzer, 2019, Merabet and Pascual-Leone, 2010, Mezzera and López-Bendito, 2016, Rauschecker, 1995). Cross-modal plasticity implies not only physiological changes such as the increased activity of the non-deprived sensory system but also the recruitment of the deprived area for compensatory senses (Voss and Zatorre, 2012). An important part of normal development is also proper sensory integration: the environment is experienced as a combination of a variety of within a single sensory modality stimulus features and between different sensory modalities. The surrounding of an organism is complex and comprises sensory inputs from several senses at the same time. Only very few day-to-day life events and stimuli present themselves as unimodal, rather as multisensory experiences, deriving from a combination of information acquired through several different sensory modalities. The brain has to integrate multisensory information to provide a complete and coherent picture of events to allow a proper and adapted behavioural response. Deprivation of one sensory modality may therefore also affect proper sensory integration (Dionne-Dostie et al., 2015). Insufficient or abnormal sensory experience early in life could lead to sensory integration dysfunction when multisensory integration is not adequately processed to provide an appropriate response to the environmental impacts. Sensory processing abnormalities (increased or decreased sensory sensitivity) were found in several neurological and neuropsychiatric disorders including schizophrenia, bipolar disorder, ASD, ADHD, and depression (Harrison et al., 2019). Altered sensory sensitivity can also be linked to seizure susceptibility. Impairments of sensory sensitivity and sensory modulation (behavioural responses to regulate sensory input, e.g., to reduce or prevent exposure to stimuli) were found in childhood epilepsy. Sensory modulation disorders were reported in 49 % of the 158 children with epilepsy. Increased behavioural responses in epileptic children were associated with sensory “sensitivity” and “sensory avoidance”, but not with “sensory seeking”. Comorbidity of childhood epilepsy with ASD and ADHD was associated with more severe sensory modulation dysfunction, although 27 % of children with epilepsy without comorbid disorders also showed a sensory processing impairment (van Campen et al., 2015a). There is no evidence of sensory impairment in childhood absence epilepsy. However, a hypersensitivity to a gentle tactile touch in the WAG/Rij rat pups with a genetic predisposition to absence epilepsy was found (see also Section 4.1). Most of the WAG/Rij pups (64 %) compared with age-matched Wistar pups (23 %) showed an active avoidance response to tactile touch (Malyshev et al., 2014). In both pre-symptomatic and symptomatic WAG/Rij rats an increased sensitivity to painful stimuli as measured in four different and commonly used pain tests was found, indicating an early deviant somatosensory system (De Caro et al., 2020). This also demonstrates that the changes in the somatosensory system precede the onset of SWDs and can be considered as a causative contributing factor to their development. The somatosensory system, especially the tactile sensations from whiskers, plays a key role in the perception of the environment in rodents. Layer IV of the primary somatosensory cortex (S1), called the barrel cortex contains a topographic representation of each facial whisker. The whisker map in the barrel cortex is established during the critical period that extends along the first postnatal week when activity-dependent development and balance of excitatory and inhibitory inputs are formed (Che et al., 2018). Neonatal sensory deprivation in rodents induces long-lasting changes in the structure and function of the somatosensory system. For instance, neonatal whisker trimming leads to enlarged excitatory and weakened inhibitory receptive fields in layer IV barrel neurons (Shoykhet et al., 2005), as well as to morphological changes in the spiny stellate neurons (larger dendritic spines and greater spine density) and behavioural alterations (higher explorative activity and more frequent social interactions) in adult rats (Lee et al., 2009). Neonatal deprivation in whisker-dependent tactile perception impaired fear/anxiety-related emotional systems of the amygdala (greater stress-induced c-fos expression) and social behaviours in the social preference and social dominance tests in mice as adults (Soumiya et al., 2016). There is convincing evidence that primary sensory cortices are anatomically interconnected. Anatomical studies in rodents revealed direct cortico-cortical projections from S1 and the auditory cortex (A1) to the visual cortex (V1). Conversely, V1 was shown to connect to S1 and only weakly to A1 (Henschke et al., 2015). Such anatomical connections indicate the presence of a functional multimodal interplay between the primary sensory cortices. These functional multimodal connections between cortical areas can cause a perceptual improvement of one sensory modality when the other sensory modality is lost. For example, blind individuals compensate the lack of visual inputs by responding to somatosensory or auditory inputs with improved sensitivity. In rats, which were visually deprived at postnatal 4 weeks (PND26-PND30), but not later (PND58-PND66), increased sensitivity of the somatosensory system involved in the whisker-dependent tactile perception was found (Abe and Yawo, 2018). It has been well-documented that the somatosensory cortex plays a crucial role in the pathogenesis of absence seizures in genetic rodent models (Ding and Gallagher, 2016, Meeren et al., 2002, Polack et al., 2007). Of interest, absence seizures models, WAG/Rij rats (Meeren et al., 2002), and GAERS (Polack et al., 2007), are albinos, and, as a result, they may have visual impairments which are characteristic of albinism. Genetic epileptic mice modeling juvenile myoclonic epilepsy are not albino, suggesting that there are other causative factors for the seizures in this model (Ding and Gallagher, 2016). Both GAERS and WAG/Rij strains originate from Wistar rats. Wistar rats are albinos: they have a lower number of rod photoreceptors (Donatien and Jeffery, 2002) and night blindness (retinopathy), starting around three months of age and getting worse in the next 3 months (Lai et al., 1975). One of the symptoms is the slow adaptation to darkness (Behn et al., 2003). Visual electrophysiological studies, both electroretinography (ERG) and evoked potentials, comparing albino (Wistar) and pigmented (Long-Evans) rats at the ages of 1.5, 4, 7, and 10 months showed markedly decreased ERG b-wave amplitudes in Wistar rats, as well as tendencies for reduced amplitudes of the VEP (visual evoked potential). Markers for post-receptor processing revealed impairments in Wistar rats as well (Heiduschka and Schraermeyer, 2008). Therefore, it can be safely assumed that albino rats have problems with their visual system and, in particular, with seeing in the dark, during their behaviourally active period. As we have already mentioned, when a sensory modality is lost or impaired, as in blindness or deafness, the adaptive reorganization will take place, associated with neuroplasticity affecting both the sensory modality that is impaired, in our case the visual one, and those that remain intact: sensory loss of one modality has striking effects on the development and function of the remaining modalities (Chabot et al., 2008, Merabet and Pascual-Leone, 2010). Problems with the visual system in albino rats are likely to cause rearrangements in other sensory systems: in rodents, the collection of information from the environment is much more dependent on their whisking touch system than in rats with normal vision. The trigeminal pathway in albino rats is stimulated by the animals using this system, consisting of the micro (short and thin hairs around the nose tip), and macro - vibrissae (long stiff mystacial hairs) projecting via the brainstem and ventro-posterior medial (VPM) nucleus of the thalamus to the cortical S1 (Adibi, 2019). Interestingly, whisking consists of rhythmic cyclic vibrissae sweeping actions, consisting of repetitive forward (protraction) and backward (retraction) movements at an average frequency of about 8 Hz (Adibi, 2019, Semba et al., 1980), the same frequency as can be found in the SWDs in thalamus and cortex, accompanying the absence seizures. There is evidence from animal studies that even partial afferent signal loss leads to paroxysmal cortical activity, including SWDs that can progress to electrographic seizures (Nita et al., 2006, Topolnik et al., 2003). Given this, it is perhaps not surprising that highly excitable cells were found in the S1 in GAERS (Polack et al., 2007), and that the somatosensory cortex was hyper-excitable in symptomatic WAG/Rij rats (D’Antuono et al., 2006, Lüttjohann et al., 2011), and that there are changes in cerebral blood flow in brain regions involved in the generation and expression of SWDs in WAG/Rij rats (Tsenov et al., 2019). Furthermore, trigeminal inputs originating from the snout and vibrissae, but not the input from the nervus facialis, are necessary for the initiation of the spontaneous SWDs (Abbasova et al., 2010) in the cortical somatosensory projection area, specifically, this region is the cortical focus of SWDs generation in both WAG/Rij rats and GAERS. Not only from the Wistar lines obtained WAG/Rij strain and GAERS, but also other Wistar lines, e.g. those obtained from Winkelmann, Germany, tested at 84–94 weeks, and Harlan Wistar rats at a much younger age, show a high number of SWDs (Perescis et al., 2019, van Luijtelaar et al., 1994). Agouti and hooded strains showed much fewer SWDs in a strain comparative study (Inoue et al., 1990). In all, we can assume that all albino lines are at least partly sensory (visual) deprived. It is therefore proposed that the increase in excitability and other neural changes observed in the cortical areas of sensory-deprived individuals, such as an increase in arborization of pyramidal cells in S1 (Karpova et al., 2005), are due to increased experience-dependent neuronal activity of the intact sensory system, in this case, the touch system during postnatal life. According to this hypothesis (Singh et al., 2018), the anatomical reorganization of cortical areas in sensory-deprived animals would be the result of heightened use of the intact senses from birth onwards and during the rest of their lives. This view would predict that sensory deprivation of the whisking touch system would further enhance the process of epileptogenesis. This hypothesis was tested in the WAG/Rij model of absence epilepsy (Sitnikova, 2011). Trimming of the whiskers from postnatal day 1–21 led to predicted increases in both the number and duration of the spontaneously occurring SWDs when the rats were 5 and 8 months old in comparison to sham-trimmed WAG/Rij control rats, showing that sensory deprivation of the whisking touch system accelerates epileptogenesis. Mechanisms, regarding the type and nature of the changes in SWDs induced by early sensory deprivation in the cortex and thalamus, await to be explored, as well as whether in adult WAG/Rij rats whisker trimming would lead to quick changes in excitability and, as a consequence, changes in the number of SWDs. Sensory deprivation and epilepsy were not often studied, there is some literature regarding the role of hearing impairments on audiogenic seizure susceptibility (Sun et al., 2011). Tympanic membrane (TM) damage in young rats showed that 2 weeks later more than 80% of the rats showed audiogenic seizure (AGS) when exposed to a loud sound (120 dB sound pressure level white noise), while none of the control animals showed this. The susceptibility to AGS lasted at least 16 weeks after the TM damage, and even the hearing loss recovered. The seizures were controlled by the GABA transaminase inhibitor vigabatrin, suggesting a role of reduced GABAergic inhibition in this process. C-fos staining showed strong staining in the inferior colliculus (IC) in the TM-damaged rats, not in the control rats, after exposure to a loud sound, indicating a hyper-excitability in the IC during AGS. These results indicate that early-age conductive hearing loss can impair sound tolerance by reducing GABA inhibition in the IC. Interestingly, many Wistar rats and also 30% of the WAG/Rij rats show increased sensitivity to audiogenic seizures (Kuznetsova et al., 1996), perhaps because their partial blindness might have caused adaptations in other sensory modalities. In the absence models, the cortex and thalamus are the main key players. Changes in excitability that are occurring in parvalbumin (PV)-containing inhibitory interneurons located in the deprived barrel columns can occur very rapidly, in some instances within hours after the initial whisker plucking (Marik et al., 2010). PV-containing neurons were found to be decreased in the neocortex of WAG/Rij rats (Arkan et al., 2019), suggesting that their gradual decrease might play a role in both epileptogenesis and its increased speed after sensory deprivation. However, this suggestion awaits experimental verification. Emerging evidence suggests that an early environmental factor such as maternal care can improve or prevent the development of several pathologic phenotypes in offspring: stress vulnerability, fear, anxiety-like and depression-like behaviours (Champagne et al., 2008, Freitas et al., 2015, Masís-Calvo et al., 2013, Singh-Taylor et al., 2018, Weaver et al., 2004), genetically determined absence epilepsy with depression-like comorbidity (Sarkisova and Gabova, 2018), and schizophrenia-relevant features of behaviour (van Vugt et al., 2014). In rats, individual differences in maternal behaviour, primarily in licking/grooming and arched-back nursing, were shown to exert multilevel effects on physiological, morphological, behavioural, and neurochemical characteristics in the offspring (Champagne et al., 2008, Champagne et al., 2004, Champagne et al., 2003, Masís-Calvo et al., 2013, Peña et al., 2014, Weaver et al., 2004, Zhang et al., 2005) Although most studies of maternal care have focused on a special role of arched-back nursing for pups’ development (Champagne et al., 2003), non-arched-back nursing or non-nutritive contact with pups was found to be also a very important component of maternal behaviour, which can exert a beneficial phenotypic effect on the offspring (Sarkisova and Gabova, 2018, van Vugt et al., 2014). Well-caring rat mothers commonly express higher levels of licking-grooming, arched-back, and/or non-arched-back nursing compared with poor-caring rat mothers. Maternal licking-grooming represents naturally occurring sensory stimulation to the skin of pups, but non-arched-back nursing or non-nutritive contact with pups constitutes warmth from skin-to-skin tactile contacts with pups or thermotactile stimulation to most of the pup body (Kojima and Alberts, 2011). The mother’s arched-back nursing is the best posture for milk delivery compared with the non-arched-back posture (Lonstein et al., 1998). However, skin-to-skin tactile contact with pups is greater during non-arched-back nursing. Although suckling stimulation and receipt of milk are effective reinforcers for rat pups, these nutritive rewards did not contribute to the development of social attachments, as seen in the rat pups’ filial huddling. Interestingly, the frequency with which mother rats exhibit non-nutritive skin-to-skin contact with pups (hovering posture over the pups) did, however, correlate with the subsequent filial preference (Kojima and Alberts, 2011), which can be regarded as a form of social attachment. Depressive WAG/Rij mothers exhibited shorter non-nutritive skin-to-skin tactile contacts with pups (non-arched-back nursing) and pup licking-grooming, but even longer arched-back nursing compared with control Wistar rats (Sarkisova et al., 2017b, Sarkisova and Gabova, 2018); see also Section 3.5). This is probably the reason why the WAG/Rij pups compared to Wistar pups showed less attachment to their mothers (mother-oriented test), despite a normal nest-seeking response caused by olfactory stimuli (olfactory discrimination test) and the same level of locomotor activity (gait reflex test) (Malyshev et al., 2014). Newborn WAG/Rij pups also exhibited hypersensitivity to a gentle tactile touch (touching the ear with a cotton swab), which may be due to insufficiency of tactile stimulation and tactile skin-to-skin tactile contacts provided by their depressive mothers and may represent a kind of tactile deafferentation. On the other hand, a hypersensitivity to a tactile touch in the WAG/Rij pups can be a manifestation of increased stress reactivity or enhanced anxiety due to reduced maternal tactile stimulation and tactile contacts early in life. It is known that poor maternal care (insufficient contact with the mother) early in development leads to increased corticosterone levels and to a hyper-reactivity to stress in the offspring (Raineki et al., 2010). Recent studies have shown that the rewarding sensation of touch in affiliative interactions could be underpinned by a specialized system of nerve fibers called C-tactile afferents, which respond optimally to slowly moving, gentle touch (McGlone et al., 2014, Pawling et al., 2017). Interestingly, it has been shown in rodents that tactile stimuli activate hypothalamic oxytocin neurons (Okabe et al., 2015). This suggests that C-tactile afferent stimulation may cause oxytocin release during affiliative tactile interactions (Walker et al., 2017). Oxytocin is released when a mother cares for her child, making the interaction pleasurable. In this context, it could not be excluded that hypersensitivity to a gentle, non-painful tactile touch in WAG/Rij pups (Malyshev et al., 2014) may be due to tactile allodynia (pain hypersensitivity) associated with a reduced C-tactile hedonic touch experience early in life. Insufficient tactile contact with the mother induced by maternal separation was shown to cause mechanical allodynia in the whisker pad skin in adulthood (Dubner et al., 2016). In several studies, tactile stimulation (stroking a pup’s body with a soft brush) was used to ‘‘mimic’’ maternal licking and grooming behaviour. It has been demonstrated that neonatal tactile stimulation improves the maturation of premature infants and newborn rats (Schanberg and Field, 1987). Additional research has reported the beneficial effects of tactile stimulation on neonatal isolation-induced anxiety-like behaviour and pain sensitivity in adult rats (Imanaka et al., 2008), and the ability of tactile stimulation to attenuate amphetamine sensitization (Muhammad et al., 2011) and improve responsiveness to diazepam (Boufleur et al., 2012). Neonatal tactile stimulation can also decrease depression-like behaviour and potentiate antidepressant sertraline action in rats (Freitas et al., 2015). It has been demonstrated that symptoms of emotional and behavioural disorders in infants, caused by maternal pre- and postnatal anxiety and depression, can be modified by tactile stimulation assessed by mothers’ self-reported stroking of their babies during the first week of life. Of note, the positive effect of maternal stroking was associated with a reduction of the glucocorticoid receptor NR3C1 gene methylation (Pickles et al., 2017, Sharp et al., 2015), leading to a reduced HPA reactivity mediated via increased glucocorticoid receptor gene expression (Murgatroyd et al., 2015). These studies provide evidence that maternal stroking in infancy has a similar beneficial effect to that reported in rodents. A recent study reported that neonatal tactile stimulation affects the genetic absence epilepsy and comorbid depression-like behaviour in the WAG/Rij model (Balikci et al., 2020). From postnatal days 3–21, neonatal tactile stimulation (by a soft baby brush) was carried out for 15 min three times per day. The tactile stimulation protocol from a video article by Mychasiuk et al., (2013) was used. The effect of tactile stimulation was compared with that of deep touch pressure and maternal separation. Rat pups that were not subjected to any stimulation were used as a control. At the age of 5 months, WAG/Rij males were tested in the open field, sucrose consumption, and forced swimming tests. At the age of 6 months, EEG recordings were carried out. The number and total duration of SWDs per hour were assessed. Tactile stimulation, deep touch pressure, and maternal separation did not change substantially locomotor activity, and the amount of 20% sucrose consumed for 15 min. The tactile stimulation and deep touch pressure significantly increased the number of approaches to the drinking bottle in the sucrose consumption test. Tactile stimulation and deep touch pressure increased the latency to immobility, but only tactile stimulation decreased the immobility time and increased the duration of active swimming in the forced swimming test. Tactile stimulation and deep touch pressure reduced the number and total duration of SWDs compared with the control group. Results show that neonatal tactile stimulation can ameliorate the genetic absence epilepsy and comorbid depression-like behaviour in adult WAG/Rij rats. Of note, deep touch pressure similar to tactile stimulation reduced the number and duration of SWDs in adult rats. However, the effect of the tactile stimulation on depressive-like behaviour in the forced swimming test was greater compared with that of the deep touch pressure. Thus, tactile stimulation decreased immobility time and increased latency to immobility and the duration of active swimming, while the deep touch pressure increased only the immobility latency (Balikci et al., 2020). The authors believed that the deep tactile pressure resembles maternal non-arched back nursing. However, deep tactile pressure mimicked only one parameter of maternal non-arched-back nursing, such as hovering over the pups associated with tactile pressure, but not a rather significant parameter of maternal care, the warmth from skin-to-skin tactile contacts with pups or thermotactile stimulation. That’s probably why the effect of the deep tactile pressure on the depressive-like phenotype in WAG/Rij rats was substantially weaker compared with that of the tactile stimulation (produced by a soft baby brush), which, was more like a maternal licking-grooming than deep tactile pressure resembled maternal non-arched-back nursing. This assumption is supported by the fact that licking-grooming and non-arched-back nursing, naturally provided by the foster Wistar mother, exerted an equally strong effect on both the absence seizures and depression-like comorbidity in adult WAG/Rij offspring (Sarkisova and Gabova, 2018). The mechanism underlying the beneficial effects of neonatal tactile stimulation on the brain is not yet known. However, it has been assumed that the fibroblast growth factor-2 (FGF-2) may be a key modulator of these effects. For example, it has been demonstrated that tactile stimulation leads to an increase in the production of FGF-2 in both skin and brain (Richards et al., 2012). As FGF-2 is known to be a potent neurotrophic growth factor, increased production of FGF-2 in response to tactile stimulation might contribute to cortical plasticity and positive outcomes. Neonatal tactile stimulation can also affect other neurotrophic factors, including BDNF, which is very essential for structural and functional plasticity during development (Lu and Naggapan, 2014). BDNF increases neuronal excitability and is localized in the brain areas implicated in epileptogenesis. Moreover, seizure activity increases the expression of BDNF mRNA and protein, indicating that BDNF may contribute to the lasting structural and functional changes underlying epilepsy development and progression. BDNF modulates neuronal excitability in the hippocampus, but it also promotes neuronal survival in models of SE. In the pilocarpine model of SE, it has been shown that the BDNF-expressing vector injected in the hippocampus increases neurogenesis, limits neuronal damage, and reduces the occurrence of spontaneous seizures (Simonato and Zucchini, 2010). BDNF regulates synaptogenesis during development and has been shown to enhance axonal branching in the hippocampus and dendritic branching in the cortex (Binder et al., 2001). Glutamate release is enhanced, whereas inhibitory transmission is diminished by BDNF. Of note, the BDNF signaling pathway is impaired in GAERS after the onset of absence seizures. Intracerebroventricular injection of BDNF significantly reduces the occurrence of SWDs in GAERS (Landweer, 2010). BDNF is known to modulate the expression of neurotransmitters, which have a potential role in epileptic seizures and affective behaviours. BDNF may be dysregulated in depression (Nestler et al., 1968). Of note, neonatal tactile stimulation can increase BDNF levels (Antoniazzi et al., 2017). Thus, converging evidence suggests that deficits in BDNF signaling may contribute to the pathogenesis of both absence epilepsy and depressive-like comorbidity, and can underlie positive phenotypic effects of the neonatal tactile stimulation. Tactile stimulation early in life affects the neuroanatomical organization, dendritic morphology, and synaptic connectivity during brain development (Kolb and Gibb, 2010, Mychasiuk et al., 2013, Richards et al., 2012). Neonatal tactile stimulation was shown to increase dendritic branching, dendritic length, and spine density in the prefrontal cortex and amygdala (Richards et al., 2012). At the same time, it was found that tactile stimulation early in life led to decreases in spine density and dendritic length in the parietal cortex (Kolb and Gibb, 2010). This means that tactile stimulation may differentially affect the synaptic organization in different cerebral regions. Morphometric analysis of neurons in epileptic WAG/Rij and non-epileptic ACI rats revealed distinctive differences between somatosensory and motor cortex. The shape of dendritic arborization, the branching, and the orientation of dendrites in the somatosensory cortex where SWDs are thought to originate were different in epileptic WAG/Rij and non-epileptic ACI rats. The number of free terminations of apical dendrites was greater in the somatosensory cortex compared to the motor cortex in non-epileptic rats. In epileptic rats, there was also a difference between the two cortical areas, however in the opposite direction (Karpova et al., 2005). Put in other words, morphometric characteristics of dendrites in the cortical focal area of WAG/Rij rats were at variance with dendritic characteristics outside the focal areas, which were functionally similar to the areas in non-epileptic controls. These morphological features might reflect the hyper-excitability of somatosensory neurons, which underlie the initiation and spreading of SWDs in WAG/Rij rats. Thus, these results allow us to assume that the seizure-reducing effect of neonatal tactile stimulation in WAG/Rij rats (Balikci et al., 2020) can be mediated by its influence on the morphologic characteristics of dendrites in the somatosensory cortex leading to a decrease in the excitability of the epileptogenic cortical focus. It cannot be excluded that artificial neonatal tactile stimulation in WAG/Rij rats, like a natural mother’s care, might affect the development of the mesolimbic DA pathway (Peña et al., 2014), which is impaired in young WAG/Rij rats (Malyshev et al., 2014) and insufficiency of which is implicated both in absence epileptogenesis (Cavarec et al., 2019) and in the expression of depression-like symptoms in adult WAG/Rij rats (Sarkisova et al., 2013). Altogether, these findings indicate that tactile stimulation in early life plays an important role in preventing the development of many disorders in the offspring, including genetically determined absence epilepsy and depression-like comorbidity. One of the most important environmental factors that can have a significant impact on the later-life phenotype in offspring is the maternal diet. Numerous epidemiological studies and data from animal models indicate that maternal undernutrition, protein restriction, or, on the contrary, energy-rich diets during the perinatal period can lead to cardiometabolic diseases such as obesity, insulin resistance, hypertension, and raised serum cholesterol levels in the offspring (Bertram and Hanson, 2001, Samuelsson et al., 2008). Interestingly, several studies have shown that a paternal low-protein or high-fat diet can also influence future metabolic disease risk in the offspring (Carone et al., 2010, Ng et al., 2010), providing experimental evidence for non-genetic intergenerational paternal transmission of metabolic phenotypes (cholesterol and lipid metabolism) to offspring. The mechanism by which alterations in maternal or paternal diet may induce a long-term trans-generational effect on metabolism and phenotype in the offspring has been suggested to involve the altered epigenetic regulation of metabolic genes (Ferguson-Smith and Patti, 2011). It has been found that a high-fat diet in dams during pregnancy and lactation induces epigenetic and phenotypic changes in the offspring, increases expression of the μ-opioid receptor and preproenkephalin in the reward-related brain structures (nucleus accumbens, prefrontal cortex, hypothalamus), and this was accompanied by the hypomethylation of the promoter regions of genes in association with long-term alterations in gene expression (DA and opioids) and behaviour (preference for palatable foods) (Pitman and Borgland, 2015, Thanos et al., 2018, Vucetic et al., 2010). Maternal obesity and diabetes have also been reported to induce changes in DNA methylation in the liver and metabolic defects in the offspring (Li et al., 2013). Although most studies have concentrated on changes in DNA methylation associated with alterations in maternal diet, there is growing evidence that early-life nutrition can also induce substantial changes in other epigenetic mechanisms, such as histone modifications and miRNAs (Sandovici et al., 2011). Alterations in the paternal diet in rodents have also been shown to produce epigenetic changes in the offspring. Protein-restricted diet of male rats before mating can produce widespread changes in DNA methylation in the liver of the offspring compared to the control. Differences in paternal diet may be transmitted to offspring through epigenetic modifications of histone code and miRNA as well (Carone et al., 2010, Lillycrop and Burdge, 2015), indicating that histone and miRNAs may also play a role in the transmission of obesity and impaired metabolic state from the high-fat-fed fathers to the offspring. Evidence that maternal diet in humans can induce long-term epigenetic and phenotypic changes in the offspring is more limited. However, persistent epigenetic alterations in the DNA methylation of a number of genes in individuals who were prenatally exposed to famine during the Dutch Hunger Winter have been reported (Heijmans et al., 2008, Roseboom et al., 2006). More recent studies have indicated widespread gender-related changes in the epigenome of offspring associated with maternal periconceptional micronutrient supplementation intake (Khulan et al., 2012). Maternal high-fat or high-fructose diet during pregnancy and lactation can affect not only metabolic state but also emotional and cognitive behaviour in offspring (Coulibaly et al., 2017, Peleg-Raibstein et al., 2012, Winther et al., 2018) Increased anxiety (reduction of entries and time spent in open arms in the elevated plus-maze test), depression-like behaviour (increased immobility time in the forced swimming test), and cognitive impairments (episodic memory in the novel object recognition task and spatial working memory in Y-maze) have been reported. Males were more affected than females (Coulibaly et al., 2017). Clinical and animal model data indicate the contribution of maternal diet to susceptibility to adult-onset disease in offspring. Emerging evidence indicates the role of DNA methylation, the best-studied epigenetic mechanism, in the pathogenesis of many neurological and psychiatric diseases, including epilepsy (Kobow et al., 2013, Kobow and Blümcke, 2012) and depression (Fuchikami et al., 2011). The process of DNA methylation depends on the availability of dietary methyl group donors. A metabolic pathway that integrates nutrients from the environment to produce multiple epigenetic modifications through DNA methylation is the one-carbon cycle (Mentch and Locasale, 2016). Its simplified scheme is illustrated in Fig. 2. S-adenosyl-L-methionine (SAM) is the end product of a one-carbon cycle and serves as the universal methyl group donor for DNA methylation. DNA methyltransferase requires SAM to establish or maintain DNA methylation patterns. The synthesis of SAM is dependent on the availability of dietary folic acid, vitamin B12, methionine, betaine, and choline (Cooney et al., 2002). Developmental choline deficiency alters SAM levels and global and gene-specific DNA methylation (Niculescu et al., 2006). Prenatal choline availability has been shown to impact neural cell proliferation during early development and learning and memory in adult rodents (Glenn et al., 2007, Meck and Williams, 2003). Additional dietary cofactor such as zinc influence the availability of methyl groups for SAM formation, and thereby affects CpG methylation (Danchin et al., 2020). Vitamin B12 (cobalamin) is a cofactor of 5-methyltetrahydrofolate-homocysteine methyltransferase (MTHF) that catalyzes the conversion of homocysteine into methionine, the direct precursor of SAM. Folic acid in the form of tetrahydrofolate is a cofactor that is used in a number of biochemical reactions such as the biosynthesis of amino acids, DNA synthesis, and repair. It is also necessary for the conversion of homocysteine to methionine. If vitamin B12 is unavailable, tetrahydrofolate is “trapped” and cannot be used to convert homocysteine to methionine (Mattson and Haberman, 2005). The bioavailability of these cofactors may influence DNA methylation by modification of the one-carbon cycle activity and production of SAM (Feil and Fraga, 2012). Moreover, methyl donors are necessary for some neurotransmitter synthesis (DA, NA, 5-HT), impairment of which plays an important role in the pathogenesis of affective disorders, including depression (Gao et al., 2018). The neurotransmitters DA and NA are synthesized from the amino acid tyrosine in a series of chemical reactions dependent on tyrosine hydroxylase. 5-HT is synthesized from the amino acid tryptophan, and the rate-limiting step is catalyzed by tryptophan hydroxylase. SAM functions as a methyl-donating cofactor in the rate-limiting step of the synthesis of the monoamines DA and 5-HT (Mischoulon and Fava, 2002, Otero Losada and Rubio, 1989). Enhancement of SAM levels permits it to act as a cofactor of COMT, decreasing COMT enzyme activity and thereby degradation of catecholamines (Tsao et al., 2011). Therefore SAM can be regarded as a treatment option for depressive disorders that increase monoamines: low levels of SAM, elevated homocysteine, and low 5-HT, DA, and NA are usually found in depressive patients (Bottiglieri et al., 2000). Moreover, it has been also demonstrated antiepileptic and memory-enhancing effects of SAM administration in a PTZ-induced kindling model of epilepsy (Dhediya et al., 2016). A maternal diet with insufficient content of methyl donors can cause impairment of DNA methylation and alteration in gene expression leading to the development of various disorders in offspring (Geoffroy et al., 2019). Maternal (McCoy et al., 2017) and paternal (McCoy et al., 2018) methyl donors-depleted diets lead to increased anxiety and depression-like behaviour in adult rat offspring. In this context, of particular interest is the use of methyl-enriched maternal diets for correction or prevention of later pathologic phenotypes in the offspring. It has been found that the maternal high fructose diet with supplementation of methyl donors and cofactors of the one-carbon cycle (folic acid, choline, betaine, L-methionine, and vitamin B12) improves anxiety-like and depression-like behaviours and induces better performance in spatial and recognition memory tests in adult rat offspring (Coulibaly et al., 2017). Supplementation of the maternal diet with methyl donors during critical periods of brain development (in utero and pre-weaning stage) counteracted the development of some of the adverse effects seen in mice offspring from dams fed a high-fat diet: excessive weight gain, increased fat preference, changes in gene expression, and global hypomethylation in the prefrontal cortex. Sex differences in the effects of the maternal diet were observed. So, fat preference and DA transporter mRNA expression in the reward-associated ventral tegmental area were significantly increased only in male offspring born to mothers fed a high-fat diet, as compared to control offspring. Methyl donors-supplemented maternal diet normalized these measures. Maternal high-fat diet-induced global hypomethylation was more pronounced in male offspring than in female offspring. This adverse effect of the maternal high-fat diet was corrected by methyl supplementation. On the contrary, µ-opioid receptor mRNA expression changes were more pronounced in female offspring compared to male offspring: in females, mRNA expression was increased both in the prefrontal cortex and nucleus accumbens, while in males only in the prefrontal cortex (Carlin et al., 2013). Although the effect of the maternal diet on the epigenome of offspring is best investigated in the field of metabolic diseases, the effect of the maternal diet on the epigenome of offspring in the area of epilepsy and associated behavioural comorbidities has only recently begun to be investigated. It has been found in rats of the Krushinsky-Molodkina strain (KM) that maternal methyl-enriched diet during the prenatal and early postnatal ontogeny decreased the intensity of the audiogenic seizure in the progeny of both sexes (Poletaeva et al., 2014). Positive effects of maternal methyl-enriched diet on the genetically-based epileptic phenotype in offspring of KM rats were associated with changes in the methylation status of several genes, which were previously shown to be hypermethylated after epileptic tolerance procedure (Miller-Delaney et al., 2012). Maternal methyl-enriched diet induced different changes (increases or decreases) in the methylation status of different genes in adult offspring depending on the animal group that differs in the intensity of audiogenic seizures (“0” – no seizures and “4” – intense seizures). The majority of these genes were related to nuclear functions, such as DNA binding and transcriptional regulation. The epileptic tolerance procedure (seizure preconditioning) decreased the intensity of SE provoked by kainate in mice. Surprisingly, genes that were hypermethylated after epileptic tolerance did not match mostly the genes that were hypomethylated after the development of epilepsy (Miller-Delaney et al., 2012). Whether maternal methyl-supplemented diet can improve behavioural comorbidities (elevated anxiety and depression-like behaviour) described in KM rats (Sarkisova et al., 2017c) remains to be investigated. The effect of maternal methyl-enriched diet on pathologic phenotype and gene expression in WAG/Rij rats was studied (Sarkisova et al., 2021, Sarkisova et al., 2020). The methyl-enriched diet contained methyl group donors (choline, betaine, L-methionine) and cofactors of the one-carbon cycle (folic acid, vitamin B12, zinc), which are highlighted in red in Fig. 2. The methyl-supplemented diet, as used in these studies, modified other genetically determined pathologic phenotypes, such as audiogenic seizures in KM rats (Poletaeva et al., 2014) and agouti coat color in mice (Cooney et al., 2002) and rats (Prasolova et al., 2009). Correction of pathological phenotypes was accompanied by changes in the methylation profile of certain genes (Cooney et al., 2002, Prasolova et al., 2009, Poletaeva et al., 2014). In WAG/Rij rats, maternal methyl-enriched diet during the perinatal period (a week before mating, during mating, pregnancy, and a week after parturition) reduced SWDs and comorbid depression in an adult 7-month-old offspring (Sarkisova et al., 2020). This beneficial phenotypic effect was greater expressed in males compared to females. In WAG/Rij males, maternal methyl-enriched diet reduced the number and total duration of SWDs, but it did not affect the mean duration of SWDs. In contrast, in WAG/Rij females, maternal methyl-enriched diet reduced the mean and total duration of SWDs, but not the number of SWDs. The reduction in the number of mature SWDs in males was accompanied by the appearance of a large number of immature discharges, which are typical for young 2–3 month old rats of this strain (Gabova et al., 2020), indicating an anti-epileptogenic effect. Of note, in 50 % of WAG/Rij males born to mothers fed methyl-supplemented diet, mature SWDs were absent, and they were replaced by immature discharges commonly recorded in young (pre-symptomatic) WAG/Rij rats. At the same time, in 100 % of age-matched males born to mothers fed a control diet mature SWDs were observed. In other words, maternal methyl-enriched diet during the perinatal period decelerated the age-related process of the progressive development of epileptic activity in WAG/Rij males. In WAG/Rij females, the anti-epileptogenic effect of maternal methyl-enriched diet was less pronounced. Maternal methyl-enriched diet caused substantial changes in the averaged power spectra of SWDs, as was established with Fast Fourier Transform analysis, only in males: a decrease in the spectral power at the fundamental frequency, and the first and second harmonics. Maternal methyl-enriched diet did not significantly affect the averaged spectral power of SWDs in female offspring. No significant differences were found between male and female offspring born to mothers fed a control diet in the number, mean and total duration of SWDs, as well as in the averaged spectral power of SWDs. The methyl-enriched maternal diet had no substantial effect on the averaged spectral power of the background EEG, indicating a selective effect on SWDs. In male and female WAG/Rij offspring born to mothers fed a methyl-enriched diet, the duration of immobility in the forced swimming test was shorter and the duration of climbing was longer compared with the corresponding values in WAG/Rij offspring born to mothers fed the control diet. Moreover, maternal methyl-enriched diet increased the duration of swimming only in females, and the number of divings only in males. Maternal methyl-enriched diet increased sucrose preference (anti-anhedonic effect) only in males and did not substantially affect the preference for sucrose in females (Sarkisova et al., 2020). The beneficial effect of maternal methyl-supplemented diet on SWDs and depression-like comorbidity in WAG/Rij offspring was associated with increases in the DNA methyltransferase 1 (DNMT1) and HCN1 ion channel gene expression in the somatosensory cortex (Sarkisova et al., 2021). Epigenetic modifications induced by maternal methyl-enriched diet in the offspring at the early stages of ontogenesis were supposed to be a possible mechanism underlying the correction of genetically determined pathologic phenotype in WAG/Rij rats. Of note, maternal methyl-supplemented diet-induced increases in the expression of the DNMT1 gene in the somatosensory cortex of WAG/Rij offspring were similar to the effect of early long-term ethosuximide treatment, a first choice anti-absence drug (Dezsi et al., 2013). Put in other words, maternal methyl-enriched diet during the perinatal period produced the same alterations in the DNMT1 gene expression in the somatosensory cortex of offspring as pharmacological long-term anti-absence therapy. Reduced expression of the HCN1 ion channel in the somatosensory cortex in WAG/Rij rats is thought to be associated with the genesis of SWDs (Blumenfeld et al., 2008, Kole et al., 2007, Schridde et al., 2006, Strauss et al., 2004). Moreover, loss of HCN1 in HCN1-knockout rats caused spontaneous bilateral SWDs accompanied by behavioural arrest, both of which were suppressed by ethosuximide (Nishitani et al., 2019), providing evidence of a causal relationship between HCN1 and absence epilepsy. A rapid decline in the expression of HCN1 channels precedes the onset of absence seizures (Kole et al., 2007). This suggests that the reduced expression of HCN1 facilitates the initiation and propagation of spontaneous generalized seizures. Given this, increased expression of the HCN1 gene in the somatosensory cortex in WAG/Rij offspring is supposed to be associated with the beneficial phenotypic effects of methyl-supplemented maternal diet. These findings are the first to indicate a new preventive therapeutic strategy based on the maternal diet during the perinatal period, targeting DNA methylation in the initiation site of SWDs, to correct symptoms of genetic absence epilepsy and comorbid depressive-like behaviour in offspring, (Sarkisova et al., 2021, Sarkisova et al., 2020). We can assume that the maternal methyl-supplemented diet prevents the decline in the expression of HCN1 in the somatosensory cortex, and therefore epileptogenesis, leading to the prevention of the development of depression-like comorbidity. Maternal methyl-enriched diet increased the expression of HCN1 and DNMT1 genes not only in the somatosensory cortex, but also in the hippocampus, as well as the expression of the TH gene in the nucleus accumbens (Sarkisova et al., 2021), a brain region critically involved in the pathophysiology of depression. Moreover, maternal methyl-supplemented diet also increased DAergic tone of the mesolimbic brain system, which contributes to epileptogenesis and comorbid depression in WAG/Rij rats (Sarkisova et al., 2022). Further studies are necessary to understand epigenetic mechanisms by which maternal dietary supplementations during the perinatal period prevent epileptogenesis and the development of depression-like comorbidity in offspring genetically predisposed to absence epilepsy. Another possible mechanism for the beneficial effect of the methyl-enriched diet on absence epilepsy and comorbid depression in the offspring of WAG/Rij rats could be its effect on the gut microbiota. It is now well established that maternal diet during the perinatal period affects the development of gut microbiota in the offspring (Al Rubaye et al., 2021) leading to specific epigenetic signatures that may predispose or prevent the development of later-life pathology. Experimental evidence suggests that maternal supplementation of dietary methyl groups is critical for the neurodevelopment of offspring (Emmerson and Jadavji, 2016). Furthermore, folate and vitamin B12 deficient diets may negatively impact both the microbiome and the brain function such as memory (Park et al., 2022). A number of studies have clearly demonstrated that the gut microbiota can modulate or contribute to neurological and psychiatric diseases, including epilepsy (Gong et al., 2020, Mengoni et al., 2021) and depression (Eltokhi and Sommer, 2022). The gut-brain axis has been shown to be able to affect the excitability in the brain and thereby modulate seizure susceptibility (Darch and McCafferty, 2022, Mengoni et al., 2021). Differences in the gut microbiota have been reported in patients with epilepsy compared to healthy controls (Gong et al., 2020) and in the WAG/Rij absence model (Citraro et al., 2021). Interestingly, microbiota transplantation from non-epileptic Wistar rats or ethosuximide-treated WAG/Rij rats induced a significant reduction in the number and duration of SWDs. Transplantation from ethosuximide-treated WAG/Rij rats was more effective than that from Wistar rats (Citraro et al., 2021). The link between depression-like behaviour and microbiota has also been described (Eltokhi and Sommer, 2022). For example, the transplantation of fecal matter from depressed patients into microbiota-depleted rats led to depressive-like behaviour (Kelly et al., 2016). Clinical evidence for the role of the microbiota in depression is provided by an alteration in the number of microbiota and their diversity in individuals with depression when compared to healthy controls. Moreover, the role of the microbiota-gut-brain axis in regulation of DAergic signaling, the dysfunction of which leads to depressive disorders, has been established (Hamamah et al., 2022). Taken together, these data allow us to assume that the maternal methyl-enriched diet leads to a balanced and diverse composition of the microbiota in the offspring, which may contribute to a favorable phenotypic effect. Further research is needed to find out whether the maternal methyl-enriched diet affects the composition of microbiota in the offspring of WAG/Rij rats and whether this really makes a significant contribution to the positive phenotypic effect of the methyl-enriched diet. Positive effects of maternal methyl-supplemented diets on the pathologic phenotype in offspring were also shown for other genetic animal models of neurological disorders, such as the Alzheimer’s disease model (Ash et al., 2014, Velazquez et al., 2020), atherosclerosis model (Delaney et al., 2013), Down syndrome model (Moon et al., 2010), and Rett syndrome model (Nag and Berger-Sweeney, 2007) in mice. Maternal choline supplementation ameliorated Alzheimer’s disease pathology in old offspring by reducing brain homocysteine levels and changing 27 genes expression related to inflammation, histone modification, and neuronal death. The maternal diet reduced amyloid-ß load, and microglia activation, and improved cholinergic signaling in the brain and cognitive deficits in old mice. A transgenerational benefit of a methyl donor-supplemented maternal diet on the development of Alzheimer-like symptoms in mice was shown (Velazquez et al., 2020). Human studies are very limited and inconsistent, some of them point to positive effects, such as the prevention of neural tube defects and mental health problems (O'Neil et al., 2014a) in offspring, but others are inconclusive (Bekdash, 2019). Although animal model studies that examined how maternal methyl-supplemented diets impact epigenome and phenotype in offspring have reported a favorable outcome, it should be emphasized that excessive or inadequate intake of methyl donors can have adverse effects (Bekdash, 2019, De Crescenzo et al., 2021, O’Neill et al., 2014b). For instance, a higher folic acid-supplemented maternal diet during pregnancy led to disruptive gene and protein expression changes in the cerebral hemispheres, as well as behavioural abnormalities in neonatal C57Bl6J mice, including increased ultrasonic vocalizations, greater anxiety-like behaviour, and hyperactivity (Barua et al., 2014). Excess folic acid-supplemented diet during pregnancy can alter cortical neurodevelopment in mouse offspring. Paradoxically, changes in the brain due to very high amounts of folic acid mimicked those associated with a deficiency of folic acid (De Crescenzo et al., 2021). Women who have given birth to a child with neural tube defects (O’Neil et al., 2014a, O’Neill et al., 2014b) or who have epilepsy and take anticonvulsants (Moore, 2005) have generally been advised to take high doses of folic acid. However, a very high amount of folic acid can be harmful to the brain development of the fetus. Human epidemiological studies showed a strong correlation (0.87) between maternal consumption of prescription prenatal vitamins (containing > 1 mg of folic acid) and ASD incidence in offspring (Beard et al., 2011). Too much folic acid might disrupt brain development and thus increase the risk of ASD (Leeming and Lucock, 2009, Rogers, 2008). However, consumption of folic acid by women beginning at the periconceptional period was associated with a reduced risk of ASD in children of mothers with inefficient folate metabolism (for example, MTHFR gene variant). A greater risk for ASD was also observed for children whose mothers had other one-carbon metabolism pathway gene variants and reported no prenatal vitamin intake (Schmidt et al., 2011). Based on the available evidence, it can be concluded that intake of dietary methyl donors and cofactors during the perinatal period may alter fetal development, thus establishing a link between early environment and disease development in the offspring later in life. However, the results presented here suggest the importance of ‘optimal’ methylation status during the perinatal period, particularly due to maternal diet, not only for reproducing healthy offspring but also for preventing the development of pathological phenotypes, even if they are genetically determined. The composition of the maternal diet, dose and duration of methyl donor supplementation at critical stages of neurodevelopment as well as genetic contribution to the components of one-carbon metabolism (for example, MTHFR gene variant) seem to be very important factors to consider. Other factors also may interact with maternal diet to influence the phenotypic outcome in the offspring: genetic background, gender, developmental windows of exposure, and tissue-specific susceptibility. It should also be taken into account that changes in the methylation status on the global level, as most likely occurs when methyl supplements are used, can affect both “beneficial” and “harmful” genes. The disease promotion or prevention may depend on the combined genes that will be affected. Of note, the beneficial phenotypic effect of the maternal methyl-enriched diet on absence epilepsy and depression-like comorbidity was associated with epigenetic modifications in the expression of genes relevant to this pathology. Therefore, we can assume that the methyl-supplemented maternal diet can be regarded as a potential new epigenetic therapeutic strategy for the treatment of human absence epilepsy and its comorbidities. This review presents the current findings indicating that environmental perturbations during the perinatal period have a major impact on the development of absence epilepsy and psychiatric comorbidities in the offspring. The developing brain is very sensitive to environmental impacts and therefore it is not surprising that the prenatal administration of drugs that directly or indirectly affect the balance of excitatory and inhibitory processes in different brain regions, often resulting from neurodevelopmental dysfunction in GABAergic cortico-thalamo-cortical and extra-thalamic limbic circuitries, might be a common cause not only of increased seizure susceptibility and epilepsy but also of multiple comorbid behavioural, emotional and cognitive impairments. Accumulating evidence suggests that early-life stress can prime seizure occurrence and epileptogenesis (Huang, 2014). Moreover, early-life stress is a major risk factor for anxiety and depression in adulthood and may contribute to neuropsychiatric comorbidities in epilepsy (Mumtaz et al., 2018). In addition, epigenetic modifications can be regarded as a shared pathogenic mechanism underlying the effect of various types of early-life stress on epilepsy and its psychiatric comorbidities. In general, the literature on early-life interventions and their impact on epilepsy and its comorbidities for a large part is aimed at the HPA axis and hippocampus, and, indeed, many changes in the limbic system are widely acknowledged. However, little attention is commonly paid to other brain structures. Early life stress has a broader impact than only on the hippocampus and the limbic system. It can also affect the excitatory/inhibitory balance in the medial prefrontal cortex (Ohta et al., 2020). Next, disruptions or alternations in network activity in one brain region could impact other networks and connected brain regions (Onat et al., 2007), which may directly or indirectly contribute to psychiatric disorders co-existing with epilepsy. Another important issue is that long-term consequences induced by early-life interventions may not be the same in various brain regions, especially in the case of GABA with its diversity of subunit compositions. So, an upregulation of the numbers of a specific type of receptors in one brain region could be accompanied by a downregulation in another part of the brain (Citraro et al., 2006, Tong et al., 2009). It is clear by now that early interventions may work oppositely in different brain regions (Ohta et al., 2020). It may be assumed that early interventions-induced changes in the brain connectivity/network activity and region-specific structural, neurochemical, and molecular abnormalities and dysfunctions could underlie psychiatric comorbidities of epilepsy, including anxiety and depression. Changes in the network functional connectivity were found in genetic absence epilepsy both in human and animal models. WAG/Rij mothers do not seem to be the best moms, at least when they are compared to Wistar mothers. In addition to absence seizures and comorbid depression-like state, WAG/Rij dams were found to exhibit reduced maternal care, as evidenced by the smaller number of tactile contacts with pups and licking/grooming or tactile stimulations. Rearing by foster Wistar mothers with a high level of maternal care reduced the number and duration of SWDs and comorbid depression in adult offspring of WAG/Rij rats (Sarkisova and Gabova, 2018). However, it is unknown whether less good and perhaps insufficient maternal care provided by depressive WAG/Rij mothers (Sarkisova et al., 2017b) causes early-life stress in offspring, and how much this could contribute to the alterations of network structure in adult WAG/Rij rats. Neonatal maternal separation is one of the commonly used laboratory methods to study early-life stress effects on the development of neurologic and psychiatric disorders in adulthood. However, there was a considerable amount of variability between studies related to the behavioural outcomes induced by neonatal maternal separation: no effects, harmful effects, or positive effects on anxiety and depression-like behaviour. In the same way, opposite results were obtained, indicating the anti-absence action of neonatal maternal separation in WAG/Rij rats (Schridde et al., 2006) and pro-convulsive effects in limbic epilepsy models (Kumar et al., 2011, Salzberg et al., 2007). This inconsistency in the data is probably related to differences in the maternal separation protocol and the lack of standardization (Wang et al., 2020). However, the opposite effects of neonatal maternal separation on absence epilepsy and TLE may also be due to the specificity of the models used, the differences in vulnerability of the affected networks in the genetic models and the induced seizure models, and the opposite effects of neonatal maternal separation in different regions of the brain. Of interest is that GABA mimetic drugs have also opposite effects in TLE and absence epilepsy, and therefore the opposite effects reported might also be region-specific and GABA-related. The effects of the neonatal maternal separation on the psychiatric comorbidities of epilepsy have not been investigated in any experimental seizure or epilepsy model, including absence epilepsy models. However, since other antiepileptogenic effects may affect the comorbidities as well, it cannot be excluded that neonatal maternal separation may also have an impact on the comorbidities. Neonatal handling is also usually used as a form of early-life intervention which can result in long-term consequences. In different studies, the duration of handling is variable, and this procedure can be repeated a different number of times. The effects of neonatal handling on epileptogenesis and seizure susceptibility were rarely studied and, if studied, these investigations were based on the assumption that this early-life intervention is stressful, and the focus was on hippocampus-related epilepsies and the role of the HPA-axis. In this case, early-life handling was pro-convulsive in the lithium-pilocarpine model (Persinger et al., 2002). However, beneficial effects of neonatal handling such as reduced anxiety/emotionality and stress responses later in life were also reported (Raineki et al., 2014), as well as the antiepileptogenic effect in the WAG/Rij rat model (Schridde et al., 2006). It is still unknown what causes the opposite effect of neonatal handling in different animal models of epilepsy: pro-convulsive in the TLE model and antiepileptogenic in the absence epilepsy model, but see our suggestions above regarding the opposite effects of neonatal maternal separation. It is well documented that neonatal handling of the pups changes the behaviour of the mother: it increases licking and grooming and does not change the arched-back nursing (de Azevedo et al., 2010, Reis et al., 2014). The increase in licking/grooming was exactly the type of maternal care that prevented the development of absence epilepsy and comorbid depression in WAG/Rij offspring in a cross-fostering study (Sarkisova and Gabova, 2018). This allows us to conclude that increased maternal care and immediate effects of infantile tactile stimulation following the return of pups to the nest could be considered as a mediating mechanism for the beneficial effects of neonatal handling on behaviour (Raineki et al., 2014), as well as on epileptogenesis in the WAG/Rij model (Schridde et al., 2006). Further research is needed to find out the role of maternal care in the development of epilepsy and its psychiatric comorbidities, and in particular in the effects of neonatal maternal separation and neonatal handling on epileptogenesis and psychiatric comorbidities in the absence epilepsy. In addition, inconsistency of data concerning the effects of neonatal maternal separation and handling in the absence epilepsy and TLE suggests that the effects of these environmental manipulations should be more carefully investigated, considering the time period of these manipulations during the postnatal development (the first weeks of life) with special attention to which of the sensory systems is currently being formed and what effects do these manipulations have on maternal behaviour. Early sensory deprivation, as induced by neonatal whisker trimming, can lead to neuroplastic changes (enlarged excitation and weakened inhibition) in the somatosensory cortex and behavioural alterations, including measures of anxiety in adults, next to accelerated epileptogenesis, as was demonstrated in the WAG/Rij model. Moreover, WAG/Rij rats and GAERS are albinos with a relative insufficiency of the visual system. This may imply that there will be neuroplastic changes in other sensory systems, namely in the somatosensory cortex, resulting in a hyper-excitability of neural cells and an increased sensitivity to painful stimuli. Both results were reported in the genetic rat models (De Caro et al., 2020, Polack et al., 2007). Mechanisms regarding the type and nature of the early sensory deprivation-induced changes in SWDs in the cortex and thalamus await to be investigated, as well as whether in adult WAG/Rij rats whisker trimming would lead to alterations in excitability and, as a consequence, the changes in the number of SWDs. Not only sensory deprivation but also intense sensory (visual) stimulation has been shown to have dramatic effects, as was found in Sprague-Dawley rats. Visual stimulation, by itself, can induce in adult rats an enduring sensitization of visual response with epileptiform SWD-like characteristics, not seen before the onset of stimulation. This work deserves to be replicated and extended, also earlier in the development. Moreover, it is necessary to investigate the consequences of intense visual stimulation early in life both in the visual system and in the somatosensory system, considering that plastic changes in one sensory system may have large consequences for other sensory systems. A recent study demonstrated rather large differences between WAG/Rij and Wistar rats in pain thresholds, indicating hypersensitivity to mechanical and thermal noxious and non-noxious stimuli in WAG/Rij rats (De Caro et al., 2020). This is important, considering that the pain system involves the same thalamic nuclei, somatosensory cortex, and thalamo-cortical circuitry as the networks in which SWDs are generated. Therefore, a study of pain thresholds in GAERS is relevant. Interestingly, hypersensitivity to a gentle, non-painful tactile touch has earlier been detected in newborn WAG/Rij pups (Malyshev et al., 2014). The fact that the changes in pain threshold were occurring in pre-symptomatic WAG/Rij rats, that is before SWDs are emerging, is important and may point toward the cause of SWDs in the genetic rat models. Pain sensitivity is known to be heritable. Therefore, research on gene expression of pain sensitivity in WAG/Rij rats seems to be the way to proceed. Hypersensitivity to a non-painful tactile touch in WAG/Rij pups may be due to tactile allodynia (pain hypersensitivity) associated with a reduced C-tactile hedonic touch experience provided early in life by depressive WAG/Rij mothers. Insufficient tactile contact with the mother induced by maternal separation caused mechanical allodynia in the whisker pad skin in adulthood (Dubner et al., 2016). Consequently, it is also necessary to investigate in the future whether early-life environmental factors, especially maternal care, might contribute to abnormal pain sensitivity in the WAG/Rij model and whether the administration of anti-pain medication early in development could be an effective way to prevent SWDs. In addition, whether neonatal tactile stimulation can correct abnormal pain sensitivity in WAG/Rij rats and whether the beneficial effect of neonatal tactile stimulation on the absence epilepsy and comorbid depression-like behaviour in WAG/Rij rats is associated with epigenetic modifications of gene expression relevant for this pathology remains to be investigated. Mounting evidence suggests that maternal care leads to epigenetic modifications of gene expression, specifically by changes in DNA methylation, which might be a cause of the long-lasting impact of maternal care on the offspring’s phenotype. Given this, a maternal methyl-enriched diet was used as an epigenetic treatment to affect DNA methylation, leading to changes in gene expression, and, as a consequence, causing correction of genetic absence epilepsy and depression-like comorbidity in WAG/Rij rats. Maternal methyl-enriched diet during the perinatal period delayed epileptogenesis: it decelerated the age-related process of the progressive development of epileptic activity and reduced depression-like comorbidity in WAG/Rij offspring (Sarkisova et al., 2020). Antiepileptogenic and disease-modifying effects of the maternal methyl-enriched diet were associated with increases in the DNMT1 and HCN1 gene expression in the somatosensory cortex (Sarkisova et al., 2021), indicating a new preventive therapeutic strategy to correct genetic absence epilepsy and comorbid depression in the offspring (Sarkisova et al., 2021, Sarkisova et al., 2020). Of note, the beneficial effect of the maternal methyl-enriched diet was accompanied by the same alterations in gene expression in the somatosensory cortex as pharmacological treatment with ethosuximide, a first-choice anti-absence drug (Dezsi et al., 2013). Therefore, the methyl-supplemented maternal diet can be regarded as a potential new epigenetic therapeutic strategy for the treatment of human absence epilepsy and its comorbidities. Whether the maternal methyl-enriched diet can alter the pain sensitivity in adult WAG/Rij offspring needs to be established. The expression of several ion channels has been identified to contribute to the aetiology of epilepsy, including absence epilepsy. These channels are critical for electrical signaling between neurons and are responsible for the regulation of neuronal excitability. Dysregulation of ion channel gene expression is highly associated with epilepsy. Although the molecular mechanisms that underlie these changes are not understood yet, it is known that many of these ion channel genes can be regulated by NRSF (McClelland et al., 2014, McClelland et al., 2011). Epileptogenesis causes downregulation of genes that are involved in epilepsy, for instance, the HCN1 ion channel gene in both genetic absence epilepsy (Nishitani et al., 2019, Strauss et al., 2004) and in various TLE models (McClelland et al., 2011). However, specific mechanisms are still unknown although there is evidence that HCN1 channelopathy derives from NRSF-mediated transcriptional repression contributing to epileptogenesis. This means that therapeutic interventions targeting NRSF to restore HCN1 gene expression can slow down the progression of epilepsy, as was shown in a TLE mouse model (McClelland et al., 2011). However, it remains to be investigated whether targeting NRSF to restore HCN1 gene expression could slow down the development of absence epilepsy in genetic absence epilepsy models. Reduced expression of HCN1 ion channels in the somatosensory cortex is thought to be associated with the genesis of absence epilepsy in the WAG/Rij model. Data presented in this review indicate that pathologic phenotype in WAG/Rij rats can be modified by early-life environmental interventions leading to epigenetic modifications of the HCN1 channel expression in the somatosensory cortex (Sarkisova et al., 2021, Schridde et al., 2006). Neonatal maternal separation and neonatal handling, presumably leading to increased maternal care, neonatal artificial tactile stimulation, mimicking maternal licking/grooming (Balikci et al., 2020), and improved maternal care of healthy foster mothers (Sarkisova and Gabova, 2018) exerted disease-modifying effects on the pathologic phenotype in WAG/Rij offspring. Increased Ih and HCN1 expression at the SWD initiation site was associated with the absence seizure reduction (Schridde et al., 2006). Ih drives the repetitive firing in nociceptive neurons mediated by HCN1, and HCN channel blockers have an analgesic action on peripheral pain (Ramírez et al., 2018). Finally, changes in Ih/HCN1 were already reported in pre-symptomatic WAG/Rij rats (Kole et al., 2007), suggesting that the reduced Ih currents and the HCN1 channel expression may be an initiating factor in the pathogenesis of SWDs in WAG/Rij rats. Therefore, it is thought that Ih/HCN1 might play a role in both epileptogenesis and pain transmission. It can be assumed that maternal separation and neonatal handling, known to affect HCN1 expression and Ih, may exert antiepileptogenic effects via alterations in maternal behaviour that the mothers display in response to environmental changes. Whether maternal care can alter HCN1 gene expression in the somatosensory cortex needs to be found out. Epigenetic modifications of relevant genes expression induced by the maternal methyl-enriched diet in the offspring at the early stages of ontogenesis are also assumed to be a possible molecular mechanism underlying the correction of genetically determined pathologic phenotype in WAG/Rij rats (Sarkisova et al., 2021). Further in-depth studies are necessary to better understand epigenetic mechanisms by which maternal dietary supplementations during the perinatal period prevent epileptogenesis and the development of depression-like comorbidity in offspring genetically predisposed to absence epilepsy. Epigenetic therapy based on maternal diet, in general, is a new promising area of research. However, many questions still need to be addressed and answered, regarding the composition of the maternal diet, dose and duration of methyl-donors supplementation at critical stages of neurodevelopment as well as genetic contribution to the components of one-carbon metabolism (for example, MTHFR gene variant). Other factors also may interact with the maternal diet to influence the phenotypic outcome in the offspring: genetic background, sex, developmental windows of exposure, and tissue-specific susceptibility. Finally, the genetic absence epilepsy models with their clear and easy quantification of EEG-based epileptic seizures are rather suitable for the study of how early environmental factors shape the brain not only in relation to epileptogenesis but also in relation to other associated phenotypic alterations, such as neuropsychiatric disorders. Epigenetic interventions for the absence epilepsy and its comorbidities are a new area of research. Results presented in this review highlight DNA methylation as an epigenetic mechanism in controlling and/or modulating absence seizures and their comorbidities, and might insight into epigenetic treatment based on a maternal methyl-enriched diet. The main conclusions and perspectives for future research are summarized in Fig. 3. Sarkisova K.Yu.: Conceptualization, Methodology, Writing – Original draft preparation, Reviewing and Editing;Van Luijtelaar G.: Supervision, Writing – Original draft preparation, Reviewing and Editing. The authors declare no conflict of interest.

PMC9649975

Waqas Mahmood,Irshad Ahmad,Mohsin Abbas Khan,Syed Adnan Ali Shah,Muhammad Ashraf,Mirza Imran Shahzad,Irfan Pervaiz,Muhammad Sajid-ur-Rehman,Umair Khurshid

Synthesis, characterization, molecular docking and biological evaluation of Schiff Base derivatives of cefpodoxime

02-11-2022

Cefpodoxime,Corona virus class,Antiviral,Schiff bases,FTIR,NMR (1H and 13C)

Synthesis of new Cefpodoxime derivatives via Schiff Bases mechanism and the efficiency of their antimicrobial and antiviral activities were addressed. They were analyzed for structural validation by using spectroscopic techniques using FTIR, 1HNMR, and 13CNMR. Molecular docking against IBV Virus papain-like protease (PLPro) was done with Auto dock tools against compounds having excellent IC50 values against IBV (Corona Class) virus. All derivatives showed strong zone of inhibition ranges from (55 ± 2.0 to 70 ± 0.8 mm) against E. coli. Compounds 1,2,4 and 6 derivatives showed remarkable activity against Stenotrophomonas maltophilia and Serratia marcescens. But For most the newly synthesized derivatives C1 (64 ± 1.60), C3 (32 ± 0.80), and C8 (64 ± 1.60) showed potential IC50 values against two variants of Corona class viruses i.e. Avian Influenza (H9) and Avian corona (IBV) viruses. The current study revealed that newly synthesized Schiff Bases possessed strong anti-viral potential. Further studies may make a breakthrough in medical sciences to tackle latest challenges such as Corona Virus Diseases.

Synthesis, characterization, molecular docking and biological evaluation of Schiff Base derivatives of cefpodoxime Synthesis of new Cefpodoxime derivatives via Schiff Bases mechanism and the efficiency of their antimicrobial and antiviral activities were addressed. They were analyzed for structural validation by using spectroscopic techniques using FTIR, 1HNMR, and 13CNMR. Molecular docking against IBV Virus papain-like protease (PLPro) was done with Auto dock tools against compounds having excellent IC50 values against IBV (Corona Class) virus. All derivatives showed strong zone of inhibition ranges from (55 ± 2.0 to 70 ± 0.8 mm) against E. coli. Compounds 1,2,4 and 6 derivatives showed remarkable activity against Stenotrophomonas maltophilia and Serratia marcescens. But For most the newly synthesized derivatives C1 (64 ± 1.60), C3 (32 ± 0.80), and C8 (64 ± 1.60) showed potential IC50 values against two variants of Corona class viruses i.e. Avian Influenza (H9) and Avian corona (IBV) viruses. The current study revealed that newly synthesized Schiff Bases possessed strong anti-viral potential. Further studies may make a breakthrough in medical sciences to tackle latest challenges such as Corona Virus Diseases. Schiff Base was initially synthesized by the Italian Scientist Hugo Schiff [1] in 1864. The condensation reaction of aldehydes/ketones with aromatic amines leads to the discovery of compounds that were later called Schiff Bases. Carbon nitrogen double bond (R1R2C = NR3) was the functional group in compounds that indicates Schiff Bases. R1 and R2 indicate a side chain of organic origin while R3 binding with nitrogen may be aryl or alkyl group [2]. The condensation of aldehydes (acetaldehyde, benzaldehyde, valeraldehyde, and cinnamon aldehyde) with aromatic amine (aniline) leads to the discovery of Schiff Base first time in the 18th century [3]. Schiff Bases have significant biological activities that are, the presence of unique electron donating and electron accepting functional moieties. Schiff bases possessed many potential biological activities as the compound having Schiff bases have both electrons-accepting and electron-donating groups. Schiff bases have significant biological activities reported that include Antimicrobial activity antidepressant activity [4], Antidyslipidemic activity [5], Anthelmintic activity [6], Antitubercular activity [7], Anticonvulsant activity [8], Anti-inflammatory activity, analgesic activity and non-ulcerogenic activity [9], Antitumor activity [10], Antioxidant activity [11], Antiviral activity [12], Anti-hypertensive activity [13] and Antidiabetic and antiglycation activities. The current study encompasses the antiviral potential of newly synthesized novel compounds (C-1 to C-9) against Avian Influenza (H9) and Avian corona (IBV) viruses. Avian Influenza (H9) viruses are viruses that contain a segmented, RNA genome encoding possessing negative-sense, 10 core proteins, and various proteins belong to the “Orthomyxoviridae” family. Different subtypes were formed as the result of the combination of hemagglutinin (HA), surface proteins, and neuraminidase (NA), for example, H9N2, H1N1, and H5N2 [14]. Avian corona (IBV) virus belongs to the coronavirus class (order Nido-virales, family Corona-viridae, genus Corona-virus) [15] that cause infection in the respiratory tract, kidney, gut, and reproduction of chicken. IBV is an enveloped coronavirus having un-segmented, single-stranded with a positive-sense genome of RNA Corona-viruses class contain the largest RNA virus genomes. Anti-microbial activity that is newly synthesized Schiff-Bases was evaluated against four different bacterial strains Bacillus subtilis (BS), Stenotrophomonas melophilia (SM), Serratia marcescens, (SM) & Escherichia coli. (EC) by using Agar Well Diffusion Method [16]. Different dilutions (200 μg/ml, 300 μg/ml, and 400 μg/ml) of newly synthesized compounds that were dissolved in (DMSO) and samples were evaluated for analysis. The prepared Petri plates having agar medium were incubated at 37 °C in the incubator for 24-h and the result was documented to measure the zone of inhibition (mm) [17]. Seven to Eleven days old Chicken embryonated eggs were purchased from the local hatchery and utilized in antiviral studies by using the inoculation method. The candling of the eggs was done before inoculation. The susceptible viruses were inoculated in the chorioallantoic fluid of the eggs. The broader ends of the candled eggs were drilled with a sterile needle for inoculation. After inoculation, the hole was blocked with the help of molten wax, and eggs were incubated at 37 °C. The allantoic fluid was collected after 48 h of inoculation subjected to the Hemagglutination (HA) test for antiviral studies [18]. In the Hemagglutination test, Chicken Blood was collected in freshly prepared Alsevior Solution and centrifugation was done for 5 min at 4000 rpm. The supernatant solution was discarded and RBCs (Red Blood Cells) were washed with phosphate-buffered saline (PBS) solution and pH was maintained at 7.2. The following step was repeated three times. 1 % suspension was prepared such that 10 μl of packed cells were mixed in 1 ml phosphate-buffered saline (PBS) solution with pH 7.2. After that, the prepared cells were used in performing the standard HA test [19]. The crystal structure of enzymes IBV-PLpro [20], (PDB ID: 4x2z), was obtained from RCSB PDB. All water molecules were removed from the crystallographic structure and polar hydrogen atoms were added utilizing Autodock tools (ADT) version 1.5.6. ADT saved the prepared file in PDBQT format. In case of IBV-PLpro, the grid box was centered on the conserved catalytic triad and surrounding amino acid residues composing ubiquitin Binding Domain and amino acid residues composing subsites 1, 2 and 3. The grid dimensions were 50 × 50 × 50 Å with points separated by 0.5 Å for IBV-PLpro and 40 × 40 × 40 Å with points separated by 0.5 Å. The new derivatives were synthesized as shown in Figure 1, and characterization was done followed by the qualitative and quantitative analysis. Physical characteristics including Color, Odor, Melting Point, and Physical state were studied as shown in Table 1. The Schiff Base formation was verified by the FTIR-spectra of Schiff-Bases. The presence of Schiff Base (–C PBM data was replaced with SVG by xgml2pxml: <glyph-data id="pc-E00C" format="PBM" resolution="300" x-size="12" y-size="9"> 000000000000 000000000000 000000000000 111111111111 000000000000 111111111111 000000000000 000000000000 000000000000 </glyph-data> <svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="20.666667pt" height="16.000000pt" viewBox="0 0 20.666667 16.000000" preserveAspectRatio="xMidYMid meet"><metadata> Created by potrace 1.16, written by Peter Selinger 2001-2019 </metadata><g transform="translate(1.000000,15.000000) scale(0.019444,-0.019444)" fill="currentColor" stroke="none"><path d="M0 440 l0 -40 480 0 480 0 0 40 0 40 -480 0 -480 0 0 -40z M0 280 l0 -40 480 0 480 0 0 40 0 40 -480 0 -480 0 0 -40z"/></g></svg> N) carbon-nitrogen bond peaks at 1692, 1660, 1661, 1667, 1677, 1622, 1633, 1647 and 1669 cm−1 found. The characteristic peaks ensured the formation of the Schiff Base and the absence of peaks of carbonyl and amine groups authenticated the completion of the reaction [21]. The solubility of all the synthesized Schiff Base derivatives was tested in different solvents that included Methanol, Ethanol, Water, Chloroform, and DMSO. The sample ligands were frequently soluble in hydrophilic solvents. Anti-bacterial assay of newly synthesized Schiff Bases C1–C9 (Figure 1) was performed to evaluate the susceptibility. The analytical activity was done on present gram-positive Bacillus subtilis (BS), Stenotrophomonas maltophilia (SM), Serratia marcescens (SM) & Escherichia Coli (EC). Results of antibacterial activity were shown in graphical form as shown in Table 3. In this assay, Antimicrobial-activity was done by the well diffusion method. In this activity, the Cefpodoxime drug was labeled as a standard drug for an antimicrobial relative study concerning all derived moieties. Four different bacterial strains were selected, and activity was checked by using three different concentrations (A), (B), and (C). 1 ml of DMSO as solvent was taken and 5 mg sample/standard dissolved and took 20 μl marked as A, 40 μl marked as B and 60 μl marked a C. 1 mg of Nine Novel compounds C1 to C9 (Figure 1) and parent drug C (Cefpodoxime) were dissolved in 1 ml DMSO solution separately by using Eppendorf tubes to prepare the stock solution. Later 100 μl solutions of each novel compound with an equal volume of viral inoculums were mixed and injected in 7–11 days old chicken embryonated eggs according to the described method. All the protocols followed in this study were approved by the departmental biosafety committee of The Islamia University of Bahawalpur, Pakistan. The newly synthesized derivatives of cefpodoxime indicated that the substitution of aromatic aldehydes and ketones showed that C1 (IC50 ± SEM (μM) = 8.29 ± 0.92) substituted with the hydroxyl group on benzene ring and C7 (IC50 ± SEM (μM) = 6.26 ± 0.62). The results are shown in Table 2 having substituted methoxy group showed active results. The crystal structure of enzymes IBV-PLpro, (PDB ID: 4x2z), was obtained from RCSB PDB. All water molecules were removed from the crystallographic structure and polar hydrogen atoms were added utilizing Autodock tools (ADT) version 1.5.6. ADT saved the prepared file in PDBQT format. In case of IBV-PLpro, the grid box was centered on the conserved catalytic triad and surrounding amino acid residues composing ubiquitin Binding Domain was centered on catalytic dyad (HIS41 and CYS143) and amino acid residues composing subsites 1, 2 and 3. The grid dimensions were 50 × 50 × 50 Å with points separated by 0.5 Å for IBV-PLpro and 40 × 40 × 40 Å with points separated by 0.5 Å. The docking studies of C1, C3, C4, C5, C8 and C9 in 2D and 3D figures were shown below in Figure 2 binding energies, residues, types of interactions, hydrophobic interactions, and electrostatic interactions of C1–C9 newly synthesized compounds during docking studies against IBV Papain like protease protein (PLpro) shown in Table 4. All chemicals with a grade of analytical standard utilized in this current research work were purchased from the following distributors. Cefpodoxime was obtained from Mega Pharmaceutical Pvt. Ltd. Lahore. Its percentage purity was 97%. All the materials, different solvents, and chemicals were purchased from Sigma-Aldrich and Merck international. All list was of analytical grade and utilized without any purification. Thin-layer chromatography (TLC) was done on Merck precoated silica gel, aluminum plates (Kieselgel) for purification. Later, TLC was visualized using a UV lamp, with wavelength of light fixed at 245nm. Melting points were measured by Gallen Kamp melting point apparatus. IR (Infrared) spectra were recorded on FTIR. 1H NMR proton spectra were obtained using NMR spectrophotometer (100 MHz), and 13C NMR spectra were obtained at NMR spectrophotometer (400 MHz). Equimolar solution of available aldehydes/ketones and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h as shown in Scheme 1. at controlled temperature in the water bath, cooled at room temperature and filtration was performed [22]. R1 and R2 of synthesized Schiff Bases were shown in Table 5 The solvent evaporation was performed by a rotary evaporator; the solid by-product was collected and dried at room temperature. Recrystallization done was performed with alcohol (ethanol). Physical Characterization was done by using different physio-chemical procedures. Equimolar mixture of Salicylaldehyde and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml). Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask. The reaction mixture was refluxed for 3 h at solvent boiling point in the water bath, cooled at room temperature and filtration was performed. Brown crystals: Yield (78%), m. p. 103–105 °C, Mol. Wt. 531.56, Elemental Analysis: (Calculated) for C22H21N5O7S2: C, 49.70; H, 3.99; N, 13.28; (Found): C, 47.25; H, 3.78; N,12.99; FTIR (cm−1), 3312, 3439, 3565 ν(NH), 2808, 2881, 2978ν(CH), 1652,1684 ν(C=N), 1608 ν(CH = CH), 1228,1271 ν(C–N), 1162 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 9.33–9.44s(2H), 4.79–4.81d(1H), 4.84–5.20d(1H), (CH), 3.21s(3H), 3.84s(2H), (−CH2), 3.53s(3H), 3.83s(3H); (−CH3), 7.11s (1H), (−NH−), 4.82s(2H), (-NH2), 9.51s,(1H) (OH),13C NMR (DMSO−d6, 400 ppm), δ 25.8 (C-1), 129.5 (C-2), 123.6 (C-3), 61.8 (C-4), 58.8 (C-5), 163.7 (C-6), 159.5 (C-7), 69.7 (C-8), 57.8 (C-9), 162.9 (C-10), 151.7 (C-11), 61.8 (C-12), 149.0 (C-13), 123.6 (C-14), 168.3 (C-15), 160.3 (C-16), 119.4 (C-17), 131.2 (C-18), 122.2 (C-19). 135.1 (C-20). 117.2 (C-21), 160.6 (C-22). Equimolar mixture of Benzaldehyde and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Yellowish brown crystals: Yield (77%), m. p. 113–115 °C, Mol. Wt. 515.09, Elemental Analysis: (Calculated) for C22H21N5O6S2: C, 51.25; H, 4.11; N, 13.58; (Found): C, 50.25; H, 4.01; N, 13.30; FTIR (cm−1), 3309, 3447 ν(NH), 2817, 2886 ν(CH), 1710,1762 ν(C=N), 1660 ν(CH = CH), 1374 ν(C–N), 1274 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 10.02 s (1H), 5.20d (1H), 4.80–4.81d (2H), (CH), 3.204s–3.208s (2H), 3.831–3.839s (2H), (−CH2), 3.84–3.85s (2H), 3.91s (3H); (−CH3), 7.47–7.49s (2H), (−NH−), 4.811–4.819s (2H), (-NH2), 7.93–7.95s, (4H) (OH), 13C NMR (DMSO−d6, 400 MHz), δ 25.7 (C-1), 129.0 (C-2), 123.6 (C-3), 61.8 (C-4), 58.8 (C-5), 163.9 (C-6), 159.5 (C-7), 69.7 (C-8), 57.8 (C-9), 162.9 (C-10), 151.8 (C-11), 61.8 (C-12), 148.9(C-13), 123.6 (C-14), 168.3 (C-15), 159.5 (C-16), 129.1 (C-17), 130.8 (C-18), 128.7 (C-19), 130.8 (C-20), 128.9 (C-21), 129.3 (C-22). Equimolar mixture of Cinnamonaldehyde and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Yellowish brown semi-solid: Yield (71%), Mol. Wt. 541.60, Elemental Analysis: (Calculated) for C24H23N5O6S2: C, 53.22; H, 4.28; N, 12.93; (Found): C, 53.25; H, 4.78; N, 12.97; FT-IR (cm−1), 3301ν(NH), 2824, 2896 ν(CH), 1621–1661 ν(C=N), 1446, 1515 ν(CH = CH), 1012-1035ν (C–N), 1066 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 9.68-9-69d (1H), 7.71–7.74t (4H), 4.79–4.93s (1H), (CH), 3.21–3.22t (3H), 3.84s (1H), (−CH2), 3.71s (2H), 3.84s (3H); (−CH3), 7.71–7.74t (4H), (−NH−), 4.14s (2H), (-NH2), 9.59–9.60d, (2H) (OH), 13C NMR (DMSO−d6, 400 MHz), δ 21.2 (C-1), 130.1 (C-2), 119.2 (C-3), 56.9 (C-4), 56.9 (C-5), 152.8 (C-6), 152.4 (C-7), 143.8 (C-8), 56.9 (C-9), 141.9 (C-10), 152.4 (C-11), 72.5 (C-12), 143.8 (C-13), 119.2 (C-14), 152.8 (C-15), 152.4 (C-16), 119.2 (C-17), 134.1 (C-18), 134.1 (C-19), 128.8 (C-20), 128.6 (C-21), 127.5 (C-22), 128.6 (C-23), 128.8 (C-24). Equimolar mixture of Benzophenone and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Yellowish brown crystals: Yield (71%), m. p. 91–93 °C, Mol. Wt. 591.66, Elemental Analysis: (Calculated) for C28H25N5O6S2: C, 56.80; H, 4.22; N, 11.87; (Found): C, 57.05; H, 4.09; N, 12.30; FT-IR (cm−1), 3273, 3352 ν(NH), 2932,2968ν(CH), 1757ν(C=N), 1667 ν(CH = CH), 1370 ν(C–N), 1268 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 7.72–7.74s (5H), 7.66–7.69s (5H), 4.79–4.81d (1H), (CH), 3.83–3.86s (2H), 3.88–3.93s (2H), (−CH2), 3.51s (3H), 3.83s (3H); (−CH3), 7.55s (1H), (−NH−), 4.15s (2H), (-NH2), 7.74s, (1H) (OH), 13C NMR (DMSO−d6, 400 MHz), 21.2 (C-1), 129.4 (C-2), 128.5 (C-3), 69.7 (C-4), 57.4 (C-5), 151.7 (C-6), 151.7 (C-7), 69.7 (C-8), 57.4 (C-9),151.7 (C-10), 151.7 (C-11), 57.4 (C-12), 137.1(C-13), 128.5 (C-14), 151.7 (C-15), 195.7 (C-16), 195.7 (C-17), 129.4 (C-18), 128.5 (C-19), 132.5 (C-20), 128.5 (C-21), (C-22), 129.4 (C-23), 129.4 (C-24), 128.5 (C-25), 132.5 (C-26), 128.5 (C-27), 129.4 (C-28). Equimolar mixture of Acetophenone and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Red semi solid: Yield (77%), Mol. Wt. 529.59, Elemental Analysis: (Calculated) forC23H23N5O6S2: C, 52.15; H, 4.41; N, 14.38; (Found): C, 52.25; H, 4.30; N, 13.30; FT-IR (cm−1), 3201, 3270 ν(NH), 2810, 2886 ν(CH), 1677 ν(C=N), 1528 ν(CH = CH), 1363 ν(C–N), 1266 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 7.94–7.95 d (5H), 4.800–4.808d (2H), 6.72–6.78d (1H), (CH), 3.15s (1H), 3.16s (2H), 3.84s (2H), (−CH2), 3.25s (3H), 3.83s (3H); (−CH3), 7.52s (1H), (−NH−), 4.80s (2H), (-NH2), 9.51s, (1H) (OH), 13C NMR (DMSO−d6, 400 MHz), δ 25.4 (C-1), 133.0 (C-2), 123.6 (C-3), 61.8 (C-4), 58.8 (C-5), 163.9 (C-6), 159.6 (C-7), 69.7 (C-8), 57.8 (C-9),162.8 (C-10), 151.8 (C-11), 61.8 (C-12), 148.9(C-13), 123.6 (C-14), 168.3 (C-15), 25.4 (C-16), 168.3 (C-17), 136.9 (C-18), 128.0 (C-19), 128.6 (C-20), 133.0 (C-21), 128.6 (C-22), 128.0 (C-23). Equimolar mixture of Formaldehyde and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Light brick crystals: Yield (79%), m. p. 86 °C, Mol. Wt. 439.06, Elemental Analysis: (Calculated) for C16H17N5O6S2: C, 41.15; H, 4.21; N, 16.38; (Found): C, 41.25; H, 4.11; N, 16.32; FT-IR (cm−1), -3275, 3473 ν(NH), 2824, 2971ν(CH), 1666 ν(C=N), 1622 ν(CH = CH), 1097 ν(C–N), 1066 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 9.51 s (1H), 5.16–5.17d (1H), 5.2d (1H), (CH), 1.25–1.26s (2H), 1.49–1.50s (4H) (−CH2), 1.03–1.24s (4H) (−CH3), 7.10s (1H), (−NH−), 4.82s (2H), (-NH2), 9.51s, (1H) (OH), 13C NMR (DMSO−d6, 400 MHz), δ 25.4 (C-1), 128.9 (C-2), 123.6 (C-3), 61.9 (C-4), 58.8 (C-5), 163.9 (C-6), 159.5 (C-7), 69.7 (C-8), 57.8 (C-9),162.8 (C-10), 151.9 (C-11), 61.9 (C-12), 148.9(C-13), 123.6 (C-14), 171.7 (C-15), 162.8(C-16). Equimolar mixture of Vanillin and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Brownish amorphous powder: Yield (74%), m. p. 96 °C, Mol. Wt. 561.59, Elemental Analysis: (Calculated) for C23H23N5O8S2: C, 49.19; H, 4.13; N, 12.47; (Found): C, 48.21; H, 4.31; N, 14.21; FT-IR (cm−1), 3201, 3429 ν(NH), 2881 ν(CH), 1633 ν(C=N), 1540 ν(CH = CH), 1370 ν(C–N), 1239 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 9.77 s (1H), 7.39–7.42d (2H), 6.95–6.96d (2H), (CH), 3.15–3.21s (3H), 3.84s (2H), (−CH2), 3.83s (3H), 3.83s (3H), 3.84s (3H); (−CH3), 7.39s (1H), (−NH−), 3.79d (2H), (-NH2), 9.77s, (1H) (OH), 13C NMR (DMSO−d6, 400 MHz), δ 25.4 (C-1), 128.8 (C-2), 123.8 (C-3), 62.0 (C-4), 57.4 (C-5), 153.7 (C-6), 153.0 (C-7), 69.7 (C-8), 57.4 (C-9),153.7 (C-10), 151.7 (C-11), 62.0 (C-12), 148.2 (C-13), 123.8 (C-14), 153.7 (C-15), 153.0 (C-16), 128.8 (C-17), 123.8 (C-18), 115.4 (C-19), 151.7 (C-20), 148.2 (C-21), 111.8 (C-22), 56.9 (C-23). Equimolar mixture of Dimethyl Formamide and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Yellowish brown crystals: Yield (78%), m. p. 102–106 °C, Mol. Wt. 482.53, Elemental Analysis: (Calculated) for C18H22N6O6S2: C, 44.80; H, 4.60; N, 17.42; (Found): C, 46.20; H, 4.50; N, 18.21; FT-IR (cm−1), 3380 ν(NH), 2911, 2977 ν(CH), 1647 ν(CN), 1615 ν(CHCH), 1374 ν(C–N), 1194 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 9.49–9.53 s (2H), 4.14–4.15d (4H), 5.19–5.20d (2H), (CH), 3.20s, 3.49s (2H), 3.84s (2H), (−CH2), 3.84t (3H), 3.83s (3H); (−CH3), 7.94s (1H), (−NH−), 4.08d (2H), (-NH2), 9.49–9.53d, (1H) (OH), 13C NMR (DMSO−d6, 400 MHz), δ 25.4 (C-1), 91.7 (C-2), 92.0 (C-3), 61.8 (C-4), 58.8 (C-5), 162.2 (C-6), 159.6 (C-7), 69.7 (C-8), 57.4 (C-9),162.2 (C-10), 151.7 (C-11), 61.8 (C-12), 92.0(C-13), 91.7 (C-14), 168.4 (C-15), 151.7 (C-16),35,7 (C-17), 35.7 (C-18). Equimolar mixture of Dimethylaminobenzaldehyde and Cefpodoxime drug were added in 250 ml capacity round bottom flask in ethanol (30 ml) used as a solvent. Few drops (Five to Six) of glacial acetic acid were also dropped in a round-bottom flask as the catalyst. The reaction mixture refluxing for 03 h at controlled temperature in the water bath, cooled at room temperature and filtration was performed. Brown crystals: Yield (79%), m. p. 130 °C, Mol. Wt. 558.63, Elemental Analysis: (Calculated) for C24H26N6O6S2: C, 51.60; H, 4.69; N, 15.04; (Found): C, 51.25; H, 4.50; N, 15.30; FT-IR (cm−1), 3308, ν(NH), 2969ν (CH), 1760 ν(CN), 1669 ν(CHCH), 1372 ν(C–N), 1271 ν(C–O), 1H NMR (DMSO−d6, 400 MHz); δ 9.67s (1H), 9.50–9.54t (4H), 4.78–4.80d (4H), (CH), 3.20s–3.21s (2H), 3.84–3.85t (4H), (−CH2), 3.84–3.85s (2H), 3.91s (3H); (−CH3), 7.66–7.81s (4H), (−NH−), 4.78–4.80d (1H), (-NH2), 8.27, (1H) (OH),13C NMR (DMSO−d6, 400 MHz), 25.4 (C-1), 130.5 (C-2), 123.3 (C-3), 60.0 (C-4), 57.4 (C-5), 168.6 (C-6), 154.2 (C-7), 69.7 (C-8), 57.4 (C-9), 168.6 (C-10), 153.9 (C-11), 62.0 (C-12), 152.3(C-13), 123.3 (C-14), 168.6 (C-15), 154.2 (C-16), 130.5 (C-17), 131.4 (C-18), 124.6 (C-19), 130.5 (C-20), 124.6 (C-21), 124.6 (C-22), 49.7(C-23), 49.7(C-24). New novel compounds were synthesized by a condensation reaction, that is a single-step reaction in a drug having a primary amine carbon group was condensed with nine different aldehydes/ketones by reflux distillation method under controlled temperature. The obtained compounds C1–C9 were dried with the help of a rotary evaporator. The percentage yield was calculated by using weighing apparatus. The physical and chemical characteristics (color, odor, physical form, solubility, and melting point) were studied. Solubility was determined in Methanol, Ethanol, Chloroform, Water, DMSO, n-Hexane, and n-Butanol by using Sonicator. The melting point was determined by the Gallen Kamp apparatus. Novel synthesized compounds C1–C9 (Table No. 5), synthesis was confirmed by spectroscopic techniques such as FTIR (Fourier Transform Infrared spectroscopy), H1-NMR (Proton Nuclear Magnetic spectroscopy), and C13 Nuclear Magnetic Spectroscopy. Antibacterial studies against four bacterial strains (Bacillus subtilis, Stenotrophomonas maltophilia, Serratia marcescens, and E. Coli) were performed and compared with standard antibacterial drug Cefpodoxime i.e. the parent drug from which all nine-novel compounds were synthesized. C1 and C2 have average antibacterial activity against Bacillus subtilis as compared to the parent drug cefpodoxime. C1, C2, C4, and C6 showed effective activity against Stenotrophomonas maltophilia in comparison to parent drug Cefpodoxime (C). Cefpodoxime as control showed average activity and C3, C5, C6, C7, C8, and C9 showed no activity on any concentration in comparison with Cefpodoxime. C1, C2, C4, and C6 showed effective activity against Serratia marcescens in comparison of Cefpodoxime (C), Cefpodoxime (C) as control showed average activity and C3, C5, C6, C7, C8, and C9 showed no activity on any concentration in comparison with Cefpodoxime. C1, C2, C3, C5, C6, C8, and C9 showed effective activity against E. coli, whereas Cefpodoxime C as control showed average activity and C7 showed no activity on any concentration in comparison with Cefpodoxime. All newly synthesized compounds C1, C2, C3, C4, C5, C6, C7, C8, and C9 showed highly strong antiviral potential against Avian Influenza (H9) and Avian corona (IBV) viruses. The grid box was centered on the conserved catalytic triad and surrounding amino acid residues composing ubiquitin Binding Domain. Most of the molecular interactions were localized in the thumb (residues 56–168), finger (residues 169–231), and palm (residues 232–310) domains of IBV PLPro. Significant interactions were also observed with residues composing the Ubiquitin Binding Domain of IBV PLPro. C1 displayed binding energy of -7.8 kcal/mol. The carboxylate –OH (O28) donated H∗ to the carbonyl group of ASN160 producing only one H-bond. Further, three weak C–H interactions were formed with residues of the palm domain i.e.; –OH of the o-cresol functional group donated hydrogen to carbons CD of PRO241 and CA of GLY240. Third, C–H linkage was established between the methoxamine group (C24) and THR238. One H-bond was donated by the ND2 amino group of ASN155 to a nitrogen atom (N21) linking the cresol group. The electron-deficient Thiazole ring system was stabilized by π-anion interaction with –OH of ASP153. The aromatic system of the o-cresol ring was stabilized by stacking with PHE 256 and the alkyl group of ILE290. The delocalized π-electron density of the indole ring of TRP156 established hydrophobic linkages with Sulphur and carbon atoms of the β-lactam ring. In the case of C2, styrene linked Thiazole portion was deeply embedded in Ubiquitin Binding Domain. PHE256 stabilized the aromatic styrene via π-stacking which was further strengthened by π-CH contact with ILE 290. H-bond was also established between the amino group of ASP153 and nitrogen atom linking styrene to Thiazole ring. The π electron density of the indole ring of TRP156 stabilized the conformation by hydrophobic contact with the sulfur atom of the Thiazole moiety. Further stability was achieved by donation of lone pair from OH hydroxyl group of SER152 to electron-deficient aromatic Thiazole ring. The carboxylate –OH (O28) further anchored by forming two H-bonds with ASP150 and adjacent residue GLY149. The amide of PHE151 donated hydrogen to the carbonyl oxygen of the β-lactam ring. Weak C–H interactions were also observed between methoxamine carbon (C24) and ASP150; beta-lactam carbon and ASN261. The Diphenylmethane moiety of C4 displayed hydrophobic interactions with Blocking Loop 1 (BL-1) composed of residues (ASP245-VAL251) linking finger and palm domains. Both rings of diphenylmethane established π-CH contacts with ALA250, CYS246, and LYS285. Further stability was achieved by an edge to face π-stacking with the aromatic ring of PHE283 and π-anion contact with –OH (OE2) of GLU248. the β-lactam ring was deeply embedded in the thumb domain with carboxylate –OH (O28) accepting H-bond from an amino group (ND2) of ASN90. Another hydrophobic contact was established between the π-electron of the thiazine ring and the alkyl group of LYS114. The β-lactam ring system of C5 sustained only weak interactions like the C–H bond with –OH (OD2) of ASP153 and π-CH contact with ALA159. The alkyl group of ethylbenzene substituent (C36) firmly anchored the molecule in the thumb and UBD Domain by maintaining three alkyl-alkyl contacts with ALA154, CYS265, and ILE290. The aromatic ring systems of phenylalanine residues 151 and 256 stabilized the conformation simultaneously via stacking with π-electrons of the ethylbenzene ring and interacting with –CH3 electrons. π-CH contact was also formed between the alkyl group of ILE290 and the π-electron density of the Thiazole ring system. C8 displayed interactions with thumb domain amino acid residues. Two H-bonds were formed by β-lactam carboxylate –OH (O28) with OG1 –OH of THR78 and the carbonyl oxygen of ILE74. β-lactam carbonyl O27 also accepted H-bond from the amino group of VAL148 while also maintaining weak C–H interaction with CA of LYS147. Methoxy oxygen O2 was also hydrogen loving from amino group NZ of LYS147. H-bond was established between N21 nitrogen atom attached to dimethylamine substituent with LYS82. CA of GLN 81 formed a weak C–H bond with N21. NE2 amino group of GLN81 donated H-bond to O14 of carbonyl group adjacent to the Thiazole ring system. π-electrons of Thiazole ring further interacted with –CH3 electrons to ILE39 to stabilize the interaction. C9 constituted only two linkages in UBD Domain via H-bonding of residues ASP153 and ASN155 with carboxylate –OH (O28) attached to the thiazine ring system. The π-electron cloud of PHE236 stabilized the Thiazole ring simultaneously via edge to face π-stacking and interacting with ring sulfur. Alkyl group electrons of ALA159 further anchored Thiazole ring by π-CH interactions whilst alkyl group electrons of ILE196 interacted with π-electrons of N, N-dimethylaniline substituent. C29 of N, N-dimethylaniline substituent exhibited weak C–H interaction with ILE296. Newly synthesized Schiff base derivatives showed potential antiviral potential against Corona class virus variant Avian Coronavirus (IBV) along with antibacterial potential against selective strains. The compounds also have hepatoprotective characteristics as compounds have significant antioxidant potential. Some of the compounds show significant anti-urease potential. Further clinical studies may lead to a breakthrough in medical sciences and the market will have potential drugs against complex infective diseases. Waqas Mahmood: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper. Mohsin Abbas Khan: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper. Irshad Ahmad: Performed the experiments; Analyzed and interpreted the data; Wrote the paper. Syed Adnan Ali Shah, Mirza Imran Shahzad, Muhammad Ashraf, Irfan Pervaiz, Muhammad Sajid-ur-Rehman and Umair Khurshid: Contributed reagents, materials, analysis tools or data. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Data will be made available on request. The authors declare no conflict of interest. No additional information is available for this paper.

PMC9649983

Dae D. Chung,Marisa R. Pinson,Amanda H. Mahnke,Nihal A. Salem,Khang T. Le,Elizabeth A. Payne,Tenley E. Lehman,Susan T. Weintraub,Rajesh C. Miranda

Dose-related shifts in proteome and function of extracellular vesicles secreted by fetal neural stem cells following chronic alcohol exposure

01-11-2022

Prenatal alcohol exposure,Fetal alcohol spectrum disorder,Extracellular vesicle,Neural stem cell,Developmental,Proteome,Exosome,Alcohol,Sex,WGCNA

Accumulating evidence indicates that extracellular vesicles (EVs) mediate endocrine functions and also pathogenic effects of neurodevelopmental perturbagens like ethanol. We performed mass-spectrometry on EVs secreted by fetal murine cerebral cortical neural stem cells (NSCs), cultured ex-vivo as sex-specific neurosphere cultures, to identify overrepresented proteins and signaling pathways in EVs relative to parental NSCs in controls, and following exposure of parental NSCs to a dose range of ethanol. EV proteomes differ substantially from parental NSCs, and though EVs sequester proteins across sub-cellular compartments, they are enriched for distinct morphogenetic signals including the planar cell polarity pathway. Ethanol exposure favored selective protein sequestration in EVs and depletion in parental NSCs, and also resulted in dose-independent overrepresentation of cell-cycle and DNA replication pathways in EVs as well as dose-dependent overrepresentation of rRNA processing and mTor stress pathways. Transfer of untreated EVs to naïve cells resulted in decreased oxidative metabolism and S-phase, while EVs derived from ethanol-treated NSCs exhibited diminished effect. Collectively, these data show that NSCs secrete EVs with a distinct proteome that may have a general growth-inhibitory effect on recipient cells. Moreover, while ethanol results in selective transfer of proteins from NSCs to EVs, the efficacy of these exposure-derived EVs is diminished.

Dose-related shifts in proteome and function of extracellular vesicles secreted by fetal neural stem cells following chronic alcohol exposure Accumulating evidence indicates that extracellular vesicles (EVs) mediate endocrine functions and also pathogenic effects of neurodevelopmental perturbagens like ethanol. We performed mass-spectrometry on EVs secreted by fetal murine cerebral cortical neural stem cells (NSCs), cultured ex-vivo as sex-specific neurosphere cultures, to identify overrepresented proteins and signaling pathways in EVs relative to parental NSCs in controls, and following exposure of parental NSCs to a dose range of ethanol. EV proteomes differ substantially from parental NSCs, and though EVs sequester proteins across sub-cellular compartments, they are enriched for distinct morphogenetic signals including the planar cell polarity pathway. Ethanol exposure favored selective protein sequestration in EVs and depletion in parental NSCs, and also resulted in dose-independent overrepresentation of cell-cycle and DNA replication pathways in EVs as well as dose-dependent overrepresentation of rRNA processing and mTor stress pathways. Transfer of untreated EVs to naïve cells resulted in decreased oxidative metabolism and S-phase, while EVs derived from ethanol-treated NSCs exhibited diminished effect. Collectively, these data show that NSCs secrete EVs with a distinct proteome that may have a general growth-inhibitory effect on recipient cells. Moreover, while ethanol results in selective transfer of proteins from NSCs to EVs, the efficacy of these exposure-derived EVs is diminished. The developing cerebral cortical ventricular (VZ) and subventricular (SVZ) zones generate most of the projection neurons of the adult cerebral cortex during a restricted developmental window, corresponding to the latter half of gestation in mouse, and to the mid-first through second trimester in humans [1]. The generation of neurons from neural stem cells (NSCs) during this critical period can be disrupted by the presence of chemical perturbagens in the maternal-fetal environment, and may lead to, for example, diminished neuron number, microencephaly, and intellectual disability. Alcohol (ethanol) exposure during pregnancy is an example of an important, health-relevant perturbagen, since ∼9.8 % or more of all pregnancies globally are exposed to alcohol [2]. Ethanol is the causal agent in a spectrum of brain and craniofacial anomalies and growth deficits that are collectively termed Fetal Alcohol Spectrum Disorders (FASD) [3], which are prevalent with between 3.1 and 9.9 % of school-aged children in the US [4]. Studies in human populations [5, 6, 7, 8] and in animal models [9, 10, 11] have shown that developmental alcohol/ethanol exposure can result in decreased brain size or microencephaly. Moreover, previous studies document that fetal NSCs in developing regions like the cerebral cortex do not die following ethanol exposure, but rather undergo a proliferative burst of rapid transit amplification and premature maturation resulting in the depletion of NSCs [12, 13, 14, 15, 16, 17, 18, 19, 20, 21]. Recent studies from our group suggest that some of the aberrant pro-maturation effects of ethanol may be mediated by microRNAs (miRNAs) secreted by NSCs within a class of small membrane-bound extracellular vesicles (EVs) [22, 23]. EVs are secreted by most cell types and are thought to participate in paracrine and endocrine transfer of proteins, lipids, DNAs, and RNAs to other cells, without any direct cell-to-cell contact [24, 25, 26, 27, 28, 29, 30]. The existence and functions of EVs were first discovered in the mid-20th century as procoagulant platelet-derived particles that could be isolated by ultracentrifugation from plasma [31] and as lipid-enriched particles [32]. Interest in EVs was amplified by discoveries in 2006–2007 that these particles could transfer RNAs between cells as a paracrine or endocrine signal [33, 34]. EVs have significant value as biomarkers, but may also be therapeutic vehicle for disease management [35]. For instance, EVs have been identified as intercellular signaling mediators in osteosarcoma and mesenchymal stromal cell niches, where they modulate a variety of regulatory pathways [36, 37, 38]. Several studies have found that the miRNA and protein contents of EVs can be taken up, and influence a recipient cell's physiology [39, 40, 41, 42, 43, 44]. In our previous studies, we reported that ethanol-induced EV miRNAs such as miR-140-3p could mediate aberrant astroglial differentiation, while suppressing neuronal and oligodendroglial lineage markers [23], and retroviral-like proteins present in EVs, like PEG10 and PNMA2, could lead to apoptosis-resistance [22]. Moreover, EV reprogramming has been implicated in microglial activation and inflammation-mediated loss of developing hypothalamic neurons [45], and in the pathogenesis of alcohol-associated liver disease [46, 47], suggesting that EVs are a component of a broader cellular and tissue sensitivity to this common perturbagen. These observations advance a novel possibility that EVs secreted by NSCs may reflect NSC reprogramming by the environment and, consequently, transfer the effects of a perturbagen between cells, thereby contributing to the spread and persistence of deleterious effects within the neurogenic niche. In our studies reported here and previously [23], we focus on EVs secreted by neural stem cells dissected from sex-specified fetal mouse iso-cortex and maintained ex vivo, as neurosphere cultures to model the early period of cortical plate neurogenesis. Assessed EVs are less than 200 nm in diameter, with a median size of 150 nm, a population that may include both exosomes and microvesicles, and express known markers for exosomes. We previously reported that ethanol exposure did not significantly change the concentration and size of EVs secreted by NSCs, nor did exposure influence the expression of at least one key EV marker, CD63 [23]. In the current study, we therefore assessed whether ethanol exposure altered the protein cargo of EVs relative to their parental NSCs. We used mass spectrometric analysis of proteins from EVs and their parental NSCs, from both male and female mouse fetuses, to assess the contribution of both genetic sex and ethanol dose on protein abundance in EVs. Exploratory weighted gene co-expression network analysis (WGCNA) identified candidate proteins and gene networks that were drivers of alterations to the proteomes. These analyses showed that the EV proteome is substantially different from the proteome of parental cells, with ethanol exposure and genetic sex as additional contributory factors to modulation of the EV proteome. EVs were also found to sequester peptides for proteins from every parental cell sub-compartment, from nucleus to cytoplasm to membrane. Surprisingly, we identified proteins, from every cellular sub-compartment, that were enriched in EVs compared to their cell of origin, suggesting that these proteins, collectively serving developmental programs, were preferentially sequestered into EVs. Ethanol exposure, in a dose-dependent manner, resulted in a preferential enrichment of specific proteins into EVs, as these proteins were decreased intracellularly and increased in EVs. This suggests a perturbagen-dependent trafficking of proteins from cells to EVs. A moderate dose of ethanol resulted in preferential enrichment of proteins in EVs related to ribosomal RNA processing and cell cycle, whereas a high dose of ethanol resulted in EV enrichment of proteins associated with mTOR activation due to amino acid deprivation and DNA replication. These data suggest the presence of a systematic cell stress response in NSCs that results in the transfer of networks of related proteins into EVs in a dose- and sex-dependent manner. The heterogeneity of EVs reported in the literature, likely reflecting the molecular diversity of their parental cells, as well as the heterogeneity in technical approaches for EV isolation [48], caused us to define EVs by their size, density, biochemical composition, and cell of origin type. Therefore, in this study, our ultracentrifuged EV population was defined as female or male mouse fetal NSC-derived, obtained under either control or ethanol-exposed conditions (Figure 1A,B). Transmission electron microscopy showed that the NSC-derived EVs were round to oval in shape and were CD63+ as indicated by immuno-gold labeling localized to the EV surface (Figure 2A). Nanoparticle tracking analysis (NTA) showed that after 72 h, culture-conditioned media contained approximately ∼109 EV per million seeded cells, or a net release of ∼10–14 EVs/hour/seeded cell, ranging in diameter from 50–200 nm with a median diameter of 150 nm, consistent with the known size range for exosomes (Figure 2B) [29, 49]. Immunoblot analyses demonstrated that, compared to parental cells, the EV fraction from our NSC supernatant was enriched in transmembrane and cytosolic proteins commonly used as EV markers: CD9, CD63, Tsg101, annexin VI, and Rab5 (Figure 2C,E and Supplementary Figure 1) [49]. In contrast, proteins such as calnexin, cytochrome c, and Ago2 which are negative EV markers were present in our parental NSCs but absent or detected at low levels in our EVs (Figure 2D,F and Supplementary Figure 2). A third group of markers, PPP2R1A, VCAM1, TIM50, and Ndufs1, was selected based on our mass spectrometric evidence for their presence in both NSCs and their secreted EVs. In all cases, immuno-reactive bands of identical molecular weight were observed in both EV and cell samples (Supplementary Figure 3), suggesting that full-length proteins from cells are packaged into EVs. These data also collectively show that our isolation method is efficient in collecting CD63+/CD9+/TSG101+ EVs from NSCs, while eliminating non-EV fractions and other cell contaminants. Using data-independent acquisition mass spectrometry for global protein identification and relative quantification, there were 4703 unique proteins in EVs and 6262 unique proteins in cells (Supplementary Table 1–3). Of these, 2765 proteins were identified in all 18 EV samples and were used for further bioinformatic analysis (Supplementary Table 4). Abundance values were imputed for proteins that were present in EVs but undetectable in parental cells in order to compare protein levels in EVs to the parent cell-of-origin. Abundance values for each protein were z-score transformed across samples (Supplementary Table 5). We matched protein names (UniProt ID) to their corresponding gene IDs (ENTREZ ID), used WGCNA to construct gene co-expression networks to identify clusters (modules) of interconnected genes and the hub genes within each module, and performed hierarchical clustering using Pearson correlation. All 36 samples including EV and parental NSC samples were included in this analysis. Five distinct modules were identified; the brown, yellow, blue, green, and turquoise modules (colors for module labeling purpose) contained 170, 120, 545, 82, and 1698 correlated genes, respectively (Figure 3A). The most highly connected hub genes for brown, yellow, blue, green, and turquoise modules were Taok3, Cpe, Itga7, Pxdn, and Snx5, respectively. Using multidimensional scaling (MDS) to visualize the co-expression networks, we observed highly correlated gene expression across brown, yellow, blue, and green modules, whereas genes contained within the turquoise module did not exhibit a correlated expression with any other module (Figure 3B). In order to assess the relationship between each module and sample traits (Location: EV vs. cell, pregnancy biological replicates denoted as set numbers, sex, and ethanol treatment), we identified modules that are significantly associated with sample traits by correlating module eigengenes (ME; 1st principal component of a given module) with traits (Figure 3C). We observed that there was a significant correlation between all MEs and sample trait of Location (EV_Cell), with eigengenes of blue (r = 0.93, p= 5 × 10−16), green (r = 0.91, p= 3 × 10−14), yellow (r = 0.65, p= 2 × 10−05), and brown (r = 0.51, p= 0.002) modules having significant positive correlations with Location (i.e., higher in EVs relative to parental NSCs) while eigengene of turquoise (r = −0.98, p= 5 × 10−25) module had a significant inverse correlation with Location (i.e., higher in parental NSCs relative to EVs). For sample trait of sex, only eigengenes of the brown module had a significant positive correlation (r = 0.52, p= 0.001). In preparation for enrichment analysis, to confirm the significant correlation between modules and the sample trait Location, we examined the correlation of MEs with Location (Supplementary Figure 4A), Location trait-based average protein/gene significance (GS) across modules (Supplementary Figure 4B), and the correlation between GS and module membership (Supplementary Figure 4C–G). GS for EV correlated with module membership in the blue module (r = 0.72, p= 3.4 × 10−88), green module (r = 0.89, p= 5.2 × 10−29), yellow module (r = 0.64, p= 3.6 × 10−15), and brown module (r = 0.22, p= 0.0039), while GS for Cell correlated with module membership (MM) in the turquoise module (r = 0.98, p < 1 × 10−200). This analysis suggests that proteins/genes highly significantly associated with EV samples were the most important element of the blue, green, and yellow modules, while those highly significantly associated with cell samples were the most important element of the turquoise module. Each protein/gene's module color, GS to Location trait, and MM can be found in supplementary materials (Supplementary Table 6). Finally, enrichment analysis was performed for proteins/genes in all five modules to study biological mechanisms (Supplementary Table 7). These analyses indicated that the EV-enriched modules, blue, green, yellow, and brown, were highly enriched for proteins related to negative regulation of cell development and differentiation, Ras protein signal transduction, immune response regulation, and protein localization and transport, respectively. Collectively, through WGCNA, we observed that proteins grouped strongly by sample type of EV and Cell, suggesting proteins loaded into EVs have a different profile compared to protein constituents of the parental cells and may support distinct biological processes. Using a reference proteome database (UniProt Knowledgebase) [50], proteins were characterized by their preferential localization within the cell to six subcellular compartments: the nucleus, cytoplasm, mitochondria, lysosomes, endosomes, and membrane. We found that proteins from each of these subcellular compartments were also localized to EVs. Moreover, we observed a number of proteins from each sub-cellular compartment (Nucleus (Figure 4A), Cytoplasm (Figure 5A), Mitochondria (Figure 6A), Lysosome (Figure 7A), Endosome (Figure 8A), Membrane (Figure 9A)) that were, surprisingly, enriched in EVs relative to their parental NSCs in Nucleus (Figure 4C vs. 4D), Cytoplasm (Figure 5C vs. 5D), Mitochondria (Figure 6C vs. 6D), Lysosome (Figure 7C vs. 7D), Endosome (Figure 8C vs. 8D), Membrane (Figure 9C vs. 9D). For protein list associated with each subcellular compartment, see Supplementary Table 8. Two-way ANOVA with FDR correction (a = 0.05) followed by a Šidák's Test for post-hoc multiple comparisons was conducted to examine global differences in average z-score transformed relative protein abundance in EV compared to parental NSCs. We found a significant interaction effect between EV vs. cell samples and female vs. male samples for many subcellular compartments (Nucleus: F(1, 32) = 15.09, p.adj = 0.002; Cytoplasm: F(1, 32) = 8.325, p.adj = 0.010; Mitochondria: F(1, 32) = 4.934, p.adj = 0.034; Lysosome: F(1, 32) = 13.969, p.adj = 0.002; Endosome: F(1, 32) = 7.527, p.adj = 0.012; Membrane: F(1, 32) = 8.587, p.adj = 0.010) (Supplementary Table 9). A Šidák's post-hoc test revealed significant pairwise mean differences (Diff) in average z-score between female-derived NSCs vs. EVs and male-derived NSCs vs. EVs, with a positive Diff indicating overall enrichment in EVs compared to cell (Nucleus: female: Diff = 0.56, p.adj < 0.0001, male: Diff = 0.95, p.adj < 0.0001 (Figure 4B); Cytoplasm: female: Diff = 0.45, p.adj < 0.0001, male: Diff = 0.80, p.adj < 0.0001 (Figure 5B); Mitochondria: female: Diff = 1.11, p.adj < 0.0001, male: Diff = 1.34, p.adj < 0.0001 (Figure 6B); Lysosome: female: Diff = -0.14, p.adj = 0.63, male: Diff = 0.71, p.adj = 0.0002 (Figure 7B); Endosome: female: Diff = -0.37, p.adj = 0.063, male: Diff = 0.27, p.adj = 0.21 (Figure 8B); Membrane: female: Diff = -0.33, p.adj = 0.016, male: Diff = 0.15, p.adj = 0.36 (Figure 9B)). Such significant differences in the average z-score between EV and cell samples by subcellular location suggest that the EV proteome varies from the cell proteome. To determine what significant biological functions these proteins enriched in EVs relative to cells collectively serve, we subjected the enriched proteins to pathway analysis. This analysis of proteins that reached criteria for EV enrichment, where averaged z-score of the protein is higher in EV samples relative to cell samples with nonoverlapping 95% confidence estimates, identified a number of significant biological pathways (Figure 10, Supplementary Figure 5 and Supplementary Table 10,11). These EV-enriched proteins were associated with pathways for endocytosis and the Endosomal Sorting Complex Required For Transport (ESCRT), both of which are essential in exosome formation. Receptor tyrosine kinases signaling and beta-catenin independent WNT signaling, both involved in intracellular and intercellular communication as signal transducers [51, 52], were also significantly overrepresented. Additionally, the planar cell polarity/convergent extension (PCP/CE) pathway and the regulation of RUNX3 expression and activity pathway, are involved in planar cell polarity for cell migration [53, 54] and in neural stem cell proliferation and differentiation [55], respectively, were significantly overrepresented. Overall, our data show that, in addition to proteins that are important for EV biogenesis, proteins enriched in EVs may serve as cell-to-cell signal transducers to regulate cell fate, proliferation, and migration during fetal development. Ethanol is consumed, and exhibits its psychological effects in the millimolar range, and can be tolerated by humans at doses in excess of 100 mM [56, 57]; therefore, our studies utilized a range of mM concentrations of ethanol to mimic both moderate and heavy exposure. At these translationally relevant doses, as used in the present study, we previously showed that ethanol does not induce cell death, but rather, aberrant maturation and loss of stem cell identity [13, 14, 17, 19, 58]. Our previous transcriptomic studies also suggested that developmental ethanol exposure can disrupt entire gene networks [12, 59]. We therefore used WGCNA to assess the contribution of ethanol exposure to alteration of protein networks in both parental NSCs and their secreted EVs. We classified samples by location (EV vs. cell) and then sub-classified by ethanol treatment (0, 120, and 320 mg/dL), and constructed WGCNA analyses for each of the resulting 6 groups (Figure 11A). The TOM heatmap plots and module counts can be viewed as an index of biological complexity. A chi-square test was performed on the ratio of modules in EVs relative to cells to examine the relationship between ethanol treatment and number of modules across locations (EV vs. cell), using the control values as percent expected (Figure 11B). The number of modules in EVs relative to cells was significantly increased in the high ethanol-exposure group compared to controls (X2 (1) = 18.316, p < 0.0001) and trended towards a significant increase for the medium exposure group relative to controls (X2 (1) = 3.247, p = 0.072). Though EV samples exhibited less overall complexity compared to parental cell samples, ethanol treatment resulted in a dose-related increase in the complexity of EV samples, without overall changes in the protein network complexity in parental NSCs. We next examined the effects of ethanol, a salient developmental perturbagen, on the proteome of NSCs and their EVs. NSCs exposed to moderate (120 mg/dL) and heavy (320 mg/dL) ethanol exhibited a significantly altered profile of proteins packaged within EVs. For the 2765 proteins expressed in all 18 EV samples, moderate and heavy ethanol exposure altered the relative abundance of 68 and 110 proteins (paired t-test, p < 0.05), respectively, not accounting for FDR correction of multiple comparisons (Figure 12A,B and Supplementary Figure 6A,B and Supplementary Table 12). We observed that, for the proteins in EVs whose relative abundances were altered, 44 proteins (∼65%) following moderate EtOH exposure and 47 proteins (∼43%) following heavy EtOH exposure, were significantly increased relative to controls. In contrast, for cells, moderate and heavy EtOH exposure altered 223 and 343 proteins, respectively, and the majority of them, 175 proteins (∼78%) in the moderate exposure group and 284 proteins (∼83%) in the heavy exposure group, were at significantly lower abundance (Figure 12C,D and Supplementary Figure 6C,D and Supplementary Table 13). In assessing the contrasting effect of ethanol, i.e., proteins at relatively higher abundance in EVs and lower in cells, we hypothesized that ethanol may have increased the transfer of specific proteins into EVs. To test this possibility, we assessed the effects of ethanol on the enrichment of proteins in EVs by calculating the EV to cell protein expression ratio (EV/cell) for each treatment group, followed by a paired comparison of the EV/cell ratio between the control group and each treatment group. For moderate ethanol exposure, the EV/cell ratio was significantly altered for 57 proteins (using the combined criteria of paired t-test, p < 0.05; Hedges’ g < -0.4 or g > +0.4 with a non-zero containing 95% confidence estimate), with the ratio significantly increased for ∼77% of the differentially altered proteins (44 proteins out of 57) relative to controls (Figure 12E and Supplementary Table 14). Likewise, for heavy ethanol exposure, the EV/cell ratio was significantly altered for 58 proteins, with 51 (∼88%) significantly increased relative to control (Figure 12F and Supplementary Table 15). We observed that both doses of ethanol consistently increased the EV/cell ratio for a number of specific proteins. Volcano plots of EV/cell enrichment effect size, g, vs. p-value (Figure 12E,F) further support our observations that ethanol exposure results in EV enrichment and/or cellular depletion of a selected group of proteins. These data imply that ethanol results in the transfer of specific proteins from parental NSC to secreted EVs. To determine whether proteins whose relative abundances were enriched in EVs but decreased in cells, due to ethanol exposure, collectively served shared biological functions, we subjected the enriched proteins to pathway analysis. This analysis of proteins that reached criteria for EV/Cell enrichment (paired t-test, p < 0.05; Hedges' g > 0.4, with non-zero containing 95% confidence estimate) identified significant biological pathways (Figure 13A, B and Supplementary Table 16–19). For moderate ethanol exposure, pathways associated with rRNA processing were over-represented. Cell cycle-related pathways including DNA replication and G1/S phases and transition, aligning with previously published literature on the developmental effects of ethanol exposure on cell cycle [19, 60], were also significantly overrepresented (Supplementary Table 16,17). Nonsense mediated decay (NMD), an error surveillance pathway that eliminates premature stop codon-containing mRNA transcripts among other functions [61, 62, 63], was also significantly overrepresented. In the case of heavy ethanol exposure, surprisingly, a criterion cutoff of Hedges' g > 0.4 did not result in the identification of enriched pathways. However, increasing stringency to Hedges’ g > 0.5 did result in the identification of stress and amino-acid deprivation pathways related to mTOR signaling, that have previously been implicated in the pathogenesis of prenatal ethanol exposure [64, 65], as well as RHO GTPase signaling, an important inhibitor of NSC migration [66] (Supplementary Table 18). Additional network analyses (Figure 13C,D) document the contribution of individual proteins to core overrepresented pathways. Overall, our data show that medium and high levels of ethanol exposure result in a dose-dependent preferential transfer of distinctly different classes of protein, that are a part of largely nonoverlapping biological pathways, from parental NSCs to their secreted EVs. We next wanted to determine the identity of cell-types, in vivo, within the developing cerebral cortex, that were likely contributors to the increased ethanol-dependent sequestration of proteins within EVs. We reasoned that the cell subtype identity would also include mRNA transcripts for ethanol-sensitive EV proteins. Using our previously published scRNAseq dataset (GSE158747) of GD14.5 fetal mouse cerebral cortical cells (Figure 14) [12], we investigated the transcript expression of upregulated ethanol-sensitive EV enriched proteins with g > 0.4 effect size for moderate (120 mg/dL) (Figure 14C) and heavy (320 mg/dL) (Figure 14D) ethanol exposures to identify potential parental cells-of-origin within the fetal neurogenic niche (Supplementary Table 13,14). We extracted cells belonging to previously identified VZ, SVZ, and transit progenitor cells (TPC) clusters (Figure 14A,B) [22], since these were the most similar cell-types to those contained in our neurosphere model. We found that the selected mRNA transcripts for EV-enriched proteins were not globally or uniformly expressed throughout the neurogenic niche, but that, collectively, transcript expression was most abundant in the fetal mouse cortical VZ (Figure 14E,F). These data indicate that, among the different cell type subpopulations in VZ, SVZ, and TPC clusters, stem cells in the fetal VZ are the principal cells-of-origin for proteins that are transferred to EVs following ethanol exposure. Since EVs are hypothesized to serve as an endocrine signal, we next examined whether NSC-derived EVs are actively taken up by cells. We found that purified EVs, directly labeled with the fluorescence reporter, MemBrite, and re-purified to eliminate free dye, were taken up by naïve NSCs and could be detected within the cytoplasm of recipient NSCs, suggesting that they were internalized by endocytosis (Figure 15A). We compared the number of labeled cells by EV uptake to a positive control, directly labeled NSCs (Supplementary Figure 7), and negative control, NSC exposure to culture medium that was mixed with MemBrite and then processed through the same purification steps as used for labeling EVs (Supplementary Figure 8). In the latter instance, no residual fluorescence was observed in recipient cells, supporting the specificity of the labeling process for EVs. Flow cytometric analysis of label-positive NSCs indicates that ∼31% of NSCs incorporate labeled EVs following incubation (Figure 15B) and that the amount of directly labeled cells was comparable to that seen with cells that uptake labeled EVs, indicating robust uptake of EVs in our neurosphere cultures. We next assessed the effects of inhibiting the biogenesis and/or release of EVs on oxidative metabolic activity and glycolysis in NSCs (Supplementary Figure 9), using three separate EV inhibitors, GW4869 (10 μM), Manumycin-A (0.5 μM), and Calpeptin (90 μM [67]). Repeated-measures one-way ANOVA with Geisser-Greenhouse correction, followed by a Dunnett's Test for post-hoc comparisons, indicated that there was a statistically significant effect of EV inhibition on oxidative metabolic activity (F(2.149,30.09) = 16.54, p < 0.0001). The Dunnett post-hoc test showed that EV inhibition, regardless of the inhibiting drug, significantly decreased oxidative metabolic rate in NSCs (GW4869: Diff = -0.5562, p.adj < 0.0001; Manumycin-A: Diff = -0.3106, p.adj = 0.0076; Calpeptin: Diff = -0.3353, p.adj = 0.0035), compared to the control NSCs that received no drugs to inhibit EV biogenesis (Figure S9A). There was a statistically significant effect of EV inhibition on glycolytic activity (F(2.237,24.61) = 10.86, p = 0.0003). The Dunnett post-hoc test showed that only EV inhibition with GW4869 significantly decreased glycolytic rate in NSCs (GW4869: Diff = -0.4586, p.adj = 0.0018; Manumycin-A: Diff = -0.1530, p.adj = 0.5360; Calpeptin: Diff = 0.3958, p.adj = 0.0587), compared to the control NSCs that received no drugs to inhibit EV biogenesis (Figure S9B). Finally, since the fetal VZ/SVZ includes both NSCs and maturing neural progenitors, the possibility exists that EVs secreted by NSCs may influence the biology of more mature cells. We therefore examined the effects of adding EVs, derived from naïve NSCs, to naïve recipient early differentiating neural cells (EDCs) (Supplementary Figure 10). For this study, NSCs were transformed into EDCs using a standard mitogen-withdrawal model with the provision of laminin to promote cell-extracellular matrix adhesion. Over 72 h, this paradigm results in the morphological transformation of NSCs into migratory bi-polar cells, and the appearance of neuronal markers and loss of NSC markers, as we have previously published [14, 17]. Paired samples t-tests were conducted to examine the effect of NSC-derived donor EVs on recipient naïve EDCs’ oxidative metabolic activity and glycolysis, while unpaired samples t-tests were used for gene expression of neuronal markers. NSC-EV exposure significantly decreased oxidative metabolic rate in naïve recipient EDCs (Diff = -0.1942, t(7) = 6.877, p = 0.0002), compared to the control EDCs that received no exogenous EVs (Figure S10A). In contrast, NSC-EV exposure significantly increased the rate of glycolysis in naïve recipient EDCs (Diff = 0.9938, t(23) = 4.734, p < 0.0001), compared to the control EDCs that received no exogenous EVs (Figure S10B). For differences in cycle threshold (ΔΔCT) for gene expression, NSC-EV exposure significantly increased ΔΔCT, which represents a significantly decreased gene expression, for Olig2 (t(67) = 2.195, p = 0.0316) and a trend towards an increased ΔΔCT, indicating a trend towards decreased gene expression, for GLAST (t(73) = 1.892, p = 0.0624) in naïve recipient EDCs, compared to controls (Figure S10C,D). There was no significant difference in gene expression for other neuronal markers: Nestin, NeuN, DCX, GFAP, and PDGFRa (Figure S10E-I). REDOX, glycolysis, and apoptosis: We next assessed the biological functionality of EVs derived from ethanol-treated NSCs (0, 120, 320 mg/dL), by transferring these purified EVs to naïve recipient NSCs (Figure 16). The sex of the donor-derived EVs and naïve recipient NSCs was matched. Repeated-measures two-way ANOVA with Geisser-Greenhouse correction, followed by a Dunnett's or Tukey's Test for post-hoc comparisons, was conducted to examine the effect of both treatment and sex on recipient naïve NSCs' oxidative metabolic activity, glycolysis, and apoptosis (Figure 16A-C). There was a statistically significant main effect of EV addition on oxidative metabolic activity (F(1.788,31.29) = 8.451, p = 0.0016), while there was no significant mean difference (Diff) for sex or an interaction effect between EV addition and sex. The Dunnett post-hoc test showed that purified EVs, regardless of the donor source, significantly decreased oxidative metabolic rate in recipient NSCs (EV 0: Diff = -0.3660, p.adj < 0.0001; EV 120: Diff = -0.2417, p.adj = 0.0005; EV 320: Diff = -0.2426, p.adj = 0.0007), compared to the control NSCs that received no exogenous EVs (Figure 16A). Moreover, there was no significant difference between the two control groups, i.e., NSCs that received filtered culture-conditioned media and NSCs that received neither purified EVs nor filtered culture-conditioned media (Diff = 0.0404, p.adj = 0.9811, ns). A secondary analysis, comparing the effects of ethanol treatment of donor EVs, did show that recipient NSCs exhibited a statistically significant and ethanol dose-related main effect on oxidative metabolism (F(1.905,40.02) = 4.400, p = 0.0202), with a Tukey post-hoc test showing that the EV 320 group (exposed to EVs from donor cells treated with 320 mg/dL ethanol), had a significantly increased oxidative metabolic rate compared to the EV 0 group (exposed to EVs from control, untreated donors; Diff = -0.1234, p.adj = 0.0345), with an intermediate effect that trended towards significance for the EV 120 group (Diff = -0.1243, p.adj = 0.0651). In contrast to oxidative metabolism, there was no significant difference in the rate of glycolysis, either due to the introduction of EVs or recipient cells' sex (Figure 16B), or caspase 3/7 activity due to EV addition (Figure 16C). As a comparison, repeated-measures two-way ANOVA with Geisser-Greenhouse correction followed by Tukey's post-hoc test was used to assess the effects of ethanol (0, 120, 320 mg/dL) on cellular oxidative metabolic activity, glycolysis, and apoptosis of NSCs (Figure 16D-F). There was a significant interaction between the effects of ethanol and sex on the oxidative metabolic rate (F(2, 36) = 4.598, p = 0.0167). A follow-up, sex-segregated analysis of the main effects of treatment showed that the high dose (320 mg/dL) of ethanol exposure significantly increased oxidative metabolic rate in female NSC samples compared to controls (Tukey's post-hoc test, p.adj = 0.0270) and to the moderate dose (120 mg/dL) group (p.adj = 0.0075), but no effect of ethanol in NSCs derived from male fetal tissue (Figure 16D). For glycolysis, there was a statistically significant interaction between the effects of ethanol and sex on the cellular oxidative metabolic activity (F(2,34) = 3.705, p = 0.0350). A post-hoc sex segregated analysis showed that medium and high dose of ethanol exposure significantly increased glycolysis rate in male NSC samples compared to controls (120 mg/dL: Tukey's post-hoc test, Diff = 0.4059, p.adj = 0.0283; 320 mg/dL: Diff = 0.5647, p.adj = 0.0056), but not in female samples (Figure 16E). Ethanol exposure did not significantly alter caspase 3/7 activity, suggesting that exposure had no significant effect on apoptosis (Figure 16F). Therefore, changes in cellular oxidative metabolism and glycolysis were not due to apoptosis and neither EV addition nor ethanol exposure significantly affected programmed cell death. Cell cycle: We further investigated the effects of EV transfer on cell cycle using flow cytometry. Using a two-way ANOVA, we found a significant main effect of EV addition (F (3, 66) = 4.181, p = 0.0090) and sex (F (1, 66) = 8.122, p = 0.0058) on the S-phase of cell cycle, compared to the control NSCs that were not exposed to purified EVs (Figure 16G,H). Dunnett post-hoc test showed that adding purified EVs to NSCs resulted in a significant decrease in the percentage of cells in S-phase (EV 0: Diff = -4.074, p.adj = 0.0157), and that adding EVs obtained from donor cells treated with 120 mg/dL ethanol similarly decreased the number of recipient cells in S-phase (EV 120: Diff = -4.643, p.adj = 0.0059). However, exposure to purified EVs obtained from NSCs treated with the high dose of ethanol did not result in a similar reduction of the number of recipient cells in S-phase (EV 320: Diff = -3.256, p.adj = 0.0725), compared to the control group that was not exposed to purified EVs. There was no significant difference for other cell cycle G0/G1 and G2/M phases (Supplementary Figure 11). EVs have gained significant attention as a novel and potential class of endocrine mediators of information transfer between cells and tissues. While EVs have been studied most extensively in tumor microenvironments and in cancer progression [68, 69, 70, 71], they may also contribute to the physiology and pathogenesis of developmental disorders, including FASD. This study showed that NSCs that were micro-dissected from fetal mouse cerebral cortex and maintained ex vivo, as neurosphere cultures, are remarkably active and have a net release of ∼10–14 EVs/hour/seeded parent cell. Moreover, our evidence shows that these EVs can be taken up by NSCs where they influence important biological outcomes like the cell cycle, cellular respiration, and cell maturation. To our knowledge, this is the first study to assess the global effects of developmental ethanol exposure on the proteome of EVs secreted by NSCs, and to show that ethanol exposure resulted in selective enrichment of specific proteins from each sub-cellular compartment into EVs at the expense of donor parent cells. This study is also to our knowledge, the first to explicitly assess the contribution of biological sex to the proteome of EVs, and to show, using WGCNA, that fetal sex was a significant contributor to at least one network of proteins expressed in EVs, specifically the brown module, though most protein networks were sex-independent. Our data also show that the EV proteome is complex. We identified peptides for over 4,700 unique proteins in EVs representing ∼75% of the parental NSC proteome. Moreover, the proteome of EVs is broadly conserved, since peptides from ∼59% of EV-containing proteins were present in all analyzed samples. Among the proteins that were tested, Western blot analyses indicated they correspond to full-length proteins of the expected molecular weights (Supplementary Figure 1–3). In addition, EV samples had increased relative abundance for proteins involved in ESCRT endosome sorting mechanism, which plays a key role in exosome biogenesis [72]. Therefore, the evidence indicates that EVs sequester full-length, and potentially functional, proteins, although the possibility exists that EVs also contain misfolded proteins or peptide products of intracellular proteolytic cleavage. The substantial number of proteins identified in the EVs suggests that the EV population is likely to be heterogenous, potentially with different, as-yet-undefined subclasses of EVs sequestering a distinct complement of proteins. Interestingly, WGCNA showed that EV gene networks could be characterized by separate modules related to negative regulation of cell development and differentiation, Ras signal transduction, immune regulation, and protein localization/transport, further supporting the possibility that EVs secreted by NSCs are heterogenous with respect to content, and perhaps function. Further studies are needed to ascertain whether these protein networks are retained within a relatively homogenous population of EVs, or spread across a diversity of EV subclasses that are released asynchronously by NSCs. Importantly, we also assessed the contribution of fetal sex to the EV proteome, and found that EVs secreted by male and female fetuses were generally similar. However, we did observe increased sequestration of lysosome-localized proteins in female EVs compared to male EVs. This study uniquely compared the proteome of EVs to their paired parent-of-origin NSCs. With this comparative analysis, we identified proteins within every subcellular location, as assigned by the UniProt Knowledgebase [50, 73], that were in greater relative abundance in EVs compared to their parental NSCs, suggesting specific enrichment in EVs. In addition to proteins like TSG101, CD63, and ESCRT-related vacuolar protein sorting proteins (VPSs) that have previously been documented to be enriched in EVs and were also enriched our NSC-derived EVs, we observed a number of proteins assigned to the subcellular compartments that were present at higher relative abundance in EVs to suggest selective loading of these proteins into EVs (Figures 4, 5, 6, 7, 8, and 9). It was surprising to find proteins enriched in EVs relative to parental NSCs that were categorized as multi-compartmental and having a nuclear location, i.e., categorized for cytoplasm and membrane compartments but also nuclear compartment. For example, LRP1 and members of the Notch family, which are multifunctional receptors and regulators of neural development [74, 75], were among the most highly enriched proteins in EV samples compared to parental cells. Interestingly, proteolytic cleavage of both LRP1 and Notch can generate both intracellular and extracellular fragments which serve as transcription regulators and decoy receptors, respectively. EV samples contained peptides for LRP1 as well as for Notch1, 2, and 3, across the length of these proteins, suggesting that their multifunctional potential is retained by EVs. Moreover, EVs were enriched for Jagged-1, a cell surface ligand that binds to Notch receptors to activate Notch signaling pathway and by that mechanism [76], controls neurogenesis. This suggests that EVs have the capacity to transmit a functional ligand-receptor entity to recipient cells. EVs were also enriched for tyrosine kinase receptors like epidermal growth factor receptor (EGFR) and fibroblast growth factor receptor (FGFR), which play essential roles in the regulation of embryonic development, cell proliferation, migration, and differentiation [77, 78]. Other neurodevelopmental-related proteins, like Wdr82, associated with Kleefstra Syndrome 2, an autosomal dominant neurodevelopmental disorder [79], and Gpc6, associated with Omodysplasia 1, an autosomal recessive skeletal dysplasia syndrome, that is also associated with intellectual disability [80], were also examples of proteins that were heavily enriched in EVs. Finally, pathway overrepresentation analysis indicated that EVs sequester proteins that control nervous system development broadly. These include, interestingly, non-canonical Wnt signaling and the planar cell polarity/convergent extension (PCP/CE) pathway which controls brain morphogenesis, including the planar orientation of neural progenitors and their directional growth to form the laminar structure of the developing cerebral cortex [81]. These and our additional findings from a scRNAseq re-analysis that mRNA transcripts for upregulated ethanol-sensitive EV-enriched proteins were unevenly expressed in the neurogenic niche, localized to VZ-type cells rather than SVZ and TPC clusters, suggest that EVs have the potential to transfer important and developmentally relevant signaling mediators among cells. These findings support a hypothesis that EVs homogenize and coordinate some behaviors of cells within the neurogenic niche by transferring proteins derived from NSCs to other cells. An important question that needs to be further explored is whether maternal-fetal perturbagens like ethanol reprogram EVs. Although ethanol inhibits fetal growth, it surprisingly does not kill fetal NSCs, but rather causes a loss of NSC renewal potential and aberrant premature maturation [9, 12, 13, 14, 17, 19, 58, 59, 82, 83]. Consequently, the effects of ethanol on EVs, a novel endocrine mediator, need to be better understood to determine if the reprogramming of EVs contributes to pathogenic processes. In the current study, we again observed that ethanol exposure did not result in an increase in apoptosis in parental NSCs, consistent with our previous observations [12, 19, 23, 83]. Because there was no change in apoptosis, it is unlikely that the release of apoptotic bodies contributed to the EV proteome signature following ethanol exposure. Our dose-response analysis showed that ethanol does result in some dose-independent effects on NSC-derived EVs, including the selective enrichment of proteins in EVs that were also depleted in parental NSCs. For example, the heat-shock protein chaperone DNAJC7 isoform 1, which regulates protein folding and protein transport [84], was significantly enriched in EVs relative to NSCs following NSC exposure to both medium and high doses of ethanol, compared to controls. Similarly, the heterogeneous nuclear ribonucleoprotein HNRNPU, which binds to pre-mRNAs and controls nuclear pre-mRNA processing and transport [85], was significantly enriched in EVs relative to cells for both ethanol-treated groups, but not controls. Pathway enrichment analysis, showed that across treatment groups, ethanol-sensitive proteins were enriched for RNA-binding and chaperone proteins involved in ribosome and translation pathways (Supplementary Table 16–19). This coordinated outcome suggests that ethanol exposure results in transfer of proteins from cells to EVs. Surprisingly, this effect is not generalized to all EV proteins, and therefore, suggests the presence of a potentially novel and selective ethanol-dependent sequestration mechanism that controls protein sorting into EVs. An intriguing finding from WGCNA was that the ethanol-dependent sequestration of proteins into EVs was accompanied by an increase in the number of networks of tightly interconnected proteins. This outcome suggests that ethanol exposure resulted in a dose-related increase in the biological complexity of EVs, without a corresponding increase in the proteomic network complexity in parental NSCs. Not surprisingly, pathway overrepresentation analysis of proteins that were enriched in EVs relative to parental NSCs due to ethanol exposure, uncovered evidence for dose-specific pathways. For example, moderate ethanol exposure favored overrepresentation of proteins related to mitosis and ribosomal RNA processing. We also observed increased sequestration of proteins associated with nonsense mediated decay (NMD) into EVs at this lower dose, an interesting finding, given that NMD is necessary for early embryo and fetal development, for stem cell renewal and maturation [86], and that disruption of NMD is embryonic-lethal [87]. In contrast, in response to the high dose of ethanol, there was increased sequestration of proteins associated with stress response to amino-acid deprivation, inhibition of DNA replication, and pathways that may be expected to impair neurogenesis [88, 89]. We also found that the high dose of ethanol, like the medium dose, resulted in sequestration of proteins related to DNA replication into EVs. However, there were also significant differences. For example, the dominant pathway overrepresentation with heavy exposure, was the mechanistic target of rapamycin (mTor) cellular response amino acid starvation, an established stress response in animal models of prenatal ethanol exposure [90, 91], that is associated with inflammation and fetal brain injury [92] and disorganized cortical lamination [93]. A secondary analysis of our previously published in vivo single cell RNAseq data of the developing cerebral cortex [12] showed that the translated mRNAs that contributed to the EV response to ethanol were predominantly localized to cells of the fetal ventricular zone and were not localized to the subventricular zone or transient progenitor cells. This analysis outcome is important because it supports the hypothesis that fetal NSCs that populate the VZ in vivo and give rise to neurospheres ex vivo are the dominant source of this ethanol response in EVs, rather than their more mature progeny. In the current study, we did find evidence for biological activity of EVs. Our data show that fluorescently-labeled EVs added to culture medium are taken up avidly by NSCs, and inhibiting the formation and release of EVs results in metabolic inhibition of NSCs. Moreover, purified EVs added to culture medium reduced the proportion of recipient NSCs in S-phase without influencing cell death. Since stem cell differentiation is coupled with cell-cycle progression (reviewed in [94]), it is possible that exogenously added EVs, from neurospheres ex vivo and NSCs from the VZ in vivo, serve to inhibit differentiation. Added support comes from our observations that exogenously added EVs derived from NSCs also decreased the rates of oxidative metabolism in both NSCs and EDCs, while increasing the rates of glycolysis and decreasing mRNA transcripts for the oligodendrocyte marker Olig2, in EDCs. Since the maturation of stem cells is accompanied by increased oxidative phosphorylation and decreased glycolysis (reviewed in [95, 96]), these data argue for a basal role for un-treated NSC-derived EVs in maintaining the stem cell niche. However, it remains to be determined whether EVs secreted by NSCs influence the biology of other spatially proximate cell populations such as those of the SVZ, or even possibly, the early neurons of the developing cortical plate. Interestingly, exogenously added EVs derived from ethanol-treated NSCs did result in additional effects in NSCs, beyond those discussed for control EVs obtained from treatment-naïve NSCs. For instance, we observed that while EVs obtained from naïve donor controls resulted in decreased oxidative metabolism and S-phase in recipient NSCs, EVs from ethanol treated donors were not as effective. One caveat that needs to be acknowledged is that endogenous EVs secreted by recipient NSCs may have competed with donor EVs from ethanol-exposed NSCs, to diminish the overall effect of ethanol exposure. While future analyses will be needed to determine whether EVs secreted by ethanol-treated NSCs confer changes in physiology and gene expression in recipient cells, these data suggest that underlying shifts in the sequestered proteome may be part of an overall picture of diminished EV efficacy due to ethanol exposure. However, these outcomes do not minimize a different possibility, i.e., that proteomic shifts in EVs due to an environmental perturbation are more consequential for parental NSCs than for recipient NSCs. It is feasible that the transfer and sequestration of proteins into EVs represent diminished/decreased functionality by secreting cells, more than a transfer of functionality to recipient cells. In conclusion, mouse-derived fetal NSCs secreted an abundance of EVs. These data collected in this exploratory study showed that the proteome of these EVs is complex and likely reflects the presence of multiple EV subpopulations. Moreover, because selected proteins from each subcellular compartment, from nuclear to cytoplasmic and membrane-bound proteins, were preferentially enriched in EVs relative to parental cells, it is likely that an active sorting mechanism shapes distinct EV proteomes. Our evidence also suggests that secreted EVs are biologically functional and may play a role in maintaining stem cell identity, though future studies in whole-animal models will be needed, to ascertain both spatial and temporal effects of EVs on neurogenesis and stem cell identity. Importantly, our data also show for the first time, that a developmental perturbagen like ethanol has a significant, yet specific impact on the proteome of EVs secreted by NSCs. Ethanol exposure resulted in a dose-dependent and selective enrichment of proteins in EVs and concurrent decrease of those proteins from parental NSCs, implying the activation of a re-sorting process in loading specific proteins into EVs. Re-sorted proteins appeared to serve specific functions, from the control of cell cycle and protein misfolding specifically at a medium dose, to resorting of stress response, DNA replication, and Rho-GTPase pathway components at a high dose of ethanol exposure. Moreover, EVs derived from ethanol-exposed NSCs did have effects beyond that observed with EVs from ethanol treatment-naïve NSCs. While further confirmatory studies in whole animal models are needed, these data do suggest that the sequestration of proteins into EVs following ethanol exposure represents a loss of function for the parental NSCs, as well as possible gain of function for recipient NSCs, that may collectively constitute a stress-adaptive response to a perturbagen like ethanol. C57BL/6J (Ai14) mice (Jackson Laboratories; Catalog # 007914) were bred in-house and time-mated overnight, from the start of the dark phase of the light-dark cycle, with the following morning defined as gestational day (GD) 0.5. Ex vivo neurosphere cultures were created from NSCs obtained from three separate pregnancies, from acutely dissociated GD 12.5 fetal mouse dorsal telencephalic neuroepithelium, which corresponds to the future isocortex, and used to model early neuronal development [17, 19]. At the time of collection, fetal sex was determined as we have previously reported [12]. Briefly, genomic DNA was obtained from fetal tissue samples using alkaline lysis and fetal sex ascertained by a rapid qPCR protocol using primers to detect repetitive sequences on X and Y chromosomes [97, 98]. Cortical neuroepithelial tissues within a single pregnancy were pooled by sex, to generate male- and female-specific cultures. Sex-specific cultures were obtained from three separate pregnancies, to generate three male and three female biological replicates. Sex-specific cultures were maintained as non-adherent neurospheres in serum-free mitogenic media and passages 7 to 10 were used in this study, as previously published [17, 99, 100] (Figure 1A). All animal procedures were performed in accordance with the Texas A&M University Institutional Animal Care and Use Committee guidelines and approval. Dispersed single NSCs were seeded at a density of 4 × 106 cells in 10 mL of culture media per T75 flask, with four flasks being defined as a single sample. Cell count and viability (∼75–85% viability) were measured using the Invitrogen Countess Automated Cell Counter (Invitrogen; Catalog #C10227; Carlsbad, CA/USA). Ethanol, unlike other psychotropic agents, is consumed, and exhibits its psychological effects in the millimolar range, and can be tolerated by humans at doses in excess of 100 mM(56, 57). We therefore used a clinically relevant dose-range of ethanol, associated with risky drinking in humans, of up to 70 mM in our studies. Previous studies have shown that this dose range does not result in the death of NSCs [19, 83], but rather, aberrant and premature maturation [17, 22, 82]. Each sample was randomly assigned to one of three ethanol treatment conditions: 0 mg/dL (control), 120 mg/dL (26 mM, moderate, level achievable by individuals that binge drink), or 320 mg/dL (70 mM, high, level achievable by individuals with chronic alcohol use [56]). Ethanol (190 proof grain alcohol) was diluted into fresh culture medium and at the time of passaging, ethanol naïve media was replaced with fresh experimental media. To model chronic alcohol exposure, neurosphere cultures were exposed to control or ethanol treatment conditions for five days, with media replacement on day 3. All flasks were tightly capped with phenolic caps and sealed with parafilm to prevent ethanol loss in the culture medium throughout the experiment. Ethanol concentrations in culture-conditioned medium were verified by gas chromatography for each experiment. The measured ethanol dose range, from moderate (100–134 mg/dL; 22 mM–29 mM) to high (250–380 mg/dL; 54 mM–82 mM), is consistent with our previously published studies [13, 17, 101, 102]. Neurospheres and culture-conditioned medium (for the concentration of EVs) were collected on day 5. EV fractions were separated and concentrated from culture-conditioned medium following an established ultracentrifugation protocol (Théry et al., 2006) with a few added steps [103]. Briefly, culture-conditioned medium was separated from the cell fraction by centrifugation at 200 x g for 5 min, and the cell pellet isolated for separate protein analysis. The culture media supernatant was next centrifuged at 2,000 x g for 10 min at 4 °C to eliminate dead cells and other debris (Figure 1B), then passed through a 0.2 μm sterile filter with a polyethersulfone membrane (VWR; Catalog # 28145–501; Radnor, PA/USA) to exclude debris with diameters greater than 200 nm. The filtered supernatant containing particles <200 nm was centrifuged through a 100 kDa molecular weight cutoff (MWCO) polyethersulfone membrane (PALL; Catalog # MAP100C37; Port Washington, NY/USA) at 4,000 x g for 30 min to concentrate the EV-enriched supernatant while filtering out any particles below 100 kDa. The material that was collected on the membrane was transferred to polypropylene thick-walled centrifuge tubes (Beckman; Catalog # 355640; Brea, CA/USA), adding chilled (4 °C) 1 x PBS buffer (Thermo Fisher; Catalog # 14190144; Waltham, MA/USA) to a total volume of 7 mL per tube. This PBS-suspended material was centrifuged at 100,000 x g for 90 min at 4 °C in a Type 70 Ti fixed-angle titanium rotor (Beckman; Catalog # 337922; Brea, CA/USA). Pellets were washed, by decanting the supernatant and resuspending the EV-enriched pellet in 1 mL chilled PBS, followed by the addition of 6 mL chilled PBS and centrifugation at 100,000 x g for 90 min at 4 °C. The concentration and size of EVs were measured by nanoparticle tracking analysis (Nanosight LM10; Malvern Panalytical; Westborough, MA/USA), as we have previously published [23]. 1 x PBS buffer (Gibco®, Thermo Fisher; Catalog # 14190144) was used for the dilution of isolated EV samples. TEM and immunogold labeling were used for the validation of EV isolation. TEM sample preparation and imaging were performed at the Texas A&M Microscopy and Imaging Center, following a protocol we have previously published [23]. For immuno-gold labeling of the EV surface marker CD63, polyclonal rabbit anti-CD63 antibody diluted 1:100 (System Biosciences; Catalog # EXOAB-CD63A-1; Palo Alto, CA/USA) was used as the primary antibody, and binding detected incubation with 12 nm goat anti-rabbit antibody conjugated with colloidal gold (Jackson ImmunoResearch Lab; Catalog # 111-205-144; West Grove, PA/USA). Protein was extracted from neurospheres and their corresponding EVs using 1 x RIPA lysis buffer (EMD Millipore; Catalog # 20–188; Burlington, MA/USA) with addition of Halt™ protease and phosphatase inhibitor cocktail (Thermo Fisher; Catalog # 78442). Extracted protein concentration was determined using Pierce™ BCA protein assay kit (Thermo Fisher Scientific; Catalog # 23225). Protein (20 μg) was size-fractionated on a 4–12 % Bis-Tris Gel, run at 150 V for 60 min, and transferred to a PVDF membrane using the iBlot transfer system (Thermo Fisher; Catalog # IB301001). The membrane was briefly washed in deionized water and then dried overnight at room temperature in the dark. Subsequently, the membrane was blocked for 1 h with Odyssey Blocking Buffer in Tris-buffered saline containing 0.1 % sodium azide (Licor; Catalog # 927–50000; Lincoln, NE/USA) and then incubated overnight at 4 °C in the dark with primary antibodies (Supplementary Table 20). The immunoblot was washed and incubated with an IRDye-conjugated goat anti-mouse or anti-rabbit IgG (LI-COR; Catalog # 926–68070 and # 926–32350) at dilution 1:15,000 for 1 h and then imaged using a LI-COR Odyssey CLx Imager. The retained EV pellet from ultracentrifugation was resuspended in 50 μL of Laemmli lysis buffer (Amresco; Catalog #M337-25ML; Solon, OH/USA), with addition of Halt™ protease and phosphatase inhibitor cocktail (Thermo Fisher; Catalog # 78442). Samples of the matched cell-of-origin NSC pellets were similarly resuspended. EV and cell samples were frozen at -80 °C, placed on dry ice, and shipped overnight to the University of Texas Health Science Center at San Antonio Mass Spectrometry Institutional Core Laboratory. Protein quantities were assessed using the EZQ™ Protein Quantitation Kit (Thermo Fisher; Catalog #R33200). Aliquots between 70–100 μg (cell samples) and 2–10 μg (EV samples) were reduced with tris(2-carboxyethyl)phosphine hydrochloride (TCEP), alkylated in the dark with iodoacetamide and applied to S-Trap mini columns (ProtiFI; Farmingdale, NY/USA) for tryptic digestion with sequencing grade modified trypsin (Promega; Catalog #V5111) in 50 mM TEAB. Peptides were eluted from the S-Traps with 0.2 % formic acid in 50 % aqueous acetonitrile and quantified using Pierce™ Quantitative Fluorometric Peptide Assay (Thermo Fisher; Catalog # 23290). Data-independent acquisition mass spectrometry was conducted on an Orbitrap Fusion™ Lumos™ Tribrid™ Mass Spectrometer (Thermo Fisher). On-line HPLC separation was accomplished with an RSLCnano HPLC system (Thermo Scientific/Dyonex): column, PicoFrit™ (75 μm internal diameter.; New Objective; Littleton, MA/USA) packed to 15 cm with C18 adsorbent (218MS 5 μm, 300 Å; Vydac/Grace; Columbia, MD/USA); mobile phase A, 0.5 % acetic acid (HAc)/0.005 % trifluoroacetic acid (TFA) in water; mobile phase B, 90 % acetonitrile/0.5 % HAc/0.005 % TFA/9.5 % water; gradient 3–42 % B in 120 min; flow rate, 0.4 μL/min. For a reference library, a pool was made of all samples in each experiment, and aliquots of the digests [0.67 μg (male EVs), 0.75 μg (female EVs, or 1 μg cells)] were analyzed using gas-phase fractionation and 4-m/z windows (120k resolution for precursor scans, 30k for product ion scans, all in the orbitrap) to create an empirically-corrected DIA chromatogram library [104, 105] by searching against a Prosit-generated predicted spectral library [106] based on a UniProt reference Mus musculus (Strain: C57BL/6J) reference FASTA database (UP000000589_10090 downloaded on 20220216). Experimental samples were blocked by source pregnancy and randomized within each block. Injections of 0.5–1.0 μg of peptides from each sample and a 2-hour HPLC gradient were employed. MS data for experimental samples were acquired in the orbitrap using 12-m/z windows (staggered; 120k resolution for precursor scans, 30k for product ion scans) and searched against the chromatogram library. Scaffold DIA (v3.2.1; Proteome Software; Portland, OR/USA) was used for all DIA-MS data processing (Supplementary Table 1–3), with a minimum of two peptides and a 1 % false discovery rate (FDR). Raw and processed data files are deposited in MassIVE repository under accession number MSV000089214 and in ProteomeXchange repository under accession number PXD033094. Proteomic data were analyzed to identify significant differences in: protein abundance between EV and cell groups, regardless of sex or ethanol dose; ethanol- and sex-dependent protein abundance between EV and cell groups; and pathways enriched in proteins that showed differential protein abundance. All computation and analysis involving protein expression were conducted in “RStudio” software (RStudio, Inc.; Boston, MA/USA), v.1.2.1335 for Windows. Only proteins detected in all 18 E V samples (2765 specific proteins) out of 36 total samples, 18 E V and 18 cell samples, were used for analysis. Each protein within a sample was given a normalized value (Normalized Value Protein X from Sample A = NX) from the ratio derived by the total exclusive peptide intensity of the protein of a sample (Total Exclusive Peptide Intensity Protein X = PX) divided by the sum of the same sample's total exclusive peptide intensity (Sum of Total Exclusive Peptide Intensity Sample A = SA): After normalization, any missing value for a protein in cell samples was imputed by a randomized number (Imputed Value Protein X from Sample A = IX), within a range of minimum, defined as the protein mean across samples subtracting three standard deviations of the protein expression value across samples (μ Protein X from all samples—3∗SD Protein X from all samples = μ X—3∗SD X) and of maximum being the protein mean across samples subtracted by two standard deviations of the protein expression across samples (μ Protein X from all samples—2∗SD Protein X from all samples = μ X—2∗SD X): The rationale for imputing undetectable values in cells stems from the hypothesis that, because EVs are derived from a cell, any protein detected in EVs must be present in their parent cells. Therefore, any protein that is detected in EVs but not in cell samples is hypothesized to be present in cells, but below the limit of detection. Any imputation that resulted in a negative number was replaced by a value of “0”. After imputation, a z-score for each specific protein was calculated across all 36 samples. Each protein was then categorized by subcellular location (Nucleus, Cytoplasm, Mitochondria, Lysosome, Endosome, Membrane), based on the UniProt Knowledgebase [50, 73]. As defined in this knowledgebase, a protein may be present non-exclusively in multiple sub-cellular compartments; therefore, for further analysis, non-exclusive proteins were assigned to all associate compartments. After classifying proteins by subcellular location, an average z-score was computed for all proteins within each subcellular location for each sample. And a 2-way ANOVA analyses (EV (or cell) x sex) with Tukey's HSD test as the post-hoc tests were computed to assess the contribution of proteins from each sub-cellular location to differences in the overall composition of EVs relative to cells (e.g., are nuclear proteins as a group more highly expressed in cell samples than EVs). Next, the subcellular location of proteins that were specifically enriched in EVs compared to cell samples was determined, and for each protein, it's average z-score and 95% confidence estimate for each sample group (EV vs cell) was computed, and proteins with non-overlapping confidence estimates were classified as statistically significantly enriched in EVs relative to parental cells. Weighted correlation network analyses (WGCNA) were conducted using R-based “WGCNA” package [107, 108]. Heat maps from the WGCNA processed data were generated using the R-based “gplots” package [109]. Using a topological overlap matrix (TOM) created under the WGCNA method, we constructed gene networks and identified modules with hierarchical clustering dendrograms (trees). Distinct modules were identified from the expression levels of 2615 gene equivalents; gene equivalents were proteins whose expression levels varied between samples, signifying expression level above “0”, and detected in at least 27 of the 36 samples to limit numbers of missing entries in genes and samples. Weighted networks were visualized by heatmaps, with the gene dendrograms and module colors displayed along the top and left side; each row or column corresponds to a single gene/protein, where light colors denote low topological overlap and dark colors denote high topological overlap. While this protocol was initially designed to examine gene expression, we are applying it to protein expression. When referring to protein expression, we will refer to the proteins; when referring to an identified hub, we will use gene as it is a measure from the WGCNA. A chi-square test of independence was performed on WGCNA EV/cell module counts to examine the relationship between ethanol treatment and number of modules across locations (EV vs. cell), using the control values as percent expected. Protein enrichment in EVs relative to parent cells was next compared between control and treatment groups. The distribution of a specific protein was assessed by comparing its value (Value Protein X from Sample A = VX) of EV sample (Value Protein X of EV Sample by Treatment per Pregnancy = VX_EV) to Cell sample (Value Protein X of Cell Sample by Treatment per Pregnancy = VX_Cell) in each pregnancy (3 pregnancies) by ethanol treatment groups (0,120,320 mg/dL), resulting in an enrichment value (R Protein X from EV sample to Cell sample by Treatment per Pregnancy = RX): Interestingly, principal component analysis showed that proteins that contributed to both the first and second principal components separated cell samples by pregnancy (Supplementary Figure 12) suggesting that protein content of EVs can vary substantially from one pregnancy to the next. Therefore, to reduce variability due to individual differences in pregnancies/mouse litters and to account for the fact that cells from each pregnancy were partitioned to three treatment groups at the initiation of ethanol exposure, we used paired samples for t-test and repeated-measures analyses for ANOVA. To identify proteins where EV enrichment was significantly altered by treatment, we used paired samples t-tests, a Hedges' g (g) effect size and a 95 % confidence estimate for the effect size using the R-based “effsize” package, v.0.8.1, to compare the means between control and treatment groups for samples from the same pregnancy. For pathway analysis, proteins whose EV to cell enrichment was altered by treatment with a p-value < 0.05, g > +0.4, and with 95 % confidence estimate for g that were non-zero containing, were assessed. Enrichment pathway analysis was conducted using the R-based “ReactomePA” package, v.1.36.0 [110], and “clusterProfiler” package, v.4.0.5 [111]. Repeated-measures two-way ANOVA with Geisser-Greenhouse correction, followed by a Dunnett's or Tukey's Test for post-hoc comparisons, were used to assess the biological functionality of EVs derived from ethanol-treated NSCs (0, 120, 320 mg/dL) on naïve recipient NSCs. All other statistical analyses were conducted using the “R” software (R Foundation for Statistical Computing; Vienna/Austria), v3.6.1 for Windows and the “GraphPad Prism” software (GraphPad Software; San Diego, CA/USA), v8.4.2 for Windows. To further understand the identity of the parental NSCs that were the source of ethanol responsive EV proteins, we analyzed our previously published single cell RNAseq (scRNAseq) data from GD 14.5 mouse cerebral cortex (NCBI GEO accession number GSE158747) [12], focusing on cell clusters that resemble the ventricular zone (VZ), the subventricular zone (SVZ), and transit progenitor cells (TPC), the three cell populations of a developing fetal mouse brain that most closely resemble cells present in our neurosphere model [12]. After cell populations with VZ, SVZ and TPC identity were visualized as previously published [22], the cell-types expressing the composite transcriptomic signature of the differentially regulated proteins in EVs (p < 0.05; g > 0.5 with a non-zero containing 95 % confidence estimate), from both the 120 and 320 mg/dL exposure conditions, were shown. To visualize the uptake of EVs to NSCs in vitro, we labeled EVs, purified as described earlier, with cyanine-based, membrane-localized, MemBrite Fix 568/580 (Biotium; Catalog # 30095-T; Fremont, CA/USA) according to the manufacturer's instructions. Briefly, isolated EVs (∼109) were incubated with MemBrite at final concentration of 200 nM in PBS (Thermo Fisher; Catalog # 14190144) for 30 min in the dark, at room temperature. PBS was then added to the EV supernatant for a total of 15 mL volume and passed through a 0.2 um sterile filter with polyethersulfone membrane (VWR; Catalog # 28145–501) to exclude any possible aggregates with diameters greater than 200 nm. The filtered supernatant containing particles <200 nm was subjected to centrifugation through a 100 kDa molecular weight cutoff (MWCO) polyethersulfone membrane filter tube (PALL; Catalog # MAP100C37) at 4,000 x g for 30 min to concentrate the small EV-enriched supernatant while filtering out unbound dye. PBS (15 mL) was added to the remaining EV supernatant and centrifuged once more at 4,000 x g for 30 min. The labeled EV supernatant was collected into a new 100 kDa MWCO filter tube in 10 mL of fresh culture media and centrifuged at 4,000 x g for 30 min, with this step being repeated with new 10 mL of fresh culture media. Finally, naïve NSCs were exposed to labeled EVs for 24 h before the cells were processed for flow cytometry or confocal fluorescence microscopy analyses. As a negative control, fresh culture media was processed for labeling using the same procedure as that used for EVs, and the resulting supernatant introduced to NSCs as described above. To directly label cells, as a positive control, MemBrite Fix 568/580 dye was diluted in 1 mL of fresh culture media for 200 nM final concentration, and ∼500,000 NSCs were resuspended in the resulting dye-containing media and incubated for 30 min in the dark at room temperature. To eliminate unbound dye, cells were rinsed three times with 5 mL of fresh culture media and centrifuged at 300 x g for 5 min. Finally, labeled cells were processed for flow cytometry or confocal fluorescence microscopy analyses. Following addition of labeled EVs or filtered culture media, cells were briefly fixed (2 % PFA, 15 min). NSC nuclei were counterstained with 300 nM DAPI (Thermo Fisher; Catalog #D1306), mounted onto glass slides (Vectashield, Vector Laboratories; Catalog # H-1200-10; Burlingame, CA/USA), coverslipped, and imaged using a confocal-laser scanning microscope (FluovView-1200, Olympus Corporation of the Americas; Center Valley, PA/USA) equipped with a 405 nm laser to excite DAPI and a 559 nm laser to excite MemBrite 568/580. Micrographs were acquired using a 60x magnification objective (UPlanSApo 60X Oil, Olympus) with 2x zoom through image spatial resolution adjustment. Cellular metabolic activity was assessed using alamarBlue™ HS (Thermo Fisher; Catalog # A50100) as a fluorometric/colorimetric indicator according to manufacturer's instructions. Briefly, 10 μL alamarBlue™ HS was added to each well with ∼30,000 cells in 100 μL media in a 96 well flat bottom clear plate, incubated at 37 °C for 4 h, then the fluorescence read at 570 nm and 600 nm wavelengths using Cytation 5 cell imaging multi-mode reader (Agilent Technologies, Inc.; Santa Clara, CA/USA). For studies on the effects of EV addition, ∼109 EV isolated from ethanol-treated (0, 120, 320 mg/dL) sex-specific NSC cultures were added to 106 naïve recipient, sex-matched NSCs (female NSC-derived EVs were delivered to female naïve recipient NSCs and male NSC-derived EVs to male recipient NSCs). Recipient cultures were incubated for 72 h before being processed for the alamarBlue assay. For studies on the effects of EV inhibition, one of three EV inhibitors, GW4869 (10 μM), Manumycin-A (0.5 μM), or Calpeptin (90 μM), in denoted concentrations that are not toxic to NSC cultures, were added to 106 naïve recipient, sex-matched NSCs. Recipient cultures were incubated for 48 h before being processed for the alamarBlue assay. For studies on the effects of naïve NSC-derived EV addition to EDCs, ∼109 EV isolated from naïve sex-specific NSC cultures were added to 106 naïve recipient, sex-matched EDCs. Recipient cultures were incubated for 72 h before being processed for the alamarBlue assay. The Lactate-Glo Assay (Promega; Catalog #J5022; Madison, WI/USA) was used according to the manufacturer's instructions to measure the glycolysis rate of viable cells. The assay detects lactate produced by glycolysis. Briefly, 50 μL of Lactate Detection Reagent was added to 50 μL of 1:50 diluted sample in a 96 well flat bottom white plate, incubated at room temperature for 1 h, then luminescence read using the Cytation 5 cell imaging multi-mode reader. For studies on the effects of EV addition, ∼109 purified EVs from ethanol-exposed (0, 120, 320 mg/dL) male and female NSC cultures were added to 106 sex-matched but naïve recipient NSCs. Recipient cultures were incubated for 72 h before being processed for glycolysis assay. For studies on the effects of EV inhibition, one of three EV inhibitors, GW4869 (10 μM), Manumycin-A (0.5 μM), or Calpeptin (90 μM), were added to 106 naïve recipient, sex-matched NSCs. Recipient cultures were incubated for 48 h before being processed for the glycolysis assay. For studies on the effects of naïve NSC-derived EV addition to EDCs, ∼109 EV isolated from naïve sex-specific NSC cultures were added to 106 naïve recipient, sex-matched EDCs. Recipient cultures were incubated for 72 h before being processed for the glycolysis assay. The Promega Caspase-Glo® 3/7 Assay Systems (Promega; Catalog #G8091) was used according to the manufacturer's instructions to quantify apoptotic cell death at day 5 of ethanol exposure and at 48 h following exposure of naïve cells to EVs. Briefly, 100 μL Caspase-Glo® 3/7 Reagent was added to each well with ∼30,000 cells in 100 μL media in a 96 well flat bottom white plate, incubated at 37 °C for 2 h, then luminescence read using the Cytation 5 cell imaging multi-mode reader. For studies on the effects EV addition, ∼109 purified EVs from ethanol-exposed (0, 120, 320 mg/dL) male and female NSC cultures were added to 106 sex-matched but naïve recipient NSCs, and incubated for 48 h, before being processed for caspase assay. For cell cycle analysis, after addition of ∼109 EV to ∼106 cells for 72 h, with ethanol-treated (0, 120, 320 mg/dL) sex-specific NSC culture-derived EVs added to respective sex-specific recipient NSCs, DNA synthesis was assessed by pulse-labeling cells with 10 μM EdU (5-ethynyl-2′-deoxyuridine) for 1 h at 37 °C. Immediately after, cells were collected and labelled using the Click-iT® EdU Alexa Fluor® 488 Flow Cytometry Assay Kit (Thermo Fisher; Catalog #C10425), in conjunction with 7-amino-actinomycin D (Thermo Fisher; Catalog # 00–6993-50), according to the manufacturer's instructions. Cell cycle analysis was performed with the BD LSR Fortessa X-20 Cell Analyzer flow cytometer (Becton, Dickinson and Company; Franklin Lakes, NJ/USA). Data were analyzed using FCS Express software v.7.12.0007 (De Novo Software; Pasadena, CA/USA). For detection of labeled EVs in NSCs, after addition of EVs, cells were briefly fixed (2% PFA, 15 min) before undergoing flow cytometry using the BD LSR Fortessa X-20 Cell Analyzer. Data were analyzed using FCS Express software v.7.12.0007 (De Novo Software). RNA from early differentiating cells was isolated using the miRNeasy™ mini kit (Qiagen; Catalog # 217004; Germantown, MD/USA). cDNA synthesis was performed using the qScript™ cDNA SuperMix kit (Qiagen; Catalog # 95048). qPCR analysis was done on an Applied Biosystem ViiA 7 Real-time PCR system (ABI/Life Technologies; Grand Island, NY/USA). For mRNA transcript quantification, presented data correspond to ΔΔCT after being normalized to β-actin. Primers for neuronal lineage markers were same primer design we have previously published [23]. Briefly described, we designed primers to span exon-exon junctions. Each primer pair's thermal stability curves were assessed for evidence of a single amplicon, with each amplicon's length being verified using agarose gel electrophoresis, and amplicon identity being verified by Sanger sequencing. A list of primers and their sequences is presented in Supplementary Table 21. Dae D Chung: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper. Rajesh C Miranda: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper. Susan T Weintraub and Amanda H Mahnke: Performed the experiments; Contributed reagents, materials, analysis tools or data. Nihal A Salem: Performed the experiments; Analyzed and interpreted the data. Khang T Le, Elizabeth A Payne and Tenley E Lehman: Performed the experiments. Marisa R Pinson: Analyzed and interpreted the data. Mr. Dae D Chung, Rajesh C Miranda, Marisa R Pinson, Nihal A Salem were supported by 10.13039/100000027National Institute on Alcohol Abuse and Alcoholism [F31AA028446; R01AA024659; F30AA027698; F99NS113423]. Data associated with this study has been deposited at https://massive.ucsd.edu. The authors declare no conflict of interest. No additional information is available for this paper.

PMC9649996

Xin Yan,Deyun Chen,Xinran Ma,Yao Wang,Yelei Guo,Jianshu Wei,Chuan Tong,Qi Zhu,Yuting Lu,Yang Yu,Zhiqiang Wu,Weidong Han

CD58 loss in tumor cells confers functional impairment of CAR T cells

23-06-2022

Visual Abstract

CD58 loss in tumor cells confers functional impairment of CAR T cells Chimeric antigen receptor (CAR) T-cell therapy has demonstrated unprecedented success in the treatment of B-cell malignancies, especially CD19-targeted CAR T-cell therapy for acute lymphoblastic leukemia (ALL) and diffuse large B-cell lymphoma (DLBCL).1, 2, 3, 4, 5 Notwithstanding some progress achieved, primary or acquired resistance to the treatment still occurs., A deeper exploration for resistance mechanisms to CAR T-cell therapy may provide diverse rationales for patient selection or potential strategies. Target antigen evasion has been confirmed as a mechanism for acquired resistance to CAR T-cell therapy.7, 8, 9, 10 Increasing evidence has suggested that the mechanisms of primary resistance to CAR T-cell therapy involve CAR T-cell defects, including impaired proliferative capacity, an exhaustion phenotype, and attenuated T-cell–mediated cytotoxicity.,, Nevertheless, intrinsic mechanisms of primary resistance of tumor cells to the treatment remain largely elusive. High-throughput clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9)-based screening is a powerful tool that provides unbiased critical genetic data to exploit for the reasons for resistance to CAR T-cell therapy or to find new biomarkers for stratifying patients. To identify tumor cell–intrinsic factors that determine resistance to CAR T-cell cytotoxicity, we performed unbiased genome-wide CRISPR/Cas9 screening with a coculture model consisting of Nalm6 cells and CD19 CAR T cells. We revealed that CD58, a ligand for the CD2 receptor expressed on T cells, plays a key role in resistance to CAR T-cell therapy in preclinical tests. Disruption of CD58 in tumor cells impaired immunological synapse (IS) formation with CAR T cells, which led to the dysfunction of CAR T cells, including attenuated CAR signal transduction, CAR T-cell expansion, and cytotoxicity. Taken together, these findings suggest a potential mechanism for resistance to CAR T-cell therapy. Additional materials and methods are provided in supplemental Materials. The process of CRISPR/Cas9 screening has been described in our previous report. Briefly, Nalm6 cells were transfected with lentivirus carrying the Brunello library, and cells exhibiting stable lentivirus integrations were selected with puromycin. Transduced Nalm6 cells were cultured with CD19 CAR T cells or control T cells at a 1:50 effector:target (E:T) ratio. Control T or CD19 CAR T cells were added to the culture at a 1:50 E:T ratio every 3 days. The cells were collected using a death cell removal magnetic bead kit (Miltenyi) for genomic DNA analysis on day 15. After removing low-quality reads from the original sequencing data, the reads were mapped to single-guide RNA (sgRNA) sequence, and each sgRNA read was counted to generate a sgRNA count table. sgRNA data in the sgRNA table was normalized and used for significance analysis. sgRNA read counts were analyzed with the MAGeCK v0.5.7 algorithm. Genes with significantly enriched sgRNAs were identified based on a log2 fold change and P value criteria. Hypergeometric distribution statistics were used to identify gene sets that overlapped with candidate genes (log2 fold change > 2 and P < .05). Cells were pretreated with blocking monoclonal antibody (mAb) against CD58 (clone TS2/9, 8 μg/mL, BioLegend) or CD2 (clone RPA-2.10, 10 μg/mL, eBioscience) or with isotype-matched control mAbs for 30 minutes. The blocking effect was detected by staining for anti–CD58-PE (BioLegend) or anti–CD2-APC-Cy7 (BioLegend). Tumor cells blocked by anti-CD58 mAbs or CAR T cells blocked by anti-CD2 mAbs were subjected to subsequent experiments, including cytotoxicity assays and degranulation assays, according to the methods described above. The sorted CD19 CAR T cells were stained with 0.2 μM CellTrace Violet dye (Thermo Fisher Scientific) in accordance with the manufacturer’s instructions. A total of 1 × 106 wild-type (WT) (mCherry+) Nalm6, 1 × 106 CD58 knockout(KO) (GFP+) Nalm6 and 1 × 106 CD19 CAR T cells (Violet+) were cocultured at 37°C for the indicated times, fixed, permeabilized, and then stained with phalloidin-AF647 (Thermo Fisher Scientific) for detecting F-actin. A total of 1 × 106 events were recorded, and samples were analyzed with an ImageStreamX MKII flow cytometer (Luminex). Image acquisition and data analysis, was performed using IDEAS software version 6.2. The calculation formula for measuring IS is as follows: F-actin enrichment at IS (%) = 100× (intensity of phalloidin at IS)/(intensity of phalloidin-stained CAR T cells) A total of 1 × 105 luciferase+ Nalm6 cells (WT or CD58KO) were IV transplanted into 4- to 6-week-old female NOD-Prkdc-scid-Il2rg-null mice (NPG/Vst, VITALSTAR). Purified CD19 CAR T cells were selected using magnetic beads (Miltenyi Biotec) 3 days post–lentiviral infection. Seven days after Nalm6 cell injection, the mice were IV injected with 1 × 106 CD19 CAR T or control T cells in 100 μl of phosphate-buffered saline (PBS) (n = 6 mice per group). Leukemia burden was monitored once per week by bioluminescence in vivo imaging (BLI) with an in vivo imaging system (IVIS) (PerkinElmer). The average flux (photons per second/area [mm2]) was used to evaluate the BLI signal. Mouse peripheral blood samples were collected through the orbital sinus and lysed using ammonium chloride-potassium (ACK) lysing buffer (Thermo Fisher Scientific). The remaining cells were stained with the indicated fluorochrome-conjugated antibodies. All studies were approved by the Institutional Animal Care and Treatment Committee of the Chinese People's Liberation Army General Hospital. Statistical analyses were conducted using GraphPad Prism 6. Statistical tests were performed using a 2-tailed t test, 1-way analysis of variance (ANOVA) test, and 2-way ANOVA test with Bonferroni correction to compare the significant differences. Survival analysis was analyzed using the log-rank test. Unless otherwise indicated, P ≤ .05 was considered statistically significant for all analyses. All group values are represented as means plus or minus standard deviation (SD) if not stated otherwise. To systematically identify critical regulators that determine resistance to CAR T-cell therapy in tumor cells, we conducted genome-wide CRISPR/Cas9 screening with a coculture model containing Nalm6 cells, CD19+ human pre-B ALL cells, and CD19 CAR T cells (Figure 1A). The Nalm6 cells were transduced with a lentiviral CRISPR Brunello library targeting ∼19 000 genes and then selected under puromycin pressure for 2 days. To better reflect the long-term high tumor burden in vivo, Nalm6 cells were supplemented with CD19 CAR T cells every 3 days for 15 days at a 1:50 E:T ratio. To avoid contamination by the enrichment of sgRNAs that are associated with tumor cell survival but not with CAR T-cell therapy, we cocultured Nalm6 cells that had been transduced with the aforementioned CRISPR library with control T cells as the control condition. The composition of the sgRNAs in surviving tumor cells under CAR T-cell or control T-cell treatment conditions was evaluated by Illumina sequencing and analyzed by MAGeCK algorithm (supplemental Table 1). Our CRISPR screening identified expected candidates among the top 10 hits, namely CD19,,, JAK2, and CASP8, consistent with known mechanisms of resistance to CD19-targeted immunotherapies (Figure 1B). Interestingly, a cluster of membrane protein genes, including CD58, ICAM1, CD81, and ITGA4 were also positively selected (Figure 1C). In order to verify our screening findings, we generated stable KO cell lines of membrane proteins (CD58, CD81, ICAM1, and ITGA4) in Nalm6 cells by the CRISPR/Cas9 technology (Figure 1D). Growth competition assays were conducted for WT or KO target cell lines with different fluorescence labels in the presence of CAR T cells. The growth competition assays revealed that CD58KO and CD81KO cells conferred progressive enrichment in the presence of CD19 CAR T cells compared with WT cells (Figure 1E-F; supplemental Figure 1A-B). However, ICAM1KO and ITGA4KO cells did not show a significant increase (Figure 1G-H; supplemental Figure 1C-D). We identified that CD81 loss induced disruption of CD19 membrane trafficking (supplemental Figure 2). This finding is similar to previous reports in which downregulation of CD81 has been identified as a mechanism for resistance to CD19-targeted therapy., In the current study, we focused on the role of CD58 in CAR T-cell therapy. To explore whether CD58-deficient tumor cells were resistant to CAR T-cell therapy with different CAR design, we also generated distinct CAR T cells including CD20 CAR T, tandem CD19/CD20 CAR T cells, and CD19.28z CAR T cells from multiple healthy donors and observed that both CD58 loss in Nalm6 cells and Raji cells were relatively resistant to CAR T cells (Figure 2A-G; supplemental Figures 3 and 4A-J). We also found that tumor cells with disruption of CD58 exhibited low sensitivity to CAR T-cell–mediated killing in a cytotoxicity assay (Figure 2H-L; supplemental Figure 2K-M). This effect was recapitulated with an anti-CD58–blocking mAb (Figure 2M-N). Notably, blockage of CD2 on CAR T cells resulted in impaired CAR T-cell–mediated cytotoxicity (Figure 2O-P). As expected, resistance was not attributed to the downregulation of CD19 expression or CD20 expression, as determined by flow cytometry (Figure 2Q-R). Additionally, we observed that knocking out CD58 had no effect on the proliferation or apoptosis of tumor cells (supplemental Figure 5A-C). However, CD58 loss did not protect tumor cells from chemotherapy-mediated killing (supplemental Figure 5D), implying that CD58 loss in tumor cells may specifically confer resistance to CAR T-cell–mediated killing. Collectively, these findings imply that the CD58-CD2 axis is necessary in cytotoxic killing by CAR T cells and that lack of CD58 in lymphoid cancer cells could induce resistance to CAR T-cell therapy. The CD58-CD2 interaction has been reported to be a crucial costimulatory signal for T-cell activation in response to target cells. Using TCGA RNA sequencing data and Tumor Immune Estimation Resource 2.0, (TIMER), we found that messenger RNA levels of CD58 were positively correlated with CD8 T-cell infiltration in many human cancer types (supplemental Figure 6A). Besides, we found that the low expression of CD58 was associated with the low expression of interferon γ (IFN-γ) and tumor necrosis factor α (TNF-α) (supplemental Figure 6B-C), which suggested that the downregulated expression of CD58 in tumor cells may be related to the dysfunction of T cells. Hence, we wondered whether resistance to CAR T-cell therapy caused by CD58 disruption in cancer cells is due to attenuated CAR T-cell function. We found decreased expansion of CD19 CAR T cells cocultured with CD58KO Nalm6 cells or CD58KO Raji cells (Figure 3A-B). We also observed that lack of CD58 in tumor cells initiated dysfunctional degranulation, as measured by CD107a level (Figure 3B; supplemental Figure 7A). In parallel, adding an anti-CD58–blocking mAb to Nalm6 cell cultures and an anti-CD2–blocking mAb to CD19 CAR T-cell cultures significantly inhibited degranulation (Figure 3C-D). Moreover, we found that loss of CD58 in tumor cells led to the reduced secretion of cytokines, such as interleukin-2 (IL-2), TNF-α, and IFN-γ (Figure 3E; supplemental Figure 7B). To further explore the effect of tumor cells with disrupted CD58 on CAR T cells, we sorted CAR T cells cultured with WT or CD58KO tumor cells and then tested their kinetic responses when cocultured with WT tumor cells (Figure 3F). Remarkably, we found that CD19 CAR T cells initially cultured with CD58KO tumor cells exhibited low expansion capacity, a reduced degranulation, and impaired ability to kill WT tumor cells again (Figure 3G-J; supplemental Figure 8A-D). We also noted increased apoptosis in CD19 CAR T cells initially cultured with CD58KO tumor cells, and this increased apoptosis was not driven by FAS or TNFR2 upregulation (Figure 3K; supplemental Figure 8E-F). To investigate the effect of chronic CD58KO tumor cell stimulation on the function of CAR T cells, we established a coculture system in which WT or CD58KO tumor cells were added to a CAR T-cell culture every 72 hours (Figure 3L). Repetitive CD58KO tumor cell stimulation attenuated CAR T-cell expansion and reduced CAR T-cell activation as measured by Ki67, CD25, and CD69 level (Figure 3M-O; supplemental Figure 8G-K). Overall, these results suggest that CD58 disruption in cancer cells conferred functional impairment of CAR T cells, including reduced CAR T-cell expansion, survival, activation, degranulation, cytotoxicity, and cytokine secretion and increased CAR T-cell death, which might be responsible for resistance to CAR T-cell therapy. The CD2-CD58 interaction is essential for the formation of effective IS, to sustaining the activation and proliferation of T cells and trigger a series of intracellular signaling pathways in T cells. Recent studies have revealed that nonclassical IS formed by CAR T cells and tumor cells has been regarded as an important indicator for predicting the effectiveness of CAR T-cell therapy.27, 28, 29 Therefore, we hypothesized that the inhibition of CAR T-cell function induced by CD58-deficient tumor cells is caused by the ineffective formation of IS and weakened CAR signaling strength. To test this hypothesis, we performed an in vitro conjugation assay. Compared with WT cells, CD58KO Nalm6 cells formed significantly fewer conjugates with CAR-expressing Jurkat cells or CAR T cells (Figure 4A; supplemental Figure 9). Next, we used high-throughput imaging flow cytometry (ImageStream) to evaluate the stability of IS (Figure 4B). CD58KO Nalm6 cells formed IS structures with potentially disadvantageous cytoskeletal properties, as measured by F-actin intensity and enrichment at the IS (Figure 4C-D), implying that CD58KO Nalm6 cells prevented effective IS formation with CAR T cells. Considering that the production of CAR T cells requires the activation of anti-CD3 antibody, the addition of anti-CD3 antibody caused high background phosphorylation level and we could not detect differences between groups. Therefore, we observed that the phosphorylation of CD3ζ-CAR, LCK, or ZAP70 in CAR-expressing Jurkat cells stimulated with CD58KO Nalm6 cells was inhibited compared with that stimulated with WT Nalm6 cells (Figure 4E). Of note, an anti-CD2 antibody can activate CAR T cells by detecting the expression of CD25 in CAR T cells (supplemental Figure 10A). However, CD58 loss in tumor cells did not show increased sensitivity to CAR T cells after the addition of an anti-CD2 antibody (supplemental Figure 10B), indicating the stable IS formed by CD58 on tumor cells is necessary for CAR T cells to successfully kill tumor cells. To decipher the underlying molecular programs accounting for CD58-deficient tumor-mediated CAR T-cell dysfunction, we leveraged the transcriptional profiles of CD19 CAR T cells cocultured with WT Nalm6 cells or CD58KO Nalm6 cells. We observed differentially expressed genes in CD19 CAR T cells stimulated with WT Nalm6 cells or CD58KO Nalm6 cells (supplemental Figure 11A). A gene enrichment analysis of these differentially expressed genes revealed that these genes were significantly enriched in regulation of cell adhesion, T-cell activation, cytokine-related signaling, and cell proliferation (Figure 4F; supplemental Figure 11B-C; supplemental Table 2). More specifically, CD19 CAR T cells cocultured with CD58KO Nalm6 cells showed marked downregulation of genes associated with activation (RIPOR2, IL1A, RUNX1, NFKB2, SDC4, and TNFRSF4)31, 32, 33, 34, 35 and significantly differentially expressed genes associated with T-cell differentiation (CXCR4, CCR7, TOX, LGALS3, and XBP1).36, 37, 38 In addition, we found that the expression of a cluster of cytokine genes (IFNG, TNF, IL17F, IL13, CCL3L3, and CCL5) was decreased in CAR T cells cocultured with CD58KO Nalm6 cells (Figure 4G). Taken together, these findings indicate that CD58KO tumor cells and CAR T cells form inefficient IS, which drives a reduction in CAR T-cell activation, resulting in CAR T-cell dysfunction. To evaluate the effect of CD58KO tumor cells on the anti-tumor ability of CAR T-cell therapy in mouse xenograft models, we examined the tumor-suppressive ability of equal amounts of CAR T cells in mice transplanted with WT tumor cells or CD58KO tumor cells (Figure 5A). Consistent with our in vitro observations, the CD19 CAR T cells in CD58KO Nalm6 cell-bearing mice showed low tumor clearance capacity (Figure 5B-C). Moreover, xenograft mice with CD58KO Nalm6 cell transplants exhibited a survival disadvantage compared with xenograft mice with WT Nalm6 cell transplants (Figure 5D). Consistently, the expansion of CAR T cells in peripheral blood in CD58KO tumor-bearing mice was significantly lower than that in WT tumor-bearing mice (Figure 5E). Besides, we found that loss of CD58 in tumor cells suppressed activation of CAR T cells, as measured by CD25 level (Figure 5F). CAR T cells in the CD58KO tumor group secreted fewer cytokines than those in the WT tumor group (Figure 5G). These results indicate that loss of CD58 in tumor cells results in reduced sensitivity to CD19 CAR T-cell therapy in vivo. Resistance to CAR T-cell therapy is a primary obstacle to its broader therapeutic use. Performing unbiased CRISPR/Cas9 screening with the Nalm6 cells, CD19+ human ALL cell line, we revealed several genetic perturbations potentially capable of mediating resistance to CAR T-cell therapy. In addition to antigen loss and T-cell dysfunction, tumor-intrinsic resistance mechanisms, such as impaired death receptor signaling and NOXA, have been recently reported.,, In the present study, we identified a potential mechanism of tumor-intrinsic resistance to CAR T-cell therapy mediated by the loss of CD58 in tumor cells. CD58 is a member of the immunoglobulin superfamily and is a ligand for the costimulatory molecule CD2 expressed in T cells. Disruption of the CD58-CD2 axis by blocking antibodies results in decreased T-cell activation, reduced IFN-γ secretion, and reduced cytotoxicity., Several reports have shown that loss of CD58 in tumor cells is an unfavorable prognostic factor and a frequent genetic abnormality in patients with hematologic malignancies., On the basis of CRISPR/Cas9 screening, a recent study has revealed that CD58 loss can confer immune evasion in tumor-infiltrating lymphocyte-mediated killing and that CD58 expression is downregulated in tumors of melanoma patients with resistance to immune checkpoint inhibitors. Likewise, CRISPR/Cas9 screening was implemented showed that CD58 deletion caused Nalm6 cell escape from natural killer cell–mediated killing in another study. These data suggest that abnormal CD58 expression in tumor cells may confer general resistance to various immunotherapies and that further investigation is needed in future studies. CD58 is essential for the formation of stable IS, maintenance of T-cell activation, T-cell survival, and T-cell–mediated killing., However, the effect of CD58 expression in tumor cells on CAR T-cell therapy remains unknown. In this study, we observed that CD58KO in 2 B-lymphoid cell lines showed significantly less sensitive to a series of CAR T cells, including CD19 CAR T, CD20 CAR T, and tandem CD19/20 CAR T cells. In addition, we found that the loss of CD58 in tumor cells triggered impaired CAR T cells, which resulted in decreased CAR T-cell proliferation, degranulation, cytokine secretion, and cytotoxic effects and increased cell death. Even if the structure of IS formed by CAR T cells and target cells is saliently distinct from that the structure of classical IS, but an increasing number of studies demonstrated that the IS formed by CAR T cells plays an important role in driving the cytotoxic function of CAR T cells.27, 28, 29,, In this study, we observed that CD58 loss in tumor cells prevented effective IS formation with CAR T cells, as measured by the intensity and enrichment of F-actin, a key component in IS., Furthermore, we found that CAR T cells saliently attenuated CAR signaling and CAR T-cell activation stimulation by CD58-deficient tumor cells. These results may shed light on the reasons that loss of CD58 in tumor cells induces resistance to CAR T-cell therapy. Unfortunately, we were unable to provide clinical relevance of CD58 protein levels to CAR T-cell therapy in hematologic malignancies. Factually, we detected CD58 protein levels in tumor specimens from 34 patients with B-cell lymphoma before infusions of tandem CD19/20 CAR T cells by immunohistochemistry., Unexpectedly, we found only 2 patients with low CD58 expression (data not shown), which is totally distinct from the 60% DLBCL patients of CD58 protein downregulation reported by others, in the Western population. In addition, we noted a large difference of CD58 mutation frequency in patients with DLBCL between the Chinese population (∼5% to 10% mutation rate), and Western population (>20% mutation rate)., Although the 2 patients with CD58 low expression had poor response in our tandem CD19/20 CAR T-cell clinical trial, few sample number limited further comparative analysis. We preliminarily speculate that the expression of CD58 may be affected by racial disparities; CD58 loss or downregulation was not a main contributing factor for resistance to CAR T-cell therapy, just a low-probability event in Chinese patients with DLBCL. Strategies to overcome tumor resistance to CAR T-cell therapy caused by CD58 loss remains to be further explored. A recent study illuminated that bypassing CD58 loss in tumor cells using a novel CAR T-cell construct integrating CD2 costimulatory domains with CAR molecules may be a potential therapeutic strategy. CD58 is regulated by both genetic and nongenetic factors. A previous study suggested that EZH2 inhibitors can restore epigenetically silenced CD58 expression on the surface of lymphoma cells, which in turn enhance IFN-γ secretion by T/natural killer cells. Reversing the functional downregulation of CD58 in tumor cells using drugs such as epigenetic modulators may contribute to novel combinatorial treatment strategies that can improve clinical responses to CAR T-cell therapy. Overall, our findings emphasize a potential molecular mechanism determining the resistance of B-cell malignancies to CAR T-cell therapy. Our observations suggest that CD58 may be a clinically predictive biomarker for evaluating response to CAR T-cell therapy in hematologic malignancies, and therefore, targeting CD58 may be a novel therapeutic avenue to enhance the sensitivity or overcome resistance to CAR T-cell therapy. Conflict-of-interest disclosure: The authors declare no competing financial interests.